id stringlengths 3 9 | source stringclasses 1 value | version stringclasses 1 value | text stringlengths 1.54k 298k | added stringdate 1993-11-25 05:05:38 2024-09-20 15:30:25 | created stringdate 1-01-01 00:00:00 2024-07-31 00:00:00 | metadata dict |
|---|---|---|---|---|---|---|
219098420 | pes2o/s2orc | v3-fos-license | Endovascular Arteriovenous Dialysis Fistula Intervention: Outcomes and Factors Contributing to Fistula Failure
Rationale & Objective Primary patency is variable with arteriovenous fistulas, and many patients require angiographic procedures to obtain patency. Accordingly, we determined postintervention patency rates and contributing factors for fistula failure following intervention to establish secondary patency in non–dialysis-dependent patients with advanced chronic kidney disease following creation of an arteriovenous fistula. Study Design Observational study from a single referral center. Setting & Participants 210 non–dialysis-dependent patients with advanced chronic kidney disease who underwent upper-extremity fistula creation for anticipated dialysis between October 1995 and January 2015 and who required subsequent endovascular therapy to establish or maintain patency were reviewed. Exposure Endovascular therapy for dialysis arteriovenous fistula primary patency failure. Outcomes Postintervention patency duration following endovascular therapy. Analytical Approach Descriptive study with outcomes determined using Cox proportional hazards models. Results Multiple fistula configurations were reviewed: 138 (65.7%) brachiocephalic, 39 (18.6%) radiocephalic, 30 (14.3%) brachiobasilic, 2 (1.0%) ulnocephalic, and 1 (0.5%) radiobasilic. There were 261 initial stenoses treated. Postintervention primary patency is defined as the time from the index intervention to repeat intervention for stenosis. Postintervention primary-assisted patency is the time from the index intervention to thrombectomy for fistula thrombosis or change in modality. Postintervention secondary patency is the time from the index intervention to fistula abandonment. Median postintervention primary patency, postintervention primary-assisted patency, and secondary patency were 2.7, 3.2, and 3.6 years, respectively. The overall 1-year primary, primary-assisted, and secondary patency rates in this cohort were 53.0%, 87.7%, and 83.5%, respectively. Compared with radiocephalic fistulas, brachiocephalic fistulas had higher risk for postintervention primary patency loss (HR, 1.90; 95% CI, 1.13-3.20; P = 0.02). Limitations Dialysis fistula revascularization techniques varied. Conclusions The radiocephalic fistula configuration had the best postintervention primary patency in this cohort. Postintervention primary-assisted patency and secondary patency were not significantly different among different fistula configurations.
Rationale & Objective: Primary patency is variable with arteriovenous fistulas, and many patients require angiographic procedures to obtain patency. Accordingly, we determined postintervention patency rates and contributing factors for fistula failure following intervention to establish secondary patency in non-dialysisdependent patients with advanced chronic kidney disease following creation of an arteriovenous fistula.
Study Design: Observational study from a single referral center.
Setting & Participants: 210 non-dialysis-dependent patients with advanced chronic kidney disease who underwent upper-extremity fistula creation for anticipated dialysis between October 1995 and January 2015 and who required subsequent endovascular therapy to establish or maintain patency were reviewed.
Exposure: Endovascular therapy for dialysis arteriovenous fistula primary patency failure.
Analytical Approach: Descriptive study with outcomes determined using Cox proportional hazards models.
Conclusions:
The radiocephalic fistula configuration had the best postintervention primary patency in this cohort. Postintervention primary-assisted patency and secondary patency were not significantly different among different fistula configurations. I n 2015, approximately 124,000 new cases of end-stage kidney disease were reported and nearly 500,000 patients were receiving dialysis treatment in the United States. 1 In the current practice environment, the Fistula First initiative emphasizes the creation of arteriovenous fistulas over grafts or catheter placement. 2,3 However, 20% to 50% of the fistulas created never mature to support dialysis. 4,5 Furthermore, 27% of fistulas fail and are abandoned within 18 months of creation. 6,7 These outcomes have significant economic consequences and constitute w8% of total Medicare end-stage kidney disease spending. 8 Fistulas are often constructed when patients are in the non-dialysis-dependent stage of kidney disease in anticipation for future dialysis. When these fistulas fail to mature, patients are often referred for endovascular management to evaluate and treat lesions that can inhibit fistula maturation, such as dilating stenoses, with angioplasty. 9 Approximately 50% of fistulas require interventions before successful arteriovenous fistula use. 10 In some cases, multiple endovascular procedures are required to aid in the maturation of the fistulas. 11 At present, there are limited data for factors contributing to postintervention patency in fistulas constructed in non-dialysis-dependent patients, and given the high incidence of endovascular procedures on immature fistulas, more investigation is warranted. 12,13 The purpose of this study was to determine the predictors of recurrent lesions in fistulas constructed in the non--dialysis-dependent population and evaluate the outcomes after endovascular therapy.
Study Population and Design
Electronic medical records were reviewed and 407 non--dialysis-dependent patients with fistulas constructed between October 1995 and February 2015 were identified. Patients were excluded if they had incomplete medical or imaging data. The final cohort consisted of 210 patients, all Complete author and article information provided before references.
Correspondence to S. Misra (misra.sanjay@mayo.edu) of whom had the fistula constructed before the initiation of maintenance hemodialysis.
Institutional review board approval was obtained (IRB # 17-008191) and informed consent was waived for this retrospective single-institution study evaluating postintervention patency rates of fistulas constructed in non--dialysis-dependent patients.
Definitions
Fistulas were divided into 4 anatomical regions based on prior definitions: juxta-anastomotic segment (JAS; artery ≤ 2 cm from the anastomosis, anastomosis, and vein ≤ 2 cm central to the anastomosis), outflow vein (vein central to cannulation zone but peripheral to cephalic arch or central vein), cephalic arch, and central vein. The central veins included any vein central to the axillary vein (eg, subclavian vein, brachiocephalic vein, or superior vena cava). [13][14][15][16] Definitions of postintervention patency duration have been previously described. 17,18 Postintervention primary patency is defined as the time from the index intervention to repeat intervention for stenosis. Postintervention primary-assisted patency is the time from the index intervention to thrombectomy for fistula thrombosis or change in modality. Postintervention secondary patency is the time from the index intervention to fistula abandonment.
Procedure
Patients were referred to Interventional Radiology for diagnostic fistulography by the Department of Nephrology if they had sonographic evidence of fistula stenosis or thrombosis, abnormal upper-extremity physical examination results including arm pain and swelling, were unable to achieve fistula maturation to sustain adequate dialysis (blood flow < 500 mL/min), or cannulation difficulties at the initiation of dialysis. Procedures were performed by 9 different interventionalists with 3 to 17 years of experience (median, 10 years). Patients were referred for surgical intervention if the fistula failed satisfactory percutaneous therapy. Fistulas were accessed in the outflow vein. The
Original Research
Kidney Med Vol 2 | Iss 3 | May/June 2020 direction of access was based on the lesion location. Fistulas were evaluated fluoroscopically with iodinated contrast. Stenoses with at least 50% visual luminal narrowing were treated with plain-balloon angioplasty to the end point of ≤30% stenosis by fluoroscopic assessment. Pressure measures were not routinely performed. Noncovered stents were not used. Covered stents were used only in cases of fistula rupture. Target lesion revascularization, defined as reintervention of recurrent stenosis at the same location, was performed on recurrent hemodynamically significant lesions as needed.
Statistical Analysis
All analyses were completed using SAS, version 9.4 (SAS Institute). Descriptive statistics of patient characteristics are reported as mean (standard deviation) for continuous variables. Non-normally distributed data was described as median (25th-75th percentile). A Cox proportional hazard model was used to assess the association between patency and patient characteristics, fistula configuration, and comorbid conditions. The alpha level was set at 0.05 for statistical significance. Median postintervention primary patency was 2.7 years, median postintervention primary-assisted patency was 3.2 years, and median postintervention secondary patency was 3.6 years. Overall 1-year primary, primary-assisted, and secondary patency rates in this cohort were 53.0%, 87.7%, and 83.5%, respectively. There were 261 stenoses treated among the 210 fistulas, establishing the baseline for postintervention patency. This included 107 (41.0%) stenoses in the JAS, 86 (33.0%) in the outflow vein, 55 (21.1%) in the cephalic arch, and 13 (5.0%) in the central veins. Table 2.
Several comorbid conditions were evaluated as potential contributors to patency loss, including diabetes mellitus, coronary artery disease, peripheral artery disease, hyperlipidemia, and hypertension. None of these were significant factors affecting postintervention primary, primaryassisted, or secondary patency. Tobacco smoking increased the risk for loss of postintervention primary patency (HR, 2.22; 95% CI, 1.12-4.40; P = 0.02), but postintervention primary-assisted and secondary patency loss were not significantly affected (P = 0.98 and P = 0.25, respectively).
Medications were reviewed, which included β-blockers, calcium channel blockers, aspirin, clopidogrel, warfarin, and statins. None of these medications were associated with lower rates of patency loss. Aspirin was associated with increased risk for loss of postintervention primary patency (HR, 1.52; 95% CI, 1.04-2.21; P = 0.03), though no association with postintervention primary-assisted or secondary patency loss was found (P = 0.99 and P = 0.64, respectively). A total of 133 (63%) patients were taking aspirin at the time of fistula construction.
DISCUSSION
This study evaluated postintervention patency in the largest cohort of dialysis fistulas constructed in the non-dialysisdependent population. Among the radiocephalic, brachiocephalic, and brachiobasilic fistulas, the radiocephalic fistula demonstrated the best postintervention primary patency. Of the 4 different fistula locations examined, the most common site for primary stenosis was the JAS segment. Recurrent stenosis at 1 year of treatment occurred most frequently at the cephalic arch, whereas central venous stenoses were the least likely to reoccur after angioplasty. The rate of restenosis was not influenced by any of the comorbid conditions reviewed. However, smoking increased the risk for postintervention primary patency loss.
The construction of radiocephalic and other fistulas in the forearm can be challenging due to small vessel size. As a result, forearm fistulas may have lower maturation rates compared with arm fistulas. 13,20,21 However, if the vessel diameter is adequate, the data in the current study demonstrate that these fistulas may be at least as durable as those created in the upper arm. Furthermore, fistulas in the wrist have the benefit of enabling the creation of upper-arm fistulas after failure of the forearm fistula, whereas it is usually impossible to create fistulas in the opposite order. 22 A retrospective study investigating endovascular therapy in 153 immature fistulas by Han et al 23 demonstrated that the radiocephalic fistula configuration was not associated with better primary (P = 0.819) or secondary patency rates (P = 0.465). However, a subsequent retrospective study by Lee et al 12 demonstrated better primary patency among 42 immature radiocephalic fistulas compared with 12 immature brachiocephalic fistulas (P = 0.003). In a study of mature fistulas, Rajan et al 24 reported that among 94 radiocephalic and 57 brachiocephalic fistulas, radiocephalic fistulas had significantly better 3-, 6-, and 12month primary patency (P = 0.004), but not secondary patency (P = 0.45).
The 1-year postintervention primary patency rate in this study was 53%. This finding is similar to mature fistulas that develop recurrent stenosis after endovascular therapy. 25,26 Dialysis fistula stenoses most often occur at or adjacent to the anastomosis, which was further confirmed in this study. 27,28 The location of stenosis recurrence varied by fistula configuration. The juxta-anastomosis was the most common location of recurrence for radiocephalic fistulas. However, recurrent central venous stenosis was the most prevalent site of recurrence among both brachiobasilic and brachiocephalic fistulas. Previously reported primary patency for treated central venous stenoses in patients with dialysis fistulas was 46% at 9 months. 29 The lower patency rate for central venous stenosis at 1 year may be due to higher elastic recoil of central venous lesions. 30 However, treatment of primary stenoses involving the cephalic arch had the worst postintervention primary patency with a 1-year probability of recurrent stenosis of 34.6%. Previously reported primary patency of cephalic arch stenosis treated using plain balloon angioplasty was 11%. 31 In contrast, central venous stenoses recurred the least often at a rate of 7.7% at 1 year after intervention in the present study.
Tobacco smoking was associated with poorer postintervention primary patency outcomes. Smoking is an established risk factor for primary failure in dialysis fistulas, though data for the effect of tobacco use and postintervention patency are limited. 32,33 The incidence of primary fistula failure may be 4.3 times greater among patients with a history of smoking. Other previously reported risk factors for arteriovenous fistula patency loss, including age, sex, and diabetes, were not found to have a significant impact on postintervention patency rates. 32,34,35 Aspirin use at the time of intervention was associated with an increased risk for loss of postintervention primary patency, but not postintervention primary-assisted or secondary patency. Consistent aspirin use has been shown to be associated with a 37% reduction in fistula failure, but no data are available for postintervention primary patency. 36 A meta-analysis by Tanner and da Silva 37 failed to demonstrate any beneficial effect from aspirin as an adjuvant therapy. The current study is the first to show that aspirin may have a deleterious effect on fistula patency postintervention. This finding may reflect the fact that recurrent fistula failure is typically a result of venous intimal hyperplasia rather than in situ thrombosis. 38 Furthermore, aspirin use can be a marker for cardiovascular disease and may signify a more comorbid population.
This study was limited by its retrospective design, resulting in variations in angioplasty technique, medical treatment, and follow-up. Angioplasty technique including balloon pressure and duration of dilation was not measured and is a major limitation. The interval between fistula construction and initiation of dialysis, whether through the fistula or a catheter, was variable. Radiocephalic fistulas in this study demonstrated better primary patency than other configurations. This may be explained by this group of patients having healthier veins or potentially less peripheral arterial disease. Vessel quality was not evaluated and may be a confounding variable for these findings.
In conclusion, among non-dialysis-dependent patients who undergo initial endovascular therapy to maintain fistula patency, the radiocephalic configuration was found to have the best postintervention primary patency compared with other configurations. Although the majority of hemodynamically significant lesions occurred at the anastomosis, lesions in the noncannulation zone of the outflow vein had the highest rates of target lesion revascularization and may require more aggressive treatment. | 2020-04-23T09:14:55.249Z | 2020-04-21T00:00:00.000 | {
"year": 2020,
"sha1": "27b51b59f64dd36542ad8df6aca0b1bc40361521",
"oa_license": "CCBYNCND",
"oa_url": "http://www.kidneymedicinejournal.org/article/S2590059520300674/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "53f8da4139c78eea00dc2613a093f689ee9d04c0",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
286211 | pes2o/s2orc | v3-fos-license | Exploiting the origins of Ras mediated squamous cell carcinoma to develop novel therapeutic interventions
The small GTPase Ras is activated in a high proportion of human cancers. Attempts to clinically block Ras activity through pharmacological means has proven largely ineffective thus far. We employed an inducible mouse model of squamous cell carcinoma (SCC) to study the effect of Ras activation and show that hair follicle stem cells (HFSCs) are a cell of origin for SCC, whereas their more restricted progeny cannot serve as cancer cells of origin and are refractory to Ras activation. We propose that by identifying the unique mechanisms by which HFSCs are mobilized to initiate Ras mediated tumorigenesis, the molecular process behind SCC can be more completely elucidated and context dependent activities for Ras more clearly defined. Here, we summarize our recent results and point to future experiments designed to create novel therapeutics by exploiting the differential sensitivities of various cells within the epidermis to Ras activation.
T he small GTPase Ras is activated in a high proportion of human cancers. Attempts to clinically block Ras activity through pharmacological means has proven largely ineffective thus far. We employed an inducible mouse model of squamous cell carcinoma (SCC) to study the effect of Ras activation and show that hair follicle stem cells (HFSCs) are a cell of origin for SCC, whereas their more restricted progeny cannot serve as cancer cells of origin and are refractory to Ras activation. We propose that by identifying the unique mechanisms by which HFSCs are mobilized to initiate Ras mediated tumorigenesis, the molecular process behind SCC can be more completely elucidated and context dependent activities for Ras more clearly defined. Here, we summarize our recent results and point to future experiments designed to create novel therapeutics by exploiting the differential sensitivities of various cells within the epidermis to Ras activation.
Introduction
Small GTPases in skin cancers. Squamous cell carcinoma (SCC) of the skin is a highly prevalent cancer with a predisposition to metastasize. Risk factors for this cancer include high lifetime UVB radiation due to sun exposure and a compromised immune system. 1,2 Specifically, organ transfer recipients with chronically suppressed immune systems have a 65-250-fold increase in SCC risk. 3 The high prevalence and significant morbidity of SCC demonstrates a need for orally or topically administered therapies to prevent or halt progression of this cancer. Previous work on murine SCC suggested that activating mutations in the small Gprotein Ras are able to drive SCC. 4,5 In fact, Ras is mutated in 30% of all human cancers including a small percentage of SCC. 6-9 The high mutation rate for Ras has led to significant effort to develop small molecule inhibitors, with little success thus far. Instead, many groups are now focusing on developing inhibitors for effectors of Ras signaling, such as the Raf family of small GTPases. 10 However, a new high-risk group for cutaneous SCC has emerged in individuals taking the B-Raf inhibitor PLX4032 as a treatment for melanoma, a cancer which is thought to be unrelated to SCC despite sharing an anatomical target. 11 As of now, it is unclear why this drug sensitizes patients to SCC or promotes its initiation, but there is evidence for crosstalk and aberrant feedback within the Ras/Raf circuit. [12][13][14] Numerous questions remain about how events downstream of Ras/Raf activation initiate various types of tumors, and how manipulation of these events could be exploited to treat cancer. Several groups, including our own, take advantage of murine models of various cancers initiated by constitutive Ras activity. Recently, we used Ras activation in specific cell types in the epidermis to identify cancer cells of origin, and to determine whether different types of cells exhibit unique molecular responses to Ras activation. 15,16 To initiate tumorigenesis in vivo, we bred mice harboring the widely-used LSL-Kras G12D knock-in allele to animals containing a floxed version of the p53 tumor suppressor gene. 17, 18 The LSL-Kras G12D allele, when combined with Cre recombinase, creates a Ras isoform that lacks GTPase activity, and is therefore continuously able to stimulate downstream pathways. To target expression of Cre to a skin stem cell population, the K15-CrePR allele was used for expression in the hair follicle (HF) stem cell niche, and the Shh-CreER knock-in allele was used to target the immediate descendants of the hair follicle stem cells, the matrix transit amplifying cells. 19,20 Comparison of the same tumorigenic insult to these two stages of the same lineage led to the conclusion that hair follicle stem cells were able to generate SCCs, whereas their Shh-expressing transit-amplifying descendants could not. From this in vivo, inducible, genetically engineered mouse model, the natural history of this cancer can be traced from its origins to its end-stage. This model provides the ability to analyze a wide variety of time-points that can be assessed for responses to Ras activation and for the development of potential interventions specific to cancer cells of origin. This model system can be also exploited as a pre-clinical testing ground for target validation of prospective SCC therapeutics.
Prevention of SCC Initiation
Targeting signaling pathways downstream of Ras. Given that patients taking PLX4032 and similar B-Raf inhibitors as a treatment for melanoma often develop SCCs, targeted therapeutics are needed in order to prevent this unfortunate side effect. Our mouse model provides a means to understand the intrinsic factors unique to HF stem cells that are necessary for SCC initiation, and thereby allows for the determination of potential targets downstream of Ras/Raf signaling for chemoprevention.
To screen for signaling pathways downstream of Ras that could be activated during the initiation of hyperplasia and/or during epithelial to mesenchymal transition (EMT) in K15-CrePR; Kras G12D and K15-CrePR; Kras G12D ; p53 ff mice, candidates were selected based on known downstream effectors (Fig. 1). We examined several signaling pathways downstream of Ras, including Map Kinases (Erk and p38) and Akt.
First, Erk1/2 activation was examined by IHC for phosphorylated Erk (p-Erk). This signaling effector of the Ras pathway was found at high levels in hyperplastic hair follicles and the basal cells of the epidermal cyst structures of skin with Kras G12D expression originating from HF stem cells. This indicates that administration of an inhibitor of MEK, an upstream regulator of Erk1/2 activity, might provide a preventative response to Kras G12D induced tumorigenesis. AZD6244 is one such inhibitor that may prove useful. 21 This potential target is further supported by previous transgenic animal studies that manipulated MEK activity. 22,23 We also examined the p38 arm of the Ras signaling by p-p38 staining. Though this marker was detected during hyperplasia and in epidermal cysts, it was also found throughout the hair follicle in control skin. This indicates that attempting to inhibit this pathway may not be useful, as it may affect normal skin homeostasis.
Second, we examined the Akt arm of Ras signaling. Using IHC for p-Akt, it was determined that Akt signaling was indeed found in some hyperplastic hair follicles and in epidermal cyst structures. Further downstream of Akt, we examined both phosphorylated mTor and phosphorylated NFkB. Phospho-mTor was evident in hyperplastic hair follicles and cyst structures at low levels compared with the robust activity of p-Akt. Rapamycin, a potent inhibitor of mTor signaling, has been suggested as a potential chemopreventative agent by studies in head and neck squamous cell carcinomas and murine chemical carcinogenesis. 24-26 Rapamycin, or a similar analog, may have some preventative effect in the initial stages in tumorigenesis in our model and in patients with Kras-inducing SCC. To examine another output of Akt signaling, we examined p-NFkB. NFkB signaling has been implicated in a wide range of tumorigenesis processes, including EMT and inflammation. 27 p-NFkB was also detected during tumorigenesis initiation in this model system. Bortezomib and Bay-117082, inhibitors of NFkB signaling, have recently been shown to be effective in inhibiting tumorigenesis in a model of lung cancer. 28 Notably, this lung cancer model utilizes the same genetic insults we used in our mouse model system. Additionally, bortezomib has been shown to have some limited effect on human cases of head and neck squamous cell carcinomas. 29 This suggests that these inhibitors may also be useful in our model of cutaneous SCC.
This examination of Ras signaling indicates that the inhibition downstream of Figure 1. Ras signaling pathways examined in SCC prone Kras G12D induced hair follicle stem cells and in SCC resistant Kras G12D induced hair follicle transit amplifying cells. P-Erk, p-Akt, p-S6 and p-p38 stained hair follicle hyperplasia and cyst structures at high levels when Kras G12D was induced in hair follicle stem cells. P-mTor, p-Ikkα/β and p-NFkB were also present, albeit at lower levels. In contrast, only p-p38 was found in the Shh-expressing hair follicle transit amplifying cells following Kras G12D induction. Darker shadings of gray represent higher activity levels. (Image adapted from White et al. 15 ).
Kras at the level of MEK, mTor or NFkB individually, or in combination, may represent a chemopreventative therapeutic regimen that can be administered simultaneous with tumor initiation in our mouse model of SCC. If successful, this may represent a potentially useful method to inhibit SCC formation in patients taking B-Raf inhibitors or in patients with compromised immune systems. However, activation of ERK, AKT and mTor (p-Erk1/2, p-Akt and p-mTor) was not detected in cells of the bona fide SCC, suggesting that these downstream pathways are not required to sustain the dedifferentiated state. This striking observation indicates that inhibition of these pathways may not be therapeutically useful following the onset of SCC in our model or in patients presenting with high grade Kras-derived SCCs. How the cancer cells evolve to shed the necessity for activity of pathways downstream of Ras or utilize alternate Ras signaling pathways warrants further investigation.
Identifying the molecular basis of sensitivity to Ras activity. In an alternate approach, since the direct descendants of the hair follicle stem cells are completely refractory to Kras G12D and Kras G12D ; p53KO induced tumorigenesis, a molecular comparison between transit amplifying cells and the parental stem cells could point toward new targets for tumorigenesis prevention. Though very closely related in hierarchy, the intrinsic properties that facilitate tumorigenesis have been lost during the transition to the transit amplifying cell type. To reveal the nature of these intrinsic properties, cell populations purified just following induction of Kras G12D expression alone and/or with p53KO could be compared in detail on the genetic, epigenetic, transcriptome and proteome levels. Novel or known mediators of Kras signaling not found in the transit amplifying population could provide targeting candidates for further exploration.
Targeting EMT to SCC progression. The K15-CrePR; Kras G12D and K15-CrePR; Kras G12D ; p53 ff mouse models are excellent systems to study Kras G12D induced epithelial to mesenchymal transition (EMT). EMT is thought to be a necessary precursor to invasiveness and metastasis, and this process results in the spindle shaped cells of the SCC produced in K15-CrePR; Kras G12D ; p53 ff mice. 30 In both K15-CrePR; Kras G12D and K15-CrePR; Kras G12D ; p53 ff skin, hair follicle stem cells undergo EMT following a brief phase of hyperplasia. This has been concluded by antibody staining of ectopic Tenascin-C, high levels of Vimentin, ectopic Keratin 8 and more recently, by ectopic N-cam staining (unpublished data). By purifying these cells from initiation of hyperplasia through induction of EMT, a detailed transcriptome and proteome profile can be generated. Since our model system can also incorporate a LSL-Yfp allele that generates YFP expression exclusively in Kras G12D expressing cells, we can be confident that cells expressing these markers were once epithelial cells. This is important in order to distinguish them from nearby cancer associated fibroblasts, which express many of the same markers. These data could yield a wealth of information from an in vivo cancer that is undergoing crosstalk with its naturally occurring microenvironment, which contrasts to traditional xenograft studies, which creates an unnatural environment with crosstalk cues that may or may not be truly representative. Theoretically, if EMT can be pharmacologically blocked, the tumor cells may revert to a more keratinocyte-like nature, which could thus be redirected from the path toward squamous cell carcinoma and instead become terminally differentiated skin cells.
Finally, though EMT is found in K15-CrePR; Kras G12D skin, these cells do not proliferate into bona fide SCC. Only in K15-CrePR; Kras G12D ; p53 ff skin do transformed cells undergo a switch to high proliferation and then to SCC development. A comparison between these two mouse models may inform on how this switch occurs. The nature of the pathways induced by Kras G12D in the context of p53KO in hair follicle stem cells may further provide novel targets for reversion back to a non-proliferating cell.
Conclusion
The methodologies outlined here provide a basis for determining potential therapeutic interventions through the detailed molecular understanding of the events that occur in squamous cell carcinoma from genesis to end-point. Squamous cell carcinomas found in patients fall under numerous sub-types, and determination of the human sub-type that most closely resembles that found in K15-CrePR; Kras G12D and K15-CrePR; Kras G12D ; p53 ff mice will be necessary, so that promising therapeutic strategies developed in this preclinical model translate more precisely to the clinic. | 2016-05-12T22:15:10.714Z | 2011-11-01T00:00:00.000 | {
"year": 2011,
"sha1": "cd0f7bd4315b30637ae4a1bc06188ad8513b068e",
"oa_license": "CCBYNC",
"oa_url": "https://www.tandfonline.com/doi/pdf/10.4161/sgtp.18088?needAccess=true",
"oa_status": "BRONZE",
"pdf_src": "PubMedCentral",
"pdf_hash": "cd0f7bd4315b30637ae4a1bc06188ad8513b068e",
"s2fieldsofstudy": [
"Biology",
"Medicine"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
272729467 | pes2o/s2orc | v3-fos-license | Drainage system analysis using Storm Water Management Model …
This research takes place in the North Siantar Subdistrict of Pematangsiantar City, where issues with the drainage system result in frequent flooding in the valley area. Previously, this phenomenon has not been extensively researched. The study aims to analyze infiltration wells on Bah Tongguran Street in the North Siantar Subdistrict of Pematangsiantar City through existing drainage simulation using SWMM software. Water pooling at this location leads to floods due to insufficient drainage capacity, exacerbated by the valley-shaped land contour and high population density. The simulation results, utilizing SWMM software with the Gumbel method for rainfall distribution and a 5-year return period, revealed that the channels on both sides of Bah Tongguran Street cannot accommodate rainwater discharge. The dimensions of the existing drainage channel are b = 0.3 and h = 0.45. To address this issue, we propose the construction of infiltration wells with a diameter of 1.4 meters and a depth of 1.5 meters. The analysis determined that the infiltration wells would be 24 wells on the left and 21 on the right. These infiltration wells will be constructed using concrete pipes, with an effectiveness rate of 65% for the left side and 58% for the right. Through this intervention, our research aims to serve as a reference for planners designing infiltration wells in valley areas and as a resource for other researchers developing sustainable flood mitigation technologies.
INTRODUCTION
Until now, no one has discussed infiltration wells as an alternative to flooding in valley areas inhabited by people.Essentially, valley areas have lower contours compared to the surrounding regions.Previous studies on planning infiltration wells are generally common, such as those conducted by Zainal [1] in residential areas, Molya [2] as an alternative to Eco-Drainage, and Nyoto [3] in the city of Jakarta.These studies focused on general research areas and were not specific to valley regions.
Flood management in valley areas requires a nuanced understanding of drainage capacity, Juliastuti [4], vulnerability assessment, Eryani [5], and real-time monitoring, Adriansyah [6].Research has shown that deficiencies in drainage capacity, such as those identified in the analysis of residential drainage systems, highlight the need for alternative solutions like infiltration wells.Studies on flood vulnerability, using various indicators, provide insights into how environmental factors affect flood risks, which is crucial for effective well placement.
Additionally, advancements in flood early warning systems, utilizing wireless sensors, can enhance the management of flood events, informing the strategic placement and operation of infiltration wells.Soil infiltration rate tests, such as those conducted in the Kathmandu Valley, further underline the importance of local conditions in determining well effectiveness.Integrating these approaches can optimize flood management strategies in valley regions, addressing the unique challenges posed by their topography [7].Additionally, there is a scientific report on hydrogeological research in the southeastern part of New York.This research, conducted by the U.S. Geological Survey (USGS) in collaboration with the New York State Department of Environmental Conservation and the New York State Department of Health, aims to protect groundwater sources used for public water supply in several villages and communities in the region [8].
However, there is still no research specifically addressing the use of infiltration wells in valley areas to mitigate floods.These studies are interconnected, as the use of infiltration wells is influenced by soil permeability.If soil infiltration conditions are impermeable, the use of infiltration well technology cannot be applied.Moreover, if groundwater conditions are less than half a meter from the soil surface, infiltration wells will not be effective.Furthermore, if the groundwater level is too high, infiltration wells run the risk of being contaminated by surface water that may be polluted by waste or pollutants [9].
The study was carried out in the North Siantar Subdistrict, Pematangsiantar City, specifically chosen to assess drainage channels in response to recurrent waterlogging and flooding.Based on the geographic location of the study in Figure 1, the city of Pematangsiantar is situated between the latitudes 2°53'20" -3°01'00" N and the longitudes 99°1'10" -99°6'35" E. The area of study for this research is in the Siantar Utara District of Pematangsiantar, covering an area of 3.65 km².Furthermore, the valley that experienced flooding in this study is situated in the Siantar Utara District, on Bah Tongguran Street.At this location, there are tertiary drainage channels on both the right and left sides of the road, with dimensions of width = 0.30 m and height = 0.45 m, and a channel length of 208.66 m, encompassing an area of 1.84 hectares.This area is designated for residential use.
The areas prone to flooding in the North Siantar Subdistrict include Bah Tongguran Street.Residents living around Bah Tongguran Street reported that the water level reached the calf of an adult, approximately 15-30 cm, and the flooding was characterized by stagnation without any flow.Flooding in drainage channels resulted in road closures and disruptions to passing vehicles, endangering road users.Figure 1.Location of Study [10] Furthermore, it forced businesses around the area to close to prevent water damage, leading to financial losses.Citizens said that the flood depth on Bah Tongguran Street ranged from 15-20 cm, covering an area of 296 m².
The North Siantar Subdistrict is a densely populated commercial area with drainage issues that need to be addressed to prevent floodrelated losses and disruptions to community activities.Therefore, this study aims to evaluate the performance of the drainage system by conducting hydrological and hydraulic analyses and simulating flow profiles using Storm Water Management Model 5.1 (SWMM) software to determine whether the capacity of the drainage channels is still able to accommodate the generated flow and runoff.This research also aims to identify methods for managing excess water and addressing flood issues in flood-prone areas of the North Siantar Subdistrict.
METHOD
The methodology of this research involves hydrological analysis, which includes calculating the average rainfall using an algebraic method with a 5-year return period [11].
The research flowchart shown in Figure 2 begins with problem formulation, followed by a literature review and data collection (both primary and secondary), which form the basis for the analysis of rainfall, flood discharge, and channel dimensions.Subsequently, an evaluation using SWMM determines the need for channel dimension improvements or infiltration well planning.If no issues are identified, conclusions and recommendations are developed.The flowchart for SWMM 5.1 modeling in Figure 3 begins with inputting data on area, elevation, and channel characteristics, followed by adding a backdrop image and setting object properties.Next, the physical network is drawn, component data is entered, and analysis settings are selected.Run the simulation; if the error is <5%, re-enter the data.If not, review the simulation results and redesign the channel if node flooding occurs.
Analyzing rainfall frequency using normal, log-normal, Log Pearson III, and Gumbel methods, calculating rainfall intensity using the monobe method, and determining the design flood discharge using the rational method [12].Hydraulic analysis is conducted by calculating the capacity of the drainage channels and simulating the hydrological and hydraulic processes using the Storm Water Management Model (SWMM) [13] software.Furthermore, improvements will be made to the flooded channels by redesigning the drainage system, and planning infiltration wells for these improvements.
RESULTS AND DISCUSSION Validation of Rainfall
Based on the calculation results in Table 1, the value of Q/√n [14] < critical Q is 0.69 < 1.14 and the value of R/√n [14] < critical Q is 0.93 < 1.28 with a 95% probability [15].Therefore, the rainfall data for Bangun and Marihat is considered valid.
Rainfall Distribution
Based on the calculation results of rainfall dispersion using the algebraic method in Table 2, it is concluded that the distribution types meeting the criteria are Gumbel and Log Pearson III.Gumbel's criteria are Cs ≤ 1.14 (0.63) and Ck ≤ 5.4 (3.37), while Log Pearson III meets additional unspecified values.
Distribution Goodness-of-Fit Test
On the one hand, in Table 3, the Gumbel and Log Pearson III distributions both meet the criteria, with a calculated maximum ΔP of 3 against a critical maximum ΔP of 5.99.Therefore, both distributions pass the Chi-Square goodnessof-fit test.Furthermore, in Table 4, the Gumbel distribution meets the criteria with a calculated maximum ΔP of 0.063 against a critical maximum ΔP of 0.41.In contrast, the Log Pearson III distribution does not meet the criteria, with a calculated maximum ΔP of 0.895 exceeding the critical maximum ΔP of 0.41 [18] [19].
Calculation of Rainfall Intensity
Calculation of rainfall intensity using the data that passed the previous test, which is the Gumbel distribution, using the Mononobe method [20].
Runoff Coefficient
The coefficient of runoff (C) is the ratio of peak runoff to rainfall intensity.The value of C is influenced by several main factors, including soil infiltration rate, slope of the land, vegetation cover, and rainfall intensity [22] [23].The flooded locations on Bah Tongguran roads consist of various land covers such as urban areas, residential areas, yards, and asphalt roads, with respective areas as shown in Table 6.Based on the calculation of the coefficient of runoff in Table 6, the coefficient of runoff for the left and right segments of Bah Tongguran road is 0.52 and 0.54, respectively.
Calculation of Planned Discharge
Based on the calculated intensity in Table 7, the planned discharge for the left segment of Bah Tongguran road is 0.093 m³/s, with an intensity of 713.11, an area of 0.0091 km², and a C value of 0.52.The right segment's planned discharge is 0.079 m³/s, with an intensity of 561.58, an area of 0.0093 km², and a C value of 0.54.
Hydraulic Analysis
The hydraulic analysis calculates the discharge capacity of the main drainage channel on Bah Tongguran Road, as illustrated in Table 8.For the left segment, the planned discharge (Qplan) is 0.093 m³/s, and the channel discharge (Qchannel) is 0.24 m³/s, indicating no flood.For the right segment, Qplan is 0.079 m³/s, and Qchannel is 0.17 m³/s, also indicating no flood.However, despite the channel's capacity, flooding occurs in Bah Tongguran because the valley area prevents water from flowing to the river.
Modeling of Drainage Channel using Storm Water Management Model (SWMM)
The drainage channel modeling for Bah Tongguran Street is carried out with a 5-year return period.It consists of 2 sub catchments, 10 junctions, 10 conduits, and 2 outfalls.Subcatchments can be represented using the subcatchment button on the object toolbar.Based on the land use, the catchment area of Bah Tongguran Street consists of residential areas with asphalt roads.The N-impervious value for non-absorbent areas is 0.011, while the N-previous value for nonwaterproof areas is 0.15, representing sparsely vegetated grassy land.The top view of the drainage network schema for Bah Tongguran Street can be seen in Figure 4.The right segment is illustrated at points JC2, JC9, JC3, and JC4 from left to right, while the left segment is illustrated at points JC6, JC10, JC7, and JC8.Additionally, Figure 5 shows that at node JC7, which is in the valley and part of the main channel of the left segment of the drainage system on Bah Tongguran Street, this section is unable to handle the water flow during the 5-year return period.This occurs because the location is in a valley or concave contour, causing the water to be unable to flow and accumulate in that lowlying area.On the other hand, Figure 6 shows that node JC3 in the main channel of the right segment of the drainage system on Bah Tongguran Street, it also cannot accommodate water during the 5-year return period.This is also due to the location being in a valley or contour resembling a basin, causing water to stagnate and accumulate in that valley area.Furthermore, there is no main river to carry the water away from nodes JC3 and JC7.
Handling Alternative for Waterlogging Using Infiltration Wells
In this research, there is a flood location point on Bah Tongguran Road, which has a topography like a basin or valley, making it impossible to redesign the drainage channel in that area to prevent floods because water will continue to accumulate in the valley area.Based on hydraulic calculations, the channel's flow rate (Q) is found to be higher than the Q planned flow rate, so the dimensions of the drainage channel are not the cause of the floods on Bah Tongguran Road.Therefore, improvement can be made by planning infiltration wells to collect and infiltrate water back into the ground, thereby reducing the burden on the drainage system.The planned infiltration wells on Bah Tongguran Road using the Sunjoto method [26] are as follows: a. Shallow infiltration wells have a diameter of 1.4 m. b.The height of the infiltration wells is 1.5 m. c.The well walls use concrete pipes with a wall thickness of 10 cm.d.The cover of the infiltration wells uses concrete pipes with a thickness of 10 cm.e.The infiltration wells are filled with coconut coir (ijuk) 10 cm and gravel 20 cm to filter the water that will be absorbed into the ground [27].f.The inlet pipe size is 110 mm in diameter.g.The overflow pipe size is 110 mm in diameter.h.The control basin uses square-shaped brick pairs with a width of 30 cm, a depth of 35 cm, and a brick thickness of 10 cm.
Type of Infiltration Wells
The detailed top view of the house featuring an infiltration well can be seen in Figure 7.The water entering the well is rainwater conveyed from the house gutter.The horizontal gutter is a half-circle with a diameter of 8 inches, and the vertical gutter has a diameter of 4 inches.The planned layout determination is adjusted according to the Indonesian National Standard (SNI) 03-2453-2002 [28]: a. Distance from the house foundation to the infiltration well is 2 meters.b.Distance from the foundation to the control tank is 100 cm.c.Distance from the control tank to the infiltration well is 50 cm.d.Distance from the infiltration well to the tree is 3 meters.Based on Table 9, it is known that infiltration wells are planned with a diameter of 1.4 m and a depth of 1.5 m, totaling 24 wells on the left segment and 21 wells on the right segment of Bah Tongguran Road.
Comparison of research results with previous studies
In contrast to previous research by Zainal [1], Molya [2], and Nyoto [3], which focused on infiltration wells in general urban or residential contexts, this study specifically highlights the unique challenges in valley areas, where conventional drainage systems are inadequate due to topographical constraints.While earlier studies established the basic effectiveness of infiltration wells in enhancing eco-drainage systems and urban flood management, they did not delve into the complexities posed by valley regions.For instance, Zainal's [1] study in residential areas did not consider the impact of topographical depressions on flood accumulation, nor did it address the specific design adaptations required for such environments.Similarly, Molya's [2] work on eco-drainage as an alternative solution did not tackle the particular issue of water stagnation in valleys, where water cannot easily escape due to the natural contour of the land.Additionally, Juliastuti's [4] research offered a nuanced understanding of drainage capacity but remained limited to urban contexts, lacking consideration of the more complex drainage challenges in valley areas.
The findings from this research, therefore, contribute a critical perspective on the application of infiltration wells in valley regions.Unlike general applications, the effectiveness of infiltration wells in valleys depends heavily on localized conditions, including soil permeability and groundwater levels.This research has demonstrated that with proper design adjustments, such as those applied on Bah Tongguran Road, infiltration wells can significantly mitigate flood risks even in challenging topographical conditions.
However, this study also highlights limitations not typically encountered in less topographically constrained areas, such as the risk of ineffective infiltration due to high groundwater levels or impermeable soil conditions, as noted in hydrogeological studies like those conducted by the USGS in New York.
CONCLUSION
The conclusion drawn from this study is based on hydrological and hydraulic calculations, as well as modeling using SWMM software.There are two channels identified that cannot accommodate rainfall discharge, namely the left and right segments of Bah Tongguran Road.The cross-sectional capacity of the channels in Siantar Utara District was determined using hydraulic calculations.The left and right segments of Bah Tongguran Road have capacities of 0.2369 m3/s and 0.1680 m3/s, respectively.
The flood control alternative for the left and right segments of Bah Tongguran Road involves planning the construction of 24 infiltration wells with a diameter of 1.4 m and a depth of 1.5 m in the left segment, and 21 wells in the right segment.The infiltration wells are planned using concrete pipes with an effectiveness of 65% in the left segment and 58% in the right segment, as shown in Table 10.
Figure 8 illustrates the site plan for the placement of infiltration well points, which are connected to the houses on Bah Tongguran Road.Additionally, Figure 9 shows the crosssection of the infiltration well.It can be seen that rainwater falling from the roof will be channeled through a downspout into a control tank, where roof sediments can first settle in the control tank.Subsequently, clean water flowing from the control tank is directed into the infiltration well.
From the infiltration well, the water will seep into the ground and become part of the groundwater storage.However, if there is an overflow in the infiltration well, the excess water will flow into the drainage system with the hope that it will gradually dry out as the drainage water returns to the infiltration well.
This study highlights the critical need for tailored design adaptations of infiltration wells in valley areas, addressing topographical challenges that were overlooked in previous research.While earlier studies confirmed the general effectiveness of infiltration wells, they did not consider the complexities unique to valleys, such as high groundwater levels and impermeable soils.This research demonstrates that with proper adjustments, these wells can significantly mitigate flood risks in challenging terrains.Future research should focus on developing advanced design models specifically for valley regions and integrating real-time monitoring systems to optimize the performance and adaptability of infiltration wells in diverse environmental conditions. of Jakarta and surrounding area," IOSR Journal of Engineering (IOSR-JEN), vol.12, no.
Figure 6 .
Figure 6.Running SWMM for the 5-year return period at nodes JC2, JC9, JC3, and JC4 sequentially from the left side of the figure
Figure 7 .
Figure 7. Top View of Infiltration Well Planning
Figure 2 .
Figure 2. Site Plan of Infiltration WellEryani[5] research on vulnerability assessment also underscores the importance of in-depth contextual analysis, particularly in identifying areas most susceptible to flooding in valley regions.Furthermore, Adriansyah[6] research focused on real-time monitoring using IoT technology has shown that the use of such technology can provide critical real-time data to monitor the effectiveness of infiltration wells in valley areas, aiding in more timely and accurate decision-making.
Figure 3 .
Figure 3. Cross Section of Infiltration Well
Table 1 .
Results of Validation Calculation for Rainfall Data in Bangun and Marihat
Table 2 .
Results of Rainfall Dispersion Calculation
Table 3 .
Summary of Chi-Square Goodness-of-Fit Test
Table 5 .
Results of Rainfall Intensity Calculation using the Mononobe method
Table 5
presents rainfall intensity calculated using data validated by the Gumbel distribution test, employing the Mononobe method.For various return periods, the intensity (in mm/hour) varies across durations: for 2 minutes, it ranges from 60.9 (2-year return) to 111.4 (100-year return); for 5 minutes, from 110.7 to 202.5; and for 10 minutes, from 69.7 to 127.5.Longer durations show corresponding intensities: 15 minutes (53.2 to 97.3), 30 minutes (33.5 to 61.3), 60 minutes (25.6 to 46.8), 120 minutes (21.1 to 38.6), 720 minutes (12 to 21.9), and 1440 minutes (10.1 to 18.61).These values are crucial for estimating rainfall intensity across different durations and essential for urban drainage planning and flood management.The Mononobe method's results are utilized directly in SWMM 5.1's rain gauge application, ensuring accurate simulation and management of stormwater in urban environments based on reliable intensity data.
Table 6 .
Area and Coefficient of Runoff for Each Land Cover in Bah Tongguran
Table 7 .
Planned Discharge of Drainage Channel in Bah Tongguran | 2024-09-20T15:22:02.589Z | 2024-07-29T00:00:00.000 | {
"year": 2024,
"sha1": "70d0b676018cdcc5bf7186d21f8e2ce0657effe2",
"oa_license": "CCBYNC",
"oa_url": "https://doi.org/10.22441/sinergi.2024.3.014",
"oa_status": "GOLD",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "cef4c7f15d0f2dadc052a899ecb4b1c27c099d1f",
"s2fieldsofstudy": [
"Environmental Science",
"Engineering"
],
"extfieldsofstudy": []
} |
251697489 | pes2o/s2orc | v3-fos-license | Evaluation of the Effect of High-Speed Sintering and Specimen Thickness on the Properties of 5 mol% Yttria-Stabilized Dental Zirconia Sintered Bodies
High-speed sintering of zirconia has become essential to single-visit dental prosthetic treatments. This important prosthetic dentistry technique demands a translucent material tougher than porcelain. Previous studies on high-speed sintered zirconia did not take heat and material thickness into consideration. We evaluated pre-sintered specimen thickness and the effect of high-speed sintering on the properties of 5 mol% Y2O3-stabilized zirconia (5Y zirconia). High-speed sintered bodies of 5Y zirconia were evaluated by density measurements, translucency measurements, three-point flexural and fracture toughness tests, X-ray diffraction (XRD), and scanning electron microscopy (SEM). High-speed sintering reduced the translucency and mechanical properties of 5Y zirconia. XRD and SEM observation results clarified that these reductions were due to the change in crystal phase composition and to the increase in residual pores, respectively, both resulting from high-speed sintering. Moreover, in high-speed sintering, as the thickness of the specimen increased, the number and size of internal pores increased, and the translucency and strength decreased. The threshold value for avoiding a reduction in translucency and mechanical properties was found to lie at ~4.4 mm. From the above results, it was concluded that 5Y zirconia is not suitable for high-speed sintering applications.
Introduction
Dental prostheses are commonly produced from zirconia due to its high bending strength, fracture toughness, chemical stability, radiopacity, and biocompatibility [1,2]. The excellent mechanical properties of the zirconia sintered body are due to the toughening mechanism associated with stress-induced phase transformation from the metastable tetragonal phase to the stable monoclinic phase [3]. Therefore, in order to obtain excellent mechanical properties, the tetragonal phase must be retained at room temperature and stabilizers (i.e., CaO, MgO, CeO 2 , and Y 2 O 3 ) must be added to the raw zirconia powder. Of these stabilizers, Y 2 O 3 is used in most dental zirconia. The stabilizer concentration greatly influences the properties of zirconia. In general, the higher the concentration of the stabilizer, the higher the translucency, and conversely, the lower the mechanical properties [4,5]. However, Kulyk et al. [6,7] reported that mechanical properties showed ambiguous changes with respect to yttrium concentration. This difference is likely due to differences in the method of adding yttria to zirconia. In general, yttria is added by dissolving the yttrium source in a precursor solution of zirconia, whereas in [6,7], yttrium was added as a powder. The powder method results in yttrium non-uniformity and could be the cause of ambiguous changes in mechanical properties. These reports suggest that not only the yttrium concentration of the entire sintered body but also the local yttrium concentration and the accompanying crystal phase composition and microstructure affect the mechanical properties. Table 2. Specimen dimensions in each test (the mark "ϕ" denotes "diameter").
Before Sintering At The Time of Measurement
Density ϕ18 mm × 18 mm ϕ15 mm × 15 mm XRD ϕ18 mm × 1.9 mm and 18 mm ϕ15 mm × 1.0 mm SEM ϕ18 mm × Zpex Smile (Tosoh, Tokyo, Japan) were used as raw materials for 5Y zirconia containing 0.05 wt% of alumina. The raw material was press-molded by a uniaxial press molding machine (Sansho Industry, OSAKA, JAPAN) to obtain a block having a diameter of 19 mm × 19 mm and a disc having a diameter of 100 mm × 17 mm. After that, CIP treatment was performed at 200 MPa using a cold hydrostatic isotropic press (Sansho Industry, Osaka, Japan).
The semi-sintered zirconia block/disk was cut with a cutting machine (Secotom, Struers, Ballerup, Denmark), and the surface was polished with SiC water-resistant polishing paper (# 1000), to reach a shape suitable for each test.
Sintering of Test Piece
Each test piece was subjected to high-speed sintering (SS: 1500 • C, 5 min) or conventional sintering (CS: 1500 • C, 2 h). CEREC Speed fire (Dentsply Sirona, Charlotte, NC, USA) was used for high-speed sintering, and AUSTROMAT 674i (DEKEMA, Freilassing, Germany) was used for conventional sintering. Each sintering process was performed according to the schedule shown in Figure 2.
Processing of Sintered Body
Each sintered body was ground by a surface grinding machine (GRIND-X PFG500II, Each sintered body was ground by a surface grinding machine (GRIND-X PFG500II, OKAMOTO, Gunma, Japan) using diamond wheels (# 325, # 800) and processed to the dimensions corresponding to each test ( Figure 1 and Table 2). Specimens for XRD and SEM were polished with dental zirconia polishing paste Zirco Shine (SHOFU INC., Kyoto, Japan), and for the three-point flexural test, they were polished with water-resistant SiC polishing paper (# 1000). Specimens for XRD and SEM analyses were then heat-treated at 1300 • C for 5 min, to remove the effects of phase transformations and residual stresses due to mechanical polishing.
Density Measurement
The density of the specimens was measured by the Archimedes method (n = 3). The measurement was carried out in accordance with JIS R 1634: 1998. From the obtained density, the relative density was calculated by the following equation: A: content of alumina (wt%); P: measured density (g/cm 3 ); ρ 0 : theoretical density (g/cm 3 ); ρ A : theoretical density of alumina; ρ x : theoretical density of zirconia containing 5.5 mol% of yttria (6.0254 g/cm 3 [17]).
Crystallographic Characterization
Crystal structure characterization of the specimens was performed by X-ray Diffraction (Multi Flex, Rigaku, Tokyo, Japan). The measurement was performed using CuKα as the X-ray source, tube current and tube voltage of 40 mA and 40 kV, respectively, 2θ range of 20-120 • , scan step of 0.02 • . A Rietveld analysis was performed for each diffraction profile using Rietan-FP [18]. For the initial structural model of Rietveld analysis, two types of tetragonal phases with different yttrium contents were used as described in a report [19] by Belli et al. (Table 3).
The tetragonality c/a was calculated from the lattice constant refined by Rietveld analysis. The yttria concentration in tetragonal zirconia was calculated using the following equations, as given by Miller et al. [20]:
Microscopic Characterization
Microscopic characterization was performed by SEM (JSM6390LA, JEOL, Tokyo, Japan) at an acceleration voltage of 15 kV. The specimens used for XRD were the same as those used for SEM observation. The specimens were Au-coated to prevent charge-up.
Translucency
The translucency was evaluated by computing the TP (Translucency Parameter) according to the following equation: L*, a*, b*: CIE L* a* b* value; Subscript w: measured with the white background; Subscript b: measured with the black background. The L* a* b* of each specimen was measured by a spectroscopic colorimeter (CM-5, Konica Minolta, Hannover, Germany). The specimens were measured on a black and white background (n = 2).
Three-Point Flexural Test
The three-point flexural test was conducted in accordance with ISO6872 (Dentistry-Ceramic materials) (n = 10). In order to investigate the effect of the specimen pre-sintered thickness on the strength, specimens made from two types of semi-sintered bodies with different thicknesses were tested ( Figure 1). The test was carried out using a universal testing machine ( P: breaking load (N); L: test span (mm); w: width of the specimen (mm); b: thickness of the specimen (mm).
Fracture Toughness Test
The fracture toughness test used the SEVNB method in accordance with ISO 6872 (n = 5). In this test, two types of specimens with different thicknesses (22 mm (l) × 4.0 mm (w) × 3.0 mm (b) and 22 mm (l) × 4.0 mm (w) × 1.5 mm (b)) were used in order to investigate the effect of the specimen pre-sintered thickness on the fracture toughness. The V-notch was applied along the thickness direction of each test piece to a depth of 0.8 to 1.2 mm, an angle of less than 20 • , and a tip diameter of less than 20 µm using a razor blade (FH-10B, FEATHER, Osaka, Japan). The test was carried out using a universal testing machine (Instron5967, INSTRON, Norwood, MA, USA) under the conditions of a span of 16 mm and a crosshead speed of 0.5 mm/min. The fracture toughness value, K 1C (MPa √ m), was calculated according to the following equation, in accordance with ISO 6872 [23]: F: fracture load (MN); S: test span (mm); w: specimen thickness (mm); b: specimen width (mm); α: relative V-notch depth (mm). α = a w a: average notch depth (mm); a = a1 + a2 + a3 3
Statistical Analysis
The t-test was used to compare 5Y-SS and 5Y-CS in relative density and TP (specimen pre-sintered thickness 1.9 mm). Flexural strength and fracture toughness were analyzed using two-way analysis of variance (two-way ANOVA). The level of significance was set at 0.05 for all analyses. Figure 3 shows the relative densities of 5Y-SS-18 and 5Y-CS-18. The relative density of 5Y-SS was significantly lower than that of 5Y-CS. This result indicates that more pores remain in 5Y-SS than in 5Y-CS, suggesting that in 5Y zirconia, high-speed sintering results in inadequate densification. Figure 4 shows the XRD patterns of 5Y-SS-1.9 and 5Y-CS-1.9. Both 5Y-SS and 5Y-CS could be identified as the tetragonal phases, high yttrium tetragonal and low yttrium tetragonal. Table 4 shows the weight ratios of the two tetragonal phases and their yttrium concentrations. The crystal phase composition of 5Y-CS is in good agreement with the values previously reported by Belli et al. [19]. In 5Y-SS, however, the difference in yttria concentration of the two tetragonal phases was smaller and their content difference was larger than that of 5Y-CS. This is considered to be due to the high heating rate and the short holding time. Matsui et al. [24] reported that in the sintering of 3Y zirconia, yttrium segregates at the grain boundaries at 1300 °C or higher and two regions with yttrium concentration of 6 to 7 mol% and ~2 mol% are formed at 1500 °C. Similar phase separation has been confirmed in 4Y zirconia [21]. Considering that yttrium segregation and phase separation occur in 5Y zirconia by the same mechanism, in 5Y-SS, yttrium segregation and phase separation are suppressed because of the high heating rate and short holding time (5 min). As a result, it is probable that the crystal phase composition of 5Y-SS differed from that of 5Y-CS.
Crystallographic Characterization
As previously mentioned, Cokic et al. reported results that differed from those of the present study regarding the effect of high-speed sintering on the crystal phase composition of 5Y zirconia [14]. In their study, high-speed sintering did not affect the crystalline phase composition of 5Y zirconia. Again, this could be due to the different types of 5Y zirconia semi-sintered bodies used (cf. Section 3.1). In high-speed sintering, the Contrary to this study, Cokic et al. [14] reported that fast sintering did not affect the density of 5Y zirconia. The cause of this difference is considered to be the difference in sample size and the difference in the type of 5Y zirconia. In this study, thick specimens of ϕ15 mm × 15 mm (ϕ18 × 18 mm before sintering) were used for density measurement, whereas the aforementioned researchers used thin specimens, 12 mm × 12 mm × 0.5 mm (15 mm × 15 mm × 3.5 mm before sintering). High-speed sintered 5Y zirconia increases residual pores as the specimen pre-sintered thickness increases (see Section 3.3). Therefore, the difference between this study and the study by Cokic et al. could be due to the difference in specimen thickness. In addition, while Zpex Smile was used as 5Y zirconia in this study, they used Katana STML (Kuraray Noritake, Tokyo, Japan). These types of zirconia semi-sintered bodies differ in zirconia primary particle size, the type and concentration of additives (i.e., alumina), and the uniformity of yttrium, which in turn affect the sintering behavior. Figure 4 shows the XRD patterns of 5Y-SS-1.9 and 5Y-CS-1.9. Both 5Y-SS and 5Y-CS could be identified as the tetragonal phases, high yttrium tetragonal and low yttrium tetragonal. Table 4 shows the weight ratios of the two tetragonal phases and their yttrium concentrations. The crystal phase composition of 5Y-CS is in good agreement with the values previously reported by Belli et al. [19]. In 5Y-SS, however, the difference in yttria concentration of the two tetragonal phases was smaller and their content difference was larger than that of 5Y-CS. This is considered to be due to the high heating rate and the short holding time. Matsui et al. [24] reported that in the sintering of 3Y zirconia, yttrium segregates at the grain boundaries at 1300 • C or higher and two regions with yttrium concentration of 6 to 7 mol% and~2 mol% are formed at 1500 • C. Similar phase separation has been confirmed in 4Y zirconia [21]. Considering that yttrium segregation and phase separation occur in 5Y zirconia by the same mechanism, in 5Y-SS, yttrium segregation and phase separation are suppressed because of the high heating rate and short holding time (5 min). As a result, it is probable that the crystal phase composition of 5Y-SS differed from that of 5Y-CS. (2) and (3)) and phase contents (Rietveld). Figure 5 shows the X-ray diffraction patterns of the specimens with thicknesses of 1.9 mm and 18 mm before sintering. In both 5Y-SS and 5Y-CS, no change was observed in the X-ray diffraction pattern depending on the specimen pre-sintered thickness. This indicates that the thickness of the specimen does not affect the crystal phase in any of the sintering methods. (2) and (3)) and phase contents (Rietveld).
5Y-SS 5Y-CS Previous Report [19]
high yttrium tetragonal tetragonality As previously mentioned, Cokic et al. reported results that differed from those of the present study regarding the effect of high-speed sintering on the crystal phase composition of 5Y zirconia [14]. In their study, high-speed sintering did not affect the crystalline phase composition of 5Y zirconia. Again, this could be due to the different types of 5Y zirconia semi-sintered bodies used (cf. Section 3.1). In high-speed sintering, the high-temperature interval of time, in which the diffusion of ions is promoted, is short, and the state of the material before sintering could greatly affect post-sintering characteristics. Figure 5 shows the X-ray diffraction patterns of the specimens with thicknesses of 1.9 mm and 18 mm before sintering. In both 5Y-SS and 5Y-CS, no change was observed in the X-ray diffraction pattern depending on the specimen pre-sintered thickness. This indicates that the thickness of the specimen does not affect the crystal phase in any of the sintering methods. Figure 6 shows SEM images of 5Y-SS-1.9 and 5Y-CS-1.9. Almost no pores were confirmed in 5Y-CS, whereas some pores were confirmed in 5Y-SS (arrow in Figure 6a). Possibly, this is because the pores were not discharged and remained due to the rapid densification and grain growth associated with high-speed sintering. In addition, the median diameter of 5Y-SS grains was larger than that of 5Y-CS grains. In particular, the number of particles 4 μm or larger increased remarkably, suggesting non-uniform grain growth. It is known that the grain size of zirconia ceramics is affected by sintering conditions such as sintering temperature and holding time [6,7,25,26]. In many reports, the lower the sintering temperature and the shorter the holding time, the smaller the grain size [27][28][29]. Similar to the results of this study, Cokic et al. reported that the grain size of 5Y zirconia increased by high-speed sintering [14]. The cause of these differences in grain growth behavior is not clear, but since only Cokic et al. used the same sintering furnace (Speedfire) as in this study, it is possible that the grain-coarsening observed in 5Y-SS is a phenomenon specific to Speedfire. Figure 6 shows SEM images of 5Y-SS-1.9 and 5Y-CS-1.9. Almost no pores were confirmed in 5Y-CS, whereas some pores were confirmed in 5Y-SS (arrow in Figure 6a). Possibly, this is because the pores were not discharged and remained due to the rapid densification and grain growth associated with high-speed sintering. In addition, the median diameter of 5Y-SS grains was larger than that of 5Y-CS grains. In particular, the number of particles 4 µm or larger increased remarkably, suggesting non-uniform grain growth. It is known that the grain size of zirconia ceramics is affected by sintering conditions such as sintering temperature and holding time [6,7,25,26]. In many reports, the lower the sintering temperature and the shorter the holding time, the smaller the grain size [27][28][29]. Similar to the results of this study, Cokic et al. reported that the grain size of 5Y zirconia increased by high-speed sintering [14]. The cause of these differences in grain growth behavior is not clear, but since only Cokic et al. used the same sintering furnace (Speedfire) as in this study, it is possible that the grain-coarsening observed in 5Y-SS is a phenomenon specific to Speedfire. Materials 2022, 15, x FOR PEER REVIEW 11 of 20 Figure 7 shows the SEM images of 5Y-SS and 5Y-CS with a specimen thickness of 1.9 mm, 4.0 mm, 8.0 mm, and 18 mm before sintering. Figure 8 shows the relationship between the specimen pre-sintered thickness and the number and diameter of pores. Counting the number of pores in the SEM image in each of the three views of each specimen, calculating the average, and converting it per unit area was the method used for pore number calculations. In both 5Y-SS and 5Y-CS, the pores increased as the thickness of the pre-sintered specimen increased, especially in 5Y-SS. This indicates that the core region of the thick specimen is not sufficiently densified, especially in 5Y-SS. In addition, an increase in pore size was observed in 5Y-SS-8.0, 5Y-SS-18, and 5Y-CS-18. Figure 8 shows the relationship between the specimen pre-sintered thickness and the number and diameter of pores. Counting the number of pores in the SEM image in each of the three views of each specimen, calculating the average, and converting it per unit area was the method used for pore number calculations. In both 5Y-SS and 5Y-CS, the pores increased as the thickness of the pre-sintered specimen increased, especially in 5Y-SS. This indicates that the core region of the thick specimen is not sufficiently densified, especially in 5Y-SS. In addition, an increase in pore size was observed in 5Y-SS-8.0, 5Y-SS-18, and 5Y-CS-18. Figure 9 shows the relationship between the specimen's pre-sintered thickness and the crystal grain size. In 5Y-CS, the grain size did not correspond to the thickness, whereas in 5Y-SS, the grain size increased as the specimen thickness increased. This suggests that in 5Y-SS, ion diffusion is promoted inside the specimen rather than near the surface. This could be because the temperature inside 5Y-SS is higher than that of the surface of the specimen. Since the Speedfire used for sintering 5Y-SS is a microwave heating furnace, the specimen itself generates heat during sintering, and there is no heat source around the test piece. Therefore, the temperature inside the test piece is inevitably higher than that near the surface, known as the "inverse temperature gradient" [30]. As a result, it is considered that the central part of the specimen was rapidly densified and grain-grown as compared with the surface, and a large number of pores were generated in that area. Charmond et al. also reported non-uniform grain growth in microwave sintering [31]. Figure 9 shows the relationship between the specimen's pre-sintered thickness and the crystal grain size. In 5Y-CS, the grain size did not correspond to the thickness, whereas in 5Y-SS, the grain size increased as the specimen thickness increased. This suggests that in 5Y-SS, ion diffusion is promoted inside the specimen rather than near the surface. This could be because the temperature inside 5Y-SS is higher than that of the surface of the specimen. Since the Speedfire used for sintering 5Y-SS is a microwave heating furnace, the specimen itself generates heat during sintering, and there is no heat source around the test piece. Therefore, the temperature inside the test piece is inevitably higher than that near the surface, known as the "inverse temperature gradient" [30]. As a result, it is considered that the central part of the specimen was rapidly densified and grain-grown as compared with the surface, and a large number of pores were generated in that area. Charmond et al. also reported non-uniform grain growth in microwave sintering [31].
Microscopic Characterization
Since only a small amount of alumina was added to the zirconia raw material used in this study, precipitation of alumina particles was not observed in any specimen. The solubility of alumina in 8 mol% yttria-stabilized zirconia is reported to be ~1 wt% at 1500 °C [32]. Although 8Y zirconia and 5Y zirconia may have different alumina solubility limits, it is considered that alumina is completely dissolved at the grain boundary of zirconia, because the concentration of alumina in this study is about 0.05 wt%, which is much lower than the solubility limit. It has been reported that, in polycrystalline zirconia, the dissolved alumina is mainly segregated at grain boundaries [33]. Therefore, we shall consider alumina as segregated at the grain boundaries in all specimens in this study. Figure 10 shows the TP of 5Y-SS and 5Y-CS. The TP of 5Y-SS was significantly lower than that of 5Y-CS, suggesting that high-speed sintering reduces the translucency of 5Y zirconia. The residual pores can be considered the main cause of this decrease in translucency. Density measurements and SEM observations suggest that 5Y-SS has more residual pores than 5Y-CS. It is probable that these residual pores became a light-scattering source and reduced the translucency. The translucency of zirconia is strongly influenced by the size and concentration of pores due to zirconia's high refractive index [34].
Translucency
The crystal phase composition is considered a factor affecting the translucency. Ceramic polycrystals generally exhibit higher translucency as they contain more optically isotropic crystal phases. From the XRD results, both 5Y-SS and 5Y-CS consist of tetragonal phases that are optically anisotropic. In the tetragonal phase, the lower the tetragonality, the closer to the optically isotropic cubic phase the sample becomes. Therefore, the light scattering due to optical anisotropy is reduced.
Zhang et al. compared the translucency of a 5Y zirconia sintered body prepared by mixing 3Y zirconia powder and 8Y zirconia powder with the 5Y zirconia sintered body prepared from the uniform 5Y zirconia powder, and reported that the former was less Since only a small amount of alumina was added to the zirconia raw material used in this study, precipitation of alumina particles was not observed in any specimen. The solubility of alumina in 8 mol% yttria-stabilized zirconia is reported to be~1 wt% at 1500 • C [32]. Although 8Y zirconia and 5Y zirconia may have different alumina solubility limits, it is considered that alumina is completely dissolved at the grain boundary of zirconia, because the concentration of alumina in this study is about 0.05 wt%, which is much lower than the solubility limit. It has been reported that, in polycrystalline zirconia, the dissolved alumina is mainly segregated at grain boundaries [33]. Therefore, we shall consider alumina as segregated at the grain boundaries in all specimens in this study. Figure 10 shows the TP of 5Y-SS and 5Y-CS. The TP of 5Y-SS was significantly lower than that of 5Y-CS, suggesting that high-speed sintering reduces the translucency of 5Y zir- conia. The residual pores can be considered the main cause of this decrease in translucency. Density measurements and SEM observations suggest that 5Y-SS has more residual pores than 5Y-CS. It is probable that these residual pores became a light-scattering source and reduced the translucency. The translucency of zirconia is strongly influenced by the size and concentration of pores due to zirconia's high refractive index [34]. Figure 11 shows the relationship between the thickness of the specimen before sintering and translucency. In both 5Y-SS and 5Y-CS, the translucency decreased as the thickness of the pre-sintered specimen increased. This decrease in translucency was mainly due to the increase in pores since the number of pores increased as the pre-sintered thickness of the specimen increased ( Figure 8). In addition, the pore size in the 5Y-SS sample with a thickness of 8 mm or more, and in the 5Y-CS sample with a thickness of 18 mm, which both show a marked decrease in translucency, has increased. This suggests a relationship between pore size and translucency.
Translucency
High translucency is the only reason to use 5Y zirconia, which is less strong than 3Y and 4Y zirconia, in dental prostheses. Therefore, it can be concluded that 5Y zirconia is not suitable for high-speed sintering applications. Figure 11. Relationship between the specimen's pre-sintered thickness and TP.
Mechanical Properties
The results of the three-point flexural test and the fracture toughness test are shown The crystal phase composition is considered a factor affecting the translucency. Ceramic polycrystals generally exhibit higher translucency as they contain more optically isotropic crystal phases. From the XRD results, both 5Y-SS and 5Y-CS consist of tetragonal phases that are optically anisotropic. In the tetragonal phase, the lower the tetragonality, the closer to the optically isotropic cubic phase the sample becomes. Therefore, the light scattering due to optical anisotropy is reduced.
Zhang et al. compared the translucency of a 5Y zirconia sintered body prepared by mixing 3Y zirconia powder and 8Y zirconia powder with the 5Y zirconia sintered body prepared from the uniform 5Y zirconia powder, and reported that the former was less translucent than the latter. Their study concluded that the higher the yttrium concentration of low yttrium tetragonal crystals, the higher the translucency when the overall yttrium concentration is the same [35].
However, this study showed the opposite result to that reported by Zhang et al.: 5Y-SS had lower translucency than 5Y-CS, despite the high yttrium concentration of the lowyttrium tetragonal phase. This could be because, under the conditions of this experiment, the difference in the number of residual pores had a greater effect on translucency than the crystallographic difference. Figure 11 shows the relationship between the thickness of the specimen before sintering and translucency. In both 5Y-SS and 5Y-CS, the translucency decreased as the thickness of the pre-sintered specimen increased. This decrease in translucency was mainly due to the increase in pores since the number of pores increased as the pre-sintered thickness of the specimen increased ( Figure 8). In addition, the pore size in the 5Y-SS sample with a thickness of 8 mm or more, and in the 5Y-CS sample with a thickness of 18 mm, which both show a marked decrease in translucency, has increased. This suggests a relationship between pore size and translucency.
pre-sintered thickness of the specimen increased ( Figure 8). In addition, the pore size in the 5Y-SS sample with a thickness of 8 mm or more, and in the 5Y-CS sample with a thickness of 18 mm, which both show a marked decrease in translucency, has increased. This suggests a relationship between pore size and translucency.
High translucency is the only reason to use 5Y zirconia, which is less strong than 3Y and 4Y zirconia, in dental prostheses. Therefore, it can be concluded that 5Y zirconia is not suitable for high-speed sintering applications. Figure 11. Relationship between the specimen's pre-sintered thickness and TP.
Mechanical Properties
The results of the three-point flexural test and the fracture toughness test are shown in Figures 12 and 13, and the results of each two-way ANOVA are shown in Tables 5 and Figure 11. Relationship between the specimen's pre-sintered thickness and TP.
High translucency is the only reason to use 5Y zirconia, which is less strong than 3Y and 4Y zirconia, in dental prostheses. Therefore, it can be concluded that 5Y zirconia is not suitable for high-speed sintering applications.
Mechanical Properties
The results of the three-point flexural test and the fracture toughness test are shown in Figures 12 and 13, and the results of each two-way ANOVA are shown in Tables 5 and 6, respectively. In the flexural strength, the sintering method and specimen thickness showed a significant difference, while cross analyses including specimen thickness and sintering method (cf. Sintering method * Specimen thickness) did not differ significantly. On the other hand, as regards fracture toughness, only different sintering methods produced a significant difference. 6, respectively. In the flexural strength, the sintering method and specimen thickness showed a significant difference, while cross analyses including specimen thickness and sintering method (cf. Sintering method * Specimen thickness) did not differ significantly.
On the other hand, as regards fracture toughness, only different sintering methods produced a significant difference. Crystal phase composition and microstructure are considered to be factors leading to low mechanical properties in 5Y-SS. Firstly, 5Y-SS has a lower content of low yttrium tetragonal and a higher yttria concentration in it than 5Y-CS. Therefore, the contribution of toughening caused by the "tetragonal → monoclinic stress-induced phase transformation" is smaller than that of 5Y-CS [19]. As a result, flexural strength and fracture toughness decreased. Secondly, 5Y-SS has more residual pores than 5Y-CS (Figures 7 and 8). Residual pores become fracture origins in polycrystalline ceramics. Therefore, flexural strength decreased in 5Y-SS.
As for flexural strength, the specimen thickness induced a significant difference: 5Y zirconia has a reduced flexural strength in thick specimens. This is because the residual pore concentration increases toward the center of the specimen. Moreover, since the fracture toughness does not change with specimen pre-sintered thickness, there is no difference in the degree of toughening caused by stress-induced phase transformation.
5Y-SS-4.4 showed a flexural strength of 712 MPa. This strength value complies with ISO 6872 Class 4 (>500 MPa; approved for use with 3-unit prostheses, those connecting 3 teeth). One Visit Treatment is not available for cases requiring prostheses larger than a 3-unit one, so 5Y-SS-4.4 seems to have a clinically suitable strength value. Even so, 5Y zirconia cannot be used for high-speed sintering because in some cases, e.g., pontic bridge, the dental prosthesis may be as thick as 10 mm. As mentioned above, the residual pores of 5Y-SS increase both in density and size as the specimen thickness increases (Figures 7 and 8). Therefore, specimens with a pre-sintered thickness greater than 4.4 mm could have significantly reduced strength.
Effect of Chemical Composition
This study shows that the translucency and mechanical properties of 5Y zirconia are greatly affected by high-speed sintering. On the other hand, previous studies on high-speed sintering did not show a significant reduction in translucency and mechanical properties of 3Y zirconia and 4Y zirconia [13,14,29]. This suggests that the higher the yttria content, the greater the high-speed sintering effect. Matsui et al. [36] reported that as yttrium concentration increased, the sintering rate in the early stage of sintering decreased, and the sintering rate after the middle stage of sintering increased. According to this report, 5Y zirconia sinters more slowly in the early stage of sintering and more rapidly after the middle stage of sintering, compared to 3Y zirconia and 4Y zirconia. In the high-speed sintering of this study, the heat is extremely high compared to conventional sintering (12 times higher at 1100 • C to 1370 • C). Because 5Y zirconia rapidly densifies after the middle stage of sintering, at such a high-speed temperature increase, pores could not be sufficiently eliminated and many remained.
A similar discussion could be made for alumina. Matsui et al. reported that the addition of alumina to yttria-stabilized zirconia favors volume diffusion over grain boundary diffusion and increases the sintering rate [37]. The addition of alumina can adversely affect high-speed sintered zirconia as well as zirconia with high concentrations of yttria.
Conclusions
This study revealed that high-speed sintering with Speedfire reduces the translucency and mechanical properties of 5Y zirconia. Crystallographic characterization by XRD and microstructure observation by SEM showed that these decreases were due to changes in crystal phase composition due to suppression of yttrium segregation and increased residual pores.
Evaluation of pre-sintered specimens of various thickness revealed that the amount and size of residual pores inside 5Y zirconia increase as the specimen's pre-sintered thickness increases, resulting in reduced translucency and mechanical properties. Especially as regards high-speed sintering, 5Y zirconia is not suitable for thicker dental prostheses because the translucency and strength are significantly reduced when the pre-sintered thickness is 8 mm or more.
Discussions on the relationship between chemical composition and the effects of highspeed sintering suggested that high yttrium concentrations and the addition of alumina could adversely affect high-speed sintered zirconia.
From the above results, we conclude that 5Y zirconia is not suitable for high-speed sintering applications.
Data Availability Statement:
The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request. | 2022-08-21T15:12:11.402Z | 2022-08-01T00:00:00.000 | {
"year": 2022,
"sha1": "74993675f1eb3769d4a800459be704e8a8f48d3d",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1996-1944/15/16/5685/pdf?version=1660821051",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "ac61088810147e8460888b6b289866af78d85055",
"s2fieldsofstudy": [
"Materials Science",
"Medicine",
"Engineering"
],
"extfieldsofstudy": [
"Medicine"
]
} |
233236023 | pes2o/s2orc | v3-fos-license | Cis-regulatory elements in conserved non-coding sequences of nuclear receptor genes indicate for crosstalk between endocrine systems
Abstract Nuclear receptors (NRs) are ligand-activated transcription factors that regulate gene expression when bound to specific DNA sequences. Crosstalk between steroid NR systems has been studied for understanding the development of hormone-driven cancers but not to an extent at a genetic level. This study aimed to investigate crosstalk between steroid NRs in conserved intron and exon sequences, with a focus on steroid NRs involved in prostate cancer etiology. For this purpose, we evaluated conserved intron and exon sequences among all 49 members of the NR Superfamily (NRS) and their relevance as regulatory sequences and NR-binding sequences. Sequence conservation was found to be higher in the first intron (35%), when compared with downstream introns. Seventy-nine percent of the conserved regions in the NRS contained putative transcription factor binding sites (TFBS) and a large fraction of these sequences contained splicing sites (SS). Analysis of transcription factors binding to putative intronic and exonic TFBS revealed that 5 and 16%, respectively, were NRs. The present study suggests crosstalk between steroid NRs, e.g., vitamin D, estrogen, progesterone, and retinoic acid endocrine systems, through cis-regulatory elements in conserved sequences of introns and exons. This investigation gives evidence for crosstalk between steroid hormones and contributes to novel targets for steroid NR regulation.
Introduction
The nuclear receptor superfamily (NRS) is composed of ligand-activated transcription factors that control important developmental and physiological processes, by regulating gene expression when bound to specific DNA sequences. The NRS is crucial for regulating cell, organ, and body homeostasis, and an alteration of expression of specific nuclear receptors (NRs) is one of the causes of many human diseases [1][2][3].
Introns are a major proportion of DNA in both the plant and mammalian genome and are considered relevant components for genome adaptation [4]. They are not simply removed after RNA processing and are responsible for chromatin modification, transcription, RNA splicing, editing, translation, and gene expression [4][5][6][7]. The presence of introns elevates gene expression in a wide range of organisms including mammals [6,8,9]. Interestingly, intronic DNA sequences act as internal promoters that can be more important than the proximal promoter and constitute unrecognized binding sites for genes transcribed by the RNA polymerase II [6,9]. Even in the absence of the promoter in some genes, mRNA accumulation can be stimulated by the presence of certain regulatory intronic sequences [8].
Several reports have suggested a positive association between the complexity of an organism and the intronic fraction in the genome, which seems to be responsible for species-specific adaptations [7,10]. Integrated genomic analysis suggests that non-coding sequences in conserved sites usually are enriched with regulatory binding sites [11]. The same high abundance of regulatory binding sites has not been observed in exons [12]. This, therefore, suggests that the prevalence of conserved non-coding sequences (CNSs) among species is related to the preservation of a specific function and/or gene regulation and RNA splicing [5,13].
The regulatory elements located in non-coding intron and exon DNA sequences (cis-regulatory elements, CREs) are often transcription factor binding sites (TFBS) or splicing factor binding sites (SFBS) [14]. Mutations in CREs result in a significant reduction in target gene transcription and predispose individuals to a wide variety of disorders such as diabetes and cancer [15][16][17]. One example of a disease caused by this alteration is prostate cancer, where the androgen receptor (AR) is overexpressed in most castration-resistant patients [18]. Ethnicity and geography are risk factors for prostate cancer [19] as well as inherited mutations of the Breast cancer type 1 susceptibility protein (BRCA1) and Breast cancer type 2 susceptibility protein (BRCA2) genes, which are connected to estrogen regulation [20]. Ethnicity and geography (high latitudes) are associated with, e.g., low serum levels of vitamin D, which leads to decreased anti-inflammatory effects as well as decreased apoptotic activity and an increased proliferation [21]. The BRCA1 gene has been reported to be a co-regulator of the AR [22] and inhibits estrogen receptor α (ERα) signaling [23]. However, the interaction between the BRCA2 gene and estrogen signaling is indirect, and mutations in the BRCA2 are associated with decreased activity of gene DNA recombination and repair processes [24].
In prostate cancer, there is an imbalance in the crosstalk between different steroid hormones (e.g., androgens, estrogens, glucocorticoids, progesterone, retinoids, and vitamin D) leading to uncontrolled cell growth. Crosstalk between steroid endocrine systems, on a genetic level as well as in signaling pathways, has been reported but not fully understood [25]. Deepening our knowledge on the crosstalk between steroid endocrine systems, through their NRs, is of importance to understand the initiation and progression of hormone-driven cancers [25]. Little is known about NR interaction with non-coding conserved nucleic acid sequences. An understanding of the conserved intronic region distribution across species will enable the identification of candidate regulatory sequences. These sequences could interact with NRs evoking a change in the NR DNA binding and the NRs regulation of expression/repression of specific genes.
Our hypothesis is that steroid endocrine systems regulate gene expression through interactions with introns and exons. This study thus aimed to investigate crosstalk between steroid NRs in conserved intron and exon sequences, with a focus on steroid NRs involved in prostate cancer etiology. For this purpose, conserved intronic and exonic NRS sequences were analyzed with a focus on cis-regulatory elements (CREs) and their involvement in fundamental processes such as growth, differentiation, homeostasis, development, and metabolism.
Material and methods
Intron and exon sequences from the 49 genes of the NRS, translated from pre-mRNA transcripts to the corresponding DNA sequence, were retrieved from the Ensembl genome database (Table S1) [26]. The transcripts selected were orthologous, having the same position and phase relative to the coding sequence, from five different mammalian species: Homo sapiens, Gorilla gorilla, Pan troglodytes, Mus musculus, and Rattus norvergicus ( Table S1). These specific mammalian species were selected based on their close phylogenetic relationship, which implies high sequence conservation probability, and their usage for modeling species to humans [13,27,28]. Since some conserved genomic regions in primates and rodents recently have been identified as unique and responsible for new emerging functions, species from both orders were included in this analysis [13].
Basic conserved sequence detection method (BCSDm)
The Basic Conserved Sequence Detection method (BCSDm) extracts the most conserved sequences (without insertions or deletions) and their location within the specific sequence region. The BCSDm implemented in Python 3.6 using Biopython [29] is based on the combination of three methods: (1) alignment of the different sequences, (2) extraction of the alignment profile and its position score matrix (PSSM), and (3) obtainment of the conserved nucleotide patterns and their position in the alignment ( Figure S1).
The alignment was performed for each gene sequence between the five mammalian species using the multiple alignment program MAFFT [30]. From the resulting alignment file, an alignment profile summarizing the alignment for the five species was obtained. The dumb consensus method was selected for extracting the alignment profile and for calculating the number of each nucleotide type at each position of the alignment for all the sequences [29]. If the percentage of the most common nucleotide type was greater than the default threshold (0.7), the nucleotide was added to the alignment profile. This method was used to avoid gaps in the extracted sequence pattern. After obtaining the alignment profile, a PSSM was calculated to represent the probabilities of the occurrence of each nucleotide in the consensus sequence. The conserved patterns were extracted from the PSSM by the following conditions. First, each position selected from the score matrix should be 100% conserved, to avoid gaps, insertions or substitutions. Second, to include TFBS and to exclude random appearance in the selection of conserved patterns, a threshold of ≥15 consecutive nucleotides was applied.
All the 49 genes from the NRS were analyzed with BCSDm. Of these, 25 genes had at least one conserved intronic pattern and were thus further analyzed. A total of 1,044 conserved intron patterns in these 25 genes were extracted using BCSDm. Conserved sequences were grouped according to their ordinal position in the transcript (called intron 1 to intron 11). The percentage of conserved sequences was calculated for each intron group. To avoid an unequal number of introns between NR genes, the conserved patterns obtained were normalized to the number of genes containing each intron. Moreover, the number of conserved patterns in each intron was normalized to their sequence length.
Exon sequences, from the same 25 NRS genes described earlier, were also analyzed with BCSDm. A total of 552 conserved exon patterns in these 25 genes were extracted and further analyzed.
Analysis of CREs
Each conserved intronic and exonic pattern in the 25 selected NRS genes was extracted by BCSDm and scanned for putative binding sites for transcription factors [31] in the CIS-BP Database [32]. The search was performed using the species parameter Homo sapiens since the aim was to find out the relevance of these patterns only for the human species. The motif model was set to the standard scoring system option which is position weight matrices (PWMs)log-odds [33]. To be more restrictive in allowing mutations and to increase the likelihood of the TFBS predicted, the log-odds threshold was set to ten. To remove sparse sequences, only matching sequences with ≥10 consecutive nucleotides were considered potential TFBS.
After identification of potential TFBS in conserved sequences, the transcription factor binding domain (TFBD) family of the transcription factors inferred to bind these TFBS sequences was analyzed. A classification depending on the domain family type was carried out for each TFBDs. Thirty-seven different family domain types were identified and the ten with the highest number of TFBS were selected for further analysis of the introns. A small fraction of all analyzed TFBDs were NR binding domains. TFBS identified as NR binding sites were further analyzed and compared between introns and exons.
The conserved intronic patterns were further analyzed using the Human Splicing Finder (HSF) database to determine whether SS motifs were contained in their sequences [34]. This tool enables the prediction of potential donor and acceptor sites for the sequence introduced. The analysis used the default prediction algorithms (HSF and MaxEnt). The consensus value (HSF) was increased from 65 to 75 to allow higher similarity and confidence of a true splice site to be obtained [34,35]. Thus, the sequences with a consensus value of ≥75 (HSF) and ≥3 (MaxEnt) were classified as containing a splice site.
The conserved intronic patterns were classified as TFBS or SS depending on their content in regulatory elements. This classification revealed that some patterns contained exclusively TFBS or SS and/or both TFBS and SS in the same sequence. Four groups were derived from these results: TFBS, TFBS-SS, SS, and not identified.
The number of TFBS and SS in non-conserved sequences from the same gene intronic regions as the conserved sequences were used as controls. These sequences were scanned into the CIS-BP Database for TFBS hits and with HSF for splicing signals, with the same parameters as for the conserved sequences analysis. Moreover, a classification of TFBD families for TFBS in non-conserved sequences was used as a control. To randomly obtain the nonconserved sequences (n = 1,044), the BCSDm program was modified to extract nucleotides from the PSSM that were less than 100% conserved while maintaining a threshold of ≥15 consecutive nucleotides to generate a sequence.
Statistical analysis
The distribution of conserved sequences among different introns was assessed by the Mann-Whitney U test. The frequency distribution of the four groups (TFBS, TFBS-SS, SS, and not identified) among introns was analyzed using the Chi-square test. The number of TFBS as well as SS in conserved and non-conserved sequences was analyzed by the Wilcoxon paired non-parametric test. The number of TFBS for each TFBD family, between conserved and non-conserved patterns, was assessed by the Chi-square test. All the statistical analyses were performed in GraphPad Prism version 7.04 (GraphPad Software, La Jolla, California, USA). Statistically significant differences were set to ns: Not significant, *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001.
Ethical approval: The conducted research is not related to either human or animal use.
Results
Out of 25 NRS genes analyzed, the first intron in each gene (intron 1) had significantly higher sequence conservation (35%) when compared with downstream introns (introns 2-9, Figure 1). This was confirmed even after normalization for the number of genes and the sequence length (Tables S2 and S3, respectively). The number of conserved patterns generally tended to decrease with an ordinal position of the intron.
Seventy-nine percent of the conserved patterns showed putative TFBS in the CIS-BP Database for Homo sapiens. In the case of the distribution of TFBS per intron, the frequency was preserved in the range from 70 to 89%, except for intron 8 that was 38% ( Table 1). The non-conserved sequences showed that 69% of putative TFBS were overlapping (Table S4). Overlapping TFBS within the conserved sequences was 79%. Hence, 10% of the TFBS sequences, detected among the conserved patterns, are associated with conservation. Thus, the number of TFBS in conserved sequences was significantly higher when compared with non-conserved sequences (P < 0.05).
SS was found in 33% of the conserved patterns of Homo sapiens in the HSF Database ( Table 1). The nonconserved sequences contained 23% SS ( Table S5) and 33% of the conserved sequences overlapped with SS. Hence, 10% of these SS are associated with conserved sequences. The number of SS in conserved patterns was significantly higher when compared with non-conserved patterns (P < 0.01).
The distribution of conserved patterns for TFBS, TFBS-SS, SS, and not identified among different introns are presented in Figure 2. The TFBS group showed a similarity with 45-67% of the patterns, except for intron 8, which showed a very low number of conserved patterns ( Figure S2). The sum of percentages for the groups TFBS-SS and SS was also conserved among introns with 28-36% ( Figure 2). The analysis of the expected frequency distribution of the four groups showed that there were no significant differences among introns, except for intron 3 (P < 0.05, Figure 2) and 8 (P < 0.05, Figure S2).
A total of 10,608 TFBS were obtained from scanning the NRS conserved patterns in the CIS-BP database. Thirty-seven different TFBD families were identified from the TFBS analysis, of which ten with the highest number of TFBS were chosen for further analysis (Figure 3). The Figure 1: Percentage of sequence conservation for each intron group (1st to 9th) from 25 genes generated from the NRS analysis. Mann-Whitney U statistical test was performed to assess the differences of conserved sequences between different groups of introns (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). predominant TFBD families among the different introns were the Homeodomain, Forkhead, and Homeodomain POU domain families. Introns 7, 8, and 9 did not share the same distribution as introns 1-6 ( Figure 3) (Figure S2). The same most numerous ten family domains as for the conserved sequences were obtained for the non-conserved sequences (Tables S6-S9). However, a lower abundancy of TFBDs than expected was identified in conserved sequences than in non-conserved sequences (P < 0.05) ( Table S10).
Further analysis of the TFBDs revealed that 4.8% of the total number of conserved sequences in introns are NR-binding domains, when compared with 16.3% in exons (Tables S11 and S12). NRs bind in clusters to the same sequence but the NRs binding to the clusters differed between introns and exons (Tables S13 and S14) ( Tables 2 and 3). Analyses of the three genes involved in prostate cancer etiology, VDR (Vitamin D receptor gene), AR, and RXRA (Retinoid X receptor alpha gene) show that there are putative binding sequences for other NRs in introns and exons (e.g., ESR, AR, PGR, RARA, RORA, RORB, RORC, RXRA, RXRB, and RXRG), suggesting crosstalk between endocrine systems (Tables 2 and 3). Furthermore, several of the TFBDs were identified as p53 transcription factor domains binding with specific TFBS in the intronic regions of five NR genes (Table S15).
Discussion
The current study suggests that intronic as well as exonic sequences may be active parts in regulating gene expression, through CREs, and may serve as a target for steroid hormones. Our results show that conserved NRS intronic sequences are more abundant in the first than other introns, are enriched with TFBS, and contain SS, which are often co-localized with TFBS. Several TFBD families for TFBS were found in intron sequences, some of which contained specific TFBS for NR and p53. The intronic NR binding domains amounted to one-third of the NR binding domains found in exons.
In concordance with previous studies, the present study revealed that the number of conserved sequences decreased with intron position, indicating that intronic conserved pattern density among several species is higher in the first intron than introns downstream [10,36,37]. However, the density of conserved patterns was higher in the present study when compared with the study by Park et al. [10]. This is perhaps due to the number and evolutionary span of species analyzed in each study (five and 46 mammalian species, respectively). Furthermore, Park et al. excluded sequences within 300 base pairs of the splice junction (the boundary between intron and exon) in their analysis [10]. The current study included such sequences since they may contain regulatory elements, such as TFBS and SS. These regulatory elements could play a role in the transcription and/or splicing process and be responsible for the expression of certain genetic splicing variants [38,39]. In contrast to the results by Park et al., where a low number of conserved patterns were found downstream of the second intron [10], the results in the current study identified high frequencies of conserved patterns downstream of the first intron (Table S5). Thus, taken together, the current study suggests that most of the introns studied may include conserved patterns containing CREs, further suggesting a role in regulating gene transcription. Consistent with the previous report [10], the present study shows that putative TFBS are more abundant within the first introns, with the highest abundance in the first intron. The proportion of TFBS relative to the conserved sequences is preserved among introns, which has not been previously reported. This suggests that the number of TFBS is directly related to the number of conserved sequences and may explain the low number of TFBS obtained in the last introns. According to the present study's analysis of the false-positive ratio, due to sequence overlap, a considerable number of TFBS were bioinformatically confirmed to be true putative TFBS. However, further experimental validations are needed, such as ChIP-on-chip or luciferase assays [40,41]. Several studies agree that the identification of TFBS is complex and usually results in a large number of both false positives and false negatives [40][41][42]. These studies reduced the false-positive ratio by scanning the genome with PWM and identified true TFBS by searching for conservation of these sequences in orthologous transcripts. A conclusion from these studies is that TFBS can be identified through studying sequence conservation alone [40][41][42]. However, not all the functional TFBS present in the genome or sequences analyzed may be identified by conservation [40][41][42]. The results in the present study are in line with previous observations [40][41][42] showing that there is a higher percentage of TFBS within conserved regions compared to non-conserved regions.
To our knowledge, this is the first study to report an equal distribution of the conserved sequence types (TFBS and SS) among introns. The underlying cause of this preservation of distribution among introns is so far unknown but may be related to the regulation of the expression of certain genes and/or their splicing. Interestingly, the present study found that the proportion of SS in conserved sequences was greater when co-localized with TFBS (TFBS-SS group) than alone. Based on this finding, we propose that specific conserved sequences from NRS can act as splice consensus sequences and that most of them are surrounded by TFBS sites. Previous studies have suggested multiple links between transcription and splicing and that there are difficulties to isolate both processes since they are closely connected [43,44]. Furthermore, crosstalk between proteins involved in both transcription and pre-mRNA splicing has been suggested, and several mammalian candidate proteins, including transcription factors, have been identified [43,45]. Thus, we suggest that TFBS closely located to SS, may act as transcription factors or splicing binding sequences, and support previous studies suggesting a link between the transcriptional and the spliceosomal complex.
Mapping of the functional domains of transcription factors is crucial to understand their molecular function [46]. In the present study, the same ten TFBD families were more numerous than other families in both nonconserved sequences and conserved sequences ( Figure S2), which suggests that these ten families are common TFBD families for intronic sequences of NRs. However, there is less consensus between introns regarding TFBD family enrichment in non-conserved sequences. Considering this, conserved sequences may be more similar in their nucleotide composition and thus bind with specific TFBD families.
Further analyses of the putative TFBS identified in this study revealed specific TFBS for p53 in intronic conserved sequences of five NR genes: The nuclear receptor subfamily 2 group F (NR2F), the estrogen receptor 1 (ESR1), the nuclear receptor subfamily 4 Group A member 3 (NR4A3), the AR and the nuclear receptor subfamily 1 group D member 1 (NR1D1). A previous study mapping p53 binding sites in the whole genome did not find p53 specific TFBS for NR intronic sequences [47]. Possible explanations for these different findings could be that Wei et al. conducted their study on cultured cancer cells, in vitro, and that their analysis was performed on expression level but not on a single molecular level [47]. Furthermore, mapping was done for the whole genome and only detected the regions highly enriched with p53binding sites [47]. On the other hand, the present study was performed at normal conditions and not under conditions influenced by transcriptional rates, where eight p53 TFBS specific for the NRs was identified. Thus, the current finding extends the knowledge about p53 binding locations and indicates that p53 is involved in regulating NR-mediated transcription. This is to our knowledge the first study to report putative interactions of steroid NRs in the case of intronic sequences. These results indicate the existence of a regulatory network involving the interaction of regulatory DNA elements located in the intronic regions of the NR genes and NR transcription factors. Of the three steroid NR genes studied, NR binding domains were demonstrated in introns 1 and 3 and exons 4, 6, 7, and 9, and thus may indicate crosstalk between endocrine systems. Crosstalk has been described for steroid NRs [25], growth factor receptors [48], steroids [49], intracellular and stressactivated kinases within the mitogen-activated protein kinase (MAPK) superfamily [50,51] as well as downstream signaling components of these kinase pathways [52,53]. Understanding the crosstalk between steroid receptors is important in the initiation and progression of hormonedriven cancers [21,25,54]. In the present study, VDR, AR, and RXRA have several TFBS for NRs in introns and exons, which are closely connected to cancer development. The current results suggest a crosstalk between the androgen endocrine system and the VDR, the ESR, the progesterone receptor (PGR), and the retinoic acid receptors as well as a crosstalk between the retinoic acid endocrine system and ESR and the thyroid hormone receptor (THR), among others. These findings are in concert with previous reports on the crosstalk between steroid receptors of prostate and breast cancer cells [21,25]. It is known that the protein interaction between VDR and RXR causes antitumoral effects in prostate cancer [55,56]. Further examples are ARs, which either have an antagonistic or a cooperative effect on the ESR binding to estradiol responsive elements dependent on the presence of dihydrotestosterone [57][58][59]. Thus, the present study provides more evidence for crosstalk between steroid NRs and contributes with novel targets for steroid NR regulation. Furthermore, this study confirms the hypothesis that individual steroids and steroid NRs rarely work in isolation but rather as a crosstalk between different receptor types, allowing activation of signaling pathways, and modulate transcriptional responses [25]. | 2021-04-15T13:29:02.739Z | 2021-01-01T00:00:00.000 | {
"year": 2021,
"sha1": "ba9b858f6540ebfcc7c83c03c9985a16f91e51e4",
"oa_license": "CCBY",
"oa_url": "https://www.degruyter.com/document/doi/10.1515/med-2021-0264/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "8528187fa7795d285bdc0ee59395500f736e636f",
"s2fieldsofstudy": [
"Medicine",
"Biology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
259162800 | pes2o/s2orc | v3-fos-license | Neuronal responses to omitted tones in the auditory brain: A neuronal correlate for predictive coding
Prediction provides key advantages for survival, and cognitive studies have demonstrated that the brain computes multilevel predictions. Evidence for predictions remains elusive at the neuronal level because of the complexity of separating neural activity into predictions and stimulus responses. We overcome this challenge by recording from single neurons from cortical and subcortical auditory regions in anesthetized and awake preparations, during unexpected stimulus omissions interspersed in a regular sequence of tones. We find a subset of neurons that responds reliably to omitted tones. In awake animals, omission responses are similar to anesthetized animals, but larger and more frequent, indicating that the arousal and attentional state levels affect the degree to which predictions are neuronally represented. Omission-sensitive neurons also responded to frequency deviants, with their omission responses getting emphasized in the awake state. Because omission responses occur in the absence of sensory input, they provide solid and empirical evidence for the implementation of a predictive process.
INTRODUCTION
Mismatch negativity (MMN) is a component of the auditory eventrelated potentials that is elicited when acoustic expectations are violated. It has been shown to correlate with behavioral and perceptual measures of deviance detection. Furthermore, the amplitude of the MMN response indexes the magnitude of the expectancy violation. The MMN response is widely considered as a neurophysiological correlate of a perceptual prediction error (1)(2)(3)(4)(5)(6)(7)(8)(9) that results from the comparison between the actual sensory (bottom-up) input and a memory trace, so-called prediction, encoded in top-down activity (10)(11)(12)(13)(14). A prediction error would arise when there is a mismatch between the internal prediction and the actual sensory input. A distinct aspect of the MMN is that the omission of an expected auditory stimulus (analogous to the deviant stimulus in an oddball paradigm) is often accompanied by elicitation of the omission MMN component in humans (15)(16)(17)(18)(19)(20)(21)(22). This is important because it confirms that MMN is not simply a passive, bottom-up process that leads to synaptic adaptation to repeated stimuli due to a constant stimulation of the same afferent pathways (8,23). Previous studies have argued that the omission response provides conclusive, empirical evidence of the predictive process, as it occurs in the absence of sensory input (19,20).
Thus, predictions can be studied by examining the neuronal signature to the omission of an expected sound. Omission responses therefore might constitute a critical and underestimated test, particularly in animal models, which may be pivotal in uncovering the substrates of predictive processes. If stimulus-evoked neuronal activity indexes the difference between a sensory signal and its topdown prediction, then the omission of a predicted sensory signal should lead to neural activity, which reflects the pure prediction signal (4,19,20,(24)(25)(26)(27). A classic functional magnetic resonance imaging study has previously examined the topic of auditory imagery by introducing gaps of silence in familiar and unknown songs, showing neural activation of auditory cortex (AC) during the gaps of silence (28). However, empirical data at the singleneuron level in animal models to support the predictive coding hypothesis in this context are elusive.
We therefore aimed to investigate whether single neurons in different auditory areas respond to the omission of stimuli similar to the human MMN response. We used two rodent species that have proven to share a similar auditory system (29,30), which allow easy experimental manipulation in anesthetized or awake preparations. We found neural responses to the (silent) omission of a predictable pure tone in a subpopulation of neurons in the AC neurons under urethane anesthesia and in the awake state. In the latter, the proportion of AC neurons exhibiting omission responses was found to be substantially larger (~36% of recorded neurons, compared to 22% in anesthetized). These neurons responded to pure-tone stimuli but showed an even larger response when tones in a sequence of identical tones were omitted. We explored responses at different rates of tone trains and found that this distinct response to the omitted stimuli mostly occurred during rapid tone trains as in human MMN studies [stimulus onset asynchrony (SOA) ≤ 150 ms]. Omission selectivity increased during the sequence, primarily because of further adaptation of the response to the standard tone. In the awake state, omission responses became more distinct from deviant responses. Spatially, strong omission responses were more abundant in the caudal portions of the primary AC (A1), posterior auditory field (PAF), suprarhinal auditory field (SRAF), and ventral auditory field (VAF) in anesthetized rats.
RESULTS
Below, we first describe the quantitative characterization of omission-related responses and their dependence upon several experimental factors such as species, brain region, brain state, SOA, and sequence predictability. Then, we compare omission-related responses in relation to corresponding stimulus-bound (evoked) responses. Moreover, we investigate the latency of the responses, their sequence dependence, relation to pure-tone deviant responses, and their laminar and field specificity.
We tested whether neurons of the auditory pathway exhibit elevated responses to the omission of sounds within a fixed-rate, repetitive sequence, i.e., whether their response signifies a violated expectation of a regularly repeating sequence. For this purpose, we recorded neuronal activity in the AC of urethane-anesthetized rats (n = 77) and mice (n = 350) as well as awake mice (n = 393), and also from the inferior colliculus (IC) (n = 48; cf. Supplementary Materials) in urethane-anesthetized rats in response to a set of long pure-tone sound sequences. The sequences corresponded to an oddball paradigm (31) where the deviant was an omitted stimulus ( Fig. 1 and fig. S1). We tested each neuron with a set of tone sequences, with fixed frequencies and intensities chosen to fall within its frequency response area. Sequences differed in their SOA (SOA: 125, 250, or 500 ms), in which the pre-activity and the analysis windows were chosen in accordance to the increasing SOAs to capture possible long latencies, and precise sequencing, either with a random number of tones before an omission or a fixed number (periodic). In a subset of the recordings in the mouse AC, we also presented classical oddball sequences of identical temporal structure with pure-tone deviants.
Omission responses were defined as a temporally localized elevation of spiking activity occurring during the periods of silence (omitted deviant; Fig. 1, asterisks in AC recording trace) when a stimulus was omitted (randomly or periodically, 10% probability in both cases) in a long sequence of tones (Fig. 1). Significance was assessed on a neuron level first before aggregating across neurons. This is essential as the classical theory of predictive coding (14) suggests that only a subset of neurons will respond to the omission of a predictable stimulus, likely signaling a prediction or prediction error (see Discussion for details), particularly inhibitory neurons (19)(20)(21). Specifically, we compared the neuronal activity recorded just before each omission trial (time windows of −25 to 5, −75 to 5, and −195 to 5 ms), with the window corresponding to the timing of the expected stimulus (5 to 120, 5 to 150, and 5 to 225 ms, respectively), using one-sided Wilcoxon signed-rank test at a Fig. 1. Experimental setup and design for recording from the IC and AC. While stimulating with sequences of pure tones, neuronal activity was recorded using microelectrodes in the urethane-anesthetized rats (AC and IC) as well as urethane-anesthetized and awake mice (AC). The long tone sequence consisted of a repeating tone (standard probability: 90%) that was occasionally replaced by an omitted tone (average deviant probability: 10%). Traces correspond to actual recordings (from the IC and AC of urethane-anesthetized rats, respectively). Red stars indicate neuronal responses during an omitted tone. SOA denotes the amount of time between the start of one stimulus and the start of the following stimulus. significance level of 0.05. We only checked for elevations in firing rate here, while we do not exclude that other neurons might show systematic reductions of their activity. To preempt a positive detection bias due to circular analysis ("double dipping") (32,33), we implemented split data analysis, i.e., half of the omission trials were used to assess whether a neuron showed a significant response, and the other half of the data were then used further on. In this way, random elevations of firing rate cannot systematically influence all subsequently analyzed measures. Following these criteria, 19.5% (15 of 77) and 21.7% (76 of 350) neurons from urethaneanesthetized rat and mice, respectively, as well as 35.6% (140 of 393) neurons from awake mice responded to omitted tones in 125 ms of SOA and are referred to as "omission-sensitive neurons." Data from the IC are only reported in the Supplementary Materials (cf. fig. S2), as the number of omission-sensitive neurons was too low for subsequent analyses.
Evidence of omission responses in AC neurons
Omission responses were originally found at an SOA of 125 ms in human electroencephalography (EEG) recordings (16). In the present experiments in rodents, we consistently find omission responses for this SOA (see Fig. 2A, red traces; representative neurons with significant omission responses, i.e., omission-sensitive neurons) under all three recording conditions. Omission responses in the AC of anesthetized rodents ( Fig. 2A, left and middle) are salient in comparison to the preceding stimulus response, which is the last stimulus in the stimulation sequence before the omission. The omission response for the awake mouse AC neuron is very distinct, both in size and timing ( Fig. 2A, right).
Next, we computed grand-average peristimulus time histograms (PSTHs) including all neurons showing significant omission responses ( Fig. 2B; see Materials and Methods for details) for the three conditions. As in the individual cells, the omission responses are more robust as we progress from anesthetized to awake animals. Significantly, while positive omission responses are present in anesthetized rodents, AC populations (Fig. 2B, left and right), they are even more pronounced in the awake mouse preparation (Fig. 2B, right). The stimulus-evoked response (blue) shows less adaptation in awake mice (Fig. 2B, right) as opposed to cortical neurons in anesthetized rats and mice (Fig. 2B, left and middle). In awake mice, the onset responses to each stimulus are visible and the omission-evoked activity (red) is readily recognizable (Fig. 2B, right). The firing rate in response to sound stimulation is low in anesthetized rodents, except for the first stimulus after the omission, and it tends to increase during omitted tones (Fig. 2B, left and middle).
The shape of the responses to the stimulus typically had an onset component, both in individual examples and at the population level. Using the 115-ms analysis window that includes most of the evoked responses (see Materials and Methods), only in the case of the anesthetized animals the response to the stimulus after the omission was larger than the response to the stimulus previous to the omission (P = 0.0125 and P < 0.0001 for rats and mice, respectively; Wilcoxon test). However, when using a 30-ms analysis window encompassing only the onset component, we found that the response to the stimulus after the omission was larger than the response to the stimulus previous to the omission under all conditions (anesthetized rat: P = 0.0496, anesthetized mouse: P = 0.0402, awake mouse: P < 0.001; Wilcoxon test). Under the awake condition, as opposed to the responses to the stimuli, the omission responses tended to be more sustained and started from the expected time of occurrence of the omitted stimulus, often continuing to the subsequent stimulus presentation.
In summary, omission-evoked responses are strongly represented at 125 ms of SOA occurring both under anesthetized and awake states in the AC, with a substantial increase in magnitude in the awake state (for comparison with our smaller dataset from the IC, see fig. S2). At longer SOA (250 and 500 ms), we still detected a small subset of omission-sensitive neurons, but their average responses (based on the second half of the split data analysis) did not provide substantial evidence for reliable omission responses at these SOAs.
Comparison between sound and omission responses
Above, we identified putative omission-sensitive neurons using the split data approach on half of the omission trials. Below, we evaluate their response on the population level and compare them to responses to stimuli (not just the preceding one) and responses to stimuli after the omission. Each of these responses was referenced to the omission preceding activity to have a common local baseline, as we expect the spontaneous rate before the stimulation sequence to overestimate the within-sequence spontaneous rate (Wilcoxon test with Holm-Bonferroni correction, P < 0.05; Fig. 3). AC neurons from anesthetized rats and mice responded with similar firing rates to stimuli and omissions (P = 0.577 and P = 0.067, respectively; Fig. 3, A and B). On the other hand, awake mice AC neurons showed a significant enhancement of omission responses relative to stimulus responses (P < 0.001; Fig. 3C; cf. Table 1 for details). We found that the response to the omission was particularly stronger than to the first stimulus after an omission under the awake condition, highlighting the omission response in the sequence (P < 0.001; Fig. 3C). The average firing rate in response to after-omission stimuli was significantly larger than the average response to the rest of stimuli in anesthetized rat and mouse neurons (P = 0.038 and P < 0.001), but not in awake mice AC neurons (P = 0.276), although it is also significant when using a 30-ms window, adapted to the brevity of the response (see above). The reduced average response to all stimuli (excluding after omissions) is probably due to an adaptation effect to the repeated stimuli (7,34), which is later released by the occurrence of the omission and subsequent after omission (see more details below in the "Omissionsensitivity dynamics depend on brain state" section).
Furthermore, we tested whether there were differences in the response to the omission under the periodic and random conditions (with a common, average probability for omissions of 10%), but there were no significant differences at the population level (P = 0.93, Wilcoxon test). As an occurrence of the omission in the periodic sequence is more predictable, we originally hypothesized a greater response under the random condition. The lack of a difference could be explained by the length of our shortest sequence under the random condition (3) in relation to the dynamics of response adaptation (see more details below in the "Omission-sensitivity dynamics depend on brain state" section and Discussion).
In summary, we find two important features regarding the omission firing rates: First, for omission-sensitive neurons, the firing rate during the omission is similar to the sound-evoked firing rate under the anesthetized condition (in both rat and mouse), but under the awake condition, the omission response is even larger than the sound-evoked activity over the analysis window. Second, the neural response of omission-sensitive neurons exhibits a timed rise at the start of the expected tone following an omission.
To quantify the comparison between sound-and omissionevoked activities, we computed an omission index (iOmi) at the cellular level [iOmi = (Omi FR − Stim FR)/(Omi FR + Stim FR)], for the omission-sensitive neurons, under the three recorded conditions, where the firing rate (FR) for stimuli was averaged from all stimulus presentations (Fig. 4A). Negative and positive iOmi values show neurons that exhibit a larger response to the sound or to the omission, respectively.
The average iOmi (Fig. 4A, vertical red dotted lines) quantifies the response selectivity to omissions for a group of neurons, here applied to the omission-sensitive neurons. Anesthetized rat AC neurons respond similarly on average to the stimulus and to the omission (iOmi = −0.089, P = 0.661, Wilcoxon test with Holm-Bonferroni correction; Fig. 4A, left), as well as anesthetized mouse AC neurons (iOmi = −0.006, P = 0.836; Fig. 4A, middle), while in awake mouse AC neurons, the average response to the omission is larger than to the stimulus (iOmi = 0.055, P < 0.001; Fig. 4A, right). Notably, the average iOmi is significantly larger in the awake mouse AC than in the anesthetized mouse AC (P = 0.0013, Wilcoxon rank sum).
Next, we compared the omission-sensitive (Fig. 4B, red) neurons to their complement, referred to as omission-insensitive (Fig. 4B, blue) neurons, in terms of the difference in activity to omissions and stimuli. In line with the definition of these groups of cells, we find that the omission-sensitive neurons have a larger difference than omission-insensitive neurons, thus relatively responding more to omissions (P = 0.028 and P = 0.016 for anesthetized rats and mice, respectively, and P = 1.7 × 10 −8 for awake mice, Wilcoxon rank sum tests). While under urethane anesthesia, the difference is negative, i.e., greater responses to stimuli than to omissions, in both species (Fig. 4B); this relation reverses for the omission-selective neurons in the awake mouse (Fig. 4B, right, red). In all comparisons above, both omission and stimulus responses were referenced to the activity in the preomission period to focus on relative changes within cells instead of absolute firing rates.
In summary, the degree of omission selectivity in the neural responses was comparable between species in the AC and increased in the awake compared with the anesthetized state in the mouse. While omission selectivity differed in omission-sensitive and omission-insensitive populations under all conditions, the difference was most pronounced in the awake state.
Response latencies of omission and sound responses
To explore the omission responses further, we measured and compared the peak latency for the omission-and sound-evoked activity (paired Wilcoxon test with Holm-Bonferroni correction, P < 0.05; Fig. 5). Response latencies for omitted tones, stimuli, and afteromission stimuli are shown in Fig. 5 for the three recorded conditions. The neurons show very similar latencies for the omission-and stimuli-evoked responses in both anesthetized preparations (Fig. 5, A and B, and Table 2), while the latency of the omission responses is larger than the latency of the stimuli-evoked responses in the awake mouse AC (P < 0.001); this is probably caused by multiple factors, including differences in the profile of the responses to the stimulus (peaking closer to the sound onset) and the omission (more sustained responses), overall larger firing rates, and reduced stimulus adaptation under the awake condition, which would contribute to faster responses. A real and genuine omission response should be time-locked to the period when the stimulus is expected to happen (but it is omitted). These latency data support the notion that the source of the omission response (i.e., putative prediction) Similarly, the white circles indicate the median and the white triangles show the 95% confidence interval for the median. Evoked firing rates have been baseline-corrected, relative to preomission activity. Asterisks directly over the violin plots indicate that the distribution is statistically different from 0 (Wilcoxon test with Holm-Bonferroni correction; see Table 1). Asterisks over horizontal lines indicate significantly different distributions (Wilcoxon test with Holm-Bonferroni correction; see Table 1). *P < 0.05, **P < 0.01, and ***P < 0.001. could differ from the stimulus response (forward response) as posited by predictive (14) coding (cf. Table 2 for details).
Comparison between omission and deviant responses
Responses to unexpected changes in tone properties, so called deviants, are very related to responses to omissions, as the latter can be considered a special case of a deviant. Their properties have long been studied under the topic of stimulus-specific adaptationoften considered a special case of predictive coding-where a certain degree of adaptation occurs for the repeated standard stimulus, while the deviant stimulus leads to a larger response, in higher auditory areas even larger than it would be in isolation (7,8,10,23,34). A natural question is whether the classical deviant responses co-occur with omission responses.
To address this question, we compare here the neural response for omission-sensitive neurons in the anesthetized and awake mouse AC (targeting A1). In the awake dataset, the deviant and omission conditions were presented for a subset of 126 (of a total of 393) cells, while in the anesthetized set, they were presented for all 350 cells, of which 55 and 76 were omission-sensitive, respectively. As usual, the deviant response (Fig. 6A, left column, orange) was larger than the standard response (P = 1.4 × 10 −9 , one-sided Wilcoxon signed-rank test; Fig. 6A, left column, blue), which, under the anesthetized condition, was nearly completely adapted. The omission response was less precisely timed than the deviant response, but was significantly larger than the preomission period ( Fig. 6A, left, red). Under this condition, both omission-sensitive and omission-insensitive neurons responded similarly to omission and deviant stimuli, as evidenced by the overlap of the cumulative distributions of the difference between deviant and omission responses (P = 0.15, Wilcoxon rank sum test between the two populations on omission-deviant responses; Fig. 6A, middle and right). The curves intersect the ordinate at 0, indicating that responses to deviants were, on average, larger than to omissions.
On the other hand, the omission-sensitive neurons (Fig. 6B) also showed a very strong response to the deviant and a clearer response to the standard than under the anesthetized condition, likely because of reduced stimulus adaptation. In addition, the awake preparation shows a clear sustained response to the omission, with an overall rate similar to the deviant response (P = 0.36). Under this condition, omission-sensitive neurons respond to omissions with larger firing rates than to deviants (P = 0.0017), which is not the case for omission-insensitive neurons (Fig. 6B, middle). This is also indicated by the "dent" in the cumulative distribution of the difference between omission and deviant responses for the omission-sensitive neurons (Fig. 6B, right, red). These response properties in the AC show that omission-sensitive neurons can also have a response to deviant sounds; however, in the awake state, the omission response of the omission-sensitive neurons becomes comparable in size and more specific to omissions (see Discussion for details).
Omission-sensitivity dynamics depend on brain state Next, we analyzed whether omission-sensitive neurons show a dependence of their response on the history of stimulation, and how this depends on the state of the animal. First, we considered the neural responses as a function of the overall stimulation sequence. As expected, the stimulus response exhibited a fast decay, within just two stimuli under both the anesthetized (Fig. 7A, left, blue) and the awake (Fig. 7A, right, blue) condition. Under the awake condition, a second time scale of adaptation existed with a time constant of~5 s. The level of the omission responses (red) stayed rather constant over the period of the whole sequence in both states. The iOmi index of omission selectivity ( Fig. 7B; see Methods) showed a slow increase, which was more pronounced under the awake case (r = 0.39, P = 1.2 × 10 −16 , Pearson correlation) than under the anesthetized condition (r = 0.1, P = 0.035). This was probably mostly driven by the adaptation of the standard response, potentially signifying a buildup of predictions (21).
Second, we studied the omission-evoked responses in relation to the number of stimuli preceding the stimulus omission. In the anesthetized case, a significant decrease [P = 0.02, Kruskal-Wallis analysis of variance (ANOVA)] is observed for the omission response ( Fig. 7C, left, red). The response to the first standard after the omission (purple) shows a similar but nonsignificant dependence. Unexpectedly, no dependence with a decrease for shorter sequences of standards is present in the awake case (Fig. 7C, right); conversely, the standard response instead shows a significant decrease (P = 0.00074). For the omission response, no dependence is present, although the omission response is on a high level relative to the standards (firing rates are normalized to the average standard response here), which suggests that the omission response might already have reached a plateau level. This possibility is supported by the rapid adaptation of the standard response after the omission (Fig. 7D), which is very salient in the anesthetized case and completes after just two standard stimuli, and even after just one in the awake case. In the latter, this behavior is only apparent if a shorter time window (30 ms) is used for the analysis (Fig. 7D, right, inset), which matches the brief, driven response (see Fig. 2B, right).
In summary, the omission sensitivity shows a significant slow increase over prolonged stimulation, particularly under awake , and awake mouse (C). The blue, red, and purple violin plots depict the distribution of the sound stimuli, omission, and after-omission latencies (see Fig. 3 for the meaning of the violin plot symbols). Asterisks denote statistical significance between stimuli and omission latencies (***P < 0.001; paired Wilcoxon with Holm-Bonferroni correction; see Table 2).
conditions. A dependence on the number of preceding standards is found in anesthetized, but not in awake. The latter might indicate that the omission response has already reached a plateau, which we cannot resolve presently, as even shorter sequences of standards (1 to 2) would be required.
Histological distribution of omission responses
The histological analysis of the electrolytic lesions in the IC demonstrates that omission-sensitive neurons are located in the nonlemniscal IC ( fig. S3). To test whether omission responses show any specific distribution across the AC fields and/or layers, we analyzed where neurons with omission responses were located in the different AC regions ( Fig. 8 and fig. S4).
We analyzed the distribution of omission-sensitive neurons across layers (Fig. 8, A to C). In the AC of anesthetized rats, the proportions of omission-sensitive neurons were 7.1, 35, and 22% for supragranular, granular, and infragranular layers, respectively (Fig. 8A, burgundy bars relative to gray). In the AC of anesthetized mice, these proportions were 26% of the neurons in the supragranular layers that were omission-sensitive, 30% in the granular, and 19% in the infragranular (Fig. 8B, burgundy bars). For awake mice AC neurons, these proportions were 40, 38, and 34% for the supragranular, granular, and infragranular layers, respectively (Fig. 8C, burgundy bars). We performed bootstrapping analysis on all three recording sets, but only found a significant reduction relative to the distribution of expected counts in the supragranular layer (P = 0.003, based on 300,000 independent draws, and P < 0.05 significance bound), although there appears to be a tendency for a higher fraction of omission-sensitive cells in supragranular layers in the awake compared with the anesthetized mouse. However, In the AC of the awake mouse, the responses are generally larger and the omission response becomes comparable in overall rate to the deviant response (P = 0.36), while the deviant response is~91% larger than the stimulus response (P = 2.7 × 10 −7 ). Many cells, particularly from the omission-sensitive (red) group, show omission responses that are larger than the deviant response (middle; red dots above diagonal). In the awake state, omission-sensitive neurons show a clear and significant separation in the difference between omission and deviant responses (P = 0.0017). In the awake state, the omission response already starts at a high level and has no significant dependence on the number of preceding standards, possibly indicating that the dependence has already plateaued. The after-omission response shows a fast and significant decay to a plateau level. (D) The response to the standard following an omission decayed quickly, within two stimuli under anesthesia and just one stimulus in the awake (see inset for 20-ms analysis window; P values based on Kruskal-Wallis ANOVA). Using our standard analysis window (5 to 120 ms), the effect is also significant, but less pronounced, because of the brevity of the response (see Fig. 2B, right).
additional research with higher-density probes will be required to fully address this question.
Last, we investigated the location of omission-sensitive neurons in subfields of the AC. Location data were only available in the rat, as the recordings in the mouse were all targeted toward A1 only. As we acknowledge that the number of cells is likely insufficient for a complete map, we provide these data in the supplement as suggestive evidence. To allocate each recorded neuron to a specific field in the rat AC, we recorded the frequency response areas and analyzed the topographical distribution of characteristic frequencies for all units. Each recording was assigned to a dorsoventral and rostrocaudal coordinate relative to bregma as in previous studies (7,8,35,36). This analysis allowed us to pool the data from all animals ( fig. S4A) and construct a synthetic map of characteristic frequencies across the entire rat AC (7,8). Similar to these previous works, we found a high-frequency reversal zone between VAF (caudally) and anterior auditory field (AAF; rostrally), a low-frequency reversal zone between A1 and PAF (dorsocaudally), and a high-frequency reversal between VAF and SRAF (ventrally). Thus, we could reliably define the lemniscal (A1, AAF, and VAF) and nonlemniscal (SRAF and PAF) auditory cortical fields as shown in fig. S4A. The omission-sensitive neurons found in this study were located mostly toward the caudal areas of the AC, within the limits of A1, PAF, and SRAF ( fig. S4A). When plotting iOmi levels for all significant sequences recorded (in this case, we did not use the average value per unit, because it would not allow to properly map the whole AC), it becomes evident that, although omission responses are distributed in all fields, the highest levels of iOmi were found in SRAF and PAF (fig. S4B).
DISCUSSION
The key finding of the current study is that a subset of neurons in the auditory brain responds to the absence of an expected stimulus, which provides empirical evidence in support of certain elements of predictive coding in a subcortical area and sensory cortex (1,(37)(38)(39)(40)(41). Specifically, we demonstrate that robust neural activity emerges during the omission of a sound in a regular sequence of repetitive stimuli. These responses are consistent with prediction error signals, as these omissions result in the violation of an expected stimulus when it should occur. We identified several key properties of the omission responses. In particular, our results demonstrate that the omission responses depend on different parameters, including the repetition rate, brain state but are similar across species. According to our findings, omission-sensitive responses are already present under anesthesia and become even more pronounced under the awake condition in both number and strength. The omission responses are likely the earliest neuronal correlate underlying the salient perception of omitted stimuli in humans and can be considered to provide additional support for a hierarchical predictive coding framework (1,37,39,42,43).
Relation to previous studies
In humans, responses to omissions of tones are primarily observed at rather high presentation rates (15,16), corresponding to the presently used shortest SOA. To resolve these on the neuronal level, the high temporal resolution available in electrophysiology is highly suited. On the basis of single-unit recordings, our study demonstrates clear omission responses at the spiking level in the auditory system and thus identifies a potential neuronal correlate for those omission responses previously recorded using noninvasive techniques such as EEG/magnetoencephalography in humans (9,16,25,(44)(45)(46)(47)(48)(49).
Multiple studies have investigated mismatches between predictions and actual stimuli on the neuronal level using calcium imaging in different sensory cortices. While calcium imaging provides the possibility for highly parallel recordings, its temporal resolution limits the use of comparably fast SOAs in continuous sequences as in the present study. In the visual cortex V1 in the mouse, multiple studies reported responses to sequence violations [e.g., (50)(51)(52)], as well as in the somatosensory system (53). In the auditory system, virtually no, if any, studies have directly looked for omission responses at the cellular level, or have not directly reported them (7,8,13,34,54,55). Two studies have reported potentially related results using calcium imaging: Li and colleagues (56) have reported neural responses at the end of a regular, slow train of repetitive stimuli as a persistent evoked response, which they termed echo responses (56,57). While those echo responses might also signal predictions, the much slower time scales (SOA: 2 to 4 s) suggest different underlying mechanisms. Possibly because of limitations imposed through the calcium dynamics, the studies did not find omission responses at fast rates. Given the shorter SOAs, we observe strong adaptation to the tones, while such adaptation appears to be absent in the longer SOAs for the echo responses, which suggests a difference in the contributing mechanisms. Other differences between omission and echo responses include latency and histological location: The latency of the present omission and stimulus responses was similar, which indicates that the omission responses are time-locked to the expected time of occurrence of the omitted stimuli. In contrast, the latency of the echo responses is significantly longer than that of the stimulus responses (237 versus 83 ms) (57). Omission responses are found in subcortical neurons, at the IC level, while the echo responses seem to only emerge at the cortical level (56). Moreover, these authors (56) also discuss very elegantly how their echo responses differ from the classical omissions recorded under the oddball paradigm as we do here. Li and colleagues (56) relate these responses to fit the time scales of motor responses. This is a key aspect of their work as their results strongly support that the neuronal circuits of the AC are critical for coding predictive information and transforming it into anticipatory motor behavior. There is also other work that is also similar in some ways to the echo responses but recorded in the IC, the so-called "long-lasting sound-evoked afterdischarge" responses or LSA neurons (58). These authors argue that, because LSA produces long-lasting firing in the absence of sound, it may be relevant to temporary or chronic tinnitus or to some other aftereffect of long-duration sound. While it thus appears that omission, LSA responses, and echo responses rely on different underlying mechanisms, the use of different stimulation paradigms and recording procedures complicates a conclusive comparison.
Earlier theoretical investigations using neural modeling had already suggested the existence of omission responses (20,59). Omission responses were here created by the difference between the activity of neurons encoding the prediction of the next stimulus and the actual input (see the "Relation to predictive coding theory" section for more details).
Relevance of the current results
We have recorded robust omission responses in the auditory pathway of anesthetized rats and anesthetized and awake mice using single-and multiunit recordings. These statistically significant responses of single neurons evoked by omitted tones had the largest proportion (~36%) observed in the awake AC and the lowest at the urethane-anesthetized AC of rats (~20%) and mice (~22%). The difference in the detected omission responses appears to be determined most strongly by brain state, because we did not find a significant difference between the omission response size in the AC between the anesthetized rat and mouse. The fact that we found omission responses also in anesthetized preparations suggests that omission detection and thus predictive coding are already partially available under the passive condition, without the need for specific attention (8,23). Furthermore, the proportion of omission responses was only substantially present at the shortest SOA of 125 ms, i.e., at relatively high presentation rates (up to 8 Hz). This time course suggests that if the sounds are close enough in time, then they form a perceptual unit and, hence, compatible with the temporal window of integration (15,16,60,61) that integrates neighboring sounds and/or omissions into a single percept.
While we could not find significant differences when the omission responses were preceded by a random number of reference stimuli (random), compared to a fixed number of reference stimuli (periodic), several studies have shown that expected omissions evoke a smaller response than unexpected omissions (19,25,45). It may be that the number of preceding repetitions (nine stimuli) in our periodic sequences is too large for producing an accurate expectation of when it should happen in our rodent models; if the animal loses track of how many repeated stimuli have happened, then these omissions would be perceived as events preceded by a random (albeit large) number of stimuli. This is supported by the rapid dynamics of adaptation or the responses to the standard stimulus following the omission, which return to baseline in just one to two stimuli (Fig. 7). Related to this, we only find a correlation between the number of preceding stimuli and the omission response under the anesthetized condition, while it may have already reached a plateau in the awake case. This finding relates to the previous work by Dürschmid et al. (62), which demonstrated that transitional probabilities are different in AC and prefrontal cortex by increasing the number of standards before a deviant occurrence. They observed that this increases the expectation of prediction errors in the prefrontal cortex, while this was not the case for the AC (62). This is in accordance with our data related to the number of preceding standard responses before an omission that did not show an enhancement but a sustained response. This is important because omission-selective responses have been reported in humans (19,27,44,63,64), suggesting the preactivation of neural circuits needed to process the actual physical inputs.
Additional related work in humans has shown an interaction between arousal (in the form of attentional set and expectation) and mismatch responses. For example, an elegant study by Auksztulewicz and Friston (65) concluded that temporal attention and sensory expectation produced opposing effects on evoked response amplitude, when orthogonally manipulated in an auditory mismatch paradigm. MMN was enhanced by attention, speaking against its supposedly preattentive nature. The distinction between expectation and attention may be useful when interpreting our results in the awake mice. In other words, there is a difference between expecting something and attending to a sensory stream, and these can have opposite effects on the amplitude of mismatch responses (and thus, presumably, omission-related responses).
Methodological considerations
An important strength in our study is that we have used two animal species and two different recording preparations in two different laboratory setups. This approach allowed us to distinguish the effect of species and brain state on the omission responses and demonstrate that the results under anesthesia do not differ significantly between mice and rats, mimicking previous results (66)(67)(68)(69)(70). In contrast, the frequency of omission-sensitive cells and the magnitude of S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E the omission responses were found to be larger in awake mice than under anesthesia. This would be in line with the differences in the level of response for humans in awake compared to unresponsive wakefulness state [see, e.g., (67)]. Furthermore, this is also consistent with our previous work comparing rat, mouse (8,71,72), and others that show a similar structural and functional organization of these rodents' auditory system [e.g., (73)(74)(75)], speaking in favor of the existence of genuine omission responses across species that can be modulated by arousal state.
It is unlikely that the omission responses we observed are not genuine responses or simply the reflection of rebound or entrainment effects as discussed elsewhere in the context of adaptation models (17,76) and our own control data shown in fig. S5. If these omission responses were just long latency activity evoked by the previous stimulus, then such activity would remain when probing the neurons with longer SOA sequences, but that is not the case in our data (see fig. S5). In addition, such an adaptation model would not account for omissions evoking a larger response than the tone they replaced as occurring in many of our responses; if anything, the effect would be opposite. Moreover, there are also models that support the predictive framework for omission responses. The modeling study by Wacongne et al. (19,20) demonstrated that omission responses were larger than stimulus responses in blocks containing quintuplets of tones where the last one was replaced by an omitted tone, whereas the adaptation model would predict equal omission responses in both instances. Moreover, entrainment should be stronger for periodic sensory stimuli (77).
Although there is no question that omission responses occur in AC, a relevant weakness in our study is that AC responses are biased toward A1 (lemniscal field) in the awake mouse and toward the nonlemniscal fields in the anesthetized rat, making it difficult to conclude if the differences in the response magnitude between the two preparations are related to the arousal state or the field location of the recordings. Although the results are in line with the hierarchical organization of prediction error that we have seen in our previous studies (8), future studies should clarify this confound.
We found a relatively small percentage (16.7%) of omission neurons in the IC of the anesthetized rat, at least compared to the amount of omission neurons in the anesthetized rats (19.5%) and mice (21.7%), as well as awake mice AC (35.6%). In the split data analysis, the IC responses are revealed to be less robust or statistical deviations. While the omission responses in the IC are thus not as robust as in the AC, the fact that the IC recordings were underpowered and were only conducted in the anesthetized state preclude a conclusion on the existence of omission-sensitive responses in the IC. For example, a recent study showed that a small proportion of IC neurons were pattern sensitive (78) and that a sparse set (~5%) of neurons in AC showed a high-rate prolonged burst firing responses to trained sounds (79). Further studies in the IC under the awake state with greater neuron numbers are required to resolve this question.
Relation to predictive coding theory
As in other sensory systems, where omission responses have been best explained by a deviation between expected and actual visual stimulus (50), our results are consistent with a predictive coding framework, in which sensory input is compared to an internal model of the environment to detect deviations from expectations. Because bottom-up inputs cannot account for omission responses, our results are consistent with the hypothesis that prediction is a hierarchical process encompassing subcortical and cortical neuronal circuitries and can be observed in two species and two brain states. Together, these findings are in agreement with two recent studies that have combined 9.4-T functional magnetic resonance imaging and high-density electrocorticographic recordings with a local-global auditory sequence paradigm to assess the hierarchical depth of auditory sequence processing in the marmoset brain (80,81). These authors demonstrate that the neural responses to omissions cannot be explained by any modulation of feedforward propagation and should contain specific information of upcoming predictive information (80,82). Hence, omission responses thus allow the study of top-down prediction signals decoupled from bottom-up input signals, i.e., without sensory input. Such a perspective supports the notion of an endogenous neural activity that could reflect the generation of predictions (83,84).
An important issue to be considered is whether responses encoding predictions can be disentangled from responses encoding prediction errors. According to predictive coding, omission responses can provide rather clear access to the representation of the prediction; as in this case, the prediction error is a, in the forward path inverted, copy of the prediction (26,85). Because the omission responses that we detected are elevations in firing rate, they likely provide access to the neural representation of prediction in the AC (64,81,86). Another distinction between prediction error and prediction responses may be feasible in timefrequency measures (87,88). Thus, positive omission responses are likely related to predictions (10,21) and within predictive coding theory represented by interneuron-mediated modulation of pyramidal output. Hence, they may share the neuronal and local circuitry of stimulus-specific adaptation that involves multiple neuron types (35,89,90), supporting feedforward and feedback inhibition and excitation of pyramidal output, which can detect deviations from regularity (35,59). Our data do not allow us to definitely distinguish whether we recorded from excitatory or inhibitory interneurons; we have observed omission responses in multiple AC layers that encompass different neuron types, suggesting that omission responses are generated and enhanced at the network level through a hierarchical process consistent with the predictive coding perspective (1, 37-39, 43, 84, 91). Canonical microcircuit formulations of predictive coding are quite consistent: They suggest that excitatory pyramidal cells encode prediction errors (92,93). Their activity is the difference between ascending stimulus information and the activity of interneurons. In this formulation, the interneurons represent the predictions, e.g., derived from regularities in the environment. The pyramidal cells are hypothesized to be located in superficial and granular layers, while the prediction carrying inhibitory neurons are hypothesized to be located in supragranular layers. However, the latter activity is also thought to be passed to infragranular pyramidal cells (92). This architecture makes it difficult to have a simple hypothesis of where omissionrelated responses should be found. This follows because they could be associated with predictions in deep pyramidal cells or the inhibitory interneurons in superficial layers. However, a more definitive interpretation would probably rest upon tracing studies or optogenetic characterization of the cells recorded. Thus, future studies are needed to separate pyramidal neurons from inhibitory interneurons to get a clearer picture of predictive coding in canonical microcircuits.
Another interesting, yet difficult, issue to resolve is how the different components described in the classical adaptation studies, i.e., prediction error and repetition suppression (8), relate to the deviant and omission responses. To address this issue, it would have been necessary to carry out additional control recordings using the cascade or many-standard sequence as in, e.g., (8,23,35,94). Unfortunately, these sequences are very time consuming and the current inclusion of other conditions (multiple SOAs, deviant) precluded the recording of these additional controls. This is definitely an interesting and open question but a limitation of the current study that should be explored in the future. In summary, the demonstration of omission responses in IC and AC, particularly in awake but also in anesthetized preparations, confirms that the auditory system does not require an external stimulus trigger to detect a deviation from expectations (95) and indicates that the auditory brain internally generates a prediction of future sensory input.
Experimental procedures
Experiments were performed on 22 anesthetized adult female Long-Evans rats (9 rats were used for IC recordings and 13 rats for AC recordings) and 6 C57/B6 mice (3 awake and 3 under urethane anesthesia). The body weight ranged from 200 to 300 g in rats (9 to 17 weeks of age) and 25 to 35 g in mice (14 to 18 weeks of age). Each individual animal was used to record from a single auditory station, either IC or AC. Experimental procedures in anesthetized rats were carried out at the University of Salamanca, and all the experimental procedures and protocols were adjusted to the directives of the Directive of the European Communities (86/609/CEE, 2003/65/CE, and 2010/63/UE) and RD53/2013 Spanish Legislation for the use and care of animals. All the details of the study were approved by the Bioethics Committee of the University of Salamanca (USAL-ID-195 and USAL-ID-574). Experiments in awake mice were performed at the Central Dierenlaboratorium at the Radboud University Medical Center. The experimental procedures and protocols were approved by the Dutch central commission for animal research (Centrale Commissie Dierproeven) and implemented according to approved work protocols from the local animal welfare body (project number 2017-0041).
Experimental procedures in anesthetized rats
The experiments were carried out inside a chamber with acoustic insulation and electrical protection. Sound stimuli were generated using the RZ6 Multi I/O processor [Tucker-Davis Technologies (TDT)]. Sounds were presented monaurally (right ear) through a closed system using a custom-made earphone (1 to 45 kHz) coupled to a custom-made cone, as a substitute for traditional ear bars. Thus, sounds were presented through the speaker under a closed-field condition to the ear contralateral to the left IC or AC.
The software was programmed with OpenEx Suite (TDT) and MATLAB. The sound system response was flattened with an onsite calibrated finite impulse response system using a one-fourthinch condenser microphone (model 4136, Brüel & Kjaer), a conditioning amplifier (Nexus, Brüel & Kjaer), and a dynamic signal analyzer (Photon+, Brüel & Kjaer). The speaker output was adjusted to ensure a flat spectrum (±2 dB) between 0.5 and 44 kHz, and the second and third harmonic components in the signal were >40 dB below the level of the fundamental at the highest output level [~76 dB of sound pressure level (SPL)]. A single neuron was recorded at a time, using a self-made glass-coated tungsten electrode whose impedance ranged from 1.5 to 3 megohm at 1 kHz.
Analog signals were digitized with an RZ6 multi-I/O processor, Medusa RA16PA preamp, and ZC16 main stage (TDT) at 25 kHz of sample rate and amplified 251×. The neurophysiological signals for multi-or single-unit activity were band-pass filtered between 0.5 and 4.5 kHz.
Experimental procedures in anesthetized and awake mice
The experiments were carried out inside an acoustically insulated and electromagnetically shielded and grounded booth. Sound stimuli were generated and recorded using a multichannel data acquisition device (USB-6351, National Instruments) through a custom-written MATLAB GUI (Controller). Sounds were presented to both ears using a speaker (T250D, Fostex) placed in front and above the animal. The output of the speaker was calibrated to have a flat spectral profile over the range of 2 to 80 kHz within 5 dB using an inverse linear filter, estimated before the experiment using an ultrasonic microphone (CM16/CMPA, Avisoft Bioacoustics) corrected for its own frequency-dependent sensitivity curve.
Surgical procedures in anesthetized rats
The initial surgical procedures were the same for the two auditory areas to be recorded (IC and AC). Electrophysiological procedures differed only in craniotomy location and recording electrode placement for each station.
Anesthesia was induced and maintained with urethane [1.9 g/kg, intraperitoneally (ip)] with supplementary doses (~0.5 g/kg, ip) administered as necessary. These replenishments were given to ensure a deep and stable anesthetic level whenever corneal or pedal retraction reflexes appeared. Urethane maintains a balanced neuronal activity better than other anesthetic agents that have a slight effect on inhibitory and excitatory synapses and that is why it was the selected anesthesia method (96).
At the beginning of surgery, both dexamethasone (0.25 mg/kg) and atropine (0.1 mg/kg) were administered to reduce cerebral edema and bronchial secretions, respectively. Lidocaine was also injected around the pinna tissue to achieve a higher anesthetic level in these areas (8). In addition, an isotonic glucosaline solution [5 to 10 ml every 6 to 8 hours, subcutaneously (sc)] was periodically administered to avoid dehydration. During all the experimental procedures, the animals were artificially ventilated, controlling the CO 2 levels, and the temperature was monitored with a rectal probe (36°to 38°C) and maintained with a heating blanket (Cibertec) placed under the animal.
Once the animal reached a surgical state of anesthesia, the ears were prepared by removing the cartilage to obtain a greater exposure of the ear canal. The trachea was cannulated to maintain ventilation artificially, and corneal and pedal withdrawal reflexes were constantly monitored to ensure that we maintained a deep anesthetic level during surgery and as uniform as possible throughout the recording procedure.
Once the animal was prepared, it was placed on the stereotactic frame; the upper jaw was attached to a bite bar and the restraint was carried out using ear bars. At this time, the speaker was also placed on the right ear of the animal, while the left ear canal was covered with plasticine.
Surgical procedures for IC (anesthetized rats)
For IC recordings, a craniotomy was performed on the left parietal bone to expose the cerebral cortex overlying the left IC. The dura was removed and the surrounding tissue was covered with cotton balls soaked in 0.9% saline solution to maintain hydration of the area.
Surgical procedures for AC (anesthetized rats)
For AC recordings, the skin and temporal muscles were removed from the left side of the skull and a 6-mm by 5-mm craniotomy was performed on the left temporal bone to expose the entire AC. The dura was removed and the cisterna magna was drained to prevent brain herniation. The exposed cortex area and surrounding tissue were covered with a transparent layer of agar to prevent desiccation and, thus, stabilize the recordings.
Surgical procedures for AC (awake and anesthetized mice)
Mice were implanted with a headpost for head fixation at the ages of 8 to 12 weeks. In the second surgery, performed when animals were 12 to 16 weeks old, mice were implanted with a silicon NeuroNexus neural probe aiming at the left primary AC, field A1, based on stereotaxic coordinates (3.2 mm post-bregma, 5.9 mm lateral from the midline) measured on the skull, estimated from (97), and confirmed preimplantation via intrinsic optical imaging of noise burst sequences (0.2-s long, 0.1 s interstimulus interval, 70 dB of SPL, 10 sounds per trial, 30 trials). A craniotomy (~1 mm by 2 mm) was opened using a microdrill centered on the part of AC that showed the strongest response in the intrinsic imaging. In the awake mice, the implant was chronic, mounted on a custom-designed drive (variant of the EDDS drive, Microprobes for Life Science), allowing the electrode array to be repositioned. The surgery was performed under general isoflurane anesthesia (1.5 to 2% dosage; E-Z Anesthesia, E-Z Systems Corporation). From 24 hours before to 48 hours after the surgery, the animal received an analgesic dissolved in water (1 ml/liter; carprofen (Rimadyl cattle), Zoetis], supplemented by another dose of carprofen at the start of the surgery (5 mg/kg, sc). In the anesthetized mice, the electrode array was inserted acutely for the period of the recordings under urethane anesthesia using one initial dose (1.7 g/kg, ip) and one to two supplementary doses (0.2 g/kg, ip). In addition, dexamethasone (0.2 mg/kg, sc) was given under both conditions after induction of anesthesia to reduce local bulging of the brain, and a ground screw was implanted over the right frontal cortex.
Electrophysiological recording procedures Auditory brainstem response recording in anesthetized rats
Once the animal was under anesthesia, before starting the surgery and therefore the recording sessions, the auditory brainstem responses were recorded using subcutaneous electrodes to ensure that the animal had normal hearing, using the RZ6 Multi I/O processor (TDT) as hardware and BioSig software (TDT). The auditory brainstem response stimuli were generated following standard procedures (0.1ms clicks presented at a rate of 21 Hz, applied monaurally to the right ear in 10 dB of upward steps between 10 and 90 dB of SPL using a closed-field speaker).
Procedures for electrophysiological recording in IC (anesthetized rats)
For measurements in the IC, an electrode was placed orthogonal to the surface of the brain (forming an angle of 20°with the horizontal plane, oriented caudally), advanced between 7500 and 10,000 μm from lambda to bregma and 500 and 3000 μm to the left ear, and then lowered into the brain between 4400 and 5600 μm using a piezoelectric micromanipulator (Sensapex) until we observed a strong spike activity synchronized with the search stimulus train.
Procedures for electrophysiological recording in AC (anesthetized rats)
Once the AC was exposed and the tissue was stabilized, an enlarged image of the exposed surface was taken (×25). The image included a pair of landmarks previously marked before the craniotomy, on the dorsal border of the temporal bone, indicating the absolute scale and position of the image relative to bregma. This image was displayed on a computer screen, and a grid was overlaid on it that would serve as a guide so we would be able to mark the recording electrode placement for each recording. The snapshot was taken with a single lens digital reflex camera (D5100, Nikon) coupled to the surgical microscope (Zeiss) through a lens adapter (TTI Medical) (7).
For AC measurements, we placed the electrode in contact with the AC and as perpendicular as possible to its surface. After this, we advanced it between 0 and 2000 μm in depth. The recording tracks were spaced 250 to 500 μm and placed through the cortical regions of interest, while blood vessels were avoided. The vascular pattern was used as a local reference to mark the position on the image where the tract was made, considering that this map varies between animals. At the end of the experiment, the limits and the relative position of the auditory fields were identified according to the pure-tone frequency response topographies in rats (7). We estimated the depth of each neuron from the reading of the micromanipulator relative to the brain surface (Fig. 8A). We used the depths of the layers in the rat AC as stated by Pérez-González et al. (35) (i.e., the granular layer would extend between 457 and 614 μm).
The characteristic frequency of each recording track was calculated as the average characteristic frequency of all neurons recorded on that track, including a rapid multiunit activity frequency response area recording performed between 400 and 550 μm in depth, corresponding to layers IIIb to IV. The reversals of the characteristic frequency progression defined the boundaries between the cortical fields (98), so that most recordings could be assigned to one specific field: primary or secondary and their corresponding subdivisions: AAF, VAF, A1, PAF, and SRAF.
Procedures for electrophysiological recording in AC (awake mice)
For the time of the recording, the animal was placed on a platform located in the middle of a booth and head-fixated using an implanted headpost. All recordings were performed using NeuroNexus silicon probes with 64 channels. The recording sites were distributed differently per probe; in particular, we used two 4x16 (A4x16-Poly2-5mm-23s-200-177 and A4x16-2.5mm-tet/lin-300/125-333-121/177-PC) configurations. Data were digitized using a combination of a 64-channel (Intan Technologies) headstage connected to an Open Ephys recording system (Open Ephys, Cambridge, USA). Data were collected at a sampling rate of 30 kS/s, band-pass filtered between 0.3 and 6 kHz, and spike-sorted to get multi-and singleunit responses using the openly available function autoSortC from the Controller package. The sorting used automatic criteria for cluster-cutting rather than human decisions, which are prone to be variable from day to day. Sorting was performed separately for each prong. The average number of cells/electrode in a single session was 0.71 after spike-sorting, which is a typical yield, e.g., in comparison to other recording techniques (0.625 cells per electrode) (99). We estimated the depth of each cell from the depth of the array relative to the brain surface and the position of the recording sites in each array configuration (Fig. 8). The depths of the layers were estimated from a coronal section of the mouse AC in the Allen Brain Atlas (100) (i.e., the granular layer would extend between 355 and 477 μm).
Location of cells and recording of the frequency response area (anesthetized rat)
The sounds used to search for neuronal activity were trains of noise or pure tones (1 to 4 stimuli/s). These stimuli had a short duration (30 ms) to avoid a strong adaptation. Sound parameters such as frequency, intensity, presentation rate, or type of sound such as white noise, narrow band noise, or pure tones were manually varied as necessary to facilitate mismatch.
For each neuron recorded, the frequency response area was calculated, which is the response magnitude map for each pair of frequency/intensity data. This was calculated in the first place to ensure that the frequencies and intensities used were contained in the response range. To obtain this frequency response area, a sequence of tones was presented at a rate of 4 Hz, randomly varying the frequency and intensity of the tones presented (three to five repetitions of each tone).
For the AC recordings, the coordinates of the neurons over the AC extension were measured from the micromanipulator as the position of the recording electrode in reference to bregma. For the IC recordings, at the end of the recording session, a 5-μA current was driven for 5 s through the recording electrode to produce an electrolytic lesion. Animals were euthanized with a lethal dose of pentobarbital, and the head was removed and immersed in a mixture of 1% paraformaldehyde and 1% glutaraldehyde in 1 M phosphatebuffered saline. After fixation, tissue was cryoprotected in 30% sucrose and sectioned in a freezing microtome to obtain 40-μmthick coronal sections. The sections were stained using 0.1% cresyl violet to facilitate identification of cytoarchitectural boundaries.
Online viewing
For the anesthetized experiments, we visualized the electrophysiological recordings online using custom-written software with MATLAB (MathWorks) and OpenEx Suite (TDT). Multiple-unit activity was automatically extracted by manually setting a unilateral action potential threshold above the background noise as an accurate estimate of neuronal population dynamics (101). Once a neuron was isolated and confirmed to be stable, the stimulation protocol described in the next section was applied. For the awake experiments, the activity was monitored online using the Open Ephys GUI, and stimuli were presented if single-unit neural activity could be visually detected.
Stimulation parameters
The sounds used for stimulation were sinusoidal tones of 75-ms duration, including up and down ramps of 5 ms, at a frequency and intensity within the response area of the neuron. In the single-electrode recordings in the rat, the frequency-intensity combination could be chosen specifically for the single neuron recorded at a time. In the many-electrode recordings in the mouse, we chose a set of frequencies that made a good match with the frequency response areas of the currently recorded cells. We used a fixed set of three intensities in the mouse ( [50,60,70] dB of SPL) that were intense enough to activate all recorded neurons. These sounds were presented in an oddball sequence of 400 tones long. The oddball sequence consisted of a repeating tone (90% standard probability) that was occasionally omitted (10% deviant probability).
The omission of the tone was presented in two different ways within the oddball sequence: periodically or randomly, always with the same probability of appearance. A minimum of three standard tones were presented before each deviant. These sequences were presented with different SOAs (125, 250, and 500 ms) to control for confusion between actual responses to omissions and responses to previous standard stimuli. To double-check that no sound was played during omissions, we recorded the output of the speaker while playing the stimulation sequence to check that there was no acoustic signal during the omissions (fig. S1).
In a second set of recordings, we compared deviant tones with omission responses in the AC of anesthetized rats and awake mice (Fig. 6). For this purpose, identical sequences were used with either deviant tones or omissions occurring at the same positions in the sequence. In the single-electrode recordings in the rat, the standard and deviant frequencies were chosen to lie symmetrically around the characteristic frequency at the same intensity to elicit similar response rates (Fig. 6A). The paradigm was run with either of the two stimuli as standard and deviant. In the multielectrode recordings in the mouse, we chose a pair of frequencies that provided the best match to the currently recorded cells and also ran the paradigm with both frequencies for standard and deviant. All subsequent analyses for omissions and deviants were identical. Comparisons were then performed on the set of omission-sensitive neurons.
The output of the system was also recorded using a one-fourthinch condenser microphone (model 4136, Brüel & Kjaer) to ensure the absence of any sound or noise during omissions ( fig. S1). The spectrogram in fig. S1A shows a portion of the presented periodic omission sequence, reflecting the intensity of the sounds that the system reproduces. The power spectrum shown in fig. S1B allows comparison between the signal produced during a real stimulus (blue) and the omitted signal (red) and confirms that no sound was reproduced during the omission.
Data and statistical analysis
After the single-and multiunit response times were established, we selected the cells that elicited responses for omissions. We did that by comparing firing rates during omitted tones to the pre-stimulus firing rates. For this purpose, comparison windows had to be defined. The pre-stimulus/stimulus windows were defined individually for each SOA. These windows were chosen after visual inspection of the responses, aiming to include the bulk of the evoked responses while leaving enough time to calculate the pre-stimulus activity for the next trial. Because of the long AC responses, for the 125 SOA recordings, there is some overlap between the prestimulus and stimulus windows. The pre-stimulus window start was defined as 25, 75, and 195 ms before the stimulus onset for 125, 250, and 500 ms of SOA, respectively. All pre-stimulus windows ended 5 ms after the start of the stimulus, because the evoked response latency was always longer great than 5 ms. This time also marks the start of the "stimulus window." Stimulus windows ended at 120, 150, and 225 ms for 125, 250, and 500 ms of SOA, respectively. Note that stimulus here refers to both tone and omission. These values are summarized in Table 3.
Firing rates were computed in all of the predefined windows and used in subsequent analyses to compare the responses under the tested conditions. PSTHs (10-ms bins; Fig. 2 and fig. S5) were aligned to show the average response around the omission. Peak latencies were calculated for each tone or omission as the time (relative to stimulus onset) of the maximum value in the stimulus PSTH (1-ms bins).
Omission responses were identified as an elevated firing rate during the omitted tones. Specifically, responses during the omission window were considered significant if their mean firing rate was significantly larger than the mean firing rate of the cell's activity, immediately preceding the omission window.
Using these time windows, we collected for each omission the firing rates of preomission activity and the firing rates of omission activity. To avoid a positive bias due to double dipping, we used split data analysis (32), i.e., only every other omission was used for statistical testing, while the complementary set was used for all subsequent analyses. These sets are referred to as "statistics set" and "analysis set." The starting point, the first or second omission, for the division into the two sets was randomized independently between cells, i.e., either using the omissions at positions [1, 3, 5, …] or [2, 4, 6,...] for the statistics set and, correspondingly, the complement for the analysis set. In the statistics set, the preomission activity was compared against the periomission activity over the length of the stimulus window (see above) using a single-sided paired Wilcoxon signed-rank test. A neuron was termed an "omission-sensitive neuron" if the P value was <0.05. False positives in this classification will not translate to elevated omission responses, based on the split data analysis, where only systematic differences, present in both splits of the data, are retained in the plotting. The significance on the population level is then subsequently tested on the part of the data not used for statistical testing (see Fig. 3).
The iOmi used to quantify the comparison between sound-and omission-evoked firing rate (Fig. 4A) was defined as: iOmi = (Omi FR − Stim FR)/(Omi FR + Stim FR). The firing rate for stimuli was averaged from all stimulus presentations.
The statistical analysis on the population level was performed using ANOVA (Kruskal-Wallis analysis for one-way comparisons) followed by pairwise comparisons between individual conditions using Wilcoxon signed-rank test with Bonferroni correction for the number of tests performed. Rates were transformed to approximately normal distributions by taking their logarithm, given that neural responses are distributed exponentially particularly at low rates. For differences between firing rates, e.g., stimulus versus omissions, we computed the logarithm of the difference in rates, before applying ANOVA.
For testing whether the percentage of omission-sensitive neurons in each layer was as expected if these neurons were randomly distributed across layers, we sampled the same number of omission-sensitive neurons according to the empirical distributions across layer groups (i.e. granular, infragranular, and supragranular) using bootstrapping (300,000 samples) and then compared the actual distribution of omission-sensitive cells to the bootstrapped densities. The P values represent their location in the distribution at the closest end of the density.
Supplementary Materials
This PDF file includes: Figs. S1 to S5 Legend for data S1 Other Supplementary Material for this manuscript includes the following: Data S1 View/request a protocol for this paper from Bio-protocol. | 2023-06-16T06:16:22.859Z | 2023-06-01T00:00:00.000 | {
"year": 2023,
"sha1": "c0d561fe3375250852aa8f48f8bf23d0849436c2",
"oa_license": "CCBYNC",
"oa_url": null,
"oa_status": null,
"pdf_src": "PubMedCentral",
"pdf_hash": "1f56b3dc920e32a9d1d0ef0023bbf87a97fc3134",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
67922758 | pes2o/s2orc | v3-fos-license | Osteoporosis: It’s More Than Calcium
Osteoporosis is a major public health concern with millions of Americans being affected. This painful and oftentimes crippling disease is multifactorial in nature. Therefore, the development of osteoporosis is related to more issues than a person’s intake of calcium. As a result, osteoporosis is not age or gender dependent. Risk factors for osteoporosis include proper nutrition including calcium intake, heredity, age, gender, physical inactivity, smoking, excessive alcohol use, and the use of several medications. With such a variety of risk factors, everyone should assume risk for disease development and thus take steps for the prevention of this bone fragility disease. © 2004 Californian Journal of Health Promotion. All rights reserved.
Osteoporosis is a major public health concern in the United States.An estimated 28 million Americans are already afflicted with this disease and approximately 1.5 million new fractures occur each year (Barefield, 1996;Boughton, 1999;Krall & Dawson-Hughes, 1999;National Institutes of Health (NIH), 2000).In addition to those already afflicted, another 18 million Americans have low bone density, placing them at risk for this disorder (NIH, 2000).
Osteoporosis is a painful, sometimes crippling disease.The incidence of osteoporosis has reached epidemic levels in the United States and is responsible for considerable death, illness, loss of independence, decreased quality of life, and associated economic costs (Gerrior, Putnam, & Bente, 1998;National Osteoporosis Foundation [NOF], 1998).Estimates for the United States indicate that 13-18% of women over the age of 50 are afflicted with osteoporosis.Even more startling is the estimate that another 37-50% of women over 50 have some degree of osteopenia or low bone density (Barefield, 1996;Gerrior, et al., 1998).
Fractures that are a result of bone fragility account for extensive morbidity, mortality, and loss of function (Lappe, 1994;Turner, Taylor, & Hunt, 1998;Ullom-Minnich, 1999).The financial costs associated with osteoporotic fractures include direct medical charges, rehabilitation, and extended treatment facilities.Osteoporosis-related hip fractures alone result in estimated costs of $12.8 billion to $17.8 billion per year.Rehabilitation and institutionalization account for approximately 40% of these costs while less than 1 percent is due to lost productivity (Barefield, 1996).Rehabilitation and institutionalization costs are the largest majority since almost half of the individuals hospitalized with osteoporosis-related hip fractures never fully recover.
Twenty-five percent of the total hip fractures result in the person being severely handicapped.Another 20% of these people die within one year of a hip fracture (Barefield, 1996;Lappe, 1994).As the population subgroup of Americans over the age of 65 grows, estimates indicate osteoporosisrelated costs will be greater than $62 billion by 2020 and $200 billion by the year 2040 (Barefield, 1996;Cummings, Rubin, & Black, 1990).
Osteoporosis is a complex multifactorial condition characterized by low bone mass and deterioration of bone tissue, with a consequent increase in bone fragility and susceptibility to fracture (Notelovitz, 1993;Siddiqui, Shetty, & Duthie, 1999).Osteoporosis is often referred to as the silent thief because osteoporotic bone loss takes place gradually throughout the course of many years without any signs or symptoms (Boughton, 1999).This reduction in bone mineral density (BMD) later manifests itself as low-trauma fractures.
To better understand osteoporosis, it is helpful to understand the process of bone formation.Bone is living tissue that is constantly being renewed in a two-stage process of formation and resorption that occurs throughout life (Boughton, 1999).During different life stages, the balance of these processes may weigh heavier in one direction than the other.For example, during adolescence, bone formation occurs at a greater rate than bone resorption.The formation stage is characterized by the building of new bone to replace old bone by cells called osteoblasts.The resorption phase is characterized by a higher level of bone breakdown than formation as a result of cells called osteoclasts (Boughton, 1999;Smith, 1993).
There are two main type of bone that are recognized: trabecular or spongy bone (25%) and cortical or compact bone (75%) (Smith, 1993).Trabecular bone forms the internal support system of the bone and is metabolically more active than cortical bone (by about ten times) (Notelovitz, 1993;Smith, 1993;Wardlaw, 1993).Cortical bone makes up the outer shell of bone and predominates in the shafts of long bones.Although each bone in the body contains both cortical and trabecular bone, the relative proportions differ.In osteoporosis, both cortical thinning and a loss of trabecular support are evident (Wardlaw, 1993).However, trabecular bone loss occurs at a greater rate thereby increasing fracture risk in areas where trabecular bone makes up the largest portion of the bone's structure such as in the vertebrae, wrist, and the ends of long bones (i.e. the hip) (Boughton, 1999;Physicians Desk Reference (PDR), 1999;Smith, 1993;Wardlaw, 1993).
Osteoporosis is not always a result of bone loss and can be characterized as either primary or secondary in nature.Primary osteoporosis can occur in both genders at all ages but often follows menopause in women and occurs late in life in men (NIH, 2000).Postmenopausal osteoporosis is known as primary osteoporosis Type I and is characterized by an increased bone resorption that primarily affects trabecular bone.Type I primary osteoporosis is directly linked to the decreased production of estrogen that coincides with menopause (Peterson, 2001;Wardlaw, 1993).Rapid bone loss is osteoclastmediated and occurs in women within the first 5 to 10 years after menopause (Peterson, 2001).Primary osteoporosis Type II is a slow bone loss resulting from a proportionate loss of trabecular and cortical bone usually due to a decrease in bone cell activity accompanying aging.This type of osteoporosis predominately afflicts men and women over the age of 70 years and is called senile osteoporosis (Glaser & Kaplan, 1997;Peterson, 2001;Wardlaw, 1993).
Secondary osteoporosis usually occurs as a result of another disease or medication.The most common medical conditions include chronic renal disease, hypogonadism, hyperthyroidism, Cushing's disease, and some forms of cancer (Wardlaw, 1993).Surgical procedures such as an early oophorectomy or total gastrectomy, can lead to bone loss.Additionally, some medications including anticonvulsants, corticosteroids, Depo-Provera, and Heparin, have toxic effects on bone and increase bone loss (Glaser & Kaplan, 1997;Kulak, Schussheim, McMahon, Kurland, Silverber, Siris, et al.., 2000;Peterson, 2001).Regardless of the cause of osteoporosis, the consequences are devastating to the financial, physical, and psychosocial aspects of one's health (NIH, 2000;Donohue, 1999;National Osteoporosis Foundation [NOF], 2000;Moon, 2000).
Researchers agree that osteoporosis is not ageor gender-dependent in who it targets.While white postmenopausal women have the highest incidence of osteoporotic fractures, and most of our knowledge about diagnosis and treatment is derived from research on this population, the disease does not go unrecognized in other ages and races, or in men (Moon, 2000).However, in comparison to men, women are at higher risk for developing osteoporosis and have a lifetime risk of an osteoporotic fracture as high as one in three (Mark & Link, 1999).One third of women aged 65 and older will have at least one vertebral fracture in their lifetime and 33% will experience a hip fracture by age 90 (Aufdemorte, 1991).However, due to the diseases' silent nature, there are few reliable statistics on how many women, including young women, are already developing low bone density and osteoporosis.A study conducted by Tokar and colleagues among a convenient sample of 165 college-aged women found 1% of the participants to already have osteoporosis and an additional 14% of the participants to have low bone density or osteopenia (Tokar, Ford, Turner, & Denny, 2003).
Osteoporosis is a condition normally associated with elderly women.However, an increase in the number of women in their 20s and 30s suffering from osteoporosis has been reported (Hart & Dip, 1996).Among postmenopausal women, factors associated with osteoporosis diagnoses include age, race, and family history (Turner et al., 1998).Additional research indicates that other factors may also play a role in the development of this disease.Among these are factors that are common among women who are younger than the age of menopause.These other factors include low calcium intake, physical inactivity, smoking, excessive alcohol intake, use of steroid medications, and eating disorders (Hsieh, Novielli, Diamond, & Cheruva, 2001;Moon, 2000).The development of osteoporosis is associated with many risk factors that transcend age.
The prevention of osteoporosis is linked to strong bones being built during childhood and adolescence and being maintained throughout adult life (Mark & Link, 1999).Since clinical manifestations of osteoporosis often do not appear until later in life, one of the most important factors in preventing osteoporosis is the attainment of an optimal peak bone mass during adolescence and young adulthood (Cromer & Harel, 2000).Peak bone mass is defined as the highest level of bone mass achieved through normal skeletal growth (Masi & Bilezikian, 1997).The achievement of peak bone mass occurs in the third decade of life; about the age of 30 (Masi & Bilezikian, 1997;NIH, 2000;PDR, 1999).During the next 10-15 years, the bone structure stays relatively stable with slight reductions in mass if certain lifestyles are practiced.
However, at the age of menopause, dramatic decreases in bone mass are lost due to changes in hormone production.Regardless of these hormonal changes, debilitating bone loss in not inevitable.The physiologic processes that lead to osteoporosis occur over much of a patient's lifespan and are amenable to interventions throughout that lifetime (Katz, & Sherman & DiNubile, 1998).
The optimization of bone health is a process that must occur throughout the lifespan because once a woman experiences a fracture due to bone fragility, no known therapy can rebuild the damaged bone to a healthy level.Therefore, measures taken to prevent bone fragility are of vital importance (Anderson & Metz, 1993;Blalock, et al., 1996).Evidence indicates that young women can increase their peak bone mineral density, promote long-term bone health, and reduce the risk of disease later in life by following effective dietary exercise and lifestyle practices (Mark & Link, 1999).Because there is currently no medical intervention to completely reverse the effects of osteoporosis, the most powerful tool to reduce the incidence of osteoporosis is prevention through health education (Mark & Link, 1999).
Osteoporosis prevention programs have traditionally been marketed toward women later in life (postmenopause).As a result, programs have emphasized nutritional changes, exercise programs, and hormone replacement therapy to prevent further bone loss.Few, if any, programs have been developed specifically for younger women as a means of preventing this debilitating disease (Blalock et al., 1996;Jamal, Ridout, Chase, Fielding, Rubin, & Hawker, 1999).This may be due to the seeming contradiction of young women having osteoporosis.However, if young women are to prevent or delay the development and onset of osteoporosis in later life, then osteoporosis prevention needs to begin decades before women experience menopause (Kasper, Peterson, Allegrante, Galsworthy, & Gutin, 1994).A major component of such a prevention effort is education about behaviors that impact skeletal growth, the importance of regular menstrual cycles, proper nutrition, adequate physical activity, and cautions about medication use, smoking, and excessive alcohol intake.
Diet and Osteoporosis
Nutritional intake is a key component of osteoporosis prevention.
Several different dietary factors play a role in either the advancement or prevention of osteoporosis while still others negatively effect bone mass.
Dietary Factors that Positively Effect Bone Health
Calcium.The roles of calcium in nature are numerous.This is also true when reviewing its roles in the human body.Calcium's most notorious role is that of structure or mechanics and is expressed in the mass, hardness, and strength of the bones and teeth (Weaver & Heaney, 1999).This is further evident with more than 99% of the calcium in the body being used and present in bones and teeth (Wardlaw, 1997).Overall, calcium accounts for 1-2% of a person's body weight (Weaver & Heaney, 1999).
The most documented and accepted health benefit of calcium is its role in bone health.In bone, calcium exists primarily in the form of hydroxyapatite (Ca 10 (PO 1 ) 6 (OH) 2 ) (Weaver & Heaney, 1999).When bones form, the calcium salts form crystals called hydroxyapatite, on a matrix of the protein collagen (Whitney & Rolfes, 2002).As the crystal structure becomes denser, the strength and rigidity of the bones increase.
Bone is a dynamic tissue that is constantly undergoing osteoclastic bone resorption and osteoblastic bone formation (National Academy Press (NAP), 2000).In growing children, bone formation exceeds resorption.This process is balanced in healthy adults while formation lags behind resorption after menopause and with aging in men and women (NAP, 2000).The skeleton has an obvious structural role and also serves as a reservoir for calcium.
Deficiency is the most widely known issue associated with calcium intake.A chronic inadequate calcium intake through diet or supplementation is one factor in the etiology of several disorders.The disorder given the most attention is osteoporosis.This disease is multifaceted with many correlated risk factors such as smoking, glucocorticoid use, and physical inactivity (Swaminathan, 1999).However, many of these risk factors influence calcium uptake and utilization.Yet the role of calcium intake in the prevention of osteoporosis can be reduced to two basic principles: build the highest peak bone mass possible and protect the bone mass that has accumulated (Heaney, 1992).
Low calcium intakes coupled with high obligatory calcium losses from the body, deplete calcium reserves.In other words, low intakes cause subnormal bone mass and strength.This is one of the contributing factors of osteoporosis.The primary strategies for reducing osteoporosis risk are to optimize bone mass during growth and to reduce bone loss later in life.The aim of both of these strategies is to maximize calcium intakes (Weaver & Heaney, 1999;Krall & Dawson-Hughes, 1999).In the past few years, evidence has been given to indicate that dietary calcium intake is positively related to bone mineral density in children and adolescents.Research indicates the higher the calcium intake, the greater the peak bone mass (Valimaki et al., 1994;Jackman et al., 1997).Additionally, research has indicated a positive correlation between bone mass and calcium intake in premenopausal adult women (Welten, Kemper, Post, & Van Staveren, 1995).Overall, there is strong evidence that calcium intake influences bone mass in all age groups (Gennari, 2001).
Adequate dietary calcium is essential for building denser, stronger bones in the first three decades of life and for slowing the rates of bone loss in later years (Gerrior et al., 1998).The importance of calcium in the human diet is evident in many government publications.Healthy People 2010 is a document created to identify the current health status of Americans and the health and lifestyle areas needing improvement.One of the goals of the Healthy People 2010 document is to increase the quality of life among Americans.This is important when evaluating calcium intake and its role in the development of osteoporosis.As a result, there are three calcium-related objectives outlined in Healthy People 2010.One of these is specific to calcium intake while the other two are related to osteoporosis.
Objective 19-11.Increase the proportion of persons aged two years and older who meet dietary recommendations for calcium.
Calcium is essential for various mechanic and physiologic functions in the body.According to the Institute of Medicine (IOM) (1997), and the new Dietary Reference Intakes, the recommendations for adequate daily intakes (AIs) of calcium are 500 milligrams for children one to three years.The recommendations for other age groups include children aged four to eight years, 800 milligrams; for adolescents aged nine to 18 years, 1300 milligrams; for adults 19 to 50 years, 1000 milligrams; and for adults older than 51 years, 1200 milligrams (IOM, 1997).
According to baseline data collected during the National Health and Nutrition Examination Survey of 1988-1994 (NHANES III), only 46% of persons aged two years and older were at or above approximated mean calcium requirements based on calcium from foods, dietary supplements, and antacids.By the year 2010, the target for adequate calcium intake is 75% of the population agedtwo years and older.
Sources of dietary calcium are numerous.The most often recognized form is that of dairy products.Milk products contribute a significant proportion of calcium to the diets of women in western societies and women who do not consume milk products are unlikely to meet their calcium needs (Fleming & Heimbach, 1994;Horwath, Bovern, Campbell, Busby, & Scott, 1995).
Several barriers to milk product consumption have been identified in research.Low consumption has been associated with a dislike for milk products (Horwath et al., 1995), adverse reactions such as lactose intolerance (Arney & Pinnock, 1993), perceived adequate intake (Chapman, Chan, & Clark, 1995), and the perception of milk products increasing dietary fat and cholesterol intake (Horwath et al., 1995).However, research has also identified that consuming more meals at home or at sit-down restaurants is associated with moderate calcium intake (Lewis & Hollingworth, 1992).
The barrier to dairy product intake related to increasing fat and cholesterol is important to highlight.Many young women engage in selfimposed energy reduction diets.As a result, they are at risk for not obtaining an adequate calcium intake level.In a study conducted in Australia, researchers found that 68% of the 18year-old university students in the study had calcium consumption levels below 800mg.The majority of these women were on energyreduction diets and at risk of a dietary calcium intake deficit at a time when calcium intake should be enhanced (Portsmouth, Henderson, Graham, Price, Cole, & Allen, 1994).Dietary modification in the form of dairy products is important to consider.A three-year prospective study found that increasing dairy product consumption retards vertebral bone loss in premenopausal women (Baran et al., 1989).Dairy products are the richest source of dietary calcium and need to be consumed to assist women in achieving adequate calcium intake.In fact, research indicates the most people obtain 50% of their calcium from dairy products (Wardlaw & Weese, 1995).
Calcium supplementation is another effort utilized to meet the published Healthy People 2010 objectives.Interest and enthusiasm in the use of dietary supplements appears to be growing in the United States.According to NHANES III data collected in 1988-1994, approximately 40% of the population took dietary supplements during the month prior to the interview (Ervin, Wright, & Kennedy-Stephenson, 1999).Characteristics of these supplement users included women, Caucasian, increasing age, higher levels of income and education, and a self-reported health status of good or fair.Of these supplement users, approximately 46 percent took a combination vitamin/mineral product (Ervin et al., 1999).The data did not indicate the proportion of people consuming calcium supplements alone.
Calcium supplements are a nondietary, alternative source of calcium (Sutton, 2000).Individuals who are unable to get enough calcium in their regular diet generally take calcium supplements.
These nondietary alternatives often come as a salt.A calcium "salt" contains calcium along with another substance, such as carbonate or gluconate (Micromedex, 2001).As a result, some calcium salts have more elemental calcium than others.
Studies examining the effects of calcium supplementation in premenopausal women have shown conflicting results.Some have shown no significant effect on bone mineral density (BMD) in premenopausal women (Smith, Gilligan, Smith, Surpos, 1989) while still others have shown significant increases in BMD (Baran et al., 1989;Rico, Revilla, Villa, de Buergo, & Arribas, 1994;Lloyd et al., 1993;Johnston et al., 1992).Although there have been discrepancies in the findings, the overall conclusion is that calcium intake, even from supplements, has some effect in increasing peak bone mass during growth and bone maturation.
Although calcium supplements have a role in achieving adequate calcium intake, it is important to remember that the best source of calcium is food.A calcium supplement is to supplement what is obtained from food sources.Additionally, it is important to remember that bone health is not based on a single nutrient (Heaney, 1992).Diets low in calcium also tend to be low in other nutrients that are essential for normal bone development such as zinc, manganese, copper, ascorbic acid, protein, and vitamin D (Holbrook, & Barrett-Connor, 1991).
The need for knowledge is extensive.Calcium intake is vital for the current and future health of individuals.Information about how to obtain adequate calcium intakes is vital for several subpopulations such as vegetarians, lactose intolerant individuals, ethnic groups, and young women.Education about calcium needs to be expanded throughout one's life because research indicates that bone mass can increase through the third decade of life (Recker, Davies, Hinders, Heaney, Stegman, & Kimmel, 1992).
Vitamin D. Vitamin D plays an important role in calcium metabolism, calcium absorption, and bone health.
The National Osteoporosis Foundation describes the relationship between calcium absorption and vitamin D as being similar to that of a locked door and key (NOF, n.d.).Vitamin D is the key that unlocks the door and allows calcium to leave the intestine and enter the bloodstream.Vitamin D also works in the kidneys to help resorb calcium that otherwise would be excreted (NOF, n.d.).
Vitamin D is a fat soluble vitamin that has been associated with bone-related disorders for centuries.Vitamin D deficiency is associated with rickets in children, osteomalacia in adults, and secondary hyperparathyroidism (Combs, 1998).The daily requirement for this vitamin is met from the diet or from synthesis in the skin.From either of these sources, vitamin D is metabolized to 25-hydroxy vitamin D (25OHD) in the liver and is then converted to the active metabolite 1,25 dihydroxyvitamin D (1,25 [OH2 D) in the kidney (Combs, 1998;Swaminathan, 1999).
Recommendations for vitamin D intake are 200 international units (IU) for women under the age of 50, 400 IU for women 51-70 years of age, and 600 IU for women over 70 years of age (Willhite, 1998).The increase in vitamin D intake recommendations indicates that ageing affects vitamin D metabolism.The conversion of 25OHD to the active metabolite 1,25(OH)2D is reduced because of an age-related decline in renal function (Swaminathan, 1999).Additionally, as adults age, the ability to make vitamin D through the skin decreases (NOF, n.d.).
There are many sources of vitamin D available, however sunshine is the major source.One study indicated that total body sun exposure could provide the equivalent of 10,000 IU of vitamin D without a person experiencing toxicity.The study suggested that the RDA for vitamin D is a physiologic limit (Vieth, 1999).It may be important to note that sources of vitamin D may be managed differently in the body and therefore, the body can handle a greater amount of sunlight produced vitamin D than vitamin D from dietary sources.However, dietary sources of vitamin D are necessary to utilize by people who do not get adequate sun exposure due to latitude, season, work environment, etc. Dietary sources of vitamin D include vitamin D-fortified dairy products, egg yolks, saltwater fish, liver, and supplements (Dowd, 2001;NOF, n.d.;Willhite, 1998).
Vitamin D enhances calcium's ability to build and maintain bones.Several studies have indicated the importance of vitamin D as a partner with calcium in the prevention of osteoporotic fractures.Vitamin D in combination with calcium has been shown to increase bone density and decrease fracture rates (Dawson-Hughes, Harris, Krall, & Dallal, 1997;Chapuy et al., 1992).Additionally, vitamin D supplementation has demonstrated an increase in bone mineral density (Dawson-Hughes, Harris, Krall, Dallal, Falconer, & Green, 1995;Ooms, Roos, Bezemer, Van Der Vijgh, Bouter, & Lips, 1995).
Vitamin D has a vital role in the health of bone.The role of this fat soluble vitamin is important for the absorption and utilization of calcium which is the primary mineral in bone.Therefore, it is important for individuals to be knowledgeable about the interaction of this vitamin and mineral for total bone health.Young women run the risk of being deficient in vitamin D if they avoid the sun or fail to consume adequate dietary sources.In other words, if women avoid consuming dairy products, they not only lose out on the calcium content but also the possibility for vitamin D consumption.
Fluoride.Fluoride has effects on the matrix of bone itself and on osteoblast function.Although fluoride has been used for dental carry prevention, it is approved for the treatment of osteoporosis in many countries around the world (Dowd, 2001;Sowers, Wallace, & Lemke, 1986).
Fluoride is capable of stimulating osteoblasts to increase bone formation.
Although the new bone has increased crystallinity, making it more resistant to resorption, the bone formed is abnormal and mechanical strength is compromised (Willhite, 1998).A clinical trial study demonstrated this compromise of mechanical strength by utilizing immediate-release fluoride.The study showed gains in spinal bone mineral density but spinal fracture rate was not decreased.Additionally, fracture risk of the hip and appendicular skeleton actually increased (Pak, Sakhaee, Rubin, Zerwekh, 1997).The authors concluded that fluoride may increase trabecular bone density at the expense of cortical bone even with adequate calcium supplementation (Pak et al., 1997).
Additional studies have been conducted using sustained-release fluoride.A study conducted by Pak and colleagues (1995) looked at osteoporotic women consuming sustainedrelease sodium fluoride and calcium citrate for four years.The study found an increase in spinal BMD by almost 5% and a decrease in vertebral fracture rate when compared to calcium intake alone (Pak, Sakhaee, Adams-Huet, Piziak, Peterson, Poindexter, 1995).
Fluoride is a compound that aids in the development of a strong bone structure.However, studies indicate that fluoride alone can actually reduce bone strength.The research further indicates that in combination with other minerals and vitamins such as calcium and vitamin D, fluoride has its greatest benefits for bone health.
Other positive dietary factors.Several other vitamins and minerals play a key role in the health of bone and therefore the prevention of osteoporosis.It is important to discuss these compounds because it supports the fact that bone formation and health is a complex and multifaceted process.It is important for one to have good overall health and to achieve such a status through adequate dietary intake.Simply taking supplements of various vitamins and minerals does not allow for adequate or appropriate nutritional balance to assist in adequate bone health.
Vitamin K has been suggested to play a specific role in osteoporosis.Vitamin K is involved in the synthesis of various proteins in the body including three in bone tissue: osteocalcin, matrix gla protein (MGP) and protein S (Dowd, 2001;Swaminathan, 1999).Vitamin Kdependent proteins contain gamma carboxyglutamic acid residues, and vitamin K is required for the carboxylation reaction of glutamic acid (Combs, 1998;Swaminathan, 1999).A lack of vitamin K will lead to a reduction in carboxylation of vitamin Kdependent proteins.
Vitamin K is also known for its role in blood clotting.
For those who are taking anticoagulants, caution needs to be exercised when taking supplemental Vitamin K because it could affect clotting.A physician should be consulted prior to taking supplemental vitamin K.The major dietary source of vitamin K is green vegetables (Combs, 1998).Although vitamin K deficiency is rare in most societies and its role in bone health is not completely clear, vitamin K cannot be ignored as a compound of importance in bone health.
Phosphorus is another compound that is important for bone health.Phosphorus is widely available in the diet and is part of the crystal structure of bones.Almost 85% of the body's phosphorus is present in crystalline form in bone as hydroxyapatite (Gennari, 2001).Since phosphorus is so abundant in the diet, a nutritional deficit is generally not a concern.However, among those with malnutrition or low dietary calcium intakes, phosphorus may bring harm by causing an increased calcium excretion (Dowd, 2001).
Phosphorus deficiency may be more important than currently recognized.Multiple nutritional deficiencies have been implicated in the concentrated number of fractures of the upper femur (Combs, 1998).With approximately 20% of elderly individuals in industrialized nations ingesting less than 60% of the RDA for phosphorus or protein, a relative hypophosphatemia might result (Heaney, 2000).This inadequate phosphorus intake could slow bone repair of osteoporotic fractures (Dowd, 2001).When searching for a good source of phosphorus, it is important to know that milk provides calcium and phosphorus in an optimal ratio for building bone tissue (Dowd, 2001).
The final mineral to be discussed in relation to bone health is magnesium.Approximately 50-60% of the magnesium in the body is in bone (Gennari, 2001).A well-balanced diet including whole grains, legumes, green leafy vegetables, and nuts contains a healthy amount of magnesium (Dowd, 2001).Although many people do not consume an adequate amount of calcium in their diet, magnesium deficiency is relatively rare (Combs, 1998;Dowd, 2001).
In experimental magnesium deficiency studies, osteoblastic and osteoclastic activity is decreased and there is a cessation of bone growth and osteopenia develops (Robbins & New, 1997).
Additionally, the magnesium content of trabecular bone in osteoporotic subjects is significantly lower than other subjects and magnesium intake has been reported to be lower in osteoporotic subjects (Combs, 1998;Schwartz & Reddi, 1979).
In addition to its content in bone, magnesium plays a role in the formation of 1,25 dihydroxyvitamin D. A magnesium-dependent hydroxylase enzyme is involved in the formation of 1,25 dihydroxyvitamin D. Therefore, if a magnesium deficiency did occur as in those with serious illness, alcoholism, prolonged vomiting and diarrhea, or intestinal malabsorption, this could possibly adversely affect calcium absorption and therefore bone strength (Combs, 1998;Dowd, 2001;Risco, Traba, de la Piedra, 1995).
Dietary Factors that Negatively Effect Bone Health
Protein.Protein-energy malnutrition during childhood or adolescence may retard growth and reduce body strength and peak bone mass (Toss, 1992).However, it has also been suggested that a high-protein diet may increase the risk of osteoporosis as a result of increases in urinary calcium excretion (Lau & Woo, 1998).The mechanism of increased urinary calcium excretion is thought to be a result of glomerular filtration rate (GFR) and the acid load from the sulphur containing amino acids methionine and cysteine (Marcus, 1982).This increase in endogenous acid production mobilizes calcium from the skeleton to form salts to neutralize the acidity (Krieger, Sessler, & Bushinsky, 1992).Concerns about high-protein diets arose when studies showed increases in urinary calcium excretion (Heaney, 1993).Although there is little evidence from clinical trials to support the observation that high dietary protein intake causes increased bone loss, result of observational epidemiologic studies do support this view.Results from two ecologic studies indicate that the incidence of hip fracture is inversely related to the per capita consumption of protein (Abelow, Holford, & Insogna, 1992;Hegsted, 1986).While it is difficult to interpret these results because of the numerable confounding effects of other dietary components, these studies still lend concern to the issue of protein intake and osteoporosis development.
The average protein intake in many industrialized countries, including the U.S., is at least 50% above recommended levels (Krall & Dawson-Hughes, 1999).An average increase in dietary protein of 1 gram results in the loss of an additional 1 milligram of calcium in the urine (NAP, 2000;Krall & Dawson-Hughes, 1999).Such calcium losses could be important for individuals with a low usual calcium intake or impaired calcium absorption.
While protein has been shown to increase calcium excretion, its roles in calcium absorption and retention are controversial.A twenty-year study conducted by Heaney (2000), on Catholic nuns, concluded that protein intakes, within the current U.S. intake levels, do not affect calcium absorption.The research related to protein's influence on osteoporosis is still controversial, but two things are known for sure: a) protein is necessary for bone and muscle health, and b) high levels of protein intake increase urinary calcium excretion (Powers et al., 1999).
Sodium.Sodium and calcium excretion are linked in the proximal renal tubule.Sodium causes an increase in renal calcium excretion (Krall & Dawson-Hughes, 1999).For each 500milligram increment in sodium excretion or intake, there is approximately a 10-milligram increase in the amount of calcium lost in urine (NAP, 2000).
A study conducted by Devine and colleagues (1995) on postmenopausal women, demonstrated the impact of sodium intake on the rate of bone loss at the hip.The study identified an increasingly negative change in hip bone density with higher urinary sodium levels and with increasing sodium intakes between 1 and 6 grams per day.The study further suggested that halving the current sodium intake of 121 mmol/d (or 2700 mg/day) would be equivalent to increasing dietary calcium by 22 mmol/d (or 890 mg/day) (Devine, Criddle, Dick, Kerr, & Prince, 1995).However, a study of pubertal females found no association between bone mineral density and urinary sodium excretion (Matkovic et al., 1995).
Studies correlating high sodium intake with a decrease in bone mineral density raise concern due to the high salt content of processed foods and the quantity of these foods that Americans consume.Reassuring information from a study by Massey and Whiting (1996) was that calcium intakes between 1000 milligrams and 2000 milligrams minimized bone loss in the hip area associated with diets high in sodium.These calcium levels are within the new Dietary Reference Intakes (DRIs) recommendations and do not exceed the Tolerable Upper Limit values (ULs).
Caffeine.The consumption of caffeine and its relationship to bone health is a controversial topic.Caffeine is the most widely consumed psychoactive substance in the world with coffee supplying greater than 80% of the caffeine consumed by adults in the United States (Barone & Grice, 1994).
Numerous studies have reported on caffeine as a possible risk factor for bone loss in adult women.The results however have been contradictory.
Several studies have reported no association between caffeine intake and fracture frequency or changes in bone density (Johansson, Mellstrom, Lerner, & Osterber, 1992;Lloyd, Rollings, Eggli, Kieselhorst, & Chinchilli, 1997;McCullock, Bailey, Houston, & Dodd, 1990;Tavani, Negri, & LaVecchia, 1995).Others, however, have reported small but significant increases in either fracture frequency or bone loss.The most notable of these studies is the Framingham study that reported a 53% greater incidence of hip fracture in those who consumed more than two cups of coffee or four cups of tea after controlling for weight, sex, age, estrogen use, smoking and alcohol (Kiel, Felson, Hannan, Anderson, & Wilson, 1990).Another noteworthy study was the Nurses Health Study of 84,000 women who were followed for 6 years.The study found that women with the highest intake of coffee and the highest intake of caffeine (800 mg or more daily) had three times the rate of hip fractures as the no coffee/no caffeine group (Hernandez-Avila, Colditz, Stampfer, Rosner, Speizer, & Willett, 1991).
Caffeine ingestion causes a short-term (within 1-3 hours) increase in urinary calcium loss, but studies have failed to document sustained effects of caffeine on urinary or fecal calcium excretion (Krall & Dawson-Hughes, 1999).Among people who have low calcium intakes the effect may be of great importance as the body fails to adequately compensate for the additional calcium loss.
The effects of caffeine on bone mass in young women have also been a controversial topic.A study conducted by Packard and Recker (1996) indicated that a moderate caffeine intake (one cup of coffee per day or 103 mg) appeared to be a safe level with respect to bone health.However, another study conducted by Conlisk and Galuska (2000) did find that caffeine consumption decreased bone mineral density at various skeletal sites.The study indicated that for every 100 mg of caffeine consumed, femoral neck BMD decreased 0.0069 g/cm2, and lumbar spine decreased 0.0119 g/cm2 (Conlisk & Galuska, 2000).
Although there was no significant difference between those who consumed low levels of calcium and those who consumed high levels of calcium in this study, it stands to reason that such decreases over many years can increase a woman's risk of osteoporosis significantly.
Although sustained effects of caffeine on calcium excretion have not been observed in these studies, there is another aspect of this consumption that needs consideration; beverage replacement.On any given day, half of all Americans drink carbonated soft drinks according to data collected for 1994-1996 (Gerrior et al., 1998).The intake of these beverages has increased drastically among teenagers, younger adults, and women drinking low-calorie beverages, since 1970.Annual food supply data show that per capita consumption of regular carbonated soft drinks increased from 22 gallons in 1970 to 40 gallons in 1994 and to 41 gallons in 1997 (Gerrior et al., 1998).Whether caffeine affects over all calcium excretion is important, but the fact that carbonated beverages are replacing calcium-rich drinks needs consideration.
Long-term effects of frequent caffeinated beverage intake must be conducted before a consensus about the true relationship between caffeine and calcium can be decided upon.If caffeine increases calcium excretion for 1-3 hours, it stands to reason that a person who consumes 3-4 caffeinated beverages a day will have a significant loss of calcium in their urine.Dietary Factors with Questionable Effects on Bone Health Soy and isoflavones.Phytoestrogens such as isoflavones, which are found in many soy foods and supplements, have a chemical structure that causes them to act in the body like the estrogenic hormone estradiol (Anderson, 1999;Messina, 1995;Messina, 1999).The lower rate of hip fracture among Japanese women in comparison to US women is often cited as providing support for the protective effect of isoflavones (Ross et al., 1991).However, this argument is without merit (Messina, 1999).
However, researchers at Creighton University found no effect on spine and femur bone mineral density in early postmenopausal women (Heaney, Dowell, Rafferty, & Bierman, 2000).Although soy has nutritional value, its effects on bone seem to be small and of tentative clinical importance.
Several studies have been conducted to compare the antiresporptive effect of estrogen replacement in postmenopausal women versus that of soy isoflavones (Heaney et al., 2000;Setchell, 2000).However, in both of these studies, there was not a greater bioavailability of soy isoflavones.In fact, calcium in fortified soy milk was not absorbed as efficiently as calcium from non-soy milk.
Although several animal studies indicate a potential benefit in the inhibition of bone resorption as a result of soy isoflavones (Anderson, Ambrose, & Garner, 1995;Brandii, 1992) there have not been consistent results when applied in human models.There are several areas left unanswered that are necessary to delineate the effects of soy fully or to allow making any recommendations (Dowd, 2001).Information regarding efficacy, safety, which isoflavones have the greatest effect, and amounts of isoflavones necessary for measurable benefit, are necessary to provide accurate recommendations.
Conditions That Effect Dietary Intake
Lactose Intolerance.About 25% of adults in the United States have lactose intolerance and develop symptoms of diarrhea and bloating after ingestion of a large dose of lactose (NAP, 2000).This condition is caused by a deficiency of an enzyme that breaks down milk sugar in the intestine called lactase.Many people believe they are lactose intolerant, but they do not know they do not have to be.Lactose intolerance is especially common in African Americans, Hispanics, Native Americans, and Asian Americans (Gerrior et al., 1998).Those who demonstrate lactose intolerance often avoid dairy products entirely.However, such foods as hard cheese (Swiss, cheddar, American), yogurt, and lactose-reduced milk have lower lactose levels and therefore can be consumed by sufferers of lactose intolerance (Dowd, 2001).In addition, studies indicate that many lactoseintolerant people can tolerate smaller doses of lactose such as the amount present in an 8-ounce glass of milk without the negative side effects (Levin, 1999;NAP, 2000).
Lactose-intolerant individuals absorb calcium normally from milk, but they are at an increased risk of calcium deficiency because of their avoidance of milk and other calcium rich dairy products (NAP, 2000).
Although lactose intolerance may influence intake, there is no evidence to suggest that it influences the calcium requirements (NAP, 2000).Therefore, to aid in calcium intake, lactose-free dairy products are available.There are also good nondairy sources of calcium such as canned salmon with the bones, fortified cereals, and calcium-fortified orange juice.
Nutritional Summary
Nutrition plays a significant role in the cause and prevention of osteoporosis.This disease can be prevented by maintaining an optimal calcium intake, maintaining an adequate vitamin D level, avoiding a high salt intake, and avoiding extremely high animal protein intakes.Nutrition through food consumption is the most adequate method to achieve good bone strength.The bone matrix is complex and simply taking a supplement or several supplements to possibly compensate for an inadequate diet, could cause a greater imbalance and still lead to a weak bone mass.
Nutritional factors alone do not completely explain the prevention puzzle of osteoporosis.
Additional factors such as physical activity, heredity, and ethnicity play a role in understanding the etiology and prevention of the disease.
Lifestyle Factors and Osteoporosis
Physical Activity Bone is living tissue that responds to exercise by becoming stronger and denser.There are two types of exercises that are important for building and maintaining bone strength and density: weight-bearing and resistance exercises (NOF, 2000).The specific characteristics of physical activity that are most important for influencing bone are not completely understood, but research indicates that high mechanical loads may be more osteotropic than low-intensity loads (Taaffe, Robinson, Snow, & Marcus, 1997).Additionally, the number of repetitions has been shown to make modest effects of bone mass (Carter, Fyrie, & Whalen, 1987;Rubin & Lanyon, 1985).
However, conclusions about the effects of exercise in the prevention of osteoporosis are still vague.Research seems to be inconclusive about the types of recommendations that should be made.
The majority of osteoporosis researchers would probably agree that recommending physical activity is important for improving balance, muscle tone, flexibility, strength, and coordination; all aspects that could prevent falls and low-trauma fractures (American College of Sports Medicine [ACSM], 1995).Yet questions remain as to what types of activities, for how long, how often, at what intensity, and should the activities be site specific?Much research exists that indicates simply physical activity in itself can help prevent osteoporosis.
As with other areas of osteoporosis research, a lot of information is available for older populations.The few studies that exist for younger, premenopausal women lend support to the information that has been acquired while studying older populations.In a study conducted by Turner and colleagues, women who participated in the NHANES III study over the age of 50 were reviewed.The study identified physical activity as a greater predictor for fracture than heredity, smoking status, alcohol use, and dairy product intake (Turner, Leaver-Dunn, DiBrezzo, & Fort, 1998).The study revealed that inactive women were 84% more likely to suffer a fracture than females who were active 2 or more times per week.
Another study conducted by Mitchell and colleagues followed 2,567 women for an average of 11 years.This study reviewed the risk of developing osteoporosis based on cardiorespiratory fitness.The study revealed that the more fit a woman was, the less likely she was to develop physician diagnosed osteoporosis during the time frame of the study.After the researchers adjusted for age, body mass index, high blood pressure, cigarette smoking, alcohol consumption, and diabetes, they found that low fit women were 1.8 times more likely to develop osteoporosis than moderate or high fit women (Mitchell, Wei, Gibbons, & Blair, 1999).
Studies reviewing the types of activities women participate in and their risk for osteoporosis have also been conducted.
One study recently published by Turner and colleagues (2002) indicated that yard work and weight training were strong and independent predictors for positive bone density.Again, data from the Third National Health and Nutrition Examination Survey (NHANES III) were used to study the relationship between exercise mode, frequency, and bone health (Turner, Bass, Ting, & Brown, 2002).Other activities were also identified as moderate predictors for positive bone density such as bicycling, aerobics, walking, and dancing (Turner et al., 2002).
Most of the data related to young adult women between the ages of 18-35 comes from crosssectional studies comparing bone mineral density (BMD) of female athletes to that of a sedentary group.Direct measurements of bone mass have shown a positive correlation between spinal BMD and reported leisure time activity in healthy young women (Kanders, Dempster, & Lindsay, 1988).In this study, calcium and physical activity were independent determinants of BMD.
There have been few prospective studies of an exercise effect on bone mass in this age group.
In a study conducted by Snow-Harter and colleagues (1992), jogging or aerobic exercise for 20-30 minutes three times a week increased lumbar spinal bone mineral density by 1 percent in premenopausal women.Additionally, a study conducted by Bassey and Ramsdale (1994) found a significant increase of 3.4% in trochanteric bone density in a group of highimpact exercisers when compared to low-impact exercisers.Although these increases may seem to be small, any increase in bone mineral density that can delay the onset of osteoporosis or the complications of low bone density are important to consider and should not be discarded as being without merit.
Studies of athletes show that the BMD of loaded bones can be more than 30% higher in most studies and between 5% and 20% higher in most sites than that of unloaded bone or of the same bones in non-athletic control subjects (Vuori, 1996).However, for young women who are high-performance athletes, physical activity may increase their risk of osteoporosis.Intensive training that results in an extreme loss of body fat, disordered eating and estrogen deficiency can lead to bone loss (Vuori, 1996).The combination of osteoporosis, menstrual irregularities, and disordered eating is known as the female athlete triad (Sabatini, 2001).Many young female athletes experience a cessation in their menstrual cycles called amenorrhea.Several studies have reported low bone densities among these women (De Cree, Vermeulen, & Ostyn, 1991;Licata, 1992;Fabbri et al., 1991).Low bone densities result in a greater risk for stress fractures and other more devastating fractures of the hip and spine (Bass, Turner, & Hunt, 2001).Furthermore, research has indicated that premature osteoporosis occurring in female athletes may be irreversible even with calcium supplementation, resumption of menses, or estrogen replacement therapy (Drinkwater, Bruemner, & Chestnut, 1990).
The extent of exercise's influence on osteoporosis and its ability to induce bone density increases is still uncertain.Results from these studies vary according to age, hormonal status, nutrition, and exercise prescription.Regardless of the uncertainty from research results, expert panels such as the American College of Sports Medicine, recommend weightbearing activity and activities that improve strength, flexibility, and coordination to prevent osteoporosis and falls respectively (ACSM, 1995).
Alcohol Intake
Studies have indicated that alcohol suppresses bone formation.Since women are more prone to osteoporosis than men, the effects of alcohol may have a greater effect on their bones (Laitinen, Karkkainen, Lalla, Lamber-Allardt, Tunninen, Tahtela, et al., 1993).The evidence for the adverse effects of alcohol on bone mineral density (BMD) comes primarily from case control studies.However, in a 14-year longitudinal study, the rate of bone loss in men who drank regularly was faster than in controls (Slemenda, Christina, Reed, Reister, Williams, & Johnston, 1992).
Similarly, BMD in premenopausal women who drank regularly was also found to be lower than in matched controls (Arden, 1997).
Alcohol is capable of increasing one's risk of osteoporosis or fracture through several different methods.First, alcohol abuse may bring about hypoestrogenism with consequent menstrual irregularities and amenorrhea (Mello, Mendelson, & Teoh, 1989;Van Thiel & Gavaler, 1990).These irregularities can depress osteoblastic activity (Lappe, 1994).This can lead to an imbalance between the resorptive osteoclastic activity and the formative osteoblastic activity progressing to decreased bone mineral density (Thomas, 1997).
Alcohol can also lead to interference with proper nutrition, especially calcium and vitamin D intake.Individuals who consume moderate to excessive amounts of alcohol often have an imbalanced diet with a decreased consumption of calcium (Wardlaw & Weese, 1995).Additionally, research has revealed that alcoholics have a reduced ability to produce 1,25 dihydroxyvitamin D in the renal tubules (Wardlaw & Weese, 1995).By consuming a calcium deficient diet and not having an adequate capacity to produce vitamin D, which is necessary for calcium absorption, the consumption of alcohol can lead to an inability of the bone to reach its maximum strength.
A final way that alcohol can increase one's risk of fractures is due to balance difficulties experienced during inebriation which can increase the risk for falls.If a woman with poor bone health falls, she is more likely to break a bone than a women with good bone health (Thomas, 1997).All in all, alcohol may not have a direct influence on bone health, but through indirect means, a lifestyle that involves alcohol intakes of more than one drink per day can increase one's risk of osteoporosis or osteoporotic fracture (Laitinen & Välimäki, 1993).
Cigarette Smoking
Smoking puts women at risk for osteoporosis because smoking decreases serum estrogen (Thomas, 1997).A loss of estrogen leads to a decreased osteoblastic action progressing to an imbalance between resorption and formation.Estrogen also plays a role in the absorption of calcium, an essential nutrient in the formation of strong bones (Wardlaw & Weese, 1995).Additionally, smokers tend to have leaner body masses perhaps because of the interference of smoking with eating (Mazess & Barden, 1991).This combination of low estrogen and low body weight can lead to an increase in risk for osteoporosis.
Smoking is recognized as a risk factor for vertebral, forearm, and hip fractures (Hollenbach, Barrett-Connor, Edelstein, & Holbrook, 1993).Previous research indicates that women, who smoke throughout adulthood will, by the time of menopause, have an average deficit of 5 to 10 percent in bone density (Hopper & Seeman, 1994).Additionally this same study found that for every 10 pack-years of smoking, a 2% decrease in lumbar spine BMD, 0.9% decrease in femoral neck BMD, and 1.4% decrease in femoral shaft BMD were observed.Such percentages are sufficient to increase the risk of fracture.
As with other diseases associated with smoking, cessation can result in some positive effects.A study conducted by Cornuz and colleagues (1999) observed that smokers are at an increased risk of hip fracture and their risk rises with greater cigarette consumption.However, the risk declined among former smokers, but the greatest benefit was not observed until 10 years after cessation (Cornuz, Feskanich, Willett, & Colditz, 1999).The authors did identify that part of the benefit realized was associated with a difference in body weight.
The influence of both alcohol and smoking on bone health is complex.In reviewing studies regarding these lifestyle factors, it is important to realize the multiple potential confounding variables: age, heredity, consumption amounts, dietary intake, physical activity, etc.Although it is unlikely we will see a heavy smoker participating in the recommended levels of physical activity for general health, one cannot ignore the confounders present when reviewing the impact of smoking and alcohol intake on bone density.
Heredity and Osteoporosis
Osteoporosis is a multifaceted disease.One of the potential risk factors associated with disease development is family history.Lindsay and Dempster (1985) state that 75% of osteoporosis cases have a family history component.However, the degree to which this is due to genetics or to environment is debatable (Wardlaw, 1988).The most commonly cited studies supporting a genetic component are twin studies.
Additionally, other twin studies have shown a genetic effect but after the age of 25, the effect was no longer significant (Dequeker, Nijs, Verstraeten, Geusens, & Gevers, 1987).This might suggest that other factors such as the environment play a role after a certain age.
Research indicates that women who have a family history of osteoporosis generally have lower bone densities putting them at greater risk for osteoporosis in the future (Evans, Marel, Lancaster, Kos, Evans, Wong, 1988;Ulrich, Georgiou, Snow-Harter, & Gillis, 1996;Francois, Benmaler, Guaydier-Souquieres, Sabatier, & Marcelli, 1999;Diaz, O'Neill, & Silman, 1997).Several studies have been conducted on mother-daughter pairs to identify familial similarities in osteoporosis risk.A study by McKay and colleagues (1994) observed positive mother-daughter correlations ranging from 0.57 (p<0.05) for the proximal femur to 0.38 (NS) for the third lumbar vertebrae.Additionally, a study done by Krall & Dawson-Hughes (1993) observed estimates of heritability at peripheral sites such as the radium and os calcis.These studies indicate that women who have a family history of osteoporosis should be encouraged to have bone density scans and to participate in preventive behaviors.
Although women with a family history of osteoporosis should be advised to care for themselves and to practice osteoporosis prevention behaviors, women without a family history should not consider themselves safe.Measures of family resemblance often do not clearly differentiate the relative contributions of shared environmental factors and genetic factors (Tudor-Locke & McColl, 2000).Genetic interactions may be as complex as inadequate vitamin D receptors to as simple as children emulating their parents' cooking and exercise behaviors.Additionally, in trying to assess one's risk of osteoporosis, simply considering the maternal side of the family tree may not be adequate.
An offspring is a combination of genes from two parents and a combination of behaviors from two parents.As a result, a person may be susceptible to the development of osteoporosis independent of family history.Consequently, every woman should be aware of her familial history related to osteoporosis, but she should also assume personal responsibility in her prevention of developing this disease.
Ethnicity and Osteoporosis
The incidence of hip fracture and osteoporosis varies widely among ethnic groups.The highest rates have been reported among whites of northern European ancestry (Luckey, Wallenstein, Lapinski, & Meier, 1996).
Research indicates that women of Asian descent are also more likely to develop osteoporosis yet are less likely to experience osteoporotic hip fracture (Hirota, Nara, Ohguri, Manago, Hirota, 1992;Mackelvie, McKay, Khan, & Crocker, 2001).Studies report differences in body size, diet, and physical activity between Caucasian and Asian girls (Mackelvie et al, 2001).Typically, Asian girls consume about 480mg of calcium per day and participate in far fewer weight bearing activities than Caucasian girls.However, Asian girls also consume more plant based foods and plant sources of protein which may serve as a protective factor (Mackelvie et al., 2001).
Other studies have substantiated the reduced risk of hip fracture among Asian women as compared to Western Caucasian women.Lifeand work-style differences between these cultures (i.e.sleeping on hard floors vs. beds, sitting on hard floors vs. couches, walking vs. driving) may be responsible for part of this reduced risk (Fujita, 1994).This is further supported with the increase in hip fractures among native Japanese women with the progressive Westernization of the Japanese lifestyle (Fujita, 1994).
Although Asian women seem to have less risk for osteoporotic hip fracture, it may not be due to the adherence to osteoporosis prevention behaviors.Studies have shown that Asian women have similar osteoporosis risk factors as Caucasian women, such as low calcium intake and lack of physical activity (Lau, Suriwongpaisal, Lee, De, Festin, Saw, et al., 2001).Additional risk factors found to be influential in the development of osteoporosis among Asian women were the low intake of dairy product in childhood (due to lactose intolerance), frequency of dieting, and skipping meals; all similar to those of Caucasian women (Hirota et al., 1992).However, the lack of osteoporotic hip fractures may be due to a lower center of gravity due to the shorter posture shared by many Asian women (Tudor-Locke & McColl, 2000).
Compared to those of African American or Hispanic ethnicities, White (non-Hispanic) and Asian women are at greater risk for osteoporosis (Pun, Chan, Chung, & Wong, 1990).However, the mechanism for this difference is not known.Some research indicates that African American and Hispanic women achieve a higher peak bone density and lose bone density after menopause at a slower rate (Luckey et al., 1996).
Although research indicates that certain ethnicities are at a greater risk for osteoporosis than others, adherence to known osteoporosis prevention behaviors should be practiced.
Ethnicity is not merely one's skin color.Ethnicity involves traditions, practices, and environmental influences (Tudor-Locke & McColl, 2000).Overall, osteoporosis is a multifaceted disease in which ethnicity is just a small fraction of the puzzle.
Medical History and Osteoporosis
As research in the area of osteoporosis continues to expand, new factors for concern are revealed.Within one's medical history, various medications have been demonstrated as protective or risk factors.Additionally, medical conditions such as eating disorders have also been shown to be potential risk factors for osteoporosis development.
Medications and Their Effects on Osteoporosis Corticosteroids/
Glucocorticoids.The association between osteoporosis and corticosteroids or glucocorticoids was made shortly after the first use of these drugs in humans in the 1950s (Eastell, 1995).Corticosteroid treatments are so strongly related to the development of osteoporosis that they negate other factors that may be protective (i.e.race) (Lucasey, 2001).Corticosteroids are used for a variety of conditions such as asthma, rheumatoid arthritis, lupus, and inflammatory bowel disease.Concern is high among health educators because the number of young people who report using steroid medications is increasing.
Glucocorticoids affect bone in many ways.They adversely affect bone formation, bone resorption, calcium entry into the body in the gut and calcium exit from the body in the renal tubule (Reid, 1997).Osteoblasts are a group of bone cells primarily affected by glucocorticoids.These compounds affect osteoblasts by decreasing their proliferation, matrix synthesis, and decreasing their life span.This effect is thought to be mediated in part by a reduction in the production of local growth factors such as insulin-like growth factor 1 (Manolagas & Weinstein, 1999).When osteoblasts activity is hindered or decreased, it stands to reason that bone density will be compromised and thereby increase one's risk of fracture.
Research has also revealed that steroid-treated patients experience a state of calcium malabsorption (Klein, Arnaud, & Gallagher, 1977;Nordin, Marshal, Francis, & Crilly, 1981).Additionally, these patients experience a state of hypercalciuria which has been reported to be double of that in non-steroid using controls (Reid & Ibberston, 1987).Various therapy methods have been proposed to address the calcium deficit and imbalance.Among these therapies is calcium supplementation of 2000mg per day.Unfortunately, doses as high as 2400mg per day have shown to have no protective effect against the damage done by corticosteroids (Eastell, 1995;Reid, 1997;Yosipovitch, Hoon, & Leok, 2001).
In addition to causing bone loss, glucocorticoids appear to lead to changes in the architectural integrity of bone (Daens, Peretx, de Maertelaer, Moris, & Bergmann, 1999).Since bone loss occurs rapidly in the first 6 to 12 months of steroid therapy with a 5% to 20% decrease in bone density, it has been estimated that between 30% and 50% of long-term corticosteroid users will experience fractures (Lukert & Raiax, 1990).Despite all the information about glucocorticoidinduced osteoporosis, few patients are provided with information about side effects and few are counseled on how to help prevent secondary osteoporosis.A study in England showed that only 6% of patients receiving glucocorticoid therapy received calcium supplementation (Peat, Healy, Reid, & Ralston, 1995) while another study out of England showed that only 14% of patients received prophylactic medication (Walsh, Wong, Pringle, & Tattersfield, 1996).A further study conducted by Buckley and colleagues (1999) found that 58% of postmenopausal women received osteoporosis preventive treatment while using corticosteroids while premenopausal women and men were less likely to be treated.Some researchers have argued that concern about steroid use should not be applied universally to all corticosteroids.There has been a lot of debate about the risks associated with oral glucocorticoids versus inhaled corticosteroids.Most of the data that exists correlates secondary osteoporosis and corticosteroid use based on the oral mode of administration.However, the few studies that have reviewed the effects of inhaled corticosteroids have identified osteoporosis risks as well.A study done Marystone and colleagues (1995) showed that oral steroid users had significant reductions in bone density when compared to non-steroid users.
However, inhaled steroid users in the study showed intermediate reductions in bone mass when compared to non-steroid users although these values were not significant.Another study performed by Wong and colleagues (2000) showed a negative relation between total cumulative dose of inhaled corticosteroid and bone mineral density in patients with asthma.
Regardless of the route of administration, women who are utilizing corticosteroids, for any number of medical conditions, should receive counseling about ways to prevent secondary osteoporosis.
Several behaviors should be adhered to and they are consistent with recommendations for non-steroid users in the prevention of osteoporosis.Women should receive primary prevention at the onset of corticosteroid therapy.This should include dietary advice to increase calcium and vitamin D while reducing sodium and caffeine intake.Women should also be counseled to participate in weight-bearing activities and to limit cigarettes and alcohol consumption.Additionally, women should receive a baseline bone density scan and be prescribed medications that have demonstrated a bone-protective role in the use of corticosteroids such as bisphosphanates (Eastell, 1995).
DMPA is a progestin-only contraceptive that contains no estrogen.In 1992, it was approved by the Federal Drug Administration (FDA) and since then has been utilized by many women who struggle with contraceptive adherence or have a difficult time remembering to use contraception (Kass-Wolfe, 2001).There are numerous side effects of DMPA including menstrual irregularities including amenorrhea, weight gain, headaches, bloating of the abdomen or breasts, mood changes, and reduced libido (Kaunitz, 1998).
Depot medroxyprogesterone acetate (DMPA) is an injectable progesterone contraceptive technique that is used by over 3.5 million women in over 90 countries of the world (Cundy, Evans, Roberts, Wattie, Ames, & Reid, 1991;Mark, 1994).It works by primarily inhibiting ovulation through the suppression of follicle-stimulating hormone (FSH) and luteinizing hormone (LH) levels and causing a decrease in estrogen levels (Kass-Wolff, 2001;Mark, 1994).As a result, ovulation stops and in many women, with continued use, they become amenorrheic (Cundy, et al., 1991).Since the recognition of estrogen deficiency as a cause of bone demineralization is widely accepted, there is growing concern about the potential development of osteoporosis and fractures with this contraceptive technique (Mark, 1994).
Contraceptive methods that decrease bone density in a population already deficient in calcium are a rising concern in women's health (Kass-Wolff, 2001).Studies indicate that women who use Depo-Provera, for a year or longer, have lower bone densities when compared to women who do not use this medication (Bahamondes, Perrotti, Castro, Faundes, Petta, & Bedone, 1999).Other studies have supported this finding.A study conducted by Cundy and colleagues (1991) demonstrated that users of DMPA for 5 or more years had a significant reduction in bone density as compared to nonusers: lumber spine -7.5% and femoral neck -6.6%.
The loss of bone density occurs rapidly when using Depo-Provera (Cundy & Reid, 1997).In a study by Cromer and colleagues (1996), the researchers compared subjects who used oral contraceptives, Norplant, and DMPA to noncontraceptive users.At the end of one year, the researchers found that bone density significantly decreased by 1.5% in the DMPA users when compared to the controls.Additionally, at the end of two years a significant bone density decrease of 3.1% was observed (Cromer, Blair, Mahan, Zibners, & Naumovski, 1996).Research indicates that the use of Depo-Provera should be approached with caution.Counseling about this contraceptive is vital for women to understand the potential detrimental side effects.Although the loss of bone mass when using DMPA is considered to be temporary (Cromer et al., 1996) and reversible after treatment is discontinued, it is important to realize that the window of bone mass accrual is small and should be maximized.As a result, if bone density is decreased for 2-3 years due to DMPA use, these women are potentially at risk for achieving a lower peak bone mass due to this setback and therefore are at a greater risk of osteoporosis and fracture.
Oral contraceptives.Many young women choose oral contraceptives for various reasons other than their primary purpose of contraception.Oral contraceptives may be chosen to reduce menstrual cramps, regulate menstrual cycles, or as partial treatment for endometriosis.Research related to the effect of oral contraceptives as a means of osteoporosis prevention is inconclusive.
Several research studies have been conducted that show positive relationships between oral contraceptive use and the prevention of bone loss.These studies have been conducted on women at various ages and at various life stages.One of the most referred to studies was conducted by Kleerekoper and colleagues (1991).This cross-sectional, retrospective study of over 2200 women found that women with a history of oral contraceptive use were significantly more likely to have high bone mineral density measurements than those who did not have a history of oral contraceptive use (Kleerekoper, Brienza, Schultz, & Johnson, 1991).Additionally, a significant increase in bone mineral density was found with greater than 10 years of oral contraceptive use.The limitations of this study was that is did not control for smoking or exercise (Kleerekoper et al., 1991).
A study conducted by Lindsay, Tohme, and Kanders (1986) compared the bone density at various stages of life in women who ever used oral contraceptives versus those who never used oral contraceptives (OCs).An increase in bone mass of about 1% per year was observed among premenopausal women who had used OCs versus nonusers while there was no difference observed between matched-controls in the postmenopausal group (Linsay et al., 1986).
While many research projects indicate that oral contraceptives have an independent positive effect on bone density (Cooper, Hannaford, Croft, & Kay, 1993;Recker et al., 1992;Van Winter & Bernard, 1998), other research indicates no effect on bone density.In a study by Mazess and Barden (1991) the effect of oral contraceptive (OC) use on bone density was evaluated in 300 women between the ages of 20 and 39.The authors controlled for calcium intake, exercise, and smoking.No association was found between OC use and bone density.
Additional studies have not found a positive association between OC use and bone density.A study by Collins and colleagues (1988), found no significant differences in the lumbar bone mineral content, central density, or bone mineral density measurements between OC users and non-users.These subjects were matched for age, weight, and height.Furthermore, this study did not find a significant difference between the groups and the duration of oral contraceptive use which was up to 84 months in duration (Collins, Thomas, Harding, Cook, Turner, & Collins, 1988).
The variation in the findings of these studies can be related to many factors including study design, mode of bone density measurement, oral contraception composition, and duration of use.Many of the studies that reported positive associations between bone mineral density and oral contraception use reviewed women who used OCs when they contained 50µg or more of estrogen.
Today's oral contraceptive is considered low-dose and contains between 20µg and 40µg of estrogen.As a result, the lower dose of estrogen may prevent OCs from having positive effects on bone mineral density.Although they may not assist by increasing bone mineral density, OCs may slow down bone loss suppressing bone resorption which is evident in lower urinary calcium excretions (Shargil, 1985).
Research regarding osteoporosis prevention and oral contraception use is controversial.Those who may improve their bone density with the use of OCs are women who are already hypoestrogenic, have irregular menstrual cycles, or women who have other conditions that interfere with their estrogen producing capabilities.Although some protection may be offered for these special populations, women should not fall under the misperception that they are completely protected from osteoporosis development due to their use of OCs.Protecting oneself from this multifaceted debilitating disease requires attention to multiple prevention strategies.
Eating Disorders and Their Effects on Osteoporosis
Disordered eating refers to the spectrum of abnormal and harmful eating patterns used in a misguided attempt to lose weight or maintain a lowered body weight (Beals, Brey, & Gonyou, 1999).Because lower bone mineral density is one potential physiological consequence of eating disorders, the risk for osteoporosis among women afflicted with these diseases is greater (Clark, 1997).Both anorexia nervosa and bulimia nervosa and their associated behaviors increase a woman's risk for the development of osteoporosis.
Anorexia nervosa is a chronic illness that affects 1% of adolescent females and is characterized by a fear of fatness, self-imposed semistarvation and weight loss.Additionally, the illness has a high morbidity and is eventually fatal in 10-15% of cases (Seeman, Szmukler, Formica, Tsalamandris, & Mestrovic, 1992).Common clinical features of anorexia nervosa are estrogen deficiency (accompanied by amenorrhea) and a significant reduction in body weight (Treasure & Serpell, 2001).Chronic anorexia nervosa is known to lead to osteopenia and osteoporosis in adults (Rigotti, Neer, Skates, Herzog, & Nussbaum, 1991;Seeman et al., 1992).Estrogen status is likely to be a major cause of osteopenia and osteoporosis in patients with anorexia nervosa (Treasure & Serpell, 2001).The occurrence of estrogen deficiency during the first three decades of life can increase the risk of osteoporosis by preventing the attainment of peak bone density and by causing accelerated bone loss (Seeman et al., 1992).
Other factors that may be related to the development of osteoporosis in women with anorexia nervosa are nutritional intake and physical activity.Due to the intense fear of being fat, many anorexics strictly limit their caloric intake.To limit their caloric intake they consume foods that are low calorie which oftentimes excludes foods rich in calcium such as dairy products (Treasure and Serpell, 2001).This extreme dietary limitation often limits their intake of other necessary nutritional components for bone health such as protein and vitamin D.
Besides severely limiting dietary intake, anorexics participate in excessive physical activity (Seeman et al., 1992).Although weightbearing exercise can help protect against osteoporosis, excessive exercising can be associated with leanness and amenorrhea.Studies have been conducted to evaluate the effect of amenorrhea in young women suffering from anorexia nervosa.A study by Davies, Hall, and Jacobs, (1990) showed that the mean bone density was 15% lower for women with amenorrhea than age-matched controls and was related to the duration of amenorrhea and the severity of estrogen deficiency.This loss in bone density is not regained upon recovery from anorexia nervosa.A study by Hartman and colleagues (2000) found that for women who had been clinically recovered from anorexia nervosa on average of 21 years, their bone mineral density did not fully return to normal.
The risk of osteoporosis occurring in other eating disorders has also been investigated.Studies reviewing women formally diagnosed with bulimia nervosa and other nonspecified eating disorders have been found to have significantly lower bone mineral densities than expected when compared to control data (Anderson, Woodward, & Lafrance, 1995).These results are not completely surprising because between 50%-60% of patients with a current diagnosis of bulimia nervosa have a previous history of anorexia nervosa (Fairburn & Hope, 1988).
While amenorrhea is a diagnostic criterion for anorexia, menstrual irregularities occur in only about half of patients with bulimia (Seidenfeld & Rickert, 2001).Although menstrual irregularities may not be enough to increase risk of osteoporosis in bulimic patients, the compensatory behaviors of self-inducing vomiting, abuse of laxatives and diuretics, abuse of diet pill, and caloric restriction can all affect bone health.Not consuming adequate calcium and other nutritional components necessary for bone health may influence the density and strength of bones later in life.However, at this time, no studies have identified an increased risk of osteoporosis fractures in previous or current bulimia nervosa patients (Seidenfeld & Rickert, 2001).
Summary
Osteoporosis is a multifaceted disease.Multiple factors have been implicated in the development of osteoporosis including dietary factors, heredity, ethnicity, lifestyle factors, medication use, and eating disorders.Although some of these factors are not alterable, the majority of them can be changed.Through the dissemination of osteoporosis prevention materials and information (see PowerPoint 1), health educators and other health professionals will increase knowledge about osteoporosis, increase osteoporosis prevention behaviors, and reframe the attitudes of women and men about their risk for developing this debilitating disease. | 2019-03-06T14:02:40.045Z | 2004-09-01T00:00:00.000 | {
"year": 2004,
"sha1": "d36378ef714d70c2f515f166af9e1581af2af650",
"oa_license": "CCBY",
"oa_url": "https://journals.calstate.edu/cjhp/article/download/877/755",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "1a13d4bbd502204fcbf47e28f607b8361c1de27a",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
256972840 | pes2o/s2orc | v3-fos-license | 3D DEM Simulations and Experiments on Spherical Impactor Penetrating into the Elongated Particles
In this study, a brass or glass spherical impactor vertically penetrating into a granular bed composed of mono-sized spherical or elongated particles was simulated with three-dimensional (3D) discrete element method (DEM). Good agreement of the particle masses in the cup before and after penetration can be found in the simulations and experiments. The effects of particle length (Lp), friction coefficient, and particle configuration on the penetration depth of the impactor, ejecta mass, and solid volume fraction describing the response of the granular bed are discussed. The penetration depth is negatively correlated with Lp as the corresponding solid volume fraction of the granular bed decreases. A smaller friction coefficient leads to a larger penetration depth of the impactor and more ejection of particles. When the impactor is penetrating the Lp = 10 mm elongated particles, the penetration depth is negatively correlated to the order parameter and solid volume fraction.
Introduction
Non-spherical particles are widely ubiquitous in nature and industry; for instance, micro-particles of various shapes have been applied in drug transportation and mixing. Particle shape has been shown to influence various processes including drug transport and crystal preparation [1][2][3][4]. When non-spherical elongated particles are packed, the internal microstructure of the granular bed has a complex and important influence on the macroscopic mechanical response of the material, which is different from that of spherical particles. Furthermore, the mechanical behavior of non-spherical particles affected by the impact will have a significant influence on transportation and industrial production. Therefore, the study of the impact into elongated particles is of great importance.
The finite element method (FEM) is an effective numerical method for continuum material, such as variational phase-field problems [5,6], quasistatic frictional contact problems [7], and elastic systems [8]. In addition, the discrete element method (DEM) is one of the most effective numerical methods for studying the dynamic properties of granular medium; it is a discontinuous numerical method first proposed by Cundall in 1971 [9]. Based on the theory of Newton's second law and the contact model between particles, the DEM can be applied to the particle dynamics and kinematics analysis by calculating the displacement and force of each particle in the granular system based on the explicit time step iteration.
Both 2D [10][11][12][13][14] and 3D [15][16][17] discrete element models have been used in numerical simulations of object impacts into target granular medium. Studies have shown that the response of the granular bed and impactor depends on many factors, such as the shape, size, angle, and penetration velocity of the impactor [10,[18][19][20][21], and the friction coefficient and porosity of the granular bed [11]. In addition, the energy dissipates drastically during the transient impact process, which has a complex relationship with the physical parameters or material properties of the granular medium [22]. Based on studies of spherical particles, the effect of elongated particles can be discussed comparatively.
Numerical models of regular non-spherical particles such as cylinders [23], spherocylinders [24,25], polyhedrons [26,27], super-ellipsoids [28,29], multi-super-ellipsoids [30,31], and arbitrarily shaped elements [32] have been developed gradually. The packing of elongated particles was discussed by changing various parameters, such as coefficient of friction and coefficient of restitution [33]. The behavior of elongated particles with varying lengths was explored by discharging them from a rectangular hopper [34]. Studies have shown that the length of the particle can be used to adjust the buffer capacity for the non-spherical particle system [35]. Besides, the effects of particle length and configuration of granular bed on impact have not been discussed, and the effect mechanisms of the elongated particle system have not been studied comprehensively. This study can provide a fundamental insight for the mechanical behavior of impact into non-spherical particles.
In this study, we carried out experiments on vertical impact into granular material and compared the ejecta masses and initial solid volume fractions of the granular bed between experiments and simulations, considering spherical and elongated particles as the objective granular medium. For the simulations based on DEM, we investigated the behaviors of the ejected particles, the particles in the cup, and the impactor. For the above three objects of study, we considered the effects of some key factors according to their different behavior characteristics, including the length of elongated particles, the friction coefficient, the shape of the granular cup, and the configuration of particles.
Numerical Model and Methodology
In this study, the translational and rotational movements of particles are governed by Newton's second law of motion. The equations are written as follows: in which m i , v i , I i , and ω i are the mass, translational velocity, moment of inertia, and rotational velocity of particle i, respectively, and F i and T i are external forces and torques exerting on it. The contact types of the sphero-cylindrical particle for the DEM simulation are classified into four groups according to the normal contact forces, as shown in Table 1. The normal contact force calculation and implemented contact detection in this study were proposed by Kidokoro et al. [25] and modified by Guo et al. [36]. The magnitudes and directions of normal and tangential contact forces are determined by contact position and the overlap of the two contacting particles. The tangential force model is the Mindlin model, which only takes into account the static process in this study.
Scenarios Models
Group Ⅰ F n = 4 in which δn is the overlap in normal direction, E * is equivalent Young's modulus, and R * is equivalent radius of two objects in contact, which are defined as in which α is determined by the shape of contact area and determined to be 0.974 here.
Scenarios Models
Group II Table 1. Contact force models used in the DEM simulations.
Scenarios Models
Group Ⅰ F n = 4 3 * * (3) in which δn is the overlap in normal direction, E * is equivalent Young's modulus, and R * is equivalent radius of two objects in contact, which are defined as in which α is determined by the shape of contact area and determined to be 0.974 here.
Group Ⅲ = ⎩ ⎨ in which θ is the angle between the two major axes of particles. = √ √ * * , when θ is equal to zero.
is the same as Equation (6).
in which k is a constant determined as 2.5, l is the length of contact area along the major axis, b is the width of contact area, = 2 * .
Tangential force model where Ft 0 and Ft are the tangential force vectors in the previous time step and the current time step, respectively. G * is governed by * = + , in which G1 and G2 are the shear moduli of the two objects in contact, ν1 and ν2 are the corresponding Poisson's ratios. is the effective radius of contact, = * , and dt represents the incremental tangential displacement in the present time step.
Note: The equations are summarized according to the proposed models [36].
For the normal and tangential damping force [36], F n d and F t d are determined by where c is equal to 1 if the normal contact force Fn is proportional to , and c is equal to if the normal contact force Fn is proportional to . m * is equivalent mass, m * = in which α is determined by the shape of contact area and determined to be 0.974 here.
Scenarios Models
Group Ⅰ F n = 4 in which δn is the overlap in normal direction, E * is equivalent Young's modulus, and R * is equivalent radius of two objects in contact, which are defined as in which α is determined by the shape of contact area and determined to be 0.974 here.
Group Ⅲ = ⎩ ⎨ in which θ is the angle between the two major axes of particles. = √ √ * * , when θ is equal to zero.
is the same as Equation (6).
in which k is a constant determined as 2.5, l is the length of contact area along the major axis, b is the width of contact area, = 2 * .
Tangential force model where Ft 0 and Ft are the tangential force vectors in the previous time step and the current time step, respectively. G * is governed by * = + , in which G1 and G2 are the shear moduli of the two objects in contact, ν1 and ν2 are the corresponding Poisson's ratios. is the effective radius of contact, = * , and dt represents the incremental tangential displacement in the present time step.
Note: The equations are summarized according to the proposed models [36].
For the normal and tangential damping force [36], F n d and F t d are determined by where c is equal to 1 if the normal contact force Fn is proportional to , and c is equal to if the normal contact force Fn is proportional to . m * is equivalent mass, m * = in which θ is the angle between the two major axes of particles.
n , when θ is equal to zero. F max n is the same as Equation (6).
Scenarios Models
Group Ⅰ F n = 4 in which δn is the overlap in normal direction, E * is equivalent Young's modulus, and R * is equivalent radius of two objects in contact, which are defined as in which α is determined by the shape of contact area and determined to be 0.974 here.
Group Ⅲ = ⎩ ⎨ in which θ is the angle between the two major axes of particles. = √ √ * * , when θ is equal to zero.
is the same as Equation (6).
in which k is a constant determined as 2.5, l is the length of contact area along the major axis, b is the width of contact area, = 2 * .
Tangential force model where Ft 0 and Ft are the tangential force vectors in the previous time step and the current time step, respectively. G * is governed by * = + , in which G1 and G2 are the shear moduli of the two objects in contact, ν1 and ν2 are the corresponding Poisson's ratios. is the effective radius of contact, = * , and dt represents the incremental tangential displacement in the present time step.
Note: The equations are summarized according to the proposed models [36].
For the normal and tangential damping force [36], F n d and F t d are determined by where c is equal to 1 if the normal contact force Fn is proportional to , and c is equal to if the normal contact force Fn is proportional to . m * is equivalent mass, m * = (6) in which k is a constant determined as 2.5, l is the length of contact area along the major axis, b is the width of contact area, b = √ 2R * δ n .
Tangential force model where F t 0 and F t are the tangential force vectors in the previous time step and the current time step, respectively. G * is governed by 1 G * = 2−ν 1 G 1 + 2−ν 2 G 2 , in which G 1 and G 2 are the shear moduli of the two objects in contact, ν 1 and ν 2 are the corresponding Poisson's ratios. a is the effective radius of contact, a = √ R * δ n , and v t c dt represents the incremental tangential displacement in the present time step.
Note: The equations are summarized according to the proposed models [36].
For the normal and tangential damping force [36], F d n and F d t are determined by where c is equal to 1 if the normal contact force F n is proportional to δ n , and c is equal to 5 6 if the normal contact force F n is proportional to δ n 3 2 . m * is equivalent mass, , and m 1 and m 2 are the masses of two contacting particles. v n and v t are the normal and tangential components of relative velocity, respectively. β = −lne √ π 2 +(lne) 2 is the contact damping coefficient, where e is the coefficient of restitution. S n is normal contact stiffness, and S t is tangential contact stiffness, given by S n = dF n dδ n and S t = 8G * a, respectively.
Experimental and Numerical Setup
The experimental apparatus used in this study is the same as what we used in the previous study [17]. Figure 1 shows the schematic of the experimental apparatus. The experiment was carried out in an airtight vacuum chamber (<200 mTorr), taking no account of the gas drag on the particles during the impact process. The impactor has a steel dot pasted on the top. Firstly, the impactor was fixed by an electromagnet at the central position of the crossbar at a height of H = 661 mm above the top of the granular bed. Then, the cylindrical particles were poured into the cup, and the surface of the granular bed was smoothed with a scraper. The total mass of the cup and the particles in it, m 0 , was weighed. The air was then pumped out until the pressure in the chamber was lower than 200 mTorr. Finally, the impactor was released from static state under gravity, and the vertical velocity of the impactor was calculated by V 0 = 2gH (where g is the acceleration of gravity). After all particles reached the stable state, the total mass of the cup and particles remaining in it, m, was weighed. The mass of ejected particles can be obtained by the equation ∆m = m 0 − m. The objective particles used in the experiments are the steel balls and steel cords cut into different lengths (L p = 4 mm and 6 mm), whose diameters are 2 mm. The impactors are brass and glass spheres, whose impact velocities are 3.60 m/s. The parameters of all the materials used in experiment are shown in Table 2.
pasted on the top. Firstly, the impactor was fixed by an electromagnet at the central position of the crossbar at a height of H = 661 mm above the top of the granular bed. Then, the cylindrical particles were poured into the cup, and the surface of the granular bed was smoothed with a scraper. The total mass of the cup and the particles in it, m0, was weighed. The air was then pumped out until the pressure in the chamber was lower than 200 mTorr. Finally, the impactor was released from static state under gravity, and the vertical velocity of the impactor was calculated by = 2g (where g is the acceleration of gravity). After all particles reached the stable state, the total mass of the cup and particles remaining in it, m, was weighed. The mass of ejected particles can be obtained by the equation Δm = m0 − m. The objective particles used in the experiments are the steel balls and steel cords cut into different lengths (Lp = 4 mm and 6 mm), whose diameters are 2 mm. The impactors are brass and glass spheres, whose impact velocities are 3.60 m/s. The parameters of all the materials used in experiment are shown in Table 2. The setup of DEM simulations is shown in Figure 2. The geometric and physical parameters of the cup, granular particles, and impactors are chosen from the experimental study. We mainly focus on the particles in the cup and the initial trajectory of ejected particles, and the cuboidal active simulation domain is big enough and will not affect the motion of ejected particles. The implementation of DEM simulation in this study was as follows: Firstly, the computing domain was divided into same-sized cuboid cells, and a certain number of particles were generated without contact and mapped into the cells. The particles were generated within the cylindrical domain with the same diameter as the cup and enough height. The generated particles then fell downward under gravity and packed in the cup until the cup was filled up with the particles. The translational and The setup of DEM simulations is shown in Figure 2. The geometric and physical parameters of the cup, granular particles, and impactors are chosen from the experimental study. We mainly focus on the particles in the cup and the initial trajectory of ejected particles, and the cuboidal active simulation domain is big enough and will not affect the motion of ejected particles. The implementation of DEM simulation in this study was as follows: Firstly, the computing domain was divided into same-sized cuboid cells, and a certain number of particles were generated without contact and mapped into the cells. The particles were generated within the cylindrical domain with the same diameter as the cup and enough height. The generated particles then fell downward under gravity and packed in the cup until the cup was filled up with the particles. The translational and rotational motions of particles were controlled by Newton's second law of motion. The normal and tangential contact forces were calculated by the overlap between two contacting objects according to various contact models. When all particles settled down, the impactor was then released with an initial vertical velocity of 3.60 m/s right above the particle bed and penetrated into the particle bed in the center. rotational motions of particles were controlled by Newton's second law of motion. The normal and tangential contact forces were calculated by the overlap between two contacting objects according to various contact models. When all particles settled down, the impactor was then released with an initial vertical velocity of 3.60 m/s right above the particle bed and penetrated into the particle bed in the center. In the DEM simulations, the particles are steel spheres and elongated steel particles of Lp = 4 mm, 6 mm, 8 mm, 10 mm with a hemisphere cap at each end. The diameters of the spheres and elongated particles are equal to 2 mm, which is the same as the cylindrical particles used in experiments. Brass and glass spherical impactors with the same diameter are used in both experiments and simulations. The diameter, the height, the density, the coefficient of friction, and the coefficient of restitution are measured experimentally. Young's modulus and Poisson's ratio are selected from a handbook [37]. The physical and geometric parameters of particles and impactors are summarized in Table 2. Contact detection of the elongated particles in this study is the same as the sphero-cylinder model in the literature [24]. The difference is that the elongated sphero-cylinder used here is rigid, and it was a flexible elastic model in the literature. In the DEM simulations, the particles are steel spheres and elongated steel particles of L p = 4 mm, 6 mm, 8 mm, 10 mm with a hemisphere cap at each end. The diameters of the spheres and elongated particles are equal to 2 mm, which is the same as the cylindrical particles used in experiments. Brass and glass spherical impactors with the same diameter are used in both experiments and simulations. The diameter, the height, the density, the coefficient of friction, and the coefficient of restitution are measured experimentally. Young's modulus and Poisson's ratio are selected from a handbook [37]. The physical and geometric parameters of particles and impactors are summarized in Table 2. Contact detection of the elongated particles in this study is the same as the sphero-cylinder model in the literature [24]. The difference is that the elongated sphero-cylinder used here is rigid, and it was a flexible elastic model in the literature.
The time step and boundary effect are two important parameters for DEM simulations; therefore, the sensitivities of these two parameters are studied before we determine the final values of them. The penetration depths of five different time steps are compared. The L p = 10 mm steel elongated particles and the glass impactor are chosen for simulation, and the parameters of materials are shown in Table 2. Table 2).
Results and Discussion
In this section, the key parameters on impact are discussed, including particle length, friction coefficient, cup shape, and particle configuration. In addition, three objects will be studied, including the particles out of the cup, the particles in the cup, and the impactor. Their fundamental parameters are analyzed, such as the ejecta mass, the penetration depth, the solid volume fraction, and the kinetic energy.
Effect of Particle Length
The effect of particle length on impact is studied in this section, and spherical particles and elongated particles of four different lengths with the same diameter are discussed. Two materials of brass and glass impactor with impact velocities of 3.60 m/s are used. The friction coefficients are the same as the base case and material parameters are shown in Table 2.
The kinetic energy of the ejected particles, the impactor, and the particles left in the cup are calculated. The kinetic energy includes rotational kinetic energy and translational kinetic energy. For particle i with mass mi and moment of inertia Ii, the translational Table 2).
Results and Discussion
In this section, the key parameters on impact are discussed, including particle length, friction coefficient, cup shape, and particle configuration. In addition, three objects will be studied, including the particles out of the cup, the particles in the cup, and the impactor. Their fundamental parameters are analyzed, such as the ejecta mass, the penetration depth, the solid volume fraction, and the kinetic energy.
Effect of Particle Length
The effect of particle length on impact is studied in this section, and spherical particles and elongated particles of four different lengths with the same diameter are discussed. Two materials of brass and glass impactor with impact velocities of 3.60 m/s are used. The friction coefficients are the same as the base case and material parameters are shown in Table 2.
The kinetic energy of the ejected particles, the impactor, and the particles left in the cup are calculated. The kinetic energy includes rotational kinetic energy and translational kinetic energy. For particle i with mass m i and moment of inertia I i , the translational velocity and rotational velocity at time t are defined as v i (t) and w i (t), respectively, and the total kinetic energy of N particles is given by The initial kinetic energy of the impactor is E k0 = 1 2 MV 0 2 , and the initial potential energy of the impactor is E p0 = MgH 0 (the position of zero gravitational potential energy is at the final stopping point of the impactor). The mass of impactor M is calculated by the density and size shown in Table 2. We take the percentage of the energy to the sum of E k0 and E p0 as the scaled energy. ticles is close to that of 4 mm, and their values are only 0.5%, indicating that the total mass of ejected particles is very small. Similarly, when the length of particles increases to 8 mm, the ejecta mass for the brass impactor is almost the same as that of 6 mm. By comparing the results of the two impactors, it can be concluded that when the mass of the elongated particle increases to a certain value, the transferred energy cannot support the ejection of more particles. Table 2). Figure 4a is the dynamic change of ejecta mass and Figure 4b is the comparison of the final ejecta masses between simulations and experiments. In Figure 4a, we set the time that the impactor touches the top of the particle bed as t = 0 ms. The end time of particle ejection for brass and glass impactors is about 90 ms. It is obvious that the ejecta masses of the brass impactor are larger than those of the glass impactor due to higher impactor energy. For the brass impactor, the ejecta masses of L p = 6 mm, 8 mm, and 10 mm elongated particles are closer and smaller than that of L p = 4 mm. For the glass impactor, the ejecta masses of L p = 4 mm, 6 mm, 8 mm, and 10 mm elongated particles are approximately equal and smaller than that of 2 mm spheres.
In Figure 4b, the agreement of solid volume fractions and ejecta masses between experiments and simulations can be found. The results show that the ejecta mass of spherical particles is much larger than that of elongated particles. In this simulation, the method of changing the aspect ratio is increasing the length of particles (keeping the diameter unchanged). In this way, the mass of a single particle also increases as a result of increasing particle length. When the length of particles increases to 6 mm, the impact energy of the glass impactor cannot make more elongated particles eject from the cup because of the large mass of a single particle. Therefore, the percentage of total ejecta mass of 6 mm particles is close to that of 4 mm, and their values are only 0.5%, indicating that the total mass of ejected particles is very small. Similarly, when the length of particles increases to 8 mm, the ejecta mass for the brass impactor is almost the same as that of 6 mm. By comparing the results of the two impactors, it can be concluded that when the mass of the elongated particle increases to a certain value, the transferred energy cannot support the ejection of more particles. Figure 5a,b show the kinetic energy of ejected particles for two impactors. The E e k reached its peak value at about 2 ms when the particles collided with the impactor and gained the maximum velocities. Some particles began to launch upwards and fly in the air along parabolic trajectories. The ejected particles were moving upward along the parabolic trajectory before 20 ms, and their velocities were hence decreasing, resulting in the decreasing kinetic energy. It is easy to find that the peak values of E e k s for spherical particles are much larger than those of elongated particles for both glass and brass impactors because spherical particles roll more easily than elongated ones. Comparatively, the E e k s of different elongated particles lengths are obviously smaller than those of spherical particles, and E e k s are slightly smaller as the particles become longer, whether for the glass or brass impactor. We can suppose that the kinetic energy transferred to the ejected particles is to support their movement and more energy will be needed to drive a particle moving if its mass becomes larger. In this study, we changed the aspect ratio of a particle by lengthening the particle and kept its diameter unchanged. Therefore, the E e k s of longer particles (with larger masses) are smaller.
air along parabolic trajectories. The ejected particles were moving upward along the parabolic trajectory before 20 ms, and their velocities were hence decreasing, resulting in the decreasing kinetic energy. It is easy to find that the peak values of s for spherical particles are much larger than those of elongated particles for both glass and brass impactors because spherical particles roll more easily than elongated ones. Comparatively, the s of different elongated particles lengths are obviously smaller than those of spherical particles, and s are slightly smaller as the particles become longer, whether for the glass or brass impactor. We can suppose that the kinetic energy transferred to the ejected particles is to support their movement and more energy will be needed to drive a particle moving if its mass becomes larger. In this study, we changed the aspect ratio of a particle by lengthening the particle and kept its diameter unchanged. Therefore, the s of longer particles (with larger masses) are smaller.
The Impactor
The responses of the impactor are investigated by DEM simulations here, including the penetration depth of the impactor (H0) and the energy of the impactor ( ). Figure 6 shows the normalized penetration depth of granular beds for spheres and elongated particles. Comparing the H0/D0 of brass and glass impactors in Figure 6a,b, it can be found that the final H0 is negatively correlated to the particle length of the granular bed. For the brass impactor, the largest H0 is about 1.5 D0 for the granular bed made of spheres, and the granular beds with Lp = 10 mm elongated particles have the smallest H0 of about 1.1 D0. The shape of the H0(t)/D0 function of the glass impactor is similar to that of the brass impactor, but its value is less than that of the brass impactor. In addition, the glass impactor stopped earlier than the brass impactor, indicating that the brass impactor with larger kinetic energy took more time to halt the penetration. The time of static state of the glass impactor is earlier than that of the brass impactor, and the static state is defined as the unchanged value of H0 of the impactor.
The Impactor
The responses of the impactor are investigated by DEM simulations here, including the penetration depth of the impactor (H 0 ) and the energy of the impactor (E i k ). Figure 6 shows the normalized penetration depth of granular beds for spheres and elongated particles. Comparing the H 0 /D 0 of brass and glass impactors in Figure 6a,b, it can be found that the final H 0 is negatively correlated to the particle length of the granular bed. For the brass impactor, the largest H 0 is about 1.5 D 0 for the granular bed made of spheres, and the granular beds with L p = 10 mm elongated particles have the smallest H 0 of about 1.1 D 0 . The shape of the H 0 (t)/D 0 function of the glass impactor is similar to that of the brass impactor, but its value is less than that of the brass impactor. In addition, the glass impactor stopped earlier than the brass impactor, indicating that the brass impactor with larger kinetic energy took more time to halt the penetration. The time of static state of the glass impactor is earlier than that of the brass impactor, and the static state is defined as the unchanged value of H 0 of the impactor. Because of the different penetration depths of the impactor, the initial potential energy is different, resulting in a slight difference of the Ek0 between the glass impactor and brass impactor. As shown in Figure 7a,b, the Ek0 for the glass and brass impactors at t = 0 ms are 98% and 96.5%, respectively, indicating that the initial total energy of the impactors dominated by Ek0 and its Ep0 is very small. It can also be found from Figure 7a,b that the Because of the different penetration depths of the impactor, the initial potential energy is different, resulting in a slight difference of the E k0 between the glass impactor and brass impactor. As shown in Figure 7a,b, the E k0 for the glass and brass impactors at t = 0 ms are 98% and 96.5%, respectively, indicating that the initial total energy of the impactors dominated by E k0 and its E p0 is very small. It can also be found from Figure 7a,b that the kinetic energy of the impactor E i k decreases rapidly from 0 ms to 5 ms. For both glass and brass impactors, the decrease of E i k slightly increases with the increasing particle length, indicating that the dissipation of E i k is faster with a larger particle length. Furthermore, we quantify the average E i k of five lengths, and it shows that the average E i k obeys an exponential-like dissipation. In addition, the brass impactor dissipates 95% of E i k in more than 10 ms, while the glass impactor dissipates 95% of E i k in less than 5 ms. The main reason is that the initial total energy of the glass impactor is smaller than that of the brass impactor. Because of the different penetration depths of the impactor, the initial potential energy is different, resulting in a slight difference of the Ek0 between the glass impactor and brass impactor. As shown in Figure 7a,b, the Ek0 for the glass and brass impactors at t = 0 ms are 98% and 96.5%, respectively, indicating that the initial total energy of the impactors dominated by Ek0 and its Ep0 is very small. It can also be found from Figure 7a,b that the kinetic energy of the impactor decreases rapidly from 0 ms to 5 ms. For both glass and brass impactors, the decrease of slightly increases with the increasing particle length, indicating that the dissipation of is faster with a larger particle length. Furthermore, we quantify the average of five lengths, and it shows that the average obeys an exponential-like dissipation. In addition, the brass impactor dissipates 95% of in more than 10 ms, while the glass impactor dissipates 95% of in less than 5 ms. The main reason is that the initial total energy of the glass impactor is smaller than that of the brass impactor.
The Particles in the Cup
To study the dynamic response of the particles in the cup, the average contact force of the granular bed (F c pp ), the solid volume fraction (φ p ), the granular temperature (T p ), and the kinetic energy are discussed.
The correlation between H 0 and L p can be explained by the average contact force between particles F c pp in the granular bed. F c pp is the average resultant force of the normal and tangential forces at all contact points, which can describe the strength to resist the penetration of the impactor, as shown in Figure 8. When the impactor penetrates the granular bed, the particles in the cup move randomly and intensively in the first 10 ms. During this period, the impactor and the particles in the granular bed will be in contact and separated from each other until most of the energy of the impactor is dissipated. In Figure 7, it is easy to find that the energy of the impactor sharply decreases in the first 10 ms. The contact forces show a zig-zagging change, which is a characteristic of the discontinuous medium after impact. The contact force is transmitted through the force chain. The particles are discrete, and the force chain structure in the granular system will change at every moment due to the frequent formation and breakage of the contact forces between particles, which leads to the zig-zagging change of the contact force. Furthermore, a longer particle can easily trigger stronger contact force between particles and has a stronger resistance to the penetration of the impactor, thus reducing the penetration depth of the impactor. This can be used to explain the phenomenon shown in Figure 6. cles are discrete, and the force chain structure in the granular system will change at every moment due to the frequent formation and breakage of the contact forces between particles, which leads to the zig-zagging change of the contact force. Furthermore, a longer particle can easily trigger stronger contact force between particles and has a stronger resistance to the penetration of the impactor, thus reducing the penetration depth of the impactor. This can be used to explain the phenomenon shown in Figure 6. For the analysis of granular temperature, solid volume fraction, and kinetic energy, the granular bed was divided into three cylindrical regions to investigate the dynamic responses of the particles in the cup, as shown in Figure 9. To distinguish the particles right under the impactor and surrounding the impactor, the radius of region Ⅰ is chosen to be close to the radius of the impactor. Meanwhile, the thickness of region Ⅲ cannot be smaller than the length of the longest particle (~2/7Rc). Therefore, the radial widths of region Ⅰ, Ⅱ, and Ⅲ are 2/7Rc, 3/7Rc, and 2/7Rc, respectively. The edge between two adjacent regions is fixed, and particles are free to enter or leave one region. For the analysis of granular temperature, solid volume fraction, and kinetic energy, the granular bed was divided into three cylindrical regions to investigate the dynamic responses of the particles in the cup, as shown in Figure 9. To distinguish the particles right under the impactor and surrounding the impactor, the radius of region I is chosen to be close to the radius of the impactor. Meanwhile, the thickness of region III cannot be smaller than the length of the longest particle (~2/7R c ). Therefore, the radial widths of region I, II, and III are 2/7R c , 3/7R c , and 2/7R c , respectively. The edge between two adjacent regions is fixed, and particles are free to enter or leave one region. Figure 10b. It can be found that the ϕps of region Ⅰ and Ⅱ are closer, and those of region Ⅲ are slightly smaller. It is interesting to find that the ϕp of the granular bed and the H0 of impactor both decrease with longer particles, which is different from previous studies on spherical particles [13,17]. Due to the random distribution of elongated particles, a 3D cage structure will be formed, and there will be more voids in the structure compared with the granular bed of spheres. Therefore, a smaller ϕp of the bed will be formed by the random packing of longer elongated particles. Figure 9. Diagram of sub-region division for granular bed. Region I: 0 < R 1 < 2/7R c ; region II: 2/7R c < R 2 < 5/7R c ; and region III: 5/7R c < R 3 < R c . (The particles count in one region if their mass centers fall in this region; R c is the radius of the cup). Figure 10a shows the comparison of solid volume fractions for granular beds composed of spheres, L p = 4 mm, 6 mm elongated particles, impacted by glass and brass impactors in simulations and experiments. The initial solid volume fractions φ p s of the granular beds are compared in Figure 10b. It can be found that the φ p s of region I and II are closer, and those of region III are slightly smaller. It is interesting to find that the φ p of the granular bed and the H 0 of impactor both decrease with longer particles, which is different from previous studies on spherical particles [13,17]. Due to the random distribution of elongated particles, a 3D cage structure will be formed, and there will be more voids in the structure compared with the granular bed of spheres. Therefore, a smaller φ p of the bed will be formed by the random packing of longer elongated particles. Figure 10a shows the comparison of solid volume fractions for granular beds composed of spheres, Lp = 4 mm, 6 mm elongated particles, impacted by glass and brass impactors in simulations and experiments. The initial solid volume fractions ϕps of the granular beds are compared in Figure 10b. It can be found that the ϕps of region Ⅰ and Ⅱ are closer, and those of region Ⅲ are slightly smaller. It is interesting to find that the ϕp of the granular bed and the H0 of impactor both decrease with longer particles, which is different from previous studies on spherical particles [13,17]. Due to the random distribution of elongated particles, a 3D cage structure will be formed, and there will be more voids in the structure compared with the granular bed of spheres. Therefore, a smaller ϕp of the bed will be formed by the random packing of longer elongated particles.
(a) (b) Figure 10. (a) Comparison of solid volume fraction ϕp as a function of particle length Lp between experiments and simulations for both glass and brass impactors. (b) Initial solid volume fraction ϕp of the whole granular bed or three regions of granular bed as a function of particle length Lp before impact. The particle length of 2 mm represents the data of 2 mm spheres here.
Granular temperature [38] Tp is calculated to describe the fluctuation of translational velocities of the elongated particles in the cup, as shown in Figure 11. The granular temperature of all particles in the cup can be defined as follows: Figure 10. (a) Comparison of solid volume fraction φ p as a function of particle length L p between experiments and simulations for both glass and brass impactors. (b) Initial solid volume fraction φ p of the whole granular bed or three regions of granular bed as a function of particle length L p before impact. The particle length of 2 mm represents the data of 2 mm spheres here.
Granular temperature [38] T p is calculated to describe the fluctuation of translational velocities of the elongated particles in the cup, as shown in Figure 11. The granular temperature of all particles in the cup can be defined as follows: T p = (T p,x +T p,y +T p,z )/3 (11) in which T p,x = (u x − u x ) 2 , T p,y = u y − u y 2 , T p,z = (u z − u z ) 2 , and u − u is the fluctuating velocities of the particles. The operator u is used to calculate the average velocity over all particles. The T p s of three regions are calculated respectively. = (T p,x + T p,y + T p,z ) / 3 (11) in which T p,x = 〈( − 〈 〉)〉 , T p,y = 〈 − 〈 〉 〉 , T p,z = 〈( − 〈 〉)〉 , and − 〈 〉 is the fluctuating velocities of the particles. The operator 〈 〉 is used to calculate the average velocity over all particles. The Tps of three regions are calculated respectively. In Figure 11, it can be found that the difference between Tps and Lps is limited, indicating that Lp has no significant effect on particle movement. The particle temperature of region Ⅰ experiences its peak value in the first 1 ms, and the other two regions are about 3 ms and 6 ms, respectively, which shows that the particles move from the middle to the periphery. The peak value of Tp of the brass impactor is about twice that of the glass impactor in region Ⅰ and region Ⅱ. In short, the granular temperature in region Ⅰ experiences the most dramatic change, followed by region Ⅱ and region Ⅲ. Figure 12a,b shows the kinetic energy of particles in the cup , which is similar to granular temperature. The particle length has little effect on the dissipation of for any region. The particles left in the cup did not move violently but slightly; therefore, the kinetic energy of the particles left in the cup will not be affected by the particle length. In addition, particle length has little effect on the dissipation of for any region. Comparing the energy dissipation of the three regions, it can be found that the of region Ⅰ has the most drastic change, while that of region Ⅲ has the smallest change. In addition, the In Figure 11, it can be found that the difference between T p s and L p s is limited, indicating that L p has no significant effect on particle movement. The particle temperature of region I experiences its peak value in the first 1 ms, and the other two regions are about 3 ms and 6 ms, respectively, which shows that the particles move from the middle to the periphery. The peak value of T p of the brass impactor is about twice that of the glass impactor in region I and region II. In short, the granular temperature in region I experiences the most dramatic change, followed by region II and region III. Figure 12a,b shows the kinetic energy of particles in the cup E c k , which is similar to granular temperature. The particle length has little effect on the dissipation of E c k for any region. The particles left in the cup did not move violently but slightly; therefore, the kinetic energy of the particles left in the cup will not be affected by the particle length. In addition, particle length has little effect on the dissipation of E c k for any region. Comparing the energy dissipation of the three regions, it can be found that the E c k of region I has the most drastic change, while that of region III has the smallest change. In addition, the total E c k of three regions of glass impactor is larger than that of brass impactor, which indicates that the impactor with less initial energy transfers its higher energy to the particles in the cup. The brass impactor with higher initial energy transfers more energy to the ejected particles, as shown in Figure 5b.
Effect of Friction Coefficient
To evaluate the influence of particle friction on penetration, the friction coefficients of the target particles (μp-p), impactor and particles (μi-p), and cup wall and particles (μw-p) were studied. We chose the steel elongated particles of Lp = 6 mm with initial friction coefficients and the brass impactor for the DEM simulation in this section. Figure 13 shows that the effect of μp-p is dominant. The ejecta mass decreases dramatically when μp-p ranges from 0 to 0.2, and it decreases gently as μp-p changes from 0.2 to 1.0 gradually, while μi-p and μw-p have very little effect on ejecta mass. Figure 13. Relationship between percentage of ejecta mass and friction coefficient μp-p, μi-p, and μw-p
Effect of Friction Coefficient
To evaluate the influence of particle friction on penetration, the friction coefficients of the target particles (µ p-p ), impactor and particles (µ i-p ), and cup wall and particles (µ w-p ) were studied. We chose the steel elongated particles of L p = 6 mm with initial friction coefficients and the brass impactor for the DEM simulation in this section. Figure 13 shows that the effect of µ p-p is dominant. The ejecta mass decreases dramatically when µ p-p ranges from 0 to 0.2, and it decreases gently as µ p-p changes from 0.2 to 1.0 gradually, while µ i-p and µ w-p have very little effect on ejecta mass.
The non-dimensional penetration depth (H 0 /D 0 ) is used to illustrate the correlation among various friction coefficients between particles, which is exhibited in detail in Figure 14a. It is interesting to find that the maximum H 0 of the impactor is about 2D 0 when µ p-p = 0 for condition I, and H 0 decreases with increasing µ p-p. The impactor will rebound to the upside and not penetrate into the granular bed while µ p-p ≥ 0.7. However, the friction coefficients of µ i-p and µ w-p have little effect on H 0 . The maximum values of H 0 for different µ i-p s are less than 0.5 D 0 , as shown in Figure 14b, and the H 0 hardly changes for various µ w-p s in Figure 14c. The vertical velocity of impactor V is shown in Figure 15a,b. When µ p-p is less than 0.7, the decrease of V is positively correlated to µ p-p , as shown in Figure 15a. The changes of V for different µ i-p s and µ w-p s are similar, as shown in Figure 15b.
of the target particles (μp-p), impactor and particles (μi-p), and cup wall and particles (μw-p) were studied. We chose the steel elongated particles of Lp = 6 mm with initial friction coefficients and the brass impactor for the DEM simulation in this section. Figure 13 shows that the effect of μp-p is dominant. The ejecta mass decreases dramatically when μp-p ranges from 0 to 0.2, and it decreases gently as μp-p changes from 0.2 to 1.0 gradually, while μi-p and μw-p have very little effect on ejecta mass. The non-dimensional penetration depth (H0/D0) is used to illustrate the correlation among various friction coefficients between particles, which is exhibited in detail in Figure 14a. It is interesting to find that the maximum H0 of the impactor is about 2D0 when μp-p = 0 for condition Ⅰ, and H0 decreases with increasing μp-p. The impactor will rebound to the upside and not penetrate into the granular bed while μp-p ≥ 0.7. However, the friction coefficients of μi-p and μw-p have little effect on H0. The maximum values of H0 for different μi-ps are less than 0.5 D0, as shown in Figure 14b, and the H0 hardly changes for various μw-ps in Figure 14c. The vertical velocity of impactor V is shown in Figure 15a,b. When μpp is less than 0.7, the decrease of V is positively correlated to μp-p, as shown in Figure 15a. The changes of V for different μi-ps and μw-ps are similar, as shown in Figure 15b. Figure 16 shows the velocity profiles of particles with μp-p = 0.8 at different times im pacted by the brass impactor. The color of the granular bed represents the velocity ma nitude, which is mainly in the range of 0.3 m/s-1 m/s. From Figures 14a and 16, it can b found that the impactor rebounds to the height of 0.5D0 at t = 50 ms. Then, the impact starts falling under gravity, and its velocity reaches a small peak at t = 90 ms, as shown Figure 15a. Some ejected particles will contact with the impactor during t = 90-100 ms, shown in Figure 16, which slows the impactor slightly. The second penetration begins t = 90 ms, corresponding to the decreasing velocity, and the contact position is higher b cause the surface of the granular bed is irregular after the first impact. The maximum H is kept at 0.15D0 when t = 150 ms, indicating the termination of the whole penetration. A a whole, the velocity direction of the impactor is not always downward during its pen tration, resulting in the spatial curve of the actual penetration path of the impactor. addition, when the impactor rebounds, the trajectory of the impactor is a parabola, whic makes the drop-point of the second penetration depart from that of the first one. Figure 16 shows the velocity profiles of particles with µ p-p = 0.8 at different times impacted by the brass impactor. The color of the granular bed represents the velocity magnitude, which is mainly in the range of 0.3 m/s-1 m/s. From Figures 14a and 16, it can be found that the impactor rebounds to the height of 0.5D 0 at t = 50 ms. Then, the impactor starts falling under gravity, and its velocity reaches a small peak at t = 90 ms, as shown in Figure 15a. Some ejected particles will contact with the impactor during t = 90-100 ms, as shown in Figure 16, which slows the impactor slightly. The second penetration begins at t = 90 ms, corresponding to the decreasing velocity, and the contact position is higher because the surface of the granular bed is irregular after the first impact. The maximum H 0 is kept at 0.15D 0 when t = 150 ms, indicating the termination of the whole penetration. As a whole, the velocity direction of the impactor is not always downward during its penetration, resulting in the spatial curve of the actual penetration path of the impactor. In addition, when the impactor rebounds, the trajectory of the impactor is a parabola, which makes the drop-point of the second penetration depart from that of the first one. Comparing the ejecta mass and penetration depth under different particle frictions (Figures 13 and 14a), we presume that the objective particles would contact each other with a large contact force and build a strong quasi-continuous medium under a strong force chain to resist the penetration of the impactor, resulting in the decrease of V, as Comparing the ejecta mass and penetration depth under different particle frictions (Figures 13 and 14a), we presume that the objective particles would contact each other with a large contact force and build a strong quasi-continuous medium under a strong force chain to resist the penetration of the impactor, resulting in the decrease of V, as shown in Figure 15a. Therefore, we can say that the ejecta mass for a larger friction coefficient is decreased by the stronger force chain, because it takes more energy to separate the particles. Based on this consideration, we study the vertical contact force between the impactor and particles F c ip , and the average contact force between particles and particles F c pp , as shown in Figure 17a,b, respectively. It is obvious that the forces F c ip and F c pp increase with increasing µ p-p , and the force chain is the strongest when µ p-p = 1. Figure 16. Velocity profiles of elongated particles (Lp = 6 mm) impacted by brass impactor. (Lp = 6 mm, μp-p = 0.8, arrow represents velocity of impactor, whose color remains unchanged.) Comparing the ejecta mass and penetration depth under different particle frictions (Figures 13 and 14a), we presume that the objective particles would contact each other with a large contact force and build a strong quasi-continuous medium under a strong force chain to resist the penetration of the impactor, resulting in the decrease of V, as shown in Figure 15a. Therefore, we can say that the ejecta mass for a larger friction coefficient is decreased by the stronger force chain, because it takes more energy to separate the particles. Based on this consideration, we study the vertical contact force between the impactor and particles , and the average contact force between particles and particles , as shown in Figure 17a,b, respectively. It is obvious that the forces and increase with increasing μp-p, and the force chain is the strongest when μp-p = 1.
Effect of Particle Configuration
To find the effects of the shape of the cup and particle configuration, we chose steel elongated particles of L p = 10 mm and the brass impactor to pack in the cylindrical and cuboid cup in the lateral and vertical directions, respectively, as shown in Figure 18. The friction coefficients and material parameters are shown in Table 2. In order to make the elongated particles more regular when they are arranged horizontally, the size of the cuboid cup is the integral multiple of the length of the elongated particles. The solid volume fractions φ p s are shown in Figure 19. It can be seen that the φ p s of the cuboid cup are larger than that of the cylindrical cup. The φ p of vertical arrangement is the largest, and the φ p of random packing is the smallest.
The orientational order of particles can be monitored by diagonalization of the symmetric traceless order tensor Q [38], as follows: where l n i(j) is the unit vector along the major axis of the elongated particle n among all N particles. The largest eigenvalue of Q is the primary order parameter S r , which quantifies the degree of alignment. When i = j, δ ij = 1, or δ ij = 0, the order parameter S r is equal to one if all particles are packed in the same direction, and S r is zero if the orientation of every particle is different. friction coefficients and material parameters are shown in Table 2. In order to make the elongated particles more regular when they are arranged horizontally, the size of the cuboid cup is the integral multiple of the length of the elongated particles. The solid volume fractions ϕps are shown in Figure 19. It can be seen that the ϕps of the cuboid cup are larger than that of the cylindrical cup. The ϕp of vertical arrangement is the largest, and the ϕp of random packing is the smallest. The orientational order of particles can be monitored by diagonalization of the symmetric traceless order tensor Q [38], as follows: (12) Figure 19. Comparison between solid volume fractions for different packing types (type 1-lateral arrangement, type 2-vertical arrangement, and type 3-random arrangement). The length of particles in the cup is 10 mm.
Although S r can describe the order of particle configuration, it cannot show the direction of particle orientation. For example, if the major axes of all particles are all in the same inclined direction, the direction of particles cannot be specifically distinguished using Equation (12). Therefore, the average orientational parameter O of all particles is used as the quantitative analysis of the particle orientation. The orientational parameter [39] of a single elongated particle is calculated by the acute angle between the major axis of the elongated particle and the vertical axis. A larger O leads to a more vertical particle. Figure 20a,b show the order parameters S r and orientational parameters O for three packing types in the cylindrical and cuboid cups. The values of S r in the cuboid cup are all bigger than those in the cylindrical cup, which is caused by the restriction of the cylindrical wall.
The orientational order of particles can be monitored by diagonalization of the sym-metric traceless order tensor Q [38], as follows: where ( ) is the unit vector along the major axis of the elongated particle among all N particles. The largest eigenvalue of is the primary order parameter Sr, which quantifies the degree of alignment. When i = j, δ ij = 1, or δ ij = 0, the order parameter Sr is equal to one if all particles are packed in the same direction, and Sr is zero if the orientation of every particle is different.
Although Sr can describe the order of particle configuration, it cannot show the direction of particle orientation. For example, if the major axes of all particles are all in the same inclined direction, the direction of particles cannot be specifically distinguished using Equation (12). Therefore, the average orientational parameter O of all particles is used as the quantitative analysis of the particle orientation. The orientational parameter [39] of a single elongated particle is calculated by the acute angle between the major axis of the elongated particle and the vertical axis. A larger O leads to a more vertical particle. Figure 20a,b show the order parameters Sr and orientational parameters O for three packing types in the cylindrical and cuboid cups. The values of Sr in the cuboid cup are all bigger than those in the cylindrical cup, which is caused by the restriction of the cylindrical wall.
(a) (b) Figure 20. The order parameters and orientational parameters of three packings in cylindrical cup and cuboid cup: (a) is the order parameters for three packing types; (b) is the orientational parameters for three packing types (type 1-lateral arrangement, type 2-vertical arrangement, and type 3-random arrangement). The length of particles in the cup is L p = 10 mm.
As shown in Figure 21, the ejecta masses of the cylindrical cup are all larger than that of the cuboid cup. The ejecta mass of lateral arrangement is the largest, and that of vertical arrangement is the smallest. The elongated particles of lateral arrangement will easily roll out of the cup from the cup edge during impact, which leads to the larger ejecta masses of lateral arrangement. As shown in Figure 22, the H 0 of the random packing is the largest, and the H 0 of the vertical packing is the smallest in the two cups. In short, the cuboid cup has larger S r and φ p , which lead to the smaller ejecta mass and penetration depth of the cuboid cup compared to those of the cylindrical cup. In addition, the vertical arrangement of elongated particles is more regular and has a stronger ability to resist penetration than the lateral arrangement. As shown in Figure 21, the ejecta masses of the cylindrical cup are all larger than that of the cuboid cup. The ejecta mass of lateral arrangement is the largest, and that of vertical arrangement is the smallest. The elongated particles of lateral arrangement will easily roll out of the cup from the cup edge during impact, which leads to the larger ejecta masses of lateral arrangement. As shown in Figure 22, the H0 of the random packing is the largest, and the H0 of the vertical packing is the smallest in the two cups. In short, the cuboid cup has larger Sr and ϕp, which lead to the smaller ejecta mass and penetration depth of the cuboid cup compared to those of the cylindrical cup. In addition, the vertical arrangement of elongated particles is more regular and has a stronger ability to resist penetration than the lateral arrangement.
Conclusions
Three-dimensional DEM simulations and experiments of spheres impacting into elongated particles at a low speed are investigated in this study. By validating the numerical simulations with the experimental results, we demonstrate the feasibility of using the
Conclusions
Three-dimensional DEM simulations and experiments of spheres impacting into elongated particles at a low speed are investigated in this study. By validating the numerical simulations with the experimental results, we demonstrate the feasibility of using the above code to understand the sphere impacting elongated particles. Close agreement between the DEM simulations and experimental results can be obtained in terms of ejecta masses and initial solid volume fraction. The effects of the particle length, the friction, and the configuration of the elongated particle bed on impact are discussed in this study.
The effect of the sphere and elongated particles of four particles lengths on penetration depth, ejecta mass, granular temperature, and contact force are studied. Then, the friction coefficients µ p-p , µ i-p , and µ w-p ranging from 0 to 1 are discussed; the impactor rebounds when µ p-p is larger than 0.7. As for the configuration of the elongated particle bed, three packing types and two different cups are compared, ejecta mass of particles and penetration depth of the impactor are studied. Based on the present studies, the following conclusions can be drawn: 1.
The effect of particle length. The ejecta mass of the spherical particle bed is obviously larger than that of the elongated particle bed. The granular bed of longer particles has a smaller penetration depth due to the spatial structure of elongated particles, although the solid volume fraction is smaller. In addition, the average contact force between particles is positively correlated to particle length. The average kinetic energy of the impactor obeys an exponential-like dissipation, and the particle length of the elongated particles has little effect on the energy allocation from the impactor to the ejected particles and particles in the cup.
2.
The effect of friction. The µ p-p has a significant effect on the ejecta mass and penetration depth of the impactor, while µ i-p and µ w-p have a limited effect. The ejecta mass and penetration depth are negatively correlated to µ p-p . The contact force between particles and particles or impactors are positively correlated to µ p-p . 3.
The effect of particle configuration. The cuboid cup can obtain a more dense and regular granular bed. The ejecta mass and penetration depth of vertical arrangement are the smallest. For the same arrangement of elongated particles, the penetration depth is negatively correlated to order parameters and solid volume fraction.
Conflicts of Interest:
The authors declare no conflict of interest.
Nomenclature
F i external forces of particle i (N) T i external torques of particle i (N·m) m i mass of particle i (g) v i translational velocity of particle i (m/s) I i moment of inertia of particle i (kg·m 2 ) ω i rotational velocity of the particle i (rad/s) δ n overlap in normal direction (mm) F n normal contact force (N) E * equivalent Young's modulus R * equivalent radius of two objects φ p solid volume fraction F c pp average contact force between particles (N) µ p-p coefficient of friction between target particles µ w-p coefficient of friction between wall and particles µ i-p coefficient of friction between impactor and particles T p,q granular temperature of particles along q (q = x, y, z) direction (m 2 /s 2 ) F c ip vertical contact force between impactor and particles (N) O orientational parameter S r the order parameter | 2023-02-18T16:16:49.182Z | 2023-02-01T00:00:00.000 | {
"year": 2023,
"sha1": "294c275a309623874443b0a2660502bfe07b5c9a",
"oa_license": "CCBY",
"oa_url": null,
"oa_status": null,
"pdf_src": "PubMedCentral",
"pdf_hash": "4e7cc877f9d25a3c01d484cf8ce733ddaf9214c3",
"s2fieldsofstudy": [
"Materials Science",
"Geology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
265634228 | pes2o/s2orc | v3-fos-license | Mass Spectrometry Analysis of Shark Skin Proteins
The mucus layer covering the skin of fish has several roles, including protection against pathogens and mechanical damage in which proteins play a key role. While proteins in the skin mucus layer of various common bony fish species have been explored, the proteins of shark skin mucus remain unexplored. In this pilot study, we examine the protein composition of the skin mucus in spiny dogfish sharks and chain catsharks through mass spectrometry (NanoLC-MS/MS). Overall, we identified 206 and 72 proteins in spiny dogfish (Squalus acanthias) and chain catsharks (Scyliorhinus retifer), respectively. Categorization showed that the proteins belonged to diverse biological processes and that most proteins were cellular albeit a significant minority were secreted, indicative of mucosal immune roles. The secreted proteins are reviewed in detail with emphasis on their immune potentials. Moreover, STRING protein–protein association network analysis showed that proteins of closely related shark species were more similar as compared to a more distantly related shark and a bony fish, although there were also significant overlaps. This study contributes to the growing field of molecular shark studies and provides a foundation for further research into the functional roles and potential human biomedical implications of shark skin mucus proteins.
Introduction
Elasmobranchs, encompassing sharks as a prominent example, have garnered considerable research focus owing to conservation endeavors.However, their molecular biology remains a subject of immense scientific interest, despite the inherent challenges associated with experimental investigations.Prior investigations have yielded noteworthy findings with potential implications for human medicine.Notably, the liver and stomach of Atlantic spiny dogfish (Squalus acanthias) sharks unveiled the presence of the antibiotic squalamine [1], while research on chloride channels within the rectal gland of these sharks [2] has proven relevant to the study of cystic fibrosis.Furthermore, the denticle patterns found on the skin of shortfin mako sharks (Isurus oxyrinchus) have proven to be valuable in enhancing the aerodynamic properties of airplanes, specifically in terms of reducing drag and improving lift, which bears resemblance to how these denticle patterns enhance the swimming abilities of sharks in water [3].
One significant contrast between fish and mammalian skin is that the dead, keratinized protective layer of skin known as the stratum corneum is absent in nearly all fish species.Instead, fish epidermis is composed solely of living cells [4,5] and is shielded by a layer of mucus.This slimy substance comprises large glycoproteins called mucins that creates a scaffold for multiple other proteins and glycoproteins, some of which exhibit antimicrobial properties that aid in preventing the entry and establishment of pathogens [4,6,7].In mammalian infection models, increased susceptibility to infections have been demonstrated in animals lacking specific mucins [8][9][10][11].The mucus layer is created both through secretion by various secretory cell types present in the epidermis as well as sloughing of dead cells [12].
Despite the well-documented proteome of the mucus layer in certain common bony fish (Osteichthyes), likely due to the demands of the fish farming industry, our understanding of the proteome within elasmobranchs, the main subclass of cartilaginous fish (Chondrichthyes) including most sharks, remains limited.Shark skin boasts unique attributes, including its tooth-like denticles, suggesting the possibility of novel proteins in the mucus layer with distinctive properties and functions, such as pathogen defense.A characterization of the proteome in shark mucus represents a crucial initial stride toward unraveling its biological significance.Using liquid chromatography-electrospray ionization tandem mass spectrometry, we herein present the most comprehensive proteome description of shark mucus to date in two shark species representing different genera: Atlantic spiny dogfish (Squalus acanthias), one of the most common shark species worldwide, and the bottom-dwelling chain catshark (Scyliorhinus retifer).
Results and Discussion
Over the preceding decade, there has been a notable increase in the number of studies focused on Chondrichthyes, encompassing sharks and rays.These investigations have sought to elucidate their biological intricacies, primarily with a focus on conservationoriented research [13][14][15].Nevertheless, despite the significant importance of exploring Chondrichthyes for conservation and preservation efforts, as well as for translating their exceptional features for potential therapeutic applications, molecular understanding of sharks remains comparatively limited when juxtaposed with bony fish and mammals.
The skin mucus of the more prevalent bony fish (Osteichthyes) and its critical role in fish health has been examined utilizing proteomics methodologies in several studies.The primary emphasis has been on the identification and characterization of innate and adaptive immune system proteins [12,[16][17][18][19]. Sharks differ from bony fish in several aspects including cartilaginous skeleton and skin structure characterized by placoid scales (denticles) that reduce fluid friction, thereby enhancing swimming efficiency [20].The mucus layer of sharks is far less researched than in bony fish and is probably different due to dissimilar skin architecture as well as extensive evolutionary separation.In sharks, proteomics studies are scarce [21][22][23][24] perhaps due to the fact that less than 1% of Chondrichthyan species have a sequenced genome [25], which complicates analysis combined with practical experimental difficulties in handling sharks.In this work, we present, for the first time, the identification of proteins from shark skin mucus using mass spectrometry, with a focus on immune-relevant molecules.
Protein Categorization
Skin protein samples were harvested from spiny dogfish and chain catsharks, as shown in Figure 1 and described in Methods.By utilizing nanoscale liquid chromatography coupled with tandem mass spectrometry (NanoLC-MS/MS) a wide range of proteins were identified from the skin mucus of both shark species.The protein fragments were matched against the Swissprot human database and Uniprot Chondrichthyes database using Mascot 2.5.1, a tool useful for proteome analyses of relatively unexplored species such as sharks, which have very little molecular data available in public databases.Overall, we identified 206 and 72 proteins in spiny dogfish and chain catsharks, respectively, described in detail in Tables 1 and 2 as well as Supplementary Files S1 and S2.These proteins could, in principle, be from either cellular sloughing, a central process in skin physiology, or actively secreted to the mucus.The fact that more proteins were identified in spiny dogfish than catsharks is consistent with a previous study of ours in which less glycans were found in chain catshark skin and may be attributed to the tissue absorption sampling method, which was developed for teleost fish, or represent true biological differences such as a thinner mucus protein layer in chain catsharks [26].Due to the importance of the mucus as a defense barrier, we focused our attention on the proteins that may have immune system roles.Proteins were grouped into eight different clusters based on biological process annotation as described in Methods (Table 3 for dogfish and catsharks, respectively).Carbohydrate and protein metabolism represented almost 40% of the proteins detected, possibly reflective of active regulatory processes such as osmoregulation, respiration, nutrition, or locomotion, as well as defense against pathogens [27], while immune-related proteins represented almost 20%, conceivably reflecting mucus antimicrobial properties.These proportions resemble previously reported data in Atlantic cod [28].We further classified the proteins based on their type and found that 84% and 72% of the proteins in dogfish and catsharks, respectively, are cytoplasmic (including organelles and nucleus residential proteins) and that 19% and 31% are secreted proteins (Table 3 under "cellular location"), perhaps not surprising due to the fact that skin constantly turns over by sloughing.The secreted proteins are particularly interesting, as they include diverse classes of molecules such as mucins, immunoglobulins, proteases, and other proteins that have well-established roles in the immune system [29].In Tables 4 and 5 (secreted proteins), we characterize in detail the secreted proteins found in the sharks' skin mucus and below we discuss their potential roles, together with other immune-related proteins from other classified groups (see Tables 1 and 2).Table 3. Classification of proteins from the shark skin mucus.Proteins from spiny dogfish and from chain catsharks identified using LC-MS/MS.The proteins were clustered into different categories based on the gene ontology category "biological process".Further classification of protein type and cellular location) was carried out using UniProt data (www.oniprot.org)for individual proteins.Some proteins can be found in more than one cellular location and can also have more than one biological classification.Therefore, the sum of proteins from different classifications and locations can exceed the total number of proteins.Highly glycosylated and gel-forming macromolecular components of mucus secretions [30].Also named vWFD domain-containing protein, exhibiting an evolutionarily-conserved von Willebrand factor type D domain (vWD), found in mucins [31].
A0A401S584
Inter-alpha-trypsin inhibitor heavy chain H3 BBBS Heavy chain subunit of the pre-alpha-trypsin inhibitor complex.This complex stabilizes the extracellular matrix through its ability to bind hyaluronic acid, found in mucins (UniProt, [31]).
A0A4W3JXK3
Glucose-6-phosphate isomerase GS Induces immunoglobulin secretion [54] A0A4W3GT93 GDP-mannose 4,6dehydratase GS This enzyme converts GDP-mannose to GDP-4-dehydro-6-deoxy-D-mannose, the first of three steps for the conversion of GDP-mannose to GDP-fucose in animals, plants, and bacteria [55,56] Mucins are large glycoproteins that cover epithelial cell surfaces and form gel-like structures, thereby able to protect against harmful molecules and microorganisms.We found mucin-5B and 5B-like (catshark, dogfish, respectively), mucin-2-like (dogfish), as well as von Willebrand factor domain (vWFD)-containing protein (dogfish, catshark), which all are large, secreted gel-forming mucins harboring a cysteine-rich domain that strengthens the mucus barrier [60]; however, the vWFD can also be found in other, non-glycosylated, proteins.Thus, despite the sharks not appearing "slimy" as bony fish, they are indeed covered by mucins albeit with a thinner layer [26].A study from 2013 on gilthead sea bream (Sparus aurata) skin showed that mucin-2 and mucin-2-like are expressed at relatively low levels and that probiotics [61] and bacterial infection [62] increased mucin-5B expression.Thus, these proteins may serve antimicrobial purposes in the shark skin.In a bioinformatic report from 2016 in which mucin protein sequences in several species were predicted from genomic sequences, mucin-5, 2, and 6 were identified in elephant shark (Callorhinchus milii), although not verified experimentally [63].To the best of our knowledge, the present study is the first time these three mucins (mucin-5B, mucin-2, and mucin-2-like) are shown experimentally in shark skin mucus.
Jawed fish, including Osteichthyes (bony fish) and Chondrichthyes, are the most primitive animals that can make antibodies, however of different subclasses than mammals.Previously, several immunoglobulins (Ig) including IgM, IgH, IgD (IgW orthologue), and IgL were identified in bony fish, whereas three heavy chain isotopes including IgM, IgW, and immunoglobulin novel antigen receptor (IgNAR) were reported in Chondrichthyes [64].Due to its small size, shark IgNAR is often referred to as a nanobody and is the primary an-tibody of a shark's adaptive immune system with a serum concentration of 0.1-1.0mg/mL.Shark IgNAR may have developed from the IgW gene [65] and was previously identified in spiny dogfish serum [66].In spiny dogfish skin, we identified only the secreted IgW heavy chain (Table 1).IgW is believed to be the primordial antibody rather than IgM [67] and was first reported in the spleen of sandbar sharks (Carcharhinus plumbeus) in the early 1990s [41], and later in serum and lymphoid tissues of other sharks [68].Although discovered before in spiny dogfish serum, as well as well as in other shark species organs such as pancreas [69], herein we report for the first time that IgW is present in the shark skin mucin as well, where it may serve an antimicrobial role.
Furthermore, we discovered proteins and enzymes such as GDP-L-fucose synthase, fucolectin tachylectin-4 pentraxin-1 (FTP) domain-containing protein (fragment) and GDPmannose 4,6-dehydratase, which are involved in glycosylation, specifically fucosylation.Fucosylation is a glycan sugar protein modification essential to biological processes such as host-microbiota communication, viral infection or immunity [70].We have previously shown that fucosylated glycans are common on spiny dogfish, chain catshark, and little skate skin mucus proteins [26].Moreover, other proteins that are commonly posttranslationally modified by glycosylation including antithrombin, fibrinogen beta chain, transferrin and serotransferrin, hemoglobin subunit alpha, syndecan binding protein, and cystatin kininogen-type domain-containing protein were also identified in this study.Apart from their well-known role in hemostasis, these proteins have a role in the activation of the immune system [37,38].For instance, Raeder et al. [71] first identified a transferrin-like molecule in Atlantic salmon (Salmo salar) mucus infected with Vibrio salmonicida.The primary role of transferrin, which is a glycoprotein, is to sequester iron in a redox-inactive form making iron unavailable to pathogens, thus starving them [72].This is probably important in the sharks' skin antimicrobial defense, as iron is very limited in sea water [73].
The complement system, a network of more than 50 plasma and membrane-associated proteins, plays a vital role in vertebrate defense against pathogens in the blood as part of the innate and adaptive immune system [74].Upon activation, the intermediate key factor, complement component (C3), acts as a chemoattractant, phagocytotic agent and as agglutinin and initiates a cascade of events leading to bacterial lysis and also acts as an inflammation mediator.While most studies on the complement system have been carried out on blood and internal organs, it has been described to be active in the skin as well since human keratinocytes infected with intracellular Staphylococcus aureus can be attacked by the complement system [75].Notably, in the skin, complement dysregulation, deficiency, and genetic polymorphisms have been associated with a number of diseases such as psoriasis and recurrent cutaneous infection [76].In dogfish, we identified C3, as well as component 1Q (C1q) and complement protein 1S.In addition, sushi domaincontaining protein, which binds complement factors, was found both in spiny dogfish and chain catsharks.The presence of these proteins points to an active immune system, in general, and active complement cascade, specifically, in the skin mucus of dogfish sharks.In fact, a report published as early as 1907 suggested the presence of complement-like activity in dogfish serum [77].Sixty years later, Legler and Evans described the serum hemolytic complement activity in three elasmobranch species including sting ray (Dasyatis americana) and two species of shark, lemon shark (Nagaparion brevirostria) and nurse shark (Ginglyraostoma cirraium) [78].The C1q protein (complement system member) was first reported in skin mucus of European sea bass (Dicentrarchus labrax) [19].To our knowledge, our data are the first description of complement components being present in shark skin mucus where it may play an antibacterial role.
Lectins are proteins that bind to carbohydrates, for example, on bacteria, and serve multiple roles including antimicrobial and developmental [79].Lectins have been reported from various tissues of many fish species including skin mucus [28].We found four lectins in the shark skin mucus, including the following: (1) L-type lectin-containing protein (dogfish), which interacts with N-glycans (components of glycoproteins) in a Ca 2+dependent manner [80].( 2) Calreticulin (dogfish, catshark), which is important for the cell surface expression of MHC class I molecules and antigen recognition [81].(3) F-type lectins such as fucolectin tachylectin-4 pentraxin-1 (FTP) domain-containing protein (catshark), which is implicated in innate immunity (Table 5, detailed explanation).Of note, F-type lectin has been discovered in several fish species in the liver, intestines, and eggs [82][83][84] but to date not in the skin and not in sharks; however, C-type lectin has been found in the skin of Japanese bullhead shark (Heterodontus japonicus) [85].( 4) Calmodulin (dogfish), which also is involved in immune and inflammatory responses [86].Several proteasomes (protease complex, "genetic information processing" group, Tables 1 and 2) were found in both shark types, whereas cysteine proteases such as cathepsin L and B protein were identified in catsharks ("protein metabolism" group, Table 2).Proteases are essential for activation of both the innate and adaptive immune systems and perform complement activation, initiation of proinflammatory responses and the generation of peptides from foreign antigens that are then presented to the major histocompatibility complex in the adaptive immune response [36,95].Proteases have been detected in fish mucus of several cold water fish [96] as well as in fish preferring warmer waters such as the greater amberjack (Seriola dumerili), in which ectoparasite infection increases protease activity [97].Proteases have also been found in the gut of bonnethead sharks (Sphyrna tiburo) where these contribute to food digestion [98]; however, these have not been studied in shark skin.
Several annexins were found in both sharks and are known to regulate the activities of innate immune cells, in particular the generation of proinflammatory mediators, as was described in Atlantic cod [99].Although the role of annexins in the shark skin mucus has not been studied before, epigonal media derived from bonnethead sharks induced apoptosis in human cancer cells, possibly due to annexin as an apoptosis inducer [100], which highlights how sharks can be useful in human medicine.
Actin is one of the most prevalent proteins in eukaryote cells and has several roles including cell movement, cytoplasmic streaming, phagocytosis, and cytokinesis [101].Several reports suggested that the presence of actin and other cytoskeleton related proteins may not simply be due to contamination from ruptured cells but may have a separate role in mucus structure and immune system [102][103][104].In both shark types, we found cytoskeletonrelated molecules (actin, filamin, tubulin, gelsolin, tropomyosin, septin, and keratin), which is not surprising, as these proteins are common and were probably sloughed off.However, these proteins may have immune-relevant function in shark mucus, as shown in Atlantic salmon for actin [103], in rainbow trout (Salmo gairdneri) for keratin [105], and in zebrafish for septin [106].In addition, extracellular actin from insects can bind to bacteria and stimulate their killing by phagocytosis [104].Thus, an intracellular protein may change role when extracellularly located on the skin surface.Moreover, cytoskeletal-related proteins identified in spiny dogfish seemed to participate in shark osmoregulatory tissues [22].All together, these findings suggest that cytoskeletal proteins could be functionally active extracellularly in the shark skin mucus as well.
The 14-3-3 proteins are acidic proteins with several isoforms that are ubiquitously expressed, participate in regulatory processes, and are indirectly involved in immune response [28,107].Ras-related proteins are involved in signal transduction, the regulation of several biological processes, and, aptly, the immune system [108].In both shark types, we identified these proteins ("cell communication" group, Tables 1 and 2).14-3-3 was present in the skin mucus of several fish types [109], but its implication in fish skin (and shark skin) has yet to be determined.Ras proteins were shown to interact with parasite proteins in the skin mucus of common carp (Ichthyophthirius multifiliis) and thus might serve as a drug target [55].As the mucus layer is formed both by secretion and cellular sloughing, proteomics will naturally identify both secretory, membranous as well as intracellular proteins.Of note is that proteins may have dual functions; for example, ribosomal proteins with antimicrobial properties have been identified in rainbow trout and cod skin [56,110].
Protein Interaction
Protein-protein interaction network analysis may shed a light on the predicted function of the identified proteins by revealing their interaction, as well as reveal how similar these interactions are to other species.For that purpose, we used the STRING database [111], and created a proteome interaction network by merging all the proteins identified from the skin mucus of the sharks investigated and compared to published orthologues from other shark and fish species (Figures 2-5).The sources for the maps include interactions from the published literature describing experimentally studied interactions as well as databases.
A confidence score for every protein-protein interaction was assigned to the network in which a higher score is assigned when an association is supported by several types of evidence.To minimize false positives as well as false negatives, all interactions tagged as "low confidence" (<0.4) in the STRING database were eliminated.Thus, the networks are composed of a set number of nodes (proteins) and edges (interactions) (Tables 6 and 7).We found a much higher number of edges when comparing the spiny dogfish and chain catsharks to the phylogenetically close shark species cloudy catshark (Scyliorhinus torazame) and brownbanded bamboo shark (Chiloscyllium punctatum) as compared to the much more distant elephant shark (Callorhinchus milii) and zebrafish (Danio rerio) [112].There is also a similar pattern in number of nodes (proteins) albeit less significant.From the STRING analysis, examining the percentage of proteins that had orthologues with other species revealed that (1) cloudy catshark overlaps 88% with dogfish and 99% with chain catsharks, which suggests that the two catshark species are closely related; (2) brownbanded bamboo shark overlaps 88% with dogfish and 93% with chain catsharks; (3) elephant shark overlaps 81% with dogfish and 89% with catshark, somewhat counterintuitive, as elephant sharks are phylogenetically distant; and (4) zebrafish overlaps 82% with dogfish and 86% with catsharks.Furthermore, up to 19% of the proteins did not have orthologues in other shark species, which could mean that they are unique in the respective sharks or, alternatively, this could be due to methodological differences.These data indicate that most of the skin mucus proteome is conserved and shared among close (although separated by millions of years of evolution) shark species and also a bony fish, and, while speculative, this argues that these proteins may serve important physiological functions.To determine whether the skin proteomes of different species evolved independently (convergent evolution) or were already present earlier in evolution, one would need to sample common ancestors such as coelacanths.Table 6.Spiny dogfish protein interaction summary table using STRING analysis.Proteins identified in skin mucus of spiny dogfish were analyzed when employing orthologues from four different species.Number of nodes depicts the number of orthologues found out of 206 proteins.Number of edges depicts the number of protein-protein interactions found with medium (0.4) confidence.Interaction source is shown both for all active sources (all) and also limited to experiments and databases for more stringent analysis.
No. of Nodes
No Protein interaction map of identified spiny dogfish (A) and chain catshark (B) skin proteins using cloudy catshark orthologues.A possible protein-protein interaction map with high edge confidence was generated using STRING.Ticker edges (line joining the nodes) represent a confidence of 0.4.Edges represent protein-protein association where association does not necessarily mean physical binding of the proteins and there could be involvement of several proteins to a shared function.Note that colored nodes represent different clusters of the query proteins, as employed by STRING software.Full protein names for the abbreviations are provided in Supplementary Files S1 and S2.Note that the larger number of proteins identified in dogfish relative to catsharks yields more interactions; for relative comparisons, see Tables 6 and 7.
Figure 3. Protein interaction map of identified spiny dogfish (A) and chain catshark (B) skin proteins using brownbanded bamboo (bbb) shark orthologues.A possible protein-protein interaction map with high edge confidence was generated using STRING.Ticker edges (line joining the nodes) represent a confidence of 0.4.Edges represent protein-protein association where association does not necessarily mean physical binding of the proteins and there could be involvement of several proteins to a shared function.Note that colored nodes represent different clusters of the query proteins, as employed by STRING software.Full protein names for the abbreviations are provided in Supplementary Files S1 and S2.Note that the larger number of proteins identified in dogfish relative to catsharks yields more interactions; for relative comparisons, see Tables 6 and 7.
Figure 4. Protein interaction map of identified spiny dogfish (A) and chain catshark (B) skin proteins using elephant shark orthologues.A possible protein-protein interaction map with high edge confidence was generated using STRING.Ticker edges (line joining the nodes) represent a confidence of 0.4.Edges represent protein-protein association where association does not necessarily mean physical binding of the proteins and there could be involvement of several proteins to a shared function.Note that colored nodes represent different clusters of the query proteins, as employed by STRING software.Full protein names for the abbreviations are provided in Supplementary Files S1 and S2.Note that the larger number of proteins identified in dogfish relative to catsharks yields more interactions; for relative comparisons, see Tables 6 and 7. Protein interaction map of identified spiny dogfish (A) and chain catshark (B) skin proteins using zebrafish orthologues.A possible protein-protein interaction map with high edge confidence was generated using STRING.Ticker edges (line joining the nodes) represent a confidence of 0.4.Edges represent protein-protein association where association does not necessarily mean physical binding of the proteins and there could be involvement of several proteins to a shared function.
Note that colored nodes represent different clusters of the query proteins, as employed by STRING software.Full protein names for the abbreviations are provided in Supplementary Files S1 and S2.Note that the larger number of proteins identified in dogfish relative to catsharks yields more interactions; for relative comparisons, see Tables 5 and 6.
Therapeutic Implications and Human Relevance
Several studies have suggested that sharks may be relevant for human medicine.A recent comparison of gene transcripts between white shark (Carcharodon carcharias) and zebrafish revealed, surprisingly, that white shark gene products associated with metabolism, molecular functions, and the cellular locations of these functions were more similar to humans than to zebrafish [113].Moreover, squalamine, a compound with a broad-spectrum antifungal, antibacterial, and antitumor activity, that was isolated from spiny dogfish tissues [1] has resulted in a phase I and phase II human trials [114,115].Proteoglycans with anti-osteoarthritic properties isolated from the bramble shark (Echinorhinus brucus) cartilage showed significant improvement in disease parameters in an osteoarthritis rat model [116].Elasmobranchs immunoglobulins and nanobodies (small monoclonal antibodies) have raised a great attention from the scientific community, as they are the earliest jawed vertebrates to possess all the components necessary to perform responses associated with the adaptive immune system [117].The topic of sharks' usefulness in human medicine was elegantly reviewed by Luer and Walsh [117].
Study Limitations
Only female spiny dogfish were sampled in this study.The mucus harvest method may have missed some proteins and did not work well in chain catsharks in which longer absorption time or scraping may be needed.Furthermore, the mass spectrometry analysis used only shows already known proteins; thus, novel proteins unique to sharks may have been missed, and complementary methods for novel protein discovery will thus need to be used in the future.
Animals
Spiny dogfish caught using hook gear were purchased from a commercial fisherman in Chatham, MA in 2022.Only female spiny dogfish were available, likely due to commercial fishing often targeting female schools [118].Chain catsharks were collected from a National Oceanic and Atmospheric Administration survey vessel by dredging in the mid-north Atlantic between 2017 and 2019.All elasmobranchs were housed in tanks with natural sea water flow-through systems, maintained year-round at 14 • C at the Marine Resources Center (MRC) at the MBL.Elasmobranchs were housed in single-species groups and fed a diet of food-grade frozen capelin (Atlantic-Pacific North Kingstown, RI, USA) and fresh, frozen, locally caught squid three days per week.Photos were taken with an iPhone 13 Pro (Apple Inc., Cupertino, CA, USA)).
Experiments were approved by the Institutional Animal Care and Use Committee (IACUC) at the MBL (protocol no 22-22).
Skin Mucus Sampling
Skin mucus were sampled using the Kleenex tissue absorption method, previously developed for salmonoids [119].Briefly, housed elasmobranchs were caught gently with a net and a Kleenex tissue was placed on the skin for 10 s whereafter the tissue was put in Spin-X tubes (Sigma-Aldrich, St. Louis, MO, USA) on ice and later spun down at 700 g in a 4 • C cooled benchtop centrifuge.Tank water controls samples was also harvested by placing the Kleenex (Kimberly-Clark, Irving, TX, USA) briefly in the tank water.The liquid samples were transferred to plastic cryotubes, snap-frozen on dry ice, and stored at −80 • C.
Sample Preparation
Protein content was determined using a colorimetric assay (Bradford protein assay, Bio-Rad, Hercules, CA, USA).Aliquots corresponding to 20 µg protein were processed using a modified version of filter-aided sample preparation (FASP) method [120].The mucus samples were reduced in reduction buffer (6 M GuHCl (guanidinum hydrochloride (ultrapure, MP Biomedicals, Santa Ana, CA, USA), 0.1 M TEAB (triethyl ammonium buffer pH 9.5), 5 mM ethylenediaminetetraacetic acid, 0.1 M dithiothreitol) for 30 min at 37 • C. The samples were transferred to 10 kDa Microcon Centrifugal Filter Units (MPE030025, polyetylensulfon filter, Millipore, Burlington, MA, USA), and washed repeatedly with 6M GuHCl, followed by alkylation with 100 µL 0.05 M iodoacetamide in 50 mM TEAB buffer for 30 min.Digestion was performed in 0.1M TEAB with the addition of sequencinggrade modified trypsin (Promega, Madison, WI, USA) in an enzyme-to-protein ratio of 1:100 at 37 • C overnight.An additional portion of trypsin was added and incubated for 4 h.Peptides were collected by centrifugation, followed by further purification using High Protein and Peptide Recovery Detergent Removal Spin Column and Pierce peptide desalting spin columns (both Thermo Fischer Scientific, Waltham, MA, USA) according to the manufacturer's instructions.
NanoLC/MS
NanoLC-MS/MS was performed on an Orbitrap Exploris 480 mass spectrometer interfaced with Easy-nLC1200 liquid chromatography system (both Thermo Fisher Scientific, Waltham, MA, USA).Peptides were trapped on an Acclaim Pepmap 100 C18 trap column (100 µm × 2 cm, particle size 5 µm, Thermo Fischer Scientific, Waltham, MA, USA) and separated on an in-house packed analytical column (75 µm × 35 cm, particle size 3 µm, Reprosil-Pur C18, Dr. Maisch) using a gradient from 5% to 35% ACN in 0.2% formic acid over 40 min at a flow of 300 nL/min.Each preparation was analyzed using MS1 scans settings, m/z 380-1500, at a resolution of 120 K. MS2 analysis was performed in a data-dependent mode at a resolution of 30K, using a cycle time of 2 s.The most abundant precursors with charges 2-6 were selected for fragmentation using HCD at collision energy settings of 30.The isolation window was set to 1.2 m/z and the dynamic exclusion was set to 10 ppm for 30 s.
Proteomic Data Analysis
The acquired data were analyzed using Proteome Discoverer 2.4 (Thermo Fisher Scientific, Waltham, MA, USA).The raw files were matched against the Swissprot human database (March 2021) and Uniprot Chondrichthyes database (142,499 entries, February 2023) using Mascot 2.5.1 (Matrix Science, London, UK) as a database search engine with peptide tolerance of 5 ppm and fragment ion tolerance of 30 mmu.Tryptic peptides were accepted with one missed cleavage, mono-oxidation on methionine was set as a variable modification, and carbamidomethylation on cysteine was set as a fixed modification.Target Decoy was used for PSM validation.Tables referring to secreted proteins are based on targeted literature searches and UniProt data (www.uniprot.org(accessed on 1 June 2023)).
The proteins identified were clustered into different categories based on Gene Ontology category, biological process.Further classification of protein type and functional hierarchies of biological entities were based on information on KEGG BRITE Database (kegg.jp/kegg/brite.html(accessed on 1 June 2023)) and UniProt (uniprot.org(accessed on 1 June 2023)) for individual proteins.As most of the proteins are not well annotated in teleost species, the Gene Ontology terms were retrieved from its human counterparts.
Protein-Protein Interaction Network Analysis
Protein interaction network maps for the sharks' skin mucus proteins was generated using STRING (https://version-12-0.string-db.org/(accessed on 1 September 2023), employing the following organism UniProt IDs: Cloudy catshark (Scyliorhinnus torazame), Brownbanded bamboo shark (Chiloscyllium punctatum), Elephant shark (Callorhinchus milii, also called Australian ghost shark), Zebrafish (Danio rerio).To achieve a more stringent analysis, the active interaction sources were limited to experiments and databases, and an interaction score >0.4 was applied to construct the protein-protein interaction network.
Chemicals
The chemicals were from Sigma-Aldrich (St. Louis, MO, USA) unless stated otherwise.
Conclusions
This is the first study that describes the skin mucus proteome of sharks.These proteins represent several basic functional groups, and while most of them are cellular proteins, a substantial minority are secreted.We propose these skin proteomes to be relatively conserved between close shark species.Further research on elasmobranch skin is warranted, especially bioprospecting studies that aim to identify completely novel molecules using protein sequencing, decipher their functions experimentally, and, if possible, translate to human clinical use albeit with shark conservation in mind.
Figure 1 .
Figure 1.Experimental setup.(A) Shark species examined.The chain catshark scale is in inches and the spiny dogfish scale is in cm.(B) Sample harvest and analysis.Proteins were harvested by wrapping wet shark skin with a Kleenex tissue for 10 s, followed by centrifugation in SpinX tubes and analysis using mass spectrometry (NanoLC-MS/MS).N = 10 for spiny dogfish and N = 10 for chain catsharks.
Figure 2 .
Figure2.Protein interaction map of identified spiny dogfish (A) and chain catshark (B) skin proteins using cloudy catshark orthologues.A possible protein-protein interaction map with high edge confidence was generated using STRING.Ticker edges (line joining the nodes) represent a confidence of 0.4.Edges represent protein-protein association where association does not necessarily mean physical binding of the proteins and there could be involvement of several proteins to a shared function.Note that colored nodes represent different clusters of the query proteins, as employed by STRING software.Full protein names for the abbreviations are provided in Supplementary Files S1 and S2.Note that the larger number of proteins identified in dogfish relative to catsharks yields more interactions; for relative comparisons, see Tables6 and 7.
Figure 5 .
Figure5.Protein interaction map of identified spiny dogfish (A) and chain catshark (B) skin proteins using zebrafish orthologues.A possible protein-protein interaction map with high edge confidence was generated using STRING.Ticker edges (line joining the nodes) represent a confidence of 0.4.Edges represent protein-protein association where association does not necessarily mean physical binding of the proteins and there could be involvement of several proteins to a shared function.Note that colored nodes represent different clusters of the query proteins, as employed by STRING software.Full protein names for the abbreviations are provided in Supplementary Files S1 and S2.Note that the larger number of proteins identified in dogfish relative to catsharks yields more interactions; for relative comparisons, see Tables5 and 6.
Table 1 .
Identified proteins from spiny dogfish skin mucus grouped into biological groups.
Table 2 .
Identified proteins from chain catshark skin mucus grouped into biological groups.
Table 4 .
Secreted proteins identified in the mucus of spiny dogfish.A literature-based distinction of their immune potential.Organism represents the protein reference species.
Table 5 .
Secreted proteins identified in the mucus of chain catshark.A literature-based distinction of their immune potential.Organism represents the protein reference species.
Table 7 .
Chain catshark protein interaction summary table using STRING analysis.Proteins identified in skin mucus of chain catshark were analyzed when employing orthologues from four different species.Number of nodes depicts the number of orthologues found out of 72 proteins.Number of edges depicts the number of protein-protein interaction found with medium (0.4) confidence.Interaction source is shown both for all active sources (all) and also limited to experiments and databases for more stringent analysis.
Institutional Review Board Statement: The animal study protocol was approved by the Institutional animal care & use committee of the Marine Biological Laboratory, Protocol number: 20-22, date of approval-22 April 2022.Not applicable. | 2023-12-05T16:52:05.097Z | 2023-11-29T00:00:00.000 | {
"year": 2023,
"sha1": "f6322dde7affc837e17c649a4cabb69a258e9ef0",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1422-0067/24/23/16954/pdf?version=1701274097",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "a99f92c5672d5b304fa30867113607b0f268e982",
"s2fieldsofstudy": [
"Biology",
"Environmental Science"
],
"extfieldsofstudy": [
"Medicine"
]
} |
239762724 | pes2o/s2orc | v3-fos-license | The Function of Siberian Fir In The Formation Of Forest Phytocoenoses In The Floodplain Landscapes of North-Eastern European Russia
The formation of spatial patterns with Siberian fir dominating on slopes and floodplains is a consequence of fir adaptability to the influence of periodic water flow due to the relief features.
Background
The peculiarity of Siberian fir (Abies sibirica Ledeb.)growth in the territory of the North-East of the European part of Russia is that its habitats are confined to certain types of landscapes, namely floodplains of streams and rivers, slopes, and watersheds.In the structure of plant communities formed in such areas, fir is generally the predominant species.The purpose of this study is to find out causes why fir tree have been successful in colonizing specific types of landscapes.
The study was conducted in the Komi Republic, Russia.The objects of the study were chosen two forest phytocoenoses with fir dominance, one of which grows on a slope and the otheron a floodplain terrace.A comparative analysis of the complex of factors determining the growth and development of these forest communities was made.
Results
Orographic conditions have been recognized as the main factor determining the species structure of phytocoenoses in the considered landscape types.The relief features of the areas where the forest stands under consideration are located contribute to the development of periodic water flows that have a significant impact on the species structure formation.Comparison of morphological and phenological features of the main forest-forming species of the Komi Republic has shown that the fir has a number of advantages contributing to its success under the conditions of a specific hydrological regime.
Conclusions
The formation of spatial patterns with Siberian fir dominating on slopes and floodplains is a consequence of fir adaptability to the influence of periodic water flow due to the relief features.
Background
Siberian fir (Abies sibirica Ledeb.) is one of the forest-forming species of the dark coniferous taiga of central northern Eurasia.In most of its habitat it grows in a flatland landscape south of the permafrost distribution boundary, occupying the most fertile forest soils (Krylov et al., 1986).In mountainous landscapes fir often rises up to the upper boundary of the forest (1200-2000 m above sea level), forming a significant part of the dark coniferous mountain taiga of the Sayan, Kuznetsky Alatau, Mountain Shoria, North-Eastern and Western Altai, and Salair (Krylov et al., 1986;Belikovich, 2017).
In the north-east of the European part of Russia (Komi Republic), the siberian fir has similar habitats.In the plain part of the southern taiga and the southern part of the middle taiga of the Komi Republic, fir forests are found in all valleys of the Vychegda, Luza and Letka river basins.Another focus of fir forests is the slopes and foothills of the Ural Mountain Range, where they are also quite widespread in river valleys and, especially, in the upper part of the forest belt (Judin, 1954).The northernmost stands with fir can be found in the Subpolar Urals and in the upper reaches of the rivers Manya and Naroda (Mamaev, 1983).In upper reaches of the Pechora river it is widely spread and grows on drained slopes, along valleys of rivers and streams (Sambuk, 1930;Smirnova et al., 2007).
The peculiarity of fir growth is its confinement to the riverside zones, stream floodplains, slopes and watersheds (Terekhov, Paramonov, 2008).Distinctive feature of such local landscapes is saturation of soil organogenic horizon with nutrients and organic matters (Chapin et al, 2011).Phytocoenoses boundaries, if they are in similar climatic conditions, usually associated with differences in a substrate and soil characteristics (Van Der Maarel, 1976).The physical matrix of soil serves as a source of water and nutrients for plants and is a physical support system in which terrestrial vegetation has taken root.For these reasons, soil physical and chemical properties strongly influence all aspects of ecosystem functioning (Chapin et al., 2011).
The observed spatial pattern of fir distribution is traditionally associated with soil fertility and humidity and the amount of moisture in the air (Mamaev, 1983;Krylov et al., 1986;Belova, Bazhina, 2007;Zhelezova et al., 2008).
However, species-response model considering quality of soil organogenic horizon as main structure-forming factor of environment does not explain reasons of fir dominance in sites with high soil fertility.Other forest-forming species, as a rule, have wider tolerance limits to edaphic conditions.
The purpose of this study is to reveal the reasons behind the characteristic spatial distribution of fir habitat and fir dominance in forest ecosystems of floodplain and slope sites.
The task was carried out in two stages.At the first stage, the key factor influencing the formation of species structure of phytocoenoses under specific growth conditions was determined.The second stage considered the features of the main forest-forming woody species of the Komi Republic, namely: spruce (Picea abies (L.) H. Karst.), pine (Pinus sylvestris L.), fir (Abies sibirica Ledeb.), birch (Betula pendula Roth.), aspen (Populus tremula L.), allowing them to adapt to the action of this factor.
Climate characteristics of the regions where the PSPs are located are given according to (https://climatecharts.net).The main climatic characteristics affecting the development of plants -temperatures and precipitation -are represented by climagrams (Fig. 1).Taylor's method (Miller, 1931) was used at construction of climagrams.On an axis ordinate the average monthly values of temperature for the period of 100 years and on an axis abscissa the average monthly quantity of precipitation for the period of 100 years were postponed.The necessary meteodata for the period from 1901 to 2001 were obtained from (https://climatecharts.net).The De Martonne seasonal aridity index (ISDM) was calculated according to (Moral et all., 2016).Data on the average monthly water discharge of the Vym (observation period 1956 to 1988) and Luza (observation period 1955 to 1999) rivers are taken from (http://www.rarcticnet.sr.unh.edu/).Classification of soils in the regions where the PSPs are located is given according to (Dobrovol'skaya et all., 2010).The assessment of soil fertility is based on the edifice requirements of ground cover plant species of the studied fir forests, which according to the Landolta ecological scales (http://www.impb.ru).
Comparative characteristics of the ecological space of forest-forming species are given on the basis of the range ecological scales of D.N. Tsyganov (http://www.impb.ru).Phenological characteristics (time and duration of seed dispersal) and morphological characteristics (size and weight of seeds, ) of the tree species under consideration are given on the basis of literature data.Most of the data on seed morphology are given as species-specific averages, while some data are region-specific.In most literature sources, seed sizes are presented as length, , and width, , thickness, , is given rarely, sometimes only one size is given.Seed sizes and weights of a particular species, given in each source, were averaged, and the average values were combined into samples.Sample mean ̅ , ̅ , ̅ , ̅ were considered as species averages.The equivalent diameter used for comparison of species as a generalized characteristic of seed size was estimated as a geometric mean (Syvitski, 2007), for seeds whose sizes are represented by length and width, according to the expression: For seeds whose sizes are represented by length, width and thickness, according to the expression:
Results
Degree of species dominance.The study has identified characteristic features of natural forest formation in the landscapes under consideration.
Abietetum oxalidoso-herbosum fir forest (PSP-2) is located on a slope formed as a result of erosion of the bank of the Vym river.The stand is mixed in composition; multiple-aged, stratification is not expressed.Fir is the most represented woody species.Two generations of fir trees are distinguished in the age structure: age class III and V.The number of viable understorey, forming clumps in the windows, is dominated by fir.The origin of the understorey is a seedling.Regeneration of other species is not expressed.The underwood is poorly developed and represented by Rubus idaeus L., Juniperus communis L., Rosa acicularis Lindl., Lonicera pallasii Ledeb.Projective coating of grass-shrub layer is 70%.The dominant species composition: Oxalis acetosella L., Fragaria vesca L., Dryopteris carthusiana (Vill.)H.P.Fuchs, Luzula pilosa (L.) Willd.The moss-lichen layer has a projective coating of about 80% and is formed by species of the g.Hylocomium and Hypnum's mosses.
Abietetum rara herbosum fir forest (PSP-5) is located on the flood plain of the left bank of the Luza river.The stand is mixed in composition, multiple-aged.Trees in the stand are not differentiated by height.Prevailing woody species into the stand is fir, which develops intensively.Two generations of fir trees are distinguished in the age structure: age class V and VIII.A viable fir understorey, of seed origin, is abundant, confined to windows and glades, under the crowns strongly oppressed.Regeneration of other woody species in the stand is almost absent, the number of their understorey, compared with fir is insignificant.The underwood is poorly developed, represented by Rubus idaeus L., Sorbus aucuparia L., Lonicera xylosteum L., Viburnum opulus L. Grass-shrub layer has a tessellation structure with dead cover inclusions formed by needles and leaves.Projective coating of this layer is 40%.The species composition is dominated by Pyrola rotundifolia L., Rubus arcticus L., Dryopteris carthusiana Vill., Solidago virgaurea L.. The moss-lichen layer with a 40% projection coating is represented by green mosses.
The main characteristics of stands are given in Table 1.Based on the obtained data it is possible to assert, that Siberian fir is the dominant species in the investigated forest stands: the forest yield of fir at PSP-2 is 58 %, at PSP-5 is 47 %.Taking into account the higher productivity of deciduous species in relation to conifers in the same conditions it is better to determine the degree of dominance of the species based on number.Evaluation of the degree of species dominance as a ratio of the number of individuals of this species to the total number of all individuals shows that the degree of dominance of fir is 72% at PSP-2, at PSP-5 is 77%., 1997).Comparison of climagrams of the study regions shows that the differences are not significant (Fig. 1): climagrams can be found at one place of the coordinate plane, their shape and area are almost the same, which indicates the coincidence of temperature and precipitation dynamics, both during the vegetation period and throughout the year.However, the average annual temperature in the PSP-5 region is 1.5 °C and in the PSP-2 region is 0.1 °C.Effect of increasing average annual temperature is to a shift in climate patterns relative to one another along the ordinate axis and points a longer vegetation season and more heat availability in the PSP-5 region.This is a natural manifestation of the uneven distribution of solar radiation on the Earth's surface and leads to an increase in the vegetation season in the PSP-5 region by an average of 10 days (Atlas ..., 1997).The absolute maximum temperature in the PSP-5 region is 37.9 °C, in the PSP-2 region is 35.7 °C and the absolute minimum temperature is 47.3 °C (48.1 °C), respectively.The displacement of climagrams along the abscissa axis indicates greater precipitation in the PSP-5 region (mainly during the vegetation season), with an average annual precipitation of 604.2 mm for PSP-5 region and 573.3 mm for PSP-2 region.However, ISDM is practically the same: 45.2 for PSP-2 region and 44.8 for PSP-5 region.
Edaphic conditions.Soil fertility is an important factor in forest formation.The orography of compared biogeocenoses differs significantly and, at first glance, cannot claim to be a general factor determining the species structure of the forest ecosystems under consideration.However, the relief effect on plant growth and development is indirect.
Despite the marked difference in reliefs, both biogeocenoses have a specific hydrological regime, namely the presence of periodic water flow.
The influence of hydrogeomorphic conditions on the structure and number of populations cover the banks along rivers and streams has been noted by many researchers (McBride, Strahan, 1984;Bradley, Smith, 1986;Vansplunder et al., 1995;Scott et al. 1996;Cunnings at all., 2015).
The species structure of plant communities in river floodplains is determined by two main processes: seed propagation and germination, which are regulated by time and speed of water flow, the morphology of the underlying surface or channel, and the peculiarities of vegetation phenology (Warren et al., 2001;Merritt, Wohl, 2002).
Since climatic and soil conditions do not favour any tree species in the circumstances under consideration, the topography can be identified as the main factor that forms the species composition of plant communities on slopes and floodplains of rivers and streams.The relief determines the progressive movement of water towards the slope, thus forming a water flow.In turn, water flow generates water erosion, the extent of which is determined by the force and law of its movement, and is manifested by the destruction, movement and deposition of soil and rock particles (Morgan, 2005;Rodrigues et al., 2012).The influence of water flow on vegetation is carried out through the transformation of environmental gradients and disturbance of normal plant physiological functions (Warren et al., 2001;Baptist, 2005;Korzhenevskij, Kvitnitskaja, 2011;Prach, Walker, 2020).The negative impact of water flow on vegetation is especially critical in the initial stage of life cycle -the seed stage.Seeds of plants caught on the surface of slopes and floodplain terraces, as well as soil particles, are exposed to the water flow.Depending on the strength of the water flow, they can be washed off the soil surface or covered with a layer of sediment.Thus, when colonizing free areas, such as newly formed slopes and floodplain terraces, the ability of the seed to anchor itself on the soil surface and germinate is of paramount importance.
The seeds of all woody plants under consideration initially have the ability to float, although their buoyancy is not the same (Andersson et all., 2000).If the seeds of spruce, pine and fir sink in 7-15 days, the seeds of birch are able to float for 235 days (Andersson, et all., 2000;Sannikov, Sannikova, 2008).No data on the buoyancy of aspen seeds are given, but the seeds of Salix sachalinensis and S. Integra seeds, which have similar morphology, remain buoyant for no more than 15 days (Seiwa et al., 2008).In our opinion, this value can be used as an estimate of the buoyancy of aspen seeds.The buoyancy of seeds is largely due to their appendages, such as wings and hair.All wings are covered with a cuticle, which helps avoid moisture (Kozlowski, 1972), and therefore maintain buoyancy.The film wings of coniferous seeds, like the hairs of aspen seeds, form independently of the seed coat and are easily separated from the coat when wet (Kozlowski, 1972;Fedorov, 1978;Gordeeva et al., 1971).The winged growths of birch seeds are formed from the same tissue as the seed coat and contain air-bearing cavities (Takhtajan, 1991).Taking into account peculiarities of seed appendages formation, two scenarios of seed interaction with water flow are possible.The first is the transfer of seeds as any physical body with buoyancy, regardless of their morphology, and the second is the transfer of seeds that have lost their buoyancy.Seeds that have not lost their buoyancy will be washed away with almost any force of the water flow and have few chance hold on to the free territory.Of greatest interest is the impact of water erosion on seeds that have lost their ability to float.
Subaqueous hydrochory, which studies the movement of seeds that have lost their buoyancy, is now seen as an important seed dispersal (plant migration) mechanism.Such seeds are comparable in their physical properties to sediment particles and therefore the sediment transport theory is also used to consider the movement of seeds that have lost their buoyancy (Markwith, Leigh, 2008).
The most important factors affecting sediment transport are size, shape, density and mechanical properties of particles.In order for a particle of sediment to move, the force exerted on it by water must be greater than all the forces opposing the movement.Simulating the movement of particles that are irregularly shaped and differ in size is not an easy task.Accurate description of the motion of such particles is impossible due to the complex nature of the phenomenon.It is common practice to perform experimental measurements to determine the parameters of motion.
A simple model that binds the velocity of the water flow at which the particle breaks off and begins to move, and its average diameter is offered in (Gordon et al., 2004).The expression has a form: √ where is the critical velocity (breakaway velocity), ms -1 ; is the average particle diameter, mm.
From expression (Equation 3) it follows that the bigger the particle size, the bigger must be the water flow speed to start the particle movement.If this model is applied to describe the movement of a seed that has lost its buoyancy, the conclusion is obvious: seeds that are larger in size are more resistant to washing out and therefore have a better chance of settling on the soil surface.
The size of a particle is usually understood as its diameter, but with this parameter only spherically shaped particles can be clearly described.In the case where a particle is irregularly shaped, the concept of equivalent diameter is introduced.By calculating the average equivalent diameters (Equations 1, 2) characterizing the size of the seeds and comparing them, it is possible to establish the tree species whose seeds are most resistant to washing by water flow.
Soil particle sizes of organogenic horizon do not exceed seed sizes and move even at such water flow rates, at which seeds rest.Simultaneously with the process of sediment movement there is a process of sedimentation, which leads to the coverage of seeds, not washed away by water flow, by soil particles.Seed germination depends on the sowing depth (Pugnaire, Valladares, 2007;Fenner, Thompson, 2005) and therefore the thickness of the sediment layer may be critical for the seed.In (Bond, et al., 1999) an alometric expression is suggested linking the weight of the seed and the sowing depth, which allows us to evaluate the adaptability of the seed to sedimentation.The model has a form: where is the maximum depth of seed placement at which a seedling still appears, mm; is the weight of the seed, mg; is a constant equal to 27.3.
From expression (Equation 4) it follows that the greater the weight of the seed, the greater the soil layer through which it can germinate.That is, seeds with a large weight are more resistant to sedimentation and, therefore, are more likely to settle in local landscapes where sedimentation is common.
Morphometric characteristics of seeds of the considered woody species, obtained from the literature sources, are given in Table 4.The sample mean sizes equivalent to diameters calculated using equations (Equations 1, 2) and the mean seed weight are given in Table 5.The mean critical flow velocity, , (Equation 3), at which seed movement begins, and the maximum soil layer depth, h, (Equation 4), at which seed germination still occurs, are given in Table 6.Comparison of average equivalent diameters and masses of seeds of the considered rocks, as well as critical parameters of the habitat, shows that Siberian fir seeds are larger in size and have a larger mass than seeds of other compared species.This means that they are more resistant to being washed away from the soil surface and can also germinate through thicker sediments.
This gives fir an advantage when settling landscapes that are subject to periodic water flow.
The effect of water flow is not constant and is periodic.Obviously, if the timing and duration of such phases of seasonal plant development as the onset of seed, mass seed dispersal and end of seed dispersal will be different from the timing and duration of spring and autumn floods, the plant will be less exposed to water flow or can completely avoid its influence.Such plant will have the advantage of colonizing landscapes that are characterised by periodic water flow.By comparing the dates of these phenological phases for different species and the dates of floods, species less affected by water flow can be identified.
Table 7 shows data on the duration and timing of the phenological phase of seed dispersal of major forest-forming species of the Komi Republic.Sinadskij, 1973;Sukachev, 1928;Krylov et al., 1986;Kapper, 1954;Fomina, 2016; Comparison of the timing and duration of seed dispersal with the graphs of average monthly water flow in the Vym and Luza Rivers (Fig. 3), as well as average monthly precipitation in the sites where the investigated forest stands are located (Fig. 4), shows that the timing and duration of seed dispatches of such species as pine, spruce and aspen coincide with the timing of spring flooding, i.e., with the period of the highest water flow intensity on slopes and floodplain terraces.In this situation, the likelihood of seeds fixation is low.Fir and birch are in a more advantageous position, the seed dispersal of these species coincides with the beginning of the autumn flood due to seasonal increase in precipitation.The intensity of hydrological events during this period, in comparison with the spring flood, is lower, therefore, the probability of fir and birch seeds fixation is higher.Taking into account the higher resistance of fir seeds to water flow and the ability of birch seeds to remain buoyant for a long time, when settling free areas fir trees have an advantage not only in relation to spruce, pine and aspen, but also in relation to the birch.Summarizing the results, we can say that morphological features of Siberian fir seeds (seed size and weight) allow them to withstand the impact (washout from the soil surface and buried by sediments) of more intense water flows than seeds of other forest-forming woody species of the Komi Republic.In addition, the fir seed dispersal phase shift relative to the spring flood date allows it to avoid the most unfavorable period in terms of seed fixation and germination.Thus, with all other conditions being equal, the revealed features of the fir provide it with an advantage in the colonization of slope and riverside landscapes.It should be noted that the adaptation of fir to the phases of water flow occurrence has a negative impact on the inhabitation of plain landscapes, where the effect of water erosion is negligible and the role of such factors as soil temperature and humidity increases.The reverse side of late fir seed dispersal is the shortening of the period of favourable germination, which reduces the competitiveness of fir.
Conclusion
Thus, the dominance of Siberian fir growing in the forest phytocoenoses of the riparian zone of rivers and streams is a consequence of a complex of morphological and phenological features of this coniferous species.
The work was carried out under the budget theme "Spatial and temporal dynamics of structure and productivity of phytocoenoses of forest and swampy ecosystems in the European North-East of Russia", state registration number AAAA-A17-117122090014-8.
Priluzskiy administrative district is a part of Sysolo-Vychegodskaja province in Luzo-Sysolskii district of typical podzolic and swampy podzolic soils(Dobrovol'skaya et al., 2010).Fertile alluvial sod-gley soils were formed in flood plains of rivers and streams of Priluzskiy district.The territory of the Kniazhpogostskogo administrative district is a part of the Sysolo-Vychegodskoi province in the Vym-Vychegodskom district of typical podzolic, peat-podzolic-gleyic illuvial-humus soils(Dobrovol'skaya et al., 2010).Taking into account the relief features of the areas under consideration, the characteristics of their rootcontaining layer at the terrain level will be determined by the soil properties of the adjacent areas from which sediments are washed away.The fertile soil layer in the area where PSP-2 is located represents sediments washed away and washed away by melt and rainwater from adjacent terraced areas because of deluvial process.At the site where PSP -5 is located, the fertile layer is formed by alluvial deposits, the composition of which is determined by the composition of soils and parent rocks of the catchment of the Luza River.Some characteristics of the root-containing layer of soils formed because of wash-off can be estimated using the indicators characterizing the plants' exactness to the soil conditions.The presence of eutrophic plants in the ground cover of the plantations under consideration (Table2)indicates the soil fertility of the root-containing layer at both PSPs.Based on the values of soil parameters necessary for normal existence of these plants, soils of the experimental sites can be characterized as medium dry, neutral in acidity (4.5÷7.5 pH), in terms of nitrogen supply varying from medium to medium wealth, rich in humus, fine sand, well aerated (particle diameter 0.002÷0.05mm).
-according to Landolta ecological scale (Baza dannykh... http://www.impb.ru).Orographic conditions.Relief is an important factor in the formation of any forest ecosystems, although it is not included in its composition.Reliefs of the studied forest stands vary significantly.The abietetum oxalidoso-herbosum fir forest (PSP-2) is located on the slope formed as a result of erosion of the terraced bank of the Vym river.It is oriented from north to south and occupies the entire width of the slope, which is approximately 20 m.The slope of the western exposure with active erosion processes has an incline from 45 to 60 °.The plot is intersected by three shallow hollows formed by temporary watercourses.The water flow on the site is intermittent and is mainly caused by snow melting in spring and precipitation in autumn.A large catchment area ensures the accumulation of large amounts of melt water and rainfall.The abietetum rara herbosum fir forest (PSP-5) is located in the flood plain of the left bank of the Luza River.It occupies an elevated central part of the flood plain area.A feature of the hydrological regime of this area is periodic (1-2 times a year) flooding.The height of the water rise, which can be determined from the traces left by the flood on the trunks of trees, is 60÷70 cm.DiscussionAnalysis.Analysis and comparison of soil-climatic conditions of the forest ecosystems under consideration shows if not complete identity of soil and climate, then at least the proximity of their main parameters.This provides an opportunity to assess the degree of influence of these factors on the formation of species structure of phytocoenoses, comparing ecological spaces (niches) of the forest-forming species under consideration.Fig.2presents the result of comparison in a graphical form, which shows that almost all considered parameters of soil and climate conditions the range of fir tolerance in comparison with other species of woody plants is the narrowest, and, accordingly, the ecological space of fir is part of the ecological spaces of other species.Consequently, if the living conditions of the fir are changed, the fir will be primarily affected by stress.Thus, there are no grounds to consider climatic and edaphic factors favorable for fir and limiting the development of other woody plant species in the local landscapes under consideration.
Fig. 2 .
Fig. 2. Comparative characteristics of the ecological space of the woody species under consideration.1lower boundary of fir ecological space; 2upper boundary of fir ecological space; 3lower boundaries of ecological space of other species; 4upper boundaries of ecological space of other species.Tsyganov's ecologycal scales: Tmthermoclimatic, Kncontinental climate, Omombroclimatic aridity-humidity, Crcryoclimatic, Hdsoil moisture, Trsalt regime of soils, Ntriches of soils with nitrogen, Rcacidity soil.
By the example of two floodplain forest phytocoenoses of the boreal zone of the Komi Republic, in which the fir is the dominant species in the stand, it is shown, that at the similarity of climatic and edaphic parameters, the main factor determining the structure of plant communities is the relief, causing the presence of periodic water flow.Plant seeds, as well as soil particles, are included in the processes of transport and sedimentation caused by water flow, which is crucial for seed establishment and germination.The study showed that wood species adapted to the influence of water erosion have an advantage when colonizing free areas of slopes and floodplain terraces.Properties that lead to adaptation to the impact of water flow are large size and weight of seeds, low buoyancy of seeds, and asynchrony of the dispersal phase in relation to the most intensive phases of water flow.Fir, due to its morphological and phenological features, is better adapted to the negative effects of water erosion.Dispersal time, of seeds are attributes of Siberian fir, providing it with better adaptation to specific forest conditions of slopes and floodplain terraces.
Table 1 .
Taxation characteristics of stands of investigated fir forestsClimatic conditions.According to Köppen's classification, the climate type of the PSPs regions corresponds to Dfc, which is equivalent to a moderately cold climate with uniform humidification and cold summers.Climate identification on the basis of long-term weather data shows that the observed forest stands grow in different climatic conditions: abietetum oxalidosoherbosum fir forest (PSP-2) is located in Vychegodskiy climatic region, abietetum rara herbosum fir forest (PSP-5) is located in Priluzkiy climatic region, which causes differences in a number of climatic parameters (Atlas ...
Table 2 .
Exactingness of plants of the ground cover of forestations to soil conditions
Table 4
Morphological parameters of seeds of the considered forest-forming species.
Table 5 .
Average morphometric parameters of seeds, considered tree species
Table 6 .
Habitat parameters affecting the development of tree species at the seed stage.
Table 7 .
Average deadlines of seed dispersal of the main forest-forming species of the Komi Republic. | 2020-08-27T09:15:06.017Z | 2020-08-25T00:00:00.000 | {
"year": 2020,
"sha1": "a904738a65afbf18891a86badf2c247a0435418b",
"oa_license": "CCBY",
"oa_url": "https://www.researchsquare.com/article/rs-51178/latest.pdf",
"oa_status": "GREEN",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "080b4797d07cfe7df923a1d767e11b4908fabc92",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": [
"Geography"
]
} |
14414689 | pes2o/s2orc | v3-fos-license | On estimating the change point in generalized linear models
Statistical models incorporating change points are common in practice, especially in the area of biomedicine. This approach is appealing in that a specific parameter is introduced to account for the abrupt change in the response variable relating to a particular independent variable of interest. The statistical challenge one encounters is that the likelihood function is not differentiable with respect to this change point parameter. Consequently, the conventional asymptotic properties for the maximum likelihood estimators fail to hold in this situation. In this paper, we propose an estimating procedure for estimating the change point along with other regression coefficients under the generalized linear model framework. We show that the proposed estimators enjoy the conventional asymptotic properties including consistency and normality. Simulation work we conducted suggests that it performs well for the situations considered. We applied the proposed method to a case-control study aimed to examine the relationship between the risk of myocardial infarction and alcohol intake.
Introduction
The problems of detecting abrupt changes at unknown points and estimating the locations of changes are known as the change point problem. The change point problem occurs frequently in medical research. For example, cancer incidence rates remain relatively stable for people at a younger age, but change drastically after a certain age threshold (MacNeill and Mao [13]). The data obtained from a group of preschool boys indicates that their weight/height ratio relates to their age in one way before a certain age but that the functional relation of the two changes afterwards (Gallant [8]). Another example arises from a study of the risk of myocardial infarction(MI), which showed a sharp decrease in risk at low alcohol intakes and a dramatic increase after reaching a certain amount of daily alcohol consumption (Pastor and Guallar [16]). Although these examples each have distinctive features of their own, the common theme here is that the relationship of the response variables and a covariate of interest is subject to an abrupt change at a certain threshold. Very often, scientists are interested in the threshold for clinical or preventive purposes. The change point model is useful in that it purposely includes a parameter to capture the notion of threshold.
In this paper, we focus on the estimation of the change point in the so-called broken-line regression models, where the regression function is assumed to be continuous at the point of change. Estimation for these type of models with normally distributed responses has been developed by various authors. Sprent [18] was among the first to discuss the estimation of the piecewise linear models. His interest in this type of model is based on the observation that a biologist would often postulate a two-phase linear model over some alternatives such as the quadratic model largely on intuitive grounds. Hinkley [10] considered the same two-phase straight line model and derived the maximum likelihood estimator (MLE) of the change point by its marginal likelihood function and presented the asymptotic distribution of the estimator. Feder [7] studied the model in a more general framework and proved the consistency of the least-squares estimators of the regression coefficients and the change point. The estimators are asymptotically normal for some special cases including models with all linear segments. Bhattacharya [2,3] presented the asymptotic properties of the change point and regression coefficient estimators using a local log likelihood process approach. Through this approach, he showed the distinctive features of the asymptotic properties of the change point with and without the continuity constraint at the point of change.
A major difficulty in estimating the change point as a parameter for regression models is the nonsmoothness of the likelihood function with respect to the change point considered as a parameter. Many authors tried to circumvent this problem by using various smooth transitions between the two linear regimes, or using other types of the function such as the quadratic function in one of the segments separated by the change point. This technique was mostly used for the model with normally distributed response variables. Gallant and Fuller [9] discussed such a model and used a modified Gauss-Newton method to obtain the least squared estimates. The simulation studies showed that the estimates obtained by this method work well, but no asymptotic properties were presented in their work. Bacon and Watts [1] proposed a model which can accommodate a smooth transition as well as an abrupt change with a Bayesian estimation procedure to determine the parameter values. To the best of our knowledge, with the exception of normal distributions, there is very little research being done where the response variables are categorical. Furthermore, the asymptotic properties of the estimators are virtually unavailable.
In Section 2, we propose an estimation procedure for the change point and the corresponding regression coefficients in the framework of generalized linear models, and present the asymptotic properties of the proposed estimators. We discuss in Section 3 some extensions of the estimation procedure to more general models involving a change point. Section 4 discusses the choice of the bandwidth parameter and the smoothing function for the proposed estimation procedure. Section 5 presents the simulation results to assess the finite sample performance of the proposed method. In Section 6, we apply the proposed method to a data example from the EURopean study on Antioxidant, Myocardial Infarction, and breast Cancer (EURAMIC, Kardinaal et al. [12]) followed by discussion in Section 7. The proofs of the propositions are given in the Appendices.
Proposed estimation procedure for the change point
Our approach is closely related to the density estimation, whereas the cumulative distribution function (CDF) is estimated by the empirical distribution function (EDF), but the density function cannot be estimated by the derivative of CDF, On estimating the change point 307 which in essence is a step function (Rao [17]). In density estimation, a kernel function (usually a known density function itself) is introduced to solve this problem by making the EDF smooth and differentiable everywhere. The proposed estimation procedure is directly motivated by the approach adopted by Horowitz [11] and in the Econometrics literature. In particular, Manski [14] showed that the maximum score estimator of the coefficient vector of a binary response model is consistent under a weak distributional assumption. However, the asymptotic distribution of the maximum score estimator was not presented since the objective function it maximizes is a step function. Horowitz [11] used a smooth version of the same objective function to make it continuous and differentiable. The procedure produced an estimator that not only converges more rapidly but also has a tractable asymptotic distribution.
For a sequence of random variables Y i , i = 1, . . . , n, having probability (density) function of the form with the natural link function, we consider the following model incorporating the unknown change point τ as where a + = aI(a > 0) and I(a > 0) is the indicator function, and β 2 = 0 for identifiability of τ (Davies [5]). Here β 0 and β 1 are the intercept and the slope relating the response variable Y , through the link function g, to the covariate x for x ≤ τ, and β 2 is the difference in slopes for the segments before and after the change point τ . With the traditional likelihood approach, the likelihood function is not differentiable with respect to the change point τ . Specifically, the indicator function I(x > τ ) is not differentiable with respect to τ . Consequently, one of the regularity conditions for the usual asymptotic theory, namely, a certain degree of smoothness of the objective function with respect to the parameters, is violated. To circumvent this critical problem, define a continuous function K(·) which satisfies 1. |K(u)| < M for some M, 0 < M < ∞ and u ∈ (−∞, ∞). We propose to estimate the change point τ as well as the regression coefficients β 0 , β 1 and β 2 by maximizing the following objective function:
Here K(·) is analogous to a cumulative distribution function rather than a density function, the latter being more commonly used in problems such as density estimation.
Note that the difference between this objective function Q n and the otherwise usual log-likelihood function is that the indicator function I(x > τ ) in the likelihood function is replaced by the smoothing function K((x − τ )/h n ). It is clear that the objective function Q n is twice differentiable with respect to all parameters, and with some suitable conditions on the distribution for x, K((x − τ )/h n ) converges to I(x > τ ) uniformly as n → ∞. As shown below, these two factors play key roles in the asymptotic behavior of the estimates which maximize Q n . Define Σ n (β, τ ) = cov β,τ S n (β, τ ) and J n (β, τ ) = −▽S n (δ).
Proposition 1 (Asymptotic normality). Letδ = (β,τ ) be the estimators which maximize the objective function Q n (β, τ ) and δ 0 be the true value of δ, then under some regularity conditions and given that (i) X i is assumed to be bounded, where i = 1, . . . , n; (ii) lim n→∞ P (|X n − τ | < ǫ) = 0 for some ǫ > 0; and the normed estimator for δ is asymptotically normal, i.e., Or equivalently, by the results of Lemma B.1 in Appendix B.
Proof of Proposition 1 is given in the Appendices.
Some extensions
The model (2.2) discussed so far involves two straight lines, which is characterized in the literature as broken line regression or joint point models. Straight line describes the abrupt change mechanism more distinctively than other types of models. As discussed in the introductory section, some authors (e.g., Pastor and Guallar [16]; Gallant [8]) used a quadratic-linear or linear-quadratic model to characterize the change point. For example, a linear-quadratic model is expressed as A quadratic-linear model can be expressed similarly by changing the indicator function to I(x i < τ ). The advantage of this type of model, specifically, is that for estimation purposes the likelihood function has the first derivative with respect to the change point. Thus the Fisher information can be derived as the covariance of the score function. However, the usual asymptotic properties still cannot go through.
Since both linear-linear and linear-quadratic (or quadratic-linear) models face the same non-differentiability problem in τ , the approach proposed in section 2 can be easily adopted and extended to the latter situation. Corollary 1. Replacing the term x i − τ by (x i − τ ) 2 in Q n , and with the regularity conditions tailored to such a replacement, the resulting estimators of (β 0 , β 1 , β 2 , τ ) for model (3.1) are consistent and asymptotically normal.
The proof is similar to that of the linear-linear model and is omitted. In model (2.2), we have assumed that there is only one independent variable, which involves the change point. This method can be easily extended to situations where adjustment for additional independent variables is needed. Specifically, suppose there are k additional variables z 1 , . . . , z k , the systematic component part of a generalized linear model can be modified as The asymptotic properties of the estimators for γ and δ are similar to those in model (2.2) without additional independent variables and are thus omitted.
Computational issues
The estimation of the parameters (β 0 , β 1 , β 2 , τ ) involves the value of the bandwidth parameter h n . For a smoothed maximum estimator of a binary response model, Horowitz [11] suggested choosing bandwidth h n ∝ n −1/(2m+1) , where m is the order of the kernel K ′ (·) defined as the integer satisfying the following In our asymptotic analysis, with K(·) appearing as a single term in some of the elements of the hessian matrix, the form of h n does not appear to affect the asymptotic properties of the estimators as long as h n → 0 when n → ∞ and meets the restriction as follows.
A sufficient condition for the uniform convergence of K(( Generally, in order to satisfy the above conditions, we need to have h n << min(|x i − x j |, 1 ≤ i, j ≤ n). Specifically, in the iteration procedure to estimate the parameters, h n needs to be small enough so that the negative hessian matrices are positive definite.
It is recognized in the density estimation literature that the choice of the smoothing function does not affect the asymptotic properties of the density estimator. In the simulation that follows, we are listing the results where K(·) is chosen as the cumulative distribution function of the standard normal distribution. We have also conducted the simulation with K(·) as the cumulative distribution function of the exponential distribution. Results concerning the distribution of the estimators and the actual estimated values are very similar for the two different K(·). However, for the same n, h n can be chosen as a function of n that approaches 0 in a slower rate with K(·) being the exponential cumulative distribution function for the algorithm to converge. This is consistent with the conditions (a) and Note that x − τ is bounded. This is equivalent to requiring h n to converge to 0 in a relatively slower rate than K ′ (∞) and K ′′ (∞). Since the exponential density at the tail converges to 0 at a slower rate than the normal density, the corresponding h n can be chosen this way as well.
We have found in our simulations that the objective function Q n when evaluated at the true values β 0 0 , β 0 1 , β 0 2 , a function of τ , is rather flat around the true change point value τ 0 . In dealing with this issue, aside from having a small tolerance level (10 −5 at most) for convergence, another crucial issue in actually carrying out the estimation algorithm is to choose the initial values of the parameters β 0 , β 1 , β 2 and τ . In our simulations in section 5 and the example in section 6, for the data generated, we first made some graphs of nonparametric models such as the LOESS model, then we chose a number of τ i , i = 1, . . . , I, which on the graph are near the potential change point. Next we consider the GLM models. With the fixed change point valued at τ i , the corresponding β(τ i ) values were then obtained using ordinary GLM fitting software. τ p = arg max τi Q n (β(τ i ), τ i ) is chosen as the initial estimate for τ , and β(τ p ) as the initial values for β = (β 0 , β 1 , β 2 ).
After replacing the indicator function in the likelihood function with the smoothing function K(·), the model can be categorized as having nonlinear parameters. For easy calculation, according to section 11.4 in McCullagh and Nelder [15], model (2.2) can be approximated by hn )} and τ 0 is an initial value of τ , which supposedly is close to true τ .
After this approximation, the coefficient estimates as well as their covariance matrix can be easily obtained by regular software. The estimated τ is thenτ = τ 0 −ĉ/β 2 , and the approximate standard error by the delta method is {(1/β 2 , −ĉ/β 2 2 )cov(ĉ,β 2 )(1/β 2 , −ĉ/β 2 2 ) t )} 1/2 . As illustrated in our simulations, the value of the standard error obtained by this approach is similar to the one obtained by the formula supplied in Proposition 1. Hence, from a practical viewpoint, the proposed method is readily implemented in software packages such as SAS and R.
Simulation studies
Simulation studies were conducted to examine the finite sample performance of the proposed estimating procedure. Specifically, normal data with identity link and binary data with logit link incorporating a change point were considered in the simulations.
For the general linear model, we assume that the response variable is normally distributed with constant variance σ 2 = 1, and that the independent variable X follows a uniform distribution with a range of [−2, 2]. We generated the data according to In this model, the change point in which the relationship between Y and x changes is set at τ = 0.5. For each of the 1000 replications, we generated a sample of n=500 independent observations from the distribution N (µ(x), 1). The mean, median and standard deviation ofδ proposed in section 2 as well as their average standard errors calculated based on Proposition 1 are presented in Table 1. As discussed in Section 4, results of the estimated values as well as the precision of the estimates are not sensitive to the choice of the smoothing function. The smoothing function used in this simulation is K(u) = 1/(2π) 1/2 u −∞ e −t 2 /2 dt, and the smoothing parameter h n is set to be n −2 . The average s.e. by Delta method in the table is referring to the standard error obtained via the formula discussed in section 4. Note that the mean and median of the estimates are very close to the true values. This, along with the fact that the average standard errors are very similar to the corresponding sample standard deviation of the estimates suggests that our estimators converge to the true values and the normal approximation for the distribution of the estimators is valid.
The empirical coverage probabilities for δ, which in this simulation are the proportion of the 95% confidence intervals (calculated by the usual estimate ±1.96se) for the 100 random samples containing δ, by and large, are close to the nominal 95% level. In addition, the bootstrapped confidence intervals are also obtained following the bootstrap resampling scheme suggested by Boukai [4]. For each of the random samples (X i , Y i ), i = 1, . . . , n, first the change point τ as well as other parameters were estimated according to the proposed estimation procedure. The bootstrapping resample (X * i , Y * i ), i = 1, . . . , n was then generated in such a way that (X is defined as the nearest integer to a. We obtained 1000 such resamples for each of the 1000 original random samples, and then the empirical 95% bootstrap confidence interval was obtained following the application of the proposed estimation procedure to these resamples. The proportions of these 1000 bootstrap confidence intervals containing the true parameter values of δ are listed in Table 1 as the coverage probabilities of the bootstrapped CI's, which are reasonably close to the coverage probabilities of the normal CI and hence the nominal 95% level.
The Q-Q plots for the estimated β and the change point τ in Figure 1 also show that the sampling distributions of these estimators are close to the normal distributions. Results in Table 1 and Figure 1 suggest that the asymptotic normality of the estimators is well supported by the findings via this simulation.
For the logistic regression, we generated the regressor X from a uniform distribution U (−2, 2), and the binary response variable Y according to As in the normal response case, we also generated a random sample of 500 from this model with 1000 replications. The smoothing parameter used here is h n = n −3 . Results as shown in Table 2 and Figure 2 suggest that similar conclusions can be obtained for binary data as well.
6. An example: The EURAMIC study In this section, we applied, as an illustration, the proposed estimating method in Section 2 to the data from the EURAMIC study (EURopean study on Antioxidants, Myocardial Infarction, and Breast Cancer). The EURAMIC study (Kardinaal et al. [12]) is an international case-control study conducted in eight European countries and Israel, which was designed primarily to evaluate the association of antioxidants with the risk of developing a first myocardial infarction (MI) in men aged older than 70. Our example focuses on the portion of the data involving the dose-response of alcohol intake and risk of myocardial infarction. For comparison purposes, we used the same sample as in Pastor and Guallar [16], who entertained the change point model. In this sample, there are 330 cases who had a confirmed diagnosis of first acute MI and 441 controls who were obtained through several random sample schemes. The primary risk factor is alcohol intake during the year before the study took place. It is well-recognized in the literature that the risk of MI as a function of alcohol intake is J-shaped. To capture this phenomenon, Pastor and Guallar [16] have fitted a quadratic-linear model to relate the risk of MI, in logit scale, to alcohol intake, adjusting for other covariates including age, smoking status, waist-hip ratio, history of diabetes, history of hypertension, family history of coronary disease and the dummy variables identifying the medical centers where the data were obtained. Table 3 shows the estimates and their confidence intervals reported by Pastor and Guallar [16]. The approach they took is likelihood ratio-based using the original likelihood which has only the continuous first derivative with regard to τ . In their approach, the authors estimated the regression coefficients and τ by maximizing the likelihood function without smoothing. Recognizing the breakdown of the conventional asymptotic results, they computed the likelihood ratio-based confidence intervals for τ and other parameters.
Also presented in Table 3 are results using the proposed estimating method. Here K(·) is chosen to be the cumulative distribution function of the standard normal random variable, and h n = n −3 , where n = 771, the total sample size for this data set. Both approaches give rise to very similar point estimates for the change point τ as well as the β coefficients. Specifically, the threshold for which the risk of MI changes its direction is estimated at 13.1(±4.68) grams per day. However, the confidence intervals of τ are rather different from each other. While the approach by Pastor and Guallar [16] did not completely address the problem resulting from the nonsmoothness of the likelihood function with respect to the change point parameter, our estimated confidence interval for τ (τ ± 1.96s.e.) is based on a valid asymptotic theory and, at least for this example, provides a much tighter interval for investigators to pinpoint the point where the well-known protective alcohol intake effect may be reversed.
Discussion
Many approaches have been suggested to capture the phenomenon of abrupt change in relating the response variable to a particular independent variable. These include non-parametric smoothing and transforming or categorizing the continuous independent variables, etc. Like other approaches, the change point model is unlikely to fully capture the underlying mechanism. This approach, however, is appealing in that a specific parameter, τ in this case, is introduced to quantify the scientific objective of interest. In the alcohol intake and risk of MI example discussed above, by estimating the threshold for alcohol intake where the risk of MI is apparently heightened and the direction altered, the clinician is in a position to inform patients at risk the maximum allowable level of daily alcohol consumption. In fact, clinical advice as such is given daily to patients at risk for various diseases. Therefore, obtaining more precise estimates for these threshold values via the change point models rather than giving a simple cutoff point via conventional wisdom or experience would have significant clinical implications for health care practice.
On the other hand, the introduction of the change point parameter into the statistical model brings intriguing theoretical difficulties in both detection and estimation of such phenomenon. Our primary objective in this paper is to provide an estimate for the change point with desirable asymptotic properties under the GLM framework. The consistency and the asymptotic normality of the proposed estimators, whose variance can be easily estimated, enhances the opportunity for researchers to make statistical inference on the change point parameter.
Our approach of using the modified objective function eliminates the nonsmoothness problem with the change point parameter in the likelihood function. Similar modification can be applied to the situation where only the first two moments of the response variable is available, and the full knowledge of the probabilistic mechanism is absent. This work will be reported elsewhere.
Appendix A: Proof of consistency
First, we state the consistency of the estimators, which is needed in the proof asymptotic normality, as Proposition A.1 (Consistency). Letδ = (β,τ ) be the estimators which maximize the objective function Q n (β, τ ) and δ 0 be the true value of δ, then under some regularity conditions and given (i),(ii) and (iii),δ converges in probability to δ 0 , i.e.,δ P → δ 0 = (β 0 , τ 0 ). To show the consistency of the estimators maximizing the objective function Q n (β, τ ), we need the following lemma. Lemma A.1. For a sequence of random variables Y i satisfying (2.2), with the same conditions as in Proposition A.1, Proof. First, under the condition lim n→∞ P (|X n − τ | < ǫ) = 0, ǫ > 0, there exists a N > 0, s.t. for n > N, P (|X n − τ | < ǫ) < η, for any η > 0. K( xi−τ hn ) → I(x i > τ ) uniformly for i > N . In addition, if |X i − τ | ≥ ǫ and K(·) satisfies equation (4.1), it can be easily shown Note that b ′ (·) = g −1 (·), Since var(Y i ) < ∞, by the weak law of large numbers, With some constant C > 0, combining this with the condition P (|X n − τ | < ǫ) → 0 as n → ∞, Since N is finite, the first term converges to 0 as n → ∞, and η can be made arbitrarily small, hence we have Proof of Proposition A.1. The log likelihood function for the observed Y ′ s is By the weak law of large numbers, 1
By Jensen's inequality, and since
); Hence we have: (B) l 0 (β, τ ) maximizes at the true parameter values (β 0 , τ 0 ). (C) If X ′ s are bounded, and (β, τ ) ∈ is a compact set, it follows that since Y (β 0 + similarly, as in the proof for consistency of maximum likelihood estimator, we have By (A), (D) and the triangle inequality, This implies that Q n (β, τ ) satisfies the hypothesis of Theorem 4.1.1 of Amemiya (1985). Hence we haveδ P → δ 0 .
Appendix B: Proof of asymptotic normality
In this appendix, we prove Proposition 1. We first introduce the expressions of Hessian matrix elements and some lemmas needed to prove Proposition 1. Aside from the conditions stated in the lemmas, the same regularity conditions as in Proposition 1 are implied as well.
For notational purposes, let where and the nonzero elements of the matrix ∂ 2 θi ∂δ 2 are Proof. Note that there are basically three different types of terms that need to be considered in both matrices J n /n and Σ n /n. First, 1 n n i=1 (g −1 ) ′ (θ i )x k i K( xi−τ hn ) for k = 0, 1, 2. Under the conditions (i), (ii) and condition (III) on the link function g(·) along with equations (2.3) and (2.4), the above terms converge to finite numbers as n → ∞, with the same arguments in proof of Lemma A.1.
The expectation of the left side of (B.1) converges to the moment-generating function of the standard normal. It follows that the expectation of the right side of (B.1) also converges to exp(ω 2 /2). Thus, λ ′ Σ | 2008-05-16T09:32:02.000Z | 2008-05-16T00:00:00.000 | {
"year": 2008,
"sha1": "b6169c9ecb6de1f962bba7f7215608b0e38e7174",
"oa_license": "implied-oa",
"oa_url": "https://doi.org/10.1214/193940307000000239",
"oa_status": "HYBRID",
"pdf_src": "Arxiv",
"pdf_hash": "b6169c9ecb6de1f962bba7f7215608b0e38e7174",
"s2fieldsofstudy": [
"Mathematics"
],
"extfieldsofstudy": [
"Mathematics"
]
} |
16897282 | pes2o/s2orc | v3-fos-license | Status of the connection between unidentified EGRET sources and supernova remnants: The case of CTA 1
In this paper we briefly comment on the observational status of the possible physical association between unidentified EGRET sources and supernova remnants (SNRs) in our Galaxy. We draw upon recent results presented in the review by Torres et al. (Physics Reports, 2003), concerning molecular gas in the vicinity of all 19 SNRs found to be positionally coincident with EGRET sources at low Galactic latitudes. In addition, we present new results regarding the supernova remnant CTA~1. Our findings disfavor the possibility of a physical connection with the nearby (in projection) EGRET source. There remains possible, however, that the compact object produced in the supernova explosion be related with the observed $\gamma$-ray flux.
A brief overview of the EGRET-SNR connection
We have recently presented a case-by-case analysis of the positional correlation between unidentified EGRET sources and supernova remnants (SNRs, Torres et al. 2003). There are 19 EGRET sources at low Galactic latitudes (|b| < 10 o ) in spatial correlation with, mostly shelltype, SNRs. The Poisson probability for the 19 coincidences to be a chance effect is 1.05 × 10 −5 , i.e., there is an a priori 0.99998 probability that at least one of the positional associations is physical. This expected chance association was computed using a numerical code described by Romero et al. (1999), and Sigl et al. (2001). With the aim to identify the most plausible candidates for an EGRET-SNR connection, we have analyzed, among other things, the variability status of the EGRET sources and the potential contribution from known radio pulsars to the high energy flux .
In order to understand the origin of all the unidentified detections, their variability status is of paramount importance (Nolan 2003. Classic known models for γ-ray sources in our Galaxy, like a SNR-molecular cloud interaction (e.g. Aharonian & Atoyan 1996), would produce non-variable detections during the timescale of EGRET observations. To classify the variability status we take into account both indices quoted in the literature, Torres et al.'s (2001) I and Tompkins' τ . They are different in nature, although results have proven to be statistically correlated (Torres, Pessah & Romero 2001). The index I is a relative classification of variability with respect to the pulsar population. Sources with I > 2.5 are more than 3σ away from the statistical variability of pulsars. Similarly, sources for which τ is at least 0.6, are also more than 3σ away from the mean value of the τ upper limit for pulsars, and are thus classified as variable. For comparison, note that the mean value of τ (I) for known AGNs is 0.9 (3.3). We find that most EGRET sources in positional coincidence with SNRs, and certainly those which represent the best cases for a physical association, are non-variable.
In order to unambiguously identify a pulsar as the origin of the γrays from an EGRET source, γ-ray pulsations must be detected at the pulsar period. However the γ-fluxes are generally very low, and nearly contemporaneous radio/X-ray and γ-ray observations are required to fold the few available photons with the correct ephemeris. Because this is no longer possible owing to the demise of the Compton Gamma-Ray Observatory, new candidate associations must mostly be judged by comparison with the properties of the six known EGRET pulsars and those of the known best candidates. We have done so for every known radio pulsar found to be coincident with these 19 EGRET sources, and have isolated several interesting candidates for an EGRET source-pulsar connection. These are mostly found with young pulsars pertaining to the recently released Parkes Multibeam Survey (see, e.g. Torres, Butt, & Camilo 2001, Torres & Nuza 2003. This scenario, then, is a competing one to the EGRET-SNR connection and has to be taken into account in judging the plausibility of the latter. The need of a case-by-case analysis is then clearly established. Several cases for a physical association, based on the analysis of γ-ray data, the molecular environment of the SNR, and the analysis of any coincident pulsar (if any), have been isolated and reviewed in Torres et al. (2003). Among the most notable cases for such a physical association are G347.3-0.5, W66, W28 and IC443. The first of these, providing probably the best evidence to date for a hadronic origin of the γ-ray source detected by EGRET, has received much attention recently , Enomoto 2002, Uchiyama et al. 2002. In addition, the case-by-case analysis shows that it is at least plausible, contrary to expectations, that EGRET has detected distant (more than 6 kpc) SNRs. There are 5 coinciding pairs of 3EG sources and SNRs for which the latter apparently lie at such high values of distance (disregarding those related with SNRs spatially close to the galactic center). For all these cases, we have uncovered the existence of nearby, large, in some cases giant, molecular clouds that could enhance the GeV signal through pion decay. It is possible that the physical relationship between the 3EG source and the coincident SNR could provide for these pairs a substantial part of the GeV emission observed. This does not preclude, however, composite origins for the total amount of the radiation detected, since some of these cases present other plausible scenarios (e.g. energetic pulsars in the EGRET field). AGILE observations, in advance of GLAST, would greatly elucidate the origin for these 3EG sources, since even a factor of 2 of improvement in spatial resolution would be enough to reject the EGRET-SNR connection.
In what follows we attempt to briefly analyze, similarly to what was done in the case-by-case analysis of Torres et al. (2003), the pair CTA 1/3EG J0010+7309, a supernova remnant lying at relatively high Galactic latitudes, and for which there has been little analysis on their CO environment.
CTA 1: General Features
CTA 1 is a shell-type SNR, but its shell is incomplete and brokenout towards the NW. This breakout phenomenon may be caused by more rapid expansion of the blast wave shock into a lower density region toward the NW. HI observations supported this interpretation (Pineault et al. 1993(Pineault et al. , 1997. The latter papers place the SNR at 1.4±0.3 kpc, the kinematic distance corresponding to the systemic velocity of a partial HI shell (−16 km/s) that the authors find to the NW. However, it is to be noticed that the systemic (unperturbed) velocity of a partially expanding shell might be subject to a great uncertainty.
Since CTA 1 is located at a relatively high latitude (b =10 o ) and it is nearby, it has a large angular size (90 arcmin), little foreground or background confusion, and it can be observed at exceptionally high linear resolution. The age of the SNR was estimated to be 10 4 yr by Pineault et al. (1993), but it could be younger by a factor of 2 (Slane et al. 1997). CTA 1 was subject of intense observational campaigns in the past years. There have been both ROSAT and ASCA X-ray observations (Seward et al. 1995, Slane et al. 1997, as well as optical, infrared and radio studies (see Pineault et al. 1997 andBrazier et al. 1998 for a review).
CTA 1 belongs to the class of composite SNRs (SNRs with central pulsar wind nebulae), which show a shell-type morphology in the radio band and are center-filled in X-rays. Five point sources were detected with ROSAT, one of which was found to coincide with the EGRET source (at the time of the analysis, 2EG J0008+7307), see Figure 3 of Brazier et al. (1998). ASCA data suggested that this source, named RX J0007.0+7302, has a non-thermal spectrum, and gave support to the idea that it is the pulsar left from the supernova explosion (Slane et al. 1997). Recent observations by Slane et al. (2004), using XMM, confirm that the X-ray spectrum of the source is consistent with that of a neutron star and is well described by a power law with the addition of a soft thermal component that may correspond to emission from hot polar cap regions or to cooling emission from a light-element atmosphere over the entire star. An extrapolation of the non-thermal spectrum of RX J0007.0+7302 to γ-ray energies would yield a flux consistent with that of EGRET source 3EG J0010+7309, supporting the proposition that these sources could be related. A Chandra image of the central X-ray source RX J0007.0+7303 (Halpern et al. 2004) reveals a point source, a compact nebula, and a bent jet, all of which are characteristic of energetic, rotation-powered pulsars.
Optical observations were carried out by Brazier et al. (1998), with a 2.12-m telescope, but no object was found within the positional error box of the X-ray source. This allowed an upper limit to be set on the optical magnitude of any counterpart to the putative pulsar. Halpern et al. (2004) also obtained upper limits in the optical at the position of the point source, corresponding to an X-ray-to-optical flux ratio larger than 100. Neither a VLA image at 1425 MHz nor a deep pulsar search at 820 MHz using the NRAO Green Bank Telescope, revealed a radio pulsar counterpart The γ-ray source 3EG J0010+7309 is non-variable under the I and τ schemes, and has a hard spectral index of 1.85±0.10, compatible with those of the Vela pulsar. This source was also detected in the Second EGRET Catalog, but with a shifted position which made it coincide with a nearby AGN -which was at the time proposed as a possible counterpart (Nolan et al. 1996). This AGN is not coincident with the 3EG source and can not be considered a plausible counterpart any longer. Based on positional coincidence, on the hard spectral index, and on physical similarities between the Vela pulsar and RX J0007.0+7302, Brazier et al. (1998) proposed that the 3EG source and this X-ray source were related. For an assumed 1 sr beaming, the observed 100-2000 MeV flux corresponds to a luminosity of 4 × 10 33 erg s −1 , compatible with other γ-ray pulsar detections. A competing explanation to the γ-ray pulsar would be given if a massive and dense molecular cloud is found to be interacting with, or being overtaken by, the supernova remnant shock. The molecular environment of CTA 1 is analyzed below to further assess this possibility. As can be noted from the figure, the EGRET source seems too far from the SNR nearest rim. However, if some SNRs interact, as expected, with nearby massive clouds producing enhanced γ-ray emission through hadronic/Bremsstrahlung interactions, cases in which there is just a marginal coincidence between the SNRs and the centers of the EGRET sources should be also considered, since the peak γ-ray emissivity will likely be biased towards the adjacent cloud.
CTA 1: Molecular environment
As discussed in Dame et al. (1987), much of the local (closer than Perseus Arm) CO emission in the second quadrant of the Galaxy bifurcates into 2 velocity components, one centered near −12 km/s and the other close to 0 km/s. Dame et al. (1987) argue that the "−12 km/s clouds" are probably associated with a string of OB associations at 800 to 1000 pc from Earth. The clouds with velocities near 0 km/s were associated with the so-called "Lindblad Ring" at 300 pc. The long thin cloud partially overlapping CTA 1 (see Figure 1) has a mean velocity of ∼ −3 km/s, suggesting that it is part of the Lindblad Ring population. In fact, these clouds appear to be part of a very large loop of local clouds (all in the range −5 to 0 km/s). Grenier et al. (1989) proposed that this loop surrounds a 4 × 10 4 yr old Type I SNR 300 pc away. Thus, the velocity of the cloud near to CTA 1 suggests that it is part of this larger system, lying at only ∼300 pc.
The total H 2 mass (calculated from CO data) of the long thin cloud near CTA1 (in the region l = 116 o to 120 o , b = 8 o to 13 o , v = −15 to 5 km/s) is 4900 M ⊙ , assuming a distance of 300 pc. However, there is no evidence of interaction between the remnant and the cloud (what is consistent with the different most plausible distances to each of these objects). Figure 1b attached, for example, shows a latitude-velocity map integrated over the longitude extent of the cloud. There is little indication of accelerated gas at the latitude/velocity of the remnant, b = 9 o to 10 o and ∼ −15 km/s, respectively. The cloud may well be, then, in the foreground, and unrelated to CTA 1. In addition to this, there is no CO emission detected within the contours of the 3EG source, which also argues against a hadronic origin of the γ-ray source. Pineault et al.'s (1993) systemic velocity of −16 km/s for the HI shell is reasonable close to that of the "−12 km/s clouds", which in turn are probably associated with the OB associations Cep OB2, Cep OB3, Cep OB4, and Per OB3 at ∼ 800 pc. This might be suggesting that the latter is the most likely distance for CTA 1, although it remains uncertain. In any case, it seems that CTA 1 and 3EG J0010+7309 might be physically related only through the compact object left by the supernova explosion, RX J0007.0+7302. A better localization of the γray source with AGILE and GLAST will certainly test this suggestion. For the moment, our CO analysis shows that a molecular cloud origin of the γ-ray source 3EG J0010+7309 (for instance, through hadronic processes similar to the case of G347.3-0.5 proposed by Butt et al. 2001) is untenable. | 2014-10-01T00:00:00.000Z | 2004-06-14T00:00:00.000 | {
"year": 2004,
"sha1": "f942755df8b05f9f6876a8e43c240c9d7485c6d4",
"oa_license": "CCBYNCSA",
"oa_url": "https://ri.conicet.gov.ar/bitstream/11336/27800/5/CONICET_Digital_Nro.a2e65653-83b0-4a7b-8b42-6df5d431fe60_B.pdf",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "f942755df8b05f9f6876a8e43c240c9d7485c6d4",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
149629882 | pes2o/s2orc | v3-fos-license | PREVALENCE AND DETERMINANTS OF BEHAVIORAL PROBLEMS AMONG ADOLESCENTS LIVING IN ORPHANAGE FACILITIES OF DISTRICT PESHAWAR , KHYBER PAKHTUNKHWA , PAKISTAN
hildhood is the most important Cand decisive period of human life and a healthy childhood is crucial for 1 future growth and development. Alongside developmental changes, the periods of childhood and adolescent are under the influence of social, emotional, physical and biological factors which also make it nerve-racking, and thus, making the children more susceptible to developing behavioral, psychological, 2,3 and emotional problems. Millions of orphans and abandoned children across the world are deprived of this crucial 4 phase of life. In the absence of the child's parents, grandparents or reluctant relatives not willing to take care of the child, orphanage act as an institution to provide care and support for these 4 unfortunate children.
INTRODUCTION
hildhood is the most important Cand decisive period of human life and a healthy childhood is crucial for 1 future growth and development.Alongside developmental changes, the periods of childhood and adolescent are under the influence of social, emotional, physical and biological factors which also make it nerve-racking, and thus, making the children more susceptible to developing behavioral, psychological, 2,3 and emotional problems.Millions of orphans and abandoned children across the world are deprived of this crucial 4 phase of life.In the absence of the child's parents, grandparents or reluctant relatives not willing to take care of the child, orphanage act as an institution to provide care and support for these 4 unfortunate children.
The orphanages can provide a protected and positive substitute to offensive and unsecure family or community environments but they can't give 5 individualized and family fostering.
Orphanage children are exclusively susceptible to many psychosocial threats of institutional care compared to 6 children living with their families.Despite having a proper shelter, children in institutional care found difficulties regarding reading, language, attention, poor physical health, poverty, attachment disorder, mental health 6,7 difficulty and inadequate social skills.Some circumstances could increase the risk for behavioral and developmental problems, for example, abuse, neglect, and malnutrition can cause effects for 8 long-lasting.
Behavioral problem is defined as emotional problem, conduct problem, hyperactivity, peer problem and lack of 9 pro-social behavior.An emotional problem affects a person's ability to be happy and control their emotions.Conduct problem is a dreary and determined example of conduct in adolescents characterized by serious rule violati ons, aggressive behavior, disobedience and dishonesty or theft.Hyperactivity is a state of excess activity that is manifested by symptoms such as excessive movement, fidgeting, jumpiness or nervousness.Peer problem involves issues in getting along with peer and experiencing negative feelings or dislikeness towards peer.Pro social behavior is a social conduct that advantages other individuals or society in general, for example, co-working, sharing, helping, Similarly, the time period of institutionalization and age at parents' demise were statistically not significant.However, there was significant association between parent's status and behavioral problems as shown below in Table III.problems.The majority of adolescents in orphanages were of age group 10 to 12.9 years and behavioral problem were also prevalent in this age group, as this is the time period of early adolescence which consists of developmental transition from childhood to adulthood, involving multiple physical, intellectual, personality, and social developmental changes.
DISCUSSION
A statistically significant association was also found between gender and behavioral problems, the prevalence of behavioral problems was highest in females 57.7% as compared to males 32.0% which is in contrast to other studies which reported highest prevalence of behavioral problems among males, except in one study in 9,13,27 which females dominated.The reason for this need to be explored in future studies, yet the possible reason could be that females are more sensitive as compared to males.
Type of orphanage and behavioral problems were significantly associated and it was found that the orphanages run by government had highest prevalence of total behavioral problems 92.5% than orphanages run by private sectors 26.6%.The facilities provided to the children in private orphanages were better in terms of residence, education, health and extra-curricular activities than those of government orphanages.The private orphanages had invested more funds to provide better quality infrastructure.They had separate spaces for living, dining and extracurricular activities.They also focus more on the children's health by having regular monthly checkups.The children were admitted in high standard private schools to provide good quality education.On the other hand, the government orphanages lack funding and attention.Their infrastructure was inadequate to provide proper space for different requirements of life.They were utilizing common space for children's living and schooling.There was no proper schedule of monitoring health of the children.Similarly, in Karachi the prevalence of abnormal behavior on foster mother-rated overall SDQ was also higher in children in conventional orphanages as compared 26 to that in SOS children.Time period of institutionalization had insignificant association with behavioral problems.However, this was in contrast to the study in which length of stay had a positive association with the behavioral 13 problems and another study in which orphanages were found to be protective 26 against behavioral problems.
No significant association was found between Age at parent's demise and total behavioral problem.The results contradicts with another study conducted in orphanages in Dhaka city where the prevalence of psychiatric disorder was highest among the children whose parents died at or before 4 years of age than those whose parents died after completion of four A high burden of behavioral problems was found among adolescents living in orphanages of District Peshawar, Pakistan.The determinants of behavioral problems found in this study, may be used as the target of intervention to reduce behavioral problems among adolescents living in orphanages.
We are grateful to Khurram Sheraz & Aliya Durrani for their kind help during preparation, proof reading and submission of the manuscript.
20 years.
In another study conducted in Bangladesh, behavioral problems were found more among the children who KMUJ 2018, Vol. 10 No.2 98 PREVALENCE AND DETERMINANTS OF BEHAVIORAL PROBLEMS AMONG ADOLESCENTS LIVING IN ORPHANAGE FACILITIES OF DISTRICT PESHAWAR, KHYBER PAKHTUNKHWA, PAKISTAN
2 PREVALENCE AND DETERMINANTS OF BEHAVIORAL PROBLEMS AMONG ADOLESCENTS LIVING IN ORPHANAGE FACILITIES OF DISTRICT PESHAWAR, KHYBER PAKHTUNKHWA, PAKISTAN
For data analysis, SPSS version 16 was used.Mean and SD were calculated for Age.Frequencies and percentages were 96 KMUJ 2018, Vol. 10 No.
2 PREVALENCE AND DETERMINANTS OF BEHAVIORAL PROBLEMS AMONG ADOLESCENTS LIVING IN ORPHANAGE FACILITIES OF DISTRICT PESHAWAR, KHYBER PAKHTUNKHWA, PAKISTAN
Among the determinants of behavioral problems age had significant association with the total behavioral problems.The age group 10 to 12.9 had the highest KMUJ 2018, Vol. 10 No. | 2019-05-12T14:24:00.142Z | 2018-06-30T00:00:00.000 | {
"year": 2018,
"sha1": "9f4a201ad17f61d8aa24ca85c87da4ab822b3547",
"oa_license": "CCBYNC",
"oa_url": "https://www.kmuj.kmu.edu.pk/article/download/17587/pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "9f4a201ad17f61d8aa24ca85c87da4ab822b3547",
"s2fieldsofstudy": [
"Sociology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
235296815 | pes2o/s2orc | v3-fos-license | Estimation of the Effect of Accelerating New Bone Formation of High and Low Molecular Weight Hyaluronic Acid Hybrid: An Animal Study
Osteoconduction is an important consideration for fabricating bio-active materials for bone regeneration. For years, hydroxyapatite and β-calcium triphosphate (β-TCP) have been used to develop bone grafts for treating bone defects. However, this material can be difficult to handle due to filling material sagging. High molecular weight hyaluronic acid (H-HA) can be used as a carrier to address this problem and improve operability. However, the effect of H-HA on bone formation is still controversial. In this study, low molecular weight hyaluronic acid (L-HA) was fabricated using gamma-ray irradiation. The viscoelastic properties and chemical structure of the fabricated hybrids were evaluated by a rheological analysis nuclear magnetic resonance (NMR) spectrum. The L-MH was mixed with H-HA to produce H-HA/L-HA hybrids at ratios of 80:20, 50:50 and 20:80 (w/w). These HA hybrids were then combined with hydroxyapatite and β-TCP to create a novel bone graft composite. For animal study, artificial bone defects were prepared in rabbit femurs. After 12 weeks of healing, the rabbits were scarified, and the healing statuses were observed and evaluated through micro-computer tomography (CT) and tissue histological images. Our viscoelastic analysis showed that an HA hybrid consisting 20% H-HA is sufficient to maintain elasticity; however, the addition of L-HA dramatically decreases the dynamic viscosity of the HA hybrid. Micro-CT images showed that the new bone formations in the rabbit femur defect model treated with 50% and 80% L-HA were 1.47 (p < 0.05) and 2.26 (p < 0.01) times higher than samples filled with HA free bone graft. In addition, a similar tendency was observed in the results of HE staining. These results lead us to suggest that the material with an H-HA/L-HA ratio of 50:50 exhibited acceptable viscosity and significant new bone formation. Thus, it is reasonable to suggest that it may be a potential candidate to serve as a supporting system for improving the operability of granular bone grafts and enhancing new bone formations.
Introduction
For years, bone tissue engineering has been faced with the challenge of reducing the amount of bone graft material used without affecting treatment efficiency. Indeed, a matrix carrier is necessary in some applications to maintain bone graft material without loss during an operation; however, the carrier used for bone graft argumentation will reduce the absolute amount of bone graft, which can result in a decrease in the effectiveness of the artificial material [1]. Thus, improving material handling ability while preserving bone graft effectiveness remains a challenge.
Hyaluronic acid (HA) is a polysaccharide consisting of alternating residues of Dglucuronic acid and N-acetylglucosamine [2] and, as a basic component of extracellular matrix, displays high biocompatibility [3]. Previous studies have reported that HA can be used in mixture with bone graft material to modify the surface of artificial bone and enhance bone cell migration and growth [4,5]. One animal study has shown that coating HA on the surface of titanium implants can increase bone formation efficiency at the implant/bone interface [2]. Several investigations have used HA as a carrier and when combined with various bone grafts for bone augmentation. Nguyen and Lee (2014), for example, fabricated a bone substitute consisting of HA-gelatin hydrogel and biphasic calcium phosphate [6], and found that this composite provided excellent cellular response and could enhance the mechanical strength of cancellous bones and increase their bearing ability when subjected to load. A similar phenomenon was reported in tests of a mixture of HA-based matrix and collagenated heterologous bone graft [7]. In previous studies, HA has been reported to be an excellent carrier for sinus augmentation without reducing the clinical effectiveness of the allograft [6,8].
However, the effect of HA on bone formation is controversial in the case of combining hydroxyapatite/β-calcium triphosphate (β-TCP) with HA. For example, while Elkarargy (2013) found the addition of HA increases osteoconduction efficiency compared to samples without HA [9], Aguado et al.'s report (2014) indicated that using linear hyaluronic acid did not result in healing of the grafted area, and the amount of formed bone was not significantly higher in samples with HA than with β-TCP granules alone [10]. This may be due to the major role of HA in the HA-hydroxyapatite/β-TCP system only being able to act as a structural enforcement of the surrounding environment [11].
The above investigations made use of high molecular weight HA (H-HA), which can act as a physical scaffold for cell migration but cannot act as a ligand to directly bind to receptors on cell surfaces [12]. Unlike the controversial effects of H-HA on cell proliferation and differentiation [13,14], low molecular weight HA (L-HA) is well known to provide positive effects for cell proliferation and differentiation [14][15][16]. In the initial stage of wound healing, HA with a molecular weight reaching 2000 kDa only accumulates in the extracellular matrix and combines with fibrinogen to form a clot. However, L-HA with molecular weight ranging from 80 to 800 kDa influences the inflammatory response and activates macrophages [17], resulting in accelerated wound repair [18,19]. Although previous in vitro cellular studies have reported that L-HA exhibits a positive effect on bone healing, the effect of L-HA combined with hydroxyapatite/β-TCP bone graft on the healing of bone defects is still unclear [15,20,21]. In addition, it has been reported that reducing HA molecular weight results in lowered viscosity, which may cause a loss of intended operational effect. Accordingly, we fabricated H-HA/L-HA hybrid composites at different mixing ratios to be used as a carrier to release L-HA and maintain hydroxyapatite/β-TCP bone grafts for this study, and evaluated their performances using an animal model.
Preparation of H-HA/L-HA Hybrids
The molecular weight of H-HA used in this study was measured at 3000 kDa. The L-HA was prepared according to a previous method [19]. Briefly, γ-rays from a cobalt 60 irradiator (Point Source, AECL, IR-79, Nordion, Ottawa, ON, Canada) were used to destroy the H-HA structure using a continuous 1 kGy/h dose of radiation at 22 • C for 20 h. After irradiation, the molecular weight of low-molecular-weight HA was 200 kDa [19]. HA hybrids were fabricated by mixing H-HA and prepared L-HA at ratios of 80:20, 50:50 and 20:80 (w/w). Briefly, 1 g of H-LA and L-HA mixture was added to 3.5 mL phosphate buffer solution (PBS). After gently stirring for 2 h, the hybrid was freeze-dried and stored in a moisture-proof environment for further use.
Physicochemical Properties Tests of the H-HA/L-HA Hybrid
The chemical structure of the H-HA/L-HA hybrids was evaluated using 1 H nuclear magnetic resonance (NMR) spectrums obtained from a 500 MHz NMR spectrometer (DRX500 Avance, Bruker BioSpin GmbH, Rheinstetten, Germany). The measurements were performed at 27 • C. D 2 O was used as the solvent in all NMR experiments. The viscoelastic properties of the prepared HA hybrids were determined using a rheologic measurement device (Anton Paar MCR 302 rheometer, Anton Paar, Graz, Austria). The rheometer was equipped with a parallel plate with a plate diameter of 25 mm and gap of 1.0 mm at 25 • C. This device was calibrated according to a previous experiment [22]. Briefly, samples to be tested were prepared as a 10 mg/mL solution. At 30 min post-loading, the complex moduli (G*) of the HA hybrid were measured using a frequency sweep from 10 to 100 Hz. HA dynamic viscosity (η*) was recorded as a function of the shear rate range from 0.005 to 10 (1/s).
Animal Experiment
In this study, six New Zealand white rabbits (average weight 3.0-3.6 kg) were used as test samples. These white rabbits were raised in a stable environment at a temperature of 25 • C and a 50% humidity and provided with solid food and water. All animal procedures were performed according to the National Research Council's Guide for the Care and Use of Laboratory Animals guidelines and protocols approved by the Institutional Animal Care and Use Committee of the National Defense Medical Center, Taipei, Taiwan (IACUC-17-236). Since the properties of 100% H-HA and L-HA were not the focus of this study and, according to comments from the committee and in acceptance of the guiding principles of Declaration of Helsinki, which supports a reduction in laboratory animal use, only H-HA/L-HA hybrids with mixing ratios of 80:20, 50:50 and 20:80 were tested in the following experiments.
Surgical procedures were performed under sterile conditions. Before surgery, general anesthesia was achieved with an intramuscular injection of tiletamine-zolazepam (Zoletil 50) at a dosage rate of 15 mg/kg. After each rabbit was deeply anesthetized, the surgical site was shaved and disinfected with povidone iodine, and the skin cut to expose the lateral femoral condyle. According to previous studies, cylindrical defects 5 mm in diameter and 10 mm in length were drilled in each rabbit's left and right leg (Figure 1a-c) under saline cooling conditions [23,24]. The filling material was prepared by mixing 1.0 g prepared HA hybrid containing various concentrations of L-HA and 1.0 g Hydroxyapatite/β-TCP substitute (with a ratio of 60:40 w/w) (Figure 1d). For each artificial defect, 1.0 g of bone graft-HA hybrid was grafted (Figure 1e,f). After grafting, the periosteum was closed with absorbable inner and outer flap sutures (Figure 1g,h) (Vicryl ® 4.0, Ethicon, Somerville, NJ, USA). Postoperative antibiotics and analgesics were administered intramuscularly for three days. After 12 weeks of healing, the rabbits were euthanized by carbon dioxide asphyxiation under anesthesia (50 mg/mL Zoletil 50 at a dosage of 15 mg/kg). Bone tissues intramuscularly for three days. After 12 weeks of healing, the rabbits were euthanized by carbon dioxide asphyxiation under anesthesia (50 mg/mL Zoletil 50 at a dosage of 15 mg/kg). Bone tissues from the surgical site were collected and fixed in a 10% formaldehyde solution at pH 7.0. Artificial bone defect grafted with HA-free Hydroxyapatite/β-TCP substitute was set as the control. Three samples in each tested group were collected for further analysis.
Micro-CT Measurements
Femoral condyles containing the artificial defects were scanned using micro-computed tomography (micro-CT) (Skyscan 1076, SkyScan, Aartselaar, Belgium) with a 0.5mm aluminum filter at an energy level of 75 kV and current of 200 μA, with a pixel resolution of 18 μA. Data were analyzed using quantifying reconstructed three-dimensional images. According to previous studies, the percentage of new bone formation in each defect was quantitated by calculating the ratio of the bone volume (BV) to the total tissue volume (TV) in the defect holes.
Histological and Histomorphometrical Evaluation
In order to observe changes in bone growth among the artificial defects treated with different filling samples, bone samples were evaluated using histological analyses. To decalcify the sample, bone blocks were immersed in decalcifier for 4 weeks. Samples were then dehydrated in alcohol with an increasing gradient concentration (60-100%), embedded in paraffin and cut into sections with a thickness of 5 μm. Finally, bone tissue specimens were stained with hematoxylin and eosin. Histological images were acquired using a microscope slide scanner (OPTIKA, Ponteranica, Italia). Areas of new bone formation, residual bone substitute and non-mineralized tissue in the defect were quantitatively analyzed using image analysis software (ImageJ, National Institutes of Health, Bethesda, MD, USA).
Micro-CT Measurements
Femoral condyles containing the artificial defects were scanned using micro-computed tomography (micro-CT) (Skyscan 1076, SkyScan, Aartselaar, Belgium) with a 0.5-mm aluminum filter at an energy level of 75 kV and current of 200 µA, with a pixel resolution of 18 µA. Data were analyzed using quantifying reconstructed three-dimensional images. According to previous studies, the percentage of new bone formation in each defect was quantitated by calculating the ratio of the bone volume (BV) to the total tissue volume (TV) in the defect holes.
Histological and Histomorphometrical Evaluation
In order to observe changes in bone growth among the artificial defects treated with different filling samples, bone samples were evaluated using histological analyses. To decalcify the sample, bone blocks were immersed in decalcifier for 4 weeks. Samples were then dehydrated in alcohol with an increasing gradient concentration (60-100%), embedded in paraffin and cut into sections with a thickness of 5 µm. Finally, bone tissue specimens were stained with hematoxylin and eosin. Histological images were acquired using a microscope slide scanner (OPTIKA, Ponteranica, Italia). Areas of new bone formation, residual bone substitute and non-mineralized tissue in the defect were quantitatively analyzed using image analysis software (ImageJ, National Institutes of Health, Bethesda, MD, USA).
Statistical Analysis
Mean values and standard deviations of the percentages of newly formed bone, residual bone substitute and non-mineralized tissue quantified using micro-CT, and histological images were calculated and presented. Differences between samples with various amounts of L-HA were investigated using one-way analysis of variance (ANOVA) with Duncan's post hoc test (SPSS Inc., Chicago, IL, USA). For all tests, statistical significance was defined as a p value less than 0.05.
Results and Discussions
Characterization Results for Fe 3
O 4 NPs
HA synthesized using common methodologies is of high molecular weight. To obtain L-HA of a specific molecular weight, both physical methods (ultrasound, ozone, electron beam, gamma-ray radiation and heat treatment) and chemical methods (enzyme and acid degradation) are used to destroy H-HA's main structure [25][26][27][28]. However, when the main bonding chain of H-HA is destroyed, its viscoelasticity and water retention properties undergo significant changes. In this study, 1 H NMR spectra were used to examine the chemical structure of the fabricated H-HA/L-HA hybrids. We found no observable changes in chemical structure among these hybrids at different H-HA/L-HA mixing ratios (Figure 2). These results are supported by a previous report which indicated that γ-ray irradiation preserves HA's fundamental structure [19,29].
Statistical Analysis
Mean values and standard deviations of the percentages of newly formed bone, residual bone substitute and non-mineralized tissue quantified using micro-CT, and histological images were calculated and presented. Differences between samples with various amounts of L-HA were investigated using one-way analysis of variance (ANOVA) with Duncan's post hoc test (SPSS Inc., Chicago, IL, USA). For all tests, statistical significance was defined as a p value less than 0.05.
Results and Discussions
Characterization Results for Fe3O4 NPs HA synthesized using common methodologies is of high molecular weight. To obtain L-HA of a specific molecular weight, both physical methods (ultrasound, ozone, electron beam, gamma-ray radiation and heat treatment) and chemical methods (enzyme and acid degradation) are used to destroy H-HA's main structure [25][26][27][28]. However, when the main bonding chain of H-HA is destroyed, its viscoelasticity and water retention properties undergo significant changes. In this study, 1 H NMR spectra were used to examine the chemical structure of the fabricated H-HA/L-HA hybrids. We found no observable changes in chemical structure among these hybrids at different H-HA/L-HA mixing ratios ( Figure 2). These results are supported by a previous report which indicated that γ-ray irradiation preserves HA's fundamental structure [19,29]. Viscoelasticity is an important property of HA that allows the use of HA in various medical applications. Our rheological tests showed that H-HA/L-HA with different mixing ratios has almost the same complex modulus value at frequencies higher than 10 Hz (Figure 3a), meaning that the H-HA/L-HA hybrids used in this study retained their elasticity at this frequency. This result is supported by a previous study which also reported that hybrid complexes created by mixing high and low molecular weight HA maintained their moduli and were suitable for the treatment of osteoarthritis [30]. Xue and coauthors (2020) also fabricated H-HA/L-HA hybrids to evaluate their potential applications in regenerative medicine and tissue engineering [31]. Their results showed that, when the mixing ratio of H-HA:L-HA reached 80:20, the moduli reduced dramatically. This may have been due to cross-linking of the H-HA and L-HA used in their experiment. In the present study, NMR spectra readings showed no observable chemical shift peaks at 1.5 and 1.8 (Figure 2), which provides evidence that the fabricated hybrids were non-cross-linked HA [31]. Complex modulus (root mean square of storage modulus and lost modulus) is a characteristic of the overall ability to resist deformation when a dynamic force is applied [22,32]. Results shown in Figure 3a indicate that 20% H-HA is sufficient to maintain the Viscoelasticity is an important property of HA that allows the use of HA in various medical applications. Our rheological tests showed that H-HA/L-HA with different mixing ratios has almost the same complex modulus value at frequencies higher than 10 Hz (Figure 3a), meaning that the H-HA/L-HA hybrids used in this study retained their elasticity at this frequency. This result is supported by a previous study which also reported that hybrid complexes created by mixing high and low molecular weight HA maintained their moduli and were suitable for the treatment of osteoarthritis [30]. Xue and coauthors (2020) also fabricated H-HA/L-HA hybrids to evaluate their potential applications in regenerative medicine and tissue engineering [31]. Their results showed that, when the mixing ratio of H-HA:L-HA reached 80:20, the moduli reduced dramatically. This may have been due to cross-linking of the H-HA and L-HA used in their experiment. In the present study, NMR spectra readings showed no observable chemical shift peaks at 1.5 and 1.8 (Figure 2), which provides evidence that the fabricated hybrids were non-cross-linked HA [31]. Complex modulus (root mean square of storage modulus and lost modulus) is a characteristic of the overall ability to resist deformation when a dynamic force is applied [22,32]. Results shown in Figure 3a indicate that 20% H-HA is sufficient to maintain the H-HA/L-HA hybrid's elastic properties. Since bone tissues are subjected to a stress environment during daily life, the excellent elastic property shown in Figure 3a suggests that the H-HA/L-HA hybrid used in this study has good applicability in orthopedics.
H-HA/L-HA hybrid's elastic properties. Since bone tissues are subjected to a stress environment during daily life, the excellent elastic property shown in Figure 3a suggests that the H-HA/L-HA hybrid used in this study has good applicability in orthopedics. This result consistent with a previous report that indicated the major role of H-HA in graft material is to maintain a stable shape rather than directly affect bone regeneration [10]. However, quantitative results show that bone defects treated with bone graft and L-HA/H-HA hybrids containing higher L-HA amounts exhibit better bone reparative processes (Figure 4). The new bone formation of bone defects treated with the bone graft-HA hybrid complex with 50% and 80% L-HA in L-HA/H-HA hybrids significantly increased new bone formation ( Figure 5a) and decreased non-mineralized tissue (Figure 5c). New bone formation in bone defects treated with 50 H-HA/50 LHA and 80 H-HA/20 L-HA hybrid complexes were 1.47 (p < 0.05) and 2.26 (p < 0.05) times higher than defects treated with HA-free bone graft (Figure 5a). In addition, non-mineralized tissue of these filling materials was decreased 1.62 fold for 50 H-HA/50 L-HA and 1.21 fold for 80 H-HA/20 L-HA hybrids (p < 0.05) (Figure 5c). A similar conclusion can also be reached from histological assessment ( Figure 6).
H-HA/L-HA hybrid's elastic properties. Since bone tissues are subjected to a stress environment during daily life, the excellent elastic property shown in Figure 3a suggests that the H-HA/L-HA hybrid used in this study has good applicability in orthopedics. This result consistent with a previous report that indicated the major role of H-HA in graft material is to maintain a stable shape rather than directly affect bone regeneration [10]. However, quantitative results show that bone defects treated with bone graft and L-HA/H-HA hybrids containing higher L-HA amounts exhibit better bone reparative processes (Figure 4). The new bone formation of bone defects treated with the bone graft-HA hybrid complex with 50% and 80% L-HA in L-HA/H-HA hybrids significantly increased new bone formation ( Figure 5a) and decreased non-mineralized tissue (Figure 5c). New bone formation in bone defects treated with 50 H-HA/50 LHA and 80 H-HA/20 L-HA hybrid complexes were 1.47 (p < 0.05) and 2.26 (p < 0.05) times higher than defects treated with HA-free bone graft (Figure 5a). In addition, non-mineralized tissue of these filling materials was decreased 1.62 fold for 50 H-HA/50 L-HA and 1.21 fold for 80 H-HA/20 L-HA hybrids (p < 0.05) (Figure 5c). A similar conclusion can also be reached from histological assessment ( Figure 6). Data are expressed as mean ± standard deviation. The symbol * denotes a significant difference from the "bone graft with HA free" group, + denotes a significant difference from the "bone graft with 80 H-HA/20 L-HA" group, # denotes a significant difference from the "bone graft with 50 H-HA/50 L-HA" group. Significance (p < 0.05) was determined using the post hoc Duncan test after one-way ANOVA. substitute and (c) non-mineralized tissue. Data are expressed as mean ± standard deviation. The symbol * denotes a significant difference from the "bone graft with HA free" group, + denotes a significant difference from the "bone graft with 80 H-HA/20 L-HA" group, # denotes a significant difference from the "bone graft with 50 H-HA/50 L-HA" group. Significance (p < 0.05) was determined using the post hoc Duncan test after one-way ANOVA. samples filled with HA-free bone graft (19.67 ± 5.21%) (Figure 7a). The residual bone substitute (33.38 ± 3.31%) and non-mineralized tissue (25.38 ± 3.21%) of the bone graft mixed with 80 H-HA/20 L-HA hybrids were also significantly decreased compared to the HA-free bone graft group (p < 0.05) (Figure 7b,c). These histological analyses also indicate that artificial bone defects filled with more L-HA results in a better bone healing process. Combined with the CT results (Figures 4 and 5) and histomorphometric analysis (Figures 6 and 7), it is r0easonable to conclude that the effect of HA on bone healing is strongly affected by its molecular weight. These results confirm that L-HA provides an osteo-regenerative effect on bone reparation. Aguado et al. (2014) also investigated the use of HA as an aqueous binder of β-TCP granules [10]. After implanting an HA/β-TCP composite mixture into artificial holes drilled in the femoral condyle of rabbit legs, they also found that β-TCP granules mixed with HA induce an increase in bone apposition. They concluded that HA's role appears to be as a vehicle only as it does not interfere with bone remodeling; however, this is because only H-HA was used in their study. The effects of high molecular weight HA on cell proliferation and differentiation remain controversial [13,33]. and 42.00 ± 4.72%, respectively. These values are significantly higher (p < 0.05) than samples filled with HA-free bone graft (19.67 ± 5.21%) (Figure 7a). The residual bone substitute (33.38 ± 3.31%) and non-mineralized tissue (25.38 ± 3.21%) of the bone graft mixed with 80 H-HA/20 L-HA hybrids were also significantly decreased compared to the HA-free bone graft group (p < 0.05) (Figure 7b,c). These histological analyses also indicate that artificial bone defects filled with more L-HA results in a better bone healing process. Combined with the CT results (Figures 4 and 5) and histomorphometric analysis (Figures 6 and 7), it is r0easonable to conclude that the effect of HA on bone healing is strongly affected by its molecular weight. These results confirm that L-HA provides an osteo-regenerative effect on bone reparation. Aguado et al. (2014) also investigated the use of HA as an aqueous binder of β-TCP granules [10]. After implanting an HA/β-TCP composite mixture into artificial holes drilled in the femoral condyle of rabbit legs, they also found that β-TCP granules mixed with HA induce an increase in bone apposition. They concluded that HA's role appears to be as a vehicle only as it does not interfere with bone remodeling; however, this is because only H-HA was used in their study. The effects of high molecular weight HA on cell proliferation and differentiation remain controversial [13,33]. The prepared 80 H-HA/20 L-HA hybrid shows a typical viscosity curve (Figure 3b). The dynamic viscosity of this HA hybrid depends on the shear rate as a non-Newtonian liquid at 0.005-10 s −1 [34]. When the amount of L-HA was increased to 80%, a shear thinning phenomenon occurred as seen in the dramatically decreased slope (Figure 3b). When the strain rate reached 2.8 s −1 , the viscosity of this 20 H-80 L HA hybrid increased with the increasing strain rate and exhibited shear-thickening. This hardening phenomenon can provide a shock-damping function and protective effect when a sudden high-load impact is applied [19]. Although the bone graft with 20 H-HA/80 L-HA hybrid demonstrated almost double the new bone formation compared to the HA-free sample (Figures 5a and 7a), the sharp decrease in viscosity (Figure 3b) reduced its adhesive properties and may The prepared 80 H-HA/20 L-HA hybrid shows a typical viscosity curve (Figure 3b). The dynamic viscosity of this HA hybrid depends on the shear rate as a non-Newtonian liquid at 0.005-10 s −1 [34]. When the amount of L-HA was increased to 80%, a shear thinning phenomenon occurred as seen in the dramatically decreased slope (Figure 3b). When the strain rate reached 2.8 s −1 , the viscosity of this 20 H-80 L HA hybrid increased with the increasing strain rate and exhibited shear-thickening. This hardening phenomenon can provide a shock-damping function and protective effect when a sudden high-load impact is applied [19]. Although the bone graft with 20 H-HA/80 L-HA hybrid demonstrated almost double the new bone formation compared to the HA-free sample (Figures 5a and 7a), the sharp decrease in viscosity (Figure 3b) reduced its adhesive properties and may limit the applications of this material. For example, guided bone regeneration (GBR) is an important dental surgery that regenerates bone mass and increases bone width for healing after artificial dental implant procedures. For years, the most commonly used material for GBR has Polymers 2021, 13, 1708 9 of 11 been a mixture of hydroxyapatite and β-calcium triphosphate (β-TCP) [35][36][37]. However, bone grafts composed of these two particles only are difficult to handle at posterior maxilla due to sagging of the filling material [38,39]. Therefore, maintaining adequate viscosity for operability and retaining the osteo-regenerative efficiency of fabricated bone graft composites has become an issue for investigation. It has been reported that when H-HA and L-HA are mixed at a ratio of 1:1, their viscosity and in vitro resistance to enzymatic hydrolysis can be improved [31,40]. Interestingly, combined data seen in Figures 3b, 4c and 6c indicate that the fabricated bone graft composite containing the 50/50 (w/w) H-HA/L-HA hybrid exhibited acceptable viscosity and significant new bone formation. This material may be a potential candidate for sinus lift augmentation of the posterior maxilla.
The most important finding of this study is a confirmation that L-HA can promote new bone formation and decrease the percentage of non-mineralized tissue during the healing process (Figures 4-7). The mechanism of this osteo-regenerative effect was not assessed in this study but has been reported by several investigations which indicate that the effect of smaller hyaluronan molecules (at molecular weights ranging from 5 to 20 kDa) on regeneration is due to inducing cytokine and inflammatory responses at an early stage [17,41]. However, an inflammatory response was not observed when current bone graft composites were implanted into the bone defect ( Figure 6), which may be due to the molecular weight of L-HA used in this study being about 200 kDa [19]. For HA with molecular weights in the 50-200 kDa range, tissue regeneration effects due to stem cell differentiation [21] and cell proliferation [16] were reported. Ariyoshi et al. (2005) indicated that L-HA promotes bone tissue engineering through enhancing the interaction of RANKL and RANK, which activates the signal transduction pathway involved in osteoclast differentiation [20]. However, besides the biological effects mentioned above, physical factors should also be taken into account. In this study, a new hypothesis was proposed to explain the positive effect of H-HA/L-HA hybrids on bone healing based on physical properties. The high elasticity (Figure 3a) of the current H-HA/L-HA hybrid allows it to exhibit a high level of resistance to hyaluronidase [17,30,31,42]. Thus, bone graft composites with an L-HA proportion above 50% provide a niche to prolong the viscosupplementation and bioactive effects in tissue engineering. In addition, low viscosity allows bone graft composites to exhibit two other benefits for bone regeneration. The first benefit is that L-HA can easily be released from the hybrid and allow bone regeneration. As a high viscosity environment may hamper cell movement, the second benefit is that reducing viscosity while maintaining elasticity the H-HA/L-HA hybrid can change the cell-area mechanical environment and allow cells increased mobility [17].
In conclusion, a novel composite mixed with H-HA/L-HA hybrid and hydroxyapatite/β-TCP can improve the osteoconductive properties of bone grafts. The results present in this study suggest that a 1:1 H-HA/L-HA hybrid mixture can serve as a supporting system to improve the operability of bone grafts and enhance new bone formation. | 2021-06-03T06:17:23.905Z | 2021-05-24T00:00:00.000 | {
"year": 2021,
"sha1": "6ff0460a41db5a327aaa7b50adc18192b74cf35e",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2073-4360/13/11/1708/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "f4e47d9047870c77e12d4888c6db7c019cbdcbac",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
224811565 | pes2o/s2orc | v3-fos-license | Demonstration of a Filterless, Multi-Point, and Temperature-Independent Fiber Bragg Grating Dynamical Demodulator Using Pulse-Width Modulation
We demonstrated in this work a filterless, multi-point and temperature-independent FBG (fiber Bragg grating) dynamical demodulator using pulse-width-modulation (PWM). In this approach, the FBG interrogation system is composed of a tunable laser and a demodulator that is designed to detect the wavelength shift of the FBG sensor without any optical filter making it very suitable to be used in harsh environments. In this work, we applied the proposed method that uses the PWM technique for FBG sensors placed in high pressure and high-temperature environments. The proposed method was characterized in the laboratory using an FBG sensor modulated in a frequency of 6 Hz, with a 1 kHz sweeping frequency in the wavelength range from 1527 to 1534 nm. Also, the method was evaluated in a field test in an engine of a thermoelectric power plant.
Introduction
For practical reasons, the control electronics for many types of power machinery are usually placed inside or close to the operating environment of the engine. For example, the monitoring modules for dynamic pressure, temperature, and knock are placed directly in contact with the engine [1]. The electronic modules that have been developed for monitoring different types of engines (naval, thermoelectric, planes, military, and automotive) with specific emphasis on durability at high-temperature operation [2][3][4].
Monitoring the instantaneous combustion chamber pressure data is required for the closed-loop control of the fuel mass fraction burned in the engines [5,6]. The pressure sensors for this control must be durable and accurate. Using closed-loop control improves engine performance and reduces the emission of pollutants. The feedback system contains an intelligent data analysis system working with an ECU (engine control unit) to precisely dose the fuel quantity in each combustion cycle of each of the engine cylinders. In thermoelectric engines, the dynamical pressure has peaks higher than 250 bar, and the temperature in the combustion chamber is higher than 300 • C. Piezoelectric sensors used to measure the pressure of the combustion chamber currently are not durable when used continuously in high temperatures (>300 • C) [5]. Thus, this application needs a robust and trusty pressure sensor. Optical fiber sensors are a good alternative to electronic sensors in many engineering applications due to some intrinsic advantages, such as high temperature and chemical resistance and potential for long-lifetime operation. A fiber Bragg grating (FBG) sensors have flexible characteristics, low cost and are readily available [7]. Examples of FBG sensor applications include structural health monitoring in civil engineering [8,9], electric power systems [10,11], railways and roadways monitoring [12,13], in oil tanks monitoring and as chemical sensors [14,15], in biomechanics and in medicine [16,17]. FBG sensors can be fabricated using optical fibers made of different materials, such as glass [18], polymers [19], or sapphire [20].
Another issue of sensing system operation in a thermoelectric power plant is regarding the FBG interrogators. Although many commercial FBG interrogators modules can measure parameters running in frequencies higher than the engine's combustion cycles, they cannot be installed close to or inside the engines. The thermoelectric engine is high power machinery that dissipates a lot of heat. The temperature can vary depending on each part of the engine casing. For example, at the pressure monitoring point, the temperature in the case is higher than 300 • C. In the other parts that have water cooling, the temperature is lower. The average temperature in the machine room reaches 55 • C. Considering that the ECU and the FBG demodulator need to be installed close to the engine or its external body, the high temperatures will affect the operation of a supposed interrogation module. In the interrogation module that uses lasers, semiconductor optical amplifiers (SOAs) and Fabry-Perot filters [21], the cooler of the optoelectronics elements of the interrogator, will work in an excessive regime of operation, causing the device to fail in advance [22]. Passive devices such as WDM multiplexers, splitters, circulators also will be affected by the high temperatures. On the other hand, the current technology of electronic devices that can be used in a high-temperature environment is much more available than optoelectronic technology [2]. In addition, standard interrogators are not able to send the signals from all FBG pressure sensors obtained on each sweep to each engine ECU. Many commercial FBG interrogators record the measured data in an external computer in csv or text files. When a single data reading must be taken for control of each cylinder, the receiver of the optical sensing system could use, for example, the optical-edge-filtering technique [23] to detect the dynamic pressure. However, the FBG center wavelength shift due to temperature variation in the combustion chamber depends on the engine load [24]. Thus, the correct positioning of the optical edge filter in real operation is critical to reproduce the exact dynamic pressure behavior of the combustion chamber. Although some techniques were implemented to solve this problem [25], realizing this measurement in a real application is still a serious challenge and the complexity of the optical source and filter control of the edge filtering technique limits its use in this application. An alternative method based on the dispersion delay effect of a dispersion-compensating fiber (DCF) can also be used to convert the FBG wavelengths into the time domain [26], but its demodulator cannot be used near the engine.
In this work, we proposed an innovative filterless, multi-point, and temperature-independent FBG dynamical demodulator using the pulse-width-modulation (PWM) technique. PWM is a modulation technique that generates variable-width pulses to represent the amplitude of an analog input signal [27]. In [28], an interrogation system is presented based on pulse-modulation, that automatically recognized reflection signals of FBGs even when the FBGs are installed in an arbitrary order or at a long distance and affected by delays. This recognition technique was realized using pulse-modulating in the wavelength-swept laser. However, in contrast to [28], our proposed scheme has a tunable laser that sweeps a pre-set wavelength band where the FBG sensors work in continuous-wave (CW) mode. The demodulator using robust electronic devices can be used near the engine, near the pressure sensor and even integrated with the ECU to transform the wavelength variation-based signal, to a PWD signal and finally in an analog intensity signal compatible with the ECU input port. This approach can be used to measure many types of parameters using FBG sensors. In this work, we applied this method for FBG sensors placed in high pressure and high-temperature environments. The system was characterized in the laboratory using an FBG sensor modulated at a frequency of 6 Hz and a tunable laser with a 1 kHz sweeping frequency and wavelength range from 1527 to 1534 nm. A commercial FBG Sensors 2020, 20, 5825 3 of 16 interrogator was used to compare the results obtained in this application. Besides, it was evaluated in a field test in an engine of a thermoelectric power plant. Figure 1 shows the entire interrogation system where the FBG-PWM demodulator is used. The optical source for this system is a tunable laser with an appropriate sweep frequency. The sweep frequency must be higher than the maximum sensor frequency response to have enough sampled points during the measurements. In the diagram of Figure 1, the laser output can be divided for many demodulators in a power plant. Each splitter output is connected to the sensor using port 2 of an optical circulator. The FBG sensor in Figure 1 is used to measure dynamical pressure inside an engine of a thermoelectric plant engine where the temperature is very high (>400 • C) and is not stable. Port 3 of the circulator is used to connect the FBG reflected signal to the demodulator input. In the demodulator unit, the FBG wavelength-shifted signal follows first to the photodetector. In the photodetector, the optical signal is converted to electrical. Next, a transimpedance amplifier amplifies and clips the signal. Next, a flip-flop type D (FFD) digital circuit transforms it in a PWM signal. Next, low pass active filters are used at the output of the FFD to smooth the pulse train into a stable analog voltage. This analog voltage is the recovered FBG wavelength shifting signal. In other words, the variation of pulse width (PWM) is converted to an analog voltage directed related to the FBG wavelength shifting that in turn, is related to the original engine cycles modulation. This signal is sent to an ECU, which analyzes the signal and provides the correct commands to the engine in closed-loop control.
The Proposed FBG-PWM Demodulator
Sensors 2020, 20, x; doi: FOR PROOF www.mdpi.com/journal/sensors thermoelectric power plant. Figure 1 shows the entire interrogation system where the FBG-PWM demodulator is used. The optical source for this system is a tunable laser with an appropriate sweep frequency. The sweep frequency must be higher than the maximum sensor frequency response to have enough sampled points during the measurements. In the diagram of Figure 1, the laser output can be divided for many demodulators in a power plant. Each splitter output is connected to the sensor using port 2 of an optical circulator. The FBG sensor in Figure 1 is used to measure dynamical pressure inside an engine of a thermoelectric plant engine where the temperature is very high (>400 °C) and is not stable. Port 3 of the circulator is used to connect the FBG reflected signal to the demodulator input. In the demodulator unit, the FBG wavelength-shifted signal follows first to the photodetector. In the photodetector, the optical signal is converted to electrical. Next, a transimpedance amplifier amplifies and clips the signal. Next, a flip-flop type D (FFD) digital circuit transforms it in a PWM signal. Next, low pass active filters are used at the output of the FFD to smooth the pulse train into a stable analog voltage. This analog voltage is the recovered FBG wavelength shifting signal. In other words, the variation of pulse width (PWM) is converted to an analog voltage directed related to the FBG wavelength shifting that in turn, is related to the original engine cycles modulation. This signal is sent to an ECU, which analyzes the signal and provides the correct commands to the engine in closed-loop control.
The Proposed FBG-PWM Demodulator
In this proposed sensing system scheme, only the pressure sensors and the demodulators need to be placed close or on the case of the engine integrated with the ECU. The tunable laser mainly can be installed in a room with controlled temperature.
The key elements to implement the filterless, multi-point, and temperature-independent FBG dynamical demodulator are a tunable ring sweep laser [29] and the FFD digital circuit plus the active filter. The tunable source for this application can be fiber lasers based on semiconductor optical amplifiers or erbium-doped fiber using the Fabry-Perot filter. These lasers have narrow linewidth (<5 pm) and high output power (>10 dBm). In this proposed sensing system scheme, only the pressure sensors and the demodulators need to be placed close or on the case of the engine integrated with the ECU. The tunable laser mainly can be installed in a room with controlled temperature.
The key elements to implement the filterless, multi-point, and temperature-independent FBG dynamical demodulator are a tunable ring sweep laser [29] and the FFD digital circuit plus the active filter. The tunable source for this application can be fiber lasers based on semiconductor optical amplifiers or erbium-doped fiber using the Fabry-Perot filter. These lasers have narrow linewidth (<5 pm) and high output power (>10 dBm).
Once the laser output changes in wavelength overtime during the sweep, the optical wavelength variation of FBG in the engine becomes an electrical time variation in the demodulator. When the FFD receives the analog time variation signal of the FBG, it transforms this signal in a digital signal form with a pulse width variation. The pulse width variation has the information of dynamical pressure modulated in the FBG. Low pass active filters (two second-order Butterworth low pass filters) are used at the output of the PWM circuit (D) to smooth the pulse train into a stable analog voltage. The electrical-active-filter removes the digital modulation of the PWM signal recovering the original FBG modulated-signal. Then, this recovered signal is sent to the ECU. No synchronization signal is necessary for this system. Figure 2 shows the signal waveforms from the tunable laser to the demodulator output. In this figure, the signal E shows a typical engine combustion cycle. The signal-A shows the electrical sweep of the tunable laser. The signal B is the laser output intensity. Although the signal intensity in B is constant in time, the wavelength increases during the positive sweep slope, and it decreases during the negative sweep slope. The signal-C is the FBG electrical signal that is modulated by the dynamical wavelength variation. The signal-D is the digital output of the FFD circuit with pulse width modulation. The signal-E is the active filter output showing an example of a typical engine combustion cycle.
Sensors 2020, 20, x FOR PROOF 4 of 16 Sensors 2020, 20, x; doi: FOR PROOF www.mdpi.com/journal/sensors digital signal form with a pulse width variation. The pulse width variation has the information of dynamical pressure modulated in the FBG. Low pass active filters (two second-order Butterworth low pass filters) are used at the output of the PWM circuit (D) to smooth the pulse train into a stable analog voltage. The electrical-active-filter removes the digital modulation of the PWM signal recovering the original FBG modulated-signal. Then, this recovered signal is sent to the ECU. No synchronization signal is necessary for this system. Figure 2 shows the signal waveforms from the tunable laser to the demodulator output. In this figure, the signal E shows a typical engine combustion cycle. The signal-A shows the electrical sweep of the tunable laser. The signal B is the laser output intensity. Although the signal intensity in B is constant in time, the wavelength increases during the positive sweep slope, and it decreases during the negative sweep slope. The signal-C is the FBG electrical signal that is modulated by the dynamical wavelength variation. The signal-D is the digital output of the FFD circuit with pulse width modulation. The signal-E is the active filter output showing an example of a typical engine combustion cycle.
Demodulator Evaluation in Laboratory
To demonstrate this technique, we first tested the proposed system in the laboratory. Figure 3a shows the scheme to simulate the temperature changes and dynamic pressures on the FBG. An arbitrary waveform generator (BK4054B, B&K Precision Corporation, Yorba Linda, CA, USA) produced a typical 6 Hz engine combustion frequency. This generator waveform voltage was amplified by one piezoelectric driver connected to a piezoelectric transducer (model PK2FQP2- Figure 3b, (Thorlabs, Newton, NJ, USA) that stressed a polyimide coated FBG coupled into the transducer. The tunable laser used a triangle waveform frequency of 1 kHz to sweep the central wavelength from 1527 to 1534 nm. Figure 3c shows more details of the demodulator-electronic-circuit. The photodetector plus transimpedance amplifier has a bandwidth of 400 kHz and, the active filter was composed of two second-order Butterworth low pass filters with 50 Hz bandwidth. The FFD used was a 74HC74 digital circuit.
The type D flip flop circuit changes the output logic level in the rising edge of the signal in the clock input port. The FBG electrical signal from the transimpedance amplifier circuit is inserted into the type D flip-flop clock input port and, the inverted flip-flop output port is connected to the input port of the same chip. Thus, when the swept light is reflected by the FBG just the rising edge of the spectrum alters the flip-flop output. When the sweep occurs from shorter to longer wavelengths (positive slope of signal A), it is the left edge that changes the flip-flop output, and, when the sweep occurs from longer to shorter wavelengths (negative slope of signal A), the right edge changes the
Demodulator Evaluation in Laboratory
To demonstrate this technique, we first tested the proposed system in the laboratory. Figure 3a shows the scheme to simulate the temperature changes and dynamic pressures on the FBG. An arbitrary waveform generator (BK4054B, B&K Precision Corporation, Yorba Linda, CA, USA) produced a typical 6 Hz engine combustion frequency. This generator waveform voltage was amplified by one piezoelectric driver connected to a piezoelectric transducer (model PK2FQP2- Figure 3b, (Thorlabs, Newton, NJ, USA) that stressed a polyimide coated FBG coupled into the transducer. The tunable laser used a triangle waveform frequency of 1 kHz to sweep the central wavelength from 1527 to 1534 nm. Figure 3c shows more details of the demodulator-electronic-circuit. The photodetector plus transimpedance amplifier has a bandwidth of 400 kHz and, the active filter was composed of two second-order Butterworth low pass filters with 50 Hz bandwidth. The FFD used was a 74HC74 digital circuit.
Sensors 2020, 20, x FOR PROOF 5 of 16 Sensors 2020, 20, x; doi: FOR PROOF www.mdpi.com/journal/sensors flip-flop. Since the FBG is varying dynamically from longer to shorter wavelengths and vice versa, variations in rising and falling edge will create a PWM modulation. The output voltage in E can be express by (1) [30]: where VO is the averaged output voltage, δ is the duty cycle of the PWM waveform and VPWM is its amplitude. Considering that the tunable laser sweep time determines the total spectral range (BW) and the FBG produces a dynamical time variation signal proportional to dynamical wavelength shift (Δλ), δ can be written as: and VO can be written as: Therefore VO can be increased without reducing the noise-signal ratio reducing the spectral range of the sensing system. Figures 4a,b show the measured signals in C (red) and D (blue) respectively for FBG position in minimum (a) and maximum PZT displacement (b). Figures 5a,b are the PWM signals in D for two distinct PZT displacement amplitudes. These signals were measured in E using an oscilloscope. The type D flip flop circuit changes the output logic level in the rising edge of the signal in the clock input port. The FBG electrical signal from the transimpedance amplifier circuit is inserted into the type D flip-flop clock input port and, the inverted flip-flop output port is connected to the input port of the same chip. Thus, when the swept light is reflected by the FBG just the rising edge of the spectrum alters the flip-flop output. When the sweep occurs from shorter to longer wavelengths (positive slope of signal A), it is the left edge that changes the flip-flop output, and, when the sweep occurs from longer to shorter wavelengths (negative slope of signal A), the right edge changes the flip-flop. Since the FBG is varying dynamically from longer to shorter wavelengths and vice versa, variations in rising and falling edge will create a PWM modulation.
The output voltage in E can be express by (1) [30]: where V O is the averaged output voltage, δ is the duty cycle of the PWM waveform and V PWM is its amplitude. Considering that the tunable laser sweep time determines the total spectral range (BW) and the FBG produces a dynamical time variation signal proportional to dynamical wavelength shift (∆λ), δ can be written as: and V O can be written as: Therefore V O can be increased without reducing the noise-signal ratio reducing the spectral range of the sensing system. Figure 4a,b show the measured signals in C (red) and D (blue) respectively for FBG position in minimum (a) and maximum PZT displacement (b). Figure 5a,b are the PWM signals in D for two distinct PZT displacement amplitudes. These signals were measured in E using an oscilloscope. A critical evaluation regarding this proposed system refers to the characteristic of the PWM demodulated signal in terms of trustworthiness to the original FBG modulated signal. We compare in Figure 6 the modulation signal of the arbitrary-waveform-generator, the FBG signal measured in E, and the signal measured of FBG using a commercial FBG interrogator (100 Hz sweep frequency si155 Hyperion from Micron Optics, Atlanta, GA, USA). The commercial interrogator signal was obtained after post-processing. We can observe in Figure 6 that the demodulated signal in E is a good copy of the generator signal waveform; however, some noise can be observed in the signal base. We will comment on the noise source in Section 4. Also, we observed that the signal of the commercial interrogator has not enough sampled points to define all the events in an engine cycle curve. Next, we evaluated quantitatively, the demodulated signal characteristics. A critical evaluation regarding this proposed system refers to the characteristic of the PWM demodulated signal in terms of trustworthiness to the original FBG modulated signal. We compare in Figure 6 the modulation signal of the arbitrary-waveform-generator, the FBG signal measured in E, and the signal measured of FBG using a commercial FBG interrogator (100 Hz sweep frequency si155 Hyperion from Micron Optics, Atlanta, GA, USA). The commercial interrogator signal was obtained after post-processing. We can observe in Figure 6 that the demodulated signal in E is a good copy of the generator signal waveform; however, some noise can be observed in the signal base. We will comment on the noise source in Section 4. Also, we observed that the signal of the commercial interrogator has not enough sampled points to define all the events in an engine cycle curve. Next, we evaluated quantitatively, the demodulated signal characteristics. These wavelengths centers represent the FBG at different temperatures inside the engine. This wavelength range (750 pm) corresponds to a temperature variation of 57 °C for an FBG sensitivity of 13 pm/°C [24]. In this measurement procedure, we were limited in the wavelength range due to the PZT voltage limitation. The demodulator could measure the signal in the entire wavelength range determined by the tunable laser sweep (7 nm). We observed that the signals' linearities in the demodulator output are enough to reproduce the original characteristics of the FBG modulated signal with good quality. We attributed the variation in the offset of the curves to the PZT technical characteristics that are not stable with the time.
Demodulator Evaluation in a Field Test
The objective of the field test was to verify the performance of the PWM demodulator in terms of signal processing using a dynamical FBG pressure sensor installed in an environment with variable temperature. It was not the test proposal to test the demodulator itself at high temperatures in this project stage. The field tests take place in Centrais Elétricas da Paraíba (EPASA), which is a thermoelectric power plant. This thermoelectric power plant has an installed power of 340 MW, obtained from 40 model 3240 engines (MAN Diesel SE, Augsburg, Bavaria, Germany). The angular speed of each motor is 720 rpm, and heavy fuel oil (OCB1) is used to combustion engines. Each engine has 18 cylinders and uses a mechanical injection pump to control the fuel oil injection. This mechanism reduces the possibility of adjustments in the injected fuel volume and the same proportion limits the better management of the engines. In the field tests, the pressure sensors were connected in a pressure monitoring point available for each engine cylinder. To compare the pressure signals, we used again the commercial FBG interrogator and the data previously obtained from a reference sensor (model HLV 4.0 from Kistler Group, Winterthur, Switzerland), which is a standard sensor used in the thermoelectric power plant.
The thermoelectric power plant has a harsh environment. The internal average temperature in the machine room is around 55 °C, and close to the engines, it can be higher, limiting the continuous uses of standard electronic equipment. These wavelengths centers represent the FBG at different temperatures inside the engine. This wavelength range (750 pm) corresponds to a temperature variation of 57 • C for an FBG sensitivity of 13 pm/ • C [24]. In this measurement procedure, we were limited in the wavelength range due to the PZT voltage limitation. The demodulator could measure the signal in the entire wavelength range determined by the tunable laser sweep (7 nm). We observed that the signals' linearities in the demodulator output are enough to reproduce the original characteristics of the FBG modulated signal with good quality. We attributed the variation in the offset of the curves to the PZT technical characteristics that are not stable with the time.
Demodulator Evaluation in a Field Test
The objective of the field test was to verify the performance of the PWM demodulator in terms of signal processing using a dynamical FBG pressure sensor installed in an environment with variable temperature. It was not the test proposal to test the demodulator itself at high temperatures in this project stage. The field tests take place in Centrais Elétricas da Paraíba (EPASA), which is a thermoelectric power plant. This thermoelectric power plant has an installed power of 340 MW, obtained from 40 model 3240 engines (MAN Diesel SE, Augsburg, Bavaria, Germany). The angular speed of each motor is 720 rpm, and heavy fuel oil (OCB1) is used to combustion engines. Each engine has 18 cylinders and uses a mechanical injection pump to control the fuel oil injection. This mechanism reduces the possibility of adjustments in the injected fuel volume and the same proportion limits the better management of the engines. In the field tests, the pressure sensors were connected in a pressure monitoring point available for each engine cylinder. To compare the pressure signals, we used again the commercial FBG interrogator and the data previously obtained from a reference sensor (model HLV 4.0 from Kistler Group, Winterthur, Switzerland), which is a standard sensor used in the thermoelectric power plant.
The thermoelectric power plant has a harsh environment. The internal average temperature in the machine room is around 55 • C, and close to the engines, it can be higher, limiting the continuous uses of standard electronic equipment. Figure 8a shows the pressure sensor scheme [31]. The FBG was fixed in two points of a stainless steel substrate. A pre-stress was applied in FBG before the fixation. According to Figure 8a, the FBG is placed outside of the engine combustion chamber, and it is stressed by a mechanism composed of one 1-mm thickness membrane and one piston. Only one side of the membrane contacts the high-pressure and high-temperature gas inside the engine's combustion chamber. When the membrane is deformed, by the pressure, it moves a piston that stresses the FBG accordingly. In the field tests, we use FBG pressure sensors connected in a point of pressure monitoring available for each engine cylinder, as we can observe in Figure 8b. Figure 8a shows the pressure sensor scheme [31]. The FBG was fixed in two points of a stainless steel substrate. A pre-stress was applied in FBG before the fixation. According to Figure 8a, the FBG is placed outside of the engine combustion chamber, and it is stressed by a mechanism composed of one 1-mm thickness membrane and one piston. Only one side of the membrane contacts the high-pressure and high-temperature gas inside the engine's combustion chamber. When the membrane is deformed, by the pressure, it moves a piston that stresses the FBG accordingly. In the field tests, we use FBG pressure sensors connected in a point of pressure monitoring available for each engine cylinder, as we can observe in Figure 8b. Figure 8a shows the pressure sensor scheme [31]. The FBG was fixed in two points of a stainless steel substrate. A pre-stress was applied in FBG before the fixation. According to Figure 8a, the FBG is placed outside of the engine combustion chamber, and it is stressed by a mechanism composed of one 1-mm thickness membrane and one piston. Only one side of the membrane contacts the high-pressure and high-temperature gas inside the engine's combustion chamber. When the membrane is deformed, by the pressure, it moves a piston that stresses the FBG accordingly. In the field tests, we use FBG pressure sensors connected in a point of pressure monitoring available for each engine cylinder, as we can observe in Figure 8b. Figure 9a shows the point in the engine where the FBG pressure sensor was installed. Figure 9b shows the demodulator, the tunable laser kit, and the interrogator installed in a control room, 50 m from the engine under test. Figure 10 shows the curves of wavelength shifting versus pressure for one FBG sensor obtained previously of the field trial, considering the Bragg wavelength in room temperature that was 1532.90 nm (sensor #1). This sensor was submitted to 3 cycles of static pressure to verify the sensor hysteresis. As we can observe, the sensor curves are linear. The R2 coefficient is 0.9943. The other pressure sensor used in the field test had the Bragg wavelength at room temperature at 1548.12 nm. Figure 10 shows the curves of wavelength shifting versus pressure for one FBG sensor obtained previously of the field trial, considering the Bragg wavelength in room temperature that was 1532.90 nm (sensor #1). This sensor was submitted to 3 cycles of static pressure to verify the sensor hysteresis. As we can observe, the sensor curves are linear. The R2 coefficient is 0.9943. The other pressure sensor used in the field test had the Bragg wavelength at room temperature at 1548.12 nm. Figure 10 shows the curves of wavelength shifting versus pressure for one FBG sensor obtained previously of the field trial, considering the Bragg wavelength in room temperature that was 1532.90 nm (sensor #1). This sensor was submitted to 3 cycles of static pressure to verify the sensor hysteresis. As we can observe, the sensor curves are linear. The R2 coefficient is 0.9943. The other pressure sensor used in the field test had the Bragg wavelength at room temperature at 1548.12 nm.
Time (s)
Interrogator signal PWM signal Figure 11. PWM signal versus time of sensor #1 that was obtained during the temperature stabilization period at the engine monitoring point. Figure 12 shows a qualitative comparison of the dynamic curves of sensor #1 in the monitoring point of engine combustion obtained by the PWM demodulator (with 10 moving average) and the commercial interrogator in terms of wavelength shifting considering the Bragg wavelength in room temperature. As we can observe, the output signal intensity of the demodulator in Figure 12 is lower than the one obtained in laboratory measurements showed in Figure 7a. In Figure 7a the peak-to-peak intensity is~0.12 V versus~0.009 V in Figure 12. This fact is partly attributed to the higher sweep range used in field tests (~20 nm) compared with the sweep range used in laboratory measurements (7 nm). This higher sweep range was implemented to measure different Bragg wavelengths of two FBG sensors. Also, we observed that the PWM signal presents the narrowest linewidth compared with the interrogator signal.
The total wavelength shifting showed in the peak bases in Figure 12 for the interrogator signal (~8380 pm) is attributed to FBG sensor substrate deformation and the temperature of FBG in the monitoring point. Based on the stainless steel FBG substrate dimensions where the FBG was fixed (70 mm), its thermal expansion coefficient 16.10 −6 • C −1 , and the FBG temperature sensitivity of 13 pm • C −1 [24], we can estimate that the temperature operation for this sensor was around 200 • C.
Finally, Figure 13 shows a comparison of the PWM signal and the reference sensor. The time scale was synchronized in order to have a better comparison of the temporal characteristics of the sensors. Except for the noise in the PWM signal base, it shows a response similar qualitatively to the reference sensor. The first hypothesis for noise was due to the tunable laser jitter [32], which would originate from the triangular signal source that sweeps the laser. In our experiments, we used the BK Precision model BK4054B waveform generator. This generator features an RMS 300 ps + 0.05 ppm cycle-to-cycle jitter in 1 kHz and 1 Vpp. Considering the sweep frequency of 1 kHz (1 ms cycle), a variation of 300 ps would have little effect on the creation of the observed noise. A second hypothesis raised would be due to the noise margin in the decision threshold of the Flip-Flop D 74LC74 logic gate, which could widen or shorten the PWM pulses due to the variation of the decision point. This hypothesis was discarded since the voltage levels provided by the optical receiver of the demodulator to the Flip-Flop D inputs were designed to work saturated. Finally, the most likely hypothesis is attributed to the residual noise originated from the demodulator's power supply. The noise frequency is close to 60 Hz.
Sensors 2020, 20, x; doi: FOR PROOF www.mdpi.com/journal/sensors As we can observe, the output signal intensity of the demodulator in Figure 12 is lower than the one obtained in laboratory measurements showed in Figure 7a. In Figure 7a the peak-to-peak intensity is ~0.12 V versus ~0.009 V in Figure 12. This fact is partly attributed to the higher sweep range used in field tests (~20 nm) compared with the sweep range used in laboratory measurements (7 nm). This higher sweep range was implemented to measure different Bragg wavelengths of two FBG sensors. Also, we observed that the PWM signal presents the narrowest linewidth compared with the interrogator signal.
The total wavelength shifting showed in the peak bases in Figure 12 for the interrogator signal (~ 8380 pm) is attributed to FBG sensor substrate deformation and the temperature of FBG in the monitoring point. Based on the stainless steel FBG substrate dimensions where the FBG was fixed (70 mm), its thermal expansion coefficient 16.10 −6 °C −1 , and the FBG temperature sensitivity of 13 pm°C −1 [24], we can estimate that the temperature operation for this sensor was around 200 °C.
Finally, Figure 13 shows a comparison of the PWM signal and the reference sensor. The time scale was synchronized in order to have a better comparison of the temporal characteristics of the sensors. Except for the noise in the PWM signal base, it shows a response similar qualitatively to the reference sensor. The first hypothesis for noise was due to the tunable laser jitter [32], which would originate from the triangular signal source that sweeps the laser. In our experiments, we used the BK Precision model BK4054B waveform generator. This generator features an RMS 300 ps + 0.05 ppm cycle-to-cycle jitter in 1 kHz and 1 Vpp. Considering the sweep frequency of 1 kHz (1 ms cycle), a variation of 300 ps would have little effect on the creation of the observed noise. A second hypothesis raised would be due to the noise margin in the decision threshold of the Flip-Flop D 74LC74 logic gate, which could widen or shorten the PWM pulses due to the variation of the decision point. This hypothesis was discarded since the voltage levels provided by the optical receiver of the demodulator to the Flip-Flop D inputs were designed to work saturated. Finally, the most likely hypothesis is attributed to the residual noise originated from the demodulator's power supply. The noise frequency is close to 60 Hz.
Discussion
To obtain a robustness FBG system for instantaneous combustion chamber control, we demonstrated a filterless, multi-point, and temperature-independent FBG dynamical demodulator using PWM, which can be installed close or inside the engine operating environment.
The demodulator uses just electronic components except for the photodetector. All these devices must be selected for operation in high temperatures. Besides, the demodulator works digitally, which reduces the error on the FBG signal conversion. Because the laser output changes in wavelength and time during the sweep, the optical wavelength variation of FBG in the engine became an electrical time variation in the demodulator. When the demodulator receives the analog time variation signal of FBG, it transforms this signal in a digital signal with a pulse width variation. The pulse width variation has the information of dynamical pressure modulated in the FBG. The active electrical filter removes the digital modulation of the PWM signal recovering the original FBG modulated-signal. This recovered signal is sent to the ECU. No synchronization signal is necessary for this system.
The maximization of the demodulator output signal can be obtained by reducing the laser sweep spectral band to the spectral band of the operation temperature of the sensors. The Bragg wavelength (in room temperature) can be the same wavelength for all sensors in the network, but this condition is not obligatory.
The laboratory tests demonstrated that the proposed demodulator reproduced the engine characteristics adequately. Also, the demodulator signal had more resolution than the commercial interrogator. The laser sweep frequency must be 10 times higher than the maximum sensor frequency response to have enough sampled points during the measurements. We also observed some noise in the signal base (in laboratory and field test) that we attribute to the residual noise originated from the demodulator's power supply. Also, we measured the linearity of the demodulator signal that is enough to reproduce the original characteristics of the FBG modulated signal.
In the field test, the demodulator was not tested close to the engine, where we have high-temperatures. The objective of the field test was to verify the performance of the PWM demodulator in terms of signal processing using a dynamical FBG pressure sensor installed in an environment with variable temperature. The design of the electronic-board of the demodulator with high-performance electronic devices is a future project. The demodulator worked well in the field test and, it was proved by the comparison of the PWM signal with the reference pressure sensor.
Conclusions
In this work, we investigated an innovative filterless, multi-point, and temperature-independent FBG dynamical demodulator using the PWM technique. The demodulator was developed to monitor the instantaneous combustion chamber pressure in closed-loop control of fuel mass fraction burned of the thermoelectric engines. The demodulator must work integrated with an ECU close to the engine, where the temperature is high.
The system was characterized in the laboratory using an FBG sensor modulated in a frequency of 6 Hz and a tunable laser with a 1 kHz sweeping frequency and wavelength range from 1527 to 1534 nm. A commercial FBG interrogator was used to compare the results obtained in this application. Besides, it was evaluated in a field test in an engine of a thermoelectric power plant. The demodulator worked well in the field test and, it was proved by the comparison of the PWM signal with one reference pressure sensor. Briefly, we will evaluate the demodulator, integrated with an ECU, very close to the engine in closed-loop control of fuel mass fraction burned of the thermoelectric engines. | 2020-10-21T13:06:18.753Z | 2020-10-01T00:00:00.000 | {
"year": 2020,
"sha1": "14a598104101e8a61e5dc26fd013c5ce76dbd632",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1424-8220/20/20/5825/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "04b410850c85f9fc58eb33b93e827a79695724d3",
"s2fieldsofstudy": [
"Engineering",
"Physics"
],
"extfieldsofstudy": [
"Materials Science",
"Computer Science",
"Medicine"
]
} |
213614559 | pes2o/s2orc | v3-fos-license | Modeling vertical fractures using the Schoenberg model on structured grids by grid-characteristic method
RENSIT | 2019 | Vol. 11 | No.3 DOI: 10.17725/rensit.2019.11.351 Modeling vertical fractures using the Schoenberg model on structured grids by grid-characteristic method Polina V. Stognii Educational Scientific and Experimental Center of Moscow Institute of Physics and Technology, http://kmipt.ru/ Dolgoprudny 141700, Moscow region, Russian Federation E-mail: stognii@phystech.edu Nikolay I. Khokhlov, Igor B. Petrov Scientific Research Institute for System Analysis of Russian Academy of Sciences, https://www.niisi.ru/ Moscow 117218, Russian Federation E-mail: k_h@inbox.ru, petrov@mipt.ru Received 28.08.2019, peer reviewed 16.09.2019, accepted 30.09.2019 Abstract. Seismiс prospecting is one of the most popular methods of revealing oil and gas reservoirs in the North seas. Geological media, which contain hydrocarbon layers, often include different heterogeneities, for example, fractures. Many scientific works are devoted to modelling fractures and ways of considering them in various mathematical problems, but the Linear Slip model, known as Schoenberg model of fracture, is still the most actual, though this model was developed at the end of the last century. This model is characterized by the presence of an additional parameter — so called fracture opening parameter, describing fractures. In this work, we present the algorithm of introducing the Schoenberg fracture model in the grid-characteristic method. The fracture is organized to be parallel to the boundaries of the modelling grid. The fracture is situated on the border of the two media with the same characterizing parameters. We describe in detail the algorithm of computing the meanings of velocity and stress tensor in the points on the border with the fracture. In order to verify the correctness of the developed approach to modelling a fracture using the Schoeberg fracture model, we carry out the test calculations of sesmic waves spread in a homogeneous media with the vertical extremely thin fracture, described by the not equal to null normal component of the fracture opening coefficient and with the equal to null tangential component of the fracture opening coefficient.
1.INTRODUCTION
Fractures are often a barrier on the way of exploring the chosen territory during conducting the geological works on oil and gas deposits extraction [1,2]. They bring in a significant summary into the seismogramms, got as a result of geological layers exploration. Therefore, it is necessary to take into account fractured media while modelling the chosen area in order to obtain more accurate results of seismic data processing.
In reality, the sizes of most fractures are of that sort that the ratio of their width to their height is equal to zero in the limit [3] . These fractures are described by an abstract model of an extremely thin fracture. In this work, we explore only models of extremely thin fractures. A fracture is described by a set of contact and border conditions in the place of its establishment in the computational area.
There are different ways of modelling the seismic waves spread in fractured media. For example, hierarhical grids are applied to modelling statical fractures [4]. In the work [5], authors use the pseudospectral method for modelling fractures with not equal to zero parameter of viscosity.
Hudson model [6] and Schoenberg model [7] are the most well-known models, used for describing fractures. In the Hudson model, a fracture is described by the parameter of effective stiffness, which is linearly or quadratic (first or second order of accuracy, accordingly) dependent from the socalled parameter of fracture density [8].
In Schoenberg model [9,10], the condition of equal normal components of the stress tensor is set on the boundary of a fracture. In addition, the way of computing the normal components through special parameters of fracture opening, dependent from the medium, filling the fracture, is supposed. The Schoenberg model was already introduced in the well-known Galerkin method, the results can be found in [11].
In this work, we introduce the method of computing geological media with a fracture with the help of the grid-characteristic method and the induced Schoenberg model for the 2D case, which is done for this method for the first time. The contact conditions on the boundary with the fracture are described in detail. In addition, we present the example of numerical modelling for the seismic reflection from the fracture on the boundary of two media.
DETERMINANT EQUATIONS
For describing the spread of elastic waves in a homogeneous medium, we used the system of linear-elastic equations [12]: where ρ is the medium density, υ is the speed of elastic waves spread in the medium, σis the Caushy stress tensor, λ and μ are the Lame parameters, determining the properties of the elastic material.
For fracture describing, we used the equations from the model of Schoenberg [13]: , In (3)-(6) the indexes «left» and «right» denote the medium on the one side (left) and the medium on the other side (right) from the fracture. K N and K T are the normal and tangential components of the parameter of fracture opening. These characteristics are known beforehand or can be calculated. The description of the theoretical computation of normal and tangential components of fracture opening are presented in the work [14], the description of calculating the parameter of fracture opening as a result of the laboratory experiment is presented in the work [15]. The width of the fracture is considered to be extremely small.
NUMERICAL METHOD
For solving the system of linear-elastic equations (1), (2) we used the grid-characteristic method [16]. For computing the homogeneous medium, we applied the Laks-Wendroff scheme of the second order of accuracy [17] : In order to compute fractures, we also used the grid-characteristic method, but with some modifications. Now, we examine the algorithm of computing points on the boundary with the fracture in more detail.
For this, we present the system of equations (1), (2) in the following form: where q = {σ xx , σ xy , σ yy , υ x , υ y }. A x and A y are matrixes, made up from the corresponding components of the equations (1), (2): As a result of adopting the method of coordinates splitting in space, we obtain two 1D systems of equations: Now, we examine the system of equations (11) for the x-coordinate: Hereinafter, we consider the fracture, situated perpendicular to the X-axis, parallel to the boundaries of the computational grid cells.
We can make analogous transformations, which are presented lower, for the case, when the fracture is located perpendicular to the Y-axis.
The system of equations (12) is hyperbolic, therefore it can be represented in the form : where Ω x is the matrix, made of the eigenvectors, Λ x is the matrix with the eigenvalues on the diagonal. The eigenvalues of the matrixes A x , A y are: {-c p , c p , -c s , c s , 0}, where c p is the longitudinal sound velocity, c s is the transverse sound velocity.
After the variable change, the system (13) will look as following: 0.
The system (14) contains five independent equations, which can be solved by any differential scheme [17].
We examine the case of computing points on the boundary of the fracture with the help of the CIR scheme [17]. The scheme looks as follows: In (15), (16) the index i signifies the number of coordinate (for example, x), the index n denotes the number of time, t is the time step, h -is the coordinate step, c is velocity. From equations (15), (16), it follows that points on the left side of the fracture can be calculated only for vectors, corresponding to negative eigenvalues (-c p , -c s ) . Points on the right side of the fracture can be calculated only for vectors, corresponding to positive eigenvalues (c p , c s ). For solving the remaining two systems of equations on the left and on the right, we need additional boundary conditions (3)- (6). Now, we explore in detail the algorithm of computing points on the left side of the fracture for positive characteristics (the algorithm of computing points on the right side of the fracture for negative characteristics is the same). For this, we introduce an additional ghost-node border+1 on the right side of the border node [18] into the examined computational grid and place such values into the node border+1, so that the conditions (3), (4) are fulfilled on the boundary of the fracture on the next time step. The conditions (5), (6) allow to calculate the values σ xx ,σ xy on the (n+1)-th time step. In this work, we solved the equations (5), (6) where Ω out is the matrix of the eigenvectors, corresponding to the outcoming characteristics (in other words, negative meanings of the characteristics for the matrixes A x , A y ).
RESULTS
We carried out the numerical modelling of the seismic response from an extremely long fracture in a homogeneous medium for the 2D formulation of the problem. The initial impulse was described through the definition of the stress tensor inside a unit circle with the help of the following formula: xx yy x y x y x y x y The homogeneous medium was described by the following parameters. The speed of the longitudinal waves was equal to 2 m/sec., the speed of the transverse waves -1 m/sec . The medium density was equal to 1 kg/m 3 . The computational area was 9 m along the X-axis and 6 m along the Y-axis. The extremely thin fracture along the whole computational grid was situated vertically on the boundary x = 3 m (Fig. 1).
In the calculations, the time step was equal to 0.005 sec., the coordinate step for Xand Y-axis was 0.025 m. The value of the normal component of the fracture opening parameter K N was equal to 400 Pa/m, the value of the tangential component of the fracture opening parameter K T was 0 Pa/m. The same medium with the parameters, described earlier in this section, is situated on the left and on the right side of the fracture. We calculated the spreads of the seismic impulse in the right and left areas relative to the fracture independently, excepting the border points, where we changed the values of the normal and tangent components of the stress tensor (according to the border conditions (3)-(6)). At this stage, we made all the calculations one after the other. Later on, the calculations will become faster with the help of parallelization, for example, using the MPI technology [20].
CONCLUSION
In this work, we introduced the realization of the Schoenberg fracture model into the gridcharacteristic method. We interposed the Schoenberg model, which is widely used for modelling fractured media, in the grid-characteristic method with the help of the additional corresponding boundary conditions in the computational area, where the fracture is situated. We presented the detailed description of calculating the points on the boundary of the fracture for the 2D case. The boundary conditions were introduced through the CIR scheme of the first order of accuracy, while the other points of the computational grid were calculated using the Laks-Wendroff scheme of the second order of accuracy.
As an example, we carried out the test calculation of the seismic waves spread in the homogeneous medium with the vertical fracture for the case, when the tangential component of the fracture opening parameter was equal to zero. Wave patterns, depicting the normal component of the stress tensor, showed the reflected wave from the fracture and further spread of the seismic wave across the homogeneous medium.
Later on, we suppose to make the parallelization of the introduced algorithm for computing the seismic waves spread in a homogeneous medium with a fracture. It will help to speed up the computations significantly. In addition, the logic continuation of this work is the development of the analogous algorithm for computing fractured media for the 3D case.
aCknowledgments
The reported study was funded by RFBR according to the research project № 19-01-00281 . | 2020-01-02T21:48:18.197Z | 2019-12-29T00:00:00.000 | {
"year": 2019,
"sha1": "75e29c603942cfce7762d6172eae490cbd79eea6",
"oa_license": null,
"oa_url": "http://en.rensit.ru/vypuski/article/297/11(3)351-356e.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "a5418af9013bd161d09ab1ddbfdb8c7064fb19f4",
"s2fieldsofstudy": [
"Geology"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
52994674 | pes2o/s2orc | v3-fos-license | Resonant Tunneling in the Quantum Hydrodynamic Model
The phenomenon of resonant tunneling is simulated and analyzed in the quantum hydrodynamic (QHD) model for semiconductor devices. Simulations of a parabolic well resonant tunneling diode at 77 K are presented which show multiple regions of negative differential resistance (NDR) in the current-voltage curve. These are the first simulations of the QHD equations to show multiple regions of NDR. Resonant tunneling (and NDR) depend on the quantum interference of electron wavefunctions and therefore on the phases of the wavefunctions. An analysis of the QHD equations using a moment expansion of the Wigner-Boltzmann equation indicates how phase information is retained in the hydrodynamic equations.
INTRODUCTION sonant tunneling of electrons in quantum emiconductor devices can be modeled by adding quantum corrections to the classical hydrodynamic equations.The leading O(h2) quantum corrections include the effects of particle tunneling through potential barriers and particle buildup in potential wells.
ers.Sequential tunneling occurs if electrons tunnel through the first barrier, undergo scattering, and then tunnel through the second barrier.Coherent tunneling may be interpreted as the constructive interference of multiply reflected electron waves within the well.Resonant tunneling in actual devices usually involves a mixture of coherent and sequential tunneling.Scattering broadens the resonances and reduces the peak to valley ratios in the current- voltage curve.In this investigation, I will present two sets of simulations of the resonant tunneling diode: in the first set, scattering takes place throughout the device, while in the second set the collision terms in the QHD model are set to zero in the barriers and the well, to insure that the electron tunneling is coherent.
Coherent resonant tunneling and negative differ- ential resistance depend on the quantum interfer- ence of electron wavefunctions, and therefore on the phases of the wavefunctions.I will indicate how phase information is retained in the hydrodynamic equations through an analysis of the OHD equa- tions using a moment expansion of the Wigner- Boltzmann equation.202 CARL L. GARDNER
THE QUANTUM HYDRODYNAMIC MODEL
The QHD equations have the same structure [1] as the classical hydrodynamic equations: the stress tensor by h2n 0 2 eij -nTtij -t-12-OxiOx log(n) + O(h4) (6) On where n is the electron density, rn is the effective electron mass, II is the momentum density, u is the velocity, Pij is the stress tensor, V= -eb is the potential energy, b is the electric potential, e > 0 is the electronic charge, W is the energy density, q is the heat flux, T O is the temperature of the semicon- ductor lattice in energy units (k B is set equal to 1), s is the dielectric constant, N o is the density of donors, and NA is the density of acceptors.Spatial indices i, j equal 1, 2, 3, and repeated indices are summed over.Eq. (1) expresses conservation of electron number, Eq. ( 2)expresses conservation of momen- tum, Eq. (3) expresses conservation of energy, and Eq. ( 4) is Poisson's equation.The classical collision terms in Eqs. (2)and (3) are modeled by the relax- ation time approximation, with momentum and en- ergy relaxation times Tp and "r w.The heat flux is specified here by Fourier's law q -KVT, where T is the electron temperature.
The quantum corrections to the classical hydrodynamic equations appear in the stress tensor and the energy density.As h --* 0, the quantum corrections can be developed in a power series in h2.The actual expansion parameter is h2/8mTl 2, where is a characteristic length scale of the problem [2, 3].For the resonant tunneling diode in section 3 with T --T O 77 K and 100 A, the expansion parameter --0.23.
In Ref. [1], I showed that for the O(h2) momen- tum-shifted thermal equilibrium Wigner distribution function, the momentum density is given by 1-I mnu (5) and the energy density by Wigner [4] derived the quantum correction to the energy density (in a different form).Ancona and Tiersten [5] proposed the quantum correction to the stress tensor on general thermodynamical grounds, and Ancona and Iafrate [2] later derived the correc- tion in the Wigner formalism.A one-dimensional version of the QHD equations was derived by Gru- bin and Kreskovsky in Ref. [6].
SIMULATIONS OF MULTIPLE RESONANCES
I will simulate a GaAs resonant tunneling diode with double Alo.8Gao.2Asbarriers and a parabolic AlxGaa_ x As quantum well.The diode consists of an n + source (at the left), an n channel, and an n / drain (at the right).The doping density No 10 TM cm -3 in the n + source and drain, and No 5 1015 cm -3 in the n channel (see Figure 1).The channel is 500/ long, the barriers are 30 , wide, and the well between the barriers is 300 , wide.
The device has 70 .spacers between the barriers and the contacts (source and drain) to enhance NDR.The barrier height ' is set equal to 0.7 eV.
In the well, the AI mole fraction X varies between 0.8 and 0, so that the potential barrier height in the well is (see Figure 2) [ ] 2 x + x.) "well 0.7 eV 1/2( XR (8) where x/ and x R are the coordinates of the left and right edges of the well.The left edge of the first barrier and the right edge of the second barrier are modeled as step functions.(Computationally the step function goes from 0 to 1 over one Ax.)For the classical momentum and energy relax- ation times, I use modified Baccarani-Wordemann models: r0 'p "rpo T (9) (3) Zp T 'w---1 + 2 (10) -mv where the low-energy momentum relaxation time "rpO is set equal to 0.9 picoseconds from the low-field electron mobility in GaAs at 77 K.For lower valley electrons in GaAs at 77 K, the effective electron mass rn 0.063 me, where m e is the electron mass, and the saturation velocity v = 2 10 7 cm/s.I set K0 0.4 in the Wiedemann-Franz formula for thermal conductivity q KVT, r, XOZponTo/m.( The dielectric constant e 12.9 for GaAs.I discretize the 1D steady-state QHD equations using the second upwind method and compute the solution using a damped Newton method (see Ref. [1] for details).The barrier height ' is incorporated into the QHD transport equations ( 1)-( 3) by replacing I/ I/+ '.Poisson's equation is not changed.
The current voltage curve for the resonant tunnel- ing diode at 77 K is plotted in Figure 3, using a grid with 300 Ax.For these simulations, scattering takes place throughout the device.There are seven reso- nances and seven regions of negative differential resistance in the current-voltage curve between 0 and 0.5 volts.Figure 4 presents the current-voltage curve for the same device with the classical collision terms in the momentum and energy equations ( 2) and ( 3) turned off in the barriers and the well.Electron tunneling in this case must be purely coherent rather than sequential plus coherent, since there is no scattering in the well or barriers.There are twelve resonances and seven regions of negative differen- tial resistance in the current-voltage curve between 0 and 1 volt.
Note that there are five resonant "shoulders" in the current-voltage curve between 0 and 0.33 volts, and that the sixth through twelfth resonances be- come more pronounced as V increases.Ref. [7] explains this effect, which was observed in an experimental device: Below 0.33 volts, electrons tunnel out of the well through the thick portion of the parabolic well in Figure 2, and the transmission resonance widths are small.As the voltage bias increases above 0.33 volts, the transmission reso- nance widths rapidly increase because (i) the right barrier height is progressively reduced and (ii) elec- trons tunnel out of the well through the thin portion of the parabolic well.
The resonant peaks of the current-voltage curve occur as the electrons tunneling through the first barrier come into resonance with the energy levels of the quantum well.The locations of the reso- nances can be qualitatively understood from the 0 0.i 0.2 0.3 0.4 0 Volts FIGURE 4 Current density in milliamps/cm vs. voltage for the resonant tunneling diode at 77 K with scattering turned off in the barriers and the well.The dots represent computed solution points.FIGURE 5 Time (solid line) in 10 -4 seconds spent by electrons in the well vs. voltage with scattering throughout the device.For reference, a scaled version of the current density (dots) is also shown.energy levels of an infinite parabolic well.For an infinite well, E,= A+ h r n x ----= A+-0.087eV (12) where x 0 300 is the width of the finite well and A is a non-negative integer.The energy levels of the infinite parabolic well are linearly spaced in A with a spacing of 87 meV, compared with the linear spacings of approximately 70 mV (with scattering throughout the device) and 100 mV (with scattering turned off in the barriers and the well)for the computed peaks in the current-voltage curves.The first peak with scattering throughout the device oc- curs at 0.030 volts, compared with E 0 = 0.044 eV.
The overall shape of the current-voltage curve in Figure 3--the rise to a peak at 0.11 volts and the fall to the valley at 0.58mis determined by the negative differential resistance of a 0.7 eV high double barrier square well with an effective width of approximately 75 A. The energy levels of the parabolic well modulate this basic shape.
The time spent by electronics in the well is shown in Figure 5.The "dwell" time has relative maxima at voltages corresponding to valleys (relative minima) of the current-voltage curve.The absolute maximum of the dwell time occurs at the global valley of the current-voltage curve at 0.58 volts; the dwell time subsequently decreases rapidly toward zero (Cf.Ref. [1], Figure 5).The dwell time mimics 1/Ijl, where j =-enu is the current density, since the peak electron density in the well increases monotonically with applied voltage.Microscopic quantum calcula- tions predict that the electron dwell time is maxi- mum at resonance.The QHD model presents a different "macroscopic" interpretation of the dwell time (see the discussion in Ref. [1]).
PHASE INFORMATION IN THE QHD MODEL
The computations of multiple resonances and re- gions of negative differential resistance naturally raise the question: how do the hydrodynamic equations, expressed in terms of macroscopic variables n, u, and T, "know" about quantum interference, which depends on the microscopic phases of the electron wavefunctions?By analyzing the first three moments of the Wigner-Boltzmann equation, I will show how the phases of the electron wavefunctions are present in the hydrodynamic equations.
For a mixed quantum mechanical state described by wavefunctions (x, t)(A 1, 2,... ) with occupa- tion numbers ax, the Wigner distribution function is defined as fw(X, p, t) (Trh) -3 Ea f day The sum of the occupation numbers Y'.xa M, where M is the total number of electrons in the system.In general the ax's may vary slowly with x and t.For the purposes of the derivation below, I will assume the aa's are constant. (For a mixed state in thermal equilibrium with a heat bath at tempera- ture T--1/fl, aa t e-te.)I will normalize the wavefunctions so that fd3x 1 1.
I will assume that the electron flow can be ap- proximated by a single-particle effective mass Schr6dinger equation with a self-consistent manybody field: o h 2 ih Exqa V2 + Vx (14) at where the wavefunctions have energies Ex and V(x) is the self-consistent potential ener.In the effec- tive mass approximation, Schr6dinger's equation with the free electron mass and the total potential ener (the periodic potential ener of the semiconductor lattice plus V) is rewritten with the total potential ener replaced by V and the electron mass re- placed by the effective electron mass.
The local average value of an obseable X is defined in the Wigner formalism as < x(x,t)) fd3px(x,p,t)fw(x,p,t). (15)e first three moments of the Wigner-Boltzmann equation are obtained [8,1] define lIi, eij, W, and q.The quantum conservation laws have the same form as their classical counter- parts.Explicit factors of h enter only at the fourth and higher moments.
To calculate the average values in the first three moment equations, in Ref. [1] I used the momen- tum-shifted version of the O(h2) thermal equilib- rium solution [4] to the Wigner-Boltzmann equation.
The momentum-shifted O(hz) fw need only approx- imate the actual Wigner distribution function closely enough for the average values in Eqs. ( 16)-( 18) to be close to the actual values.Using this approach, I derived the three-dimensional QHD equations ( 1)- (3) with Hi, Pij, and W given by Eqs. ( 5)- (7).
Since here I am concerned with how phases ap- pear in the QHD equations, I will instead evaluate average values by writing the wavefunction q,x in terms of its magnitude A a and phase 0" qa(x, t) Ax(x, t)e i(x't). (23 Using this decomposition of the wavefunction, the density is given by n fd3pfw(X, p, t) E a f x (rh) 3d3yA:t(x + y)Aa(x y) Xe-iOa(x+y) +iOa(x-y)e2ip.y/ h EaaA2(x,t)= En (24 where the time-dependence in A a and 0 h is left implicit except when needed for the sake of clarity. n is the electron density for state A, and the total electron density is obtained by summing over all states.The momentum density equals < Pi) Hi fd3pPifw( x, P, t) ) A;(x + y)A(x-y)e -iOa(x+y)+iO"(x-y) o OYi {Ax(x + y)A,(x y)e -i^(x+y)+ia(x-y)} h EaxA2(x, t)V0(x, t) .,mnxuxiwhere I have integrated by parts and where the velocity u is defined by h u a --VO a.
(26) m This definition agrees with the standard expression for the semiclassical momentum pa mu hVO, of an electron in state A. Next I calculate From Eq. ( 27), the energy density equals uiW ujPij + qi dap 2m 2 fw(x, P, t) OYiOY {A,(x + y)Aa(x y) e 0x(x + y) + 0^(x y) ., u ;t mn u A The results in Eqs. ( 24)-( 29) are exact, and do not involve an expansion in h.However the moment equations ( 16)-( 18) with expressions (24)-( 29) do not form a closed set of equations in terms of hydrodynamic state variables (say n, u and T or n, II, and W).Note that although the only explicit occurrence of h is h2, the sums over A will in general involve a functional dependence on h.Spatial derivatives of the wavefunction phases 0x(x, t) appear in the expressions for H i, P/y, W, and q through ux.Only partial phase information is contained in the first three moment equations, since the fluid dynamical quantities (25) and ( 27)-(29) involve weighted sums of spatial derivatives of 0 h.To make further progress, define u, u + f (30 where u is the macroscopic fluid velocity and fi is the velocity with respect to the macroscopic fluid flow.I will assume that the fix are random, i.e., that nxa 0 (31) A and that the temperature is defined to leading order by the average of the microscopic kinetic energies of motion relative to the macroscopic fluid flow: _.,na-mfi -nT + O( hZ). (32) Then I obtain Hi Emn(ui + Ai) --mui (33) Pi uilI _.,(ui + txi)mn,(uj + taj) x 8m h2n --/VT V211 -1--O(h2). (36) Eq. ( 35) follows if the Wigner distribution function in the classical limit h 0 is a momentum-shifted Maxwell-Boltzmann distribution (see e.g.Ref. [9], pp.99-100).In writing the last equation, I have assumed that E xna 72 log(nA) 0, EfiajnyV.log(ha) 0, EnxV2fia 0. (37) A a The classical heat flux vanishes if the distribution function is a momentum-shifted Maxwell-Boltzmann distribution.The Fourier term -KVT in Eq. ( 36) is obtained if the Wigner distribution function in the classical limit is the momentum-shifted Maxwell-Boltzmann distribution plus the first-order Chapman-Enskog correction (see Ref. [9], pp. 99-100and 103-106).Note that I have not used the quantum correction to the heat flux in the simulations presented here.
Even with these simplifications and with q rVT, the moment equations ( 16)-( 18) with ex- pressions ( 24) and ( 33)-( 35) do not form a closed set of equations in terms of hydrodynamic state vari- ables, due to the quantum terms in Pij and W.
The quantum terms in the .stresstensor and the energy density were evaluated in Ref. [1] to leading order in h 2 by Eqs. (6)and (7).Then the moment equations ( 16)-( 18) do form a closed set in terms of hydrodynamic state variables (say n, u, and T).
The moment equations simplify to a great extent for a pure quantum mechanical state, in which all the ax's except one (say a 1) vanish, and al 1.For the pure state, The expression (42) for q is written down in Ref.
In this case, the moment equations ( 16)-(18) with expressions (25) and ( 40 Note that T 0 for the pure state and that the conservation of energy equation (45) follows from the equations for conservation of particles (43) and momentum (44).
For the pure state, then, we have the "hydrodynamic" formulation (see e.g.[11])of the quantum mechanics of a pure state: and reproduces the bound-state solutions for q,x(x, t) for 0 < E < U, where E(A 0, 1, 2,...) is the energy of the particle (see e.g.Ref. [12], pp. 152-155).The solution in the square well may be thought of as resulting from the interference of a right moving and a left moving wave.In an analo- gous but more complicated way, the full quantum hydrodynamic equations "sense" the width of the quantum well in the resonant tunneling diode.
CONCLUSION
The macroscopic quantum hydrodynamic equations are capable of modeling effects in quantum semi- conductor devices that depend on quantum interfer- ence.The interference effects appear in the hydro- dynamic expressions for the velocity, stress tensor, energy density, and heat flux through sums involving the semiclassical velocities u hVO/m.
The QHD simulations of multiple resonances and multiple regions of NDR in the parabolic well reso- nant tunneling diode clearly demonstrate the effects of quantum interference and present an intuitive macroscopic picture of electron dwell times in the quantum well.
FIGURE 2 FIGURE 3
FIGURE 2Barrier and well heights in eV.
1
)-(42) form a closed set in terms of hydrodynamic state variables n and u.Eqs.
V
46) and (47) are exact, and follow directly from Schr6dinger's equation.to the "hydrodynamic" equations (46) and (47) for the 1D finite square potential well U x< -a | 2018-09-19T07:31:48.579Z | 1995-01-01T00:00:00.000 | {
"year": 1995,
"sha1": "96c59ec00ecb28eb291152b2f8b840613b112063",
"oa_license": "CCBY",
"oa_url": "https://downloads.hindawi.com/archive/1995/054301.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "6a0dbe8a0bf62e1419e6a9fe8281180ad6b51292",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
258919446 | pes2o/s2orc | v3-fos-license | Adapting Single-Image Super-Resolution Models to Video Super-Resolution: A Plug-and-Play Approach
The quality of videos varies due to the different capabilities of sensors. Video super-resolution (VSR) is a technology that improves the quality of captured video. However, the development of a VSR model is very costly. In this paper, we present a novel approach for adapting single-image super-resolution (SISR) models to the VSR task. To achieve this, we first summarize a common architecture of SISR models and perform a formal analysis of adaptation. Then, we propose an adaptation method that incorporates a plug-and-play temporal feature extraction module into existing SISR models. The proposed temporal feature extraction module consists of three submodules: offset estimation, spatial aggregation, and temporal aggregation. In the spatial aggregation submodule, the features obtained from the SISR model are aligned to the center frame based on the offset estimation results. The aligned features are fused in the temporal aggregation submodule. Finally, the fused temporal feature is fed to the SISR model for reconstruction. To evaluate the effectiveness of our method, we adapt five representative SISR models and evaluate these models on two popular benchmarks. The experiment results show the proposed method is effective on different SISR models. In particular, on the Vid4 benchmark, the VSR-adapted models achieve at least 1.26 dB and 0.067 improvement over the original SISR models in terms of PSNR and SSIM metrics, respectively. Additionally, these VSR-adapted models achieve better performance than the state-of-the-art VSR models.
Introduction
Numerous videos are captured every day; however, due to the different capabilities of sensors, the quality of captured videos can vary greatly, which affects the subsequent analysis and applications [1][2][3][4]. Recently, computer technologies have been applied to many fields [5][6][7][8]. In particular, video super-resolution (VSR) is a technology for improving the quality of captured video. It produces high-resolution (HR) video frames from their low-resolution (LR) counterparts. The VSR problem is challenging due to its ill-posed nature, but its applications include video display, video surveillance, video conferencing, and entertainment [9].
VSR models take consecutive frames as input. Single-image super-resolution (SISR) methods process only one image at a time. So, VSR models take both spatial information and temporal information into account, while SISR models only exploit spatial information for super-resolution (SR) reconstruction. Thus, many VSR methods adapt SISR models for spatial information extraction. For example, Haris et al. [10] introduced RBPN, which employs blocks from DBPN [11] in a recurrent encoder-decoder module to utilize spatial and temporal information. Tian et al. [12] adapted EDSR [13] as the main design for the SR reconstruction network in TDAN. Liang et al. [14] utilized residual Swin Transformer blocks from SwinIR [15] in their proposed RVRT. Although these works have adapted SISR models, each method utilizes only one SISR model. Applying SISR techniques to the VSR models would require considerable effort and they may not perform as effectively as specialized VSR models.
Meanwhile, several VSR methods do not rely on SISR models. For instance, Xue et al. [16] proposed TOF, which estimates task-oriented flow to recover details in SR frames. Wang et al. [17] proposed SOF-VSR, which estimates HR optical flow from LR frames. SWRN [18] can be utilized in real time on a mobile device. However, the development of a VSR model without adapting SISR methods is very costly, as the model needs to capture both temporal and spatial information. Moreover, compared with SISR methods, they may be less effective in utilizing spatial information.
To alleviate the above issues, we propose a plug-and-play approach for adapting existing SISR models to the VSR task. Firstly, we summarize a common architecture of SISR models and provide a formal analysis of adaptation to achieve better effectiveness of different SISR models. Then, we present an adaptation method, which inserts a plug-andplay temporal feature extraction module into SISR models. Specifically, the temporal feature extraction module consists of three submodules. The spatial aggregation submodule aligns features extracted by the original SISR model. The alignment is performed based on the result of the offset estimation submodule. Then, the temporal aggregation submodule is applied to aggregate information extracted from all neighboring frames.
To evaluate the effectiveness of the proposed method, we adapt five representative SISR models, i.e., SRResNet [19], EDSR [13], RCAN [20], RDN [21], and SwinIR [15], and the evaluations are conducted on two popular benchmarks, i.e., Vid4 and SPMC-11. On the Vid4 benchmark, the VSR-adapted models achieve at least 1.26 dB and 0.067 improvements over original SISR models in terms of peak signal-to-noise ratio (PSNR) [22] and structural similarity index (SSIM) [23], respectively. On the SPMC benchmark, the VSR-adapted models achieve at least 1.16 dB and 0.036 gain over original SISR models in terms of PSNR and SSIM, respectively. Moreover, the VSR-adapted models surpassed the performance of state-of-the-art VSR models.
For this paper, the main contributions are as follows: (1) We propose a plug-and-play approach for adapting SISR models to the VSR task. Instead of adapting one SISR model, the proposed method is based on a common architecture of SISR models. (2) A plug-andplay temporal feature extraction module is introduced. Thus, the adapted model gains the capability to exploit temporal information. (3) Extensive experiments are conducted to evaluate its effectiveness.
Single-Image Super-Resolution
The SISR problem is an ill-posed problem, and learning-based methods have significantly improved the performance in terms of accuracy [13,15,[19][20][21]24,25] and speed [26][27][28][29]. In 2014, Dong et al. [30] introduced a learning-based model, namely SRCNN, into the SISR field. Inspired by ResNet [31], Ledig et al. [19] proposed SRResNet in 2017. SRResNet [19] accepts LR images directly and achieves high performance and increased efficiency. Kim et al. [13] improved the SRResNet by removing unnecessary batch normalization in residual blocks and expanding the number of parameters. In 2018, Zhang et al. [21] employed a densely connected architecture. All extracted features are fused to utilize hierarchical information. Subsequently, Zhang et al. [20] introduced the channel attention mechanism that adaptively weights features channel-wisely. In 2021, Liang et al. [15] proposed SwinIR by making use of the Transformer [32]. Additionally, SwinIR uses the Swin Transformer [33] variation, which is more appropriate for computer vision tasks. By appropriately employing convolution layers and Swin Transformer modules, SwinIR can capture local and global dependencies at the same time, resulting in SOTA performance.
Video Super-Resolution
In recent years, deep-learning-based models have been used to solve the VSR problem, and have become increasingly popular [9]. We roughly divide VSR models into two categories: (1) Models adapting SISR models: Sajjadi et al. [34] proposed FRVSR, which takes EnhanceNet [35] as the subnetwork for SR reconstruction. Haris et al. [10] applied the iterative up-and downsampling technique [11] in RBPN. The representative deep learning SISR model, EDSR [13], is utilized by many VSR models. Tian et al. [12] applied a shallow version of EDSR [13] in TDAN. EDVR [36] and WAEN [37] both employed the residual block and upsampling module from EDSR [13] in the reconstruction module. Inspired by [12], Xu et al. [38] adapted EDSR as the reconstruction module. EGVSR [39] applied ESPCN [26] as the backbone for the SR net. The recently proposed RVRT [14] utilized the residual Swin Transformer block, which is proposed in SwinIR [15].
(2) Models without adapting SISR models: DUF [40] reconstructs SR frames by estimating upsampling filters and a residual image for high-frequency details. Kim et al. [41] employed 3D convolution to capture spatial-temporal nonlinear characteristics between LR and HR frames. Xue et al. [16] proposed a method, namely TOF. It learns a task-specific representation of motion. Wang et al. [17] proposed SOF-VSR, which estimates HR optical flow from LR frames. To better leverage the temporal information, TGA [42] introduced a hierarchical architecture. Recently, Chan et al. [43] proposed BasicVSR by investigating the essential components of VSR models. Liu et al. [44] applied spatial convolution packing to jointly exploit spatial-temporal features. For better fusing information from neighboring frames, Lee et al. [45] utilized both attention-based alignment and dilation-based alignment. Lian et al. [18] proposed SWRN to achieve real-time inference while producing superior performance.
Because VSR models have to capture both temporal and spatial information, proposing a VSR method requires more effort. Thus, many researchers turn to adapting SISR models. Based on SISR models, proposing a VSR method can focus on capturing temporal information. However, these models either utilize a SISR model as a subnet or adapt modules from a SISR model to extract features. Additionally, they may be less effective than those methods that do not adapt SISR methods. Our work proposed a plug-and-play approach to adapt SISR models to the VSR task. The proposed method works on different SISR models as it follows the common architecture of SISR models we have summarized. The spatial information and temporal information are both extracted in the proposed method.
Methodology
In this section, we first summarize the common architecture of SISR models. Then, we provide a formal analysis of adaptation. Following that, a general VSR adaptation method is proposed. Finally, we present a plug-and-play temporal feature extraction module.
Revisit of Single-Image Super-Resolution Models
For the effectiveness on different SISR models [13,15,[19][20][21]46], we first summarize a common architecture, as shown in Figure 1. For simplicity, some operations such as element-wise addition and concatenation are omitted. As shown in Figure 1a, the common architecture of SISR models can be divided into three modules: shallow feature extraction (FE) module, deep FE module, and reconstruction module. Figure 1b-e illustrate the details of four SISR models. As one can see, the shallow FE module takes one LR image as input and extracts features by a few convolution layers. The deep FE module consists of several submodules or blocks, where advanced techniques, such as dense connection [21], channel attention [20], and self-attention [15], are applied. Thus, the deep FE module is where the key novelty of SISR models lies. Finally, the features from the deep FE module are fed to the reconstruction module to produce the SR image. Thus, given an LR image y ∈ R H×W×3 , these SISR models can be generalized using the following representation:
LR
where Method SISR (·) is the SISR model. x ∈ R sH×sW×3 represents the SR result with upscale factor s. H and W denote the height and width of LR image, respectively. According to the common architecture of SISR models, Equation (1) can be expanded as where the shallow and deep FE modules are noted as FE shallow (·) and FE deep (·), respectively. The reconstruction module is denoted as Recons(·). Different from the SISR problem, the VSR methods have to exploit both spatial and temporal information. Thus, we make use of sliding window framework [12] to capture temporal dependency. Given consecutive 2n + 1 LR frames Y = {y t−n , · · · y t−1 , y t , y t+1 , · · · y t+n }, the representation of VSR models is formulated as where the VSR method is Method VSR (·). x t represents the reconstructed SR frame, the frame index of which is t. Note that the main difference between Equations (1) and (3) is the input, and Equation (2) is an expanded representation of Equation (1). In order to adapt existing SISR models to the VSR task, a straightforward method is to modify the shallow FE module. Then, the adapted model can be represented as where FE shallow (·) is the modified shallow FE module.
Proposed Video Super-Resolution Adaptation Method
According to the analysis in Section 3.1, we propose a general method to easily adapt SISR models to the VSR task. As shown in Figure 2, the architecture of the proposed VSR-adapted models consists of 4 modules. Firstly, the VSR-adapted model applies the shallow FE module FE shallow (·) to obtain low-level features F s,i ∈ R H×W×C for each LR frame y i . The subscript i represents the relative index of the center frame. The center frame is denoted as 0 , and C stands for the number of channels in a feature. The shallow feature of center frame F s,0 is skip-connected to the output of the deep FE module with element-wise addition for global residual leaning. Secondly, the temporal FE module FE temporal (·) is employed to exploit spatial-temporal information. It takes LR frames to estimate the offsets of pixels. It also takes shallow features which will be spatially aggregated based on the offsets. In order to enable the deep FE module to leverage information from all LR frames, spatial-aggregated features are temporally aggregated in the temporal FE module. Thirdly, the deep FE module FE deep (·) is responsible for estimating accurate residual features with advanced techniques. Finally, the reconstruction module Recons(·) upsamples features with specific scale factors and produces SR frames. The architecture can be represented as F T = FE temporal (F s,−n , · · · , F s,0 , · · · , F s,n , y −n , · · · , y 0 , · · · , y n ), where i denotes the relative index of the target frame, ranging from −n to n. The temporal feature F T ∈ R H×W×C is the output of temporal FE module. For adapting different SISR models, the proposed method maintains the shallow FE module, deep FE module, and reconstruction module unmodified. Furthermore, we employ the temporal feature extraction module between the shallow FE module and the deep FE module in accordance with accuracy and latency concerns.
LR
From an accuracy perspective, the main difference between an input LR frame and its ground truth HR frame is the high-frequency content. Thus, the better the residual feature that is extracted, the better the achieved performance. The proposed architecture takes advantage of the deep FE module, where the key novelties of SISR models lie [46]. Further, with the information from neighboring frames, the deep FE module is able to extract more accurate features for reconstruction. Thus, the temporal FE module is employed before deep FE module.
From a latency perspective, the temporal FE module aggregates the features extracted from all input frames. It requires previous modules to complete their processing for each frame. To minimize the overall computation time, the proposed temporal FE module is employed after shallow FE module because its relatively small number of layers has a negligible impact on inference latency.
Plug-and-Play Temporal Feature Extraction Module
In order to exploit spatial-temporal information, the temporal FE module is proposed. The detailed architecture is illustrated in Figure 3, which consists of three submodules, i.e., offset estimation, spatial aggregation, and temporal aggregation.
The shallow feature F s,i and the estimated offset F o f f ,i are then fed into the spatial aggregation submodule. Here, a variation of deformable convolution is used to extract features F s,i , which takes F o f f ,i for offset. This allows the offset feature F o f f ,i to guide the alignment in the spatial aggregation submodule. Another deformable convolution is applied for refinement, resulting in output feature F T,i ∈ R H×W×C . The spatial aggregation submodule can be given by where DConvA(·, ·) is the variation of deformable convolution. The variation of deformable convolution DConvA(·, ·) takes the first input for feature extraction and the second input for offset. After spatial aggregation, the temporal aggregation submodule fuses these spatialaggregated features F T,−n · · · F T,n . For fusing a feature with (2n + 1) × C channels, a simple convolution layer is not sufficient. Therefore, a residual channel attention block [20] is employed to adaptively weight these features channel-wise. A convolution layer for channel reduction is then applied. The channel shrinkage is performed in two steps to minimize information loss: first reducing to twice the SISR features' channels and then reducing to once. The temporal aggregation submodule can be represented as where RCAB 1 (·) and RCAB 2 (·) are residual channel attention blocks. The number of channels of the features output by Conv 3 (·) and Conv 4 (·) is 2 × C and C, respectively. The temporal-aggregated feature is F T ∈ R H×W×C . Overall, the spatial aggregation aligns neighboring features based on the result of the offset estimation submodule. Then, the temporal aggregation submodule fuses the spatial-aggregated features, resulting in an output containing information from all input LR frames. Finally, the plug-and-play module extracts feature F T , which contains spatialtemporal information from all input frames. Further, we summarize the detailed algorithm of the VSR-adapted method with plug-and-play temporal feature extraction module in Algorithm 1. For easy understanding, we divided the loop into multiple ones. Input : Consecutive low-resolution frames y i . i is relative index to the center frame ranging from −n to n. Output : Super-resolution center frame x 0 . // Shallow FE module from SISR model 1 for i = −n, −n + 1, · · · , n do 2 F s,i = FE shallow (y i ) ; 3 end // Offset estimation submodule of temporal FE module 4 for i = −n, −n + 1, · · · , n do 5 F o,i = RB 5 (· · · RB 1 (Conv 1 (y i )) · · · ) ; 6 F o f f ,i = DConv 2 (DConv 1 (Conv 2 (CAT(F o,i , F o,0 )))) ;
Datasets
Following previous studies [12,16,47], we utilized the widely used Vimeo90K dataset for training. This dataset includes videos with different scenarios, such as moving objects, camera motion, and complex scene structures. It consists of 90,000 video clips with a resolution of 448 × 256. As per the official split, we use 64,612 video clips for training. The HR frames of these videos were used as the ground truth. For training, we randomly cropped these HR frames to patches with the size of 256 × 256, and these patches were bicubically downsampled to the size of 64 × 64 using the Matlab function imresize. We randomly flipped and rotated the data during training.
For testing, we evaluated the effectiveness of our proposed model on two public benchmarks, i.e., the Vid4 [48] and SPMC-11 [47]. The quantitative metrics were PSNR [22] and SSIM [23], computed in the luminance (Y) channel. We also cropped 8 pixels near the image boundary, similar to the previous approach [12].
In our implementation of SRResNet [19], we removed all batch norm layers. We used the EDSR baseline [13] with a feature channel count and block count of 64 and 16, respectively. For SwnIR [15], the LR patch size was 48 × 48, and the GT patch size was 192 × 192. We used a smaller patch size for SwinIR for lower memory consumption. The batch size for training all models was 16. We empirically set n = 2, indicating that a VSR-adapted model takes five frames as input. For SISR models, the number of input frames was one. Each SISR model and its VSR-adapted model were trained from scratch using the same setting except for the number of input frames.
We used the mean square error (MSE) as the loss function, defined as Loss = HR − SR 2 . The parameters were updated using the Adam optimizer [49] with β1 = 0.9 and β2 = 0.99. The learning rate was initialized as 1 × 10 −4 and halved for every 1 × 10 5 iterations. We trained the models for 3 × 10 5 iterations. All experiments were implemented in Pytorch and ran on a server with NVIDIA GPUs.
Effectiveness on Different Single-Image Super-Resolution Models
To evaluate the effectiveness of the proposed method, we conducted experiments on five representative SISR models. Table 1 displays the quantitative results on two popular benchmarks. The PSNR and SSIM metrics of VSR-adapted models improved by at least 1.16 dB and 0.036, respectively. It demonstrates that the proposed method works effectively on various SISR models. Moreover, the performance of the VSR-adapted models is positively correlated with the capacity of the original models. In the SISR task, EDSR [13] is better than SRResNet [19] but underperforms RCAN [20] and RDN [21]. The performance of RCAN and RDN is on par, and SwinIR [15] has the best performance. As shown in Table 1, the VSR-adapted models exhibit similar trends. We use the suffix "-VSR" to represent the VSR-adapted models. The performances of SRResNet-VSR and EDSR-VSR are weaker than those of RCAN-VSR and RDN-VSR, and SwinIR-VSR achieves the best results on both benchmarks. Moreover, we computed the PSNR metric on the Vid4 benchmark during training. As illustrated in Figure 4, the VSR-adapted models benefit from the information aggregated from neighboring frames, and they performed better in the early iterations during training. Thus, the proposed method is effective on different SISR models, and the plug-and-play temporal feature extraction module enables the VSR-adapted models to exploit spatial and temporal information.
Further, we visualized the results of the Vid4 and SPMC-11 benchmarks for qualitative comparison. Several processed frames are shown in Figures 5 and 6. We can observe that the VSR-adapted models provide visually appealing results. By contrast, the original SISR models produce blurry SR frames and incorrect textures. Overall, the VSR-adapted models reconstruct results with clearer text, richer textures, and fewer artifacts. Among the results of the VSR-adapted models, SRResNet-VSR and EDSR-VSR produce more artifacts than other VSR-adapted models. This is consistent with the capabilities of original SISR models.
Comparisons with State-of-the-Art Methods
We compared these VSR-adapted models with 10 state-of-the-art VSR algorithms, i.e., STAN [50], EGVSR [39], TOFlow [16], STMN [51], SOF-VSR [17], ST-CNN [44], TDAN [12], D3Dnet [47], FRVSR [34], and WAEN [37]. Table 2 shows the quantitative metrics on the Vid4 and SPMC-11 benchmarks. The values with † are reported in [47]. As shown in Table 2, the VSR-adapted models achieve competitive performance on both Vid4 and SPMC-11 benchmarks. All VSR-adapted models perform better than D3Dnet. Compared with D3Dnet, the SRResNet-VSR and EDSR-VSR achieve comparative performance. The performances achieved by RCAN-VSR and RDN-VSR are between FRVSR and WAEN. Among them, the SwinIR-VSR outperforms all models in terms of PSNR metrics. For a finer quantitative comparison on the Vid4 benchmark, we illustrate the PSNR metric of each frame in Figure 7. For simplicity, we select four models, i.e., TDAN [12], FRVSR [34], EDSR-VSR, and SwinIR-VSR. Compared with TDAN, the EDSR-VSR achieves similar performance. Note that the first two and last two frames show a greater difference between TDAN and EDSR-VSR. Because there is less neighboring information for VSR models to exploit, the VSR models exhibit poor performance at the beginning and end of a video. Compared with FRVSR, the SwinIR-VSR achieved better performance on the Calendar and Walk. As the frame index increases on the Calendar, the gap between SwinIR-VSR and FRVSR becomes smaller. Additionally, the performance of SwinIR-VSR is lower than that of FRVSR after the first five frames on the City. This is because the SwinIR-VSR makes use of neighboring frames in a sliding window scheme while the FRVSR utilizes them in a recurrent scheme. For a qualitative comparison, we compared the VSR-adapted models to SOF-VSR [17], TOF [16], TDAN [12], D3Dnet [47], and FRVSR [34]. As shown in Figure 8, the VSR-adapted models reconstruct visually attractive results. The text on the Calendar is now easier to read and the details of the City are clearer. Additionally, the clothes in the Walk image are more recognizable. Moreover, we observed similar trends in the SPMC-11 benchmark, as illustrated in Figure 9. The quality of the reconstructed results of EDSR-VSR is equivalent to that of the compared methods. The RDN-VSR and RCAN-VSR provide results with better quality. The result of SwinIR-VSR has the least artifacts.
Comparisons of Temporal Consistency
To evaluate the temporal consistency of the proposed method, we generated temporal profiles according to [34] for visualization. As shown in Figure 10, the positions of temporal profiles are highlighted with red lines. The heights of temporal profiles vary due to the video length. As shown in the Calendar, the temporal profiles demonstrate that the original SISR models perform poorly because they are unable to capture temporal information. By contrast, the VSR methods and VSR-adapted models produce results with fewer artifacts. However, inappropriate aggregation of temporal information can lead to degraded results. As illustrated in the City, the original SISR models and our VSR-adapted models exhibit better temporal consistency than VSR models.
Ablation Study
We used EDSR [13] as the baseline in the ablation study to evaluate the effectiveness of the proposed temporal feature extraction module, which consists of offset estimation, spatial aggregation, and temporal aggregation submodules. We evaluated three models to determine the effectiveness of each submodule. The first variation is denoted as Model 1. We fed shallow features from neighboring frames to the spatial aggregation submodule without the support of the offset estimation submodule. The neighboring features were then fused with a convolution using a 1 × 1 kernel. Model 2 is referred to as the second variation. We introduced the offset estimation submodule, which makes use of the center frame and neighboring frames to guide the spatial aggregation. The third variation, denoted as EDSR-VSR, combines all the components, including channel attention and progressive channel shrinking. Table 3 indicates that relying solely on the spatial aggregation submodule does not lead to performance improvement. However, with the support of the offset estimation submodule, there is a significant performance improvement. Furthermore, the temporal aggregation submodule further improved the performance. Three submodules play an irreplaceable role in our presented temporal feature extraction module. To evaluate the efficiency of the proposed method, we conducted a comparison on the Vid4 benchmark. We evaluated three models, i.e., EDSR [13], EDSR-VSR, and EDSR-VSR 2. The EDSR-VSR 2 employs the temporal feature extraction module after the deep feature extraction module. Table 4 shows the performance and average latency of inference. As we can see, the EDSR-VSR is about 1.6× faster than the EDSR-VSR 2. Although the EDSR-VSR is slower than EDSR [13], it reaches 24 frames per second. Specifically, we analyzed the latency of each part of EDSR-VSR. Overall, 0.89% of the latency is consumed by the shallow feature extraction module from the SISR model. The subsequent offset estimation submodule, spatial aggregation submodule, and temporal aggregation submodule occupied 21.25%, 39.99%, and 15.21% of the latency, respectively. Additionally, 22.66% of the time is spent on the deep feature extraction and reconstruction module from the SISR model. Note that the temporal feature extraction module has to process all input frames, so each submodule takes a longer time to complete the computation. Thus, the proposed method balances the accuracy and latency.
Discussion and Limitation
The proposed method builds a bridge between the SISR model and the VSR model. We revisited many SISR models and summarized a common architecture of SISR models. The proposed method leverages the inherent similarities and differences between the two tasks, and the plug-and-play temporal feature extraction module is presented to allow the VSR-adapted model to utilize information from neighboring frames. We applied it to five representative SISR models to evaluate our method, including a generator of GAN [19], three representative SISR models [13,20,21], and a Transformer-based model [15]. Compared with state-of-the-art VSR models, our VSR-adapted models achieve competitive performance.
There are several strong points of the proposed method. Firstly, the proposed architecture of VSR-adapted models provides a novel scheme to develop VSR models. As long as a SISR model follows the common architecture, it can be easily adapted to a VSR model. It reduces the delay of applications of new SISR technologies. Secondly, with the development of VSR, better temporal feature extraction techniques will be proposed, leading to better VSR performance. It divides the development of the VSR model into two independent tasks. Thirdly, the plug-and-play characteristic enables a single model to perform both SISR and VSR tasks.
Although the VSR-adapted models show promising results, we observed some failure cases in experiments. As illustrated in Figure 11, these models fail to recover tiny details. In these cases, the contrast is low in the ground truth, and the contrast is further reduced in LR frames, making SR reconstruction very challenging. Furthermore, all VSR-adapted models fail to provide clear results. Ground Truth Figure 11. The Qualitative Comparison of Details in Low-Contrast Areas.
Conclusions
In this paper, we propose a method for adapting SISR models to the VSR task. For effectiveness on various SISR models, we summarize the common architecture of SISR models. The VSR-adapted models leverage the capability of SISR models to learn the mapping between LR and HR images. Then, the proposed plug-and-play temporal feature extraction module allows VSR-adapted models to access spatial-temporal information.
Thus, the performance in the VSR task is improved by the incorporation of the SISR model and the temporal feature extraction module. The experiments on several SISR models and benchmarks show that VSR-adapted models surpass the original SISR models. The achieved performance is positively related to the capacity of SISR models, indicating the effectiveness of the proposed method. Further, the VSR-adapted models achieved better results than the SOTA VSR models. In the future, we plan to solve the problem of poor performance in low-contrast areas. | 2023-05-27T15:17:40.345Z | 2023-05-24T00:00:00.000 | {
"year": 2023,
"sha1": "15e3a2b7af22d028ee7806ca5f7369c092c53d28",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.3390/s23115030",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "a8774a723c98e7920929dc7c3f125afd750efa75",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": []
} |
246681377 | pes2o/s2orc | v3-fos-license | High-throughput widefield fluorescence imaging of 3D samples using deep learning for 2D projection image restoration
Fluorescence microscopy is a core method for visualizing and quantifying the spatial and temporal dynamics of complex biological processes. While many fluorescent microscopy techniques exist, due to its cost-effectiveness and accessibility, widefield fluorescent imaging remains one of the most widely used. To accomplish imaging of 3D samples, conventional widefield fluorescence imaging entails acquiring a sequence of 2D images spaced along the z-dimension, typically called a z-stack. Oftentimes, the first step in an analysis pipeline is to project that 3D volume into a single 2D image because 3D image data can be cumbersome to manage and challenging to analyze and interpret. Furthermore, z-stack acquisition is often time-consuming, which consequently may induce photodamage to the biological sample; these are major barriers for workflows that require high-throughput, such as drug screening. As an alternative to z-stacks, axial sweep acquisition schemes have been proposed to circumvent these drawbacks and offer potential of 100-fold faster image acquisition for 3D-samples compared to z-stack acquisition. Unfortunately, these acquisition techniques generate low-quality 2D z-projected images that require restoration with unwieldy, computationally heavy algorithms before the images can be interrogated. We propose a novel workflow to combine axial z-sweep acquisition with deep learning-based image restoration, ultimately enabling high-throughput and high-quality imaging of complex 3D-samples using 2D projection images. To demonstrate the capabilities of our proposed workflow, we apply it to live-cell imaging of large 3D tumor spheroid cultures and find we can produce high-fidelity images appropriate for quantitative analysis. Therefore, we conclude that combining axial z-sweep image acquisition with deep learning-based image restoration enables high-throughput and high-quality fluorescence imaging of complex 3D biological samples.
1 Point-by-point response to comments 1
.1 Associate Editor's Comments
Associate Editor: Comments to the Author: When submitting your revision, we need you to address these additional requirements. 1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at https://journals.plos.org/plosone/s/file? id=wjVg/PLOSOne_formatting_sample_main_body.pdf and https://journals.plos.org/plosone/ s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf 2. In your Methods section, please include information on the source or supplier of the cell lines used. Author response: Thank you for this suggestion. We have added the information in bold in the paragraph directly under the Experiments heading: "To establish proof of concept across multiple biological protocols, we tested our proposed workflow on three common protocols for culturing live 3D tumor spheroids, each of which present different imaging challenges. Four common immortalized cancer cell lines were used across the experiments: A549 (lung cancer), MCF-7 (breast cancer), MDA-MB-231 (breast cancer) and SK-OV-3 (ovarian cancer). Commercially available cell lines stably expressing a nuclear-restricted mKate2 FP were used for A549 and MCF-7, while standard SK-OV-3 (EACC) and MDA-MB-231 (ATCC) cell lines were purchased and stably transfected with the Incucyte® NucLight Red lentivirus reagent (EF1 Alpha Promoter; Sartorius) per manufacturer's instructions. Because the FP is nuclear restricted, it can be used to measure cell viability, i.e. a loss of viability results in a compromised nuclear membrane to cause a loss of fluorescence. The sections below describe these three protocols."
General questions
1. Is the manuscript technically sound, and do the data support the conclusions?
The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.
Reviewer 1: Yes Reviewer 2: Partly 2. Has the statistical analysis been performed appropriately and rigorously? Reviewer 1: Yes Reviewer 2: No Author response: In response to these questions 1 and 2, we have added two additional metrics (MS-SSIM and MSE) to the results to further support our conclusions. During the process of calculating these new metrics, we also found a bug in our metric calculation script for the single spheroid data and have updated the metrics accordingly.
3. Have the authors made all data underlying the findings in their manuscript fully available?
The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data-e.g. participant privacy or use of data from a third party-those must be specified.
Reviewer 1: No Reviewer 2: No
Author response: Thank you for noticing this. While we previously only included the repository information in our submission letter, we have now added a sentence to the Experiment section where we point the reader to a link on figshare where the data is available (https://figshare. com/projects/Dataset_of_fluorecent_3D-samples_projected_to_2D/126629).
Is the manuscript presented in an intelligible fashion and written in standard English?
PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.
Reviewer 1: Yes Reviewer 2: No Author response: We noted a number of grammatical mistakes and typos in our previous submission, and have gone through and corrected them to the best of our ability. We hope that our manuscript will now fulfill the language requirements for the journal.
Reviewer 1
1. Comparison study is not provided. Author response: We have added a number of additional references in the Introduction (see response to comment 1 from Reviewer 2) to studies that are also combining deep learning methods with fluorescent imaging. A direct comparison study is hard to come by since our acquiring method of z-sweep without hardware modifications or several z-slices has not been done before. We have made this clearer by stating that this gap exists in the field of fluorescent imaging of complex 3D samples.
2. As you mentioned that 96 images were used for research. It is not available in public. Author response: Thank you for acknowledging this. We have added a sentence in the Experiment section where we point the reader figshare https://figshare.com/projects/Dataset_of_ fluorecent_3D-samples_projected_to_2D/126629 where the data is available.
3. The deep learning model techniques, results need to compare. Author response: We have expanded on the metrics in Table 1 to further compare the deep learning models results. Also, as mentioned in comment 1 above, we have added references containing similar works. Although no direct comparison study could be found. 4. Provide in detail about OSA-CGAN. Author response: We thank you for this comment, as it made us realized that we did properly define the OSA-CGAN architecture. This has been remedied and is now addressed in the whole last paragraph in the Conditional Generative Averserial Network section. We hope this addition more clearly and explicitly defines both the OSA-CGAN and U-Net models: "We investigated the use of two different generators for our CGAN architectures, an U-Net generator and an OSA-U-Net variant, both described in section Image Generator. The corresponding architectures are refereed to as CGAN (U-Net generator) and OSA-CGAN (OSA-U-Net generator) respectively and both versions uses the PatchGAN discriminator described in above section." 5. Overall the idea was good. Author response: Thank you!
Reviewer 2
Comments to the Author Widefield fluorescence microscopy is commonly used to study biological phenomena. However, fluorescent microscopic imaging of complex 3D samples, such as tumor spheroids, is burdensome since it often entails the collection of a stack of multiple optical sections along the z-dimension, which is time-consuming and has high risk of causing phototoxicity. To alleviate this problem, we propose a workflow combining axial z-sweep acquisition and deep learning-based image enhancement. This paper is novel and the contributions are good for a journal article. The revised version of the paper may be considered for publication in this journal.
1. The literature of the paper is poor, the authors may be consider the most recent articles for literature. Author response: We have added the information in bold in the Introduction: "Whereas traditional PSF-based image restoration is incredibly challenging for complex 3D samples, convolutional neural networks (CNNs) are now routinely used for such difficult computer vision problems. Across various 3D fluorescence imaging studies, CNNs are used to segment nuclei [11,12] and cell bodies [13,14], image restoration [15,16], image super-resolution [17,18], as well as blind deconvolution in 2D widefield fluorescence imaging [19,20]. In the 3D fluorescence imaging space, recent studies have outlined methods to use CNNs for virtual refocusing to predict a user-defined 3D surface from a single plane 2D fluorescent image [21,22], using recurrent neural networks (RNNs) to reconstruct a complete 3D volume with far fewer optical sections than typically needed [23], and optimization of phase mask filters for PSF-engineering to capture 2D extended depth-of-field images or 3D volumetric images [24,25]. While these notable advancements demonstrate the potential of CNNs to increase throughput for 3D fluorescent imaging, they have primarily been limited to high-resolution, subcellular imaging. These studies do not address the need for improving throughput capabilities for imaging large 3D models, such as tumor spheroids and organoids, which can be larger than a millimeter in diameter. Therefore, there exists a gap in the field whereby high-throughput fluorescent imaging of large, complex 3D samples can be accomplished without complicated hardware or acquiring multiple z-slices of samples." 2. The reasons to achieve the superior performance of the article may be included in the revised version of the article. Author response: We have clarified the advantage of our approach in the first paragraph of the discussion: ". . . Notably, for our embedded tumor multi-spheroid samples, which required acquisition of data across a 1.5mm volume, the conventional z-stack approach took 101 seconds for a single sample whereas the z-sweep only took 1.2 seconds, nearly achieving a 100-fold speedup. . . " We have also expanded on the reasons why neural networks are suitable for this task in the third paragraph of the discussion as: "Regardless of the neural network architecture used, images that result from a z-sweep acquisition are well-suited for a deep learning based restoration workflow. While indeed appearing low in quality, z-sweep images effectively represent the fluorescent intensity of the sample integrated over the z-axis. Therefore, we speculate that the z-sweep sufficiently provides a neural network with enough real information about the sample to accurately predict the restored unintegrated fluorescent signal." 3. List the limitations of the proposed work. Author response: The limitations of the proposed work is expanded on in comment 2 above where the acquisition speed is 1.2 seconds per sample. The almost 100-fold speedup compared to the z-stack approach is a big step in the right direction but still requires 1.6 minutes per 96 well plate. This is enough for a significantly higher throughput but one can always aim for even faster acquisition in the future. Additionally, in comment 4 below we expand on the post processing of the images with the proposed work. Furthermore, we have trained the model on one of the three common types of experimental preparations for culturing spheroids we consider and have noted a transferability across these three different cell cultures. This is surely promising for further exploration of the z-sweep approach on other types of cell cultures, FPs, organoids etc. but not a guarantee. This is also discussed in the Discussion. 4. Discuss the complexity of the proposed model and how efficient it is over the existing ones. Author response: A z-sweep is acquired in 1.2 seconds compared to the 101 seconds a z-stack takes to acquire resulting in an almost 100-fold speedup. Therefore, the big advantage in our approach is the acquisition of the images. However, the trained DL-model does not cause a bottleneck either. An OSA-CGAN model evaluates a z-sweep image in 0.33s on a laptop equipped with an NVIDIA Quadro RTX 3000 and requires 4015MiB VRAM. Running the same model on an Intel Core i7-CPU @ 2.30GHz requires 10s per image. 5. Provide the dataset details or citations. Author response: Thank you for acknowledging this. We have added a sentence in the Experiment section where we point the reader figshare https://figshare.com/projects/Dataset_of_ fluorecent_3D-samples_projected_to_2D/126629 where the data is available.
6. There are several performance metrics in the literature, whereas the authors consider only few in the paper. Author response: We have considered two additional metrics from the literature which we consider relevant. MS-SSIM to extend the analysis on multiple scale and capture information about details from multiple resolutions and MSE since it's one of the most common metrics to asses the difference of individual pixels. These are added to Table 1 in the results as well as discussed in the Discussion. | 2022-02-10T14:17:08.380Z | 2022-02-07T00:00:00.000 | {
"year": 2022,
"sha1": "70a95a64e98476f1337c77f8dc2d2bbfd4068b72",
"oa_license": "CCBY",
"oa_url": "https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0264241&type=printable",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "8632ba1add3d29fd55fcc8d9fe26f77f22bff602",
"s2fieldsofstudy": [
"Computer Science",
"Engineering",
"Materials Science"
],
"extfieldsofstudy": [
"Medicine",
"Biology"
]
} |
119124132 | pes2o/s2orc | v3-fos-license | Remarks on the nonexistence of biharmonic maps
In this short note we study nonexistence result of biharmonic maps from a complete Riemannian manifold into a Riemannian manifold with nonpositive sectional curvature. Assume that $\phi:(M,g)\to (N, h)$ is a biharmonic map, where $(M, g)$ is a complete Riemannian manifold and $(N,h)$ a Riemannian manifold with nonpositive sectional curvature, we will prove that $\phi$ is a harmonic map if one of the following conditions holds: (i) $|d\phi|$ is bounded in $L^q(M)$ and $ \int_M|\tau(\phi)|^pdv_g<\infty, $ for some $1\leq q\leq\infty$, $1<p<\infty$; or (ii) $Vol(M)=\infty$ and $ \int_M|\tau(\phi)|^pdv_g<\infty, $ for some $1<p<\infty$. In addition if $N$ has negative sectional curvature, we assume that $rank\phi(q)\geq2$ for some $q\in M$ and $\int_M|\tau(\phi)|^pdv_g<\infty, $ for some $1<p<\infty$. These results improve the related theorems due to Baird et al.(cf. \cite{BFO}), Nakauchi et al.(cf. \cite{NUG}), Maeta(cf. \cite{Ma}) and Luo(cf. \cite{Luo}).
Introduction
Let (M, g) be a Riemannian manifold and (N, h) a Riemannian manifold without boundary. For a W 1,2 (M, N ) map φ, the energy density of φ is defined by e(φ) = |∇φ| 2 = Tr g (φ * h), where φ * h is the pullback of the metric tensor h. The energy functional of the mapping φ is defined as The Euler-Lagrange equation of E is τ (φ) = Tr g∇ dφ = 0 and τ (φ) is called the tension field of φ. A map is called a harmonic map if τ (φ) = 0. The theory of harmonic maps has many important applications in various fields of differential geometry, including minimal surface theory, complex geometry and so on(cf. [15]). Much effort has been paid in the last several decades to generalize the notion of harmonic maps. In 1983, Eells and Lemaire [4](see also [5]) proposed to consider the bienergy functional E 2 (φ) = 1 2 M |τ (φ)| 2 dv g of smooth maps between Riemannian manifolds. Stationary points of the bienergy functional are called biharmonic maps. We see that harmonic maps are biharmonic maps and even more, minimizers of the bienergy functional. In 1986, Jiang [9] derived the first and second variational formulas of the bienergy functional and studied biharmonic maps. The Euler-Lagrange equation of E 2 is , ∇ is the Levi-Civita connection on (M, g) and∇ is the induced connection on the pull back bundle φ −1 T N , and R N is the Riemannian curvature tensor on N .
The first nonexistence result of biharmonic maps was obtained by Jiang [9]. He proved that biharmonic maps from a compact, orientable Riemannian manifold into a Riemannian manifold of nonpositive curvature are harmonic. Jiang's theorem is a direct application of the Weitzenböck formula. If φ is biharmonic, then The maximum principle implies that dτ (φ) = 0 and so by we deduce that div dφ, τ (φ) = |τ (φ)| 2 , then integration by parts, we have τ (φ) = 0. If M is noncompact, the maximum principle is no longer applicable. In this case we can use the integration by parts argument, by choosing proper test functions. Based on this idea, Baird et al.(cf. [2]) proved that biharmonic maps from a complete Riemannian manifold with nonnegative Ricci curvature into a nonpositively curved manifold with finite bienergy are harmonic. It is natural to ask whether we can abandon the curvature restriction on the domain manifold and weaken the integrable condition on the bienergy. In this direction, Nakauchi et al.(cf. [13]) proved that biharmonic maps from a complete manifold to a nonpositively curved manifold are harmonic if(p = 2) (i) M |dφ| 2 dv g < ∞ and M |τ (φ)| p dv g < ∞, or (ii) V ol(M, g) = ∞ and M |τ (φ)| p dv g < ∞.
Later Maeta(cf. [12]) generalized this result by assuming that p ≥ 2. In this paper, we will further generalize this result to the following: then φ is harmonic.
then φ is harmonic.
For a better understanding of theorem 1.1 the readers could consult the papers [1] and [10] for examples of proper biharmonic maps(that biharmonic maps which are not harmonic).
We must point out that part (ii) of theorem 1.1 is in fact implictly contained in the proof of Theorem 3.1 of [2], which is obviously not realized by the authors of [13] and [12]. That is also the motivation for us to search further in this direction.
When the target manifold has negative sectional curvatures, we have This theorem was proved by Oniciuc(cf. [14]) under the assumption of |τ (φ)| is a constant.
Remark 1.4. Theorem 1.3 is a generalization of Theorem 1.3 in [11]. But the right statement of Theorem 1.3 in [11] should be added an additional assumption of rankφ(q) ≥ 2 at some point q ∈ M .
The rest of this paper is organized as follows: In section 2 we give some preliminaries on harmonic maps and biharmonic maps. In section 3 our theorems are proved. In section 4 we give some applications of our results to biharmonic submersions.
Harmonic maps and biharmonic maps
In this section we give more details on the definitions of harmonic maps and biharmonic maps.
Let φ : (M, g) → (N, h) be a map from an m-dimensional Riemannian manifold (M, g) to an n-dimensional Riemannian manifold (N, h). The energy of φ is defined by where we denote by ∇ the Levi-Civita connection on (M, g) and by∇ the induced Levi-Civita connection on φ −1 T N . τ (φ) is called the tension field of φ.
To generalize the notion of harmonic maps, in 1983 Eells and Lemaire [4](see also [5]) proposed to consider the bienergy functional In 1986, Jiang [9] calculated the first and second variational formulas of the bienergy functional. The E-L equation of E 2 is where {e i , i = 1, ..., m} is a local orthogonal frame on M and R N is the Riemann curvature tensor of (N, h). τ 2 (φ) is called the bitension field of φ. A map φ : (M, g) → (N, h) is called a biharmonic map if τ 2 (φ) = 0.
Gaffney's theorem
In the subsequent section we will use the following Gaffney's theorem(cf. [7]).
Biharmonic maps into nonpositively curved manifolds
In this section we will prove theorem 1.1 and theorem 1.3. First let's prove a lemma. Proof. Here most part of the proof is the same with that of theorem 3.1 in [2]. For the completeness of this paper and the convenience of the readers we give all the details here. Let ǫ > 0 and a direct computation shows that Therefore we obtain Since φ is biharmonic, from the biharmonic equation we see that where we used the assumption that R N ≤ 0. Combining the above two inequalities we obtain From inequalities (3.1) and (3.4) we deduce that Then we have where in the second equality we used∇τ (φ) = 0. Now by Gaffney's theorem we have that which implies that c = 0, a contradiction. Therefore we must have c = 0, i.e. φ is a harmonic map. This completes the proof of theorem 1.1. ✷ Proof. of theorem 1.3. By lemma 3.1, |τ (φ)| = c is a constant. We only need to prove that c = 0. Assume that c = 0, we will get a contradiction. Then by the biharmonic equation and the Weitzenböck formula we have at q ∈ M : where in the first and fourth equalities we used lemma 3.1. Since the sectional curvatures of N is negative, we must have that dφ(∂ x i )//τ (φ) at q ∈ M ∀i, i.e. rankφ(q) ≤ 1. Therefore we must c = 0, a contradiction. This completes the proof of theorem 1.3. ✷
Biharmonic submersions into nonpositively curved manifolds
In this section we give some applications of our result to biharmonic submersions. First we recall some definitions(cf. [3]). Assume that φ : (M, g) → (N, h) is a smooth map between Riemannian manifolds and x ∈ M . Then φ is called horizontally weakly conformal if either A map φ is called horizontally weakly conformal on M if it is horizontally weakly conformal at every point of M . If furthermore, φ has no critical points, then we call it a horizontally conformal submersion. Note that if φ : (M, g) → (N, h) is a horizontally weakly conformal map and dimM < dimN, then φ is a constant map.
If for every harmonic function f : V → R defined on an open subset V of N with φ −1 (V ) non-empty, the composition f • φ is harmonic on φ −1 (V ), then φ is called a harmonic morphism. Harmonic morphisms are characterized as follows(cf. [6,8]). This proposition generalizes (i) of proposition 4.2 in [11]. | 2016-04-04T10:29:08.000Z | 2015-11-23T00:00:00.000 | {
"year": 2016,
"sha1": "2760fe42a84802825fd32e51aea48ea10a35b3ab",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/1511.07231",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "2760fe42a84802825fd32e51aea48ea10a35b3ab",
"s2fieldsofstudy": [
"Mathematics"
],
"extfieldsofstudy": [
"Mathematics"
]
} |
246942416 | pes2o/s2orc | v3-fos-license | Analysis of the efficacy of a modified posteromedial approach for Klammer III posterior Pilon fractures
Purpose To analyze the curative effect and technical points of a modified posteromedial approach in the treatment of Klammer III posterior Pilon fracture. Methods A retrospective analysis of patients with Klammer III posterior Pilon fractures were conducted in our department from January 2018 to December 2019. Before the surgery, the patients were fully relieved of swelling and pain, and a comprehensive examination was carried out. The posteromedial approach exposed the posterior and medial fracture block of the distal tibia. According to the fracture of external malleolus, it is determined whether to combine a lateral incision and protect tendons and vascular nerves by a retractor, and then perform a fracture reduction and internal fixation. Postoperatively, the patients were treated with analgesia, detumescence, anticoagulation and rehabilitation exercise. The American orthopaedic foot and ankle society (AOFAS) score and visual analogue score were recorded at regular follow-up after surgery. A t-test was used for the comparison of the preoperative and final AOFAS score. Results There were 7 male and 13 female (n = 20) included in the study, aged 22 to 88 years (average age 54.2 years). The injury mechanisms were falling from a height (n = 7), traffic accident (n = 6), walking injury (n = 2) and heavy injury (n = 5). The postoperative follow-up duration was 12–24 months (mean 16.95 months). The AOFAS score of the 20 patients before and after surgery were compared. The preoperative AOFAS score was 38.90 ± 3.91, and the final AOFAS score was 80.55 ± 4.20, (p < 0.001). The mean final visual analogue scores at rest, active and weight-bearing walking were 0.30, 0.85 and 1.70, respectively. One patient reported poor postoperative wound healing and required a return to hospital for debridement and anti-infection treatment. Conclusion In the treatment of Klammer III posterior Pilon fractures, the modified posteromedial approach can fully expose the fracture block and the collapsed articular surface of the medial malleolus, achieve good reduction and internal fixation with limited injury of the tendon and vascular nerves, and have a better prognosis.
Introduction
Trimalleolar fracture is a clinically common type of traumatic fracture. The mechanism of damage is rotational violence. With the development of the construction industry, ankle lesions caused by falling are increasing. A special type of trimalleolar fracture is caused by a vertical merger force, as a result of which the fracture line of the posterior malleolus along the distal tibia coronal surface extends to the medial malleolus hill. Often ankle joint dislocations and injuries of the ankle joint cartilage surface are merged. Modern orthopaedic medicine has named these "posterior Pilon fractures". 1 The Klammer classification is commonly used to clinically evaluate such fractures. 2 The type I fracture involves the entire posterior malleolus and has a long oblique base towards the posterolateral side; the type II fracture line of the posterior malleolus extends to the inner side of the posterior colliculus; the type III consists of separate medial and lateral bone masses, involving the posterior and anterior hills of medial malleolus, which are accompanied by dislocation and serious soft tissue damage. 3 Because posterior Pilon fractures extend both backwards and forwards, the ankle commonly used in the posterolateral approach cannot be fully exposed. 4,5 Since 2013, our department has used the modified posteromedial approach to handle such fractures.
The author conducted a retrospective analysis of 20 patients with Klammer III posterior Pilon fractures who underwent improved internal fixation surgery in the author's department from January 2018 to December 2019. The summary analysis is as follows.
Clinical data
The general information Inclusion criteria were: (1) meeting post-Pilon fracture diagnostic criteria; (2) meeting the Klammer III fracture classification; (3) preoperative and postoperative imaging data were available; (4) fully reducing swelling and improving soft tissue situation; (5) surgery at 3 days to 3 weeks from the time of injury; and (6) duration of follow-up at least 18 months. Exclusion criteria were: (1) the surgery was 3 weeks or < 3 days from the time of injury; (2) the follow-up time was less than 1 year; (3) patients were not able to tolerate surgery because of system diseases (e.g severe cardiopulmonary insufficiency, severe autoimmune diseases, severe coagulation dysfunction, etc.).
Twenty patients were aged 22 to 88 years (average 54.2 years), 7 cases were men and 13 were women. Among them, 7 patients fell from high places, 6 sustained injuries in traffic accidents, 2 sustained injuries while walking, and 5 were injured by heavy objects.
Postoperative functional evaluation
According to the scoring system of the American association of foot and ankle society (AOFAS), combined with preoperative and postoperative imaging data, and some aspects of patient's pain, ankle joint activity, shoe requirements, maximum walking distance, ground requirements and gait feedback, it is comprehensively scored with 7 rating categories. 6 The score of 90e100 points as excellent, 75 -89 points as good, 50 -74 points as available, and < 50 points as poor. The healing quality of fracture reset is determined according to picture archiving and communication system imaging evaluation. 7 A visual analogue scale was used to evaluate ankle joint pain when walking and resting. Regular follow-ups were carried out at 4e6 weeks, 12e16 weeks, 6 months, 1 year and 18 months.
Statistical analysis is performed by means of SPSS 13. The ankle and back foot scoring criteria of AOFAS and are represented by the X ± SD before and after follow-up. The two-mean comparison is based on t-test analysis and p < 0.05 is considered statistically significant.
Surgical techniques
According to the symptoms and fracture displacement, all patients were treated with manual reduction and plaster fixation or calcaneal traction before surgery, then admitted to the hospital, and given symptomatic treatment such as ice compresses, detumescence and pain relievers. In patients with serious soft tissue injuries, such as exudate or hematoma blisters, the wounds will be rinsed and partially sutured in the first stage. Surgical treatment would be postponed until the swelling had subsided completely.
For patients with open commuted Pilon fractures or with fracturedislocations of the ankle, an external fixator was preferred to stabilize the fracture after a preliminary manual reduction, which facilitated the protection of soft tissue and secondary internal fixation.
Intraoperative anesthesia was performed using a laryngeal mask combined with a lower limb nerve block on the affected side. The patient was placed in the prone position, and a balloon tourniquet was used, then performing routine disinfection and laying a towel. For patients with fibula fractures, the distal lateral approach for open reduction and internal fixation was used first. A lateral longitudinal incision of about 6e8 cm was made to treat the lateral malleolus fracture (Fig. 1A). A longitudinal incision for a modified posteromedial approach was extended distally along the medial side of the Achilles tendon (AT), for about 8e10 cm in length (Fig. 1B). The soft tissue was separated layer by layer after the skin tissue was cut open, and the flexor hallucis longus (FHL) tendon was used as a "protective sleeve" of the vascular nerve bundle. The FHL tendon was pulled medially along with the vascular nerves, while the AT was pulled laterally to expose the distal tibia, posterolateral bone mass, and part of the posteromedial bone mass of the posterior malleolus (Fig. 1C). The FHL tendon and the neurovascular bundle together with the AT were pulled together to expose the medial bone mass of the posterior malleolus (Fig. 1D). The direction of pull force was determined by the size of the medial and lateral bone mass of the posterior malleolus. It also formed a "double window" exposure approach to assist with the reduction and fixation. The hematoma and bone fragments between the fracture sites were cleaned. Posterior Pilon fractures usually involve the posterior colliculus of the medial malleolus and are often broken into posterolateral and medial segments, resulting in larger and deeper bone mass. In order to move the periosteum as little as possible, the posterolateral bone block was turned outwards and laterally, and the posteromedial bone block was turned inwards and backwards to avoid injuring the inferior posterior tibiofibular ligament and deltoid ligament. After opening and cleaning the bone, the damage to the articular surface of the distal tibia and the posterior articular surface of the talus was assessed. Anatomical reduction was performed on the posterior malleolus bone block and temporarily fixated with a 1.5 mm K-wire. If there was compression or collapse of the articular surface, a small periosteum stripping device was used to pry the collapse gently. The C-arm was used to confirm that the articular surface of the distal tibia was smooth during the operation. After that, 2 locking support plates in appropriate size were selected to fix the bone blocks respectively. The incision was sutured layer by layer, the negative pressure drainage tube was inserted, and the incision was bandaged.
Postoperative management
To prevent infection, reduce swelling and relieve pain, anticoagulants were administered after surgery. Patients were further encouraged to move their toes and carry out functional training on the progressive ankle. Non-weight-bearing joint activity began at 2e3 weeks after surgery, and weight-bearing training began at 6e8 weeks. Some patients were assisted with weight-bearing training or orthopaedic shoes for support. Regular follow-up after surgery was necessary to monitor bone healing by X-ray and to adjust the training plan.
Results
Postoperative telephone visits and hospital visits were performed for 12e24 months (mean 16.95 months). According to the picture archiving and communication system imaging evaluation criteria, the ankle surface of the all 20 patients met the standards for anatomical reduction. The medium foot and ankle-hind foot scales of AOFAS were used to evaluate the efficacy of the foot and -bearing walking 1 24 0 1 2 37 83 2 20 0 0 1 40 88 3 22 0 1 2 42 78 4 20 0 0 1 44 84 5 21 1 1 3 38 77 6 18 1 1 2 35 80 7 19 1 2 3 32 76 8 20 0 0 0 39 87 9 18 0 1 2 43 79 10 15 0 1 2 41 78 11 ankle functions of patients. According to Table 1, the results of 18 cases were good and 2 cases were adequate. The preoperative AOFAS score of the 20 patients with the long-term follow-up was 38.90 ± 3.91, and the final score was 80.55 ± 4.20, which was higher than that of the preoperative score (p < 0.01), indicating a statistically significant difference (Table 2 and Fig. 2). The final visual analogue score at rest, active and walking was 0e1, 0e2 and 0e3, with an average score of 0.30, 0.85 and 1.70, respectively. One case of poor wound healing due to infection had to return to hospital for debridement and anti-infection treatment (previous history of diabetes mellitus, poor blood glucose control, and failure to follow doctor's advice to take care of the incision on return home after surgery). Postoperative screw loosening occurred in 1 case, and the loose screw was re-fixed in hospital. There were no other postoperative complications, such as bone nonunion, implant foreign body reaction, and so on.
Typical case (Figs. 3e6)
A 32-year-old female was admitted to the hospital due to swelling and pain in her right ankle because of traffic accident occurred 5 h ago. The patient suffered a trimalleolar fracture combined Klammer III posterior Pilon fracture. Open reduction and internal fixation was performed 5 days after admission with a modified postromedical and lateral approach in the prone position. The tibiofibular screw was removed 8 weeks after surgery. After 1 year of follow-up, the final AOFAS score was 87 points.
Discussion
Ankle joint fracture is a common fracture type in clinical settings, commonly occurring in sports field or accidents, which accounts for 9% of all fractures. 8 A person suffers from an ankle fracture, always relating to severe pain, swelling, deformity and secondary subcutaneous congestion. About 46% of the ankle fractures are Weber B or C type ankle fractures and dislocations of posterior distal tibial fracture, 9 which are considered to be posterior malleolus fracture or posterior Pilon fracture. 10,11 Damage mechanism and imaging features of posterior Pilon fracture The concept of a Pilon fracture was first proposed by Destot in 1911, 12 and the concept of the posterior Pilon fracture was first proposed by Hansen in 2000. 1 These refer to intra-articular fractures of the distal posterior tibia caused by the combination of a vertical compressive force and a horizontal rotational force. Gardner et al. 13 believed that the posterior Plion fracture was caused by a vertical force in the plantar-flexion position of the ankle, Wang et al. 14 suggested that the combined effect of a rotational and vertical force in the plantar-flexion position of the ankle joint can lead to a medial ankle bone block after it separated from the medial malleolus. When the ankle joint was hit by a strong vertical and rotational combined violence in the protonation position, the highenergy injury can cause the fracture block and fracture line of the posterior ankle to extend to the medial malleolus. All these posterior malleolus fracture has no direct relation with medial Pilon fractures, which is not rare clinically. Forberger et al. 16 found a significant proximal displacement in the posterior half of the talus. On anteroposterior and lateral radiographs of the ankle, the fracture line can be seen at the posterior margin of the tibia, with or without compression leading to a collapse of the distal articular surface of the tibia. The characteristic image of the posterior Pilon fracture can be seen in the medial malleolus, which is called "bilateral sign", but neither the anteroposterior nor lateral radiographs can clearly show the direction of the fracture line, as well as the size and number of fracture fragments. In such instances, a CT examination should be performed after the presence of a posterior Pilon fracture has been confirmed by X-ray. Switaj et al. 17
Modified posteromedial approach
For Klammer type III posterior Pilon fractures, the lack of ligaments in the medial posterior malleolus bone block means that the posterolateral approach requires removal of the posterior medial soft tissue, resulting in large soft tissue injury and difficulty in reduction and fixation. The improved posteromedial approach, combined with the advantages of the posteromedial approach and the posterolateral approach is more effective. The incision was made along the inner margin of the AT, and the FHL tendon was served as a "protective sleeve" of the vascular nerve plexus. The FHL tendon and the vascular nerve bundle were lifted and pulled medially, and the internal and posterolateral bone blocks were measured after full exposure. With this approach, the entire posterior end of the distal tibia is in the field of vision, and the direction of the fracture line, the number of fracture fragments and the displacement can be clearly observed. 17,18 Since the fracture can be corrected and reduced under direct vision, the uneven articular surface can be better reduced, thereby reducing the incidence of postoperative complications such as traumatic arthritis. 19 For the collapse of the articular surface, the posterior malleolus bone block can be gently pried up and the articular surface can be reduced by using a lever under direct vision. Therefore, for trimalleolar fractures combined with posterior Pilon fractures, we adopted the posteromedial approach combined with the lateral approach for reduction and fixation, the lateral approach for the fibula fracture, and the posteromedial approach for the posterior bone mass (both posteromedial part and posterolateral part). The effect of reduction and fixation was good, and a sufficient skin bridge was beneficial to the healing of the postoperative incision.
In summary, posterior Pilon fractures are severe ankle fractures caused by a combination of rotational and vertical forces. The Klammer III type covers the entire posterior malleolus, and the fracture line extends even to the anterior colliculus of the medial malleolus. Surgical treatment for this kind of fracture is difficult and cannot be accomplished effectively by the previous surgical approach used. 20 A modified posteromedial approach exposes the posterior internal and external fracture fragments and completes reduction smoothly, reducing soft tissue injury and postoperative complications. In this paper, 20 patients with Klammer III type fractures were included and the modified posteromedial approach was adopted to complete the exposed reduction and internal fixations. Through follow-up statistics, the postoperative AOFAS scores were found to be significantly improved compared with the preoperative scores. Referring to the literature at home and abroad, the incidence of postoperative complications of traditional posterolateral approach combined with the medial approach is 4%e 13%, and the low incidence of the modified posteromedial approach was similar to the traditional approach, further indicating that this surgical technique is worthy of promotion.
Ethical statement
This study was approved by the Fujian Provincial Hospital in Fuzhou.
Declaration of competing interest
All authors confirm that they have no financial and personal relationships with any commercial party. No financial relationships with any organizations that might have an interest in the submitted work in the previous five years.
Author contributions
Zheng-Chao Zhang: Manuscript data Collection and analysis, manuscript writing and revision. Wu-bing He: Control the theory, surgical techniques and put forward specialist guidance. Hao Lin: Data analysis and English writing guidance. | 2022-01-22T16:44:57.391Z | 2022-01-01T00:00:00.000 | {
"year": 2022,
"sha1": "e413f38382256e3d3555b52df5dd7abcbc747db6",
"oa_license": "CCBYNCND",
"oa_url": "https://doi.org/10.1016/j.cjtee.2022.01.003",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "4f25cfc2841e15cf99c50ecf56434e5d377f8a06",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
253129328 | pes2o/s2orc | v3-fos-license | Antimicrobial Peptides Mediate Apoptosis by Changing Mitochondrial Membrane Permeability
Changes in mitochondrial membrane permeability are closely associated with mitochondria-mediated apoptosis. Antimicrobial peptides (AMPs), which have been found to enter cells to exert physiological effects, cause damage to the mitochondria. This paper reviews the molecular mechanisms of AMP-mediated apoptosis by changing the permeability of the mitochondrial membrane through three pathways: the outer mitochondrial membrane (OMM), inner mitochondrial membrane (IMM), and mitochondrial permeability transition pore (MPTP). The roles of AMPs in inducing changes in membrane permeability and apoptosis are also discussed. Combined with recent research results, the possible application prospects of AMPs are proposed to provide a theoretical reference for the development of AMPs as therapeutic agents for human diseases.
Introduction
In recent decades, the incidence and mortality of malignant tumors have remained high, and there are difficulties, such as low treatment efficiency and drug resistance [1]. In addition, the emergence of resistant pathogenic microorganisms might cause a variety of infectious diseases [2], which seriously threaten human health. Antimicrobial peptides (AMPs) have good inhibitory effects on bacteria, fungi, viruses, and even tumor cells, without altering drug resistance [3,4]. This advantage provides a prerequisite for the development of AMPs as therapeutic agents. Previous studies have shown that AMPs can not only directly act on cell membranes and cell walls to kill cells, but can also enter cells and target organelles to achieve bactericidal effects by interfering with normal metabolic processes and playing an immunomodulatory role [5][6][7]. Many AMP drugs have been developed for clinical treatment, such as lipopeptide A21978C for the treatment of skin infection and sepsis and peptide Glutoxim/NOV-002 for the treatment of tuberculosis and non-small-cell lung cancer [8]. To address the resistance of antibiotics, the design and production of effective AMP preparations has become inevitable. However, with changes in the structure, hydrophobicity, and amphipathic properties of AMPs, the mechanism and target of action would also change [5,9]. The therapeutic mechanism of AMPs that exerts antitumor or antifungal effects through the cell membrane or wall is well-understood [10,11]. Nevertheless, studies of AMPs that cause apoptosis through the intracellular pathway are still insufficient. Therefore, sorting out the mechanism of AMPs for the treatment of tumors and infectious diseases is convenient for screening AMPs for better efficiency and to develop them as therapeutic drugs.
Through the investigation of the mechanism of action of different AMPs, it was found that some AMPs cause damage to mitochondria when exercising their biological functions [12][13][14]. Mitochondria is a site for energy production in cells, playing an important role in regulating cell growth and death [15]. The integrity of the outer and inner mitochondrial membrane is crucial for mitochondrial apoptosis, and even cell apoptosis. Changes in mitochondrial membrane permeability can directly activate the mitochondrial apoptosis pathway [16,17]. Although there are abundant reports on the effects of AMPs on mitochondria, most of the research results are merely descriptions of the phenomena, such as mitochondrial membrane potential depolarization and oxidative stress, causing mitochondrial dysfunction [18]. Only few reports have studied the effect of AMPs on mitochondrial permeability transition pore (MPTP) in detail [19], and there are no detailed classifications and mechanism analyses of AMPs on mitochondrial membrane damage. In fact, in the process of the endogenous pathway of apoptosis (the mitochondrial pathway), both the outer and inner membranes of mitochondria are damaged to varying degrees [20]. Combined with the current research reports, the damage to mitochondrial integrity is simply divided into three categories: mitochondrial outer membrane permeability (MOMP), mitochondrial inner membrane permeability (MIMP), and mitochondrial permeability transition (MPT). Separately discussing the effects of AMPs on mitochondrial membranes, especially those targeted at mitochondria [21], can provide a clearer understanding of their therapeutic mechanisms and facilitate the study of the mechanisms of unknown AMPs. This paper systematically discusses the ways in which AMPs alter mitochondrial membrane permeability. The effect of AMPs during apoptosis will be analyzed from the perspective of MOMP, MIMP, and MPT. Some structural features of AMPs targeting the mitochondrial outer and inner membrane are also discussed, which may provide new ideas for the artificial modification of AMPs.
AMPs and Apoptosis
As a regulated and active programmed death process, apoptosis is very important for the body to maintain homeostasis [22]. There are three main mechanisms regulating apoptosis: the mitochondrial pathway, the death receptor pathway, and the endoplasmic reticulum pathway [14,23]. AMPs are a class of small molecular peptides with no more than 100 amino acids [24]. Most AMPs exert their antimicrobial activity by disrupting cell membranes [25], and some AMPs with cell-penetrating ability can exert antitumor or antifungal activities in cells by different mechanisms. For example, they can bind nucleic acids and inhibit DNA or RNA synthesis to kill bacteria [26]. Alternatively, they can inhibit enzyme activity in the process of nucleic acid and protein anabolic metabolism [7]. In fact, we observed that AMPs that enter the cell or exert their effects are often able to achieve programmed cell death by mitochondrial swelling, outer mitochondrial membrane (OMM) rupture, and by stimulating the activation of apoptotic markers ( Table 1). As mentioned above, mitochondrial membrane damage is inevitable during the mitochondrial apoptotic pathway. Therefore, it is necessary to describe the mechanism of AMP-induced apoptosis from the perspective of mitochondrial membrane damage.
AMPs and MOMP
MOMP refers to the process by which proteins existing in the intermembrane space (IMS) (cytochrome c and apoptosis-inducing factor) enter the cytoplasm through the OMM, activate the apoptotic pathway, or trigger inflammatory reactions, thereby causing mitochondrial and cell apoptosis [45][46][47]. MOMP, which does not involve the inner mitochondrial membrane (IMM) and MPTP, is mainly regulated by members of the Bcl-2 family proteins, which contain one or more BH domains. Activated proapoptotic proteins Bak and Bax oligomerize and form pores in the OMM, thus initiating MOMP. Bcl-2 and Bcl-XL are anti-apoptotic proteins that inhibit the activities of Bak and Bax [48][49][50]. PETK, which is a Bax-derived peptide that contains the same BH domain as Bax, can change the permeability of the OMM by the same oligomerization mechanism and subsequently activate the apoptosis program mediated by caspase-3 [33]. This kind of artificially modified AMPs, such as PETK targeting the OMM, are good candidate drugs for the treatment of tumors and infectious diseases [21]. LL-37 accumulates in the mitochondria of human osteogenic MG63 cells and promotes the release of cytochrome c (Cyt c) and apoptosis-inducing factor (AIF) from IMS into the cytoplasm. Meanwhile, we could detect leaked Cyt c and AIF in the cytoplasm, while cytochrome c oxidase IV (COXIV), located in the IMM, was not detected [30]. This finding indicates that LL-37 may only activate the permeability of the OMM, without damaging the IMM structure [30]. However, the specific mechanism of the change in the permeability of the mitochondrial outer membrane remains to be explored.
As a derivative peptide of LL-37, 17BIPHE2 can regulate the expression of Bax and Bcl-2 in cancer cells by activating the ERK pathway to induce cell apoptosis [27]. Notably, the mitochondrial crest structure was damaged during this process, which may be attributed to the increase in intracellular reactive oxygen species (ROS) and Ca 2+ concentration caused by 17BIPHE2. It is possible that MOMP, MPTP, and MIMP occur simultaneously, and the three processes act together to disrupt the structure of mitochondria. Humanin (HN) is an effective mitochondrial-derived peptide that inhibits apoptosis. It has been reported that HN is able to induce structural changes in Bax, interact with structurally altered Bax, and produce fibers, thereby sequestrating Bax and preventing the initiation of MOMP [28,29]. Besides, some studies found that, during MOMP, factors such as Cyt c, Smac, and AIF are first released into the cytoplasm. After a series of reactions, Cyt c and Apaf-1 form apoptotic bodies, which activate the caspase cascade and lead to apoptosis [51]. At the same time, Smac can prevent the anti-apoptotic effect of XIAP. If caspase is absent or inactivated during MOMP, several apoptotic factors can induce apoptosis via inflammation ( Figure 1) [52].
According to the mechanisms of action of PETK, LL-37, 17BIPHE2, and HN [27,28,30,33], we can conclude that AMPs initiate MOMP through two distinct pathways, leading to apoptosis. One way is the direct participation of AMPs in MOMP to promote cell apoptosis with the same specific framework of apoptotic proteins regulating MOMP. These AMPs are usually artificially modified proapoptotic proteins [33]. The other way is to regulate the expression of apoptotic proteins, in order to indirectly regulate MOMP and subsequent apoptosis [27,31]. In the absence of caspases, AMPs can also trigger inflammation with the release of pro-apoptotic factors. Although increased expression of activated caspase-3 in the apoptotic pathway was detected [33,43], it is not clear whether this is caused by the apoptotic chain reaction (the combination of Cyt c with Apaf-1 to trigger apoptosis) induced directly by AMPs or by the AMP-induced expression of caspase genes. Because of the composition and structural characteristics of AMPs and the mechanism of their multi-target action, the simultaneous existence of these two causes is not excluded, but this still needs to be supported by more thorough evidence. For now, we can draw a conclusion that most AMPs promote apoptosis indirectly by promoting MOMP. ally altered Bax, and produce fibers, thereby sequestrating Bax and preventing the initiation of MOMP [28,29]. Besides, some studies found that, during MOMP, factors such as Cyt c, Smac, and AIF are first released into the cytoplasm. After a series of reactions, Cyt c and Apaf-1 form apoptotic bodies, which activate the caspase cascade and lead to apoptosis [51]. At the same time, Smac can prevent the anti-apoptotic effect of XIAP. If caspase is absent or inactivated during MOMP, several apoptotic factors can induce apoptosis via inflammation ( Figure 1) [52].
AMPs and MIMP
MIMP is considered to be a process that occurs after MOMP, during which substances in the mitochondrial matrix are released into the cytoplasm across the IMS. Theoretically, when MOMP is initiated, as Bak/Bax continues to oligomerize in the OMM, the forming pores will also continue to expand. The IMM is squeezed and protrudes from the pores on the OMM, which eventually leads to an increase in the permeability of the IMM and the release of mitochondrial DNA (mtDNA) [53,54]. Of course, if the degree of MIMP is mild, some of the ions in the mitochondrial matrix are allowed to reach the cytoplasm, causing changes in mitochondrial membrane potential. Under pathological conditions, mtDNA can activate the cGAS-STING signaling pathway in the cytoplasm, the signal of which is transmitted to the nucleus to generate interferon β (IFN-β) and trigger an inflammatory response (Figure 1) [55,56]. Moreover, it has been reported that mtDNA release is independent of mitochondrial fission and MPT [57]. Therefore, the leakage of mtDNA is one of the markers to determine the occurrence of MIMP. In fact, apoptotic caspases can inhibit inflammatory reactions during apoptosis [58], which explains why it is difficult to observe inflammatory reactions during MOMP, even if the apoptotic factors involved in inflammation are released. Some Arg-rich artificial peptides, such as KLAKLAK2 (KLA), can cause a decrease in mitochondrial membrane potential, but without initiating MOMP. Moreover, the ability of KLA to penetrate the outer and inner mitochondrial membrane is dependent on the mitochondrial membrane potential [35]. Under normal circumstances, any impairment of mitochondrial function will affect the mitochondrial membrane potential. Thus, the decrease of mitochondrial membrane potential is considered to be a preceding event to apoptosis [59,60]. Early research on cecropin A-melittin, a short hybrid peptide, found that it could pass through the IMM and allow the entry and exit of some factors, but it would destroy the structure of the IMM [34]. In a study of mitochondria-mediated apoptosis, when cells were treated with the BH3-mimetic ABT-737, which can act as an inhibitor of Bcl-XL and Bcl-2, we observed that MOMP was followed by mtDNA leakage [57]. However, it is not clear whether AMPs have a direct effect on MIMP. How mtDNA can cross the IMM and reach the cytoplasm without damaging the IMM structure remains our focus of attention. According to the premise that AMPs induce MOMP, it is speculated that AMPs can also induce MIMP and amplify its permeability. This idea could help explain why mtDNA leaks. Or perhaps, contrary to our suspicion, the leakage of mtDNA is accompanied by structural disruption of the IMM. After all, there are huge temporal differences between MOMP and MIMP [57], whereas severe MOMP-induced mitochondrial dysfunction is sufficient to disrupt mitochondrial structure. In either case, we need to continue to explore the specific mechanisms of MIMP.
AMPs and MPTP
MPTP contains three important components: the voltage-dependent anion channel (VDAC) located in the OMM, adenine nucleotide transporter (ANT) located in the IMM, and cyclophilin (Cyp-D) located in the mitochondrial matrix ( Figure 2). These three components are essential for maintaining mitochondrial osmotic balance and mitochondrial membrane permeability [61][62][63]. Under normal physiological conditions, MPTP is in a transient open state. If it is in a continuous open state, MPT occurs, which can lead to the extensive swelling of the mitochondrial matrix, accompanied by the rupture of the OMM, collapse of mitochondrial membrane potential, depletion of cell ATP, and, ultimately, cell necrosis or apoptosis [64]. Dithiothreitol (DTT) can inhibit MPTP caused by sulfhydryl oxidation of the OMM proteins, while cyclosporin A (CsA) can interact with Cyp-D, both of which are typical MPTP opening inhibitors ( Figure 2) [65]. CGA-N12, a derivative peptide of CGA-N46, can eliminate their inhibitory effect and improve mitochondrial permeability [38]. In addition, CGA-N12 maintains the MPTP opening by inducing the accumulation of intracellular ROS, thus leading to cell apoptosis [65]. Ca 2+ overload in the mitochondrial matrix is also an important factor contributing to the sustained opening of the MPTP. BIRD-2 is a synthetic peptide that interferes with the inositol 1,4,5-triphosphate receptor (IP3R), which independently induces apoptosis and appears to promote mitochondrial Ca 2+ uptake, followed by sustained opening of the MPTP [36,66].
Compared with mitochondrial membrane damage involving only MOMP, MPT induces relatively severe damage and other changes. The continuous opening of the MPTP will also directly cause MOMP and MIMP, and even damage the structure of mitochondria. In addition, there are many factors affecting the opening of the MPTP, and mitochondrial damages are also the result of a combination of factors. On the one hand, tilapia antimicrobial peptide 3 (TP3) can induce an increase in mitochondrial and intracellular ROS, thus promoting the opening of the MPTP. On the other hand, TP3 affects mitochondrial fission, leads to mitochondrial dysfunction, and increases the probability of MPTP opening [43]. Finally, TP3 leads to enhanced MPT, thereby prompting the activation of caspase-3/9 [39], which greatly facilitates apoptosis. Litopenaeus vannamei hemocyanin-derived peptide can also change mitochondrial permeability and mediate apoptosis by reducing the mitochondrial membrane potential and increasing the expression of caspase-3/9 and Bax [40]. ol. Sci. 2022, 23, x FOR PEER REVIEW 8 of 14
Challenges for Antimicrobial Peptides
Drugs that exploit the death receptor pathway to trigger the death of tumor cells and antibiotics used to treat infectious diseases can cause cells to become resistant [5,67]. Mitochondria are key to the growth and metastasis of tumor cells and play an important role in the growth inhibition of pathogens [68]. Although AMPs can replace antibiotics to exert therapeutic effects [69], the disadvantages of AMP application are very obvious. AMPs extracted from natural biological resources have low content and high extraction cost. Therefore, synthetic peptides may become a new breakthrough. By using various modifications, such as acetylation, glycosylation, protein fusion, and cyclization, the structure of AMPs can be optimized, the difficulty of industrial synthesis can be reduced, and the therapeutic effect can be improved [70,71]. For AMPs that damage the mitochondrial membrane, but cannot penetrate it, we can choose appropriate methods to render them membrane permeant. For example, AMPs modified by hydrophilic arginine-glycine-aspartate (RGD) sequence by protein fusion have the ability to penetrate cell membranes. This modification allows for the transport of AMPs from the extracellular to the intracellular space [72]. Therefore, it is feasible to rationally design AMPs to render them membrane permeant or to enhance their binding to mitochondrial targets [73], which is of great significance for improving their therapeutic effect.
In addition, almost all AMPs that can change mitochondrial membrane permeability have α-helices or can form ring structures, and their targets may be related to the structure and activity of AMPs [74] (Table 1). AMPs are difficult to obtain in large quantities, due to their low content in living organisms. Moreover, some AMPs have problems of cytotoxicity and bioavailability in clinical application [75]. Because the cytotoxicity of an AMP is positively correlated with its effect on mitochondrial outer and inner membrane permeability [37], we need to select an AMP that can alter mitochondrial membrane permeability without severely damaging mitochondrial structure and without affecting other organelles. We can, therefore, select less cytotoxic AMPs, according to the degree of mitochondrial structure damage. One of the known mechanisms of AMP membrane targeting is the cationic interaction of AMPs with negatively-charged cell membranes to increase their All the above reports indicate that AMPs could induce continuous MPTP opening and promote apoptosis. The greater the effect of AMPs on mitochondrial permeability, the stronger their cytotoxicity [37]. After treatment with AMPs, Cyt c is released, ROS is increased, and the mitochondrial membrane potential is disrupted [41]. Altogether, these results confirm that MPT induced by AMPs exhibits the typical characteristics of MOMP and MIMP, besides inducing apoptosis in the same manner. In addition, AMPs strengthened the necessary link between MPTP and MOMP ( Figure 1). We can infer that MIMP occurs during MPT. It is not clear whether MIMP is directly caused by the continuous opening of MPTP or by MOMP.
Challenges for Antimicrobial Peptides
Drugs that exploit the death receptor pathway to trigger the death of tumor cells and antibiotics used to treat infectious diseases can cause cells to become resistant [5,67]. Mitochondria are key to the growth and metastasis of tumor cells and play an important role in the growth inhibition of pathogens [68]. Although AMPs can replace antibiotics to exert therapeutic effects [69], the disadvantages of AMP application are very obvious. AMPs extracted from natural biological resources have low content and high extraction cost. Therefore, synthetic peptides may become a new breakthrough. By using various modifications, such as acetylation, glycosylation, protein fusion, and cyclization, the structure of AMPs can be optimized, the difficulty of industrial synthesis can be reduced, and the therapeutic effect can be improved [70,71]. For AMPs that damage the mitochondrial membrane, but cannot penetrate it, we can choose appropriate methods to render them membrane permeant. For example, AMPs modified by hydrophilic arginine-glycine-aspartate (RGD) sequence by protein fusion have the ability to penetrate cell membranes. This modification allows for the transport of AMPs from the extracellular to the intracellular space [72]. Therefore, it is feasible to rationally design AMPs to render them membrane permeant or to enhance their binding to mitochondrial targets [73], which is of great significance for improving their therapeutic effect.
In addition, almost all AMPs that can change mitochondrial membrane permeability have α-helices or can form ring structures, and their targets may be related to the structure and activity of AMPs [74] (Table 1). AMPs are difficult to obtain in large quantities, due to their low content in living organisms. Moreover, some AMPs have problems of cytotoxicity and bioavailability in clinical application [75]. Because the cytotoxicity of an AMP is positively correlated with its effect on mitochondrial outer and inner membrane permeability [37], we need to select an AMP that can alter mitochondrial membrane permeability without severely damaging mitochondrial structure and without affecting other organelles. We can, therefore, select less cytotoxic AMPs, according to the degree of mitochondrial structure damage. One of the known mechanisms of AMP membrane targeting is the cationic interaction of AMPs with negatively-charged cell membranes to increase their permeability [76]. Both the cell membrane and OMM have similar components (phospholipids and proteins) and electric charge, so it is not clear whether similar interaction of AMPs with the cell membrane can also occur on the OMM. Only after we understand the mechanism of action of AMPs can we modify them into therapeutic drugs. In addition, increasing the yield of AMPs and optimizing the technology of the modification of AMPs are among the challenges that we are currently facing.
Conclusions
We conclude that AMPs are potent inducers of apoptosis. If the AMPs are stable and persistent to induce MOMP, the pores formed on the OMM will persist for a long time and show a trend of continuous expansion until the AMPs are inactive, so MIMP is inevitable. On the one hand, the change in mitochondrial membrane potential will cause the abnormal continuous opening of the MPTP [38]. On the other hand, AMPs affect the opening of the MPTP by regulating the production of ROS, the concentration of Ca 2+ , or targeting the MPTP components [44]. Because MPTP spans the outer and inner membranes of mitochondria, it must also affect MOMP and MIMP [41]. In summary, the three pathways of membrane permeability changes induced by AMPs are interconnected, and even mutually reinforced (Figure 3). The mechanism of apoptosis caused by these three pathways has been confirmed by many experiments: apoptosis is mediated by the caspase protein cascade, triggered by the release of Cyt c [77] and the caspase-independent pathway, which is mediated by inflammatory signals, such as AIF or mtDNA [78,79].
After entering cells, AMPs directly or indirectly change the permeability of the mitochondrial outer and inner membranes, thereby activating the mitochondrial pathway [32]. Irrespectively of whether AMPs affect MOMP, MIMP, or MPT first, these three changes may eventually occur simultaneously in mitochondria. This may be one of the reasons why some AMPs have strong therapeutic effects and occasionally lead to the disruption of the mitochondrial structure. Therefore, AMPs that mediate mitochondrial apoptosis often simultaneously change the permeability of the outer and inner mitochondrial membrane in a synergistic manner [42]. In this review, the possible mode of action of AMPs in changing mitochondrial permeability is described, and the relationship between AMPs and MOMP, MIMP, and MPT is analyzed. This can facilitate the investigation of the mechanism of action of unknown AMPs and the screening and development of known AMPs. Undeniably, there are many unrevealed relationships regarding the effect of AMPs on apoptosis, such as whether the increase in caspase-3 expression is directly related to AMPs and whether AMPs are directly related to MIMP. All of these questions deserve further research and exploration. | 2022-10-27T15:32:42.348Z | 2022-10-22T00:00:00.000 | {
"year": 2022,
"sha1": "bb60479c83cfe2888091c830703cc625fee7429c",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1422-0067/23/21/12732/pdf?version=1666429687",
"oa_status": "GOLD",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "85df3bf376a465fceebf7c9660face661250cc5f",
"s2fieldsofstudy": [
"Medicine",
"Biology",
"Chemistry"
],
"extfieldsofstudy": [
"Medicine"
]
} |
14631791 | pes2o/s2orc | v3-fos-license | Baryon Structure and the Chiral Symmetry of QCD
Beyond the spontaneous chiral symmetry breaking scale light and strange baryons should be considered as systems of three constituent quarks with an effective confining interaction and a chiral interaction that is mediated by the octet of Goldstone bosons (pseudoscalar mesons) between the constituent quarks.
Introduction
Our aim in physics is not only to calculate some observable and get a correct number but mainly to understand a physical picture responsible for the given phenomenon. It very often happens that a theory formulated in terms of fundamental degrees of freedom cannot answer such a question since it becomes overcomplicated at the related scale. Thus a main task in this case is to select those degrees of freedom which are indeed essential. For instance, the fundamental degrees of freedom in crystals are ions in the lattice, electrons and the electromagnetic field. Nevertheless, in order to understand electric conductivity, heat capacity, etc. we instead work with "heavy electrons" with dynamical mass, phonons and their interaction. In this case a complicated electromagnetic interaction of the electrons with the ions in the lattice is "hidden" in the dynamical mass of the electron and the interactions among ions in the lattice are eventually responsible for the collective excitations of the lattice -phonons, which are Goldstone bosons of the spontaneously broken translational invariance in the lattice of ions. As a result, the theory becomes rather simple -only the electron and phonon degrees of freedom and their interactions are essential for all the properties of crystals mentioned above.
Quite a similar situation takes place in QCD. One hopes that sooner or later one can solve the full nonquenched QCD on the lattice and get the correct nucleon and pion mass in terms of underlying degrees of freedom: current quarks and gluon fields. However, QCD at the scale of 1 GeV becomes too complicated, and hence it is rather difficult to say in this case what kind of physics, inherent in QCD, is relevant to the nucleon mass and its lowenergy properties. In this lecture I will try to answer this question. I will show that it is the spontaneous breaking of chiral symmetry which is the most important QCD phenomenon in this case, and that beyond the scale of spontaneous breaking of chiral symmetry light and strange baryons can be viewed as systems of three constituent quarks which interact by the exchange of Goldstone bosons (pseudoscaler mesons) and are subject to confinement.
It is well known that at low temperature and density the approximate SU (3) L × SU (3) R chiral symmetry of QCD is realized in the hidden Nambu-Goldstone mode. The hidden mode of chiral symmetry is revealed by the existence of the octet of pseudoscalar mesons of low mass which represent the associated approximate Goldstone bosons. The η ′ (the SU (3)-singlet) decouples from the original nonet because of the U (1) A anomaly [1], [2]. Another consequence of the spontaneous breaking of the approximate chiral symmetry of QCD is that the valence quarks acquire their dynamical or constituent mass [3], [4], [5], [6] through their interactions with the collective excitations of the QCD vacuum -the quark-antiquark excitations and the instantons.
We have recently suggested [7], [8] that beyond the scale of spontaneous breaking of chiral symmetry a baryon should be considered as a system of three constituent quarks with an effective quark-quark interaction that is formed by a central confining part and a chiral interaction mediated by the octet of pseudoscalar mesons between the constituent quarks.
The simplest representation of the most important component of the interaction of the constituent quarks mediated by the octet of pseudoscalar bosons in the SU (3) F invariant limit is Here {λ F i } represents the flavor SU (3) Gell-Mann matrices and the sums run over the constituent quarks.
Because of the flavor dependent factor λ F i · λ F j the chiral boson exchange interaction (1) will lead to orderings of the positive and negative parity states in the baryon spectra, which agree with the observed ones in all sectors. For the spectrum of the nucleon and ∆ the strength of the chiral interaction between the constituent quarks is sufficient to shift the lowest positive parity states in the N =2 band (the N(1440) and ∆(1600)) below the negative parity states in the N =1 band (N(1520) − N(1535) and ∆(1620) − ∆(1700)). In the spectrum of the Λ, on the other hand, it is the negative parity flavor singlet states ( Λ(1405) − Λ(1520)) that remain the lowest lying resonances, again in agreement with experiment. The mass splittings between the baryons with different strangeness, and between the Λ and the Σ which have identical flavor, spin and flavor-spin symmetries arise from the explicit breaking of the SU (3) F symmetry that is caused by the mass splitting of the pseudoscalar meson octet, and the different masses of the u,d and the s quarks.
This lecture has the following structure. Section 2 contains the proof why the commonly used perturbative gluon exchange interaction between the constituent quarks leads to incorrect ordering of positive and negative parity states in the spectra. In Section 3 we outline the importance of the spontaneous breaking of chiral symmetry for low-energy QCD, and in Section 4 we present a short historical sketch of the role of chiral symmetry in quark based models. Section 5 contains a description of the chiral boson exchange interaction (1). In Section 6 we describe the spectra of the nucleon, the ∆ resonance and the Λ hyperon as they are predicted by the SU (3) F symmetric interaction (1). The effect of SU (3) F breaking in the interaction is considered in Section 7. Section 8 is devoted to an "exact" three-body description of baryons where the interaction (1) is taken into account to all orders. In Section 9 we discuss the role of the exchange current corrections to the baryon magnetic moments that are associated with the pseudoscalar exchange interaction. Finally, in Section 10 some recent lattice -QCD results are discussed.
Why the Gluon Exchange Bears no Relation to the Baryon Spectrum
It was accepted by many people (but not by all) that the fine splittings in the baryon spectrum (in analogy to atomic physics they are often called "hyperfine splittings" as they arise from the spin-spin forces) are due to the gluon-exchange interaction between the constituent quarks [9], [10]. Now I shall address myself to a formal consideration why the one gluon exchange interaction cannot be relevant to the baryon spectrum.
The most important component of the one gluon exchange interaction [9] is the so called color-magnetic interaction where {λ C i } are color SU (3) matrices. It is the permutational color-spin symmetry of the 3q state which is mostly responsible for the contribution of the interaction (2). Indeed, the corresponding two-body matrix element is Thus the symmetrical color-spin pairs (i.e. with a [2] CS Young pattern) experience an attractive contribution while the antisymmetrical ones ([11] CS ) experience a repulsive contribution. Hence the color-magnetic contribution to the ∆ state ([111] CS ) is more repulsive than to the nucleon ([21] CS ) and the ∆ becomes heavier than the nucleon. The price is that α s should be larger than unity, which is bad. In addition there is no empirical indication in the spectrum for a large spin-orbit component of the gluon-exchange interaction [9] implied by this big value of α s . The crucial point is that the interaction (2) feels only colour and spin of the interacting quarks. Thus the structure of the N and Λ spectra has to be the same as these baryons differ only by their flavour structure. If one looks at the Particle Data Group tables, however, one immediately sees a different ordering of the positive and negative parity states in both spectra. In the N spectrum the lowest states are 1 The interaction (2) cannot explain why the two-quantum excitations of positive parity N(1440), ∆(1600), Λ(1600) and Σ(1660) lie below the onequantum excitations of negative parity N(1535) − N(1520), ∆(1620)−∆(1700), Λ(1670) − Λ(1690) and Σ(1750) − Σ(?), respectively. For instance, the positive parity state N(1440) and the negative parity ones N(1535) − N(1520) have the same mixed ([21] CS ) color-spin symmetry thus the color-magnetic contribution to these states cannot be very different (a small difference is only due to the different radial structure of the positive and negative parity states). But the N(1440) state belongs to the N = 2 shell, while the N(1535) − N(1520) pair is a member of the N = 1 band, which means that the N(1440) should lie approximatelyhω above the N(1535)−N(1520). In the ∆ spectrum the situation is even more dramatic. The ∆(1600) positive parity state has a completely antisymmetric CS-Young pattern ([111] CS ), while the negative parity states ∆(1620) − ∆(1700) have a mixed one. Thus the color-magnetic contribution to the ∆(1600) is much more repulsive than to the ∆(1620) − ∆(1700). In addition the ∆(1600) is the N = 2 state, while the pair ∆(1620) − ∆(1700) belongs to the N = 1 band. As a consequence the ∆(1600) must lie much higher than the ∆(1620) − ∆(1700). All these features are well seen in the explicit 3-body calculations [11].
Spontaneous Chiral Symmetry Breaking and its Consequences for Low-Energy QCD
The QCD Lagrangian with three light flavors has a global symmetry if one neglects the masses of current u,d, and s quarks, which are small compared to a typical low-energy QCD scale of 1 GeV. The U (1) A is not a symmetry at the quantum level due to the axial anomaly. If the SU (3) L ×SU (3) R chiral symmetry of the QCD Lagrangian were intact in the vacuum state we would observe degenerate multiplets in the particle spectrum corresponding to the above chiral group, and all hadrons would have their degenerate partners with opposite parity. Since this does not happen the implication is that the chiral symmetry is spontaneously broken down to SU (3) V in the QCD vacuum, i.e., realized in the hidden Nambu-Goldstone mode. A direct evidence for the spontaneously broken chiral symmetry is a nonzero value of the quark condensates for the light flavors which represents the order parameter. That this is indeed so, we know from three independent sources: current algebra [12], QCD sum rules [13], and lattice gauge calculations [14]. There are two important generic consequences of the spontaneous chiral symmetry breaking. The first one is an appearance of the octet of pseudoscalar mesons of low mass, π, K, η, which represent the associated approximate Goldstone bosons. The second one is that valence quarks acquire a dynamical or constituent mass. Both these consequences of the spontaneous chiral symmetry breaking are well illustrated by, e.g. the σ-model [15] or the Nambu and Jona-Lasinio model [16]. We cannot say at the moment for sure what the microscopical reason for spontaneous chiral symmetry breaking in the QCD vacuum is. It was suggested that this occurs when quarks propagate through instantons in the QCD vacuum [5], [6]. For the low-energy baryon properties it is only essential that beyond the spontaneous chiral symmetry breaking scale new dynamical degrees of freedom appear -constituent quarks and chiral fields. The low-energy baryon properties are mainly determined by these dynamical degrees of freedom and the confining interaction. This is quite in contrast to pseudoscalar mesons. In the chiral limit, m 0 u = m 0 d = m 0 s = 0, all members of the pseudoscalar octet (π, K, η) would have zero mass, which is most clearly seen in the Gell-Mann-Oakes-Renner [12] relations that relate the pseudoscalar meson masses to the quark condensates and current quark masses m 0 . Thus the nonzero masses of mesons are determined by the nonzero values of the current quark masses. In the baryon case, even in the chiral limit, baryons would have approximately their actual masses of the order of 1 GeV, as these masses are mostly determined by the dynamical (constituent) masses, the Goldstone boson exchange interaction among them, as well as a confining interaction. The dynamical (constituent) masses are in turn determined mainly by the quark condensates, which is most clearly seen from the gap equations of the Nambu and Jona-Lasinio model, and only weakly dependent on current quark masses. The role of the current quark masses in baryons is to break just the SU (3) F symmetry in the baryon spectrum.
Chiral Symmetry and the Quark Model (Historical Sketch)
The importance of the constraints posed by chiral symmetry for the quark bag [17] and bag-like [18] models for the baryons has been recognized early.
In bag models with restored chiral symmetry on the bag surface, or bag-like models, the massless current quarks within the bag were assumed to interact not only by perturbative gluon exchange but also through chiral meson field exchange. In these models the chiral field has the character of a compensating auxiliary field only, rather than a collective low frequency Goldstone quarkantiquark excitation. The possibility of a nonzero quark condensate was also not addressed. As it was discussed in a previous section, it is the quark condensate which is the most important characteristic determining the baryon properties. According to the bag philosophy, however, there is a perturbative QCD phase inside a bag, where all condensates vanish by definition. A general limitation of all bag and bag-like models is, of course, the lack of translational invariance, which is of crucial importance for a description of the excited states. Common to these models is that the breaking of chiral symmetry arises from the confining interaction. This point of view contrasts with that of Manohar and Georgi [4], who pointed out that there should be two different scales in QCD with 3 flavors. At the first one of these, Λ χSB ≃ 4πf π ≃ 1 GeV, the spontaneous breaking of the chiral symmetry occurs, and hence at distances beyond 1 ΛχSB ≃ 0.2 fm the valence current quarks acquire their dynamical (constituent) mass (called "chiral quarks" in [4]), and the Goldstone bosons (mesons) appear. The other scale, Λ QCD ≃ 100 − 300 MeV, is that which characterizes confinement, and the inverse of this scale roughly coincides with the linear size of a baryon. Between these two scales then the effective Lagrangian should be formed out of the gluon fields that provide a confining mechanism, as well as of the constituent quark and pseudoscalar meson fields. Manohar and Georgi did not, however, specify whether the baryons should be described as bound qqq states or as chiral solitons.
The chiral symmetry breaking scale above fits well with that which appears in the instanton liquid picture of the QCD vacuum [5], [6]. In this model the quark condensates (i.e., equilibrium of virtual quark-antiquark pairs in the vacuum state), as well as the gluon condensate, are supported by instanton fluctuations of a size ∼ 0.3 fm. Diakonov and Petrov [6] suggested that at low momenta (i.e., beyond the chiral symmetry breaking scale) QCD should be approximated by an effective chiral Lagrangian of the sigma-model type that contains valence quarks with dynamical (constituent) masses and meson fields. They considered a nucleon as three constituent quarks moving independently of one another in a self-consistent chiral field of the hedgehog form [19]. In this picture the ∆ appears as a rotational excitation of the hedgehog, and no explicit confining interaction is included. A very similar description for the nucleon was suggested in [20], [21]. These types of models are now called "quark-soliton models" [22], [23].
The spontaneous breaking of chiral symmetry and its consequences -dynamical quark mass generation, appearance of the quark condensate, and pseudoscalar mesons as Goldstone excitations -are well illustrated by the Nambu and Jona-Lasinio model [16], [24]. This model lacks a confining interaction, which, as argued below, is essential for a realistic description of the properties of baryon physics.
The Chiral Boson Exchange Interaction
In an effective chiral description of the baryon structure, based on the constituent quark model, the coupling of the quarks and the pseudoscalar Goldstone bosons will (in the SU (3) F symmetric approximation) have the form igψγ 5 λ F ·φψ (or g/(2m)ψγ µ γ 5 λ F ·ψ∂ µ φ), where ψ is the fermion constituent quark field operator, φ the octet boson field operator, and g is a coupling constant. A coupling of this form, in a nonrelativistic reduction for the constituent quark spinors, will -to lowest order -give rise to a Yukawa interaction between the constituent quarks, the spin-spin component of which has the form Here m i and m j denote the masses of the interacting quarks, and µ that of the meson. There will also be an associated tensor component, which is discussed in ref. [8]. At short range the simple form (7) of the chiral boson exchange interaction cannot be expected to be realistic and should only be taken to be suggestive. Because of the finite spatial extent of both the constituent quarks and the pseudoscalar mesons the delta function in (7) should be replaced by a finite function, with a range of 0.6-0.7 fm, as suggested by the spatial extent of the mesons. In addition, the radial behaviour of the Yukawa potential (7) is valid only if the boson field satisfies a linear Klein-Gordon equation. The implications of the underlying chiral symmetry of QCD for the effective chiral Lagrangian (which in fact is not known), which contains constituent quarks as well as boson fields, are that these boson fields cannot be described by linear equations near their source. Therefore it is only at large distances, where the amplitude of the boson fields is small, that the quark-quark interaction reduces to the simple Yukawa form.
The latter point is rather important and has to be clarified. The radial dependence in (7) is a direct consequence of a free (q 2 − µ 2 ) −1 Green function for the boson field, and a pseudoscalar igψγ 5 λ F · φψ or pseudovector g/(2m)ψγ µ γ 5 λ F · ψ∂ µ φ coupling in both quark-meson vertices. Making use of the usual "static approximation" and neglecting the recoil corrections at both vertices in a nonrelativistic reduction for the constituent quark spinors, one arrives at (7). The free Green function above comes from the well-known Lagrangian for a free boson field 1/2∂ µ φ∂ µ φ − 1/2µ 2 φ 2 , which means that the boson field satisfies a linear Klein-Gordon equation.
On the other hand, the underlying SU (3) The "kinetic term" in this case is which in its expansion in powers of φ contains not only a free Klein-Gordon kinetic term but also other terms of higher powers. For instance, the terms of fourth order in the meson field give rise to π − π scattering [26], etc. The full chiral Lagrangian describing the Goldstone boson field is unknown and should contain a quartic term involving derivatives ∂ µ U and higher order terms. It is clear that a dressed Green function for the field φ should be very different compared to the free Klein-Gordon Green function since the selfinteraction of the Goldstone boson field is important. Only far away from its source (the fermion current), where the amplitude of the boson field φ is small, the lowest term in powers of φ becomes dominant, and therefore only at large distances the quark-quark interaction reduces to its simple Yukawa form.
At this stage the proper procedure should be to avoid further specific assumptions about the short range behavior of V (r) in (1), to extract instead the required matrix elements of it from the baryon spectrum, and to reconstruct by this an approximate radial form of V (r). The overall minus sign in the effective chiral boson interaction in (1) corresponds to that of the short range term in the Yukawa interaction.
The flavor structure of the pseudoscalar octet exchange interaction in (1) between two quarks i and j should be understood as follows : The first term in (11) represents the pion-exchange interaction, which acts only between light quarks. The second term represents the Kaon exchange interaction, which takes place in u-s and d-s pair states. The η-exchange, which is represented by the third term, is allowed in all quark pair states.
In the SU (3) F symmetric limit the constituent quark masses would be equal (m u = m d = m s ), the pseudoscalar octet would be degenerate and the mesonconstituent quark coupling constant would be flavor independent. In this limit the form of the pseudoscalar exchange interaction reduces to (1), which does not break the SU (3) F invariance of the baryon spectrum. Beyond this limit the pion, Kaon and η exchange interactions will differ (V π = V K = V η ), because of the difference between the strange and u, d quark constituent masses (m u,d = m s ), and because of the mass splitting within the pseudoscalar octet (µ π = µ K = µ η ) (and possibly also because of flavor dependence in the mesonquark coupling constant). The source of both the SU (3) F symmetry breaking constituent quark mass differences and the SU (3) F symmetry breaking mass splitting of the pseudoscalar octet is the explicit chiral symmetry breaking in QCD.
The Structure of the Baryon Spectrum
The two-quark matrix elements of the interaction (1) are: From these the following important properties may be inferred: (i) At short range, where V (r ij ) is positive, the chiral interaction (1) is attractive in the symmetric FS pairs and repulsive in the antisymmetric ones. At large distances the potential function V (r ij ) becomes negative and the situation is reversed.
(ii) At short range, among the FS-symmetrical pairs, the flavor antisymmetric pairs experience a much larger attractive interaction than the flavorsymmetric ones, and among the FS-antisymmetric pairs the strength of the repulsion in flavor-antisymmetric pairs is considerably weaker than in the symmetric ones.
Given these properties we conclude, that with the given flavor symmetry, the more symmetrical the FS Young pattern is for a baryon the more attractive contribution at short range comes from the interaction (1). For two identical flavor-spin Young patterns [f ] FS the attractive contribution at short range is larger for the more antisymmetrical flavor Young pattern [f ] F . Consider first, for the purposes of illustration, a schematic model which neglects the radial dependence of the potential function V (r) in (1), and assume a harmonic confinement among quarks as well as m u = m d = m s . In this model If the only interaction between the quarks were the flavor-and spinindependent harmonic confining interaction, the baryon spectrum would be organized in multiplets of the symmetry group SU (6) FS × U (6) conf . In this case the baryon masses would be determined solely by the orbital structure, and the spectrum would be organized in an alternative sequence of positive and negative parity states. The Hamiltonian (13), within a first order perturbation theory, reduces the SU (6) FS × U (6) conf symmetry down to SU (3) F × SU (2) S × U (6) conf , which automatically implies a splitting between the octet and decuplet baryons. These matrix elements alone suffice to prove that the ordering of the lowest positive and negative parity states in the baryon spectrum will be correctly predicted by the chiral boson exchange interaction (13). The constant C χ may be determined from the N−∆ splitting to be 29.3 MeV. The oscillator parameterhω, which characterizes the effective confining interaction, may be determined as one half of the mass differences between the first excited 1 2 + states and the ground states of the baryons, which have the same flavor-spin, flavor and spin symmetries (e.g. N(1440) -N, Λ(1600) -Λ, Σ(1660) -Σ), to behω ≃ 250 MeV. Thus the two free parameters of this simple model are fixed and we can make now predictions. In the N and Σ sectors the mass difference between the lowest excited 1 2 + states (N(1440) and Σ(1660)) and This simple example shows how the chiral interaction (13) provides different ordering of the lowest positive and negative parity excited states in the spectra of the nucleon and the Λ-hyperon. This is a direct consequence of the symmetry properties of the boson-exchange interaction discussed at the beginning of this section. Namely, the [3] FS state in the N(1440), ∆(1600) and Σ(1660) positive parity resonances from the N = 2 band feels a much stronger attractive interaction than the mixed symmetry state [21] FS in the N(1535) -N(1520), ∆(1620) -∆(1700) and Σ(1750) -Σ(?) resonances of negative parity (N = 1 shell). Consequently the masses of the positive parity states N(1440), ∆(1600) and Σ(1660) are shifted down relative to the other ones, which explains the reversal of the otherwise expected "normal ordering". The situation is different for Λ(1405) -Λ(1520) and Λ(1600), as the flavor state of Λ(1405) -Λ(1520) is totally antisymmetric. Because of this the Λ(1405) -Λ(1520) gains an attractive energy, which is comparable to that of the Λ(1600), and thus the ordering suggested by the confining oscillator interaction is maintained.
Consider now, in addition, the radial dependence of the potential with the SU (3) F invariant version (1) of the chiral boson exchange interaction (i.e., V π (r) = V K (r) = V η (r)). If the confining interaction in each quark pair is taken to have the harmonic oscillator form as above, the exact eigenvalues and eigenstates to the coinfining 3q Hamiltonian are where N is the number of quanta in the state, the Elliott symbol (λµ) characterizes the SU (3) harmonic oscillator symmetry, and L is the orbital angular momentum. The spatial (X), flavor-spin (FS), flavor (F), and spin (S) permutational symmetries are indicated by corresponding Young patterns (diagrams) [f ]. All these functions are well known (see, e.g., [27]). Note that the color state [111] C , which is common to all the states, has been suppressed in (17). By the Pauli principle
Here ϕ nlm (r 12 ) represents the oscillator wavefunction with n excited quanta.
As we shall only consider the baryon states in the N ≤ 2 bands, we shall only need the four radial matrix elements P 00 , P 11 , P 20 and P 22 for the numerical construction of the spectrum. The contributions to all nucleon, ∆ and Λ-hyperon states from the boson exchange interaction, in terms of the matrix elements P nl , are listed in Tables 1 and 2. In this approximate SU (3) F -invariant version of the chiral boson exchange interaction the Λ − N and the Ξ − Σ mass differences would solely be ascribed to the mass difference between the s and u,d quarks, since all these baryons have identical orbital structure and permutational symmetries. The states in the Λ-spectrum would be degenerate with the corresponding states in the Σ-spectrum which have equal symmetries.
The oscillator parameterhω and the four integrals are extracted from the mass differences between the nucleon and the ∆(1232), the ∆(1600) and the N(1440), as well as the splittings between the nucleon and the average mass of the two pairs of states N(1535)−N(1520) and N(1720)−N(1680). This procedure yields the parameter valueshω=157.4 MeV, P 00 =29.3 MeV, P 11 =45.2 MeV, P 20 =2.7 MeV and P 22 =-34.7 MeV. Given these values, all other excitation energies (i.e., differences between the masses of given resonances and the corresponding ground states) of the nucleon, ∆-and Λ-hyperon spectra are predicted to within ∼ 15% of the empirical values where known, and are well within the uncertainty limits of those values. Note that these matrix elements provide a quantitatively satisfactory description of the Λ-spectrum even though they are extracted from the N − ∆ spectrum.
The relative magnitudes and signs of the numerical parameter values can be readily understood. If the potential function V (r) is assumed to have the form of a Yukawa function with a smeared δ-function term that is positive at short range r ≤ 0.6 − 0.7 fm, as suggested by the pion size < r 2 π > = 0.66 fm, one expects P 20 to be considerably smaller than P 00 and P 11 , as the radial wavefunction of the excited S-state has a node, and as it extends further into the region where the potential is negative. The negative value for P 22 is also natural, since the corresponding wavefunction is suppressed at short range and extends well beyond the expected node in the potential function.
The SU (3) F Breaking Chiral Boson Interaction
The model described above has relied on an interaction potential function V (r) in (1) that is flavor independent. A refined version takes into account the explicit flavor dependence of the potential function in (11) (V π = V K = V η ). In the following we show how this explicit flavor dependence provides an explanation of the mass splitting between the Λ and the Σ which have the same quark content and the same FS, F and S symmetries, i.e., they are degenerate within the SU (3) F version (1) of the chiral boson exchange interaction.
Beyond the SU (3) F limit the ground state baryons will be determined by the π-exchange radial integral P π 00 , the K-exchange one, P K 00 , and by the η-exchange integrals, P uu 00 = P ud 00 = P dd 00 , P us 00 and P ss 00 , where the superscripts indicate quark pairs to which the η-exchange applies. As indicated by the Yukawa interaction (7) these matrix elements should be inversely proportional to the product of the quark masses of the pair state. Thus P us nl = mu ms P uu nl , P ss nl = ( mu ms ) 2 P uu nl . We also assume that P us 00 ≃ P K 00 , which is suggested by the fact that the quark masses are equal in the states in which these interactions act, and by the near equality of the Kaon and η masses, µ η ≃ µ K . Thus we have only two independent radial integrals.
To determine the integrals P π 00 , P K 00 and the quark mass difference ∆ q = m s − m u we consider the Σ(1385) − Σ, ∆ − N and Λ − N splittings: in good agreement with the empirical values of 77 MeV and 125 MeV, respectively.
A description of the other parts of the Σ, Ξ and Ω spectra can be found in [8].
Three-Body Faddeev Calculations
In the previous sections we have shown how the Goldstone boson exchange (GBE), taken to first order perturbation theory and without explicit parameterizing the radial dependence, can explain the correct level ordering of positive and negative parity states in light and strange baryon spectra, as well as the splittings in those spectra. A question, however, arises about what will happen beyond first order perturbation theory. In order to check this we have numericaly solved three-body Faddeev equations [28]. Besides the confinement potential, which is now taken in linear form, the GBE interaction between the constituent quarks is now included to all orders. These results further support the adequacy of the GBE for baryon spectroscopy.
In addition to the octet-exchange interaction we include here also the flavor-singlet (η ′ ) exchange. In the large N C limit the axial anomaly becomes suppressed [29], and the η ′ becomes the ninth Goldstone boson of the spontaneously broken U (3) L × U (3) R chiral symmetry in addition to the octet of pseudoscalar mesons [30].
For the GBE the spin-spin component of the interaction between the constituent quarks i and j reads: where λ a , a = 1, ..., 8 are flavor Gell-Mann matrices and λ 0 = 2/31l. In the simplest case, when the boson field satisfies the linear Klein-Gordon equation, one has the following spatial dependence for the meson-exchange potentials in (25): (γ = π, K, η, η ′ ), with quark and meson masses m i and µ γ , respectively.
Eq. (26) contains both the traditional long-range Yukawa potential as well as a δ-function term. It is the latter that is of crucial importance for baryon physics. We already discussed in Section 5 that it is strictly valid only for pointlike particles, and that it must be smeared out since the constituent quarks and pseudoscalar mesons have finite size, and in addition the boson fields in a chiral Lagrangian should in fact satisfy a nonlinear equation. Furthermore it is quite natural to assume that at distances r ≪ r 0 , where r 0 can be related to the constituent quark and pseudoscalar meson sizes, there is no chiral boson-exchange interaction, since this is the region of perturbative QCD with the original QCD degrees of freedom. The interactions at these very short distances are not essential for the low-energy properties of baryons. Consequently we use a two-parameter "representation" for the δ-function term in (26) Following the arguments above one should also cut off the Yukawa part of the GBE for r < r 0 . The πq coupling constant can be extracted from the phenomenological pion-nucleon coupling [8] as g 2 8 4π = 0.67. For simplicity (and to avoid any additional free parameter), the same coupling constant is assumed for the coupling between the η-meson and the constituent quark. This is exactly in the spirit of unbroken SU (3) F symmetry. For the flavor-singlet η ′ , however, we must take a different coupling , as the η ′ decouples from the pseudoscalar octet due to the U (1) A anomaly. This fact is illustrated best by the failure of the Gell-Mann-Oakes-Renner relations [12] for the flavor singlet [1]. Lacking a phenomenological value, we treat g 2 0 /4π as a free parameter. The constituent masses of the u and d quarks are taken to be 340 MeV, as suggested by the nucleon magnetic moments.
In the present calculation we neglect tensor meson-exchange forces. We expect their role to be of minor importance for the main features of the baryon spectra [8] (mainly due to the absence of the strong δ-function part in this case).
Our full interquark potential is thus given by While all masses and the octet coupling constant are predetermined, we treated r 0 , α, (g 0 /g 8 ) 2 , and C as free parameters and determined their values to be: Notice that we do not need any constant V 0 , which is usually added to the confining potential. In fact only four free parameters suffice to describe all 14 lowest states of the N and ∆ spectra, including the absolute value of the nucleon (ground state). At the present stage of determining the qq potential due to GBE we were led by the principle of working with the smallest possible number of free parameters. Therefore we took the octet coupling constant g 2 8 /4π, and likewise the constituent quark mass m, as predetermined. Of course, one may expect that subsequent studies within the GBE model will put further constraints on the parametrization of the qq potential.
The qq potential (28) constitutes the dynamical input into our 3-body Faddeev calculations of the baryon spectra. We show our results in Fig. 1 for the parametrization of the qq interaction as given above. It is well seen that the whole set of lowest N and ∆ states is reproduced quite correctly. In the most unfavourable cases deviations from the experimental values do not exceed 3%! In addition all level orderings are correct. In particular, the positiveparity state N(1440) (Roper resonance) lies below the pair of negative-parity states N(1535) -N(1520). The same is true in the ∆ spectrum with ∆(1600) and the pair ∆(1620) -∆(1700). We emphasize again that the qq potential (28) is able to predict also the absolute value of the nucleon mass. In previous models an arbitrary constant was usually needed to achieve the correct value of 939 MeV. At the present stage of our investigation of the baryon spectra with the GBE interaction we have left out the tensor forces. Therefore the finestructure splittings in the LS-multiplets are not yet introduced. However, it is clear from the observed smallness of these splittings and from the arguments given above, that the tensor component of the GBE can play only a minor role. Here we also note that the Yukawa part of the interaction in (26) is only of secondary importance. In fact, the pattern of Fig. 1 could also be described with the "δ-part" (27) alone (and a slightly modified set of parameters).
It is instructive to learn how the GBE affects the energy levels when it is switched on and its strength (coupling constant) is gradually increased (Fig. 2). Starting out from the case with confinement only, one observes that the degeneracy of states is removed and an inversion of the ordering of positiveand negative-parity states is achieved, both in the N and ∆ excitations. From Fig. 2 also the crucial importance of the chiral interaction V χ becomes evident. Notice that the strength of our confinement, C= 0.474 fm −2 , is rather small and the confining interaction contributes much less to the splittings than the GBE. The relative "weakness" of our effective confining interaction could be due to a partial cancellation between the much stronger color-electric confinement and the σ-exchange since they are of opposite sign. Due to the same reason one cannot expect that the effective confining interaction between the constituent quarks is strictly of linear form.
Exchange Current Corrections to the Magnetic Moments
A flavor dependent interaction of the form (1) will imply the presence of an irreducible two-body exchange current operator, as seen, e.g., directly from the continuity equation, by which the commutator of the interaction and the single particle charge operator is equal to the divergence of the exchange current density [31].
The general form of the octet vector exchange current operator, that is associated with the complete octet mediated interaction (11), is [8] HereṼ π (r) andṼ K (r) are dimensionless functions that describe π and K exchange, respectively, and which include both the pionic (kaonic) current and the pair current term at long range. Consider a simplified nonrelativistic constituent quark model. The impulse approximation expressions for the magnetic moments of the ground state octet baryons and their experimental values are listed in Table 3 (columns "IA" and "exp", respectively). A natural approach is to determine the mass ratios m N /m u and m N /m s to fit the experimental values of the magnetic moments of the Σ − and Ξ − octet and the Ω and ∆ ++ (µ Ω = −2.019±0.054 µ N , µ ∆ ++ = 4.52 ± 0.50 µ N ) decuplet baryons, which are unaffected by the exchange current operator (29). While with only two independent variables it is not possible to fit all four experimental magnetic moments exactly, the best overall fit, µ Σ − = −1.00 µ N , µ Ξ − = −0.59 µ N , µ Ω − = −2.01 µ N , µ ∆ ++ = 5.52 µ N , happens to be obtained with precisely the ratios m N /m u = 2.76 and m N /m s = 2.01, which were used for the constituent quark masses to fit ground state baryons (m u = 340 MeV and m s = 467 MeV).
We find (see Table 3) that the meson exchange current contributions systematically improve the predictions of the naive constituent quark model (i.e., with one-body quark currents only) for all known magnetic moments.
As the constituent quarks are not too heavy, both their electromagnetic and axial current operators have significant relativistic correction terms. Their effect is to reduce the magnitude of the predicted values of both the axial coupling constants and the magnetic moments of the baryons that are given by the static quark model. This correction reduces the standard overprediction of the axial current coupling constant of the nucleon (5/3 vs 1.24) and the strange baryons, but it worsens the mostly satisfactory predictions for the magnetic moments of the baryons that are obtained with the static quark model. In ref. [32] it is shown that the exchange current corrections associated with the chiral boson exchange interaction between the quarks can compensate for the relativistic correction in the latter case, while leaving it operative in the case of the axial coupling constants. This then makes it possible to obtain at least qualitatively satisfactory simultaneous description of both the magnetic moments and the axial coupling constants.
Instead of a Conclusion
Instead of a conclusion we discuss some important recent lattice QCD results in this last section. It was shown already a few years ago that one can obtain a qualitatively correct splitting between ∆ and N already within a Table 3. Magnetic moments of the baryon octet (in nuclear magnetons). Column IA contains the quark model impulse approximation expressions, column "exp" the experimental values, column I the impulse approximation predictions, column II the exchange current contribution with < ϕ000(r12)|Ṽπ(r12)|ϕ000(r12) >= −0.018 and < ϕ000(r12)|ṼK(r12)|ϕ000(r12) >= 0.03, and column III the net predictions.
quenched approximation (for a review and references see [33]). Within the quenched approximation to QCD the sea quark closed loop diagrams generated by gluon lines are neglected. Thus in the quenched approximation for baryons one takes into account only 3 continuous valence quark lines and full gluodynamics. This quenched approximation contains, however, part of antiquark effects related to the Z graphs formed of valence quark lines. One can even construct diagrams within the quenched approximation which correspond to the exchange of the color-singlet isospin 1 or 0 qq pairs between valence quark lines [34]. It is also important that these diagrams contribute to the baryon mass to leading order (∼ N C ) in a 1/N C expansion [35] (their contribution to the ∆ − N splitting appears, however, to subleading orders).
From the quenched measurements [33] it is not clear what were the physical reason for the ∆ − N splitting: gluon exchanges, instantons, or something else. To clarify this question, Liu and Dong have recently measured the ∆ − N splitting in the quenched and a further so-called "valence approximation" [36]. In the valence approximation the quarks are limited to propagating only forward in time (i.e., Z graphs and related quark-antiquark pairs are removed). The gluon exchange and all other possible gluon configurations, including instantons, are exactly the same in both approximations. The striking result is that the ∆−N splitting is observed only in the quenched approximation but not in the valence approximation, in which the N and the ∆ levels are degenerate within error bars. Consequently the ∆ − N splitting must receive a considerable contribution from the diagrams with qq excitations, which correspond to the meson exchanges, but not from the gluon exchange or instanton-induced interaction between quarks (to be precise, the instanton-induced interaction could be rather important for the interactions between quarks and antiquarks). | 2014-10-01T00:00:00.000Z | 1996-09-06T00:00:00.000 | {
"year": 1996,
"sha1": "7b590ed6078d043770875d25bf83e413cd9b8838",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/hep-ph/9609278",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "7b590ed6078d043770875d25bf83e413cd9b8838",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
237938937 | pes2o/s2orc | v3-fos-license | Role of Diagnostic Imaging in Chronic Recurrent Multifocal Osteomyelitis (CRMO) in Children: An Observational Study
The aim of this single-center observational study was to analyze the applicability of various imaging studies to the diagnosis and further evaluation of patients with chronic recurrent multifocal osteomyelitis (CRMO). The analysis included the data of 10 patients with CRMO treated between 2016 and 2021. The mean ages of the patients at the first manifestation of CRMO and ultimate diagnosis were 10 years and 7 months and 11 years and 10 months, respectively. Conventional radiography demonstrated focal loss of bone density in only 30% of the patients. Computed tomography showed disseminated foci with non-homogeneous osteolytic/osteosclerotic structure, with a massive loss of cortical layer and strong periosteal reaction. On magnetic resonance imaging (MRI), most patients presented with multifocal hypodense areas on T1-weighted images, with the enhancement of signal on T-weighted and STIR sequences. The duration of follow-up varied between 3 months and 3 years. In 40% of the patients, both clinical symptoms and the abnormalities seen on MRI resolved completely, whereas another 50% showed partial regression of clinical and radiological manifestations. MRI findings, co-existing with characteristic clinical manifestations, play a pivotal role in establishing the ultimate diagnosis of CRMO. MRI can also be used to monitor the outcomes of treatment in CRMO patients.
Introduction
Chronic recurrent multifocal osteomyelitis (CRMO) is a rare form of non-infectious inflammation of the bones in children and young adults. The disease was first described by Giedon et al. in 1972 as a "syndrome of subacute, chronic and symmetric bone lesions" [1]. According to European registries, CRMO represents 2-5% of all cases of osteomyelitis, and its incidence does not vary geographically. The prevalence of CRMO among children is estimated at 1 per 160,000 to 1 per 2,000,000 [2]. Initial manifestations of the disease are typically observed between 7 and 12 years of age, with a median age at the onset of 10 years, and a four to one female to male patient ratio [2]. Although the first case of the disease was described nearly 50 years ago, the etiology of CRMO is still unknown. However, co-occurrence of CRMO with autoimmune disorders, such as inflammatory bowel disease, coeliac disease and psoriasis, implies that the disease is associated with a systemic inflammation with a disrupted balance between anti-inflammatory and proinflammatory cytokines and resultant immune deregulation. Research on the etiopathogenesis of CRMO included various cytokines, such as interleukins (IL), IL-1, IL-6 and IL-10, and, in particular, tumor necrosis factor (TNF) alpha, known for its pro-apoptotic effect on osteoclasts [3,4]. It is likely that the etiopathogenesis of CRMO also includes a genetic predisposition [5]. In children with Majeed syndrome, associated with a mutation of the LPIN2 gene on chromosome 18p, and patients with the deficiency of IL-1 receptor antagonist caused by a mutation of the IL1RN (DIRA) gene, the signs of non-bacterial osteomyelitis appear already in early childhood [3]. CRMO is a pediatric variant of SAPHO (synovitis, acne, pustulosis, hyperostosis, osteitis) syndrome in adults, described first by Chamot in 1987.
The first manifestation of CRMO is transient pain in the affected bone or adjacent joint. Usually, the symptoms exacerbate at night causing awakening and can be attenuated with non-steroid anti-inflammatory drugs. Rarer symptoms of the disease include swelling or redness with increased temperature of the skin. In 25% of patients, CRMO may be associated with skin lesions (acne, psoriasis, palmoplantar pustulosis), low-grade fever and loss of weight. The disease can involve any bone, and the lesions are usually symmetrical. Unifocal CRMO is found rarely, typically in the epiphyses of long bones, such as the femur, tibia and fibula (34%) or the clavicles (24%) and can be misdiagnosed as a neoplastic process [6]. Other, less frequent locations of CRMO include vertebral bodies, and the mandible and pelvic bones [7].
No specific diagnostic markers of CRMO have been identified thus far. Laboratory tests play merely an adjunct role. Inflammatory markers are usually within respective normal limits or slightly elevated. Additionally, the levels of rheumatoid factor and autoantibodies specific for systemic connective tissue disorders are normal, and no link has been found between CRMO and the HLA B27 antigen. The results of microbiological and serological tests for bacterial, viral and fungal infections are negative. Histopathological examination shows non-specific subacute or chronic osteomyelitis [5,8], with the infiltration of bones with neutrophils, plasma cells, lymphocytes, macrophages and histiocytes.
Imaging studies are vital for the diagnosis of CRMO. Bone pathologies may not be visualized at the early stages of the disease. During the first phase of CRMO, imaging findings include osteolysis, followed by bone sclerosis and bone remodeling. Another characteristic feature of the disease are periosteal reactions. The presence of osteolytic foci may raise a suspicion of a neoplastic process.
While the role of computed tomography (CT) in the diagnosis of CRMO is limited, CT can visualize small bone lesions within the sternum, spine and pelvis. Given the lack of exposure to ionizing radiation, the safest and, simultaneously, the most effective diagnostic imaging modality in detecting CRMO, monitoring its outcomes and treatment effectiveness is magnetic resonance imaging (MRI). Together with technetium scintigraphy, MRI can detect latent, asymptomatic lesions [3].
Despite many attempts, no unified diagnostic criteria of CRMO have been defined thus far. The most widely accepted diagnostic criteria were proposed by Jansson et al. [8] and Roderick et al. [9]. The algorithm proposed by Jansson et al. [8] includes four major (radiologically proven osteolytic/sclerotic bone lesions, multifocal bone lesions, palmoplantar pustulosis or psoriasis (PPP), sterile bone biopsy with signs of inflammation and/or fibrosis, sclerosis) and six minor criteria (normal blood count and good general state of health, C-reactive protein (CRP) and erythrocyte sedimentation rate (ESR) mildly-to-moderately elevated, observation time > 6 months, hyperostosis, association with other autoimmune diseases apart from PPP and psoriasis, and grade I or II relatives with autoimmune or autoinflammatory disease). CRMO is diagnosed whenever a patient satisfies at least two major criteria or at least one major and three minor criteria [8]. Alternatively, a diagnostic score can be calculated based on the observation that CRMO patients typically present with low levels of inflammatory markers, normal body temperature and bone lesions in typical locations. The patient is likely to suffer from CRMO if the overall score is 39 points or higher (Table 1) [10]. Table 1. Clinical score for a diagnosis of CRMO according to Jansson et al. [10].
In 2016, Roderick et al. [9] proposed the so-called Boston diagnostic criteria based on clinical manifestation, radiological findings, the location of bone lesions and the results of laboratory tests. Patients who satisfy four criteria (typical clinical symptoms, typical imaging findings, multifocal lesions or unifocal lesion within the clavicle and CRP < 30 g/L) are likely to present with CRMO, whereas in those satisfying the first two criteria but presenting with unifocal extraclavicular lesions and elevated CRP, the diagnosis should be confirmed by bone biopsy (plasma cell infiltration, fibrosis, sclerotization and negative result of microbiological testing).
Another algorithm facilitating the qualification of patients with suspected CRMO to require a bone biopsy was proposed by Taddio et al. [11]. According to this algorithm, the biopsy is advisable in patients in poor general condition, with elevated inflammatory markers, abnormal complete blood counts, non-specific imaging findings or unifocal disease. The decision to perform a biopsy should be made on an individual basis after carefully weighing all potential benefits and risks.
Given the non-specific course of the disease, CRMO is a diagnosis of exclusion. In patients with fever, elevated inflammatory markers and poor general condition, the primary diagnosis is infectious osteomyelitis. The unifocal character of the disease with the evidence of osteolysis on imaging studies raises suspicion of primary bone malignancy (osteosarcoma), especially if the lesions are located in the epiphyses of long bones. Skin manifestations and concomitant nail psoriasis warrant evaluation for psoriatic arthritis as a cause of bone/joint pain. Pain in the lumbosacral spine may be a manifestation of a spondyloarthropathy.
CRMO is characterized by periodical exacerbations and remissions, and, hence, the identification of patients in whom the disease will resolve spontaneously and those who require intensive long-term anti-inflammatory treatment can be challenging. The first line of treatment includes non-steroid anti-inflammatory drugs (NSAIDs). In 80% of the patients, this treatment results in complete or partial clinical and radiological remission within 6 months [10]. According to the consensus statement from 2018, patients who do not respond adequately to the NSAID therapy may additionally receive glucocorticoids (prednisone with the initial dose of 2 mg/kg/day, no more than 60 mg/day, followed by the maintenance dose of 0.1-0.2 mg/kg/day) with the evaluation of response after 6 weeks. In 65% of the patients, the treatment results in remission within 3 to 6 months [12,13]. Whenever the activity of the disease is high, the patients may be switched to diseasemodifying antirheumatic drugs (DMARDs, usually methotrexate or sulfasalazine), TNF inhibitors or bisphosphonates. The therapy should be continued for at least 12 months [14], except for bisphosphonates, which should be administered for 3-6 months [15].
The aim of this single-center observational study was to analyze the applicability of various imaging studies to the diagnosis and further evaluation of patients with CRMO.
Materials and Methods
The study included 10 patients, 7 girls and 3 boys, treated at the Department of Pediatrics, Hematology, Oncology and Rheumatology of the Regional Children Hospital in Bydgoszcz between 2016 and 2021.
The patients were referred to the Department because of non-specific complaints, the etiology of which could not be explained otherwise. The ailments reported by the patients and their legal guardians typically included persistent joint and/or bone pain without a history of previous trauma, with concomitant swelling and/or fever or without.
The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Local Bioethics Committee of The Nicolaus Copernicus University, Collegium Medicum in Bydgoszcz, Poland (protocol code: KB401/2021, date of approval: 15 June 2021). All diagnostic images shown in this paper were fully anonymized; the legal guardians of the patients gave their written consent to publish the images.
The analysis included the documentation of outpatient and inpatient treatment and the results of imaging studies: conventional radiography, MRI, CT and ultrasound. The diagnostic process varied depending on the patient's clinical condition, availability of imaging data from the referring center and legal guardians' consent for further procedures. Hence, not all patients underwent all three types of diagnostic studies at our center. Similarly, bone biopsy was not carried out in some eligible cases given the lack of consent from the patient's legal guardians.
All imaging studies were carried out at the Department of Imaging Diagnostics, Regional Children Hospital in Bydgoszcz. Conventional radiograms were obtained with an YSIO X-ray system (Siemens, Munich, Germany). CT scans were acquired with a 64-slice SOMATOM Definition AS scanner (Siemens, Munich, Germany). MRI was conducted with a 1.5 T Magnetom Essenza DOT device (Siemens, Munich, Germany). Follow-up MRIs were performed at various frequencies and time intervals (range: 3-24 months), depending on the patient's clinical condition. The results of all imaging tests were interpreted by the first author (M.K.). Qualification for bone biopsy was based on the algorithm proposed by Taddio et al. [11].
The results were subjected to statistical analysis with Statistica 10 package (Stat-Soft, Tulsa, OK, USA). Statistical characteristics of continuous variables were presented as arithmetic means and ranges, and the distributions of discrete variables as numbers and percentages.
Results
Mean age at the first manifestation of the disease was 10 years and 7 months, with mean ages of male and female patients of 12 years (range 8-17) and 10 years (range 9-15), respectively. The mean time elapsed from the first clinical manifestation of the disease to the final diagnosis was 15.4 months (Figure 1). The mean age at diagnosis was 11 years and 10 months. Alternative diagnoses considered in the study patients included infectious osteitis, Ewing's sarcoma, juvenile idiopathic arthritis and viral infection.
Unilateral lesions were found in eight patients (80%) and bilateral in two (20%). In three patients, the lesions were located in two anatomical areas (e.g., hip joint and tarsal joint), whereas, in the other seven patients, the disease was limited to only one anatomical region. Inflammatory lesions were most often located in the pelvic bones (33%) and lumbar spine (16%). Unifocal lesions were most commonly found in the tibia, fibula, clavicle, thoracic spine and sacrum (Table 2). Unilateral lesions were found in eight patients (80%) and bilateral in two (20%). In three patients, the lesions were located in two anatomical areas (e.g., hip joint and tarsal joint), whereas, in the other seven patients, the disease was limited to only one anatomical region. Inflammatory lesions were most often located in the pelvic bones (33%) and lumbar spine (16%). Unifocal lesions were most commonly found in the tibia, fibula, clavicle, thoracic spine and sacrum (Table 2). In 80% of the patients, the initial manifestation of the disease was bone pain. In 30% of the patients, conventional radiography demonstrated focal loss of bone density demarcated by an osteosclerotic zone; the radiograms of the other patients were normal (Table 3). In 80% of the patients, the initial manifestation of the disease was bone pain. In 30% of the patients, conventional radiography demonstrated focal loss of bone density demarcated by an osteosclerotic zone; the radiograms of the other patients were normal (Table 3). CT of the affected anatomical region was carried out in 30% of the patients (Table 3). CT scans demonstrated the presence of disseminated foci with a non-homogeneous osteolytic/osteosclerotic structure, with a massive loss of cortical layer and strong periosteal reaction. In the case of the thoracic and lumbar spine, CT showed a slight decrease in the height of vertebral bodies.
All patients with a presumptive diagnosis of CRMO underwent MRI (Table 3). Most patients presented with multifocal hypodense areas on T1-weighted images, with the enhancement of signal on T-weighted and STIR sequences. Additionally, the areas involved with the inflammatory process showed an intensive contrast enhancement. Furthermore, MRI demonstrated swelling of adjacent soft tissues and bone marrow edema. The abnormalities described above, in particular bone lesions, were found in all patients diagnosed with CRMO. Typical MRI findings in the thoracic and lumbar spine included the decreased height of vertebral bodies and bone marrow edema.
The duration of follow-up of the study patients varied between 3 months and 3 years. In 40% of the patients, both clinical symptoms and radiological abnormalities resolved completely, whereas another 50% of patients showed partial regression of clinical and radiological manifestations. In the remaining 10% of the patients, follow-up was too short to confirm progression/regression of the disease (Figures 2-4).
Discussion
Due to a greater awareness of CRMO as a potential diagnosis in children with bone pain, the proper diagnosis may be reached earlier. In our study, the mean time elapsed from initial clinical manifestations of the disease until the ultimate diagnosis was 15.4 months; this is a relatively short period given that the lack of specific clinical, laboratory and imaging markers hinders the diagnostic process substantially.
A characteristic radiological feature of CRMO is the presence of osteolytic foci, frequently surrounded by a hyperdense bone structure. However, none of our patients presented with such lesions on baseline radiograms. This implicates a limited applicability of conventional radiography in the detection of osteolytic foci typical for CRMO. However, conventional radiography remains a useful tool in patients with suspected CRMO as it may identify other causes of skeletal pain to be considered on differential diagnosis. The CRMO-specific lesions could be visualized on CT. Our experiences indicate that CT has higher diagnostic value than conventional radiography in detecting CRMO. However, it
Discussion
Due to a greater awareness of CRMO as a potential diagnosis in children with bone pain, the proper diagnosis may be reached earlier. In our study, the mean time elapsed from initial clinical manifestations of the disease until the ultimate diagnosis was 15.4 months; this is a relatively short period given that the lack of specific clinical, laboratory and imaging markers hinders the diagnostic process substantially.
A characteristic radiological feature of CRMO is the presence of osteolytic foci, frequently surrounded by a hyperdense bone structure. However, none of our patients presented with such lesions on baseline radiograms. This implicates a limited applicability of conventional radiography in the detection of osteolytic foci typical for CRMO. However, conventional radiography remains a useful tool in patients with suspected CRMO as it may identify other causes of skeletal pain to be considered on differential diagnosis. The CRMO-specific lesions could be visualized on CT. Our experiences indicate that CT has higher diagnostic value than conventional radiography in detecting CRMO. However, it needs to be stressed that although, due to technological advancements, the dose of ionizing radiation absorbed during CT is relatively low, it is still high enough to interfere with the development of pediatric patients.
MRI can visualize bone marrow edema within the involved bone, which typically manifests as hypointense areas on T1-weighted images and hyperintensity on T2-weighted and STIR images with strong contrast enhancement [16]. In this study, MRI proved to be the most accurate diagnostic option, demonstrating the presence of inflammatory lesions in all examined patients, even those with no abnormalities on ultrasound and conventional radiography. These observations imply that MRI plays a vital role in the evaluation of patients with presumed CRMO. The diagnostic value of the method seems to be higher compared with conventional radiography and CT.
MRI also remains the most accurate diagnostic option in patients with relapse. In 30% of patients included in this study, MRI demonstrated progression of primary foci as well as secondary foci, usually in the proximity of the primary foci or in the thoracic/lumbar spine. Clinically, such patients presented with pain of the involved anatomical area and/or impaired function of adjacent joints. Our study also confirmed the role of MRI as a gold diagnostic standard in patients with remission. In such patients, MRI showed a substantial decrease in the size of inflammatory foci within the bone, regression of bone marrow edema and resolution of lesions in adjacent soft tissues. Such radiological presentation correlated with partial or complete remission of clinical symptoms.
The results presented above confirm the leading role of MRI in establishing the correct diagnosis of CRMO, monitoring of treatment outcomes and detection of new lytic foci during relapse and progression of the disease. It needs to be stressed that, apart from being an accurate diagnostic option providing high-quality images, MRI is not associated with exposure to ionizing radiation and contrast agents are generally safe. Optimally, each patient with suspected CRMO should undergo a whole-body MRI as the gold standard [17]. This examination allows for the detection of all skeletal inflammatory foci, including clinically latent ones, in a relatively short time. State-of-the-art MR scanners can produce such high-quality whole-body images within 40 min. This will facilitate the planning of the therapeutic process and monitoring of its outcomes [18].
Notably, effective cooperation between pediatricians and radiologists will facilitate the proper diagnosis of CRMO without exposing patients to an invasive procedure, such as bone biopsy. No biopsy is needed if the presentation of the disease on imaging studies is typical [19], with multiple inflammatory foci [14,15]. Biopsy should be considered mainly in less typical cases with unifocal location and non-characteristic presentation on diagnostic imaging. In such cases, the lack of neoplastic cells in the biopsy specimen, and the presence of lymphocyte infiltrate, osteonecrosis, and/or bone marrow edema/fibrosis support the diagnosis of a non-bacterial inflammatory disease. Our series included three patients eligible for bone biopsy (patients no. 3, 5 and 7 in Table 3). In two of those cases (patients no. 3 and 7), the biopsy was not carried out given the lack of the legal guardians' consent and the fact that the patients did not satisfy some criteria proposed by Taddio et al. [11]. In patient no. 5, with changes in the left clavicle, a bone biopsy was carried out at another center before referral to our department, which excluded malignant character of the process.
An unquestioned strength of this study stems from the fact that it included a relatively large group of patients with CRMO, a rare clinical condition. However, this study also has some limitations inherent to its observational character and associated primarily with the lack of standardization of the diagnostic process. Hence, the results presented herein should be considered as guidance for further research rather than a basis for ultimate conclusions regarding clinical practice. An important question which needs to be addressed in the nearest future is how to expedite the diagnostic process in patients with suspected CRMO. With no doubt, the process needs to be standardized through a set of evidence-based guidelines and should be minimally invasive yet cost-effective. Another direction of future research should be the identification of a highly accurate diagnostic marker of CRMO, whether an imaging finding or a laboratory parameter.
Conclusions
MRI remains the gold standard in the diagnosis and monitoring of CRMO. While bone biopsy has unquestioned value in cases were the imaging findings are atypical, its applicability can be limited in pediatric patients given frequent lack of parental consent for this procedure. This puts particular emphasis on effective cooperation between clinicians and specialists in diagnostic imaging during the course of the diagnostic process. Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.
Data Availability Statement:
No new data were created or analyzed in this study. Data sharing is not applicable to this article.
Conflicts of Interest:
The authors declare no conflict of interest. | 2021-09-28T05:15:23.143Z | 2021-09-01T00:00:00.000 | {
"year": 2021,
"sha1": "417767627097394406bc38af3d5996e529cd7afe",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2227-9067/8/9/792/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "417767627097394406bc38af3d5996e529cd7afe",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
210496803 | pes2o/s2orc | v3-fos-license | Beyond the Black and White: Female Domestic Servants, Dress and Identity in France and Britain, 1900-1939
ABSTRACT Far from being a trivial detail, clothes fundamentally define who we are and how we are perceived by others. Drawing on a large sample of French and British servants’ memoirs, this article explores how dress served a crucial but contrasting role in the way French and British servants articulated their identities within and outside the home between 1900 and 1939. It argues that servants’ dress was deeply linked to the nature of the occupation in each country and the structure of their respective female labour markets.
Introduction
Autobiographies are important sources for the study of marginal groups whose voices do not boom through the official archives. In particular, they are one of the only avenues available to recover the experiences of female domestic servants. These documents have helped historians understand the economic struggles faced by these women and the social trajectories they followed. 1 But autobiographies are also bursting with revealing digressions and significant details, memories of everyday life, of what was eaten on a certain day, which film was seen, what dress was worn. Rather than overlooking these banal recollections, this article seizes upon them to explore the everyday lives of domestic servants. It is based upon the premise that to understand marginal groups we must actively read at the margins of their stories.
One example is the case of Irene Phillips, who decided to write her memoirs about her childhood in an English village at the turn of the century and her work as a housemaid in the imposing English country houses of the interwar period. As she sat down to write, the dress she had donned on her first day of work vividly came back to her memory. She opened her memoir with the following: 'A servant girls' life was very hard and you were expected to wear stripped cotton dresses and aprons and caps for morning and afternoon black dress, white lace apron and lacy cap.' 2 In an interview about her time in service, Paulette Belaire, a French maid in the 1920s, was asked whether she wore a uniform and struggled to answer: 'I don't know, I dressed like I wanted. Did I wear an apron? Of course, I did, like any fourteen years old girl.' 3 These passing remarks on dress set alongside a broader discussion of these women's occupational trajectories and work conditions are easy to miss. However, they reveal that, far from a simple dressing of the main text, the contrasting discourses that the two women created around their dress can help us better understand what it meant to be a domestic servant, to wear a carefully selected white lacy cap and apron every day or, on the contrary, to done a work apron without giving it a second thought. That Irene Phillips returned to her dress years after the events highlights the strong emotions that her uniform evoked. Paulette Belaire's experience in France was entirely different. She seemed bemused by the question regarding the existence of a servant's uniform. While these two women had the same occupation, there was a stark contrast between Irene Phillips who, by putting on her black and white uniform, took on the distinctive identity of the domestic servant in Britain and Paulette Belaire, who painted herself as a more elusive figure dressed like any fourteen-year-old French girl.
This article explores how dress served a crucial but contrasting role in the way French and British servants defined their identity. 4 Clothes provide protection from the elements, but they also fundamentally define who we are and how we are perceived by others. 5 They can indicate the wearer's occupational and regional identity, class, gender, age and religion. They can empower and confer respectability, but they can also be used to punish, shame, ridicule and exclude. The bulk of the historiography on dress has focused on the clothing of the elite or items with more public symbolism such as ceremonial dresses, court clothing and military uniforms. 6 Historians of clothing have adopted this narrow focus because of the role of the upper classes as trend setters with a more vibrant, extravagant, and varied fashion and because upper-class clothing has been more carefully preserved. Notable exceptions to this trend include Daniel Roche's seminal work on the wardrobes of the rich and the poor in Ancien Régime France, as well as Vivienne Richmond and John Styles' ground-breaking studies of the clothing of the poor in England in the eighteenth and nineteenth centuries, which show the important symbolism of dress in all categories of society. 7 The article builds on Roche, Richmond and Styles' approach to clothing by focusing on the way servants articulated their identity through their dress.
Domestic manuals and employers' accounts mention the servants' dress solely as a conspicuous item to forward the social status and respectability of the employers. This article, in contrast, is based on research drawn from the largest sample of servants' autobiographies collected to date, composed of 103 British and 97 French female servants' autobiographies and interviews. 8 While this type of qualitative method has already been used on a smaller scale by historians of domestic service in Britain, the French historiography has largely relied on employers' accounts and descriptions of servants in novels to build its picture of domestic service. 9 The article introduces new materials to the study of French domestic service by tapping into the rich resources of the Association pour l'autobiographie et le patrimoine autobiographique, which houses more than twenty unpublished twentieth-century servants' autobiographies. These sources provide the opportunity to analyse the importance of dress in the selffashioning of servants, on which conventional sources have very little to say. Servants' autobiographies are key sources to study their work dress, because, as Jennifer Craik points out 'anecdotes about uniforms involve formative moments of self-hood.' 10 The evidence from the autobiographies and oral interview transcripts is nevertheless corroborated with other types of sources such as newspaper articles and official reports to place it in historical context. Servant's discourses about their clothes are considered in connection with the societal debate called 'the servant problem' or 'servant crisis'. The period between the early twentieth century and the start of the Second World War saw a fall in the number of domestic servants which lead commentators to speculate about the future of the occupation. The extent of the decrease in size of the servant labour force and the scale of the debate, however, were much more important in Britain than in France. While French commentators discussed the insubordination of domestic servants and the growing tension with their employers, there was a real sense of crisis in Britain as employers thought that the rise of new employment opportunities in the service sector and changing attitudes regarding the social hierarchy had sealed the fate of domestic service. 11 The debate about domestic service was carried in the news and filled pages of government reports but it also struck at the heart of British households. Servants and employers negotiated new definitions of domestic service, argued over personal boundaries and, ultimately, debated the extent of the employers' authority over the servants' lives. Because dress was a critical tool to define, control and classify workers, it was at the core of what contemporaries called the 'servant problem'.
This paper contributes to our understanding of the nature of domestic service in France and Britain by showing that far from a trivial detail, French and British servants' dress played a crucial role in defining their identity and stood as a visual reminder of the status of their occupation in their respective society. The first section of this paper sets the scene by outlining the way domestic service in France remained connected to the productive sphere much later than in Britain where service was an occupation solely dedicated to domestic tasks following the separation of home and work. The second section provides a description of the uniform commonly worn by British servants and argues that its use was justified by the hierarchical and conspicuous nature of domestic service in the British economy. In contrast with the British servants' uniform, it is shown that most French servants wore ordinary clothes because of their place within the French family-economy. In the third section, it is argued that the presence or absence of a uniform fundamentally shaped the type of identities that French and British servants adopted within their employer's home. While domestic service might stand as a 'reference point for the establishment of class difference' in Britain, the study of servants' dress reveals that this narrative had little relevance in France where rural identities predominated. Finally, the fourth section compares the servant's uniform to other female uniforms in both countries and reflects on how British servants struggled with the symbolism of their uniform in a society where female workers enjoyed an increasing amount of independence.
Home and work: the nature of domestic service in France and Britain Domestic service was staple part of many young women's lives between 1900 and 1939. In Britain, domestic service was the primary employer of women until the Second World War. Fewer women were employed in service in France as the agricultural sector continued to recruit a significant share of the population. Nevertheless, around fourteen percent of the French female labour force was in service. 12 Servants were employed in a range of households, from grand upper-class mansions with a large staff to middle and lower middle-class households requiring just one or two servants. Up to the Second World War, most servants lived in their employers' home. This close proximity where constant contact and intimacy were inevitable meant that domestic service was often perceived as more than just a job. It was a key site where people from different social classes, generations and regions met each other and negotiated their respective identities. Domestic servants, therefore, had 'a compelling presence' in French and British economic, social and cultural life. 13 While domestic service was an important sector of employment and a key cultural institution in France and Britain, the occupation was shaped distinctly by their respective national economies. The industrialisation of Britain in the eighteenth and nineteenth century meant that the model of the family economy, in which all family members and household residents played a role in the survival of the family as an economic and social unit, disappeared. 14 While historians have agreed that this process of separation between workplace and home was rather 'protracted and tortuous', by the twentieth century, the distinction was undeniable and goods were no longer produced within the household. 15 The impact of industrialisation in Britain has been examined mostly in regard to the agricultural and industrial sector, but the consequences on service were also significant. Domestic servants had been traditionally defined in broad terms as 'persons employed by men of trades and professions under them to assist them in their particular callings; or, such persons as others retain to perform the work and business of their families'. 16 Servants were employed for a variety of tasks: on farms, they helped with the dairy and nursed children; in shops, they sold the goods and cleaned the counter. Because servants could be found in all sorts of trades, domestic service was more characterised by a type of relationshipworking for a master and living within his householdthan an occupation dedicated to a specific task. Servants' participation in domestic and productive activities alongside the family unit, however, disappeared as work was exported outside the home and rationalised within larger structures. In the agricultural sector, the employment of female servants who took care of the home while helping to raise hens or weed the fields was incompatible with intensive capitalist arable farming, which relied instead on day labour. 17 In parallel, the rise of chain stores and lock-up shops contributed to the dissociation between the homes of shopkeepers and their place of work. 18 While in family-businesses, servants might help make and sell goods while taking care of various domestic tasks, these two types of activity were increasingly separate in the twentieth century. A shop assistant was employed to help in the shop while a servant was hired to take care of the shopkeeper's suburban home. In 1900s, Jessie Henderson, for example, took care to ensure that 'everything was in order' before her employer, a grocer, left his suburban house to work at his shop where his assistants awaited. 19 Shopkeepers who could not afford a suburban semi-detached villa and had to live above or next to their shop also made sure to maintain a careful distinction between rooms identified by functions. In these conditions, the servant had little to do with the shop as she dusted the backrooms. 20 In the household of a Jewish shopkeeper in the East End of London where sons and daughters helped their parents run the shop in the 1930s, Winifred Foley, a general maid, solely took care of domestic tasks: 'When business was good', she wrote, 'they would work fifteen hours a dayand I worked fifteen hours a day for their domestic comfort.' 21 By the twentieth century, therefore, domestic service was an occupation solely dedicated to the care of the home in Britain.
In contrast, France experienced a slower rate of urbanisation and industrialisation. In 1901 more than half of the French population still lived in the countryside and around 40 percent of female labourers worked in agriculture. Not only was the agricultural sector far bigger in France than in Britain, it also offered more opportunities of employment to women. In contrast with the capitalist, intensive and specialised style of farming which came to dominate British agriculture, 61 per cent of the agricultural labour force in France worked on their own account in 1906. 22 This different type of agricultural organisation had long-term consequences for women's employment. On small French farms, the family economy continued to thrive and required the work of both men and women. Women laboured in the fields, at home, and also often sold the production on nearby markets. 23 The French economy also contained a significant share of small familyowned shops and businesses in which women did all sorts of domestic and productive tasks. In 1939, half of the retailers in France were small structures located in rural areas far from the modern lights of the Bon Marché department store in Paris. 24 Many French servants recalled in their autobiographies that they worked in the shop, sold food in the market, distributed oats for the coachmen's horses, made deliveries by bicycle, alongside cleaning the house and caring for the children. 25 Yvonne Cretté-Breton, for example, worked as a maid in a bakery. 26 She cleaned the house and helped with various tasks in the shop: 'On Sunday, I also had to cook twenty roasts; on Monday, I had to put some cinders into bags with a shovel; on Tuesday, I had to clean the copper and marbles with the help of the bread delivery girl.' 27 A deposition for a court case made by a servant who was accused of stealing money from her employers in Lyon in 1934 also highlights the difficulty of trying to narrow down the servant's occupation in France. When asked what her occupation was, Marie Bouget gave the following answer to the jury: 'I served M. Michalon who lived at 135 Roanne street, Central Café, the whole time he was in charge of this business (. . .) I was a waitress and I also did the housework. As a matter of fact, I was a maid-of-all-work.' 28 Marie Bouget did not say that she did waitressing or housework as a side activity. Both seemed of equal importance to her and we cannot reduce her to a servant who did a bit of waitressing on the side of her domestic service activities. Yvonne Cretté-Breton and Marie Bouget's experiences show the multifaceted aspect of domestic service in France as servants cared for children and animals, washed the dishes and served the clients at the shop. Domestic service was a specialised type of occupation dedicated to the care of the home and its inhabitants in Britain, while the same category in France corresponded to an older understanding of servicecommon in Britain until the eighteenth centuryin which servants were not only those who performed domestic tasks but anyone who left their home to live and work in someone else's.
Dressed for the job: the servant's uniform
The way that French and British servants dressed was directly connected to the different nature of the occupation in their respective economies. Each of the 103 British female servants in my sample described themselves wearing a uniform. It was a common experience regardless of the social standing of the households which employed them. In wealthy households that employed more than one servant, the uniform varied according to the servant's position within the staff hierarchy. A parlour maid wore a morning uniform to do the cleaning and an afternoon uniform to serve at the table and welcome guests. Eileen Balderson started her career as a between-maidi.e. a junior housemaid at the bottom of the staff hierarchyin a country house in 1931 and later moved up to the position of housemaid. She recalled that her employer asked her to wear 'a striped print dress, usually blue and white, but [sometimes] pink or mauve and white' with 'a large white apron and a cap' and 'black shoes and stockings'. In the afternoon, she changed into 'a black dress with the same apron and cap as in the morning'. 29 In contrast, cooks, kitchen maids and scullery maids did not change outfits for the afternoon as they stayed in the kitchen all day and were not on display. Nannies, housekeepers and lady's maids also sported different styles of dress to underline their higher position in the servants' hierarchy. Most servants worked as maids-of-all-work in middle-class households and wore a similar uniform to the housemaid. Elsie Oman, for example, was a maid in a 1920s middle-class household that had gone bankrupt and in which the daughter of the house was helping with domestic tasks. Despite working in a more modest household, she was still required to wear a print dress for the morning and a black dress with 'white collars and cuffs' for the afternoon. 30 Urbanisation and industrialisation contributed to a public anxiety concerned with the difficulty of distinguishing people according to their rank in the anonymous crowd of the city. In this context, the uniform was used to create a distinction between the mistress and her maid, to underline who was inferior and who had power in the household. 31 As the nature of the relationship between servants and employers became less intimatethey did not work alongside each other anymorethe servant's uniform was designed to articulate visible status distinctions. 32 The famous black and white uniform that servants wore in the afternoon was also symbiotically linked to the 'domestic' nature of the occupation. As domestic servants stopped being involved in productive activities, their role became increasingly one of social display. While servants spent backbreaking hours cleaning, washing and dusting the various rooms of their employer's house, they also helped their mistress entertain her visitors. For the idle if not slightly bored middle-class wife waiting for her husband to return from work in the evening, calling on friends and acquaintances in the afternoon was a way to pass time as well as an important social practice to maintain her place in the social circle. 33 As a result, servants wore an afternoon uniform so they would be suitably dressed to impress potential callers when they greeted them at the door. The material of the afternoon uniform was delicate if not extravagant to represent the status of the employers. The white apron, collar and cuffs also highlighted the neatness and cleanliness of the servant. 34 Being able to employ a maid solely dedicated to menial domestic tasks that would not stain her pristine white uniform was a mark of respectability.
While representations of domestic servants in literature, paintings, films and domestic manuals might lead us to think that French servants wore a similar uniform, servants' accounts tell a different story. 35 The popular figure of the French maid in a becoming little black dress and a dainty white apron was a carefully constructed cultural figure that was used to export the fashion and sophistication of the French bourgeoisie to both national and foreign audiences. This classy and often erotic figure in uniform, however, was far from representative of French domestic service. In reality, few French servants remembered wearing specific clothes for work. Those who worked for the provincial bourgeoisie, shopkeepers, artisans, or on farms, did not wear a uniform. For example, Juliette, who worked for a small middle-class household in Lyon, was told by her mother that her old dress, a patched skirt and a smock 'would do' for her work dress. When she arrived at her employers' house, they only requested that she wear an apron over it. 36 While white aprons were the most common, some women also recalled wearing grey and black aprons. 37 The cap, one of the key elements of the distinctive dress of the domestic servant in Britain, was mentioned only once in the sample of autobiographies. Even then, the servant who mentioned it strongly suggested that it had disappeared from use in most households by the 1920s, for she denounced her employer as being 'stuck in the past'. 38 The rules regarding servants wearing caps, black dresses and dainty white aprons were only followed by the urban elites. For example, when Marie worked in Paris for an upperclass courtesan after a few jobs in some provincial households, she was asked to wear a short afternoon apron decorated with embroidery. This outfit made her feel like 'an operetta lady's maid with this small apron that doesn't even reach my knees. I don't feel comfortable.' 39 Marie's reaction suggests that she saw aprons as a regular accessory that was useful to protect a servant's dress and make her look clean, which in turn highlights that decorative aprons were not regularly used in smaller provincial households. The only mention of a black dress in the sample of autobiographies came from Marie Tual who worked for a countess in the 1930s. Having worked in smaller households beforehand, she thought that the imposition of a black dress was an odd request and 'laughed' at what she called a 'costume'. 40 By referring to the distant world of the stage to make sense of their new uniforms, these two women's accounts show their unfamiliarity with the performative and deferential aspect of domestic service in upper-class households.
In the countryside, where half of the French population lived until the Second World War, the maid was often part of an informal network which made the need for a uniform irrelevant as she was already known to be a maid within the social structure of the village. Moreover, in the countless small family businesses spread around the four corners of France, women did all sorts of domestic and productive tasks alongside their employers for which they often simply wore an apron over their everyday dress. Servants were employed more for the hard work that they could do, whether it was in the kitchen, the henhouse or at the shop's counter, than for representing the social standards of the household by wearing a uniform. This finding complements Diana Crane's research on the fashion expenditures of the wives of farmers and skilled workers in late nineteenth-century France. 41 Using Frédéric Le Play's collection of household budgets, she discovered that, in contrast with the process of emulation among the lower-middle class in Britain, the wives of farmers, some with a significant income, and those of artisans and craftsmen living in the French provinces did not buy many fashion items to copy the urban middle class. In a society where women continued to play a productive role within the home, there was little sartorial emulation and thereby, little need for a servants' uniform.
Domestic servants and their dress: a feather in their cap?
Whether or not servants had to wear a uniform led French and British women to have drastically different experiences of domestic service and build different types of identities around their occupation. Because working-class women paid a particular attention to their clothes as a publicly visible form through which they could demonstrate their respectability, the extent of servants' ability to control what they wore had a profound impact on the way they made sense of their position in society. 42 Some British servants were proud of their uniform, especially in the few exceptional cases where it was provided by rich employers rather than made or bought by the servant herself. Mrs Woodburn, a housemaid in Lancaster in the 1900s, was thrilled with the new uniforms the maids were given for a family wedding: 'We were all dressed in grey, lovely silver grey, shiny like alpaca, in them days, nice plain pinnies, no lace, but happen a lot of tucks round the bottom.' 43 In these conditions, the uniform was the mark that one served a prestigious household and was well-placed in the hierarchy of domestic servants. The acquisition of new items of clothing was also an exciting event for those women who often owned very few clothes and were used to wear hand-me-downs from sisters or mothers; although this practice was subsiding by the 1930s. 44 Other servants like Mollie Moran, a kitchen maid in an upper-class London in the 1930s, were proud of their uniforms for different reasons: I've heard of some scullery maids who were ashamed to wear their uniforms outside, for fear of being seen as a skivvy. Not me! I was proud of it and I wore my apron like a badge of honour. I had a job and was sending money home to my mother. That meant I was respectable. 45 The uniform marked Mollie Moran's place in a class of respectable workers, who, even if they did not earn much, were nonetheless able to provide for their family in contrast to the badly clothed, unemployed and charity-dependent poor. It contributed to her pride in becoming a wage earner and, thereby, gaining an adult or independent status. While Mollie Moran was a severe critic of the hierarchical system of domestic service in other parts of her autobiography, she nonetheless had a positive discourse about her uniform.
Her reaction illustrates what Lucy Delap calls the 'complex emotional landscape' of domestic service and the wide range of ways servants made sense of their personal experiences in service.
For many British servants, nevertheless, the uniform carried a darker symbolism. They resented having to wear a garment that was used to create a distinction between the mistress and the maid. While it was respectable to be employed and earn a wage, for many, there was little respectability in complying with the rituals of middle-class employers to the point of losing their sartorial and personal freedom. Violet Firth was a middleclass woman who experienced domestic service during the First World War. She joined the Women's Land Army and worked as a gardener for private households. She wrote her memoirs during the interwar period, hoping to contribute to the 'servant problem' debate and raise awareness about the harsh social and material conditions of domestic service. Having been on both sides of the mistress-servant relationship, she felt even more strikingly the way that clothing drew a boundary between two women, giving authority to one and forcing the other into subservience. She recalled: The woman in the fashionable silk frock tells the woman in the washed-out print dress that she cannot afford to give higher wages. In the evening, the woman in the kitchen uses her scanty leisure to patch the dress, and so make it hang together a little longer. Then she goes into the dining room to serve dinner and sees that the silk frock has been changed for a satin frock. Next evening, she patches the print dress again, because the material, never strong, is rotten with wear and washing. When she goes into the dining-room she sees that the satin frock is replaced by a lace frock. Is there any need for me to detail the feelings of that woman towards her employer? 46 The dress of the servant became a signifier of class, highlighting the difference between the middle-class woman and her servant. The afternoon apron and cap, often called by servants 'a badge of servitude', were further symbols of the servants' inferiority. Having long fallen out of fashion by the 1880s, the cap was viewed as an anachronistic imposition on servants that singled them out and emphasised the power that employers had over them. 47 When Jean Rennie learned that she would have to give up her hope of a secondary education to become a servant, she was apprehensive about having to 'submit to the badge of servitudea cap and apron'. 48 Hannah Mitchell, who would later become an active member of the socialist and suffragette movement, insisted on the distinction between normal aprons and decorative aprons used in domestic service. She argued that while she was used to wear an apron to work, she could not stand the idea of having to wear a special cap and apron to serve at her employers' table. She 'absolutely refuse[d] to don the muslin badge of servitude' because of the inferior and servile status it embodied. 49 Far from a simple whim, servants' reluctance to wear caps was commented upon in newspapers, governmental reports and manifestos dedicated to the future of the occupation. As early as 1891, Reynolds's Newspaper noted that 'the servants' cap, that badge of servitude stops many girls from becoming servants.' 50 This belief was put more strongly in the 1916 report on domestic service by the Women's Industrial Councila organisation promoting the interests of women at work -, which found that 'the caps were generally referred to as the trademarks of modern slavery'. 51 The denunciation of caps as badges of servitude was still in vogue during the interwar period and served as a rallying point in the campaign for the modernisation of domestic service. The Joint Committee of Industrial Women's Organisations at the National Conference of Socialist Women in Blackpool in 1931, for example, put forward a proposal to abolish 'the servant's cap as a badge of servility'. 52 It was argued that a more modern relationship between servants and employers included a practical uniform that did not single out the servant only to show off the employer's status.
Servants also experienced a sense of inferiority because of how the uniform physically felt to wear. In the 1916 Women's Industrial Council report on domestic service, some women complained that the black dress they were required to wear in the afternoon was too warm in the summer. 53 In servants' autobiographies, Jean Rennie noted that when working long hours, sweating from carrying buckets of water, her black woollen stockings were uncomfortable as the wool absorbed the moisture and became itchy and heavy. 54 Monica Dickens who worked as a general maid in London also criticised the unfunctional cap which kept falling in front of her eyes when she was trying to work in the 1930s. 55 Servants felt their inferior position through their impractical and uncomfortable uniforms, which were maintained only for the sake of their employers.
Because the main aim of the uniform was to enforce a strict social hierarchy and to represent the social standard of the household, some servants felt that it reduced them to a fixed identity, thereby suppressing their individuality and self-esteem. Violet Firth argued that the uniform constructed the servant as an 'other': 'that is a shock to them [the employing classes] to find human nature akin to their own concealed by a servant's apron.' 56 Dorothy Fudge, a parlour-maid in the 1920s, was angry when her employers used her uniform to stress her inferior status and suppress her identity, even when she was off-duty: We went to Bournemouth with them [her employers], and on arrival, we were told they would meet us at a certain restaurant on the seafront. When the time came, we were having tea in the 'best room', and feeling very grand, when the colonel's sister came in. She told us Mr and Mrs Adams would be late picking us up and then, in a very loud voice, said: 'I hardly recognised you! You look so different without your caps!' We could have hit her. 57 With her cap, Dorothy Fudge was a faceless maid, hence the surprise of her employers when they saw her off-duty in her every-day clothes. These women were alienated from their own appearance in order to help their mistress and master maintain a respectable image.
Finally, the uniform was not only a symbol of servants' inferior status, it was also an important financial struggle for the many women who had to pay for their uniform out of their meagre income. One servant, for example, wrote to The Common Cause journal in 1911 and complained that she was required to buy a uniform out of her £10 annual wage, leaving very little spare money. 58 It is relatively difficult to estimate the cost of the uniform as many servants bought some fabrics to make their own rather than purchased readymade clothing. Servants' testimonies, nevertheless, highlight that in many cases, these women worked for free for the first months of their employment as the money they earned barely covered their travel and clothing expenses. 59 In 1904, Kate Taylor, a general maid who was paid 15d per week, took six months to repay 'the necessary print for dresses and hessian for aprons'. 60 By the interwar period, buying the uniform was still an important financial hurdle. In 1933, Mrs Slade who earned £18 a year as an 'in-between maid' had to repay £12 for her uniform. 61 This system contributed to the power imbalance between servants and employers and restricted servants' opportunities, for they could not leave service early without empty pockets.
In contrast with the heavy emotional weight that the uniform bore on British servants, few French servants thought about their work dress as a fraught marker of class within the employer-employee relationship. This attitude partly came from the different composition of the servant labour force. Since around 33 per cent of female servants came from farming families, many of these women were not the obvious inferiors to their employers. 62 The daughter of a farmer who could return to inherit part of the property after her marriage, for example, was not the social subordinate of the next-door farmer or shopkeeper for whom she worked.
The numerous servants who travelled to a town or city to find a new position nonetheless found that their clothing cast them apart. This feeling of inferiority, however, was linked to their rural identity rather than their social class. In Britain, a large share of servants came from urban areas and those who migrated from the countryside did not tend to have a very distinctive style or identity as the communities they left were often barely different from those in nearby towns by the interwar period. 63 In 1921, whitecollar workers were the third largest occupational group in English county districts after domestic servants and agricultural workers. 64 In contrast, it was a much more striking experience for French servants to leave the tightly-bonded agricultural communities, some with their own dialect and style of dress. The French servant's slightly oldfashioned clothes, her accent and her unfamiliarity with urban ways marked her as an 'other'. 65 Marie was told by a friend as she stepped out of the train in Paris that her wide brimmed hat looked like a shepherdess' hat and that she should wear a fashionable cloche hat instead. 66 Others noted that their black dresses were old fashioned compared to the more colourful outfits of the Parisian women or that their long-braided hair contrasted with the trendy cropped hair of the 1920s. 67 While the daughter of a small-property holder farming family might be superior to an urban employee in the strict economic sense of 'class', her rural origins could be used to make her feel inferior. The rural-urban continuum was an alternative system of identity within servants' lives. When Jeanne Bouvier made the mistake of dropping the clean laundry on the floor, for example, her employer severely reprimanded her and called her 'you peasant!'. 68 This incident suggests that her employer used Jeanne Bouvier's rural origins to put distance between the maid and the rest of the family. It also reveals that the insult was meaningful for Jeanne Bouvier who distinctly remembered this precise incident and the humiliation she felt afterwards. While both French and British servants had to come to terms with feelings of inferiority in domestic service, their experiences were mediated through different discourses about class and rurality.
A new look: comparing servants' and female workers' dress
The distinctive uniform of domestic servants in Britain stigmatised them both within the household and made them feel inferior to other workers outside the house. As servants dealt with suppliers and delivery men to fill their employer's pantry or as they accompanied their mistress on errands, they felt self-conscious about their uniform. The garment marked the distinct status of the domestic servant compared to other female workers. Young women's employment opportunities beyond the domestic sphere radically expanded between 1900 and 1939 as the number of women working in the retail and the clerical sector grew. 69 These new jobs did not always provide women with higher wages but they offered increased independence and more free time to workers who were not expected to live on their place of work. 70 For example, while most British servants only enjoyed one half-day off during the week and every other Sunday afternoons, most workers had their evenings to themselves and the entire Sunday off. The significance of those changes on the female labour market was heightened by the campaign for women's political enfranchisement which associated women's political emancipation with the public world of work and its promise of social and financial independence. While the factory or the shop girl fitted with the suffrage representation of 'modern and emancipated womanhood', the domestic servant was looked down as a passive and servile worker stuck in the private sphere. 71 From both a social and political standpoint, therefore, domestic servicewith its lack of private space, sparse free time and nosy mistresseswas casted as a dated occupation.
However, it is not the case that domestic servants felt singled out because they were the only ones to wear a uniform in the crowd of modern female workers. On the contrary, there was an increasing number of workers who wore uniforms to advertise the company they worked for, to distinguish between ranks in organisational hierarchies or to assert their professional status in the twentieth century. 72 However, these new uniforms were fundamentally different from those of domestic servants. Diana Crane argues that the nature of uniforms changed in the twentieth century as they became increasingly linked to an occupational status with little bearing on the workers' identity beyond the workplace. 73 In this context, the servants' uniform was dated not only because of its style but because of the broader message that it vehiculated about the servants' place in the social hierarchy. While modern uniforms highlighted the skills or the types of job performed by female workers, the servant's uniform was more encompassing and cumbersome as it defined these women's social status and hierarchical relationship with their employers. The contrast between the status of domestic servants and other female workers fed the heated debate around the 'servant crisis'. A 1919 Ministry of Reconstruction report, based on interviews with a panel of domestic servants and employers, noted that: 'the fact cannot be denied that domestic workers are regarded by other workers as belonging to a lower social status. The distinctive dress which they are required to wear marks them out as a class apart, the cap being generally resented.' 74 A chambermaid felt similarly distressed about the way her uniform was looked down upon by other female workers: 'The greatest trouble with service is having to wear a cap and apron. Shop girls and business girls look down upon servants for that reason.' 75 Domestic servants, therefore, suffered from the comparison of their uniform with the dress codes of the new professions and the more popular occupations in the service sector.
In many British servants' autobiographies, the authors admired the nursing profession and considered its uniform as a standard to emulate. For Anne Kynoch, leaving domestic service for nursing was a personal triumph: I had already made several applications to different hospitals for entry as a probationer nurse. Nothing has given me a greater feeling of achievement than the receipt of a letter from the matron of a Yorkshire hospital for infectious diseases. All the world was mine! After drudgery and struggle the ladder was in place now for the climb. 76 The comparison between nursing and domestic service was also frequently made in the press because nursing was seen as an occupation that had successfully transformed itself. 77 In the nineteenth century, the 'new nurse' rose in status from that comparable to a maid-of-all work in contact with bodies and dirt, to a respectable middle-class woman who was in a position to heal not only the physical diseases but also the moral sins of her patients. The profession also became highly desirable for lower-class girls like Anne Kynoch who had aspirations for a respectable and skilled occupation. 78 The desire to become a nurse was paired with an admiration of the nurses' uniform. Elsie Oman, for example, commented that before she became a servant, she and her friend 'both thought we would like to be nurses and wear a smart uniform.' 79 Margaret Powell, a kitchen-maid in the 1920s, drew a contrast between the despised servant's dress and the respectable nurse's uniform: 'My mistress wanted me to wear a cap, but I wouldn't. It always struck me as a badge of servitude. I know nurses wear caps but somehow it's different with them.' 80 She thought that the uniform of the nurse had a different and more attractive meaning to the servant's. Rather than a 'badge of servitude', the Red Cross uniform, in particular, was 'the proudest badge'. 81 During the interwar period, the memory of women's important contribution to the war effort during the First World War gave a heightened meaning to the nurse and her uniform, while the maid was associated with domesticity far from the front.
The position of shop assistant was another object of envy for servants. When Lily Kerry explained what job she first wanted to do before ending up working as a servant in 1937, she painted a sort of hierarchy of the type of employment a young girl could aspire to: When I was about 12 or 13 years old, our teacher asked us what we wanted to do on leaving school. (. . .) I said I wanted to be a nurse, but my teacher said I would be a domestic servant. Unfortunately, you couldn't start training to be a nurse until you were 17. What I would have liked to have done before then was to work in a shop selling clothes and hats, but my mother had other ideas. 82 This hierarchy of employment, which placed shop work before domestic service, indicates the popularity of the figure of the shop assistant amongst young women. This trend was bitterly commented upon in the introduction of the 1916 report on domestic service: 'partly owing to the multiplication of other openings for women's work (. . .) the most promising girls are apt to prefer lower wages, less material comfort, and much less security of employment in shops (. . .) to the often-quoted advantages of domestic service.' 83 With the rise of the department store at the end of the nineteenth century and the feminisation of shop work after the First World War, an increasing number of young and single women were employed in big stores and chains. 84 The number of women working as shop assistant rose from 87,000 in 1861 to 500,00 by 1914. 85 Shop work was desirable because of the independence it gave to these young women, especially after the First World War, when the employees did not have to live at the shop. 86 Sally Mitchell's study of girls' magazines between 1880 and 1915 highlights how shop work was valued over domestic work as it replaced paternalism with a market relationship. In one fictional story published in Girl's Own Paper, a character who first worked as a companiona sort of genteel domestic servantbefore becoming a shop girl explains that she 'would much rather there was nothing but a "money" bond between me and my employer. (. . .) From ten to six I am at his order, but all the rest of the time I am independent, and I needn't feel grateful to him, because my salary is fairly earned and just a matter of business.' 87 If the shop girl worked in some drapery establishment, she might model the products which were on sale, feeding the fantasy of upward mobility. 88 In most cases, however, the shop assistant wore an elegant although discreet uniform. She was required to dress all in black or white, sometimes specifically in black silk. 89 While the colour of the shop girl's uniform was subdued, black silk gave it a sophisticated air. 90 Ultimately, the shop girl's uniform was only worn in the shop and did not encroach on her identity while the servant's uniform was an important part of the servant's life, defining her appearance from dawn till dusk. In consequence, many women chose to become shop assistants because it was perceived as a more glamorous, socially enhancing and freer occupation than domestic service. British servants' rejection of their uniform, therefore, was not solely based on their personal relationship with their employers, but it was also strongly linked to the evolution of the status and appearances of women on the labour market. As female occupations were increasingly articulated outside of the private spherewith the advantages it entailed in terms of personal freedomand specialised to the same extent as men's, British servants felt the disadvantages of their occupation and its uniform more strikingly.
In France, in contrast, there were few occupations with a distinctive dress to which comparisons could be made. Some shop assistants and waitresses in large towns and cities wore uniforms and enjoyed a relatively independent lifestyle. Parisian shop assistants in department stores, for example, were referred to as 'the queens of the urban proletariat' because of their lavish clothing. 91 However, most shop assistants kept working for small family-run shops. 92 In these stores, most of the workforce was either family or servants, dressing very far away from the glamorous standards of the shop assistant in a department store and enjoying less independence. 93 These women often simply wore an apron for what was deemed a traditional female occupation in the sphere of the family economy.
It was also unlikely for French servants to compare themselves to nurses as the professionalisation of nursing was incomplete and the occupation remained dominated by nuns well into the interwar period. 94 The religious roots of nursing were apparent in the nurse uniform, which closely resembled the dress of a nun with a long white veil. 95 The slow transformation of the occupation was linked to a deep suspicion of women's waged labour in French society. As the ideal of a professional woman clashed with the Third Republic's conservative gender values, the state favoured nuns as nurses instead of female civilians, despite the government's secular crusades in other fields. 96 The few nurses who were not religious were unskilled women from working-class or peasant backgrounds and were mostly treated as domestic servants. 97 This was the case for Suzanne Ascoët, whose experience in a hospital in the 1930s seemed no different to her previous job as a maid, and for Joséphine Rochet, who left her job as a maid-of-all-work to work without a diploma or qualification in an hospital in Grenoble during the First World War. 98 In consequence, few women looked up to nursing as a modern aspirational occupation. 99 The relative absence of popular female occupations with distinctive uniforms that servants could look up to highlights the lack of specialisation of women's work in France. In a society where women's work was still significantly happening within the family economy, it was difficult for French servants to make sense of their experiences as a distinct occupational group.
Conclusion
While both groups of workers in each country were called similar names -'domestiques', 'servantes', 'domestic servants'and performed tasks inside their employers' home in exchange for a wage as well as boarding and lodging, this article has shown that what it meant to be a servant in the first half of the twentieth century was different in France and Britain. The contrast between the uniform in Britain and the simple work apron worn by domestic servants in France provides an interesting avenue through which to explore servants' identities in both countries. In Britain, domestic service was an occupation that was detached from 'productive' activities and mainly dedicated to domestic tasks. As servants cleaned their employers' house, their role was also conspicuous as their presence signalled the respectable status of the employers who could afford such help. The uniform that servants were required to wear was a key part of this conspicuous display by making them more visible and distinctive from other members of the household. As a result, this garment made many servants feel inferior and frustrated by the way it restricted the sartorial expression of their sense of self. Servants also struggled with the symbolism of their uniform beyond the home as they compared themselves to other uniformed female workers. As employment opportunities for women expanded and new service jobs offered more free time and independence to workers, servants suffered from the way their uniform symbolised a less flexible and more intrusive type of relationships with the employers. The uniform, therefore, was a source of inferiority and alienation for many British servants. In contrast, domestic service retained a connection to the productive sphere much later in France. Within small family production units, servants worked alongside their employers and other employees to keep the household running. Their appearance and occupational identity were not distinct from that of other female workers who also toiled within the home. They simply wore a rough apron to protect their clothes as they went about their day. For the servants who left the countryside to find work in towns and cities, nonetheless, clothing did play an important role in their sense of self. Instead of marking their occupational or class identity, however, these women's clothes often helped to define their status as rural migrants.
The comparison of servants' attitudes towards their dress in France and Britain helps us recast the 'servant problem' in new terms. The strength of the debate around the servant problem in Britain compared to France was not the sign that dated paternalist behaviours were more prevalent in the former. On the contrary, the revolt of British domestic servants against bad working conditions and their uniform was fostered by the modernisation of the female labour market. The uniform was a symbol of servitude and exploitation for many domestic servants, but it also was a product of the increasing specialisation of female occupations which gave the possibility for distinct groups of female workers to argue for their rights. While attempts at creating servants' trade unions from the late nineteenth century onwards were relatively unsuccessful because of the difficulty of mobilising a fragmented labour force, the uniform was nevertheless a rallying point for many women who otherwise worked in drastically different environments. 100 The recognisable black and white uniform and the distinctive cap and apron also helped make servants' complaints more easily identifiable in the public sphere. While French servants suffered similar levels of exploitation and a lack of regulation of their work, it was much more difficult to foster a debate around a 'servant crisis' when few women saw the work they did in their employers' home as a distinct occupation. French servants' integration within the family economy might have saved them from painful feelings of inferiority and alienation, but it also made it more difficult for them to argue for the improvement of their rights as workers. Because service work was invisible and malleable, these women's voices have been harder to identify and promote up to the recent times. In the 1970s historians of domestic service in France claimed that servants 'were not generators of sources' and 'did not have a voice'. 101 This deficiency makes it all the more important to show that it is possible to retrieve French servants' testimonies and explore the complexity of domestic servants' experiences beyond the elite circles of the Parisian capital. | 2019-11-22T00:36:53.563Z | 2019-10-20T00:00:00.000 | {
"year": 2019,
"sha1": "40796d893fb92f3b953b38bde40f5f3e3ef59807",
"oa_license": "CCBY",
"oa_url": "https://www.tandfonline.com/doi/pdf/10.1080/14780038.2019.1691475?needAccess=true",
"oa_status": "HYBRID",
"pdf_src": "TaylorAndFrancis",
"pdf_hash": "b83b615f483f26b0faf0ff6db7d361de50dc49c3",
"s2fieldsofstudy": [
"History",
"Sociology"
],
"extfieldsofstudy": [
"History"
]
} |
73666722 | pes2o/s2orc | v3-fos-license | Ultrastructural analysis of the shells of Anodontites trapesialis ( Lamarck ) and Anodontites elongatus ( Swaison ) ( Mollusca , Bivalvia , Etherioidea ) from the Mato Grosso Pantanal Region , Brazil 1
Based on optical and SEM microscopic observations, the projections of the outer surface of the periostracum and inner micro-structures of the shell are described and redefined. The outer surface of the periostracum is practically smooth in both species. Considering a mesoscopic view of the periostracum, A. trapesialis (Lamarck, 1819) presents regular corrugations in the form of radial sequences of arches on the disk region, isolated rays or horizontal sequences of rays on the anterior lower region. A. elongatus presents corrugations formed by series of oblique arches on the disc and oblique rays on the carina. Under SEM, micro ridges were more evident in A. elongatus, but a wide diversity of shapes and patterns of micro fringes were observed in A. trapesialis, especially in young individuals. Considering the profile of the shell layers, the periostracum is relatively thin and apparently simple in A. trapesialis and thinner in A. elongatus. The prismatic layer is thick in both species, consisting of a single series of elongated prisms and wedge-shaped prisms close to the outer surface. The nacreous layer consists of very fine lamellae without pattern or with a slight staircase-like; in A. elongatus this layer is divided by a laminar inclusion of conchiolin. The fringes are abundant and diversified in A. trapesialis, a species less resistant to desiccation due to the presence of a wide intervalvar gap. The existence of a greater density of micro fringes and spikes in juveniles may be related to the orientation of the animal in order to search for an ideal site for development or for escape from regions subject to seasonal droughts, like Pantanal.
To describe, compare and differentiate the macro-and micro characteristics of the periostracum, as well as the arrangement of the remaining inner layers forming the shells of freshwater bivalves goes beyond a strictly descriptive objective, and represents an important tool to be used in different approaches ranging from Systematics to Ecotoxicology.BØGGILD (1930) characterized the outer surface and the organization of inner structures that compose the shell and described seven morphological types, commenting on the organization of these structures in 50 families of bivalve mollusks.For Unionidae ( = Naiadidae sic.), the author described a shell consisting of three layers.The outer layer is the periostracum, a fine organic layer basically consisting of protein, which covers the entire extension of the shell.Below the periostracum is the prismatic layer, consisting of considerably elongated and juxtaposed prisms, and the nacreous layer, consisting of the deposition of horizontal lamellar plates.Using light and scanning electron microscopy, KOBAYASHI (1969KOBAYASHI ( , 1971) ) detected 11 different basic types of aggregation of the elements that form the calcareous part of the shell of bivalves.A crystallographic characterization of mollusk shells was published by CARTER (1990 a, b) who described and illustrated patterns of the organization of the periostracum and of the calcareous structures of various groups.
External periostracal projections recognized by BOTTJER & CARTER (1980) as of periostracal origin or as adventitious (formed later), were described for some Bivalvia like Arcoida, Mytiloida and Veneroida, and classified according to their morphology and function.
Studies on shell formation and processes of biomineralizatin are relatively few.SALEUDDIN & PETIT (1983) suggested a mechanism for biomineralization in unionid bivalves, summarized as follows: The periostracum is composed by three layers.The outer and middle layers are formed within the periostracal groove and the inner layer by the mantle epithelium.The prismatic shell layer is formed inside the middle layer of the periostracum through a process of vacuolization and antrum formation.By a simmilar process, the nacreous layer is formed from the inner peristracum.More recently, CHECA (2000) proposed a different model for the origin of the periostracum and mode of formation of shell layers in Unionidae, that consists: 1) the periostracum is formed by two layers, an outer, thin layer, which is secreted within the periostracal groove, and an inner, thicker layer, which is secreted by the epithelium of the outer mantle fold; 2) calcification initiates within the inner periostracum as fibrous spherulites, which protrude from the periostracum, coalesce and compete mutually, transforming into composite prisms; 3) nacreous tablets begin to nucleate directly on the ends of prism fibers by epitaxy.Transition from composite prisms to nacre is probably induced by merely crystallographic processes.
Little information is available about South American naiads (Hyriidae and Mycetopodidae).MARSHALL (1925) started the characterization of the periostracum surface for the major Unionoida families and concluded that radial "micro sculptures" seem to be typical of Mutelidae and Mycetopodidae, both belonging to the superfamily Etherioidea, according to KABAT (1997).BONETTO & EZCURRA (1965) defined three basic patterns of striation occurring in the periostracum: folds, macroscopic "sculptures" and microscopic "sculptures".On the basis of this systematization, BONETTO (1966) described and illustrated the pattern of the periostracum folds of Monocondylaeinae and observed that the genus Fossula Lea, 1870 had characteristics similar to those of Anodontites Bruguière, 1792.According to BONETTO (1967) both genera have rare and sparse periostracum striations and a well-developed prismatic layer.According to a phylogenetic study conducted by BOGAN & HOEH (2000) and HOEH et al. (2001), the genera Monocondylaea Orbigny, 1835 and Anodontites represent related groups in the family Mycetopodidae.
The aim of the present study is to describe and to compare the form and pattern of outer projections of the periostracum and the arrangement of the inner shell structures of Anodontites trapesialis (Lamarck, 1819) and Anodontites elongatus (Swainson, 1823), to complement taxomic, biological and ecological studies conducted on both little known Mycetopodidae species, from the Mato Grosso Pantanal (Wetland) region.
Specimens of Anodontites trapesialis (Figs 1 and 2) and
Anodontites elongatus (Figs 13 and 14) were collected from the lake Baía do Poço (Well Bay) in the municipality of Santo Antônio do Leverger, State of Mato Grosso, Brazil, from May 1998 to April 1999.The shells selected for the study were washed and placed in plastic bags without being submitted to any fixation process.Twenty individuals for each species representing different length classes were analyzed in order to recognize and describe a structural pattern.For scanning electron microscopy, shells from juvenile and adult individuals were washed with distilled water for 24 hours and dried at room temperature.The material was fragmented with a hammer and appropriate portions were selected under a stereoscopic microscope.Fragments from the carina region were used for external analysis of the periostracum, and fragments from the central region of the shell were used for analysis of inner structures.The surface to be fixed was sanded and the fragments were submitted to two consecutive baths under ultrasound for approximately 30 seconds and then fixed on stubs with silver glue or with doublefaced carbon adhesive tape.The pieces were sputtered with gold and images were obtained with a scanning electron microscope, model Phillips XL30 and processed in the Microscopy and Microanalysis Center (CEMM) of PUCRS.The used scale for magnifications ranged from 2 to 500 µm.Biometric analysis was performed using the Image Tool software, version 2.0, developed by UTHSCSA (WILCOX et al. 1997).The images thus obtained were used for measurements of periostracum, prismatic and nacreous layers thickness, the number and density of folds, ridges and the distances between them as well as the density of Revista Brasileira de Zoologia 22 (3): 724-734, setembro 2005 the projections on the periostracum (Tab.II).Mean and standard deviation were calculated for each unit sample and used as a parameter for species comparison.
The projections and structures of the periostracum were recognized following BOTTJER & CARTER (1980) with adaptations for the specific features observed on the periostracum of the studied species.The terminology "macro" and "microsculture" used by MARSHALL (1925) and BONETTO & EZCURRA (1965) for impressions and also for projections on shell surface brought some difficulties concerning periostracum characterization.The terminology sculpture is now suggested only for strongly impressed radial beak sculpture present in most Hyriidae.In view of those divergences observed in the literature a redefinition was necessary and it is summarized in table I.
The calcareous structures were classified according to the criteria proposed by CARTER (1990 a, b).The identification of the species was based mainly on morphological characters following HEBLING (1976) and SIMONE (1994) for A. trapesialis, and SIMONE (1997) for A. elongatus.The species were considered only at the species level according to the taxonomical revisions of SIMONE (1994SIMONE ( , 1997)).Voucher specimens were deposited in the mollusks collections of Museu de Ciências e Tecnologia da Pon-tifícia Universidade Católica do Rio Grande do Sul, Porto Alegre, Rio Grande do Sul, under the numbers MCP 8524, 8525 and 8526, and in the Laboratório de Ecologia Animal -NEPA -from Instituto de Biociências of the Universidade Federal de Mato Grosso, Cuiabá, Mato Grosso, Brazil.
RESULTS
The shells of Anodontites trapesialis and Anodontites elongatus consist of three different layers: the periostracum, the prismatic layer, and the nacreous layer.
Periostracum -outer surface
Anodontites trapesialis (Figs 1 and 2).With a thickness of 20.1 ± 3.9 m, the periostracum is olive green in colour, with straw yellow shades in smaller and younger individuals.Larger and older individuals become darker, with a greenish brown colour.The surface presents folds, corrugations, micro ridges and micro fringes (Fig. 3), as described below.
Folds (Macroscopic).The periostracum is practically smooth on the disk region (Figs 3 and 4), with the rare and sparse occurrence of concentric folds (at most 15 per centimetre in radial direction), almost always low and/or worn out.In the posterior slope the folds are closer (60 per centimetre), higher and overlapping, with a minimum distance of 0.35 mm between them.In juvenile individuals the folds are closer and more conspicuous (Fig. 4), and may be visualized also in the anterior region at an approximate density of 30 folds per centimetre.Corrugations (Mesoscopic).There are radial sequences of short slightly convex arches (Fig. 3) arranged in rows containing about 70 arches per centimetre.Each individual may present three to seven rows always arranged on the disk region.They are clearly visible not only on the periostracum but also demarcated on the underlying calcareous layer.Bas-relief rays (±10 per mm) occur separately, between the rows of arches or forming sequences perpendicular to the folds, always being more frequent in the anteroventral portion.They are visible only under a magnifying lens and confer a dull aspect to this region of the shell.
Microscopic structures.Ridges with a radial, oblique or horizontal orientation in relation to the folds were occasionally observed (Fig. 3).They do not follow the pattern of the vertical parallel micro ridges observed in most Mycetopodidae, but form more commonly not continuous rows of horizontal ridges or fishbone like structures.
Scanning electron microscopy permitted the observation of very coarse periostracal projections forming rows of fringes (Figs 4 to 8) covering the outer surface of the shell.These minute projections take on different shapes and orientations according to where they are located.They may be simple and cylindrical like fingers (Fig. 5), with the aspect of an aculeus or forming serrated bands or rows of spikes like small triangular flags (Fig. 6), or they form simple and smooth low bands connecting separated spikes like flags with triangular extremities or fissured extremities, taking on a digitiform aspect (Fig. 7).The number of projections ranged from 1700 to 6000 units/mm², with the highest incidence being always associated with the anterior region of the shell of juveniles, which always presented values higher than 5000 projections/mm².In adult individuals, the projections were sparse and worn out or often partially covered by encrustations not secreted by the mollusc.In this case, the distal extremities of the projections are only visible in form of little spikes (Fig. 8).
Anodontites elongatus (Figs 13 and 14).This species presents a very fine periostracum with a thickness of 7.3 ± 1.3 m.All individuals are dark brown in colour, which can even be black in some larger specimens.Folds, corrugations, micro ridges and projections occur in this species.
Folds (Macroscopic).The periostracum is smooth on the disk region, with only a few folds being observed in the distal portion (26 per centimetre).However, in the region of the posterior slope a high density of high juxtaposed folds (Fig. 15) with an exfoliated aspects was always observed, with values exceeding 50 per centimetre.Lower folds, almost always worn out, occurred in the anterior region, with a density of 40 to 60 folds per centimetre.In juveniles, there was a higher density of Table I.Definition of folds, corrugations and projections of the periostracal surface of Anodontites trapesialis and A. elongatus observed respectively by unaided yes, under stereomicroscope and by scanning electron microscopy (SEM).
Periostracum projections Definitions
Folds Macroscopically identifiable periostracal commarginal folds.Frequently described as "concentrically" organized in relation to the umbos, but not recommended (COX et al. 1969).They may present an exfoliated aspect as if they were superposed scales, or may simply appear as longitudinal segmentations of the periostracum (but not interrupted).
Corrugations (Fig. 3, 13) Mesoscopic* observable periostracal corrugations oriented in radial bands perpendicular to the umbos or parallel to the folds.They may appear in the forms of short bands or concentric arches forming stair-like radial rows, or as oblique arches with one end touching a neighbour ray, forming fishbone-like figures.Those regular corrugations reflect a very soft topography of the underlying shell layer (BOTTJER & CARTIER 1980).They are normally displaced around the disc but little far from the beaks and very different from the macroscopic radial beak sculpture of the Hyriidae that, if present, is strongly impressed on the shell surface.
Micro ridges (Fig. 3, 15,16) Regular, superficial, periostracal structures that are expressed only on the outer surface of the periostracum (BOTTJER & CARTIER 1980).Radial, oblique or parallel and very close microstriae mainly present on the carina region.They are visible only at high magnification under the stereoscopic microscope or the SEM.They are usually present at high density, possibly conferring a dull aspect to the surface.
Micro fringes or spikes 16) Minute projections observable only by SEM, preferentially on the entire region that covers the carinae.They are organized longitudinally or obliquely in relation to the axis of the micro ridges or may be dispersed throughout the periostracum surface.These micro-ornaments take on varied shapes that may be simple, like aculei, with a triangular or cylindrical base and acuminate or rounded extremities, or compound forming fringes with a digitiform appearance or like small flags.
* Mesoscopic observations were defined as those located at the limit of visualization between the unaided eye and a magnifying lens.
Revista Brasileira de Zoologia 22 (3): 724-734, setembro 2005 folds with a well-preserved exfoliated appearance.Corrugations (Mesoscopic).Peculiar radial bands occur on the disk region associated with the umbo (Fig. 13).They are formed by two parallel series of oblique arches, forming figures resembling Christmas trees.Each series of oblique arches consist of about 100 corrugations per centimetre.When they occur, the rays are always orientated obliquely to the folds both in the anterior and the posterior regions of the shell.The highest density of these radial bands is associated with the carina, where rays without a defined pattern give an undulated aspect to the periostracum.
Micro ridges are present on the entire shell surface (Fig. 15).Always orientated perpendicular to the folds, they are visible only at high magnification.There are at least 90 microstriae mm -1 and the distance between them does not exceed 20 µm.
The projections were only represented by small isolated spikes in this species (Fig. 16), with a constant morphological pattern with a broad base and a tapering extremity and resembling aculei.Their density was always about 2,600 spikes/mm 2 in A. elongatus.
A greater occurrence of spikes, as well as a better integrity of folds and sculptures, was always more evident in juveniles than in adults, as also observed in A. trapesialis.
Internal structures. Prismatic and nacreous layers
The prismatic and nacreous layers were distinguishable below the periostracum in the shells of A. trapesialis and A. elongatus.
It was possible to characterize for both species a uniform prismatic layer consisting of elongated prisms in a parallel arrangement perpendicular to the external surface.The prismatic layer of the two species was classified as regular simple prismatic layer according to the criteria of CARTER (1990 a, b) and fibrous aragonitic composite by CHECA (2000), with the presence of almost always pentagonal and sometimes hexagonal prisms (Figs 10 and 18).Most of the prismatic units that form this layer are elongated and continuous, extending from the periostracum to the beginning of the nacreous layer.However, smaller wedge-like prisms were also observed, always starting from the outer surface immediately below the periostracum.They are typical competition features according to UBUKATA (1994).The surface of the prism is covered with a reticulate lining of a spongy aspect which is visualized only at approximations of more than 2 µm (Figs 11 and 19).
In A. trapesialis the prismatic layer is apparently more fragile, formed of smaller prisms measuring around of 450 µm in length and 23 µm in width (Fig. 9).Few wedge-like prisms were observed in the images obtained for these species, and the exist- elongatus have a larger prismatic layer (Tab.II), with prisms reaching as much as 580 µm in length and 40 µm in width (Fig. 17).
The presence of wedge-like prismatic units was constant (Fig. 18), with prisms of different length reaching as much as two thirds of the length of the main prisms.
The nacreous layer consists of minute superimposed plates (Fig. 12), most of the time without a defined organization.Some shell regions of both species may show a symmetrical oblique arrangement.These characteristics are of the sheet nacreous structure type, among those described by CARTER (1990 a, b), where these lamellae may be organized as the steps of a staircase or may have no defined pattern.In A. elongatus the thickness of the nacreous layer was, on average, 650 µm, corresponding to almost three times the thickness of the shells of A. trapesialis, whose minimum thickness is about 220 µm (Fig. 21).The size and thickness of each lamellar plate varied between species (Tab.II).An intermediate membrane, the conchiolin layer, very thin, of an irregular pattern, similar to the texture of the periostracum, and running longitudinally along the nacreous layer (Fig. 22), was observed only in A. elongatus.
DISCUSSION
During analysis of the studied species, special attention was devoted to the periostracum, since this structure presented more divergences between species.Considered to vary little among unionids, the thickness of the periostracum is 15 to 50 µm in most of the species of this group, classified by TEVESZ & CARTER (1980) as 'intermediate thickness'.Varying around 20.1 µm, the periostracum of A. trapesialis perfectly fits this classification.However, the mean thickness of the periostracum of A. elongatus was about 7.3 µm, a fact considered to be rare within the superfamily by the cited authors, who believe that an excessively thin periostracum is inefficient in protecting the shell from dissolution.
Basically consisting of fibrous proteins and quinones, the periostracum also contains some carbohydrates and lipids, which together form the periostracal units.The biochemical synthesis giving origin to these units is called sclerosis and occurs in the periostracal groove, between the two most external mantle margin folds and located in the outer fold of the mantle.The newly formed periostracum is distally compressed and expelled in the form of a multilamellar entity (SALEUDDIN & PETIT 1983).It is prob- ably at this time that the sculptures are molded, thus defining the periostracum structure typical of each species.
In both species here studied, the periostracum was practically smooth in a superficial view.The corrugations that could be observed with the unaided eye or under the stereoscopic microscope, were sparse and more frequent in A. elongatus than in A. trapesialis.This situation was also observed by BONETTO & EZCURRA (1965), who described a periostracum with barely developed "macro sculptures" (actually considered as corruga-tions) arranged in a linear series of arches.BONETTO (1966) also emphasized the similarities existing in the periostracum of the genera Anodontites and some Monocondylaeinae like Fossula, which also presents rare and sparse series of corrugations.
The micro ridges of A. elongatus were fine, compactly arangend and clearly visible under SEM, always radially organized and perpendicular to the folds, like the definition of "fine radiating threads" given by MARSHALL (1925) and BONETTO & EZCURRA (1965) to Mycetopodidae micro ridges.Nevertheless those last cited authors stated for A. elongatus as quite poor in terms of "micro sculptures", a fact contradicting our observations.BONETTO & EZCURRA (1965) commented on the fact that "micro sculptures" were practically imperceptible in A. trapesialis, as also were the folds formed by the periostracum, contradicting also the results of the present researches.
SEM permitted the identification of another structural pattern of minute projections that had not been previously described for this species.Denoted as fringes and or spikes, the minute projections present on the periostracum respectively of A. trapesialis and A. elongatus, occurred frequently and showed structural diversity, with intra-and interspecific variations.
Anodontites trapesialis is more vulnerable to environmental alterations.This species has a considerable intervalvar gap through which the foot extend and never succeeds in fully closing its valves.This fact prevents this species from isolating itself in the presence of adverse environmental situations.Considering that the specimens here studied originate from the Mato Grosso Pantanal region, an ecosystem subjected to a rigorous cycle of drought and flooding characteristic of wetlands (JUNK 2000), the presence of fringes suggests that the function of these structures may be associated with an increased sensitivity to water flow, especially among juveniles.These present a greater density and a distinct morphological pattern of spikes when compared to adult individuals.Thus, the juveniles of A. trapesialis may possess a high specialised sensor to water movements.According to BOTTJER & CARTER (1980) the periostracal projections can be functional for: "achieve stability in the substratum, to minimizing settlement by encrusting epizoans, reducing the hazard of shell penetration by endolithic organisms, camouflaging and protecting the shell margins from predators and extending the range of tactile sensory perceptions away from mantle edge".The authors (op.cit.) considered the latter advantage "as important only in the Mytiloidea, which apparently lack visual perception in the posterior mantle, and which secrete especially long adventitious hairs in certain species".The observed periostracal fringes and spikes of juveniles Anodontites are in fact not sensitive structures and not long enough such as the hairs of some mytilids, but rather projections that should promote some kind of vibration do to water current.These signs could be interpreted by the sensitive system of the animal as an order to move away.During sampling in Pantanal Region it was observed that juveniles are rarely found together with adults.The juveniles live mainly on margins and places of low water that are more subjected to the risk of desiccation.
No variation in the structural profile of the periostracum was observed in the studied species, both presenting a simple membrane.However, WAITE (1977) demonstrated a difference in the structure of the periostracum, which appears to consist of two different layers: an upper amorphous layer and a weakly fibrous lower layer.In agreement with this, biochemical studies conducted by SALEUDDIN & PETIT (1983) suggest that the newly formed periostracum goes through a remodelling stage through the deposition and polymerization of glycocalcite, processes that are repeated several times.
The importance of the periostracum in the calcification process is shown by the active performance of this membrane (SALEUDDIN & PETIT 1983), which is responsible for: a) nucleation, b) crystal orientation, c) crystal size, and d) crystal type in the formation of the remaining layers forming the shell (WILBUR & SAULEUDDIN 1983).CHECA (2000) recognized only two layers by SEM observations on unionid shells.The external part is composed internally of micron-sized sublayers.
There are several definitions for the different configurations of the prismatic layer among bivalves.BØGGILD (1930) categorized them as normal, complex, and compound prismatic structures.KOBAYASHI (1969KOBAYASHI ( , 1971) defined them as prismatic structures and fibrous prismatic structures.In fact and according to the classification of UBUKATA (1994) A. trapesialis and A. elongatus present composite aragonitic prisms that are encapsulated by a surrounding organic matrix (KOBAYASHI 1969(KOBAYASHI , 1971)).WILBUR & SALEUDDIN (1983) explained that this organic matrix, in addition to the attributes mentioned earlier, also has the function of crystal to crystal adherence, constructing a unified structure.
The structure and arrangement of the prisms are specific for the family and may vary in some details among genera and species (TOLSTIKOVA 1974).This fact was not observed in A. trapesialis and A. elongatus, with the only significant difference concerning prism length, which was approximately on average, 480 µm for A. trapesialis and 560 µm for A. elongatus.However, the length of these may vary within the same species according to their location on the shell and shell size.In the anterior part the prisms are shorter and in the posterior region close to the margin of the shell, they are longer, as also described by TOLSTIKOVA (1974).
Considering the nacreous layers there was a clear difference in the thickness between the two species.Anodontites elongatus (650 µm) has a nacreous layer almost three times thicker than A. trapesialis (240 µm), explaining the resistance of the shell of A. elongatus to external agents, offering better protection to the organism.BØGGILD (1930) observed that the lamellar or tabular units that compose the nacreous layer are extremely thin, with values always lower than 0.001 mm.SEM images permitted the exact determination of the height (or thickness) of these units, which measure approximately 0.20-0.30µm, with more than 2000 superimposed units being necessary to form a 650 µm layer like that of A. elongatus.SALEUDDIN & PETIT (1983) described that the formation of this layer is also associated with the periostracum, in the time when products from cell glycocalices adhere to the newly formed organic film, coming from distal cells of the mantle fold.The deposition of this material in randomly arranged plates represents the beginning of the nacreous layer.CHECA (2000) describes that nacreous tablets nucleate on previously formed crystals of prismatic layer and the epitaxial growth explain the common orientation of nacreous tablets.Only A. elongatus presented seg-mentation in the nacreous layer by a continuous horizontal layer very similar to the periostracum.KAT (1983) observed within the shell of Margaritiferidae and some Unionidae one or more types of layers named as conchiolin layers, which according to him, closely resembles the periostracum in ultrastructure.These inclusions of organic matter were first described by BØGGILD (1930) so that, due to the intercalation of the organic matrix, the nacreous part is dissolved more slowly by acids.TEVESZ & CARTER (1980) and KAT (1985) described this event and suggested that mollusks with an excessively thin periostracum are inefficient in terms of protection against the dissolution of the calcareous structures of the shell.On the basis of these statements, we may suggest that the presence of these organic sublayers may act by strengthening and protecting the nacreous structure since the periostracum of A. elongatus is much thinner (7.0 µm) than that of A. trapesialis (20 µm).TEVESZ & CARTER (1980) also stated that the intermediate layers of conchiolin are frequent in Unionidae species, especially those living in environments subjected to acidification, or may be a response of the organisms to exposure to contaminants that penetrate between the shell and the mantle.This question was the starting point for IMLAY (1982) when he suggested the use of freshwater bivalve shells to monitor not only the presence of heavy metals, but also the levels of turbidity, low oxygen concentrations, temperature and pollutants in a general manner.
always short, not exceeding the upper third of the prismatic layer.With a more robust structure, the shells of A.
Table II
. Mean (± standard deviation) values concerning the structures that compose the layers of the shell of Anodontites trapesialis and A. elongatus from the Baia do Poço (Well Bay), municipality of Santo Antônio do Leverger, Mato Grosso State.Analyses were carried out using scanning electron microscope (SEM). | 2018-12-31T21:45:41.683Z | 2005-09-01T00:00:00.000 | {
"year": 2005,
"sha1": "dc588f6417874699437aeacce8036fc001f460c0",
"oa_license": "CCBYNC",
"oa_url": "https://doi.org/10.1590/s0101-81752005000300033",
"oa_status": "GOLD",
"pdf_src": "ScienceParseMerged",
"pdf_hash": "dc588f6417874699437aeacce8036fc001f460c0",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": [
"Biology"
]
} |
15022955 | pes2o/s2orc | v3-fos-license | Angiopoietin-Like Protein 3 Promotes Preservation of Stemness during Ex Vivo Expansion of Murine Hematopoietic Stem Cells
Allogeneic hematopoietic stem cell (HSC) transplantations from umbilical cord blood or autologous HSCs for gene therapy purposes are hampered by limited number of stem cells. To test the ability to expand HSCs in vitro prior to transplantation, two growth factor cocktails containing stem cell factor, thrombopoietin, fms-related tyrosine kinase-3 ligand (STF) or stem cell factor, thrombopoietin, insulin-like growth factor-2, fibroblast growth factor-1 (STIF) either with or without the addition of angiopoietin-like protein-3 (Angptl3) were used. Culturing HSCs in STF and STIF media for 7 days expanded long-term repopulating stem cells content in vivo by ∼6-fold and ∼10-fold compared to freshly isolated stem cells. Addition of Angptl3 resulted in increased expansion of these populations by ∼17-fold and ∼32-fold, respectively, and was further supported by enforced expression of Angptl3 in HSCs through lentiviral transduction that also promoted HSC expansion. As expansion of highly purified lineage-negative, Sca-1+, c-Kit+ HSCs was less efficient than less pure lineage-negative HSCs, Angptl3 may have a direct effect on HCS but also an indirect effect on accessory cells that support HSC expansion. No evidence for leukemia or toxicity was found during long-term follow up of mice transplanted with ex vivo expanded HSCs or manipulated HSC populations that expressed Angptl3. We conclude that the cytokine combinations used in this study to expand HSCs ex vivo enhances the engraftment in vivo. This has important implications for allogeneic umbilical cord-blood derived HSC transplantations and autologous HSC applications including gene therapy.
Introduction
Hematopoietic stem cells (HSCs) have the ability to self-renew and to give rise to cells of all blood lineages. This makes HSCs a valuable source for treatment of patients with genetic blood disorders through cell-or gene-based therapies [1][2][3][4]. These therapies are restricted by the limited availability of suitable, human leukocyte antigen (HLA)-matched donors. Additionally, if autologous cells for genetic modification are concerned, limited numbers of cells can be retrieved per patient. Umbilical cord blood (UBC) potentially provides an alternative and abundant source of donor HSCs, if the number of HSCs could be increased in vitro [5,6]. Optimization of in vitro expansion protocols would therefore facilitate successful transplantations using UCB-derived HSCs or genetically-modified autologous HSCs [7,8].
Early attempts to expand HSC in vitro resulted in a preferential expansion of committed progenitor cells without preserving stemness, resulting in defective long term hematopoiesis [9]. However, the knowledge on hematopoietic stem cell expansion has increased, and new methods for promoting expansion of stem cells while retaining stemness have been developed. Ectopic expression of the transcription factors, such as homeobox B4 (HoxB4) or apoptotic regulators such as Bcl-2 have been investigated and can result in robust HSC expansion [10,11]. However, the long term consequences of constitutive activation of anti-apoptotic pathways triggered by specific factors such as Bcl-2 or HoxB4 is not yet fully investigated. Another obstacle is the delivery of these proteins, which may require vector-based vehicles, which should be efficient and not genotoxic [12][13][14][15]. To circumvent these problems, it would therefore be preferable to develop methodology to expand HSC without the introduction of foreign DNA sequences.
Several growth factors have been identified over the years that enhance the self-renewal capacity of mouse HSCs, including ligands for various pathways such as Notch1 [16], stem cell factor (SCF) [17], thrombopoietin (TPO) [18,19], fms-like tyrosine kinase-ligand (Flt3-L) [20], fibroblast growth factor (FGF-1) [21,22] or WNT-pathway factors like Wnt3a [23]. The Lodish group identified a fetal liver stromal cell population that produces high levels of insulin growth factor-2 (IGF-2) and angiopoietin-like proteins in addition to SCF and delta-like NOTCH1 ligands. These factors were shown to support HSC expansion in vivo [24][25][26]. The combination of IGF-2, angiopoietin-like 2 (Angptl2), and angiopoietin-like 3 (Angptl3) growth factors also support HSC expansion in vitro [25,26]. Various studies support a pivotal role for Angptl3 in regulating HSC self-renewal capacity [26][27][28][29]. This was confirmed by results from the Angptl3 knock-out mouse model that demonstrate reduced numbers of quiescent HSCs as well as reduced repopulation capacity in transplantation experiments [27]. Angptl3 is expressed by endothelial and other stromal cells in the bone marrow and binds as an extrinsic factor to receptors on HSCs [27]. At present, the receptor for Angptl3 is not clear as the immune-inhibitory receptor human leukocyte immunoglobulin-like receptor B2 (LILRB2) and the mouse orthologue paired immunoglobulin-like receptor (PIRB) have been identified as receptors for several angiopoietin-like proteins (Angptls) including Angptl2, 25, and 27, but this is unclear for Angptl3 [28]. Binding of Angptls to its receptor results in reduced expression of Ikaros and activate self-renewal capacity [27]. Overexpression of Ikaros in HSCs was shown to diminish repopulation capacity [27]. The combination of saturated levels of SCF, TPO, IGF2, FGF1 and Angptl3 has been proven as a suitable cocktail that promotes expansion of long-term repopulating HSCs (LT-HSCs) numbers up to ,30-fold [26].
In the present study, we tested the preservation and expansion of long-and short-term HSCs in vitro in serum-free culture conditions in the presence of SCF, TPO, IGF2 and FGF-1 (STIF) [26] or SCF, TPO and FLT3-L (STF) [9]. Long-term repopulating capacity was investigated for cultured HSCs under various conditions followed by transplantation into sub lethally irradiated mice. We investigated a potential additive effect of mAngptl3, that may exert a direct effect on HSCs. We also tested the potential leukemogenic or toxic effects of ectopic expression of Angptl3 in transplanted mice.
Mice
Female a-thalassemic BALB/c mice between 8 to 12 weeks of age were used as bone marrow (BM) recipients, and healthy male littermates were used as donors for HSCs. Mice were bred and housed under specific pathogen free (SPF) conditions at the Experimental Animal Facility of Erasmus Medical Center (Rotterdam, the Netherlands). All experiments have been approved by the local ethical committee for animal experiments and are in accordance with national legislation.
Stem cell isolation
Lineage negative (Lin 2 ) cells were purified from BM using the BD IMag Mouse Hematopoietic Progenitor Cell Enrichment Set (BD Biosciences, Breda, The Netherlands) according to the manufacturer's instructions. HSC were further enriched from the Lin 2 cell population by sorting Sca-1 + /c-kit + (LSK) cell populations using a BD FACS Aria flow cytometry (BD Biosciences). For this, Lin 2 cells were incubated with c-kitallophycocyanin (APC; BD Biosciences) and Sca-1-R-phycoerythrin (PE; BD Biosciences), and washed once with Hank's solution supplemented with HEPES (300 mOsm) prior to sorting.
Production of lentiviral vectors Lentiviral (LV) particle production was done by transfecting the LV-Angptl3-GFP vector in combination with pMDL-g/pRRE, pMD2-VSVg and pRSV-Rev helper plasmids into HEK 293T cells using standard calcium phosphate as previously described [31,32]. Lentiviral particles were concentrated through ultracentrifugation for 2 hours at 20,000 rpm and collected at 4uC. The multiplicity of infection (MOI) of LV-Angptl3-GFP or LV-GFP was determined on HeLa cells by serial dilution, and the percentage of eGFP-positive cells was estimated by flow cytometry. Angptl3 protein expression was detected by a standard Westernblot procedure using the mouse monoclonal anti-ANGPTL3 (1D10) antibody (Novus Biologicals, Cambridge, United Kingdom).
Lentiviral hematopoietic stem cell transduction and sorting
Lin 2 donor cells were transduced overnight with LV-Angptl3-GFP or the LV-GFP at a cell density of 106 cells/ml using a MOI of 10. During this transduction procedure, cells were maintained in serum-free DMEM medium that was supplemented with various growth factors including murine SCF (100 ng/ml, R&D), murine TPO (20 ng/ml, R&D) and murine IGF-2 (20 ng/ml, R&D). The following day, cells were diluted to 5610 4 cell/ml and cultured for another 24 hrs. Subsequently, Lin 2 GFP + cells were flow-sorted with a purity of .90%. Sorted cells were used in experiments directly or following preculturing in STIF media. All of the cells were incubated at 10% CO2 levels and 37uC.
In vitro clonogenic progenitor assays
Frequencies of HSCs and progenitor cells were estimated using semi-solid colony assays. For this, freshly sorted or cultured (transduced) Lin 2 (2610 3 ) or LSK (0.2610 3 ) cells were plated in 35 mm culture dishes (BD BioCoat Collagen IV, tissue-culture treated polystyrene) that contained 1 ml of enriched DMEM culture medium that was supplemented with 0.8% (wt/vol) methylcellulose (Methocel A4M Premium Grade, Dow Chemical, Barendrecht, The Netherlands) as described [33,35]. All primary data is shown in table S1 and S2.
For colony-forming unit granulocyte-macrophage (CFU-GM) differentiation assays, Lin 2 /LSK cells were cultured in methylcellulose-enriched DMEM medium that was further supplemented with 10 ng/ml mouse interleukin-3 (mIL-3), 100 ng/ml m-SCF and 20 ng/ml granulocyte macrophage colony-stimulating factor (GM-CSF). For burst-forming erythroid unit (BFU-E) assays, Lin 2 /LSK cells were incubated in methylcellulose-enriched DMEM medium that was supplemented with 100 ng/ml m-SCF and 4 U/ml human erythropoietin (H-EPO, Behringwerke, Marburg, Germany). Cells were maintained for 14 days prior to microscopic analysis and the total numbers of colonies were counted. Each experiment was carried out in duplicate.
Colony forming unit spleen (CFU-S)
Lin 2 or LSK donor cells were transplanted into lethally irradiated (8 Gy) BALB/c female recipient mice (n = 6-10 per group). A total of 1,000 or 3,000 of freshly isolated Lin 2 cells were transplanted or equivalents of 100 or 1,000 cells that were cultured for 4 to 7 days in STF, STFA3, STIF or STIFA3 media. For LSK cell transplantation, a total of 30 or 100 freshly isolated LSK cells were transplanted directly or following culture for 7-days in STF, STFA3, STIF or STIFA3 media. For LSK cell transplantations, 1000 irradiated non-selected BM cells (50 Gy) were co-transplanted to improve homing. No splenic colonies were observed in control mice that were transplanted with 1000, 50 Gy-irradiated BM cells only. All primary data is shown in table S1 and S2.
In addition, Lin 2 cells were transduced with LV-GFP or LV-Angptl3-GFP. The GFP + population was sorted 2 days after transduction, and 1000 Lin 2 GFP + cells were intravenously transplanted into lethally irradiated (8 Gy) BALB/c recipients (n = 7-10 mice per group). Twelve days after transplantation, mice were sacrificed and the spleens were incubated in fixation buffer (70% ethanol supplemented with 5% acetic acid and 2% formalin). The number of spleen colonies (CFU-S) was counted.
Long-term repopulation ability assay (LTRA)
Lin 2 or LSK male donor cells were transplanted into sublethally irradiated (6 Gy) female a-thalassemia mice. Blood was collected monthly for six months following transplantation, and blood cell counts were measured with a Vet Animal Blood Counter hematology analyzer (Scil Animal Care Company GmbH). Red blood cell (RBC) chimerism were measured following preparation of mouse peripheral blood in 1 ml 0.6% NaCl buffer, and then populations of microcytic thalassemia RBCs and healthy RBCs were determined by flow cytometry (FACS Calibur, BD Biosciences). Percentage of chimerism was calculated using the following formula: [20.6+SQUART((0.62-460.0026(10.43-donor cells %)))/0.004] [36].
In another experiment, donor Lin 2 cells were transduced with control LV-GFP or LV-Angptl3-GFP. GFP + cells were sorted 2 days after transduction as described above. Lin 2 , GFP + cells (10,000 cells) were transplanted into sub lethally irradiated (6 Gy) female a-thalassemia recipient mice. Blood was collected at one, four, six and nine months following transplantation to determine the percentages of GFP + peripheral blood cells. Similarly, percentages of GFP + white blood cell types in BM or spleens were measured 9 months after transplantation, using antibodies against Sca-1, c-Kit, CD4, CD8, CD19, and CD11b (Miltenyi Biotec, BD Biosciences).
Y-chromosome Q-PCR
Y-chromosome Q-PCR was performed to detect the percentage of donor leukocyte chimerism in recipient mice following transplantation. DNA was extracted from BM using a DNeasy Blood & Tissue Kits (Qiagen, Germany). Specific primers for the Sry locus in the Y-chromosome were designed using Beacon software (
Statistical analysis
Data are expressed as the mean 6 standard deviation (SD). Statistical significance between nominal data point comparisons was determined using the Mann-Whitney-U test. Standard deviations of colony counts were calculated on the assumption that crude colony counts show a normal Poisson distribution.
Results
The effect of Angptl3 on ex vivo expansion and differentiation capacity of murine hematopoietic stem cells The expansion of murine hematopoietic stem cell (HSC) was tested using various media conditions; Lin 2 or purified LSK cells were cultured ex vivo in media supplemented with STF, STFA3, STIF or STIFA3 growth factor cocktails (see materials and methods). Culturing LSK cells for 7 days in STF-or STIF-media led to ,35-fold expansion in total cell number ( Figure 1A). The LSK phenotype was best preserved in STIF media (at a level of 4563%) compared to STF media (at 3064%) ( Figure 1B). Addition of mAngptl3 did not result in significant increases in the preservation of LSK phenotype during expansion. The expansion rate of Lin 2 cells also did not differ between STF or STIF media ( Figure S1A). Again, supplementing the media with Angptl3 did not result in a significant increase in total cell numbers. Culturing Lin 2 cells in STFA3 media for ten days resulted in a near 60-fold expansion of total cell numbers.
We next tested the differentiation capacity of in vitro expanded hematopoietic progenitor cells by carrying out various colonyforming unit assays on Lin 2 cells (Figure S1B-C) or sorted LSK cells cultured in STF, STFA3, STIF or STIFA3 media (Figures 1C and 1D). BFU-Es colony forming units were boosted 1161 fold for LSK cells cultured in STF and STIF media for 7 days compared to non-cultured LSK cells ( Figure 1C). Addition of Angptl3 significantly increased the total number of colony forming units 1661 and 1561 fold for STFA3 and STIFA3 media, respectively. For granulocyte-macrophage progenitor cells, culturing in STF or STIF media for 7 days expanded the number of CFU-GM colonies by 9-fold and 7-fold, respectively. Again, culturing with Angptl3-supplemented media further increased the number of CFU-GM colonies to 1362 and 1461 fold using STFA3 and STIFA3 media, respectively ( Figure 1D). Similar results were obtained from the number of BFU-E and CFU-GM colony forming units in the Lin 2 subset (Figures S1B-C).The colony forming unit spleen assay (CFU-S) was used to assess the effect of Angptl3 on short-term HSC (ST-HSC) in Lin 2 and LSK cells ( Figure 1E, Figure S1D). LSK cells pretreated in STF or STIF media showed a 2862 and 3062 fold increase in splenic colonies, respectively. Culturing LSK cells in STFA3 or STIFA3 media boosted CFU-S colony formation to 3864 and 3763-fold relative to control mice. In case of STFA3 the increase was significantly higher relative to STF.
Effect of Angptl3 on long-term HSCs expansion from LSK cells
We next investigated whether culturing of HSCs in the presence of Angptl3 would potentiate long-term hematopoiesis in vivo. Long-term repopulating ability (LTRA) assays were performed by transplanting sorted male Lin 2 (3000, 1000 or 300 cells) or LSK cells (200 cells) into sub-lethally irradiated female a-thalassemia mice with or without prior culturing in STF, STFA3, STIF or STIFA3 media for 7 days (Figure 2A, Figure S2). Near full donor engraftment at 7 months following transplantation of LSK cells was identified in the BM and PB compartments of mice regardless of prior incubation conditions (Figure 2A). One million BM cells from these primary transplanted mice were then retransplanted into secondary recipients, and these mice remained healthy for over 6 months without any symptoms of disease. The percentage of erythrocyte chimerism in the peripheral blood of secondary recipients that received uncultured LSK cells was 43%. However, culturing of the LSK cells prior to transplantation in the first recipient using STF or STIF media increased chimerism levels to 6063% and 6565% in the secondary transplanted mice, respectively. Incubation with Angptl3-containing media further increased chimerism levels for STFA3 and STIFA3 media to 7464% and 7764% respectively ( Figure 2B). The percentages of leukocyte chimerism-established by Y-chromosomal QPCR in bone marrow samples-was similar to the pattern of RBC chimerism levels in peripheral blood in these mice. To quantify the differentiation capacity of cultured LSK cells compared to freshly sorted LSK cells, primary recipient mice were transplanted with 12 or 120 LSK donor cells directly or with offspring cells from 12 or 120 LSK cells following a 7 day culture in STF, STFA3, STIF or STIFA3 media. Transplanting 12 freshly-isolated LSK Angptl3 Preserves Stemness of HSCs PLOS ONE | www.plosone.org cells let to a 2564% donor erythrocyte chimerism level in the peripheral blood 6 months after transplantation ( Figure 2C). Culturing of 12 LSK cells in STF or STIF media prior to primary transplantation resulted in donor erythrocyte chimerism of 3765% or 4064% in secondary transplanted mice, respectively. Culturing 12 LSK cells in STFA3 or STIFA3 media reconstituted 4864 and 5665% of donor erythrocyte chimerism levels, respectively. Again, leukocyte chimerism in the bone marrow phenocopied erythrocyte chimerism levels in peripheral blood of primary transplanted mice. Based on the results from the serial dilution transplantation experiment, culturing LSK cells in STF or STIF media prior to transplantation enhanced ,10 or ,6-fold long-term repopulation activity of HSCs. Culturing LSK cells in presence of Angptl3 enhanced number of LT-HSC ,3-fold, therefore STFA3 and STIFA3 media resulted in ,17-and ,32fold increase in long-term repopulation activity of HSCs compared to non-pretreated LSK cells, respectively ( Figure 2D). For Lin 2 ells, culturing in STF media significantly increased erythrocyte chimerism that was already visible 1 month after transplantation, and became more evident 6 months after transplantation. These results also demonstrated that 7 day culture period seems most optimal. Again, addition of Angptl3 further increased erythrocyte chimerism levels as well as leukocyte chimerism levels ( Figures S2A and S2B). Based on these limiting dilution transplantation experiments, we conclude that culturing HSCs in STF media for 7 days increased long-term repopulation capacity ,10-fold, and was increased by .300-fold for STFA3-incubated Lin 2 cells ( Figure S2C).
The effect of Angptl3 on ex vivo expansion and differentiation capacity of Lin 2 cells is direct
To further show that Angptl3 promotes expansion and reconstitution of HSCs, Angptl3 was ectopically expressed in Lin 2 hematopoietic progenitors by a lentiviral vector with GFP (LV-Angptl3-GFP). As a control, Lin 2 cells were transduced with a control vector solely expressing GFP (LV-GFP). Western blotting showed that sorted cells expressed the Angptl3 protein 2 days after transduction with LV-Angptl3-GFP whereas the control cells remained negative ( Figure 3A). Sorted, transduced cells were then cultured in STIF media and cell numbers were counted after 4 and 7 days. Expression of Angptl3 resulted in a significant increase in cell numbers ( Figure 3B), and preservation of progenitor cells and ST-HSCs compared to control cells as assessed by BFU-E and CFU-GM assays ( Figure 3C) or CFU-S assay ( Figure 3D), respectively.
We then assessed the reconstitution capacity of long-term hematopoiesis for Lin 2 cells using long-term repopulation ability assays (LTRA). For this, 10,000 sorted, male Lin 2 GFP + cells were transduced with the control LV-GFP or LV-Angptl3-GFP vector and transplanted into female a-thalassemia recipient mice. All mice remained healthy for over 9 months, and none of these mice exhibited tumor growth or elevated WBC counts. One month post-transplantation, we detected less than 2% GFP + cells in the Figure 4A). Over the following 8 months, the percentage of GFP + cells in the LV-GFP control group remained stable and varied between 6.5-8%. The percentage of GFP + cells in recipients that were transplanted with LV-Angptl3-GFP-transduced Lin 2 cells increased over months to 1362% nine months after transplantation. Percentages of GFP + cells were also significantly elevated in the bone marrow of these mice (2363%) compared to control mice (1562%) ( Figure 4B). Hematopoietic differentiation was also measured in GFP + cells in the peripheral blood, bone marrow and spleen compartments (Figures 4C-D-E). In the BM and PB, the LV-Angptl3-GFP group contained more GFP + cells with Sca-1 + and Sca-1 + /c-kit + stem cell markers than the LV-GFP control group. Thus Angptl3 may preserve the immature state of HSCs in vivo. No difference was observed for the percentage of mature CD4 + or CD8 + T cells or CD11b + myeloid cells within the GFP + population, but the percentage of CD19 + B cells was significantly decreased in the LV-Angptl3-GFP group compared to the control group.
We next sorted Lin 2 , GFP + cells from BM pools of recipient mice that were transplanted with LV-Angptl3-GFP or LV-GFPtransduced Lin 2 cells nine months before, and performed progenitor-and short-term colony forming assays. No significant difference in the frequency of BFU-E progenitor cells was identified in the BM Lin 2 , GFP + cell population from both groups (,12 vs. ,10 BFU-E colonies/2610 3 Lin 2 , GFP + cel ls) , re sp e ct iv e ly ( Figure 5A ) . Howe v e r , t he nu m ber of CFU -GMs was sig n ifi ca ntl y hig her for mice transplanted with LV-Angptl3-GFPtransduced Lin 2 cells relative to the control mice (3161.6 vs. 2162 CFU-GM colonies/26103 Lin 2 GFP + ) ( Figure 5B). In addition, the number of ST-HSCs was 2 fold higher for the LV-Angptl3-GFP group compared to the LV-GFP control group (761 vs. 361 CFU-S/103 BM Lin 2 , GFP + cells, respectively) ( Figure 5C).
Discussion
Over the past two decades, many attempts have been made to increase the quantity of long-term HSCs by in vitro culturing conditions. Although sufficient HSCs are obtained from donors for conventional bone marrow transplantations, expansion of HSC may become more and more relevant for transplantations relying on umbilical cord blood HSCs or transplantation of limiting, genetically-modified HSCs. Serum free expansion cultures of HSCs using SCF, TPO and Flt3L-supplemented media (STF) were shown to maintain the number of murine long term HSCs but led to increased numbers of human primitive hematopoietic progenitors with preserved engraftment potential [9,37,38]. Zhang et al. reported a new combination of growth factors that included SCF, TPO, IGF-2 and FGF-1 (STIF) that was supplemented with angiopoietin-like protein 2 or 3 (Angptl2, Angptl3) that supported ex-vivo expansion of murine long-term HSC frequencies by 24-to 30-fold in 10 days [26,27]. Furthermore, IGF-binding protein 2 (IGFBP2) and Angptl5 (A5) were introduced as additional factors that support human HSC expansion [39]. Using SCF [40], TPO [41], and FGF-1 supplemented with IGFBP2 and Angptl5, the number of human stem cells that can repopulate NOD-SCID mice increased ,20-fold compared to non-cultured HSCs [42].
In recent years, other factors have been identified that support in vitro expansion of HSCs. These stimulate the Wnt [43] and Notch [44] pathways that have been implicated in the regulation of HSCs fate [45]. Wnt signaling may inhibit glycogen synthase kinase 3 (GSK-3) thereby stabilizing b-actin that supports expansion of HSCs. However, inhibition of GSK-3 also results in the upregulation of the mammalian target of rapamycin (mTOR), which promotes the proliferation of committed progenitor cells. It was shown that a dual inhibitor for both GSK-3 and mTOR resulted in maintenance and expansion of HSCs in vitro, even in the absence of cytokines [45]. Microenvironmental factors such as pleiotrophin may also enhance HSC expansion in vitro and improved HSC repopulating capacity by ,10-fold using competitive transplantation assays [46]. Chemical compounds may have also have an effect: All-trans retinoic acid (ATRA) in combination with SCF, FLT3L, IL-6, and IL-11-enriched medium prolonged the repopulating capacity of HSCs [47]. The Cu 2+ -chelator tetraethylenepentamine (TEPA) enhanced ex vivo expansion of CD34 + CD38 2 and CD34+ Lin 2 subsets isolated from umbilical cord blood samples, as well enhanced their shortterm repopulating activity in NOD-SCID mice [48]. The histon deacetylase inhibitor (HDI) valproic acid [49] and StemRe-genin1-a small molecule antagonist of the Aryl hydrocarbon receptor [50]-were both able to promote long-term hematopoiesis following transplantation of cultured HSCs. Prostaglandin E2 (PGE2) may also be useful for ex vivo expansion of HSC [51,52]. Taken together, several factors can be utilized to optimize most prominent ex vivo expansion conditions for HSCs.
The optimal mix of cytokines and culture conditions that warrant most optimal ex vivo HSC expansion is not yet clear. We therefore explored ex vivo expansion of HSCs using two different combinations of growth factors (STF and STIF) with or without the addition of Angptl3. Angptl3 may provide optimal preserva-tion of stemness and promote long-term hematopoiesis without provoking leukemogenic or toxic effects following transplantation. The Angptl3 polypeptide (455 amino acids) has all the characteristic features of angiopoietins, and includes a signal peptide, an extended helical domain that forms dimeric or trimeric coiled coils, a short linker peptide and a globular fibrinogen homology domain (FHD). Angptl3 is expressed by BM-endothelial and other stromal cells, and binds directly to the cell-surface on HSCs [27]. For this, HSCs have the immune-inhibitory receptor human leukocyte immunoglobulin-like receptor B2 (LILRB2) that can bind various ANGPTLs. In mice, the orthologue paired immunoglobulin-like receptor (PIRB) has also been identified as a receptor for various Angptl proteins [28]. Binding of Angptl3 to the receptors of HSCs provoke differences in the expression of cell cycle regulators and transcription factors like repression of Ikaros but upregulation of Hes1 and Hoxa9, which are both important regulators for HSC self-renewal and differentiation [27]. Angptl3null mice were shown to have a 3-fold higher level of Ikaros [27]. Ectopic expression of Ikaros in mouse HSCs severely reduced hematopoietic reconstitution capacity following transplantation [27], which suggests that Angptl3 may be one of the most important regulators for HSC stemness [27].
Incubation of HSCs in STF and STIF media has a similar effect on the expansion of overall cell numbers as well as progenitors and short-term HSC numbers. We now demonstrate that STIF is superior over STF to preserve the total number of LSK cells, resulting in improved long-term hematopoietic repopulation results following transplantation. The addition of Angptl3 does not change the overall expansion rate, but improves the preservation of HSC stemness that support short-term and longterm hematopoiesis. Culturing LSK cells in the presence of STFA3 for 7 days improves short-term hematopoiesis in CFU-S assays by ,40-fold compared to the control group. The long-term repopulating capacity of LSK cells is enhanced ,32-fold following a 7 days incubation in STIFA3 media as estimated from peripheral blood erythrocyte chimerism levels. The effect of Angptl3 is clear percentage of GFP positive cells was determined, 9 months after re-transplantation in BM, and spleen in an average of 5 mice per group. (C) Nine months post transplantation, the percentage of different blood lineages (Sca-1, c-kit, Sca-1/c-kit, CD4, CD8, CD19, and CD11b cells) in donor-derived cells (GFP + ) in PB, (D) in BM, and (E) in spleen was measured by flow cytometry. *P#0.05. doi:10.1371/journal.pone.0105642.g004 both on Lin 2 cells as well as on highly purified HSCs (LSK-cells), indicating that the effect of Angptl3 is directly on HSCs and preserves stemness. In addition to a direct effect of Angptl3 on stem cells, we observed that the repopulating capacity of long-term hematopoiesis after transplantation was better for Lin 2 cells than for highly purified LSK cells, implying that Angptl3 may also exert an additional effect on accessory cells that support the maintenance of HSCs.
The effect of Angptl3 on preserving stemness of HSCs was further supported by finding reduced numbers of circulating CD19 + B-cells following transplantation of Angptl3-expressing HSCs. As overexpression of Angptl3 down-regulates Ikaros in HSCs [27], and since Ikaros deletions are frequently observed in B-cell acute lymphoblastic leukemias [53], this may provide an alternative explanation for the reduced circulating mature B-cells following transplantation of Angptl3-overexpressing HSCs. In our experiments, we did not find any evidence for leukemia or other types of cancer or toxicity at nine months following transplantation of Angptl3-overexpressing donor HSCs.
Conclusions
To conclude, we showed that the combination of five growth factors [SCF, TPO, IGF-2, FGF-1 and Angptl3 (STIFA3)] yielded a very significant expansion of stem cells that were able to provide long-term hematopoiesis in mice. This combination may promote superior engraftment over existing methods when using minimal stem cell numbers in the case of genetically-modified HSCs to treat genetic diseases or transplants that rely on umbilical cord blood stem cell donors. | 2017-06-03T02:41:16.784Z | 2014-08-29T00:00:00.000 | {
"year": 2014,
"sha1": "921b838b5aedd1f68a10fe06f9c1b0bcbfbf302e",
"oa_license": "CCBY",
"oa_url": "https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0105642&type=printable",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "921b838b5aedd1f68a10fe06f9c1b0bcbfbf302e",
"s2fieldsofstudy": [
"Biology",
"Medicine"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
267354941 | pes2o/s2orc | v3-fos-license | Characteristics of Building Envelope that Influence the Value of Operative Temperature on Office Buildings Based on Jakarta Climate Data
Architecture is accused of being an energy user and the most significant contributor to global warming, so it is the architect's responsibility to ensure that indoor comfort conditions are achieved with little or no traditional energy. This study focuses on knowing the best Range in building envelope configurations to obtain Operative Temperature comfort values. Simulation methods is used for model this research. Software that b used for this research are EnergyPlus v8.1 and Open Studio plugin version 0.7. The results show that the temperature of operative temperature (Top) does not reach a comfort value of 25 ° C by SNI 03-6572-2001 standards. If the air temperature setting is smaller, the energy consumption of the building will be higher. In the WWR variation, intensity of energy consumption per Ta was reduced by 1 ° C by 3.68 kWh / m2. Then in the variation of SHGC, increase energy consumption value of each Ta is reduced by 1 ° C by 3.44 kWh / m2. While for shading variations, the increase in value is 3.57 kWh / m2.
ABSTRACT
Architecture is accused of being an energy user and the most significant contributor to global warming, so it is the architect's responsibility to ensure that indoor comfort conditions are achieved with little or no traditional energy.This study focuses on knowing the best Range in building envelope configurations to obtain Operative Temperature comfort values.Simulation methods is used for model this research.Software that b used for this research are EnergyPlus v8.1 and Open Studio plugin version 0.7.The results show that the temperature of operative temperature (Top) does not reach a comfort value of 25 ° C by SNI 03-6572-2001 standards.If the air temperature setting is smaller, the energy consumption of the building will be higher.In the WWR variation, intensity of energy consumption per Ta was reduced by 1 ° C by 3.68 kWh / m2.Then in the variation of SHGC, increase energy consumption value of each Ta is reduced by 1 ° C by 3.44 kWh / m2.While for shading variations, the increase in value is 3.57 kWh / m2.
INTRODUCTION
Architecture is accused of being an energy user and the most significant contributor to global warming, so it is the architect's responsibility to ensure that indoor comfort conditions are achieved with little or no traditional energy [1].Energy-efficient architecture is an architecture with the lowest possible energy requirements that can be achieved by reducing the number of resources that make sense.The heat source in the building itself can be divided into two, namely external and internal heat loads [2].The heat that comes with natural light is an example of external heat, while heat derived from body metabolism, lighting, and equipment is an example of internal heat.This aspect has a significant share in contributing to energy consumption in buildings.Thermal comfort we can talk about [3], indeed will not be separated by making the temperature indicator a benchmark.The actual indicator that must be considered in achieving the comfort level of humans inhabiting that space is the Operative Temperature value, where Operative Temperature is the average value of the sum between water temperature and mean radiant temperature and is the temperature directlyfelt by skin [4].
This study explains in more detail the use of heating systems that oppose cooling systems.This shows that the farther away the area is from the site of the outermost opening, the higher the temperature will be felt.In the case of buildings with a heating system and a sub-tropical climate with four seasons.If the building uses a cooling system, the farther the distance from the opening area, the lower the temperature felt [5].
So far, storeys buildings that use artificial ventilation just pay attention to air temperature in the cooling system / HVAC [6].The Operative Temperature value must be achieving human comfort level, where Operative Temperature is the temperature directly felt by human skin (Borgstein, 2014).This study aimed to determine the characteristics of building envelopes to attain optimal operative temperature values based on climate data.In addition, it is also associated with the amount of energy consumption in buildings when affected by a decrease in the Operative Temperature value if applied in the city of Jakarta [7].
RESEARCH METHOD
This research uses simulation methods using modelling on computer The simulation in this research was carried out computerized with the help of the superior software used [8], namely EnergyPlus v8.1 and the Open Studio plugin version 0.7 which was run on Google Sketchup v7.0 software.Modeling in this system is carried out on the middle floor with modeling only on typical floors [9].In Figure 1, the hypothetical floor dimensions are 40m x 40m.The division of zones in a typical floor model can be explained as follows, four office activity zones (air-cooled) and one core zone in the middle of the building (not air-conditioned).The naming of office zones includes east, south, west and north zones.The top and bottom floors of the simulation floor will be modeled adiabatically [10].This type of building modelling is divided into six parts from each orientation, namely two parts of the corner or angle which gets sunlight from two directions and four parts of depth.Between depth is taken as a 3.5m drive because, at that distance, the Top temperature data is obtained, significantly different from the temperature of the Top zone in the first [11].The function of the building to be studied is an office with an open plan layout set to simplify the calculation of air-cooling load (cooling load).Work time is set 5 days a week starting at 08.00-18.00.
The simulated variables include (applied to each orientation) the area of glass (WWR 30%; 40%; 50%; 60%; 65%).In this study WWR 10% -20% was very rarely applied in building design so it was not applied as a variable.Besides that, it also shows water temperature as a variable, where the variations include Ta 20 ° C, 18 ° C, 15 ° C. Then horizontal shade element variables (VSA 30,VSA 50,VSA 70,VSA 90).
Effect of WWR Variation on the Radiant Temperature Mean Value
The size of the opening is very influential on the value of Mean Radiant Temperature (TMRT) [12] .In this simulation, the application of a large number of openings from 30% to 65%, by locking the glass SHGC at a value of 0.4 (Stopsol Dark Blue type) and the value of water temperature in the HVAC setting of 25 ° C without using shading.This comparison is done to show the effect of increasing the WWR number on changes in the value of TMRT.If the larger WWR means the wider the area of the glass and the smaller the size of the massive wall.The larger the area of the glass will give effect to raise heat that get into the room which will directly affect the value of TMRT [13].Based on the figure 2, the simulation results show that the greater the ratio of the glass field will the greater the value of the value obtained.In addition, it can be seen that the simulation results showthe influence of WWR on the value of each orientation obtained that is highest in the northwest orientation and lowest in the southern orientation.The range of TMRT values when carried out by WWR variations is at a temperature of 30.49° C-39.26 ° C. From a 10% drop WWR can reduce the temperature of the Tm by approximately 1 ° C-1.5 ° C.
From the explanation above it can be explained on the Table 1 that the glass field provides the greatest contribution to indoor heat compared to a massive field or wall.Thus, increasing the area of the wall will reduce the value of TMRT so that the heat will get into the building less.Further simulation analysis is carried out by looking at the energy consumption consumed by the building.The size of the temperature or temperature is closely related to the size of the energy consumed by the building, this is because the lower the perceived value, the greater the consumption of energy used to cool the room, but despite the greater energy consumption but the occupants of the building will reach the optimal level of comfort.The table below is a table of obtaining energy consumption from WWR 40 compared to WWR 65 in conditions of HGC still 0.4, Ta 25 ° C and without using shading.
The difference in the amount of energy consumption between WWR 40% and WWR 65% is 5 kWh / m2 explain ini Table 2 dan 3.These results show the greater the WWR the greater the energy consumption.This is becausethe bigger the WWR will be the greater the heat that will be received by the building.With this condition, the greater the energy needed to be able to cool the building.
Table 3.The Figure 5 is a graph of Ta modification in WWR 30% -65% on the western wall.When the Ta setting changes to 20 ° C that meet the Top value on SNI standards, the occupant's comfort value of 25 ° C is only WWR 30%.Then changes were made again to the Ta setting to 18 ° C, the results obtained were those that were able to meet the Top SNI standards, namely WWR 35%, WWR 40% and WWR 50%.The last modification done is to change the Ta setting to 15 ° C and the whole WWR variable from 35% -65% can reach the Top value ± 25 ° C.
The graph above is a simulation result by locking the WWR variable at 40%, SHGC 0.4 and without shading.The results can be seen if the conditions can reach Top ± 25 ° C in all orientation directions, namely in the Ta setting of 18 ° C and 15 ° C. Based on the explanation above, it can be summarized that the WWR is almost always fulfilled to reach the comfort level of Top 25 ° C, which is between 30% -40%.
The Effect of Decreasing Air Temperature on Energy Consumption Intensity The level of energy consumption is influenced by the air temperature settings carried out.Decreasing the air temperature to reach air temperature has a direct impact on the energy consumption intensity of the building.The smaller the Ta temperature setting, the performance of the HVAC system will increase, resulting in higher energy consumption in the building but comfortably will have a very good effect to people that used the building.The Figure 6 below is a graph of the simulation results in several variations of WWR can show the level of increase in energy consumption from changes in Ta settings.Source: Personal Analysis, 2019.
If seen from the table 5, it can be seen that every decrease in regulation Ta will increase energy consumption.The range of increases in energy consumption per 10% WWR increase in the reduction of 25 ° C to 20 ° C is 18 kWh / m2 -20 kWh / m2.Then an increase of about 24 kWh / m2 -27.5 kWh / m2 each increases 10% WWR when the Ta is decrease to 18 ° C from 25 ° C.After that there was an increase in energy consumption of approximately 33.5 kWh / m2 -38 kWh / m2 in the reduction of Ta from 25 ° C to 15 ° C at every 10% increase in WWR.
Analysis of Variations in Shading on the Radiant Temperature Mean Value
In an effort to determine the effect of using shading on TMRT values, a simulation was carried out by locking WWR at 40%, SHGC 04, and Ta setting 25 ° C. while the simulated free variable is VSA 30, VSA 50, VSA 70 and without shading.The existence of this shading is able to reduce the value of surface inside temperature significantly.With that it is also able to reduce the TMRT value of the room.Based on the figure 12 and 13 can be seen that shading can reduce the value of surface inside temperature, the wider the shading will be the smaller surface inside temperature.The wider the shading, the more constant the temperature or temperature is from each zone.This is because the wider the shading will be more able to withstand the heat or radiation of the sun entering the building.On the Table 9 and Figure 14 with less light entering the food, it will further reduce the Indirectly from the explanation on figure 14, if the wider the shading will decrease the surface inside temperature, the TMRT value will also decrease by applying shading.The detailed on the figure 15 and table 10.Based on Table 12, the size of the Top is closely connected to the building energy consumption.The results obtainedshowed that shading was able to reduce energy between 4 kwh / m2 11.5 kwh / m2.This depends on the length of the short shading used.The longer the shading is used, the lower the Top in the buildingwill be, so the smaller energy consumption will be smaller.
Based on the results of the TMRT and Top simulation, it was concluded that with the setting of water temperature (Ta) fixed at 25 ° C, nothing has been fulfilled to achieve a comfortable Top for occupants of 25 ° C because the lowest temperature value is still ± 28 ° C.However, based on simulations with a fixed Ta value of 25 ° C, it is sufficient to reduce or decrease the Top value by making changes to SHGC because it can reduce by ± 01 ° C-1.7 ° C.
With reference to the comparison table 13, the acquisition of Decrease Range results in theWWR variation gets the smallest value, which is in the range 0.1 ° C-0.6 ° C (every 10% reduction in WWR).This shows that this WWR variation has a big influence in increasing the Top value in the room.The larger the glass used will not only increase the cooling load value but will also create thermal discomfort in the room.
Figure 2 .
Figure 2. Graph Value of the WWR Intermediate Radiant Temperature Mean Source: Personal Analysis, 2019
Figure 3 .
Figure 3. Graph of Correlation between WWR and Operative Temperature Value Source: Personal Analysis, 2019.
Effect of Air Temperature on Operative Temperature ValueThe lowest Top temperature value when using the Ta setting of 25 ° C [14] which is 28.25 ° C.This shows on figure4that what has been done so far in designing comfort in space does not meet the appropriate standards.Then the next simulation is carried out by using air temperature (Ta) as a variable.The process carried out in the experiment is by changing the air temperature setting.Air Temperature variable values include 20°C, 18°C and 15°C.The target comfort value to be achieved is depends on the comfort value or operative temperature stated in SNI 03-6572-2001, namely 25°C.
Figure 4 .
Figure 4. Graph of Operative Temperature Value When Decreased in Air Temperature Settings Source: Personal Analysis, 2019.
Figure 5 .
Figure 5. Operative Temperature Value Chart When Performed in Decreasing Inter-WWR Temperature Water Settings Source: Personal Analysis, 2019.
Figure 6 .
Figure 6.Graph of the Value of Increase in Energy Consumption Intensity After the Decreasing of Temperature Water Settings Source: Personal Analysis, 2019.
Figure 12 .
Figure 12.Graph the value of Surface Inside Temperature and Mean Radiant Temperature at VSA 30 Source: Personal Analysis, 2019.
Figure 13 .
Figure 13.Graph of Surface Inside Temperature and Mean Radiant Temperature When Not Using Shading Source: Personal Analysis, 2019.
Figure 14 .
Figure 14.Graph Value Surface Inside Temperature between Shading Source: Personal Analysis, 2019.
Figure 15 .
Figure 15.Graph Value of Temperature Radiant Mean between Shading Source: Personal Analysis, 2019.
Table 1 .
Comparison of the Maximum Value and Minimum Surface Inside Temperature with Mean Radiant Temperature Table of Acquisition of Energy Consumption between WWR Table Comparison of Operative Temperature Values between Variables
Table 4 .
The Intensity Value of Energy Consumption (kWh / m2) between WWRs is based on a Decrease in Water Temperature Settings
Table 5 .
Difference in Value of IKE or Energy Consumption Intensity (kWh / m2) between WWR
38 Characteristics of Building Envelope that Influence the Value of Operative Temperature on Office Buildings Based on Jakarta Climate Data Nurina
Vidya Ayuningtyas, Istiana Adianti surface inside temperature and TMRT of the room.The existence of shading can reduce the surface inside temperature of approximately 1 ° C-3 ° C.Table9.Value of Surface Inside Temperature between Application Shading Source: Personal Analysis, 2019.
Table 10 .
Value of Temperature Radiant between Shading Applications
of Building Envelope that Influence the Value of Operative Temperature on Office Buildings Based on Jakarta Climate Data
Nurina Vidya Ayuningtyas, Istiana Adianti lowest Top Value at 28.61 ° C and the highest Top Value at 30.83 ° C.Figure 16.Graph of the Operative Temperature Value between Shading Source: Personal Analysis, 2019.
Table of Energy Consumption Gain between Shading
Table 13 .
Table Comparison of Operative Temperature Values between Variables Source: Personal Analysis, 2019. | 2024-02-01T16:30:25.158Z | 2024-01-29T00:00:00.000 | {
"year": 2024,
"sha1": "ef71e80dded848e252fac216ac1da6d552b6ed75",
"oa_license": "CCBYSA",
"oa_url": "https://library.matanauniversity.ac.id/ojs/index.php/marka/article/download/224/162",
"oa_status": "HYBRID",
"pdf_src": "Anansi",
"pdf_hash": "41c80468f9795a0eb92a1b13b249453b648b52c4",
"s2fieldsofstudy": [
"Environmental Science",
"Engineering"
],
"extfieldsofstudy": []
} |
10756236 | pes2o/s2orc | v3-fos-license | Mucus trail tracking in a predatory snail: olfactory processing retooled to serve a novel sensory modality
Introduction The rosy wolfsnail (Euglandina rosea), a predatory land snail, finds prey snails and potential mates by following their mucus trails. Euglandina have evolved unique, mobile lip extensions that detect mucus and aid in following trails. Currently, little is known of the neural substrates of the trail-following behavior. Methods To investigate the neural correlates of trail following we used tract-tracing experiments in which nerves were backfilled with either nickel-lysine or Lucifer yellow, extracellular recording of spiking neurons in snail procerebra using a multielectrode array, and behavioral assays of trail following and movement toward the source of a conditioned odor. Results The tract-tracing experiments demonstrate that in Euglandina, the nerves carrying mucus signals innervate the same region of the central ganglia as the olfactory nerves, while the electrophysiology studies show that mucus stimulation of the sensory epithelium on the lip extensions alters the frequency and pattern of neural activity in the procerebrum in a manner similar to odor stimulation of the olfactory epithelium on the optic tentacles of another land snail species, Cantareus aspersa (previously known as Helix aspersa). While Euglandina learn to follow trails of novel chemicals that they contact with their lip extensions in one to three trials, these snails proved remarkably resistant to associative learning in the olfactory modality. Even after seven to nine pairings of odorant molecules with food, they showed no orientation toward the conditioned odor. This is in marked contrast to Cantareus snails, which reliably oriented toward conditioned odors after two to three trials. Conclusions The apparent inability of Euglandina to learn to associate food with odors and use odor cues to drive behavior suggests that the capability for sophisticated neural processing of nonvolatile mucus cues detected by the lip extensions has evolved at the expense of processing of odorant molecules detected by the olfactory system.
Introduction
Chemical senses are arguably the oldest and most important sensory modalities in the animal kingdom. The earliest animals on the planet most likely navigated their environments by responding to chemical cues, and even now animals of all phyla rely on some type of chemosensation to obtain food, avoid predators, and find mates. Land snails and slugs are highly sensitive to odors and display robust associative conditioning to olfactory cues (Gelperin 1975;Kemenes 1989;Alkon and Nelson 1990;Sahley et al. 1990;Sahley and Crow 1998;Balaban 2002). These characteristics combined with their simple and relatively accessible nervous systems make them useful model systems for studying the neural substrates of sensory processing and learning, particularly in the chemosensory modality.
In many species of snails and slugs, the receptor cells of the olfactory epithelia (located on the two optical tentacles) send axons through olfactory nerves to a pair of cerebral ganglia (Hubendick 1955). Electrophysiological and imaging analyses have demonstrated that olfactory information processing and olfactory learning in many species of slugs and snails occurs in the procerebrum located at the point where the olfactory nerve joins the cerebral ganglion (Chase 1985;Gelperin and Tank 1990;Kimura et al. 1998;Straub et al. 2004;Ierusalimsky and Balaban 2010). The procerebrum consists of a layer of small, densely packed cell bodies and two separate layers of neuropil. The procerebrum shares several characteristics with the olfactory bulb of mammals, including large, spontaneous oscillations in the local field potential (Delaney et al. 1994) that are changed in frequency and amplitude by odor stimulation (Gelperin and Tank 1990;Gervais et al. 1996;Gelperin 1999). Work with the slug, Limax maximus, has shown that odor-cued associative conditioning alters the activity of procerebral neurons in a spatially specific way (Kimura et al. 1998;Teyke et al. 2000). Given the small size of the nervous systems of snails and slugs:~80,000-100,000 cells, approximately 75% of which are in the procerebra (Gelperin and Tank 1990;Balaban 2002), it is likely that the procerebrum plays a critical role in sensory processing in general, not just olfactory processing. Investigating a snail model in which a sensory modality other than olfaction is a significant determinant of behavior can shed light on the extent that the procerebrum is involved in processing of information in other sensory modalities.
Snails, similar to other gastropods, secrete mucus from their foot which aids in locomotion, acting as both glue and a lubricant (Denny 1980a,b). The mucus is left behind by the animal, forming a trail. Many species of gastropod have been reported to follow mucus trails of their own and other species to find mates, return to a "home" location, and in some cases to catch prey (for review see (Wells and Buckley 1972;Ng et al. 2013).
Euglandina rosea, the rosy wolfsnail, is a carnivorous land snail native to the Southeastern U.S. It tracks down its prey (other snails and slugs) as well as potential mates by following the mucus trails they leave behind. Euglandina snails follow mucus trails using a sophisticated chemosensory system that is separate from olfactory sensing (Chiu and Chou 1962). Previous work has shown that the sensory epithelia adapted for detecting mucus are on the long, mobile lip extensions that are absent in other snail species (Cook 1985a,b;Clifford et al. 2003). While tracking prey, the Euglandina constantly touch their lip extensions to the trail being followed. In laboratory experiments, they do not appear to detect mucus trails at a distance, and amputation of the optic (olfactory) tentacles had no effect on trail following while amputation of the lip extensions caused a large deficit in trail following (Cook 1985b;Clifford et al. 2003). The dependence of Euglandina on their lip extensions for mucus trail following is particularly striking given that other snails and slugs are able to follow trails of odors or mucus using their optic tentacles (Chase and Croll 1981;Cook 1985c).
In the field, Euglandina are voracious predators that, except for a specific, possibly distasteful slug, are known to eat almost any molluscan prey they encounter (Cook 1985b(Cook , 1989Kinzie 1992;Gerlach 1999Gerlach , 2001Meyer and Cowie 2010;Davis-Berg 2011). In the laboratory, Euglandina easily distinguish mucus of prey snails from that of other Euglandina. Although mucus trails from other Euglandina are followed at approximately the same frequency as prey snails (~90% of all trails encountered) adult Euglandina rarely attack other Euglandina. Similarly, prey snails that have been covered with Euglandina mucus are usually ignored after a brief inspection, while Euglandina that have been covered with prey mucus are rapidly attacked by the predator snails (Shaheen et al. 2005).
Euglandina also show robust chemosensory learning. They will follow artificial trails of novel, nonvolatile chemicals after only one or two trials of eating a prey snail coated with the chemical, but they do not learn to follow the artificial trails if exposure to test compounds is not paired with feeding on a prey snail (Clifford et al. 2003). Not only are Euglandina able to learn to follow artificial trails associated with food they also learn to follow trails of novel chemicals that have been paired with exposure to a conspecific (Shaheen et al. 2005). These results show that, in the mucus sensing modality, Euglandina have a sophisticated associative learning ability in which both food and access to potential mates can act as a reward to reinforce a voluntary behavior (following a trail of a novel compound). While previous work has demonstrated the centrality of mucus sensing to Euglandina behavior, it is not known how neural processing of mucus stimuli is carried out in the central ganglia. In addition, while the presence of odorants has been shown to disrupt trail following (Clifford et al. 2003) very little is known about the role of olfactory sensing in driving the voluntary behavior of Euglandina.
In this study, we sought to identify the neural pathways and processing that are important for mucus trail chemosensation in Euglandina and compare them to those involved in odor processing in a similarly sized, herbivorous land snail species, Cantareus aspersa. While there has been a report of trail following by Cantareus snails, trail following is not a prominent part of their behavioral repertoire. This is in sharp contrast to Euglandina, who follow nearly all trails encountered regardless of their state of hunger or satiety, and even if characteristics of the snail that they are following make it unsuitable for consumption (Cook 1985b(Cook , 1989Kinzie 1992;Gerlach 2001;Clifford et al. 2003;Davis-Berg 2011).
We also compared the relative significance of odor and mucus cues in directing Euglandina behavior by attempting to train the animals to orient toward food-associated odorants, and comparing their learning performance with that of Cantareus snails exposed to identical training paradigms.
Methods Anatomy
Tract-tracing experiments with Cantareus and Euglandina nerves using nickel-lysine or Lucifer yellow as back-filling dyes were done according to the methods of (Fredman 1987;Hernadi 2000). Euglandina and Cantareus snails were anesthetized by injecting 1-3 mL of cold 50 mmol/L MgCl 2 into the neck. The central ganglia connected to the optical and oral tentacles, and the lip extensions were dissected out of the snail, and the nerve coming from the sensory epithelia was cut and sucked into a micropipette containing a solution of 10 mmol/L nickel-lysine. The nerves were left in the nickel-lysine overnight at 4°C; then the snail brain was developed for one half hour with 3 mmol/L rubeanic acid, fixed overnight with 2% paraformaldehyde, dried with an alcohol series, and then cleared with methyl salicylate (wintergreen oil). In staining the superior tentacle nerves of both Cantareus and Euglandina, the optical nerve was separated from the olfactory nerve, and only the olfactory nerve was sucked into the pipette with the nickel-lysine. In staining the lip extension nerve of Euglandina, the nerve was cut between the cerebral ganglia and the joining of the lip extension nerve and inferior tentacle.
For backfilling experiments with Lucifer yellow, we followed the same procedure except that the backfilled brains were not developed, but simply fixed and dried with the alcohol series before visualizing with a fluorescent microscope (Olympus IX-71).
Euglandina central ganglia were stained with toluidine blue by modifying the procedures of (Altman 1980). Snail brains were dissected out of the snail and mounted on slides using Meyer's albumin fixative. After staining, the tissue was dehydrated and cleared with methyl benzoate.
Electrophysiology
Local field potential (LFP) oscillations were recorded from Cantareus and Euglandina procerebra using the Panasonic MED64 multielectrode recording system (Automate Scientific, Berkeley, CA). The MED64 probes contain 64 electrodes in an 8 9 8 matrix with interelectrode spacing of 75 lm. The electrodes are embedded in the center of a transparent glass dish.
Euglandina and Cantareus snails were anesthetized by injecting 1-3 mL of cold 50 mmol/L MgCl 2 into the neck. A single procerebrum connected to the superior and inferior tentacle nerves (Cantareus), and the lip extension nerves (Euglandina) was dissected out of the snail. The skin of the optical or lip extensions that contain the sensory epithelium was left intact and attached to the nerves.
The ganglion was laid across the electrode grid and pressed onto the grid with a slab of 2% agarose with the nerves and sensory epithelium uncovered by the agarose. To assess the effect of sensory stimulation, dilute solutions of odorant or mucus in water were applied to the sensory epithelia and the response of the neural networks was measured. Electrode traces were sampled at 20 kHz and the data were preprocessed by applying IIR Butterworth filters to remove 60 Hz power interference harmonics. High-frequency components (>5 kHz) that do not correspond to biological processes were removed using FIR LF filter with linear phase.
Paired association procedure
Each snail was tested for a baseline attraction to a dilute solution of each odorant before any other exposure to the odorant. After the initial test, each Euglandina was fed a prey snail (juvenile Cantareus, their regular diet in the lab) and 1-2 drops of a dilute odorant solution were dropped onto its radula as it ate. Because the procerebra were laid whole across the electrodes, the electrodes recorded neural activity from superficial cells in the cell mass layer.
Dilute solutions of four naturally occurring odorants were used. We chose 10% solutions of cinnamon oil, almond oil, bay oil, and anise oil as these are complex mixtures with multiple volatile compounds, and since they are used in food were likely to be safe for the snails to eat. A different odorant solution was used for each behavioral experiment so that the odor would be novel in the baseline condition. The snails were housed in a different room from where the feeding trials took place, which was also different from the room in which the test trials were run. The radula is the tooth-lined tongue that snails use to draw food into their mouths. Cantareus snails were fed minced carrots as their regular diet in the lab, and for the experiments, 1-2 drops of the dilute odorant were dropped on their radulas as they ate the carrot. The snails were tested again for attraction to the odorant 24-48 h after each training session in which eating was paired with exposure to an odorant.
Tests for formation of olfactory associations
The ability of Euglandina and Cantareus to learn to approach a novel odor through association of the odor with food was tested using three methods. In the first method, a cotton swab soaked in odorant (a 10% solution of either cinnamon oil or almond oil) was placed at the upper left corner of a 21 9 27.5 cm transparency sheet. The test snail was placed in the lower right corner of the same sheet facing the swab and at least 20 cm away ª 2013 The Authors. Brain and Behavior published by Wiley Periodicals, Inc. from it. The snails were allowed to crawl until they left the transparency sheet. The mucus trails of the snails were visualized by sprinkling the sheets with charcoal powder and rinsing under running water. The snail's sticky mucus trails trapped the dark powder so it remained on the sheet as the rest of the powder was washed away (Karowe et al. 1993). The distance from the snail's trail to the swab was measured and the closest point on the sheet was recorded. If the snails left the transparency sheet without moving toward the odorant, the closest distance to the swab was the starting point-20 cm away from the swab. Data from all snails tested were included in the analysis, regardless of whether they initially moved toward the swab or away from it. Significance of the data was tested with an ANOVA.
For the previous experiment, the snails were placed facing the odorant, and so might have a bias to move toward it that would affect the results. To ensure that the direction the snails faced was not the deciding factor in the decision to move toward the odor, we used a different approach to measure the attractiveness of the test odor. In the second type of odor learning experiment, a cotton swab soaked in a different odorant (10% bay oil) was placed in the middle of a 21 9 27.5 cm transparency sheet. The test Euglandina or Cantareus snail was placed 10 cm from the swab and facing the opposite direction. The test snails were allowed to crawl until they left the transparency sheet, and the trails were visualized with charcoal powder. Experiments were scored "attracted" versus "not attracted" based on whether the test snail turned around and moved toward the swab. Snails that turned around and traveled toward the swab past the point where the back of their shell had been placed at the start were scored as "attracted." To be scored as "attracted" the snails had to travel back past the point there they were originally placed within about three body lengths (~10 cm) distance from that point. Snails which did not turn around or did not travel past the point where they were placed at the start of the experiment within 10 cm were scored as "not attracted" (see Fig. 1C and D for examples). Significance of the data was tested with Logistic regression.
The ability of Euglandina to learn to follow artificial trails of an odorant chemical was tested by painting a streak of 10% anise oil on a transparency sheet, placing the snail 5 cm away from the chemical trail and allowing it to crawl across it. After the experiment, a marker pen was used to mark where the odorant trail was laid and the movement of the snail was visualized by sprinkling the sheet with charcoal powder and rinsing off the excess. After the first test of following the artificial trail, the snail was fed a prey snail while the anise solution was dropped on its radula, and the snails were tested for following of the trail again in 24-48 h. Snails were judged to have followed the trail if their mucus trail was superimposed over or paralleled the anise trail for at least three body lengths (approximately 10 cm). Significance was tested with a Logistic regression. For examples of typical olfactory and trailfollowing olfactory association experiments see Figure 1.
All behavior tests were conducted in conditions that were as consistent as possible. The experiments were performed under artificial lighting in the same windowless room. The testing arenas (the transparency sheets) were arranged on the same bench and in the same orientation for every trial. Snails from both species were oriented in the same direction relative to the room at the start of each trial for all of the behavioral tests.
Materials
All chemicals were purchased from Sigma-Aldrich Chemicals (St Louis, MO).
Neuroanatomy
Euglandina follow the mucus trails of prey and conspecific snails using the lip extensions visible in Figure 2A. As shown in Figure 2B, two large nerves connect the lip extension to a ganglion in the neck of the snail, and this ganglion is also connected to the oral tentacle nerve. A single large nerve, separate from the olfactory nerve, connects this "lip extension ganglion" to the cerebral ganglion, which enters laterally on the procerebrum while the olfactory nerve enters anteriorally (see Fig. 2C).
Backfilling experiments were done by dissecting the olfactory or lip extension nerve from the tentacle, cutting the end and dipping it into a solution of either nickellysine or the fluorescent dye, Lucifer yellow. Developing the nickel-lysine back-filled brain with rubeanic acid gives a blue color where the nickel accumulated, while fluorescent imaging reveals the accumulation of Lucifer yellow. Figure 3 shows the results of two different backfilling experiments in which the lip extension nerve of a Euglandina was backfilled with nickel-lysine or Lucifer yellow ( Fig. 3A and B). Both dyes show an accumulation in the procerebrum. In the Lucifer yellow backfill, the higher contrast possible shows that the labeling is in the crescent shape that is typical of the cell body layer of the procerebrum. As sensory nerves project from the periphery to the center, these backfilling experiments indicate that neurons carried in the lip extension nerve synapse in the procerebrum of the cerebral ganglion. with nickel-lysine ( Fig. 3C and D). The blue accumulated in the procerebrum just as with experiments in which the lip extension was backfilled.
Similar backfilling experiments were performed with Cantareus snails. Both the optical and oral tentacles were backfilled with nickel-lysine. As shown by the deposition of nickel from the backfilling, in Cantareus, the oral tentacle nerve enters laterally on the cerebral ganglia and innervates the procerebrum (Fig. 4A). When olfactory nerves of Cantareus are backfilled, deposits of nickel and Lucifer yellow appear in the procerebrum as well, but cover a larger area than the labeling when the inferior tentacle is backfilled (Fig. 4B). The crescent shape of the labeling of the procerebra of both Euglandina and Cantareus is consistent with the shape of the cell body layer in the procerebrum (Nagy and Sakharov 1970;Ermentrout In C, the snail turned to approach the swab, and the trial was scored as "attracted." In D, the snail did not turn to approach the swab within 10 cm of the initial location so the trial was scored as "not attracted." (E, F), Following an artificial trail of odorant (at location of pen line). Snails initial location at left corner of the sheet. In E, the test snail was scored as following the trail. In F, the test snail was scored as not following. The pen lines to mark the location of the swabs, the odorant trails and the "attracted/not attracted" line were all added after the conclusion of the behavioral experiment.
et al. 1998) suggesting that neurons in the optical, oral, and lip extension nerves synapse in the cell body layer of the procerebrum.
Electrophysiology
Oscillations in the local field potential (LFP) that change in frequency and amplitude in response to odor stimulation have been recorded from the cerebral ganglia in a number of mollusks including the slug, Limax maximus (Gelperin and Tank 1990) and the snail Helix pomatia (Chase 1981;Pin and Gola 1987;Sch€ utt et al. 1999). As shown in Figure 5, separate electrodes of the MED64 are able to record oscillations from Cantareus ganglia that are increased in frequency by the application of an odorant (10% bay oil) to the sensory epithelium of the tentacle. Interestingly, electrodes at the lateral edge of the procerebrum (#25 and #34) record a different pattern of LFP oscillations than an electrode placed more medially, and maintain a separate rhythm even after odor stimulation. Fifteen active electrodes were recorded from the cerebral ganglia of four different snails. Average spike frequency was 0.32 AE 0.04 Hz before odorant application and 1.48 AE 0.31 Hz after (P < 0.05; Kruskal-Wallis test). Similarly, recordings from Euglandina ganglia (Fig. 6) show an increase in both frequency and amplitude of LFP oscillations after stimulation of the lip extension epithelium with a mucus solution. As with Cantareus ganglia, the pattern of the oscillation varies in different parts of the procerebrum. Notice that before mucus stimulation, each electrode has a slightly different pattern of activity, even the electrodes closest together (numbers 14-16). After mucus stimulation an oscillating activity of frequency 3-8 Hz develops. The oscillation is largely synchronized across all electrodes, although some electrodes miss some peaks of the oscillations. Thirteen active electrodes were recorded from the cerebral ganglia of two different snails. Average spike frequency was 0.81 AE 0.53 Hz before odorant application and 2.84 AE 0.55 Hz after (P < 0.05; Kruskal-Wallis test).
Interestingly, in Euglandina ganglia that were stimulated by mucus applied to the lip extension, the neural activity recorded by neighboring electrodes alternated between periods of synchronization and desynchronization (Fig. 7). Even when the activity recorded by neighboring electrodes followed different rhythms, there was frequently a regular pattern of spikes that were synchronized (e.g., every third or fourth spike).
Odor conditioned behavior
Cantareus snails and Euglandina showed remarkable differences in their response to having a novel odor paired with consumption of food. Euglandina were tested with three different odor association paradigms using dilute solutions of three different complex, naturally occurring odorants. Compared to Cantareus snails, Euglandina were markedly less efficient at learning to approach conditioned odors. In the first test for odor association, snails were assessed for changes in their approach to a cotton swab containing an odorant that had been paired with food. In the baseline trial, prior to any feeding exposure, both Cantareus and Euglandina, on the average approached within 7-9 cm of the swab before leaving the sheet. After several paired feedings, the Cantareus snails came much closer to the swab, averaging only 2 cm away by the fifth trial. In contrast, with two different odorants, the average closest distance that Euglandina approached the swab did not change, even after seven paired feedings (Fig. 8A).
In a second odor-learning test with a different odorant, Cantareus and Euglandina were placed facing away from a swab coated with the odorant, allowed to crawl, and the direction that they crawled was scored as "attracted" if they turned around to crawl toward the swab and "not attracted" if they did not turn around. As with the distance to swab test, the Cantareus snails performed much better. In the baseline test, about 30% of the snails (Cantareus or Euglandina) turned toward the odorant.
After the second paired feeding, more than 50% of the Cantareus snails turned around and moved toward the odorant, and after eight paired feedings, 100% of the Cantareus test snails turned around to approach the odorant. In contrast, the Euglandina's performance never got above chance. At best, only 50% of the Euglandina snails turned toward the odor (after nine paired feedings), and there was no trend with increasing numbers of paired feedings (Fig. 8B).
The apparent inability of Euglandina to learn to travel toward novel odors associated with food is in marked contrast to their ability to learn to follow artificial trails of novel chemicals. Previous experiments with nonvolatile compounds showed that Euglandina learn to follow novel trails after one to three paired feedings (Clifford et al. 2003), and can learn to follow artificial trails paired with exposure to a potential mate as well as exposure to food (Shaheen et al. 2005). To rule out the possibility that the Euglandina's poor learning performance might be due to an inability to detect the volatile compounds that were used, we tested their ability to learn to follow an artificial trail of a new odorant molecule. After a baseline trial with an artificial trail of 10% anise oil, we fed test Euglandina prey snails paired with a solution of 10% anise oil. Twenty-four hours later, the snails were placed near an artificial trail of dilute anise oil and observed for trail following. Similar to what we have previously observed with nonvolatile artificial trails (Clifford et al. 2003), after a single paired feeding 50% of the test snails followed the artificial trail, with 80% of them following after two paired feedings.
Discussion
Laboratory experiments with the predatory snail Euglandina have shown that these snails have a highly developed ability to detect mucus from other snails and slugs and to select a response to mucus cues from a repertoire of several behaviors. Previous work has shown that based on cues in mucus, Euglandina can distinguish between prey snails and conspecifics as well as favored and unfavored prey species (Cook 1989;Clifford et al. 2003;Meyer and Cowie 2010) reacting differently to mucus trails depending on the identity of the trail layer. In the laboratory, the snails can tell the directionality only of conspecific trials, apparently by distinguishing the right side of the trail from the left (Shaheen et al. 2005), while in the wild, the snails appear to be able to determine trail directionality from prey trails as well (Davis-Berg 2011). Euglandina also rapidly learn to follow trails of novel chemicals associated with either prey snails or potential mates (Shaheen et al. 2005). While it is not certain that snail mucus does not contain volatile compounds that could contribute to trail attractiveness, previous work has shown that mucus trail following does not require olfaction, as amputating the optical tentacles has little impact on mucus trail following, while amputating the lip extensions eliminates trail following in most Euglandina tested (Cook 1985a;Clifford et al. 2003). While the olfactory systems and olfactory learning abilities of several species of slugs and snails have been extensively studied (Chase 1981(Chase , 1985Chase and Tolloczko 1993;Gelperin 1994;Gervais et al. 1996;Sahley and Crow 1998;Balaban 2002), almost nothing is known about the anatomy and physiology of mucus trail chemosensation. This study identifies connections between the lip extensions that mediate mucus trail detection and the cerebral ganglia, and demonstrates that mucus stimuli detected by the lip extensions are processed in the same central ganglia and in the same manner as odor molecules detected by the olfactory system. Our anatomical and tract-tracing experiments show that in the Euglandina, the nerve from the inferior tentacle joins with the nerve from the lip extension, and the combined nerve connects to the procerebral lobe where neurons from the lip extension synapse in the cell body layer. While a large swelling at the point where the lip extension nerve and oral tentacle nerve comes together suggests a ganglion, it is unlikely that afferent nerves from the sensory epithelium terminate at this point as nickel-lysine and Lucifer yellow taken up by the distal ends of lip extension nerve are transported past this point to the cerebral ganglion. Our results suggest that the connectivity and processing of input from the lip extension may have arisen in the Euglandina as an elaboration of the neural processing dedicated to the oral tentacle in other snails and slugs. This hypothesis is supported by our observation that in Euglandina, the lip extension nerve and oral tentacle nerve join, and the joined nerve enters the cerebral ganglion in the mid-lateral area where the oral tentacle enters in other land snails.
In the Euglandina, backfilling the lip extension nerve produces extensive labeling of the procerebrum appearing to cover the entire procerebrum, and resembles the results of backfilling of the Cantareus olfactory nerve more than the Cantareus oral tentacle nerve backfilling. Labeling of the Euglandina olfactory nerve produces labeling of the procerebrum that looks substantially the same as the labeling produced by backfilling the lip extension nerve.
In addition to the similarity between the anatomical labeling, the neuronal activity of the Euglandina procerebrum is similar to neuronal activity recorded from the procerebra of other land snails. The activity is characterized by a widespread oscillation in local field potential with a frequency of 0.1-0.3 Hz, and stimulation with odorants changes the frequency and amplitude of the oscillations (Chase 1981;Gelperin and Tank 1990;Kimura et al. 1993;Delaney et al. 1994;Ermentrout et al. 1998).
Multielectrode recording from Euglandina ganglia shows large increases in the frequency of the oscillations after mucus was applied to the lip extension, and an increase in the synchronization of the activity across a wide area of the procerebrum. Similar results were obtained with multielectrode recordings from Cantareus ganglia, where stimulating with an odorant on the olfactory epithelium increased the frequency of the synchronized field potential oscillation across a large stretch of the procerebrum.
Most intriguingly, our results demonstrate that Euglandina have paid a price for their highly developed responsiveness to mucus. Euglandina are very efficient at learning to follow trails of novel compounds associated with eating a prey snail or contact with a potential mate, (A) Results from conditioning experiment where the distance from snail trail to an odorant-soaked swab was measured at the closest point. Seven Euglandina were tested for learning for each of two different odors, while seven Cantareus were tested on a single odor. Data are means AE SEM. Significance was tested by ANOVA. Points at which the average distance to the swab was significantly different than the baseline at the P < 0.05 level are indicated with *. (B) Percent of snails turning toward odor on a swab was measured with nine Cantareus snails and 10 Euglandina. Significance was tested with a logistic regression. For the significant data points (indicated by *) the following values were obtained: four paired feedings v 2 = 4.22, P = 0.040; six paired feedings v 2 = 5.12, P = 0.023; seven paired feedings v 2 = 4.89, P = 0.027; nine paired feedings v 2 = 4.24, P = 0.039. (C) Seven Euglandina were tested for following of an artificial trail of an odorant (bay oil). The 0 time point represents baseline with no prior exposure to bay oil. Significance was tested with logistic regression, and is indicated with *. For the significant data points, the following values were obtained: two paired feedings v 2 = 4.77, P = 0.029; three paired feedings v 2 = 4.98, P = 0.026. as long as they can contact the compounds with their lip extensions (Clifford et al. 2003;Shaheen et al. 2005). However, they are strikingly ineffective at learning to orient or move toward novel odors detectable only with the olfactory sense on their optic tentacles, even when those odors have been repeatedly associated with food. Their lack of ability to learn that an odor is associated with a food source is in striking contrast to the abilities of Cantareus aspersa, another land snail of similar size, which learns to move toward a conditioned odor in just a few trials. The Euglandina's lack of ability to learn from odors is unlikely to be due to an inability to detect them, as earlier results have demonstrated that the presence of a strong odor can disrupt mucus trail following (Cook 1985a;Clifford et al. 2003). While not all of the specific odors tested in this study are in the native range of Euglandina, studies of olfaction in numerous species support the hypothesis that odor detection and olfactory transduction involve basic mechanisms that are universal across most species in most phyla (Hildebrand and Shepherd 1997,) so it is very unlikely that Cantareus snails could detect these odors while Euglandina individuals could not. Moreover, Euglandina are as efficient in learning to follow trails of volatile compounds as they are with nonvolatile compounds, once they are able to touch the trail with their lip extensions. This suggests that it is route of detection that is crucial, not the specific odors being tested.
Another possibility is that the dilute solutions of cinnamon, almond, and bay oils that we used as odorants are somehow aversive to Euglandina, and that prevents them from approaching the odors even when associated with food. Even if that is the case, similar studies with Limax maximus, have demonstrated appetitive conditioning to odors that were initially aversive to the slugs (Sahley and Crow 1998), suggesting that initial aversion can be overcome by pairing an odor with food.
The Euglandina has developed sophisticated central mechanisms to process mucus cues and use them to drive its behavior, and our data show that this seems to have occurred at the expense of processing of olfactory cues using the olfactory epithelium on the optic tentacles.
Our learning experiments demonstrate that, unlike other land snails and slugs, Euglandina appear to draw only limited information from odor stimuli, and the central processing capability devoted to odors does not appear to be enough to support associative learning. Given that land snails have only~40,000 neurons in each procerebrum (Gelperin and Tank 1990;Balaban 2002), the limited processing power available may have forced this trade-off during the evolution of the trail-following behavior. | 2018-04-03T02:46:30.442Z | 2013-12-18T00:00:00.000 | {
"year": 2013,
"sha1": "7310c0da39eae4037c33544a42dc9fdb3861af2e",
"oa_license": "CCBY",
"oa_url": "https://onlinelibrary.wiley.com/doi/pdfdirect/10.1002/brb3.198",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "7310c0da39eae4037c33544a42dc9fdb3861af2e",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
261629847 | pes2o/s2orc | v3-fos-license | Possible Mechanisms of Dark Tea in Cancer Prevention and Management: A Comprehensive Review
Tea is one of the most popular drinks in the world. Dark tea is a kind of post-fermented tea with unique sensory characteristics that is produced by the special fermentation of microorganisms. It contains many bioactive substances, such as tea polyphenols, theabrownin, tea polysaccharides, etc., which have been reported to be beneficial to human health. This paper reviewed the latest research on dark tea’s potential in preventing and managing cancer, and the mechanisms mainly involved anti-oxidation, anti-inflammation, inhibiting cancer cell proliferation, inducing cancer cell apoptosis, inhibiting tumor metastasis, and regulating intestinal flora. The purpose of this review is to accumulate evidence on the anti-cancer effects of dark tea, the corresponding mechanisms and limitations of dark tea for cancer prevention and management, the future prospects, and demanding questions about dark tea’s possible contributions as an anti-cancer adjuvant.
Introduction
Nowadays, the incidence and death rate of cancer are increasing year by year.Cancer is not only the leading cause of death but also the main obstacle to extending human life in various countries [1].According to recent statistics from the World Health Organization (WHO), there were 19.3 million new cancer cases and 10 million cancer-related deaths worldwide in 2020.The four most common cancers in the world are lung cancer, female breast cancer, bowel cancer, and prostate cancer, which account for four out of ten of all cancers diagnosed worldwide.By 2040, 27.5 million new cases of cancer are expected to occur globally each year [2].Given this trend, the prevention and management of cancer have urgently become important issues worldwide.The occurrence of cancer is closely related to people's lifestyle, dietary habits, environmental changes, genetics, and other factors [3,4].Previous studies have reported a negative correlation between tea consumption and cancer [5,6].
Tea can be divided into green tea, black tea, oolong tea, white tea, yellow tea, and dark tea according to the different degrees of fermentation.Dark tea belongs to post-fermented tea.According to the origins, processing technology, and different fermentation strains, there are several types of dark tea in China, mainly including Yunnan Pu'er tea, Guangxi Liupao tea, Hunan Fubrick tea, Hubei Qingbrick tea, Sichuan Kangbrick tea, and so on.Different dark teas are different in appearance, aroma, and content of active ingredients.For example, Fu brick tea has a higher content of catechin and flavonoid glycosides, while Pu'er tea has a higher content of catechin derivatives, flavonoids, and alkaloids [7].Studies have confirmed that dark tea has a high potential for anti-cancer activity for human health [8][9][10][11], and it is recommended as a nutritional supplement for the prevention and management of cancer [12][13][14].Therefore, this review summarized the possible anti-cancer mechanisms of dark tea in the past 10 years through articles on the Web of Science, PubMed, and Google Scholar.The following string terms are used to search for articles: "dark tea" and "anti-inflammation" or "anti-oxidation" or "anti-cancer".We explore the potential of dark tea for preventing and managing cancer and also discuss the corresponding mechanisms and limitations.
Bioactive Components in Dark Tea
Dark tea contains a variety of bioactive compounds, including catechins, phenolic acids, flavonols, flavones, alkaloids, terpenoids, and their derivatives [9].It has unique flavor substances and active ingredients that are different from black tea, green tea, and oolong tea.Microbial fermentation is believed to be a key factor responsible for shaping the sensory properties and beneficial health effects of dark tea [15].
The polyphenols in dark tea include catechins, flavonoids, anthocyanins, and other compounds, which have biological activities such as anti-oxidation, antisepsis, and antiinflammation [16][17][18].Dark tea is also rich in amino acids, especially theanine and γaminobutyric acid, which have biological activities such as enhancing immunity, antifatigue, improving memory, and calming [19][20][21].In addition, dark tea contains short-and medium-chain saturated fatty acids, phenolic acids, organic acids, alkaloids, nucleotides, etc. Alkaloids such as caffeine have been reported to lower blood cholesterol, excite nerves, and benefit cardiovascular health [22].Polysaccharides are also abundant in dark tea and have biological activities such as regulating immunity, lowering blood sugar, and improving intestinal flora.Last but not least, theabrownin, a heterogeneous polyphenolic compound, is one of the most active and abundant pigments in dark tea, and it is also the characteristic constituent of dark tea.In recent ten years, theabrownin in dark tea has been reported to possess various kinds of activities, including anti-oxidation, anti-inflammation, reducing plasma lipids and body weight, anti-cancer, and prevention of diabetes [23][24][25][26].
Anti-Cancer Effects of Dark Tea and Its Components in Different Cancer Models
Active components in dark tea have been shown to possess anti-cancer effects in different cancer models, such as lung, liver, skin, breast, and pancreatic cancers, in vivo and in vitro, but few population experiments have been conducted.
The water-soluble polysaccharides isolated from the dark brick tea significantly inhibited the proliferation of lung cancer cells A549 and liver cancer cells SMMC7721 [27].Theabrownin was reported to inhibit the proliferation of lung cancer cells in vitro and in vivo [28].A new epigallocatechin gallate derivative isolated from Anhua dark tea enhanced the apoptosis of lung cancer cells HCC827-Gef [29].The extract of Ya'an Tibetan tea has an anti-proliferation effect on the human hepatoma cell line HepG2 [30].Theabrownin significantly promoted the apoptosis of human melanoma cells A375 and inhibited the growth of zebrafish A375 xenograft tumors [31], and also inhibited the proliferation of Huh7 liver cancer cells by activating the Jun N-terminal Kinase pathway, as well as inhibited the tumor growth in Huh7 xenografted zebrafish [32].Dark tea contains caffeine, theophylline, and theobromine, and theophylline inhibits the proliferation of breast cancer cells MDA-MB-231 and MCF-7 and cervical cancer cells HeLa by down-regulating the expression of SRSF3 [33].Pu'er tea extract can inhibit the proliferation of human tongue cancer TCA8113 cells and has been proven to prevent oral mucosa cancer in mice [34].Dark tea extract inhibited the proliferation of human pancreatic cancer cells SW1990, PANC-1, and human colon adenocarcinoma cells SW1116.After subcutaneous injection of PANC-1 cells into athymic nude mice, dark tea inhibited the growth of cancer cells in a xenograft tumor model [35].
The Possible Anti-Cancer Mechanisms of Dark Tea
As shown above, the anti-cancer effects of dark tea have been extensively studied, and their possible mechanisms are shown in Figure 1
The Possible Anti-Cancer Mechanisms of Dark Tea
As shown above, the anti-cancer effects of dark tea have been extensively studied, and their possible mechanisms are shown in Figure 1 and Table 1 below.
Anti-Inflammation
Research shows that the development of cancer is closely related to the inflammatory response, and many cancers are caused by chronic inflammation [36,37].Nuclear factorkappa B (NF-κB) is a transcription factor involved in a variety of biological processes, including immunity and inflammation [38,39], and it can facilitate the aggressive phenotype and transcription of oncogenes in cancer cells.In a prospective cohort study, a long-term anti-inflammatory diet was found to improve survival for breast cancer survivors [40].Dark tea has been shown to have an anti-inflammatory effect [41], and the prevention and treatment of cancer from the perspective of anti-inflammatory has been evaluated in many clinical studies and reports [42].In the analysis of the therapeutic effect of six kinds of tea on the liver injury induced by carbon tetrachloromethane (CCl4) in mice, dark tea can inhibit the NF-κB pathway to be anti-inflammatory and reduce liver injury, while green tea, yellow tea, oolong tea, and white tea have no such effect [43].TB extracted from Pu'er tea regulates the immunity of RAW264.7 macrophages and inhibits inflammation by inhibiting the NF-κB/MAPK/PI3K-AKT signaling pathway [44].Dark tea has also been shown to alleviate colitis induced by dextran sodium sulfate and reduce the expression level of inflammatory factors, mainly by regulating the NF-κB and HIF-1α signaling pathways, regulating gut bacteria, and enhancing the synthesis of short-chain fatty acids such as butyrate [45].In DSS model mice, the intervention of Pu'er tea extract (100 mg/kg) and the positive control drug 5-aminosalicylic acid (100 mg/kg) were both found to reduce the levels of TNF-α, IL-6, and other pro-inflammatory cytokines in the serum and colon of colitis mice, inhibit the activation of the NF-κB pathway, and down-regulate the expres-
Anti-Inflammation
Research shows that the development of cancer is closely related to the inflammatory response, and many cancers are caused by chronic inflammation [36,37].Nuclear factorkappa B (NF-κB) is a transcription factor involved in a variety of biological processes, including immunity and inflammation [38,39], and it can facilitate the aggressive phenotype and transcription of oncogenes in cancer cells.In a prospective cohort study, a long-term anti-inflammatory diet was found to improve survival for breast cancer survivors [40].Dark tea has been shown to have an anti-inflammatory effect [41], and the prevention and treatment of cancer from the perspective of anti-inflammatory has been evaluated in many clinical studies and reports [42].In the analysis of the therapeutic effect of six kinds of tea on the liver injury induced by carbon tetrachloromethane (CCl4) in mice, dark tea can inhibit the NF-κB pathway to be anti-inflammatory and reduce liver injury, while green tea, yellow tea, oolong tea, and white tea have no such effect [43].TB extracted from Pu'er tea regulates the immunity of RAW264.7 macrophages and inhibits inflammation by inhibiting the NF-κB/MAPK/PI3K-AKT signaling pathway [44].Dark tea has also been shown to alleviate colitis induced by dextran sodium sulfate and reduce the expression level of inflammatory factors, mainly by regulating the NF-κB and HIF-1α signaling pathways, regulating gut bacteria, and enhancing the synthesis of short-chain fatty acids such as butyrate [45].In DSS model mice, the intervention of Pu'er tea extract (100 mg/kg) and the positive control drug 5-aminosalicylic acid (100 mg/kg) were both found to reduce the levels of TNF-α, IL-6, and other pro-inflammatory cytokines in the serum and colon of colitis mice, inhibit the activation of the NF-κB pathway, and down-regulate the expression of HIF-1α [46,47].Pu'er significantly down-regulated the levels of inflammatory pathway proteins (MyD88, TLR4, p38MAPK, and p-NF-κB p65); therefore, drinking aged Pu'er tea (10 mg/kg BW per day for mice and 7 g/kg BW per day for humans) has a practical effect on alleviating and preventing the development of intestinal inflammation [48].Fuzhuan brick tea extract (30 mg/kg and 60 mg/kg) significantly reduced DSS-induced rectal bleeding, shortened colon length, and reduced the production of inflammatory cytokines in mice [49].Fuzhuan tea polysaccharide reduced the level of inflammatory cytokines IL-1β, IL-6, IFN-γ, and TNF-α in DSS-induced mice [50].The anti-inflammatory effect of Fuzhuan brick tea has been shown by Dai et al. to alleviate diarrhea in mice, significantly reducing the level of the pro-inflammatory factor 5-hydroxytryptamine (5-HT) and increasing the expression of sodium hydrogen exchanger 3 (NHE-3).Compared with berberine (60 mg/kg) in the positive control group, the effects of Fuzhuan brick tea (2530, 1260 mg/kg) on diarrhea were more notable [51].Research shows that colitis can lead to colon cancer.Hu et al. found that Pu'er tea can not only diminish the development of inflammation and reduce the level of pro-inflammatory factors but also promote the synthesis of short-chain fatty acid butyrate (BA).Additionally, BA protects the intestinal barrier, which can significantly reduce inflammatory factors (IL-1β and IL-6) and increase anti-inflammatory cytokine release (IL-10 and IL-22).In addition, BA helps reduce the up-regulation of PI3K/AKT/NF-κB pathway proteins by DSS [52].Overall, these findings indicate that dark tea may have potential as a natural approach to preventing colon cancer by reducing inflammation.
In the model of inflammation induced by fat accumulation, dark tea has been shown in many studies to reduce the levels of pro-inflammatory factors and relieve the inflammatory response in the body.Liu et al. observed that Pu'er tea decreased the expression of the proinflammatory factor IL-6 in mice fed a high-fat diet (HFD) [53].Xiao et al. found that ApoE knockout mice treated with Pu'er tea for 16 weeks had decreased relative mRNA expression levels of the pro-inflammatory factors IL-6, IL-12, and TNF-α in aortic tissues, and the activity of NF-κB was significantly reduced.This shows that Pu'er tea can inhibit lipid deposition in the blood vessel wall, reduce the inflammatory response, and finally inhibit plaque formation in the arterial wall [54].Zhu et al. found that after Pu'er tea intervention, the levels of IL-6 and TNF-α were significantly reduced in both animal experiments and cell experiments, which reversed the inflammation caused by fat accumulation in the liver [55].Similar conclusions were also verified by Cai et al., who showed that Pu'er tea significantly reduced the levels of TNF-α, Mcp1, and IL-1β in the liver of HFD-fed mice, which were regulated by IL-6/STAT3 signaling in the liver of mice [56].Dark tea is also rich in tea polyphenols; it (TPs50%; 0.50 g/kg) can inhibit liver inflammation by inhibiting the expression levels of COX-2 and iNOS in an obese dog model.Increased COX-2 expression is seen in a variety of inflammatory diseases and human cancers, including lung and liver cancer [57].Pu'er tea also significantly reduced the levels of IL-8, IL-1β, IL-6, and TNF in peripheral blood mononuclear cells (PBMC) obtained from healthy people's blood samples (male = 6 and female = 6) [58].The above content shows that dark tea can help reduce inflammation, possible pathways are shown in Figure 2 and the reduction of inflammation may reduce the risk of cancer, but it is not limited to this.It is worth noting that most research has focused on colitis and inflammation caused by fat accumulation.In view of this, future studies can focus on the anti-inflammatory effects of dark tea on other organ sites.
Antioxidation
Oxidative stress occurs when the production and removal of oxygen-free radicals in the body or cells are unbalanced.Oxidative stress is associated with the occurrence of a variety of diseases but also affects the evolution of cancer [59], stimulates tumorigenesis, and promotes the transformation and proliferation of cancer cells.Studies have demonstrated the antioxidant properties of dark tea [60,61], and antioxidants are negatively correlated with the occurrence of cancer [62].Therefore, the consumption of dark tea may help prevent cancer by reducing oxidative stress in the body.Numerous researchers have used in vitro experiments to verify the antioxidant properties of dark tea.By using ABTS, FRAP, DPPH, HSA, SSA, and other experiments, the antioxidant properties of dark tea were confirmed.Zhao et al. found that dark tea had a strong antioxidant capacity; the total phenol content of dark tea was (81.43 ± 40.92 mg GAE/g DW), the FRAP value was (1472.27± 691.91 µmol Fe 2+ /g DW), and the TEAC value was (715.99 ± 352.02 µmol Trolox/g DW) [63].Among them, polyphenols, including tea polyphenols, five catechins (EC, GC, GCG, CG, and EGCG), and two flavonoids (rutin and kaolin), showed a highly significant positive correlation with in vitro antioxidant activities [64].Compared to mature Pu'er, Roda et al. found that the average antioxidant activity of raw Pu'er was higher, and the main contributors were polyphenols and flavonoids [65,66].Zhang et al. found that phenolic components extracted from Pu'er have stronger antioxidant activity than vitamin C, and Pu'er may be an ideal natural antioxidant [67].Guo et al. planned to compare the antioxidant capacity of tea polysaccharides extracted from different tea varieties and found that the total phenol and antioxidant activity of Pu'er tea polysaccharide (TP-4) were the highest through an in vitro antioxidant assay [68].Zheng et al. extracted tea polyphenols and tea polysaccharides from Ya'an Tibetan tea, which were found to have antioxidant activity in vitro using DPPH [69].In addition to the significant contribution of polyphenols to antioxidation, Su et al. isolated thealenol A from Pu'er tea and determined the antioxidant activity of teadenol A by DPPH and T-AOC in vitro [70].The polysaccharides extracted from Fu Brick tea and Qingzhuan brick tea also have strong antioxidant activity in vitro [71,72].Overall, these studies show significant antioxidant activity in dark tea, which is
Antioxidation
Oxidative stress occurs when the production and removal of oxygen-free radicals in the body or cells are unbalanced.Oxidative stress is associated with the occurrence of a variety of diseases but also affects the evolution of cancer [59], stimulates tumorigenesis, and promotes the transformation and proliferation of cancer cells.Studies have demonstrated the antioxidant properties of dark tea [60,61], and antioxidants are negatively correlated with the occurrence of cancer [62].Therefore, the consumption of dark tea may help prevent cancer by reducing oxidative stress in the body.Numerous researchers have used in vitro experiments to verify the antioxidant properties of dark tea.By using ABTS, FRAP, DPPH, HSA, SSA, and other experiments, the antioxidant properties of dark tea were confirmed.Zhao et al. found that dark tea had a strong antioxidant capacity; the total phenol content of dark tea was (81.43 ± 40.92 mg GAE/g DW), the FRAP value was (1472.27± 691.91 µmol Fe 2+ /g DW), and the TEAC value was (715.99 ± 352.02 µmol Trolox/g DW) [63].Among them, polyphenols, including tea polyphenols, five catechins (EC, GC, GCG, CG, and EGCG), and two flavonoids (rutin and kaolin), showed a highly significant positive correlation with in vitro antioxidant activities [64].Compared to mature Pu'er, Roda et al. found that the average antioxidant activity of raw Pu'er was higher, and the main contributors were polyphenols and flavonoids [65,66].Zhang et al. found that phenolic components extracted from Pu'er have stronger antioxidant activity than vitamin C, and Pu'er may be an ideal natural antioxidant [67].Guo et al. planned to compare the antioxidant capacity of tea polysaccharides extracted from different tea varieties and found that the total phenol and antioxidant activity of Pu'er tea polysaccharide (TP-4) were the highest through an in vitro antioxidant assay [68].Zheng et al. extracted tea polyphenols and tea polysaccharides from Ya'an Tibetan tea, which were found to have antioxidant activity in vitro using DPPH [69].In addition to the significant contribution of polyphenols to anti-oxidation, Su et al. isolated thealenol A from Pu'er tea and determined the antioxidant activity of teadenol A by DPPH and T-AOC in vitro [70].The polysaccharides extracted from Fu Brick tea and Qingzhuan brick tea also have strong antioxidant activity in vitro [71,72].Overall, these studies show significant antioxidant activity in dark tea, which is due to the presence of various active components such as polyphenols, flavonoids, and tea polysaccharides.
Several studies have investigated the antioxidant activity of dark tea in vitro.Hou et al. found that Pu'er tea extract diminished the level of MDA in kainic acid-treated PC12 cells, and inhibited the production of cellular ROS and lipid peroxidation [73].Wang et al. demonstrated that Pu'er tea can remove ROS produced by human cancer cells (Caco2 and HepG2 cell lines) [74].The 2S,3R-6-methoxycarbonylgallocatechin isolated from Anhua dark tea protected NRF2/ARE HEK293 cells from ROS via NRF2 activation [75].The 8-C N-ethyl-2-pyrrolidinone-substituted flavan-3-ols extracted from Pu'er tea also possessed significant antioxidant activity and could prevent HMEC damage caused by H 2 O 2 [76].
In vivo data also confirmed the antioxidant effect of dark tea.Dark tea extract reduced ROS levels in hematopoietic cells by inhibiting the expression of NOX4 and significantly improved the survival of mice exposed to 7.0 and 7.5 Gy total body irradiation [77].Pu'er tea extract can improve the activities of antioxidant enzymes such as glutathione peroxidase (GSH-Px) and superoxide dismutase (SOD) in HFD rats and decrease the level of malondialdehyde (MDA), a lipid peroxidation product, in obese rats [78].It can also inhibit oxidative stress and lipid peroxidation [79].Braud et al. found that the Pu'er tea extract mixture decreased the production of the ROS marker O 2 − in rat hepatocytes and also prevented t-BHP-induced mitochondrial oxidative stress [80].Pu'er tea extract reduces serum MDA levels induced by quinocetone and increases the activities of GSH, SOD, and GPx.The main mechanism is to improve ERK phosphorylation levels and thus increase Nrf2/HO-1 pathway expression [81].Similar results have also been confirmed.Wang et al. found that in QCT-induced mice, under the action of Pu'er tea extract, ROS accumulation is reduced, antioxidant activity is enhanced, and levels of antioxidant enzymes (SOD, GPx, and CAT) as well as non-enzymatic antioxidant GSH are also increased [82].Zheng et al. found that Pu'er tea powder could reduce MDA and GSH levels and increase SOD and GSH-PX levels in rats induced by cisplatin [83].Contrary to the findings of in vitro investigations, Cao et al. observed that dark tea had greater antioxidant capacity than green tea in vivo; dark tea extract reduced the level of malondialdehyde and increased the levels of superoxide dismutase, glutathione peroxidase, and glutathione in the liver of mice, indicating that tea has antioxidant and hepatoprotective activities [84].
There are many studies on the antioxidant properties of dark tea in vitro, and more in-depth cell and animal experiments should be carried out to further verify its antioxidant properties.Dark tea antioxidant summary are shown in Table 2.In animal experiments, there are many studies on high-fat diet modeling, and more disease models can be designed in the future to verify antioxidant activity.Moreover, most of the antioxidant studies only focus on the apparent enzymes without further exploring the pathways and mechanisms of antioxidants.
Inhibiting the Proliferation of Cancer Cells
Dark tea inhibits proliferation by inhibiting the cell cycle.In the study of breast cancer, proline dehydrogenase (PRODH) induced epithelial-mesenchymal transformation of cancer cells and increased cell proliferation.Xie et al. also focused on breast cancer and found that the proliferation of the human breast cancer cell line MDA-MB-231 was inhibited by Pu'er tea and that its mechanism was through activation of the JNK pathway [85].
Tumor cell growth is inhibited by down-regulating the S phase and causing stagnation in the G1 and G2 phases [86].The inhibition of cell proliferation by Pu'er tea has been involved in numerous cancer studies.Pu'er tea induces G1 phase block in HepG2 cells, possibly through the activation of AMPK to improve the expression level of P21 [87].EGCG inhibits the proliferation of oral cancer HSC-3 cells, causing them to stall in the G1 phase of the cell cycle [88].In the gastric cancer cell line SGC-7901, Pu'er tea blocked the S phase of the cell cycle and significantly reduced the G2/M phase [89].The new epigallocatechin gallate derivatives isolated from Anhua dark tea showed better antitumor activity than EGCG and inhibited the proliferation of lung cancer cells HCC827-Gef [29].In human melanoma cells A375, TB in Pu'er tea inhibited their proliferation, causing DNA loss and inducing apoptosis; meanwhile, in a zebrafish xenotransplantation model of A375 cells, TB inhibited tumor growth, and the effect of TB was cancelled only by simultaneously eliminating p53 and p65.It is confirmed that it is mediated by the p53/NF-κB signaling pathway [31].TB (500 µg/mL) can also significantly inhibit the proliferation of colon cancer cells (HT-29) through long-term oxidative stress [90].Chen et al. conducted a population study and found that long-term Pu'er tea consumption down-regulates α9-nicotine-acetylcholine receptor (α9-nAchR) expression and protects against the formation of foam cells induced by nicotine inhalation due to smoking, which is beneficial to reduce the risk of lung cancer [91].
Promoting Cancer Cell Apoptosis
Microbial fermentation improves the bioavailability of dark tea components.It was found that methylxanthine isolated from Pu'er tea contributed to the anti-tumor activity of Pu'er tea and showed concentration-dependent growth inhibition on two cancer cell lines, MDA-MB-231 and HT-29 [92].Zhao et al. found that fermented Pu'er tea increased the number of HT-29 colon cancer cells and human umbilical vein endothelial cell apoptotic bodies by DAPI staining and flow cytometry analysis.The expression of Bax, caspase-9, and caspase-3 mRNA increased, and the expression of Bcl-2 decreased [93].Yao et al. used a weak tea polyphenol base to prepare theabrownins and found that it could induce apoptosis in HT-29 cells.They also explored the possible mechanism of inducing REDOX imbalance in cancer cells [94].In addition, Liang et al. conducted in vivo experiments to explore the anti-cancer effect of theabrownins.TB promoted apoptosis of cancer cells, and the pathway was found to reduce phosphorylation of phosphatidylinositol 3-kinase (PI3K) and protein kinase B (Akt) [95].Xiao et al. studied the effect of TB on lung cancer and found that TB induced apoptosis of lung cancer cell H1299 and inhibited tumor growth in zebrafish, which was a p53-independent mechanism mediated by the activation of the MAPK/JNK signaling pathway [96].Zhao et al. found that Pu'er tea can inhibit the proliferation of human tongue cancer TCA8113 and induce its apoptosis through the decrease in matrix metalloproteinase (MMPs) and the increase in tissue inhibitor of metalloproteinase (TIMPs) mRNA transcription.It was also verified in a mouse experiment that the tumor volume of mice treated with Pu'er tea was significantly reduced, and apoptosis was induced by up-regulating Bax and down-regulating Bcl-2 [34].Xu et al. concentrated their views on the prevention and treatment effects of TB on liver cancer and found that TB induced apoptosis of liver cancer cells (Huh7), up-regulated the expressions of ASK1, p-JNK, p-c-Jun, and Bax, and down-regulated Bcl-2.The JNK signaling pathway was also verified in in vivo experiments with zebrafish [32].Zhang et al. showed that using Pu'er tea in combination with cancer immunotherapy is an effective way to fight cancer [97].
In conclusion, dark tea can prevent and treat cancer by inhibiting the proliferation, metastasis, and apoptosis of cancer cells.Possible pathways are shown in Figure 3.However, most of the studies only stay at the cellular and animal levels, and more population studies should be carried out.
Anti-Obesity
Obesity has become a widespread problem [98].Various variables contribute to its occurrence, such as diet, metabolic abnormalities, and genetics [99].In recent years, the number of obese people has been on the rise, and obesity is also closely related to the occurrence of certain cancers [100,101].In a retrospective cohort study of 22,198 patients who underwent bariatric surgery between 2005 and 2012, patients who underwent bariatric surgery had an overall reduction in cancer incidence over 3.5 years of follow-up compared with matched participants who did not undergo surgery (HR = 0.67, 95% C.I. = 0.60-0.74,p < 0.001) [102].Studies have indicated that excessive obesity can increase the risk of colon cancer and breast cancer, and obesity can promote tumorigenesis by stimulating inflammatory responses in the body [103].Dark tea has been reported to lower lipids, which provides another perspective for the prevention and treatment of cancer [104,105].
Anti-Obesity
Obesity has become a widespread problem [98].Various variables contribute to its occurrence, such as diet, metabolic abnormalities, and genetics [99].In recent years, the number of obese people has been on the rise, and obesity is also closely related to the occurrence of certain cancers [100,101].In a retrospective cohort study of 22,198 patients who underwent bariatric surgery between 2005 and 2012, patients who underwent bariatric surgery had an overall reduction in cancer incidence over 3.5 years of follow-up compared with matched participants who did not undergo surgery (HR = 0.67, 95% C.I. = 0.60-0.74,p < 0.001) [102].Studies have indicated that excessive obesity can increase the risk of colon cancer and breast cancer, and obesity can promote tumorigenesis by stimulating inflammatory responses in the body [103].Dark tea has been reported to lower lipids, which provides another perspective for the prevention and treatment of cancer [104,105].
Fu Brick Tea (200 mg/kg BW for 4 weeks) can reduce the levels of TG and TC in the serum of mice and alleviate the occurrence of alcohol-induced hepatic steatosis [106].In the study of Pu'er tea, it was found to significantly reduce serum TG, TC, and LDL-C levels and increase NO content in mice fed a high-fat diet [107].Similar conclusions were also verified in the zebrafish model, where different structures of TB from Pu'er tea were found to have lipid-lowering effects on high-fat-induced zebrafish.TB (1000 µg/mL) was able to reduce lipid levels in high-fat zebrafish to 51.57%, better than positive control simvastatin (0.06 µM) [108].Ma et al. found that Pu'er tea can combine with bile salts to expel cholesterol from the body and stimulate the continuous conversion of cholesterol into bile salts to achieve a lipid-lowering effect [109].Most studies on the lipid reduction of Pu'er tea have been conducted from the perspective of the regulation of obesity-related gene expression, such as farnesol X receptor (FXR), which is involved in the control of bile acid (BA) synthesis and hepato-intestinal circulation, and the activation of liver FXR and the inhibition of intestinal FXR are beneficial to obesity-related metabolic diseases [110].Liupao tea (200 mg/kg BW for 8 weeks) can also reduce the levels of TG, TC, and LDL-C in the serum of obese mice and increase the level of HDL-C.The expression levels of PPAR-α, LPL, CPT1, and CYP7A1 were up-regulated, while the protein expression levels of PPAR-γ were down-regulated [111].Xu et al. found that TB and tea polysaccharides Fu Brick Tea (200 mg/kg BW for 4 weeks) can reduce the levels of TG and TC in the serum of mice and alleviate the occurrence of alcohol-induced hepatic steatosis [106].In the study of Pu'er tea, it was found to significantly reduce serum TG, TC, and LDL-C levels and increase NO content in mice fed a high-fat diet [107].Similar conclusions were also verified in the zebrafish model, where different structures of TB from Pu'er tea were found to have lipid-lowering effects on high-fat-induced zebrafish.TB (1000 µg/mL) was able to reduce lipid levels in high-fat zebrafish to 51.57%, better than positive control simvastatin (0.06 µM) [108].Ma et al. found that Pu'er tea can combine with bile salts to expel cholesterol from the body and stimulate the continuous conversion of cholesterol into bile salts to achieve a lipid-lowering effect [109].Most studies on the lipid reduction of Pu'er tea have been conducted from the perspective of the regulation of obesity-related gene expression, such as farnesol X receptor (FXR), which is involved in the control of bile acid (BA) synthesis and hepato-intestinal circulation, and the activation of liver FXR and the inhibition of intestinal FXR are beneficial to obesity-related metabolic diseases [110].Liupao tea (200 mg/kg BW for 8 weeks) can also reduce the levels of TG, TC, and LDL-C in the serum of obese mice and increase the level of HDL-C.The expression levels of PPAR-α, LPL, CPT1, and CYP7A1 were up-regulated, while the protein expression levels of PPAR-γ were down-regulated [111].Xu et al. found that TB and tea polysaccharides inhibited lipase [112].TB extracted from Fuzhuan tea can reduce the weight and white adipose tissue weight of HFD mice [113].Fuzhuan tea and Ya'an Tibetan tea can improve obesity, prevent dyslipidemia, and reduce weight gain [114,115].Lv et al. used different substances to extract the active components from Pu'er tea and found that flavonoids in Pu'er tea contributed the most to lipid lowering through high-throughput screening of cell models PPARδ, PPARγ, FXR, and LXR [116].The view that polyphenols in Pu'er tea inhibit cholesterol synthesis in HepG2 cells was confirmed by Lu et al. [117].Yue et al. found that Pu'er tea could inhibit the activity of pancreatic lipase in high-fat mice, down-regulating SREBP-1c and FAS and enhancing LDLR and CPT-1α to promote fat consumption [118].Similarly, Ye et al. found that Pu'er tea also down-regulates the expression of LXRα, FAS, SREBP-1c, and PPARγ [119].
In addition to the mouse model, rat, rabbit, and Caenorhabditis elegans models were also used.Liang et al. found through rats and male rabbits that Pu'er tea can mediate lipid metabolism by inhibiting key enzyme activities, and Pu'er tea down-regulates Lp-PLA2, HMGR, and PL and up-regulates LCAT activity [120].Huang et al. found that Pu'er tea decreased FAS expression and increased AMPK phosphorylation to improve fructose-induced hyperlipidemia [121].LO2 cells and Caenorhabditis elegans were studied by Su et al., and it was found that Pu'er tea could relieve the expression of lipid accumulation and metabolism transcription factors such as PPARα, CD36, Plin2, Scd1, activate the SIRT1-FOXO pathway to inhibit the expression of SREBP-1c and FAS, and inhibit systemic lipidation in Caenorhabditis elegans [122].Cao et al. focused their research on preadipocytes and found that Pu'er tea inhibited the proliferation and differentiation of 3T3-L1 preadipocytes because the transcription factors peroxisome proliferator-activated receptor-γ and CCAAT/enhancer binding protein-α were down-regulated during differentiation [123].Moreover, the lipidlowering effect of Pu'er tea is also closely related to intestinal flora, which was verified by Kuang et al. [124].Pu'er tea was orally administered to obese mice for 9 weeks at a dose of 1000 mg/kg BW (equivalent to drinking 13-15 g tea per day for adults), and atorvastatin was administered at a dose of 10 mg/kg BW per day to inhibit obesity [53].
Dark tea has also been shown to inhibit obesity in human studies.Pu'er tea extract (3 g/day) taken daily for 20 weeks can significantly reduce body weight and improve blood lipid status in people with hyperlipidemia [125].In healthy men aged 24-32 years, taking Pu'er tea powder (50 mg/kg) daily for 4 weeks can reduce hyperlipidemia [23].In addition, Pu'er tea extract (Pu'er tea polyphenol 32.48 mg, eq.gallic acid) can also reduce blood sugar levels after meals [126].In a multicenter, cross-sectional study, it was found that about 62% of Chinese people drink tea, mainly dark, black, and green tea.Daily tea consumption was inversely associated with the risk of diabetes in women, the elderly, and the obese.The prevalence of newly diagnosed diabetes in green, black, and dark tea drinkers was 9.8%, 5.8%, and 5%, respectively.Those who drank dark tea (Pu'er) had the lowest prevalence of newly diagnosed diabetes [127].
In general, dark tea has a lipid-lowering effect, which can reduce the production of pro-inflammatory factors in obese people, inhibit the growth and spread of tumors, and reduce the risk of cancer.Possible pathways are shown in Figure 4.It can also reduce the risk of cancer by reducing oxidative stress and DNA damage.The specific mechanism by which obesity reduces the incidence of cancer needs further study.
Regulating Intestinal Flora
It is shown that the imbalance of ecological disruption and the subsequent gut microbes in the intestine may lead to a variety of pathological changes and also lead to the occurrence of cancer [128,129].While dark tea has been shown to modulate gut microbiota [130,131].
The polyphenols in dark tea can regulate the intestinal microbiota in animal and human experiments, it has been confirmed by many people that tea can increase the α and β diversity of the gut microbiota in animals [132].Gao et al. found that active ingredients such as polyphenols and caffeine in Pu'er tea increased the abundance of Akkermansia muciniphila and Faecalibacterium prausnitzii in mice on a high-fat diet, thereby improving obesity and inflammation [133].Additionally, TB plays an important role in regulating the intestinal flora.Deng et al. found that TB can promote the proliferation of beneficial microbiomes, such as Lactobacillus and Lachnospiraceae_NK4A136_group [134].Yue et al. found that Pu'er tea increased the composition of Actinobacteria and Proteobacteria in rats with a high-fat diet, and the effect of TB on diabetic rats mainly depended on the targeted regulation of intestinal microorganisms [135].Yue further induced rats with high sugar, high fat, and high salt and found that the prognosis of TB and the ratio of Firmicutes to Bacteroidetes (F/B) decreased significantly, which promoted the propagation of Bacteroidetes and inhibited Firmicutes [136].Su et al. found that Pu'er tea can regulate the gut microbes in colitis mice and increase the abundance of beneficial bacteria, such as Muribaculum, Lactobacillus, Rikenellaceae, and Lachnospiraceae; the abundance of harmful bacteria, Romboutsia and Turicibacter, decreased, which provided a safe and effective new strategy for the prevention and treatment of colitis and reduced the probability of colon cancer [137].In addition, Huang et al. found that, in addition to regulating the composition and proportion of intestinal microbes, SCFAs produced by intestinal microbiota also played an important role [138].
Regulating Intestinal Flora
It is shown that the imbalance of ecological disruption and the subsequent gut microbes in the intestine may lead to a variety of pathological changes and also lead to the occurrence of cancer [128,129].While dark tea has been shown to modulate gut microbiota [130,131].
The polyphenols in dark tea can regulate the intestinal microbiota in animal and human experiments, it has been confirmed by many people that tea can increase the α and β diversity of the gut microbiota in animals [132].Gao et al. found that active ingredients such as polyphenols and caffeine in Pu'er tea increased the abundance of Akkermansia muciniphila and Faecalibacterium prausnitzii in mice on a high-fat diet, thereby improving obesity and inflammation [133].Additionally, TB plays an important role in regulating the intestinal flora.Deng et al. found that TB can promote the proliferation of beneficial microbiomes, such as Lactobacillus and Lachnospiraceae_NK4A136_group [134].Yue et al. found that Pu'er tea increased the composition of Actinobacteria and Proteobacteria in rats with a high-fat diet, and the effect of TB on diabetic rats mainly depended on the targeted regulation of intestinal microorganisms [135].Yue further induced rats with high sugar, high fat, and high salt and found that the prognosis of TB and the ratio of Firmicutes to Bacteroidetes (F/B) decreased significantly, which promoted the propagation of Bacteroidetes and inhibited Firmicutes [136].Su et al. found that Pu'er tea can regulate the gut microbes in colitis mice and increase the abundance of beneficial bacteria, such as Muribaculum, Lactobacillus, Rikenellaceae, and Lachnospiraceae; the abundance of harmful bacteria, Romboutsia and Turicibacter, decreased, which provided a safe and effective new strategy for the prevention and treatment of colitis and reduced the probability of colon cancer [137].In addition, Huang et al. found that, in addition to regulating the composition and proportion of intestinal microbes, SCFAs produced by intestinal microbiota also played an important role [138].Whether there are individual differences in the regulation of intestinal flora and the specific mechanism of action between the active components of dark tea extract and intestinal flora needs to be further explored.The content and molecular weight of uronic acid may be the important factors affecting the oxidation resistance of DTPS [71] ROS, MDA determination Gamma-aminobutyric acid (GABA) in Pu'er tea Pu'er tea extract (1, 10 µ g/mL) and GABA (0.1, 1, 10 µM) decreased ROS production and lipid peroxidation in PC12 cells in a dose-dependent manner [73]
8-C N-ethyl-2-pyrrolidinone substituted flavan-3-ols
The 8-C N-ethyl-2-pyrrolidinone substituted flavan-3-ols possessed significant antioxidant activity and could prevent HMEC damage caused by H 2 O 2 [76] Mice exposed to 7.0 and 7.5 Gy total body irradiation Dark tea extract Dark tea extract reduced ROS levels in hematopoietic cells by inhibiting the expression of NOX4 [77] Obese rat model Pu'er tea Pu'er tea increased the activity of antioxidant enzymes such as SOD and GSH-Px, while decreased the level of lipid peroxidation product MDA in obese rats [78,79] t-BHP oxidative stress-induced rat hepatocyte model Pu'er tea Pu'er tea extract decreased the production of ROS marker O 2 − in rat hepatocytes and prevented t-BHP induced mitochondrial oxidative stress [80] SD rats Balb/c mice Pu'er tea Pu'er tea reduced quinocetone-induced oxidative stress [81,82] Wistar rats Pu'er tea Pu'er tea decreased MDA and GSH, and increased SOD and GSH-Px levels [83] Mice with acute alcoholic liver injury
Different dark tea extracts
Dark tea has greater in vivo antioxidant activity than green tea [84]
Expectation and Prospect
To sum up, the fact that dark tea can prevent and treat cancer has been proven by many studies, but there are still problems to be solved and improved.There are many studies on colitis and colon cancer in dark tea, and more research on the prevention and treatment of cancer in other parts of the body should be carried out in the future.At the same time, most of the studies focus on the apparent but fail to deeply explore the pathways of cancer prevention and treatment, and there is a lack of holistic connection between each anti-cancer pathway.
In a prospective study of 532,949 participants, higher levels of tea drinking were associated with a lower risk of bladder cancer (compared with no tea consumption: HR = 0.87, 95% C.I. = 0.77-0.98 for low consumption; HR = 0.86, 95% C.I. = 0.77-0.96for moderate consumption; HR = 0.84, 95% C.I. = 0.75-0.95for high consumption) [139].In the European Prospective Investigation into Cancer and Nutrition, 201 cases of hepatocellular carcinoma were identified among 486,799 men/women followed for 11 years, and it was found that increased tea intake was associated with a lower risk of hepatocellular carcinoma [140].Studies have found that drinking three cups of fermented dark tea a day can reduce the risk of coronary heart disease and diabetes [141].And there have been many prospective studies on the anti-cancer effects of black tea and green tea.Increasing the intake of green tea and black tea can reduce the risk of lung cancer, gynecological cancer, bladder cancer, etc., and its anti-cancer effect is mainly attributed to tea polyphenols, such as epigallocatechin-3gallate and theaflavins [142][143][144][145].However, there are few prospective studies focusing on the anti-cancer effect of dark tea, which can be increased in the future to make dark tea more convincing in preventing and managing cancer.
Conclusions
Cancer has become the primary problem threatening human health, and the number of cancer patients is also increasing rapidly [146].Therefore, the prevention and treatment of cancer have become two of the most promising directions.At present, the main treatment for cancer is radiotherapy and chemotherapy [147,148], but these methods are harmful to the human body [149], and it is more realistic and reliable to prevent cancer through daily dietary intake.
This paper reviewed the effect of dark tea on cancer prevention and treatment and established the interaction and corresponding mechanism.The summary shows that the main mechanisms of dark tea preventing cancer are anti-inflammatory, antioxidant, inhibiting cancer cell proliferation, inducing cancer cell apoptosis, anti-obesity, and regulating intestinal flora.Daily consumption of dark tea can achieve cancer prevention by inhibiting pro-inflammatory cytokines TNF-α, IL-1β, and IL-6 and increasing anti-inflammatory cytokines IL-10 and IL-22, which is mainly attributed to the down-regulation of the NF-κB signaling pathway.The treatment of dark tea can also effectively prevent cancer by increasing the activity of the antioxidant enzymes SOD, CAT, and GSH-Px, removing ROS produced by cells, and decreasing the level of MDA.The antioxidant pathway is mainly attributed to increasing the level of ERK phosphorylation and increasing the expression of the Nrf2/HO-1 signaling pathway.Dark tea's inhibition of cancer cell proliferation and induction of apoptosis have also been verified by many studies, which are mainly mediated by the MAPK/JNK signaling pathway.Dark tea also reduces the levels of serum TC, TG, and LDL-C, mainly inhibits the synthesis of fatty acids and cholesterol in the body, and promotes the realization of excretion, thus reducing obesity-related metabolic diseases and preventing cancer.At the same time, dark tea can also regulate intestinal flora, improve the diversity of intestinal microbes α and β, promote the reproduction of bacteroides, and inhibit firmicutes to effectively prevent cancer.
Theabrownin, tea polyphenol, and tea polysaccharide are the main active substances in dark tea.Although a large number of in vitro and in vivo studies have confirmed the positive effect of dark tea extract on human health, complete research on the exact molecular mechanisms of various active ingredients related to their corresponding anti-cancer abilities is still lacking.In addition, as is known to all, the bioavailability of dark tea in the human body is poor [150,151], and there are few studies on the anti-cancer effects of specific active substances in dark tea.Whether the anti-cancer effect of dark tea is more effective with a single component or the synergistic effect of different components needs to be further confirmed.It is important to note that dark tea, a natural food, can be highly recommended as a nutritional supplement for cancer prevention and management with few side effects.
Figure 1 .
Figure 1.Possible anti-cancer mechanisms of dark tea.
Figure 1 .
Figure 1.Possible anti-cancer mechanisms of dark tea.
Figure 2 .
Figure 2. The anti-inflammatory pathway of dark tea.
Figure 2 .
Figure 2. The anti-inflammatory pathway of dark tea.
Figure 4 .
Figure 4. Dark tea reduces the production of pro-inflammatory factors by inhibiting obesity.
Figure 4 .
Figure 4. Dark tea reduces the production of pro-inflammatory factors by inhibiting obesity.
and Table 1 below.
Table 2 .
Antioxidation of dark tea and its pathways. | 2023-09-10T15:39:10.726Z | 2023-09-01T00:00:00.000 | {
"year": 2023,
"sha1": "ffcedf72e0db2de306a835fd3234a13543a9a8f7",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2072-6643/15/18/3903/pdf?version=1694087900",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "70ef4b8340a9791009ad8b6142418afefbe64a1f",
"s2fieldsofstudy": [
"Agricultural And Food Sciences"
],
"extfieldsofstudy": [
"Medicine"
]
} |
262141406 | pes2o/s2orc | v3-fos-license | Hedgehog signaling is required for endometrial remodeling and myometrial homeostasis in the cycling mouse uterus
Summary Decades of work demonstrate that the mammalian estrous cycle is controlled by cycling steroid hormones. However, the signaling mechanisms that act downstream, linking hormonal action to the physical remodeling of the cycling uterus, remain unclear. To address this issue, we analyzed gene expression at all stages of the mouse estrous cycle. Strikingly, we found that several genetic programs well-known to control tissue morphogenesis in developing embryos displayed cyclical patterns of expression. We find that most of the genetic architectures of Hedgehog signaling (ligands, receptors, effectors, and transcription factors) are transcribed cyclically in the uterus, and that conditional disruption of the Hedgehog receptor smoothened not only elicits a failure of normal cyclical thickening of the endometrial lining but also induces aberrant deformation of the uterine smooth muscle. Together, our data shed light on the mechanisms underlying normal uterine remodeling specifically and cyclical gene expression generally.
Highlights Hedgehog signaling components are similarly enriched across the estrous cycle
Smoothened is required for uterine endometrial remodeling Smoothened is required for uterine myometrial homeostasis
INTRODUCTION
Repeated cycles of gene expression are central to diverse aspects of biology, from the Period and Cryptochrome (PER/CYC)-mediated cycles that drive circadian rhythm 1 to the Notch-mediated clock and wavefront that generates the blocks of embryonic somites. 2 It is interesting then that while decades of research have defined the crucial roles of steroid hormone cycles in control of the mammalian estrous cycle, 3,4 far less is known of the cyclical patterns of gene expression that link those hormones to cyclical changes in the structure of the reproductive tract.
Indeed, the uterus is unlike other adult tissues as it repeatedly undergoes extensive remodeling during the estrous cycle, implantation, and pregnancy.In some mammals, such as humans, the uterine lining is cyclically shed in the process of menstruation.However, most mammals do not menstruate and instead undergo a reiterative process of tissue remodeling across the estrous cycle.In the mouse, the endometrial lining of the uterus thickens and elaborates when estrogen increases during the first half of the estrous cycle, and this lining dramatically thins and simplifies in the second half, when progesterone increases. 3,4Over this time frame, new studies have shown that luminal epithelium turns over 5 and the ratio between stromal and epithelial cells shifts dramatically. 6However, the gene regulatory mechanisms that act downstream of cycling hormone signals to drive the tissue behaviors underlying cyclical uterine remodeling are not known.
For example, endometrial Indian hedgehog (Ihh) is a direct target of progesterone and drives remodeling during pregnancy and implantation in the rodent, [7][8][9][10] and others have shown that Ihh is cyclically controlled during the menstrual 11 and estrous cycle. 12,13But the function of Hedgehog signaling in the non-pregnant, cycling uterus is unknown, and previous bulk RNA sequencing (RNA-seq) studies 12,13 have focused on only two stages.Moreover, while these genomic datasets of the cycling endometrium in humans [14][15][16][17][18] and rodents 6 have begun to illuminate the transcriptome of the non-pregnant uterus, our knowledge of how downstream signaling pathways contribute to cyclical remodeling and homeostasis is sparse.Given the power of mouse genetics to explore mammalian biology, this gap in knowledge presents a key hurdle.
Here, we fill that gap by determining the transcriptomic landscape of the cycling adult mouse uterus at every stage of the estrous cycle.Interestingly, these data suggest that uterine remodeling requires robust re-use of genetic systems more commonly associated with embryonic development, including Hox family transcription factors and Wnt, transforming growth factor b (TGF-b), and Hedgehog signaling.For the latter case, we show that the majority of the signaling pathway undergoes robust cyclical transcription and that loss of Hh signaling not only severely perturbs normal cyclical remodeling of the uterus but also induces an aberrant remodeling of the smooth muscle tissue in the myometrium.Our work on the normal estrous cycle may shed light on other aspects of myometrial biology, like pregnancy and parturition, or the progression of uterine smooth muscle tumors.
Quantification of uterine remodeling across the mouse estrous cycle
We sought to understand the molecular/genetic changes that accompany the iterative remodeling of the normal mouse uterus, so we first extended previous studies 3,4 by developing a comprehensive set of morphometrics to quantify normal uterine morphology across the estrous cycle.We therefore visualized gross tissue morphology of uteri at each estrous cycle stage using H&E staining (Figures 1A-1D, n = 8 per stage) and independently quantified diverse uterine compartments.
The total uterus area was thickest at proestrus and then steadily decreased until diestrus (Figure 1E), and the area of the lumen followed a similar pattern (Figure 1F).The endometrium was also thickest at proestrus and thinnest at diestrus (Figure 1G).We also quantified the number of uterine glands and the luminal epithelial cell height in the endometrium across the estrous cycle; both were highest at proestrus and lowest at diestrus (Figures 1I and 1J).By contrast, while the myometrium displays cyclical spontaneous contractile activity during the estrous cycle, 19 overall myometrial area was not significantly altered across the estrous cycle (Figure 1H).Thus, the normal cycling endometrium displays several cyclical changes in gross morphology across the estrous cycle while the myometrium does not.
The murine uterus is transcriptionally dynamic across the estrous cycle
We hypothesized that downstream of steroid hormones, a variety of signaling pathways might be utilized in the adult uterus to regulate cyclical tissue remodeling.To test this hypothesis, we performed 3 0 TagSeq of the whole uterus at each stage of the estrous cycle (n = 3 per stage).Principal-component analysis (PCA) of three biological replicates per estrous cycle stage showed clear separation of estrous cycle replicates, with PC1 and PC2 accounting for 49% and 22% of the variance, respectively (Figure 1K).While most samples within each stage cluster well together, we noted that the metestrus samples (in yellow) do not, which we expect is due to either biological variability or inaccurate estrous stage prediction.Regardless, the stages clearly form a cycle within the PCA, indicating that the stages are transcriptionally distinct (Figure 1K).
To better visualize the cyclical changes in DEGs, we plotted the top 15 enriched DEGs of each cycle stage in a heatmap, showing a cascade of gene expression during the estrous cycle (Figure 1P).This highly dynamic transcriptional landscape of the cycling uterus provides a rich dataset with which to explore the molecular mechanics that may link hormone cycles to cyclical morphogenesis in the uterus.
Genes associated with embryonic development are cyclically expressed in the adult mouse uterus during the estrous cycle
We were curious if the DEGs at each cycle transition were enriched in particular gene classes, so we performed Gene Ontology (GO) analysis. 20Curiously, developmental biology GO terms were prevalent at metestrus.The top GO terms included limb development; appendage development; limb morphogenesis; appendage morphogenesis; tube development; and embryonic limb morphogenesis (Figure 2C).At proestrus, the top GO terms included sensory perception of smell; sterol biosynthetic process; and epithelium development (Figure 2A).At estrus, the top GO terms included response to external stimulus and response to other organism (Figure 2B).
Using the GO term analysis as a partial guide, we manually curated the developmental genes and signaling pathways that were differentially expressed in the uterus across the estrous cycle and assembled heatmaps and a table (Table S2) showing differential gene expression.For example, Hox family genes that have known roles in implantation [21][22][23][24] such as Hoxa10, Hoxa11, Hoxd11, and Msx1 were all upregulated when progesterone is high during metestrus and diestrus (Figure 2D; Table S2).Conversely, HoxB4 is known to be expressed in uterine glandular epithelial cells, 25 and we found Hoxb4 to be cyclically expressed with higher expression in early stages when gland numbers peak (Figures 1I and 2D; Table S2).
Likewise, dynamic expression of Wnt ligands has been suggested during the estrous cycle, 26 with Wnt7b being required for uterine gland development. 27,28Notably, we found Wnt7b also cycled with higher expression during the early stages (Figure 2E; Table S2), while Wnt11 and several Wnt signaling transducers displayed the converse pattern, peaking late in the cycle (Figure 2E; Table S2).Finally, TGF-b signaling is also required for implantation and uterine gland and myometrial development, 29,30 but its role during the estrous cycle is unknown.In our dataset, we identified differential expression of genes involved in TGF-b signaling: Tgfbi is enriched at metestrus and diestrus, while Lrg1 is enriched at estrus (Figure 2F; Table S2).Thus, the expression of many of genes required for either implantation or uterine development are dynamically expressed across the estrous cycle, though the relationship between this cyclical expression and cyclical uterine remodeling has yet to be determined.
Hedgehog signaling components are cyclically co-regulated across the estrous cycle
Our most striking finding was that the genes encoding nearly all members of Hedgehog signaling pathway (Figure 3A) were cyclically expressed together and enriched at diestrus (Figures 3B-3M; Table S2).The major Hh ligand in the uterus, Ihh, and the known downstream target genes of the pathway, such as Gli1 and Ptch1, are significantly enriched (denoted with an asterisk) (Figure 3B; Table S2).Given that Ihh is a direct target of progesterone at implantation, 8,10,31 enrichment of Ihh and downstream target genes Gli1 and Ptch1 at metestrus or diestrus when progesterone is high was expected. 4However, the remaining signaling components, including Gli2, Gli3, Ptch2, Smo, Sufu, Gpr161, Kif7, Evc, and Evc2 are not typical downstream targets of Hh signaling, yet they also trend upward or are significantly enriched (denoted with an asterisk) at diestrus (Figure 3B; Table S2).We confirmed the significant enrichment of most of these genes by qPCR on an additional 5 animals for a total of n = 8 (Figures 3C-3M).When evaluated with n = 8, some of the pathway components differ from the RNA-seq dataset (Figures 3E-G, 3I, and 3L).Thus, uterine expression of the genes encoding most of the Hh signaling are co-regulated across the mouse estrous cycle.
Hedgehog signaling is required for iterative uterine remodeling across the estrous cycle
While Hedgehog signaling has been extensively studied at implantation, specifically in the endometrium, less is known about possible roles for Hedgehog signaling in uterine remodeling or homeostasis across the normal estrous cycle.We therefore made a conditional knockout of the major activator, smoothened (Smo), using the progesterone receptor cre (PR-Cre, Figure 4A) which is expressed throughout the adult uterus. 32,33The controls (ctrl) were PR-Cre +/À or À/À Smo fl/+ or +/+ , and the conditional knockouts (Smo cKO) were PR-Cre +/À Smo fl/fl .We confirmed that Smo is indeed significantly decreased or absent in the cKOs compared to the controls by qPCR (Figure 4B).Smo is the major activator of the Hh signaling pathway; therefore, the Smo cKOs should not be able to activate downstream Hh signaling target genes.We confirmed this by checking for the expression of Gli1, a downstream target gene of the pathway.Indeed, while Gli1 is upregulated at diestrus compared to estrus in our controls, that fails to occur in the cKOs (Figure 4C).5][36] To confirm that our cKO is infertile, we performed breeding studies on the controls and Smo cKOs.The control mice had an average of 9 pups per litter, while the Smo cKOs failed to ever have pups (Figure 4D).We also confirmed that the Smo cKO animals cycled normally compared to controls using three methods.With vaginal cytology, we found that the cKOs progress through the estrous cycle relatively normally, albeit with a significantly shorter diestrus stage (Figure 4E).We also looked at levels of progesterone in the serum with the University of Virginia Ligand Assay and Analysis Core and showed no significant difference between controls and cKOs (Figure S1A).Finally, we show that both controls and Smo cKO ovaries have corpus lutea at diestrus, suggesting that ovulation is occurring in all animals (Figure S1B).Despite this relatively normal cycling, uteri lacking Smo displayed radically altered morphologies (Figures 4F-4I).Smo cKO uteri failed to thicken during the estrogen-dominant early stages of the cycle (i.e., estrus).Measurement of both the total uterine area and the endometrial area confirmed this result as both were significantly reduced in the mutants (Figures 4F, 4G, 4J, and 4K).Thus, Hedgehog signaling is required for normal morphogenetic changes in the uterine endometrium across the mouse estrus cycle.
Hedgehog signaling controls homeostasis of longitudinal smooth muscle in the uterus
In contrast to the obvious endometrial phenotype at estrus, Smo cKO mice displayed a more subtle but interesting phenotype during the progesterone-dominant later stages (i.e., diestrus), when the endometrium is normally quite thin (see Figure 1).Specifically, the size of the myometrium, composed of inner circular and outer longitudinal layers of muscle, 37 was increased relative to the endometrium in Smo S2.Each column represents one sample, and all heatmaps use z-scores of the DESeq2 variance stabilized counts.cKO mice (Figure 4L).Moreover, the appearance of the tissue was altered, seemingly thrown into aberrant deformations not observed in control uteri (Figure 5).This stage-specific change in the myometrium was surprising as the myometrial area does not remodel across the cycle in normal mice (see Figure 1).This result prompted us to further quantify these myometrial phenotypes.We found that while the Smo cKO longitudinal smooth muscle layer trended toward increased thickness at diestrus, it did not reach statistical significance (Figure 5A).However, when we quantified deformation using ''smoothness'' (i.e., ratio of uterine perimeter length [Figure 5B') to circumference of the best fit ellipse [Figure 5B'']), we observed a significant decrease specifically at diestrus in the Smo cKO (Figures 5C-5F).These data suggest that in the absence of Hh signaling the mouse uterus undergoes aberrant deformation of the outer muscular layer.
Muscle fibers are larger in the absence of Smoothened
The aberrant tissue deformation observed in the myometrium of Smo cKO at diestrus is reminiscent of normal folding observed in some developing organs, such as the exterior of some mammalian brains.There is evidence that such folding results from differential growth of the outer gray matter and the inner white matter, resulting in mechanical buckling. 38The inverse is true in the developing vertebrate gut, where the growing inner endoderm and mesenchyme are restrained by the outer layers of smooth muscle resulting in folding. 39We hypothesized that the aberrant deformation of uterine smooth muscles in the Smo cKO at diestrus may be caused by an increase in longitudinal muscle fiber number or size compared to the relatively unchanging endometrial area (Figure 4L).
We therefore quantified the size and number of muscle fibers using Masson Trichrome histology, which stains muscle in red and collagen in blue (histology in Figures 5 and 6).Given that uterine smooth muscle fibers are embedded in extracellular matrix, 40 we were able to assess smooth muscle fibers using Ilastik, which takes advantage of color and texture differences to differentiate structures, 41 to produce masks of S2.
(C-M) qPCR of the same core components of Hh signaling with n = 8 samples.See also Table S1.t tests were performed, where * = p < 0.05; n.s.= not significant.
muscle fibers (Figures 6A-6D and 6A''-6D 00 ).While the average number of muscle fibers was unchanged, we observed a striking and significant increase in the size of individual muscle fibers in Smo cKO uteri, specifically at diestrus (Figures 6E and 6F).A more granular view of the data revealed a curious pattern to this effect.When we binned fibers by area, we found that the Smo cKO phenotype was driven by changes in fibers at the extremes of size (Figure 6G).These data suggest that changes in muscle fiber size may underlie the aberrant deformation of the uterine myometrium that accompanies loss of Hedgehog signaling in the cycling adult mouse uterus.
DISCUSSION
One of the first studies to show that the human uterus undergoes cyclical remodeling (rather than pathological changes) during the menstrual cycle was published in 1908. 424][45][46][47][48] However, it would be nearly 100 years before a similar study on mouse uterine remodeling relative to circulating steroid hormone levels was published, 4 despite this animal's power and widespread use as a tractable genetic model for mammalian biology.5][56][57][58] By contrast, the normally cycling mouse uterus remains far less understood.Several studies have assessed morphological changes across the estrous cycle, 3,4 and others have compared the transcriptomes of two cycle stages. 12,13owever, as Wood et al. remarked, several parameters in their study showed correlations with specific estrous cycle stages. 4Therefore, we wanted to expand the transcriptional landscape of the cycling uterus.
We determined the cyclical transcriptome of the whole uterus across the four key stages of the estrous cycle.We found that hundreds of genes are differentially expressed at each cycle stage comparison, demonstrating that the mouse uterus is highly transcriptionally dynamic across the relative short span of 24 h, which roughly characterized each stage (Figure 1).This result contrasts sharply with the mouse oviduct, which we recently showed undergoes minimal transcriptional change across the estrous cycle. 59mong the most interesting results of this analysis was that many developmental pathways are differentially expressed across the estrous cycle, including Hox genes, Wnt signaling, and TGF-b signaling (Figure 2) and Indian Hedgehog signaling (Figure 3).Ihh is a direct target of progesterone signaling at implantation 8,10 and has also been identified as differentially expressed between proestrus and estrus, 12 and estrus and diestrus. 13Therefore, Ihh itself was a positive control in our dataset.However, it was striking to us that most of the system, including receptors and transducers, displayed a clear cycle of expression peaking at diestrus (Figure 3).And while Hh signaling is required for uterine development 60 and implantation, 10 a role in the cycling uterus has not been described.
We took a conditional genetic approach, which revealed a key role for Hh signaling in uterine remodeling and homeostasis across the normal estrous cycle.Conditional deletion of smoothened with the progesterone receptor Cre resulted in a failure of endometrial expansion during estrus (Figure 4).We found it curious that the upregulation in gene expression was at diestrus and yet the endometrial phenotype was at estrus.We speculate that progesterone-regulated Hh signaling primes the endometrium for the remodeling that occurs under estrogen.In support of this speculation, it has recently been shown in human endometrial biopsies that the progesterone receptor binds the Ihh promoter during the proliferative phase of the menstrual cycle, when estrogen is high. 14Finally, while the underlying basis for this phenotype remains to be discovered, it is of special interest because Ihh signaling is required for endometrial stromal cell proliferation to support implantation in the pregnant uterus. 10n addition to this failure of endometrial remodeling, loss of Smo also severely disrupted the homeostasis of the myometrium.This muscular layer normally changes very little across the cycle (Figure 1), but in the absence of Hh signaling, we observed aberrant deformation accompanied by an increase in muscle fiber size (Figures 5 and 6).2][63][64][65][66][67] Moreover, the deformed appearance of the myometrium of Smo cKO mice at diestrus is remarkably similar to that observed in H&E-stained post-partum mouse uteri which are undergoing involution, or the healing process after birth by which the myometrium remodels to return to a non-pregnant state. 68Our hypothesis moving forward is that Hh signaling in the myometrium is downstream of progesterone signaling (similar to the endometrium at implantation) and is required for myometrial quiescence.0][71] In the future, we will investigate the role of Hh signaling in myometrial quiescence and remodeling that occurs during late pregnancy or involution.Finally, our data provide another lens through which to view the general phenomenon of cyclical patterns of gene expression.Such cycles are most well defined in the context of circadian rhythm 1 but control diverse biological processes in adults and in embryos.Interestingly, Hedgehog signaling has been implicated in the hair growth cycle in adult mammals 72,73 and many Hh-related genes display long (seasonal) cycles of gene expression. 74,75On shorter time scales, elements of the Hh signaling pathway also display daily cycles of expression in the adult mouse liver that are linked to circadian rhythm. 76Our data therefore shed new light on important aspects of the reproductive cell biology of uterine remodeling and also more generally on the issue of cyclical control of gene expression.
Limitations of the study
There are several caveats to the work presented here.The PGR-Cre line is expressed in additional tissues including the ovary, oviduct, hypothalamus, and pituitary.While we cannot rule out their involvement in our phenotype, the smo cKO mice do cycle normally based on vaginal cytology which indicates that estrogen and progesterone are being cyclically secreted, progesterone is secreted at concentrations similar to controls, and the smo cKO ovary shows corpus lutea similar to controls demonstrating that ovulation is occuring.We performed transverse sectioning instead of sagittal sectioning.This precludes us from assessing the entire length of the uterus.To overcome the limitations of this, we quantified at least four sections from over 500 mm along the length of the uterus.Finally, we performed bulk RNA-seq on the whole uterus.The main limitation here is that we do not have single-cell resolution and have missed potential gene expression changes (i.e., if one cell type downregulates a signaling pathway, but another cell type upregulates it, we have likely not identified either change in our dataset).
Figure 1 .
Figure 1.Cyclical mouse uterine remodeling occurs concomitant with transcriptional changes (A-J) H&E staining of mouse uterine sections at (A) proestrus, (B) estrus, (C) metestrus, and (D) diestrus where the dotted white line denotes boundary between the endometrium and myometrium.Scale bar = 250mm.Quantitation of (E) uterine area, (F) lumen area, (G) endometrial area, (H) myometrial area, (I) uterine gland number, and (J) luminal epithelial height during those same time points.Each point on the graphs represents one uterus and is the mean quantification of 4 or more sections; all error bars are standard deviation of the mean.(K-O) (K) Principal-component analysis (PCA) plot of 3 0 TagSeq dataset.Volcano plots showing differential gene expression between (L) diestrus and proestrus, (M) proestrus and estrus, (N) estrus and metestrus, and (O) metestrus and diestrus.(P) Heatmap of the most variable differentially expressed genes at each estrous cycle stage, where each column represents one sample.All heatmaps use z-scores of the DESeq2 variance stabilized counts.t tests were performed, where * = p < 0.05; n.s.= not significant.
Figure 2 .
Figure 2. Development genes are cyclically expressed in the uterus during the estrous cycle (A-F) GO terms of differentially expressed genes enriched at (A) proestrus, (B) estrus, and (C) metestrus.Heatmaps of (D) Hox genes, (E) Wnt signaling, and (F) TGF-b signaling.See also TableS2.Each column represents one sample, and all heatmaps use z-scores of the DESeq2 variance stabilized counts.
Figure 3 .
Figure 3.The Hedgehog signaling pathway is cyclically regulated across the estrous cycle (A) The biochemical pathway of Hh signaling.(B) Heatmap of the core components of Hh signaling across the estrous cycle, where the Z score of the DESeq2 variance stabilized counts is used.Each column represents one sample.See also TableS2.(C-M) qPCR of the same core components of Hh signaling with n = 8 samples.See also TableS1.t tests were performed, where * = p < 0.05; n.s.= not significant.
Figure 4 .
Figure 4. Hedgehog signaling is required for cyclical uterine remodeling (A) Experimental flow to determine role of Hh signaling in the adult uterus during the estrous cycle.(B) qPCR for Smoothened in the controls and Smo cKOs.(C) qPCR for Gli1 in the controls and Smo cKOs.(D) Breeding trial of controls and Smo cKOs.(E) Estrous cyclicity of controls and Smo cKOs.(F-I) H&E-stained uterine sections from control and Smo cKOs.(J-L) Quantitation of (J) total uterine area, (K) endometrial area, and (L) myometrial-to-endometrial area ratio.Each point on the graphs represents one uterus and is the mean quantification of 4 or more sections.t tests were performed, where * = p < 0.05; n.s.= not significant.Scale in image = 250mm.All error bars are standard deviation of the mean.
Figure 5 .
Figure 5.The myometrium is significantly folded in the absence of Smoothened (A-E) (A) Quantitation of longitudinal muscle thickness.Each point on the graph represents one uterus and is the mean quantification 4 more sections.(B-E) Trichrome histology of the mouse uterus.(B 0 0 ) Outline of highlighting the ruffled nature of the Smo cKO.(B 00 -E 00 ) Best fit ellipse of the outline.(F) The smoothness of each sample was determined by taking the ratio of the true outline to the best fit ellipse.Each point on the graph represents one uterus and is the mean quantification of 4 or more sections.t tests were performed, where * = p < 0.05; n.s.= not significant.Scale bar = 250mm.All error bars are standard deviation of the mean.
Figure 6 .
Figure 6.Myometrial fibers are significantly bigger in the absence of Smoothened (A-D) Masks of the longitudinal muscle fibers were produced using Ilastik.(A 0 -D 0 ) Masson Trichrome histology was used to identify the myometrial muscle fibers.(A 00 -D 00 ) Longitudinal fibers were identified throughout the entire section.(E-G) (E) Quantitation of the fiber number in each sample.Each point on the graph represents one uterus and is the mean quantification of 4 or more sections.(F) Quantitation of the fiber area in each sample.Each point on the graph represents one uterus and is the mean quantification of 4 or more sections.(G) Fiber areas were binned in 1000mm 2 increments.t tests were performed, where * = p < 0.05 and n.s.= not significant.Scale in main image = 50mm.Scale in 'and ''images = 250mm.All error bars are standard deviation of the mean. | 2023-09-23T15:01:59.257Z | 2023-09-01T00:00:00.000 | {
"year": 2023,
"sha1": "401ed5c862dc03652e31bfa3cce13eddbe2d5ce1",
"oa_license": "CCBYNCND",
"oa_url": "https://doi.org/10.1016/j.isci.2023.107993",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "45b809cca65c68e2b851b8a35d20b587271d8894",
"s2fieldsofstudy": [
"Biology",
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
119648831 | pes2o/s2orc | v3-fos-license | Symmetries of handlebodies and their fixed points: Dihedral extended Schottky groups
A Schottky structure on a handlebody M of genus g is provided by a Schottky group of rank g. A symmetry (an orientation-reversing involution) of M is known to have at most (g+ 1) connected components of fixed points. Each of these components is either a point or a compact bordered surface (either orientable or not) whose boundary is contained in the border of M. In this paper, we derive sharp upper bounds for the total number of connected components of the sets of fixed points of given two or three symmetries of M. In order to obtain such an upper bound, we obtain a geometrical structure description of those extended Kleinian groups K containing a Schottky group Γ as finite index normal subgroup so that K/Γ is a dihedral group (called dihedral Schottky groups). Our upper bounds turn out to be different to the corresponding ones at the level of closed Riemann surfaces. In contrast to the case of Riemann surfaces, we observe that M cannot have two different maximal symmetries.
Introduction
A closed Riemann surface is symmetric if it has an anticonformal involution (called a symmetry). Under the well known equivalence between compact Riemann surfaces and smooth irreducible projective complex algebraic curves, the symmetric ones correspond to real algebraic curves, that is, curves which may be defined over the real numbers. Harnack's theorem [7] asserts that a symmetry of a closed Riemann surface of genus g has at most (g + 1) connected components of fixed points; each of these being a simple loop and called an oval (or mirror). Sharp upper bounds for the total number of ovals of given k ≥ 2 nonconjugated symmetries of Riemann surfaces are well known (see, for instance, [2,3,4,6,24]). In [11] it was proved that given integers k 1 , k 2 ∈ {0, 1, . . . , g}, there is a symmetric Riemann surface admitting two symmetries τ 1 and τ 2 so that τ j has k j ovals. In [14] there is derived a sharp upper bound on the number of ovals of two symmetries. Most of that study is consequence of the well known structure of non-Euclidian crystallographic (NEC) groups [15].
A lowest regular planar cover of a closed Riemann surface S of genus g ≥ 1 is known to be provided by a Schottky group Γ of rank g (i.e., a purely loxodromic Kleinian group, with non-empty region of discontinuity, isomorphic to a free group of rank g). If Ω is the region of discontinuity of Γ, then M = (H 3 ∪ Ω)/Γ is a compact manifold homeomorphic to a handlebody of genus g (we say that Γ induces a Schottky structure on M), its interior M 0 = H 3 /Γ carries a natural complete hyperbolic structure (that is, a complete Riemannian structure with constant negative curvature) and S = Ω/Γ is its conformal boundary.
A symmetry of the handlebody M is an orientation reversing self-homeomorphism τ of order two, whose restriction to M 0 is a hyperbolic isometry; we also say that M is symmetric. By lifting τ to the universal cover, we obtain an extended Kleinian group Γ whose orientation preserving half is Γ (called an extended Schottky group). A geometrical structure of the extended Schottky groups, in terms of the Klein-Maskit combination theorems [16,17], was obtained in [5]; it provides the type of the corresponding symmetry in a similar way as NEC groups do for the case of symmetries of Riemann surfaces. It follows from such a geometrical description (see also [12]) that the locus of fixed points of τ has at most g + 1 connected components and that each of such connected components is either an isolated point (in the interior M 0 ) or a 2-dimensional bordered compact surface (which may or not be orientable) whose border is contained in S . The interior of the quotient orbifold M/ τ = (H 3 ∪ Ω)/ Γ fails to be a manifold exactly at the projection of the isolated fixed points of τ (locally looks like a cone over the projective plane) and its conformal boundary is the compact Klein surface S / τ = Ω/ Γ. At this point it is important to note that the presence of isolated fixed points of τ is not detected by its restriction to S ; so the above upper bound is not a direct consequence of Harnack's result for symmetric Riemann surfaces. We say that the symmetry is maximal (for M) if the number of connected components of its locus of fixed points is g + 1.
Let us note that a symmetry of M induces a symmetry on its conformal boundary closed Riemann surface S , but a symmetry on S may not be extended to a symmetry of M. Nevertheless, a symmetry η of S always keeps invariant a collection of pairwise disjoint simple loops which cut-off the surface into planar surfaces. It follows that there is a new handlebody N (with a suitable Schottky structure) admitting S as its conformal boundary so that η is induced by a symmetry of N. We should remark that if S has two different symmetries, then it is not clear the existence of a common handlebody N admitting two symmetries, each one inducing the given ones on S ; this makes the difference on the study of symmetries at the level of handlebodies with that on Riemann surfaces.
Let τ 1 , τ 2 be two different symmetries of a handlebody M with Schottky structure induced by the Schottky group Γ (of rank g ≥ 2). As said before, by lifting the symmetry τ j , for j = 1, 2, to the universal cover space, we obtain an extended Schottky groups Γ j containing Γ as its orientation preserving half. These two groups generate an extended Kleinian group K (called a dihedral extended Schottky group). In Theorem 4.1 we provide a geometrical structural description, in terms of the Klein-Maskit combination theorems, of these dihedral extended Schottky groups. As a consequence of such a geometrical structure description we are able to obtain sharp upper bounds for the number of connected components of k ∈ {2, 3} symmetries on a handlebody (Theorems 2.1 and 2.4). Such upper bounds are slightly different as those ones for Riemann surfaces. In particular, we obtain that a handlebody has at most one maximal symmetry (Corollary 2.2). Let us robserve that at the level of Riemann surfaces, Natanzon [22] proved that if a hyperelliptic Riemann surfaces admits a maximal symmetry, then it admits two of them. In the same paper it was proved the uniqueness of maximal symmetry in the non-hyperelliptic case.
Upper bounds for components of fixed points of symmetries
Sharp upper bounds for the number of ovals of two symmetries on a symmetric Riemann surface of genus g were provided in [3]: if q is the order of the product of these two symmetries, then such an upper bound is 2(g −1)/q +4 (if q is odd) and 4g/q +2 (if q is even). If moreover, q ≥ 3 (i.e., the two symmetries do not commute) and q does not divides g − 1, then in [14] it was obtained the upper bound where [ ] stands for the integer part and q, and, moreover, such an upper bound is sharp. Next result provides a sharp upper bound at the level of symmetric handlebodies.
Theorem 2.1. Let M be a handlebody of genus g ≥ 2, with a Schottky structure, and let τ 1 , and τ 2 be two different symmetries with m 1 and m 2 connected components of their fixed points, respectively. If q is the order of τ 1 τ 2 , then Moreover, for every positive integer q ≥ 2, the above upper bound is sharp for infinitely many values of g.
The proof of the above is done in Section 6 and in Section 8 we provide examples to see the sharp part. Let us observe that, for instance, for g ≥ 2 and q = 2, the sharp upper bound at the level of Riemann surfaces is 2g + 2, but at the level of handlebodies this is g + 3, from which we obtain the following fact (already observed in [10] if both symmetries only have isolated fixed points).
Corollary 2.2. A handlebody of genus g ≥ 2 admits at most one maximal symmetry. Theorem 2.1 asserts that the upper bound m 1 + m 2 = g + 3 may only occur if q = 2, that is, when τ 1 , τ 2 = Z 2 2 (explicit examples are provided in [10]); so the following fact holds. Corollary 2.3. Let M be a handlebody of genus g ≥ 2, with a Schottky structure, and let τ 1 , and τ 2 be two symmetries with m 1 and m 2 connected components of their fixed points, respectively. If m 1 + m 2 = g + 3, then τ 1 , τ 2 = Z 2 2 . For the case of three symmetries on a closed Riemann surface of genus g, in [23] it was proved that number of ovals is bounded above by 2(g + 2) and that such upper bound is sharp. The following provides the corresponding situation for the case of handlebodies. We denote by D r the dihedral group of order 2r.
Theorem 2.4. Let M be a handlebody of genus g ≥ 2 with a Schottky structure and let τ 1 , τ 2 and τ 3 be three different symmetries and set H = τ 1 , τ 2 , τ 3 . If m i is the number of connected components of fixed points of τ i then if g ≥ 4 and H Z 2 × D r for any r.
Moreover, the above upper bounds are sharp for g = 2, 3 and for infinite many values of g ≥ 4.
Preliminaries and previous results
In this section we briefly review several definitions and basic facts we will need in the rest of the paper. More details on these topics may be found, for instance, in [18,20].
3.1. Extended Kleinian groups. We denote by M the group of Möbius and extended Möbius transformations (the composition of a Möbius transformation with the complex conjugation) and by M its index two subgroup of Möbius transformations. The group M can also be viewed, by the Poincaré extension theorem, as the group of hyperbolic isometries of the hyperbolic space H 3 ; in this case, M is the group of orientation-preserving ones. Möbius transformations are classified into parabolic, loxodromic (including hyperbolic) and elliptic. Similarly, extended Möbius transformations are classified into pseudo-parabolic (the square is parabolic), glide-reflections (the square is hyperbolic), pseudo-elliptic (the square is elliptic), reflections (of order two admitting a circle of fixed points on C) and imaginary reflections (of order two and having no fixed points on C). Each imaginary reflection in M has exactly one fixed point in H 3 and this point determines such reflection uniquely. If K is a subgroup of M not contained in M, then K + = K ∩ M is its canonical orientation-preserving subgroup.
A Kleinian group is a discrete subgroup of M and an extended Kleinian group is a discrete subgroup of M necessarily containing extended Möbius transformations. If K is a (extended) Kleinian group, then its region of discontinuity is the subset Ω of C composed by the points on which it acts discontinuously. Note that K is an extended Kleinian groups if and only if K + is a Kleinian group; both of them with the same region of discontinuity.
(1) (Free products) Let K j be a (extended) Kleinian group with region of discontinuity Ω j , for j = 1, 2. Let F j be a fundamental domain for K j and assume that there is simple closed loop Σ, contained in the interior of F 1 ∩ F 2 , bounding two discs D 1 and D 2 , so that, for j ∈ {1, 2}, Σ ∪ D j ⊂ Ω 3− j is precisely invariant under the identity in K 3− j . Then (i) K = K 1 , K 2 is a (extended) Kleinian group with fundamental domain F 1 ∩ F 2 and K is the free product of K 1 and K 2 (ii) every finite order element in K is conjugated in K to a finite order element of either K 1 or K 2 and (iii) if both K 1 and K 2 are geometrically finite, then K is so.
(2) (HNN-extensions) Let K 0 be a (extended) Kleinian group with region of discontinuity Ω, and let F be a fundamental domain for K 0 . Assume that there are two pairwise disjoint simple closed loops Σ 1 and Σ 2 , both of them contained in the interior of F 0 , so that Σ j bounds a disc D j such that (Σ 1 ∪D 1 )∩(Σ 2 ∪D 2 ) = ∅ and that Σ j ∪ D j ⊂ Ω is precisely invariant under the identity in K 0 . Let T be either a loxodromic transformation or a glide-reflection so that T (Σ 1 ) = Σ 2 and T (D 1 ) ∩ D 2 = ∅. Then (i) K = K 0 , f is a (extended) Kleinian group with fundamental domain F 1 ∩ (D 1 ∪ D 2 ) c and K is the HNN-extension of K 0 by the cyclic group T , (ii) every finite order element of K is conjugated in K to a finite order element of K 0 and (iii) if K 0 is geometrically finite, then K is so.
3.3. Kleinian 3-manifolds and their automorphisms. If K is a Kleinian group and Ω is its region of discontinuity, then assocated to K is a 3-dimensional orientable orbifold M K = (H 3 ∪ Ω)/K; its interior M 0 K = H 3 /K has a hyperbolic structure and its conformal border S K = Ω/K has a natural conformal structure. If K is torsion free, then M K and M 0 K are orientable 3-manifolds and S K is a Riemann surface; we say that M K is a Kleinian 3-manifold and that M K and S K are uniformized by K. Now, if K is an extended Kleinian group and K = K + , then the 3-orbifold M K admits the orientationreversing homeomorphism τ : M K → M K of order two induced by K − K and M K / τ = (H 3 ∪ Ω)/ K.
Let M be a Kleinian 3-manifold, say M = (H 3 ∪ Ω)/Γ, let S = Ω/Γ be its conformal boundary and let M 0 = H 3 /Γ be its interior hyperbolic 3-manifold. An automorphism of M is a self-homeomorphism whose restriction to its interior M 0 is a hyperbolic isometry. An orientation-preserving automorphism is called a conformal automorphim and an orientation-reversing one an anticonformal automorphism. A symmetry of M is an anticonformal involution. We denote by Aut(M) the group of automorphisms of M and by Aut + (M) the subgroup of conformal automorphisms. Let π 0 : H 3 → M 0 be the universal covering induced by Γ. Clearly, π 0 extends to a universal covering π : H 3 ∪ Ω → M with Γ as the group of Deck transformations. If G ⊂ Aut(M) is a finite group and we lift it to the universal covering space H 3 under π 0 , then we obtain an (extended) Kleinian group K containing Γ as a normal subgroup of finite index. The group G contains orientation-reversing automorphisms if and only if K is an extended Kleinian group.
3.4. Schottky groups and their Kleinian manifolds. The Schottky group of rank 0 is just the trivial group. A Schottky group of rank g ≥ 1 is a Kleinian group Γ generated by loxodromic transformations A 1 , . . . , A g , so that there are 2g disjoint simple loops, C 1 , C ′ 1 , . . . , C g , C ′ g , with a 2g-connected outside D ⊂ C, where A i (C i ) = C ′ i , and A i (D) ∩ D = ∅, for i = 1, . . . , g. The region of discontinuity Ω of Γ is known to be connected and dense in C, that S = Ω/Γ is a closed Riemann surface of genus g and that the associated Kleinian manifold M = (H 3 ∪ Ω)/Γ is a handlebody of genus g. In this case, its interior M 0 = H 3 /Γ carries a geometrically finite complete hyperbolic Riemannian metric with injectivity radius bounded away from zero. Conversely, those geometrically finite hyperbolic structures for which the injectivity radius is bounded away from zero in the interior of a handlebody are provided by Schottky groups. As a consequence of the retrosection theorem [1,13], every closed Riemann surface can be uniformized by a Schottky group.
It is well known that a Schottky group of rank g can be defined as a purely loxodromic Kleinian group of the second kind which is isomorphic to a free of rank g [19]. It also follows that every Kleinian structure on a handlebody is provided by a Schottky group. A Schottky group of rank g can also be defined as a purely loxodromic geometrically finite Kleinian group which is isomorphic to a free of rank g (essentially a consequence of the fact that a free group cannot be the fundamental group of a closed hyperbolic 3manifold). It follows that every Kleinian structure on a handlebody is provided by a Schottky group. For this reason, we say that a Kleinian structure on a handlebody is a Schottky structure.
If M is a handlebody of genus g ≥ 2, with a Schottky estructure, then it is known that Aut(M) is finite; moreover, Aut(M) has order at most 24(g − 1) and Aut + (M) has order at most 12(g − 1) [25,26]. Each conformal (respectively anticonformal) automorphism of M induces a conformal (respectively anticonformal) automorphism of the conformal boundary S and the later determines the former due to the Poincare extension theorem.
Let M be a topological handlebody of genus g and let H be a finite group of homeomorphisms of M. It is well known that there are a (extended) Kleinian group K, containing as a normal subgroup a Schottky group Γ of rank g, and an orientation preserving homeomorphism f : M → M Γ , where M Γ = (H 3 ∪ Ω)/Γ is the handlebody uniformized by Γ, with f H f −1 = K/Γ. This is in really consequence of the fact that a handlebody is a compression body (see also [26]). In this way, to obtain examples of handlebodies with groups of sutomorphisms, one may just work at topological constructions. These topological constructions may be produced by fattening up some symmetrical graphs; for instance Cayley graphs. The examples we produce in this paper are done in terms of Schottky groups, but they can be obtained as before by using (finite extensions of) dihedral groups and their Cayley graphs.
3.5. Extended Schottky groups. An extended Schottky group of rank g is an extended Kleinian group whose canonical orientation-preserving subgroup is a Schottky group of rank g. Those extended Schottky groups containing no reflections are called Klein-Schottky groups (these were previously considered in [10]) and the others are called reflection Schottky groups. As a consequence of the results in [5,12], it is possible to obtain a geometric structural description of extended Schottky groups in terms of the Klein-Maskit combination thorems as follows.
). An extended Schottky group is the free product (in the Klein-Maskit combination theorem sense) of the following kind of groups: (i) cyclic groups generated by reflections, (ii) cyclic groups generated by imaginary reflections, (iii) cyclic groups generated by glide-reflections, (iv) cyclic groups generated by loxodromic transformations, and (v) real Schottky groups (that is groups generated by a reflection and a Schottky group keeping invariant the corresponding circle of its fixed points).
Let us consider a handlebody M, with a Schottky structure induced by the Schottky group Γ. If τ : M → M is a symmetry, then the lifting of τ to the universal cover space produces an extended Schottky group Γ containing Γ as its index two orientation preserving part and so that Γ/Γ = τ . Conversely, every extended Schottky group is obtained by the lifting of a symmetry of a handlebody with a Schottky structure. Direct consequence of Theorem 3.2 is then the following. Corollary 3.3. Let Γ be an extended Schottky group constructed, as in Theorem 3.2, using α groups of type (i), β groups of type (ii), γ groups of type (iii), δ groups of type (iv) and ε groups of type (v). If Γ is the canonical orientation-preserving subgroup of Γ and M = (H 3 ∪ Ω)/Γ, then Γ induces a symmetry of M whose connected components of fixed points consist of α two dimensional closed discs, β isolated points, and ε two dimensional non-simply connected compact surfaces. In particular, If τ is a symmetry of a Kleinian manifold homeomorphic to a handlebody of genus g, and n 0 is the number of isolated fixed points of τ, n 1 is the number of total ovals in the conformal boundary and n 2 is the number of twodimensional connected components of the set of fixed points of τ, then n 0 + n 1 , n 0 + n 2 ∈ {0, 1, . . . , g + 1}.
3.6. Lifting criteria. Next we recall a simple criterion for lifting loops which we will need in the structure description of the dihedral extended Schottky groups. This is a direct consequence of the Equivariant Loop Theorem [21], whose proof is based on minimal surfaces, that is, surfaces that minimize locally the area. In [9] a proof which only uses arguments proper to (planar) Kleinian groups is provided. A function group is a pair (G, ∆), where G is a finitely generated Kleinian group and ∆ is a G-invariant connected component of its region of discontinuity.
3.7.
A counting formula. Let K be an extended Kleinian group containing a Schottky group Γ of rank g as a normal subgroup of finite index (the last trivially holds if g ≥ 1). Let G = K/Γ and let us denote by θ : K → G, the canonical projection. If there is some involution τ ∈ G , which is the θ-image of an extended Möbius transformation, then Γ = θ −1 ( τ ) is an extended Schottky group whose canonical orientation-preserving subgroup is Γ. As consequence of Theorem 3.2, the group Γ is constructed using α reflections, β imaginary reflections and ε real Schottky groups. Theorem 3.5 below asserts that values α, β and ε can be obtained from K, G and θ. First we provide some necessary definitions.
Let C = {c i : i ∈ I} be a maximal collection of anticonformal involutions (i.e. reflections and imaginary reflections) in K which are non-conjugate there. The group K is geometrically finite, as it is a finite extension of a Schottky group, so one can prove that the set C is finite. However, due to the Theorem 3.2, we will not need it explicitly in the proof of the next theorem and so we do not go into details. We call the set C a complete set of symmetries of K and we shall refer to its elements as to canonical symmetries. For each i ∈ I we set I(i) ⊂ I defined by those j ∈ I so that θ(c i ) and θ(c j ) are conjugate in G (in particular, i ∈ I(i)). Note at this point that it may happens that for j ∈ I(i), c j can be imaginary reflection even if c i is a reflection and viceversa. Such situations occur when θ(c i ) is a symmetry containing both isolated fixed points and two-dimensional components of the set fixed points. We set by J(i) the subset of I(i) defined by those j for which c j is an imaginary reflection. We also set by F(i) ⊂ I(i) − J(i) for those j for which c j has a finite centralizer in K and E(i) = I(i) − (J(i) ∪ F(i)). Note that, as Γ has a finite index in K, a reflection c ∈ K has an infinite centralizer C( K, c) in K if and only if it has an infinite centralizer in Γ.
Structural description of dihedral extended Schottky groups
In this section we provide the structural picture of dihedral extended Schottky groups in terms of the Klein-Maskit combination theorems. (1) A dihedral extended Schottky group is the free product (in the sense of the Klein-Maskit combination theorems) of the following groups (see Figure 1) (i) α cyclic groups generated by reflections; (ii) β cyclic groups generated by imaginary reflections; (iii) γ cyclic groups generated by loxodromic transformations; (iv) δ cyclic groups generated by glide-reflections; (v) ε groups generated by a reflection and some finite number of elliptic transformations and reflections, each of them commuting with the previous reflection, so that α + β + δ + ε > 0.
(2) Let K be any extended function group constructed by the previous groups. Then, K is a dihedral extended Schottky group if and only if there is a surjective homomorphism from with kernel a Schottky group so that ϕ(K + ) is the cyclic group of order p, for some positive integer p > 1.
If K is a dihedral extended Schottky group and Γ ⊳ K is a Schottky group of rank g which is a normal subgroup of K so that K/Γ D n , then where g is as in (a). (2). Also, in Part (3), if t jk is elliptic of order two, then σ j t jt is an imaginary reflection commuting with σ j . In this way, we only need to provide the proof of Parts (1) and (2) of the above theorem. (1) and (2) Let us consider a dihedral extended Schottky group G, generated by two different extended Schottky groups G 1 and G 2 , so that G + 1 = G + 2 = G. Let us denote by G + the index two orientation preserving half of G. It follows that G, G 1 , G 2 , G and G + all have the same limit set, and hence the same region of discontinuity, Ω, on C. We set S + = Ω/G; S 1 = Ω/G 1 ; S 2 = Ω/G 2 ; S + = Ω/ G + ; and S = Ω/ G. Since G is a finite extension of G, and S + is a closed Riemann surface, S is a compact not necessarily orientable orbifold, with a finite number of orbifold points and possible non-empty boundary, and S + is a closed Riemann surface with some finite number of orbifold points. On S + we have an anticonformal involution, preserving the finite set of orbifold points, so that quotient of S + by it is S . Proof. (i) follows at once from the fact that [ G : G] < ∞ and G contains no parabolic elements. As G + is a geometrically finite function group, (ii) follows from [8].
Remark 4.2. Let us note that Parts (3)(a) and (3)(c) are direct consequence of Riemann-Hurwitz formula and that Part (3)(b) is consequence of Part
Since G is a subgroup of index 2 in G i , i = 1, 2, we have that each orientation reversing element η i ∈ G i induces an orientation-reversing conformal automorphism τ i of S + , so that S i = S + / τ i . As η 2 i ∈ G, we have that η i is either: (i) a reflection or (ii) an imaginary reflection or (iii) semi-hyperbolic.
Remark 5.2. In the case the extended Schottky group G j is a Klein-Schottky group, we have that η j is either an imaginary reflection or semi-hyperbolic.
Let J be the group of automorphisms of S + generated by τ 1 and τ 2 ; then S = S + /J. Since S + is closed, and τ 1 τ 2 , there is some smallest positive power p > 1 so that (τ 1 • τ 2 ) p = 1, that is, J is a dihedral group of order 2p. It follows that every orientation-reversing element of J is conjugate in J to either τ 1 or τ 2 . Lifting to Ω, we see that every orientation-reversing element of G either acts without fixed points on Ω or they are reflections.
Proposition 5.4. Every orientation-reversing element of G is either semi-hyperbolic or an imaginary reflection or a reflection.
Proof. Every orientation-reversing element of G is a lift of a conjugate of either τ 1 or τ 2 , each one either a reflection or an imaginary reflection. It follows that if a ∈ G is orientation-reversing, then a 2 ∈ G. In this way a 2 = 1, in which case a is either a reflection or an imaginary reflection, or a 2 is a loxodromic transformation, then a is neither semi-parabolic nor semi-elliptic. Proof. Denote the projection from Ω to S by π, and denote the projection from Ω to S + by π + . Choose any point z ∈ Ω and any element a ∈ G. Then, since π(z) = π(a(z)), there is a j ∈ J so that π + (a(z)) = j(π + (z)). It is clear that this map is continuous in z; hence j depends only on a. Hence we have defined Φ(a) = j. Now suppose that Φ(a i ) = j i , i = 1, 2. Then π + (a i (z)) = Φ(a i )(π + (z)), for all z ∈ Ω. So Φ(a 2 a 1 )(π + (z))) = π + (a 2 (a 1 (z)) = Φ(a 2 )(π + (a 1 (z))) = Φ(a 2 )Φ(a 1 )(π + (z)).
This shows that Φ is a homomorphism; it is obvious that its kernel is exactly G. It is also obvious that the image of Φ is contained in J. Since every element of J lifts to Ω to yield an element of G − G, Φ maps onto J.
The next result make explicit some properties of Φ we will need. First, we need some remark. The dihedral group J has order 2p. If p is odd, then every involution on J is orientation reversing; but if p is even, then there are involutions in J that preserve orientation. In the particular case that p = 2, we have that any two involutions in J generate it. Proof. The first statement follows at once from the fact that the kernel of G is a Schottky group, which is free. The second statement follows at once from the fact that if a 1 , a 2 ∈ Φ −1 (α), then a 1 a 2 ∈ Ker(Φ) = G, that is, both reverse orientation or both preserve it. As we have assumed that α is orientation reversing, we must have some orientation reversing Möbius transformation in Φ −1 (α). The argument is similar for β. The last statement is now easy to see.
Remark 5.7. Observe that Φ( G + ) is the cyclic group J + of order p generated by αβ in Proposition 5.6. As G ⊳ G + , Φ restrict to a surjective homomorphism from G + onto the cyclic group J + with kernel the Schottky group G.
We need some topological information about the group G. We first need to generalize the decomposition of a function group into its structure subgroups to the case of an extended Kleinian group. Let us consider the regular planar Schottky covering (Ω, G, P : Ω → S + ). By construction, we have that the group J lifts under P and such a lifting is the group G. As a consequence of Theorem 3.4 we have the existence of a collection of pairwise disjoint loops F in S + so that it divides S + into genus zero surfaces (since we are dealing with a Schottky covering), each loop lifts to a simple loop on Ω and so that F is invariant under the action of J.
Let A be a connected component of S −F and let J A its stabilizer in J. As A is a genus zero surface, we have that J A is a finite subgroup of M. We also have that J A is subgroup of a dihedral group of order 2p, which restrict the possibilities for J A as subgroup of M. If we set σ = τ 1 τ 2 the conformal automorphism of S + of order p, then J A is either: (i) a cyclic conformal group generated by a power of σ; or (ii) a cyclic group of order two generated by an anticonformal involution conjugated to either τ 1 or τ 2 ; or (iii) a dihedral subgroup of J. In either case (ii) or (iii) we have that A is invariant under an anticonformal involution τ ∈ J. If τ is either (a) a reflection containing a loop of fixed points on A or (b) an imaginary reflection, then we may find a simple loop β ⊂ A which is invariant under τ; moreover, if τ is reflection, then β is formed of only fixed points of it. We may add such a loop and its J-translated to F without destroying the conditions of Theorem 3.4; we still denoting such a collection by F . In this way, we may assume that in either (ii) or (iii) the anticonformal involution is necessarily a reflection whose circle of fixed points is not completely contained in A (it intersects some boundary loops). Now we can delete loops from this new collection F so that we do not destroy all the above properties in order to make it minimal in the sense that eliminating from it any finite collection of loops will destroy any of the above properties. We have then, as a consequence of the above, the following existence fact.
Proposition 5.8. There is a collection of pairwise disjoint simple loops F in S + so that: (
i) each loop in F lifts to pairwise disjoint simple loops on Ω; (ii) F divides S + into genus zero surfaces; (iii) F is invariant under the action of J;
(iv) each connected component A of S + − F has stabilizer J A being either: (iv.1) trivial; or (iv.2) a cyclic conformal group generated by a non-trivial power of σ = τ 2 τ 1 ; or (iv.
3) a cyclic group of order two generated by a reflection with no oval of fixed points completely contained inside A; or (iv.4) a dihedral group generated by a non-trivial power of σ and a reflection with no oval of fixed points completely contained inside A. (v) F is minimal in the sense that eliminating from it any finite collection of loops will destroy any of the above properties (i)-(iv).
Remark 5.9. In case both extended Schottky groups G 1 and G 2 are Klein-Schottky groups, as we know that in this case all anticonformal involutions in G are imaginary reflections, Proposition 5.8 asserts that J A is either trivial or a cyclic group generated by a non-trivial power of σ.
Let us denote by G the loops on Ω obtained by the lifting of those in F (as in proposition 5.8). We call these simple loops the structure loops and the connected components of Ω − G the structure regions.
The loops in F are called the base structure loops and the connected components of S + − F are called the base structure regions. By the construction of the base structure loops, we may see that The set of structure loops G and the set of structure regions are invariant under G.
If we consider a structure region R so that P(R) = A, then we have that P : R → A is a conformal homeomorphism, in particular, the stabilizer in G of the structure region R, say G R is isomorphic to J A . In particular, the following holds.
Proposition 5.10. The stabilizer in G of any structure region R is either: (i) trivial; (ii) a finite cyclic conformal group; (iii) a cyclic group of order two generated by a reflection whose circle of fixed points is not completely contained on R; (iv) a dihedral group generated by a reflection τ with its circle of fixed points intersecting some boundary loop of R and an elliptic element η of finite order so that (ητ) 2 = 1.
Proposition 5.11. If R ⊂ Ω is a structure region whose stabilizer is trivial, then the restriction to R of the projection map from Ω to S = S + /J = Ω/ G is a homeomorphism onto its image.
Proof. If a ∈ G, then either a ∈ G R = {I}, in which case, a(R) = R, or a G R , in which case, a(R) ∩ R = ∅.
Let us now consider some structure loop β ∈ G and denote its stabilizer in G by G β . As our group J has no a dihedral subgroup inside the cyclic subgroup J + , the orientation perserving half of J, we have the following fact.
Proposition 5.12. The orientation preserving half of G β is either trivial or a finite cyclic group generated by some elliptic element.
We have from Proposition 5.4 that an orientation reversing transformation inside G is either an imaginary reflection or a reflection or a semi-hyperbolic transformation. Clearly, as a structure loop is contained in Ω, we have that a semi-hyperbolic transformation cannot keep such a structure loop invariant. In particular, the only orientation reversing transformations in G that keep invariant some structure loop can be either an imaginary reflection or a reflection. Also, a structure loop can be stabilized by at most one imaginary reflection, for the product of two distinct imaginary reflections is always hyperbolic, and the entire structure loop must be contained in Ω. A structure loop can be stabilized by a reflection in two different manners. One manner is that it fixes it point-wise, that is, the structure loop is the circle of fixed points of the reflection. The second manner is that the structure loop is not point-wise fixed by the reflection, in which case, there are exactly two fixed points of the reflection on the structure loop. These two points divide the loop into two arcs which are permuted by the reflection. All the above gives us the following possibilities.
Proposition 5.13. Let β ∈ G a structure loop and G β its stabilizer in G. Then, we have the following possibilities: (i) G β is trivial; (ii) G β is a cyclic group generated by some elliptic element of finite order whose fixed points are separated by β; (iii) G β is a cyclic group generated by an elliptic element of order two with its fixed points on β; (iv) G β is a cyclic group of order two generated by some imaginary reflection; (v) G β is a cyclic group of order two generated by some reflection and β is the circle of fixed points of the reflection; (vi) G β is a cyclic group of order two generated by some reflection and β is not the circle of fixed points of the reflection. In this case, β should contain exactly two fixed points of the reflection; (vii) G β is generated by a reflection that has exactly two fixed points on β and an elliptic involution with same both fixed points. In this case, the composition of these two is an imaginary reflection and G β is isomorphic to Z 2 2 ; (viii) G β is generated by an elliptic involution whose fixed points are separated by β and a reflection that has β as its circle of fixed points on. In this case, G β is isomorphic to Z 2 2 ; (ix) G β is generated by an elliptic involution with both fixed points on β and a reflection for which β is its circle of fixed points. In this case, G β is isomorphic to Z 2 2 ; and (x) G β is generated by an elliptic involution with both fixed points on β and a reflection for which β intersects the circle of fixed points of the reflection at two points separating the fixed points of the conformal involution. In this case, G β is isomorphic to Z 2 2 ; and (xi) G β is generated by a reflection with exactly two fixed points on β and elliptic involution with both fixed points on the circle of fixed points of the reflection. In this case, G β is isomorphic to Z 2 2 ; The structure loops divide Ω into structure regions; the stabilizer of every structure region is either (i) trivial or (ii) elliptic cyclic or (iii) a cyclic group of order two generated by a reflection whose circle of fixed points is not completely contained on the structure region or (iv) a dihedral group generated by an elliptic transformation and a reflection whose circle of fixed points is not completely contained on the structure region; hence every structure region has a finite number of structure loops on its boundary. Moreover, by Proposition 5.13, we also know all possibilities for the stabilizers of these structure loops. The following fact is easy to see.
Proposition 5.14. Let R and R ′ be any two different structure regions with a common boundary loop β.
Then, they are equivalent under G if and only if (i) G β contains an element which does not belong to G R or (ii) there is another boundary loop β ′ of R and an element t
Next result is related to those structure regions with stabilizer either trivial or a cyclic group generated by a reflection whose circle of fixed points is not completely contained inside the structure region.
Proposition 5.15. Let R ⊂ Ω be a structure region with stabilizer G R either trivial or a cyclic group of order two generated by a reflection whose circle of fixed points is not completely contained on R. If β is a boundary loop of R, which is not invariant under a reflection in G R , then there is a non-trivial element k ∈ G − G R so that k(β) still a boundary loop of R.
Proof. Let β be a structure loop on the boundary of the structure region R. We assume that it is not invariant under a reflection that stabilizes R. The hypothesis on G R and Proposition 5.13 gives us two general situations for G β .
First case: The stabilizer G β of β in G contains a cyclic group generated by either (i) a reflection whose circle of fixed points is β or (ii) an imaginary reflection or (ii) an elliptic element of order two with both fixed points on β. In this case we may choose for k the generator of such a cyclic subgroup of G β , which clearly does not belong to G R . Second case: The stabilizer G β is either: (i) trivial or (ii) a cyclic group generated by a reflection that does not fixed it point-wise. In case (ii) we have that the reflection is the generator of the stabilizer of R; by our hypothesis this case does not happen. Then we have that G β is trivial. In this case, the projection of β on S + is a simple loop β * which has trivial stabilizer in J. We have that β is free homotopic to the product of the other boundary loops of R. If none of the other boundary loops of R is equivalent to β under G, then we may delete β * and its J-translates from F , contradicting the minimality of F .
The other kind of structure regions R always contain an elliptic cyclic group H = G + R inside its stabilizer G R (there are two cases, the stabilizer is either cyclic or a dihedral group). In the case that G R is a dihedral group, we know that the anticonformal involution in G R is a reflection whose circle of fixed points intersect some boundary loops of R; that is, each boundary loop of R is either kept invariant (with exactly two fixed points of the reflection) or it is permuted with other boundary loop (in which case it has no fixed points of the reflection) by such a reflection. In any of the two possibilities for G R , we have that R has either 0, 1 or 2 structure loops stabilized by H on its boundary; it has also some number of structure loops stabilized only by the identity in H on its boundary. It is clear that if no structure loop on the boundary of R is stabilized by H, then both fixed points of H lie in R. It is also clear that a structure loop is stabilized by H if and only if it separates the fixed points of H. Also, observe that if G R is a dihedral group, as the reflection in G R has no its circle of fixed points completely contained on R, then we must have that in the case that R contains only one of the fixed points of H, such a reflection should fix that point, in particular, such a reflection should commute with a generator of H; this obligates to have G R Z 2 2 .
Proposition 5.16. If R ⊂ Ω is a structure region with non-trivial conformal stabilizer G +
R = H and there is a fixed point of H in R, then both fixed points of H lie in R.
Proof. Suppose there is only one fixed point of H in R. Then there is a unique structure loop W on the boundary of R stabilized by W. Then every other structure loop on the boundary of R is stabilized by the identity in H. If G R = H, then it follows that if were to fill in the discs bounded by the other structure loops on the boundary of R, then W would be contractible; that is, if we delete the projection of W and their J-translates from our list of base structure loops, this would leave unchanged the smallest normal subgroup containing the base structure loops raised to appropriate powers. Since we have chosen our base structure loops to be minimal, this cannot be. If G R H, then we have a reflection τ ∈ G R whose circle of fixed points is not completely contained in R, and G R = H, τ a dihedral group. In this case, we should have that the fixed point of H in R is also fixed by τ; then, both fixed points of H are fixed by τ. As G R Z 2 2 , we have H Z 2 . In this case, we may also delete the projection of W and its J-translates from F in order to get a contradiction to the minimality of F . Proposition 5.17. Assume we have a structure region R ⊂ Ω with stabilizer G R containing a non-trivial conformal cyclic group, say generated by k. If β 1 , β 2 are two different boundary loop of R which are invariant under k, then there is a (non-trivial) element k ∈ G so that k(β 1 ) = β 2 .
Proof. Assume this is not the case. We observe that R/ G + R ⊂ S + is an annulus with some holes cut out of it. If we fill those holes with discs, then the projections of β 1 and β 2 become freely homotopic; the projection of β 1 , raised to the appropriate power, is freely homotopic to a product of the projection of β 2 , raised to the same power, and the projections of the other structure loops on the boundary of R. Hence we can delete the projection of β 1 on S + and its J-translates, which contradicts the minimality of this set.
Structure regions with trivial stabilizers. Let G be a dihedral extended Schottky group admitting a structure region R with trivial stabilizer G R = {I}. As consequence of Proposition 5.15, every other structure region is equivalent under G to R and R is a fundamental domain for G and the boundary loops are paired by either (i) reflections, (ii) imaginary reflections, (iii) loxodromic transformations or (iv) semi-hyperbolic transformations (see Figure 1).
Structure regions with non-trivial stabilizers. Let us now consider the case our dihedral extended Schottky group G has not a structure region with trivial stabilizer. Proposition 5.14, and the fact that S is compact and connected, permits us to construct a connected set R obtained as the union of a finite collection of non-equivalent structure regions (some of them may have non-trivial stabilizer) together their boundary loops which projects onto S . Some of the boundary loops of R have a non-trivial stabilizer which sends R at the other side of the boundary loop an the other boundary loops are equivalent in pairs under G.
(1) Assume one of the structure regions R ⊂ R has stabilizer a cyclic group of order two generated by a reflection τ we have the following. The circle of fixed points C τ of τ intersects some structure loops in the boundary of R. Let us denote by γ the intersection of C τ with R. Let us denote by R 1 the union of one of the components of R−γ with γ. Inside R 1 we have boundary loops, which, as a consequence of Proposition 5.15 are paired by either (i) reflections, (ii) imaginary reflections, (iii) loxodromic transformations or (iv) semi-hyperbolic transformations. The other boundary loops intersect C τ . Let β be any structure loop in the boundary of R which intersects C τ , necessarily at two points. If β belongs to the boundary of R, then we already noted that either (a) there is an involution κ ∈ G (conformal or anticonformal) with κ(β) = β and κ( R) ∩ R = β; or (b) there is another boundary loop β ′ of R and an element σ ∈ G so that σ(β) = β ′ and σ( R) ∩ R = β ′ . If β is in the interior of R, then we have another structure region R ′ ⊂ R with β as one of its border loop. In this case, as consequence of Proposition 5.14, we should have G β = τ and τ ∈ G R ′ .
(2) Assume one of the structure regions R ⊂ R has stabilizer a cyclic group generated by an elliptic transformation η. In this case we have two possibilities: either (i) both fixed points of η belong to R or (ii) there are two boundary loops β 1 and β 2 of R, each one invariant under η, and there is some κ ∈ G with κ(β 1 ) = β 2 , κ( R) ∩ R = β 2 . We have that each other boundary loop β of R is either (i) the boundary of another structure region inside R (so it has trivial stabilizer in G) or (ii) it belongs to the boundary of R (in which case we have either (a) or (b) above). 5.1. Geometrical Constructions: Part (1) of Theorem 4.1. All the above asserts that G should be constructed, by using standard combination theorem techniques [18], by the groups in the first part of Theorem 4.1.
Remark 5.18.
(i) If both G 1 and G 2 are Klein-Schottky groups, then in G we have no reflections, then in the above construction we only have to use, in the previous constructions, groups of type (ii), (iii) and (iv) [10]. In particular, G + is a Schottky group. (ii) If G is constructed with groups of types (i), (ii), (iii), (iv) and (v), then we have that G + is the free product of a Schottky group and a finite number of order two elliptic groups. (iii) In the general situation we have that G + is a free product of a Schottky group and a finite number of elliptic groups of finite order; the orders of these elliptic elements should be divisors of some fixed positive integer p > 1.
5.2.
Part (2) of Theorem 4.1. Let us consider an extended function group K which is constructed by Klein-Maskit's combination theorems as described in the first part of Theorem 4.1. In general, it may happens that K is not a dihedral Schottky group. By hypothesis, there is a surjective homomorphism ϕ : K → D p with kernel a Schottky group Γ so that ϕ(K + ) is the cyclic group of order p, for some positive integer p > 1. Choose a, b ∈ D p , both of order 2, both generating D p . As K 1 = ϕ −1 ( a ) and K 2 = ϕ −1 ( b ) are both extended Schottky groups (in both cases Γ is its index two preserving half) and K = K 1 , K 2 , we obtain that K is in fact a dihedral extended Schottky group. The other direction is clear from the definition of a dihedral Schottky group.
Examples
Example 8.1 (Sharp upper bound in Theorem 2.1). Let q ≥ 2 and let us consider an extended Schottky group K, constructed by Theorem 3.2 using exactly r + 1 reflections E 1 ,. . . , E r+1 . The orbifold uniformized by K is a planar surface bounded by exactly r + 1 boundary loops. Let us consider the surjective homomorphism Let Γ = kerθ. If we set L = E 2 E 1 and C j = E r+1 E j+1 , for j = 1, . . . , r − 1, then it is not difficult to see that Γ = L q , C 1 , . . . , C r−1 , LC 1 L −1 , . . . , LC r−1 L −1 , . . . , L q−1 C 1 L −q+1 , . . . , L q−1 C r−1 L −q+1 is a Schottky group of rank g = (r − 1)q + 1. Let us consider the extended Schottky groups As the group K contains no imaginary reflections nor real Schottky groups, it follows that Γ j is constructed, by Theorem 3.2, using α i reflections and β i loxodromic and glide-reflection transformations. The handlebody M = (H 3 ∪ Ω)/Γ, where Ω denotes the region of discontinuity of K, admits two symmetries, say τ 1 and τ 2 induced by Γ 1 and Γ 1 . The number of connected components of fixed points of τ i is exactly α i . As consequence of Theorem 2.1, we should have α 1 + α 2 ≤ 2(r + 1). Next, we proceed to see that in fact we have an equality, showing that the upper bound in Theorem 2.1 is sharp. In order to achieve the above, we use Theorem 3.5. A complete set of symmetries in K is provided by E 1 ,. . . , E r+1 . We also should note that C( K, E i ) = E i , for every j, and that J(i) = ∅ and I(i) = F(i).
In this case Γ = kerθ is a Schottky group of rank g = 6n + 4. The handlebody produced by Γ admits the symmetries τ 1 = a, τ 2 = b and τ 3 = bab. It is clear (again by direct inspection or by using Theorem 3.5) that m 1 = m 2 = m 3 = 2n + 3. In this case Γ = kerθ is a Schottky group of rank g = 2r + 1. The handlebody produced by Γ admits the symmetries τ 1 = c, τ 2 = a and τ 3 = b. It is clear (again by direct inspection or by using Theorem 3.5) that m 1 = 2r and m 2 = m 3 = 4. | 2019-04-22T13:08:13.697Z | 2017-10-20T00:00:00.000 | {
"year": 2017,
"sha1": "d017b7505a4f60b4313eb8cdde933f9f4b9a2056",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "afc9b50f54ad3a401cd0dd7b0fbe0e4ae653e3d4",
"s2fieldsofstudy": [
"Mathematics"
],
"extfieldsofstudy": [
"Mathematics"
]
} |
259433477 | pes2o/s2orc | v3-fos-license | The effects of digital marketing, word of mouth, and service quality on the purchase decisions: An empirical study of food SMEs products
The purpose of this study is to analyze the influence of digital marketing, word of mouth, and service quality on purchasing decisions through a quantitative questionnaire using an accidental sampling method. The research is designed to find out the relationship between the influence of digital marketing, word of mouth and service quality on consumer purchasing decisions using a quantitative approach method. The variables in this study consist of independent variables and dependent variables. The variables studied include digital marketing, word of mouth, service quality and consumer purchasing decisions. The research was conducted at food SMEs in Jakarta, Indonesia. Sources of data in this study were primary data including consumer responses to digital marketing, word of mouth, service quality and purchasing decisions obtained from the results of distributing online questionnaires. The sample size used in this study was 680 people. The data collection method used in this study was an online questionnaire distributed by social media. The data were analyzed using SPSS software and structural equation modeling (SEM) with SmartPLS software tools. The results of this study indicate that the higher the digital marketing, word of mouth, and service quality, the higher the purchasing decision. SMEs must further optimize the use of digital marketing in marketing their companies such as uploading interesting content on one of the existing social media. From a word-of-mouth point of view, companies must promote more to their closest circle of benefits what they get from using SEMs products. Service quality must continue to provide excellent service to consumers or customers so that these consumers feel comfortable and satisfied.
Introduction
Information technologies are currently presenting increasingly diverse information media.Technological changes from the existence of print media, television, and radio, have developed into increasingly sophisticated technology resulting in media that can always be used, namely the internet.The internet is an information medium that can provide facilities and convenience to obtain the information needed.With advances in information and communication technology, the world no longer knows boundaries, distance, space or time.By using the internet, businesspeople no longer experience difficulties in obtaining information to support business activities, even now various kinds of information can be obtained, so information must be filtered to obtain appropriate and relevant information.
In the era of the industrial revolution 4.0 and digital technological developments, consumers use technology to meet their daily needs.This can be seen from how much consumers use information technology facilities in fulfilling the information they need, one of which is the use of the internet.According to Aditi et al. (2023), the internet is growing rapidly and has become one of the sources of information that can be accessed easily which has resulted in a steady stream of consumer changes in fulfilling lifestyles.The changing times have brought people's shopping lifestyles to shift from conventional ones to online sales site transactions which are affected by the changing lifestyles.The flow of changes in people's shopping styles has resulted in many online sales sites which have resulted in people getting more information and transacting easily without having to go to the place where the goods or services are sold.Changes in consumer behavior are due to the benefits of online purchases which are more practical and efficient in time, effort and cost (Hagen et al., 2022) .As business is currently developing, challenges and competition are getting sharper in seizing market share, companies are required to be able to empower their resources effectively and efficiently so that the company can have a competitive advantage.According to Barykin et al. (2022), every organization is expected to be able to develop a marketing strategy to survive in the face of competition, one of which is having a marketing strategy in the form of a marketing mix with digital marketing.The company's business strategy is expected to have an impact on financial, non-financial benefits, to survive in the industry, and to achieve the organization's long-term goals (He et al., 2023).Digital marketing is one of the marketing media that is currently in great demand by the public to support various activities carried out.They are gradually starting to leave the conventional/traditional marketing model and switch to modern marketing, namely digital marketing (Khan et al., 2023).
Digitalization provides a shift so that customers have a bigger role to play in determining the factors that influence their purchase intention for a product, banking products are no exception.This makes banks have to be more digitally involved with their customers and dig deeper to find out what influences their customers' buying intentions for launched banking products.Companies have traditionally utilized marketing to push their products to buyers, and digitalization of marketing allows companies to influence customer purchase intentions.The company's management integrates digital marketing with traditional marketing to provide profitable added value for customers as well as for the company.
Digital marketing communications and transactions can be done at any time in digital marketing.A survey conducted by Ayachi et al. (2022) found that 132.7 million Indonesians were connected to the internet, and currently the internet plays an important role in determining consumer purchasing decisions.The increase in the number of internet and social media users is a huge opportunity for businesspeople to market any product.Alsoud et al. (2023) and Ayachi et al. (2022) use the term red-hot, to describe such a huge opportunity for marketers in marketing via the internet.Internet product marketing can be more helpful because the internet allows a more effective marketing process, faster response and lower costs.Low costs and fast information dissemination are expected to increase sales so that they can achieve targeted turnover.Digital marketers can also find out consumer responses to the products offered by looking at the testimonials or comments posted.Good communication can strengthen good relations with consumers.This indirectly will be a separate satisfaction for consumers because consumers feel cared for.When consumers are satisfied, consumers will give positive testimonials and then recommend them to others (Mukaromah et al., 2022;Fahmi et al., 2022).
Recommendations can be made via social media or from word of mouth (WOM).WOM communication is where individuals exchange information, especially about good things, so that they can influence consumer decisions and will make businesses achieve success.This statement is in line with research conducted by Al-Gasawneh et al. (2023) and Barykin et al. (2022) which confirms that WOM conversion rate was 85% and using WOM as a source of information to change decisions is 67%.The effect of WOM is very large, many marketers have even taken advantage of it because it comes from a trusted source.The company is expected to be able to provide good service in providing feedback on every comment post so that consumers feel well served.In making a purchase decision, the quality of service is an important thing to consider.The quality of services provided by the company, which aims to provide convenience for consumers.Every time they carry out business processes, both goods and services, consumers must be given good service because service quality has a close relationship with purchasing decisions (Mittal et al. ,2022).Companies can maintain their business and be able to compete with other competitors if they provide good service and satisfying customer service must be the main mission in the service sector where customer satisfaction is the main priority.This study aims to determine the effect of digital marketing, word of mouth and service quality on purchasing decisions.The aims of this research are: (1) to analyze the effect of digital marketing on purchasing decisions, (2) to analyze the influence of word of mouth on purchasing decisions, and (3) analyze the effect of service quality on purchasing decisions.
Digital Marketing
Digital marketing is a type of activity in marketing that is used to promote or market a product or service and to reach potential customers using digital media.According to Aditi et al. (2023), Asnawati et al. (2022) and Barykin et al. (2022), digital marketing is a type of marketing that is widely used to promote products or services and to reach consumers using digital channels.In the era of globalization, it is very important to implement digital marketing since the scope is broad and easy to use.Research conducted by Asnawati et al. (2022) and Barykin et al. (2022) analyzed the effect of the e-marketing mix strategy on buying decisions at the online store study on zalora.co.idBandung consumers.Research conducted by Alsoud et al. (2023), Ayachi et al. (2022), Asnawati et al. (2022) and Barykin et al. (2022) examine the influence of digital marketing on purchasing decisions.Based on these studies, the formulation of the hypothesis used is: H1: Digital marketing has a positive and significant effect on purchasing decisions.
Word of Mouth
Word of mouth communication is a communication process in the form of providing recommendations both individually and in groups for a product or service that aims to provide personal information.According Haudi et al. (2022), WOM is an effort to market a product or service using viral marketing so that customers discuss, promote, and recommend a product and service to others enthusiastically and voluntarily.This technique is very well used since consumers before buying goods will seek recommendations from friends, relatives or family towards product purchases.Juwaini et al. (2022) reported that WOM communication has a significant effect on purchasing decisions.This is in line with other studies (Chauhan et al., 2023;Csordás et al., 2022) who found that WOM had a positive and significant effect on consumer purchasing decisions.Research conducted by Ginting et al. (2023), Bartschat et al. (2022) and Haudi et al. (2022) examined WOM communication and batik purchasing decisions.The results of the study stated that the factors that most influenced the purchasing decision of Bangkalan Batik were brand awareness and public reception.While the factors that do not affect the purchase decision is WOM communication (Novitasari, 2022;Araujo et al., 2022).Based on this research, the formulation of the hypothesis used is: H2: WOM has a positive and significant effect on purchasing decisions.
Quality of service
Service quality is the expected level of excellence and control over that level of excellence to meet customer desires) Haudi et al., 2022).There are two main factors that influence service quality: expected service and perceived service.The implication is that good or bad service quality depends on the ability of service providers to meet customer expectations consistently (Zhang et al., 2022).Some studies indicate that digital marketing activities and consumer online behavior may increase consumer awareness and purchasing decisions (Chauhan et al., 2023;Csordás et al., 2022).Service quality also has a positive and significant effect on consumer purchasing decisions (Bartschat, et al., 2022;Haudi et al., 2022).Research conducted by Csordás et al. (2022) and Ginting et al. (2023) show that promotion has a positive and insignificant effect, while service quality has a negative and insignificant effect on consumer purchasing decisions.Price, Promotion, Location and Service Quality simultaneously have a significant effect on consumer purchasing decisions.Therefore, we propose, H3: Service Quality has a positive and significant effect on purchasing decisions.
Method
This research method is quantitative, questionnaire using accidental sampling method and it is designed to find out the relationship between the influence of digital marketing, WOM and service quality on consumer purchasing decisions using a quantitative approach method.The variables in this study consist of independent variables and dependent variables.The variables studied include digital marketing, WOM, service quality and consumer purchasing decisions.The research was conducted at SMEs food in Jakarta, Indonesia.Sources of data in this study were primary data including consumer responses to digital marketing, WOM, service quality and purchasing decisions obtained from the results of distributing questionnaires.The population of this research is food SMEs customers.The technique of determining the sample in this study was the accidental sampling method.The sample size used in this study was 680 people.The data collection method used in this study was an online questionnaire distributed by social media.Fig. 1 shows the structure of the proposed study.
Word of mouth
Purchase decision
Results and discussion
The validity test is carried out by correlating the factor scores with the total score and if the correlation of each of these factors is positive (r> 0.3), then the research instrument can be said to be valid Table 2 shows each of the Cronbach's Alpha values for each of these instruments is greater than 0.6 (Cronbach's Alpha > 0.6).This shows that all instruments are reliable so they can be used to conduct research.The purchase decision variable in this study is the dependent variable.The results of validity testing using structural equation modeling (SEM) with the SmartPLS software tools are as follows:
Fig. 2. Validity Testing by SmarPLS
Based on the results of the SmartPLS analysis, the loading factor value for all indicators is more than 0.7, so it can be concluded that all indicators are valid or meet the validity requirements.The multiple linear regression analysis model is used to obtain the regression coefficient which will determine whether the hypothesis will be accepted or rejected.The results of this analysis refer to the results of the influence of the Digital Marketing variable, the WOM variable, the Service Quality variable on Purchasing Decisions at SMEs Product Regional Office VII Denpasar.The results of the regression analysis with the program (SPSS) version 21.0 for Windows can be seen in Table 3 below.Based on The results of hypothesis testing using structural equation modeling (SEM) with the SmartPLS software tools are as follows:
Fig. 3. Hypothesis Testing by SmarPLS
Based on the structural equation modeling (SEM) test with the SmartPLS software tool, the significance level of each variable is obtained at t value > 1.96 and p value < 0.050 so it can be concluded that there is a significant relationship.Based on the test value, the first hypothesis is confirmed.The coefficient of the variable X1 is a positive value of 0.324 meaning that digital marketing has a positive effect on purchasing decisions.If Digital Marketing increases while Word of Mouth and Service Quality remains unchanged, the Purchase Decision is expected to increase.The results of this study are in accordance with research conducted by Ginting et al. (2023), Bartschat et al. (2022), Haudi et al. (2022) and (Zhang et al., 2022).Research conducted by Asnawati et al. (2022) and Barykin et al. (2022) analyzed the effect of the e-marketing mix strategy on buying decisions at online stores and found a positive and significant effect.
Our results have also confirmed the second hypothesis.The coefficient of the variable X2 is positive 0.170, meaning that WOM has a positive effect on Purchase Decision.If WOM increases while Digital Marketing and Service Quality remain unchanged, the Purchase Decision is expected to increase.The results of this study are in accordance with research conducted by Alsoud et al. (2023) and Purwanto et al. (2023).
Finally, our results have also confirmed the third hypothesis meaning that service quality has a positive effect on purchasing decisions.If the Quality of Service increases while Digital Marketing and WOM remain unchanged, then the Purchase Decision will increase.The results of this study are in accordance with research conducted by Bartschat et al. (2022) and Haudi et al. (2022).
In digital developments, Indonesian people who were used to shopping directly at stores have changed to do their shopping using digital media or shopping online.The ease of transactions, ordering and searching for information makes people more interested in shopping online.So that every company must be able to adjust its marketing strategy by moving to online media or commonly called digital marketing.In fact, the more digital media is used by consumers, the easier it will be for companies to achieve their targets.Digital marketing has a good impact on companies.Digital marketing influences consumer buying interest and consumers feel motivated to buy products as a result of digital marketing carried out by companies.Consumers consider digital marketing to be more informative in explaining products, easy to browse and more attractive.For this reason, companies can take advantage of digital marketing to attract potential customers.
The research found that there was a positive relationship between customer satisfaction and WOM communication from one customer to another customer to visit a fast food seller.Furthermore, Oscarius Yudhi Ari Wijaya (2021) also used WOM but through electronic media to determine the magnitude of the influence on brand image and purchase intention.In the context of online electronic word of mouth (e-WOM), consumers are very interested in writing, sharing experiences, and reading pleasant and unpleasant experiences in virtual communities.Consumer reviews are a source of information.The purchase decision is preceded by a desire.Desires or intentions are influenced by many factors including information.The information consumers need is obtained through word of mouth (e-WOM) conversations.Electronic Word of Mouth (e-WOM) is a form of peer-to-peer non-commercial interpersonal communication on the internet, where one person can share positive (or negative) things about a product or service with another person.Based on the results of the research and data analysis that has been done, it can be concluded that digital marketing variables have a positive and significant influence on consumer buying interest, even though the effect is low because the rest is influenced by other variables that are not included in this research model.The website is an indicator of the most influential digital marketing variable in increasing consumer buying interest.Digital marketing creates added value for consumers, which in turn has a positive effect on consumer purchase intentions.Companies that have adapted to digital marketing experience increased business revenue, reduced marketing costs and increased visibility in the digital ecosystem.All of this is reflected in increased consumer purchase intentions for digitally marketed banking products.Nonetheless, further research is needed to examine whether the purchase intent generated from digital marketing is permanent, sustainable, and can be replicated to other parts of marketing, or even other companies (De Guzman et al., 2022).
Service quality is needed for a company because if a company does not provide optimal service and does not meet the expectations of consumers, it causes consumers to be unsatisfied, if they maximize their services, consumers will repurchase service quality as a level of excellence as expected from consumers.If a company can make a standard, namely the quality of its service is able to keep up with consumer expectations, it will be large that consumers will make repeat purchases since the company's quality image is not assessed by the company itself but based on the consumer's point of view because they enjoy the company's products.However, a service cannot be assessed if the consumer has not experienced a product from a company, so the components to be assessed include reliability, responsiveness, assurance, and empathy.While the price is related to repurchase intention, the price is used as an alternative for consumers to buy.If a company can provide the right price or in a sense according to consumer expectations in viewing its product, it is not impossible that repurchasing interest will increase.Companies also must consider the benefits they get offset by competitive competition in the market.Price indicators include the price offered in accordance with product quality, affordable prices, price compatibility with benefits and finally price competitiveness.
The theoretical implications of the results of this research provide evidence for the development of consumer behavior and marketing science, especially regarding Digital Marketing, Word of Mouth, Service Quality and Purchase Decisions.Thus, the results of this study provide empirical support and can be stated to strengthen the results of previous studies.In addition, the results of this research can practically be a reference for other researchers who want to do some studies on Digital Marketing, WOM, Service Quality and Purchasing Decisions.Theoretically, this research also provides an understanding that Digital Marketing, Word of Mouth and Service Quality can significantly increase Purchasing Decisions.By increasing Digital Marketing, Word of Mouth and Service Quality, Purchasing Decisions will increase.
Conclusion
The results of this study have indicated that there was a positive and significant influence between Digital Marketing and Purchase Decisions.This means that the use of Digital Marketing can improve Purchase Decisions.There is a positive and significant influence between WOM on Purchasing Decisions.This means that the use of WOM can improve Purchasing Decisions.There is a positive and significant influence between Service Quality on Purchasing Decisions.This means that the use of better Service Quality can improve Purchasing Decisions.Based on the limitations contained in this study, the researcher proposes a suggestion that SMEs should further optimize the use of Digital Marketing in marketing their company, such as uploading interesting content on one of the existing social media.From a WOM point of view, companies must promote more to their closest circle of benefits that can be obtained from using SMEs products.From Service Quality, we must continue to provide excellent service to consumers or customers so that these consumers feel comfortable and satisfied.
For future researchers, it is hoped that future researchers will not stick to the factors in this study, namely Digital Marketing and WOM, but can add other factors that might influence Purchasing Decisions.Future researchers are expected to be able to add or use other research subjects.
Fig. 1 .
Research Model . A valid instrument is an instrument that can be used to measure what should be measured.The level of validity indicates the extent to which the data collected does not deviate from the description of the variable in question.The results of the validity test of this study are shown in Table1 below: Table1shows that all the correlation coefficients of the variable indicators tested have a value greater than 0.30 (r > 0.3).These results indicate that all the indicators contained in this study proved to be valid.The reliability test is used to measure the consistency of a variable in research.The instrument is said to be reliable if it has a Cronbach Alpha value > 0.60.The results of the reliability test of this study are shown in Table2 below:
Table 3
The constant value of -0.098 indicates that if Digital Marketing, Word of Mouth and Service Quality are equal to 0 (zero), then the Purchase Decision decreases.X1 = +0.319indicatesthatDigital Marketing has a positive effect on Purchase Decisions, if Digital Marketing increases, Purchase Decisions will increase.X2 = +0.181indicatesthatWord of Mouth has a positive effect on Purchasing Decisions, if Word of Mouth increases, Purchasing Decisions will increase.X3 = +0716, indicating that service quality has a positive effect on purchasing decisions, if employee service quality increases, purchasing decisions will increase Determination analysis was carried out to determine the extent to which the independent variables varied, namely X1 (Digital Marketing), X2 (Word of Mouth) and X3 (Quality of Service) on the Purchase Decision variable (Y).Based on the SPSS results which can be seen in Table4.
Based on these results it is known that the value of R2 = 83.8percent, which means that 83.8 percent of decisions are influenced by the variables Digital Marketing (X1), Word of Mouth (X2), and Service Quality (X3) and the remaining 16.2 percent is influenced by other variables not examined in this study.The F test is used to determine whether simultaneously (simultaneously) all independent variables (Digital Marketing variables, Word of Mouth variables, and Service Quality variables), have an influence on the dependent variable. | 2023-07-11T02:35:11.407Z | 2023-01-01T00:00:00.000 | {
"year": 2023,
"sha1": "58bd72fab0ad1fb8608fc4e4d4e58b413d37d4c3",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.5267/j.ijdns.2023.5.012",
"oa_status": "CLOSED",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "a43e09307e048a38b0e1afc5d94babaa7175a27b",
"s2fieldsofstudy": [
"Business"
],
"extfieldsofstudy": []
} |
260359516 | pes2o/s2orc | v3-fos-license | Ethnobotanical study of medicinal plants used by the population of Ain Chkef (North central Morocco)
Background: Medicinal and aromatic plants (MAPs) use in the folk medicine is widely spread in Morocco. This work aims to study their use in Ain Chkef; a rural area in North central Morocco. It also documents multiples data concerning the ethno-medico-botanical traits of the most cited species. Methods: An ethnobotanical survey was carried out through the interview of 183 individuals. The responses were analyzed using several plant citation indexes (RFC; FIV; PPV and FL). Correlations between plants and humans of different ages, genders, educational levels, and marital status, were investigated by multiple components analysis (MCA) performed using XLSTAT software. Results: 93.44% of the studied population use MAPs. Among the 48 used species, belonging to 27 botanical families, the most frequently cited were Origanum compactum Benth . Chenopodium ambrosioides L, Cuminum cyminum L., Mentha pulegium L. Leaves were the dominant part used. The infusion and decoction were the principal preparation methods. MCA showed that people not or low using MAPs (< 3) are sharing common traits: Male, not married, high educational level, < 40 years aged. People highly (> 7) or moderately (4 to 7) using MAPs are mainly: Female, married, with low educational level, >40 years aged. Conclusion: The present study highlighted the spread use of MAPs among the interviewed population in Ain Chkef as traditional remedies. Thus, this work will provide researchers with an important ethnobotanical database which can be exploited in the development of pharmacognosy.
Background
In spite of the development of medicine and the availability of modern therapies, an important interest to traditional pharmacopoeia remains remarkable (Abouri et al. 2012).
Medicinal and aromatic plants (MAPs) are widely used in the preparation of herbal remedies especially in rural areas (Abouri et al. 2012), since they are affordable, easily accessible and own an efficient healing potential (Al-Adhroey et al. 2010).
Morocco is one of countries where the art of healing diseases using MAPs has reigned (Jouad et al. 2001).Indeed, thanks to its geographical position (between Sahara in the South, the Atlantic Ocean on the West and the Mediterranean Sea in the North), Morocco presents very rich and diversified ecosystems (Ouhaddou et al. 2014).Moreover, Moroccan flora is characterized by more than 7000 species and subspecies including about 800 species of MAPs (Ouhaddou et al. 2014).
The knowledge of ethnomedicinal uses of MAPs is extremely important since it constitutes the first pillar of ethnopharmacology (Süntar 2020).In fact, based on traditional uses of MAPs as remedies, many researchers have demonstrated their pharmacological properties as well as their richness in bioactive compounds (Rates 2001), which can therefore be used for the development of natural-based drugs (Zareef et al., 2023).
In order to contribute to ethnopharmacology and to preserve Moroccan heritage regarding the preparation of traditional remedies based on MAPs, several ethnobotanical studies have been conducted in different regions of Morocco (Abouri et al. 2012, Bellakhdar 1997, Jeddi et al. 2021, Sijelmassi 1993, Tahraoui et al. 2007), generally demonstrating the wide use of MAPs in the studied areas.In order to contribute to this research effort, we carried out this work focused on the Ain Chkef area (north central Morocco), where to the best our knowledge, no ethnobotanical survey has been done.The main objectives of the present work were to explore the extent of use of MAPs in this area; to identify the MAPs species used, to determine the diseases treated by these MAPs, and to investigate the existence of determining relationships between those collected data and diverse socio-demographic characteristics (age, gender, educational level...).The study was completed by a bibliographic review documenting the described bioactive molecules and pharmacological properties of the most used MAPs.Furthermore, this work matches with the current global concern related to sustainability.It corresponds to the research of nature-based solutions and the valorization of natural resources and ancestral knowledge linked to human wellbeing.
Study area
Ain Chkef is a rural commune located in the Fez-Meknes region (Figure .1), precisely in the Moulay Yaacoub province (Kingdom of Morocco) (HCP Morocco 2012).The geographical coordinates of its centre are 33° 57' N, 5° 1' 41'' W, with an altitude of 499 m and a Mediterranean climate (Db-city.com 2023).Ain Chkef (a peri-urban area of Fez city) is covering an area of 146.352Km 2 with a population density of 238 inhabitants / km 2 (Ministry of urban planning and territory development (MUPTD), 2014).This population benefits from easy access to superior services through the relatively developed urban network of the metropolis (MUPTD 2014).
In addition to the agricultural potentialities residing in the production of cereals and fodder (MUPTD 2014), the studied area is also distinguished by the planted forest Ain Chkef; a green corridor of 60 ha sheltering diverse native and exotic plant species (Benamar 2011).
Data collection
The ethnobotanical survey was carried out at Ain Chkef area.
This study is based on the completion of a questionnaire concerning the socio-demographic profile of the people surveyed (age, gender, level of education, marital status and socio-economic level) on the one hand and the plants they use for medicinal purposes (scientific and local name, parts used, targeted diseases, preparation methods and doses) on the other hand.
The study sample consisted of respondents randomly interviewed at the entrance and exit gate of the main weekly popular market (souk), in the Ain Chkef area.The total number of people interviewed was 183, and the time spent with each interviewee varied between 15 and 20 minutes.
The protocol followed in this work was previously validated by the Council of our Laboratory and applied in several published studies (Jeddi et al. 2021, El Hachlafi et al. 2020, Benkhaira et al. 2021).
Plant species identification
The plant species used by the interviewed population were identified with the help of Moroccan botanical books (Bellakhdar 1997, Fennane et al. 1999, Sijelmassi 1993) and through the consultation of other ethnobotanical studies carried out in Morocco (El Hachlafi et al. 2020, Jeddi et al. 2021, Ouhaddou et al. 2014).
Figure 1.Geographical location of the studied area in Morocco.
Data analysis
Collected data were statistically analyzed using Microsoft Office "Excel 2013".Socio-demographic collected data of respondents were analyzed by descriptive statistics (frequencies and percentages).Concerning ethnobotanical data, they were interpreted through quantitative citation indexes: RFC, FIV, PPV and FL.
Moreover, in order to understand the structure of relationships among plants and humans differing by their age, educational level, gender and marital status, multiple component analysis (MCA) was performed by XLSTAT software.
Relative frequency of citation (RFC)
This index reveals the relative importance of each MAPs species mentioned by respondents.It is calculated by dividing the number of people who mentioned the use of the species (FC), by the total number of respondents using MAPs (N), according to Tardio and Pardode-Santayana (2008) formula: RFC=FC/ N With: 0 < RFC < 1
Family Importance Value (FIV)
FIV shows the importance of the families used.It is designed to evaluate the biological taxonomic value of plants; it is calculated by the following formula (Sreekeesoon & Mahomoodally 2014):
FIV=FC family / Ns
Where FC family (FC family=RFC) refers to number of informants citing the family, and Ns is the total number of cited species in each family.
Fidelity level (FL)
Fidelity level reveals the efficiency of a plant species against a particular disease.It was calculated according to the following formula according to Sreekeesoon & Mahomoodally (2014): FL=IP/ Lu x 100 IP: refers to the citations number of a given plant species used in the treatment of a particular disease.Lu: indicates the total number of citations of this species used for the treatment of any disease.
Plant part value (PPV)
Plant part value indicates the use frequency of each plant's organ.It is calculated following this formula (Gomez-Beloz 2002):
PPV=RU plant part/ RU
RU plant part: the sum of uses reported per a given plants part.RU: the number of uses reported for all plant parts.
Multiple component analysis (MCA)
Multiple component analysis was performed by XLSTAT software to understand the structure of relationships among plants and humans differing by their age, educational level, gender and marital status.Since the MCA is generally applied for qualitative variables, the age of respondents and the number of MAPs they use are quantitative variables, they were converted on to qualitative variables using different classes: -For the age: -Age-Low: Age < 20 years.
-Age-High: Age > 40 years.-For the number of MAPs used: -MAPs used zero: People not using MAPs.
-MAPs used-High: MAPs used > 7. Concerning the educational level (EL), in the MCA, the university level wasn't considered among the modalities of this variable, because of the low number (4) of informants corresponding to this modality.These four cases were added to the number of the secondary level informants.
Two parameter's Kolmogorov-Smirnov test
A statistical analysis of the population distribution was carried out in the context of the comparison of informant populations using and not using MAPs, grouped into age, sex, educational attainment, socio-economic and marital status categories.The two-parameter Kolmogorov-Smirnov test was performed using the XLSTAT software, for each category, to determine whether the two samples came from a population of the same distribution (equality hypothesis accepted) or not (equality hypothesis rejected) at p = 0.05.For each category, the samples correspond to the frequency of each of the two parameters (MAPs use or not) for informants of a given age between 12 and 74 years (62 values).For the «age» category, the sample is divided on two parts (< 42 years and ≥ 42 years) of 31 values each one.
Review study
In order to highlight the ethnomedicinal uses of the most cited MAPs species, a literature review was performed regarding the pharmacological properties of these species, as well as their traditional uses reported in other ethnobotanical studies.Data bases used for this purpose were: PubMed, Science Direct, Scopus, and Google Scholar.In addition to the name of the species, the keywords used for this research were: 'ethnomedicinal uses', 'biological activities', and 'pharmacological properties.
Chemical structures of different bioactive compounds of these species were drawn using the ChemDraw software.
Use of MAPs by respondents
In the studied area, only 6.55% of respondents have declared not using MAPs for medicinal purposes, while the majority (93.44%) have stated that they use MAPs for the treatment of various ailments (Figure 2).This result could be explained by the attachment of the studied population to its traditional heritage regarding the use of MAPs (Benamar et al. 2023).
Figure 2. Distribution of respondents according to their statements regarding their use or not of MAPs.
Socio-demographic data of respondents Age
As shown in figure 3, the use of MAPs in the studied area is widespread among people in the age group > 40 years (47.95%),followed by those who are between 20 and 40 years (41.52%).While the lowest percentage of MAPs use is noted among young people < 20 years (10.52%).Our results are similar to those obtained in other ethnobotanical studies conducted in different regions of Morocco (Jeddi et al. 2021, Kachmar et al. 2021).The dominance of the use of MAPs by the elderly could be explained by the fact that these people have a better knowledge and highest conviction of the efficiency of the traditional medicinal use of MAPs (Benkhaira et al. 2021).
Figure 3. Distribution of respondents using or not MAPs according to their age.
Gender
The use of MAPs in the Ain Chkef area varies according to informant gender (Figure 4); indeed, women (84.21%) use them more than men (15.78%).This could be attributed to the fact that women have more knowledge about MAPs, their preparation and administration modes, in comparison with men.Similar findings were observed in other ethnobotanical studies at national scale (Alaoui et al. 2018, Bencheikh et al. 2021, Jeddi et al. 2021).As for people not using MAPs, women and men represent almost the same percentage: 58.33% and 41.66% respectively (Figure 4).
Educational level
In the studied area, among the medicinal plant users, the majority (57.89%) are illiterate, while 23.39% of them have a primary educational level and 16.37% have a secondary level.The lowest percentage (2.33%)correspond to users having a university level (Figure 5).Similar results were obtained in Morocco by Bencheikh et al. (2021), El Hachlafi et al. (2020) and Jeddi et al. (2021).As regard people not using MAPs, they are composed equally (50%) of illiterate and educated individuals.
Socio-economic status
In Ain Chkef area, most of people using MAPs have a medium socio-economic level (81.28%) followed by those with a low level (36%), while only 1.75% have a high level (Figure 6).This may explain the use of herbal medicine by the indigenous population in the treatment of diseases as an accessible and effective mean.Our findings are consistent with those obtained in other ethnobotanical surveys (Benkhaira et al. 2021, Jeddi et al. 2021).
As for people not using MAPs, they are composed of only 16.66% of individuals having a low socio-economic level, while the majority of them (83.33%) have a medium level.This result suggest that the latter may be users of modern medical treatments in spite of their cost.Figure 6.Distribution of respondents using or not MAPs according to their Socio-economic level.
Marital status
According to the present ethnobotanical survey (Figure 7), MAPs are used much more by married people (81.28%), than by single people (18.71%).This could be explained by the fact that families with medium or low incomes tend to use MAPs in order to reduce or avoid the high costs of medical visits and treatments (Benkhaira et al. 2021, Jeddi et al. 2021).Regarding people not using MAPs, they include married and single persons with percentages of 66.66% and 33.33% respectively.
Figure 7. Distribution of respondents using or not MAPs according to their marital status.
Multiple component analysis of collected data
The results of the MCA are presented in figure 8. Overall, the two axes F1 and F2 account for 41.06 % of the total variation in our data.F1 presents 25.73 % of the total variation and reveals two categories of people according to their use of MAPs: people not using MAPs (MAPs used zero), or with a low use of MAPs (MAPs used < 3) and people with a medium (4 < MAPs used < 7) or a high (MAPs used > 7) use of MAPs.The second axis F2 accounts for 15.34 % of the total variation and separates people according to their gender (male and female).It can be noted from this figure that people with zero or low use of MAPs are sharing common traits: marital status (MS-Single), educational level (EL-Secondary), age group-Low and gender-Male.As regard people with a medium or a high use of MAPs, they have other characteristics: MS-Married, EL-illiterate and EL-Primary, Age-Medium / Age-High and gender-Female.These findings are suggesting the following explanations: People with low educational level who are associated with a medium or high use of MAPs (Figure 8) are tightly attached to their traditional heritage and have high trust on the efficiency of MAPs.However, it must be noticed that these people are probably not aware of MAPs danger when used at high doses.As for people with higher educational level (EL-Secondary), they are associated with zero of low use of MAPs, may be because they are more conscious of health risks that can result from the random use of MAPs.The correlation between the medium and the high use of MAPs with people of medium or high age can be attributed to the high knowledge of these persons of MAPs uses modes in the treatment of several diseases and their attachment to their traditional heritage, unlike the case of young people.Regarding the gender, the MCA allows also to note that women are associated with medium or high use of MAPs unlike men.This can be justified by the fact that women have more expertise in the preparation of different MAPs recipes that are appropriate for the treatment of various ailments.
Informant population's socio-demographic and cultural traits analysis by Kolmogorov-Smirnov test
Although the current ethnobotanical study was based on a randomized sampling of informants, we considered interesting to supplement the MCA results by a statistical comparison of the distribution of the different population categories according to the frequency of their use or non-use of MAPs at each age between the minimum and maximum encountered (12-74 years).The results of this analysis performed by the two parameter Kolmogorov-Smirnov test, presented in Table 1, show that the hypothesis of equal distribution of the populations of informants using or not the MAPs (MAPs-using and MAPsnot-using) was rejected for three categories based on age, gender and educational level, which is in conformity consistent with the MCA results (Figure 8) showing for example that Medium or High-age is associated with other characteristics of the population highly or moderately using MAPs, particularly gender-Female; MS-Married; EL-illiterate and EL-Primary.This means that populations using MAPs in these categories could not have an an age distribution equal to that of populations not using MAPs.In contrast, the Kolmogorov-Smirnov test shows that the hypothesis of equal distribution of populations using or not MAPs is accepted for "low socioeconomic level" and "unmarried" categories.
Thereby, these results corroborate with those of multiple component analysis, which support the findings that age, gender, marital status, socio-economic and education levels are important demographic variables affecting the use of herbal medicinal plants, in the studied area.
Floristic analysis
The ethnobotanical survey conducted in Ain Chkef area has allowed us to identify 48 species of MAPs, belonging to 27 botanical families, used for the treatment of various ailments.
All the plants listed are presented in table 2 where are indicated for each of them: The scientific name of species and botanical families, the vernacular name, parts used, targeted diseases, preparation method and the data related to FC, RFC and FIV; for each of them.Among the 27 botanical families cited by respondents, the most representative are Chenopodiaceae (FIV=0.251),Verbenaceae (FIV=0.181),Lamiaceae (FIV=0.110),Apiaceae (FIV=0.078)(Figure 9).Other studies have also showed the preponderance of ethnomedicinal use of plant species belonging to these families (Alaoui et al. 2018, Chaachouay et al. 2020, El Hachlafi et al. 2020).
This shows the importance of these plant species in the treatment of various ailments such as diabetes, osteoarticular diseases, fever, stomach disorder, cold, cough etc (Table 2).
Figure 10.Distribution of the most cited species according to their Relative frequency of citation (RFC).
Ethnopharmacological plants traits
Several in vitro and in vivo studies have previously demonstrated the pharmacological potential of MAPs used in the studied area.Table 3 shows different ethnomedicinal uses reported in other ethnobotanical works of the most cited species.Moreover, it highlights their pharmacological properties according to different scientific studies.Chemical structures of the main bioactive compounds present in these species are given in Figure 11.-Pyelonephritis, kidney stones (Bencheikh et al. 2022).
-Diabetes, cardiovascular diseases and pathologies of the digestive system (Bouyahya et al. 2017).
Preparation mode
In order to extract and benefit from bioactive compounds present in MAPs, several preparation methods are used by the studied population.According to figure 13, Infusion (38.20%) and decoction (33.26%) represent the most used preparation methods, followed by other modes; raw (23.39%), cataplasm (3.43%), maceration (0.86%), fumigation (0.64%) and cooked preparations (0.21%).The frequent use of infusion and decoction can be explained by the fact that these two methods allow the extraction of the maximum of active substances and reduce or cancel the toxic effect of plant remedies (Benaiche et al. 2019, Salhi et al. 2010).
Figure 13.Frequency of preparation methods of the used MAPs by respondents.
Dose used and duration of treatment
Concerning doses of MAPS used, among informants, 74.51% use spoons of MAPs, 18.57% use handle and 0.86% use pinch (Figure 14).As for the duration of treatment, the majority of people (85.96%) use MAPs for one day, whereas 8.42% use them until recovery and 5.62% for one week (Figure 15).These results revealed that MAPs are used randomly in the studied area with different quantities.Unfortunately, this can threaten life of the indigenous population, since the administration of phytochemicals at unspecified and non-rational doses may lead to adverse effects on human health (Benkhnigue et al. 2010, Jeddi et al. 2021).
The lowest FL value was obtained for Marrubium vulgare L. (FL=26.31%).This can be explained by the multiple uses of this species by the studied population for the treatment of different ailments.Moreover, Marrubium vulgare L. has been found according to many studies to exhibit a multitude of pharmacological properties, such as antioxidant, hepatoprotective, antiproliferative, immunomodulatory, antidiabetic and anti-inflammatory activities (Aćimović et al. 2020).
Table 4. Efficiency of the most used plants species by the respondents against a given disease according to their FL.F: fever; SD: stomach disorder; NSD: nervous system disease; H: headache; DT: diabetes; CO: cold; OD: oral disease; KD: kidney.
Conclusions
The use of MAPs plays a very important role in the therapy of the indigenous population of Ain Chkef area and is considered as an alternative to synthetic medicines.In fact, 93.44% of people resort to MAPs for the preparation of traditional remedies.Among MAPs used, 48 species were identified belonging to 27 botanical families.
Figure 4 .
Figure 4. Distribution of respondents using or not MAPs according to their gender.
Figure 5 .
Figure 5. Distribution of respondents using or not MAPs according to their educational level.
Figure 11 .
Figure 11.Chemical structures of the main bioactive compounds of MAPs having the highest RFC values.
Figure 12 .
Figure 12.Plant parts used by respondents in treatment of various ailments.
Figure 14 .
Figure 14.Used doses of MAPs for the preparation of traditional remedies by respondents.
Figure 15 .
Figure 15.Duration of treatment using MAPs by the studied population.
Table 1 :
Results of the two parameters Kolmogorov-Smirnov test applied to the datasets of different categories of informant's populations represented in figures 3-7.The nul hypothesis was that the two samples come from a population of the same distribution (p = 0.05).
Table 3 .
Literature data on ethnomedicinal uses and pharmacological properties of the most cited species in this study.
various ailments according to literature Pharmacological properties / Main bio-active compounds
As regard relationships among plants used and population studied, people with no or low use of MAPs (MAPs used < 3) are sharing common traits: marital status (MS-Single), educational level (EL-Secondary), age group-Low (< 40 years) and gender-Male.Whereas people with a medium or a high use of MAPs (4 < MAPs used < 7 and MAPs used > 7 respectively), have other characteristics: MS-Married, EL-illiterate and EL-Primary, Age-Medium (40 < Age < 60 years) and Age-High (Age > 60 years) and gender-Female.The most cited plant species are Origanum compactum Benth.(RFC=0.538),Chenopodiumambrosioides L. (RFC=0.251),Cuminumcyminum L. (RFC=0.233),Menthapulegium L. (RFC=0.222)and the dominant families are Chenopodiaceae (FIV=0.251),Verbenaceae(FIV=0.181),Lamiaceae(FIV=0.110),Apiaceae(FIV=0.078).Leaves are the most used plants parts (PPV=52.08%)whileinfusion (38.20%) and decoction (33.26%) are the dominant preparations methods.Doses of MAPs are not precise and vary between people.Several species were reported to be effective against a given disease, among which eight species present a FL of 100%: Allium sativum L., Artemisia herba alba Asso, Camellia sinensis (L.) Kuntze, Carum carvi L., Coriandrum sativum L., Cuminum cyminum L., Salvia officinalis L., Trigonella foenum graecum L.In conclusion, the results of this study enabled to discover and illustrate traditional medicinal knowledge of the population of Ain Chkef.Besides, they constitute an important ethnobotanical database which can be exploited by researchers in the development of herbal medicines.Furthermore, this article matches with several sustainable development goals (SDGs) such as SDG 3, 11, and 15 (https://fr.unesco.org/sdgs). | 2023-08-02T15:12:20.128Z | 2023-07-30T00:00:00.000 | {
"year": 2023,
"sha1": "4c079c4af060dfd63818a0e79a4d82e543b6061a",
"oa_license": null,
"oa_url": "https://ethnobotanyjournal.org/index.php/era/article/download/4667/1725",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "069c26ad8478b5b1080f4b4c3189afad4a52bdbb",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": []
} |
252544902 | pes2o/s2orc | v3-fos-license | A two-parameter deformation of the quasi-shuffle\\ and new bases of quasi-symmetric functions
We define a two-parameter deformation of the quasi-shuffle by means of the formal group law associated with the exponential generating function of the homogeneous Eulerian polynomials, and construct bases of $QSym$ and $\WQSym$ whose product rule is given by this operation.
Introduction
The Hopf algebra of quasi-symmetric functions [5] is historically the first example of a combinatorial Hopf algebra extending that of symmetric functions. While there is no general agreement on what should be the formal definition of a combinatorial Hopf algebra, its is clear that to be considered as combinatorial, such an algebra should, in addition to combinatorial product and coproduct rules, also have at least two bases related by a combinatorial rule.
Quasi-symmetric functions entered the scene with these requirements, having from the beginning two bases M I (monomial) and F I (fundamental). The product rule for the M I is the so-called quasi-shuffle (apparently first described in [28]), and its coproduct is given by deconcatenation.
At the time, the algebra of symmetric functions had already a lot of known bases (elementary, complete, Schur, power-sum, Hall-Littlewood, Jack, Macdonald etc., see [15]), and the subsequent realization of the dual of QSym as noncommutative symmetric functions quickly led to the introduction of a lot of new bases on both sides (see e.g., [12,7,23,14,25,27,18,24,21] and the survey [16] for a sample).
Originally motivated by the study of posets, quasi-symmetric functions were next related to the combinatorics of descents in permutations and to the Solomon descent algebra, then to the 0-Hecke algebras and certain degeneracies of quantum groups [13]. More recently, they played a crucial role in the theory of Macdonald polynomials, via the quasi-symmetric expansions of certain LLT polynomials [6].
Another source of interest in quasi-symmetric functions stems in the theory of multiple zeta values (MZV), which are the specializations x n = 1 n of the monomial functions M I . Thus, the MZV satisfy a product formula given by the quasi-shuffle, but their representation as iterated integrals naturally labelled by binary words leads to another expression of the product, given by the ordinary shuffle of binary words.
The equality between both expressions encodes algebraic relations satisfied by the MZV [29].
A natural question is therefore: does there exist bases of QSym naturally labelled by binary words which multiply by the ordinary shuffle of binary words? This question has been answered affirmatively in [18], where a general method for constructing such bases has been presented, and a special case investigated more in depth.
Recently, a variant of the quasi-shuffle called the block shuffle has been introduced by Hirose and Sato, also in relation with the MZV. It has been proved by Keilthy [11] that the block shuffle algebra is isomorphic to the ordinary shuffle algebra, providing an isomorphism analogous to Hoffman's exponential [8], where the exponential is replaced by the hyperbolic tangent.
Actually, Hoffman's classical proof of the isomorphism between shuffle and quasishuffle algebras is equivalent to the fact that QSym admits bases whose product rule is given by the ordinary shuffle of compositions 1 . The existence of such bases comes itself from the fact that Sym is the free graded associative algebra over a sequence of primitive elements.
Once more, a natural question is now to construct bases of quasi-symmetric functions whose product rule is given by the block shuffle. The existence of such bases is implied by the result of [11]. We shall reprove this result in a simpler way, starting on the dual side with a sequence of primitive generators, and investigate the bases obtained by a particular choice of these generators, corresponding to the Solomon idempotents of the descent algebra.
For a word w = a 1 ⋯a n over the alphabet of positive integers, and an integer x, define ζ x (w) = (a 1 + x)a 2 ⋯a n , where a 1 + x is the sum of integers. We shall see that the ordinary shuffle, the quasi-shuffle and the block shuffle are special cases of a two-parameter family of associative products defined by the recurrence associated with the formal group law This is proved by introducing a basis of QSym satifying this product rule. This basis is then lifted to WQSym.
The block shuffle
The block shuffle of words over the alphabet of positive integers is defined by where ǫ is the empty word, and for letters a, b where ζ a (bv) = (a + b)v (a + b is the sum of the integers) and ζ a (ǫ) = 0. Suppose that X I is a basis of QSym satifying X I X J = X(IˆJ) (where X is regarded as a linear map on the vector space spanned by compositions, and X(I) ∶= X I ). Let Y I be the dual basis of X I . From the product formula of the X I , we can calculate the coproduct of Y I . In particular, to get ∆Y n , we see that X n can appear in X I X J if and only if ℓ(I) − ℓ(J) = 1, and that its coefficient is a sign (−1) r , where r = 1 2 (ℓ(I) + ℓ(J) − 1). Hence, Assuming that Y I = Y I is a multiplicative basis, and setting Y = ∑ n≥1 Y n , this can be rewritten as Starting with a sequence Π n of primitive generators of Sym, and setting we have from the addition formula is the generating series of (5). Thus, defining Y n in this way does indeed yield a basis of Sym whose dual basis X I mutiplies by the block shuffle.
A special choice
Let us now choose Π = ϕ, where Thus, are the noncommutative Eulerian polynomials of [4,20]. Therefore, with the notation of [4,20] (18) n , and using [22, (16)], we can compute Lemma 3.1. The coefficient of A(n, r) in the above sum is Proof -One easily checks that This equation is also obviously satisfied by the right-hand side of (23). Moreover, the initial conditions are Finally, we arrive at the following expression: On the ribbon basis, or, equivalently For example, Actually, the sums F (n, k, r) are the coefficients of the "amazing matrix" of [10]. It is proved in this reference that
Some transition matrices
The construction of the basis X I is a special case of the following one. Start with a multiplicative basis ϕ I of Sym, for example ϕ = log σ 1 . Choose a formal series f (z) = ∑ n c n z n , and define Y ∶= ∑ n Y n = f (ϕ). We have then Let φ I be the dual basis of ϕ I and X I be the dual basis of Y I . By duality, (33) translates as By construction, the coproduct of X I is given by deconcatenation, since it is dual to a multiplicative basis, so that the map Ψ f ∶ φ I ↦ X I is a morphism of coalgebras. It is proved in [3] that all morphisms of coalgebras are actually of this type [3,Th. 2.2] and that more generally the group of coalgebra automorphisms of a shuffle algebra is isomorphic to the group of invertible formal power series in one variable under composition (in particular, This construction has been rediscovered in [9], where the following convenient notation has been introduced. If we encode compositions by words over an auxiliary alphabet Z = {z i i ≥ 1}, endowed with the operation z i ◇ z j ∶= z i+j , and define a linear operator P by P (z I ) ∶= P I for any basis P I of Qsym, (35) can be recoded as in [9] (36) where z = ∑ i z i and f ◇ (z) means that powers of z are evaluated with the ◇ operation 2 If we choose f such that c 1 = 1, we have X n = φ n . We can then write Indeed, the l.h.s. is the Cauchy kernel for the pair of bases (φ, ϕ) specialized at the virtual alphabet Λ defined by ϕ n (Λ) = λ for all n. Then, and (39) In the notation of [9,11], this would be (40) X(exp * (λz)) = φ 1 1 − λz where * is the product rule for the X-basis, for example the quasi-shuffle if f (z) = e z and the block shuffle of [11] if f (z) = tanh z.
Generalisation
In order to obtain interesting bases, the series f (z) should have an addition theorem of the type F is a formal group law). Tractable examples are rather scarce. To the list e z , tanh z and tan z, one can add and provides a deformation of the quasi-shuffle [3] satisfying the recurrence These examples can be interpolated by the (known) formal group law This is the exponential generating function of the homogeneous Eulerian polynomials where r(σ) is the number of rises of σ and d(σ) its number of descents.
Define Y n by Y = f (ϕ). Proof -By definition, and since whence the result.
The expansion of the ribbon basis follows from the identity Another expression can be obtained as above in terms of the amazing matrix. Imitating the calculation of Section 3, we have (treating s − t as a scalar) Theorem 5.2. Let X I be the dual basis of Y I . Then, where ⋆ is defined on words over the integers by the recursion In particular, the operation ⋆ is associative and commutative.
Proof -By (43), the coproduct of Y is Writing compositions as words to be in line with the notation of the theorem, we have where the map ∂ x Y a 1 a 2 ⋯ar ∶= Y (a 1 −x)⋅a 2 ⋯ar is the adjoint of ζ x , so that finally In all other cases, it is clear that ⟨X au X bv , Y cw ⟩ = 0, whence the recurrence (56).
Clearly, X 1 = φ 1 = F 1 = M 1 is the sum of the variables. It is always interesting to expand its powers on a new basis. Since φ 1 n = φ n 1 , it follows from (35) that For I = 1 n , X 1 n is a symmetric function. The coefficients of its monomial expansion are One can give a closed formula for the product X 1 p X 1 q .
Proposition 5.3. For a composition I, let ℓ 0 (I) denote its number of even parts, and ℓ 1 (I) its number of odd parts. Then Proof -By duality, Now, if p = q, so that n = p + q is even, β n− 1 2 if p − q = 1, so that n = p + q is odd.
Thus, the coefficient of X I is X 1 p X 1 q is equal to the number of 2 × r integer matrices (80) p 1 p 2 ⋯ p r q 1 q 2 ⋯ q r with row sums p, q, column sums i 1 , . . . , i r , and such that p k = q k if i k is even, and p k − q k = 1 of i k is odd. On can form such a matrix by adding to the matrix p ′ k = q ′ k = ⌊ i k 2 ⌋ a matrix of 0 and 1, with 0 in the columns of the even i k , and exacly one 1 in the columns of the odd i k , such that the sum of the first row is p − p+q−ℓ 1 (I) 2 and, equivalently, such that of the second row is q − p+q−ℓ 1 (I)
Extension to other algebras
The algebra of quasi-symmetric functions has a noncommutative version WQSym (Word Quasi-Symmetric functions) consisting of the invariants of Hivert's quasisymmetrizing action of the symmetric group on the free associative algebra [2]. The product of the monomial basis of WQSym is described by convolution of packed words M w which can also be described in terms of the bigger algebra MQSym (Matrix Quasi-Symmetric functions) [1], based on packed integer matrices. The product of two basis elements MS P MS Q , where P is a p × r matrix and Q a q × s matrix is obtained by forming the block matrix (82) P • Q = P 0 0 Q regarded as a word over the alphabet of rows, and taking the quasi-shuffle of the word P 1 ⋯P r formed by its first p rows with the word formed by its last q rows, the contractions being given by vector addition of the rows. Packed words can be encoded by packed 0 − 1-matrices with exactly one 1 in each column. For example, u = 21321 is encoded by where m ij = 1 iff the jth letter is an i. Then, packed convolution corresponds to the quasi-shuffle of these matrices, and so can be described by a recurrence where A ′ and B ′′ denote A and B completed by the appropriate number of zeros on the right or on the left. The basis X I can be lifted to WQSym by means of the (reverse) refinement order on packed words: if To describe the product X u X v , we shall embed WQSym and MQSym in the limit ℓ → ∞ of the level ℓ quasi-symmetric functions defined in [19], where the noncommutative product of these algebras can be realized as a shifted version of the commutative product of a bigger algebra.
Let Ω be the semigroup of nonnegative integer sequences i = (i 0 , i 1 , . . .) with finite sum i . We denote by Sym (N) the Hopf algebra of N-colored noncommutative symmetric functions, which is defined as the free associative algebra over indeterminates S i , with S 0 = 1, endowed with the coproduct (86) ∆S n = i+j=n S i ⊗ S j .
Departing from the notation of [19], we represent color sequences i as row vectors, and regard the label I of the basis element S I = S i 1 ⋯S ir as an r × ∞ matrix. The multidegree I of I is defined as its column sum sequence. The Hopf algebra QSym (N) of N-colored quasi-symmetric functions is defined as its graded dual with respect to this multidegree. Both algebras admit polynomial realizations, in terms of two colored alphabets In terms of A, the generating function of the complete functions can be written as This realization gives rise to a Cauchy formula which in turn allows one to identify the dual of Sym (N) with the limit of an algebra introduced by S. Poirier in [26].
j , j ≥ 1}, be an N-colored alphabet of commutative variables, also commuting with A. Imitating the level 1 case (see [1]), we define the Cauchy kernel with I = (i 1 , . . . , i m ). These functions form a basis of a subalgebra QSym (N) of K[X], which we shall call the algebra of N-colored quasi-symmetric functions .
We can now define ϕ x = log σ x , and get a family of primitive generators of Sym (N) by setting ϕ n = coefficient of x n in ϕ x .
With a formal series f (x) = x + O(x 2 ) as above, we can now define a basis Y I by setting Y n = coefficient of x n in f (ϕ x ). The coproduct of Y will the be given by the formal group law associated with f .
Thus, if f is as above chosen as the exponential generating functions of the E n (s, t), the product rule of basis X I will be the (α, β)-quasi-shuffle over the alphabet Ω.
The subspace of QSym (N) spanned by the M I where I is a packed matrix (followed by zero columns) is stable for the product. This is not the case of the span of the matrices encoding packed words, but both are stable for the shifted product: if I is an r × p packed matrix, and J an s × q packed matrix, define This is therefore an encoding of the product X u X v in WQSym, which is thus given by the shifted (α, β)-quasi-shuffle of the corresponding matrices. Since in this case no two contractions can be equal, the product is multiplicity-free, and the cofficient of each X w is just a single monomial α i β j , which can be explicitly computed. Let f w uv be this coefficient. If there exists j = k such that u j = u k and w j = w k , or v j = v k and w j+ u = w k+ u , then f w uv = 0. Otherwise, let a i be the number of different values u j such that w j = i, and b i be the number of different values v j such that w = j + u = i. Then, For example, the coefficient of X 11211 in X 123 X 11 is αβ, since a 1 = 2, b 1 = 2 and a 2 = 1, b 2 = 0. | 2022-09-28T06:44:41.591Z | 2022-09-27T00:00:00.000 | {
"year": 2022,
"sha1": "30f41d41cfb5f84494eb055ccbb6ed9f7dd4da85",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "30f41d41cfb5f84494eb055ccbb6ed9f7dd4da85",
"s2fieldsofstudy": [
"Mathematics"
],
"extfieldsofstudy": [
"Mathematics"
]
} |
34205687 | pes2o/s2orc | v3-fos-license | Influence of site index on the relationship between forest net primary productivity and stand age
Previous studies show that forest net primary productivity (NPP) varies pronouncedly with stand age, and these variations play a crucial role in determining forest carbon sinks or sources at regional scales. Some forest carbon cycling models, eg. InTEC (The integrated terrestrial ecosystem C-budget model), calculates annual forest NPP in the long term according to normalized NPP-age relationships and the reference forest NPP at a given age. Therefore, the accurate NPP-age relationship is important for forest NPP estimation. In this study, NPP at various stand ages for twelve major forest stand types in Heilongjiang Province in northeast China is derived from yield tables with consideration of the total biomass increment and foliage and fine-root turnovers. Similar to previous studies, our results also show that forest NPP increases quickly at young ages, reaches the maximum value at middle age (10–40 years old), and then decreases to a relative stable level at old ages. However, we additionally found that forests under better site conditions have faster growth rates in young ages and steeper declines after reaching the maximum. Therefore, when the NPP-age curves for different site indices are normalized against the maximum value of each curve, there are significant differences among them. These differences have implications on the methodology for estimating the spatial distribution of forest carbon sources and sinks.
Introduction
Net primary productivity (NPP) is defined as the difference between total photosynthesis and autotrophic respiration [1], and is an important component of the terrestrial carbon cycle. In addition to environmental factors (such as climate, CO 2 and nitrogen deposition), disturbance (such as fire, insect and harvest) has strong impacts on the carbon cycle by directly releasing carbon to the atmosphere and by accelerating the respiratory processes [2][3][4][5]. It also alters the forest age structure and species composition, therefore modifies the spatial distribution of NPP. In general, disturbance has dominant influence on the forest carbon cycle over other factors [6][7][8][9][10]. To quantify the full impacts of disturbance on the forest carbon cycle, NPP-age relationships are needed for estimating regrowth after disturbance [11][12][13][14][15]. This is why it is imperative to analyze the controls in regulating the relationship between NPP of different forest stand types and stand age. Different methods, e.g. forest inventory data analysis [16,17], empirical models [6] or process-based models [18][19][20][21][22], have been developed to estimate forest NPP. Some models for quantifying forest NPP only consider the effects of environmental factors without taking into account of the influence of forest stand age [23][24][25][26]. However, the spatial distribution of forest carbon sources and sinks depends more on forest stand age than environmental factors [12,15,27], and therefore it is important to consider forest stand age in carbon budget estimation [28,29].
NPP-age curves vary among different species and are also affected by the site condition index (SCI) because forest growth depends on local soil, topographical and hydrological conditions, in addition to meteorological conditions [13,30,31]. SCI, defined as the average tree height at the specific age (usually at the age of 50). Younger ages are sometimes used for plantations, short-lived species, or species managed on shirt rotations [32][33][34][35]. However, so far only mean NPP-age relationships are derived under long-term mean climate conditions without consideration of site conditions. In northeast forests of China, either the constant agerelated NPP curves or coarse age-NPP relationships derived using data over entire China were used for forest carbon cycle modeling [14,21]. Fewer studies are reported on NPP-age curves under different site indices. Chen et al. [36] used stand yield tables to develop a non-linear function of age-NPP relationship stratified by a site index for several Canadian boreal black spruce forests. In northeast forest of China, age-NPP relationships are neither stratified by SCI nor developed for specific species. In this paper, we developed functions to simulate the relationships between NPP and age for twelve major forest stand types under different SCIs in northeast China, using yield tables, biomass equations for four different tree biomass components, as well as foliage and fine-root turnover rates. The maximum NPP magnitude and the age at which the maximum NPP occurs will also be analyzed against SCI. This information will be helpful for foresters to decide the most suitable curve for use in a particular area and also will be useful for carbon cycle modelers to decide specific NPP-age curves based on site condition indices.
platyphylla forests, Populus plantation, mixed coniferous forests, mixed broadleaf forests and coniferous-broadleaf mixed forests in this study. These yield tables consist of the necessary information for estimating wood volume and biomass, and usually include age, diameter at breast height, tree height, stand density for living trees, and increment of gross total volume. The age ranges of the above stand types in yield table are 0-50, 0-50, 0-40, 0-80, 0-100, 0-70, 0-80, 0-70, 0-30, 0-150, 0-90, 0-120, respectively. Using these data, the total tree biomass (both aboveground and belowground) can be obtained by summing up its components (stem, branch, foliage, root), and used for stand NPP calculation by multiplying the tree NPP at average height and diameter with stand density. For mixed forest types consisting of various species, NPP of a forest stand is estimated based on its species composition and their areal fractions available from the inventory data, since they are not included in yield tables. The inventory data in Heilongjiang province covering the period of 1990-2005 involve 19,500 permanent field samples that provide growth, site information, species composition, stand age, forest type, stand density, site condition index and so on. Site condition index is calculated according to national standard, "the main forest types harvest table in cities and countries" (Number: DB23/ T 1377-2010), in China. It is quantified by Eq 1 for Pinus koraiensis and Pinus sylvestrisvar forests, by Eq 2 for other stand types. Here, SCI refers to site condition index, t 1 refers to stand specific age, k and c are parameters. Stand specific age refers to the age at where stand growth trends to be stable. Average stand height is calculated by Eqs 3 and 4 (Eq 3 for Pinus koraiensis forests and Pinus sylvestris forests, Eq 4 for other stand types, where TH refers to average stand height, t refers to stand specific age, A, k and c are parameters). Parameters used in the above model were obtained from yield table compilation calculating from inventory data ( Table 1). The larger value of site condition index indicates better site condition. Influence of SCI on the relationship between NPP and age
NPP calculation
In this study, NPP is calculated based on normal yield tables, biomass allometric equations, inventory data, and leaf and fine-root turnover rates from other studies. NPP is modeled as the sum of biomass increments, litter-fall and fine-root turnovers [36] for twelve forest types in Heilongjiang province, northeast China: where dB/dt is the increment of dry matter in total living biomass per year at the specific age with the assumption that the fine root don't change in two successive years, and it is estimated from the yield tables; L f and L fr are the annual foliage and fine-root turnovers. The methods to compute the biomass increment in living trees and foliage or fine-root turnovers will be described in the following paragraphs. Understory NPP and tree mortality are not included in this NPP estimation because of the lack of understory vegetation data and mortality statistics. In this study, the total biomass (B t ) is separated into four components: B s for stem, B b for branch, B f for foliage, and B r for coarse root. Dong et al. [37] developed empirical models for the total biomass and the biomass of six components including aboveground, underground, stem, crown, branch, and foliage biomass for 15 major tree species (or species groups) based on a non-linear error-in-variable modeling approach and 516 sampling trees in Heilongjiang Province. These models are adopted in our study. Models and their coefficients for eight species are provided in appendix A and Table 2. The accuracy rate for biomass estimation was not calculated in this work, it was from the results of Dong's work [37], since the estimation of biomass was used by the models in his work. In order to avoid the infinite increases of foliage biomass estimation from some of the empirical functions with limited sample ranges of stand age, we made an assumption that the maximum foliage biomass appears at the stage of fastest growth of living biomass.
This method solves the problem that the sum of four components of biomass does not equal to total biomass. In these models, the tree diameter at breast height and the tree height are taken as independent variables affecting each biomass component. We use these models to calculate total tree biomass in chronological series in yield tables for each species. Stand biomass per unit area varies with age and is estimated by multiplying the single tree biomass with tree density. [38][39][40][41][42][43]. All these equations show a growth pattern like "S" curve, i.e. the slow-fast-slow-stable pattern. Here, 5 growth models were used separately to fit the relationship between biomass and age for different species in Heilongjiang Province. It was not convergence for parameter estimation by using Mitscherlich equation for almost all species except for Pinus koraiensis. Korf model cannot be used to fit the biomass-age relations of Pinus koraiensisvar since it also has convergence problems. However, for coniferous-broadleaf mixed forest, Korf model was the best with the minimum root mean square error (RMSE) and maximum determination coefficient (R 2 ) of 38.76 g•m -2 and 0.9998. The fitting precisions of the Logistic, Gonmpertz and Richards models did not appear much difference. In order to avoid the effects of uncertainty of biomass estimation by using different models on NPP results. Richards equation (Eq 6) with the average minimum RMSE and maximum R 2 was chose at last. RMSE and R 2 of Richards model for all species were showed in Fig 1.
where B is the estimated biomass (g•m -2 ) as a function of stand age (t in years) and its true value is calculated from B t . A, r, and c are parameters representing the maximum biomass, growth rate and assimilating capacity and can be determined as coefficients of biomass-age models (Eq 6). Furthermore, various nutrition conditions, climate changes and management policies impact tree growth curves. However, the average diameter and height inputs to the growth equation represent the smooth growth under average environmental conditions. In previous studies, the average annual biomass increment was taken into account for the first part of NPP calculation in Eq 5 that may smooth the growth rate of plants or even miss the important change at certain stages, and was not a successive increment changing with age. Therefore, dB/dt, the derivative of the growth equation, is better to capture small changes in the plant growth rate.
Litter-fall and fine-root turnovers are the other two components of NPP and are estimated separately because they are not parts of the yearly biomass increment. Litterfall is quantified as age-dependent foliage biomass times its corresponding turnover rate (Eq 7). For deciduous broad-leafed species, obviously their foliage turnover rates are equal to 1. However, for coniferous species, the needle life span varies in relation to many factors, including temperature, light, water, nutrients, insects, diseases, and air pollution [44], and is rather difficult to measure or quantify. We use specie-specific leaf turnover rates published in White et al. [45]. Fine-root turnovers are estimated according to the foliage turnovers by introducing the allocation ratio between new fine root C and new leaf C (Eq 8). The principle for linking leaf and fine-root turnovers is presented in Thornton (1998). Values of the ratio of new fine root carbon to new leaf carbon allocation and foliage turnover rates are chosen from White et al. [45] for different species (Table 3).
Where T f is the foliage turnover rate (yr -1 ); C f is the ratio of carbon to dry matter; and e is the allocation ratio between new fine-root carbon and new leaf carbon. We use 0.44 for C f according to our previous work on forest carbon ratios among biomass components in Northeast China forest area [46]. The carbon content coefficient of the total tree was 0.45. Influence of SCI on the relationship between NPP and age The relationship between NPP calculated from the above methods and forest stand age from yield tables is modeled using Eq 9, where M, b, d, and g are coefficients to be determined [12]. Normalization of a NPP-age curve against the maximum NPP produces the pattern of NPP variation with age and can be used in carbon cycle models to reconstruct the historical NPP variation with age.
Results and discussion
Age effects on biomass and NPP
The variations of the total biomass, total NPP and its components with stand age under the average site condition index (SCI = 14) for each stand types in Heilongjiang Province are shown in Figs 1 and 2. The curve of the total biomass changing with age is well fitted using Richard's equation. NPP increases rapidly before reaching its maximum and thereafter decreases to a relatively steady state. At younger ages, carbon is mostly accumulated in stems, branches and coarse roots so the total NPP is dominated by living biomass increments. The decline of NPP with age is mainly caused by the decreasing rate of living biomass increment. At older ages, NPP-age curves are dominated by leaf and fine-root turnovers since carbon allocations to these two components are larger than the other parts. The stable state of NPP is due to the assumption that leaf turnovers remain constant after reaching the maximum value. Our previous study shows that foliage turnovers decline a little at older ages for many stands [13]. Therefore, we realized that this assumption should be examined further when chronosequences of litter-fall data are available. It is shown from Fig 2 that different stands have distinct NPP magnitudes and NPP-age patterns. The maximum NPP showed large variations among different stands, ranging from 108 to 631gCm -2 yr -1 . For pure broad-leaf forest, the maximum NPP ranged from 108 to 260 gCm -2 yr -1 , with the lowest value for Betula davurlastands (108 gCm -2 yr -1 ). While for pure coniferous forests, the value changed between 303 to 631 gCm -2 yr -1 with the highest value for Larix gmelinii stands (631gCm -2 yr -1 ). The maximum NPP of coniferous forests is much larger than that of broad-leaf forests in pure forest stands, while plantations are more capable to sequester carbon than natural forests. The carbon accumulation capacity of coniferous-broadleaf mixed forests is greater than those of mixed coniferous and mixed broad-leaf forests, with Influence of SCI on the relationship between NPP and age the maximum value of 256 gC m -2 yr -1 . Coniferous forests grow to mature more slowly and its NPP reaches the maximum value at later ages, i. e. 20 years or more. NPP of Pinus sylvestrisvar stand reaches the maximum at 30 years. NPP of broad-leaf forests reaches the maximum at younger ages, i. e. 10 year old. Among three natural mixed forests, the mixed broad-leaf forests reach the maximum NPP earliest at the age of 17 years, followed by 23 years for coniferousbroadleaf mixed forests and 43 years for mixed coniferous forests. The carbon sequestration in coniferous-broadleaf mixed plantations can be two times or more greater than natural coniferous-broadleaf mixed forests, similarly to results of Thomas et al. [47]. Wang et al. [14] showed that forest NPP reaches its maximum at an older age than that of our study, possibly because, (1) they studied five forest biomes over the China, which are incomparable to the few specific species within Heilongjiang province with only 37 samples of all species, and (2) the forests in Heilongjiang grow to mature earlier with fertile soil.
Effects of age and site index on volume increment
The volume increment is defined as the average volume growth of a stand within a certain time period and it varies with stand age. Fig 3 shows age-related volume increments derived for twelve stand types under various SCIs. Volume increments at different ages are estimated from the yield tables for stands of 10-150 years old. For all species, the better site condition, the larger is the volume increment and the greater are the increment changes around the peak age. Based on the classification of sites and their corresponding volume increment curves, we can infer their capacities to produce wood. These volume increment curves for each species are not proportional to the site condition index, suggesting that we should stratify proper volume increment data according to the site condition index before analyzing the data. In conclusion, both the SCI and age affect the volume increment and are important to carbon cycle estimation. Volume increment is a useful indicator for biomass increment since many studies show that timber volume is linearly related to biomass. It is also a measure of growth status influenced by environmental factors, such as water content, nutrient condition, radiation and temperature. Among different stand types, there is a large difference in the volume increment. Populus davidiana, the fastest volume growth among all broad-leaf natural forests, reaches the maximum volume increment earlier than Betula platyphylla and Tilia amurensis. Populus plantation is two times more productive and two times faster than natural Populus davidiana to reach the peak. This is perhaps due to forest management. Larix gmelinii plantations have more volume increments and reach the maximum values later than Pinus koraiensis and Pinus sylvestris var. For natural mixed forests, the maximum volume growth of mixed coniferous forests is smaller than coniferous-broadleaf mixed and mixed broad-leaf forests. Three types of mixed forests have the largest volume growths at different ages: about 17 years old for mixed broad-leaf stands and 20-30 years old for mixed coniferous and coniferous-broadleaf mixed stands. However, the relationships between volume increment and age can only explain the volume or woody biomass growth, or show the discrepancies in the carbon uptake abilities among trees of the same mono species under different environmental conditions. A large amount of volume growth for some stands does not always indicate more carbon accumulations. Additionally, NPP may be large in small volume growth stands, or small in large volume growth stands because it is affected by many factors in addition to volume growth, such as carbon to dry matter ratio, and leaf and fine-root turnover rates. Therefore, it is necessary to calculate NPP for carbon uptake estimation and comparison among different stands under different SCIs. Effects of site index on NPP-age and normalized NPP-age curves Stand NPP changes with age under different SCIs as shown in Fig 4 and have similar patterns to relationships between volume increment and age. Better site conditions produced greater NPP, faster growth rates in young forests and steeper declines after reaching the maximum value. The peak age is smaller in better site conditions. This general pattern is consistent with results of Chen et al. [36].
The site condition index has a great impact on mono-species stand NPP. The changes of NPP with age are not proportional to SCI. The largest NPP variations caused by SCI differ from 50% to 500% indicating that more attention should be paid to different site conditions index for determining NPP-age relationships. The relative variations of NPP with age at poor sites are smaller than those at good sites. Therefore, it is necessary to choose a suitable NPPage curve for carbon cycle modeling based on SCI in order to get more accurate estimation of NPP if the spatial distribution of SCI is available in the future for large areas, and it is helpful for the foresters to use SCI to decide the most suitable NPP-age curves for forest management in a particular area.
In regional carbon cycle estimation, the absolute NPP value in a pixel in a recent year can be modeled based on remotely sensed LAI and land cover information in combination with soil and meteorological conditions [48], while normalized NPP-age relationships are used to extend this NPP value to other years in long-term carbon cycle modeling [12]. The normalization is made by dividing a NPP-age curve by the maximum NPP in the curve. It is therefore of interest to know if the NPP-age curves under different site condition indices can be normalized to one curve. Normalized NPP-age relationships under different SCIs for twelve kinds of stands are shown in Fig 5. Comparing to the absolute NPP-age relationships (Fig 4), the normalized ones are much less sensitive to SCIs. The normalized NPP-age variations under different SCIs are within 10% for most stands except for Populus plantation and Pinus sylvestrisvar forests, which differ by about 30% and 22%, respectively, across different SCIs at old ages. Interestingly, the same variation is not noticed for natural Populus davidiana forests. Since NPP values of the same mono specie forests at old ages, whether planted or natural forests, are more influenced by physiological limitations than forest management, the planted and natural Populus davidiana have the similar NPP values at old ages. Meanwhile, Populus plantations grow faster than natural Populus davidiana forests, and produce much more NPP at growing boom ages under the same site condition. Therefore, the differences in normalized NPP curves of Populus plantations are larger than that of natural Populus davidiana. Unlike the Populus davidiana, Pinus sylvestrisvar forests have longer life span, and are more adaptable to site conditions, more resistant to cold and drought conditions. Under poor site conditions, NPP of Pinus sylvestrisvar at old ages does not change too much from that under good site conditions (Fig 4), whereas at peak growing stages, NPP is much larger under better site conditions. Thus, Pinus sylvestrisvar has more different normalized NPP-age curves under different SCIs than those of other coniferous species. Comparing to mono specie forests, the variations of NPPage relationships with SCIs of coniferous-broadleaf mixed forests was the smallest. For the three types of mixed forest stands, the normalized NPP-age curves vary less than 10% under different SCIs. These results suggest that the NPP-age curve normalization is an effective way to extract the general NPP-age curve shape for a species or a forest type for regional carbon cycle modeling. Applying only one normalized curve for a stand type to different SCIs could generally incur an error in NPP estimation at old ages by less than 10%, but for some stands, the error can be as large as 30%. More studies are needed to quantify this error.
For most carbon cycle models, historical NPP was calculated based on constant NPP-age relationship, ignoring the effects of site conditions. In this study, we choose NPP-age Influence of SCI on the relationship between NPP and age relationships under different site condition indices (SCI equals to 12, 14 and 16) to estimate NPP in 2009 (Fig 6(a), 6(b) and 6(c)) and analyze the differences. Some NPP changes can be found in different SCIs shown in Fig 6. Forest NPP at old ages in southern and northern parts of Heilongjiang province shows a slight reduction as SCI increases. However, in the middle of Heilongjiang province, forest NPP at younger ages shows some increases when site condition becomes better. These results also suggest that NPP under better site conditions has faster increases at younger ages and rapid decreases at old ages, and has larger relative changes than NPP under poorer site conditions. Relationships between NPP and age for different stands NPP-age curves under the mean SCI (SCI = 14) are shown in Fig 7(b). The fitted coefficients for Eq 9 and R 2 (determination coefficient) for specie-specific functions are listed in Table 4. Pure coniferous plantations have larger NPP than natural broad-leaf forests. Natural coniferous forests can accumulate more carbon than natural mixed broad-leaf and mixed coniferous forests. Coniferous forest NPP decreases substantially after reaching to its maximum value, while the decrease is not as pronounced for broad-leaf forests except the Populus plantations. These results are similar to those of He et al., [13]. NPP of Populus plantations shows an obvious decline in old ages. More productive species have greater changes of NPP with age and show a quicker increase to the maximum and steeper decrease afterwards. Our study shows that NPP of old forests (> 100 years) maintains about 30-60% of the maximum NPP occurred at mid-ages (15-40 years). Previous studies indicated that NPP in old forests generally decreases to about half or one-third of its maximum value [49]. Gower et al. [50] stated that the aboveground NPP commonly decreases by 0-76% as stands mature, similar to the result of about 0-63% decreases by He et al., [13].
The decrease of NPP at old ages is mainly due to the declining carbon allocation to wood components, in addition to increased autotrophic respiration for sapwood maintenance, decreased photosynthesis efficiency and declining N-availability to trees [49]. In addition to these factors affecting the performance of individual trees, changes in forest structure, such as self-thinning and wind damage, would also negatively impact forest NPP at old ages [51]. For old age forests, leaf and fine root turnovers take a large part of photosynthetic productions [52]. Accurate estimates of leaf and fine root turnovers and carbon allocation ratio of new fine roots to new leaves are of importance to NPP calculation.
Generally, the fine root turnover consumes a larger proportion of net primary productivity in coniferous forests than that in broadleaf forests. Our study shows that it costs 5.6%~36.3% and 5.8%~37.8% of NPP for leaf and fine root turnovers in Pinus sylvestrisvar plantations at age older than 10. These values for Pinus koraiensis plantations, broadleaf forests are 2% 40.3% and 2.7%~56.4%, 13.7%~31.3% and 16.8%~38.2% respectively. These values not only vary with site condition indices but also with forest ages. Larix gmelinii plantations allocate 32.5% of NPP to fine roots at 20 years old under the mean site condition index (SCI = 14). It's larger than that in Mei's work, that fine root consumed about 12 percent of NPP in Larix gmelinii plantations [53]. This is possibly because: (1) the calculation of fine roots in her work only includes fine roots with the diameter less than 1 mm, while the definition of the fine root diameter in our study is less than 2 mm; and (2) the site condition chosen in Mei's study is better than that of the mean SCI. Larger fine root turnovers occur under worse site conditions since it can produce more nutrients. Other research showed that fine root turnovers returned 18.1% Influence of SCI on the relationship between NPP and age of NPP to the soil carbon in Larix gmelinii plantations, while litterfall returned 18.4%to the soil carbon, nearly the same [54]. It means the carbon allocation ratio between new fine root and new leaf equals 1, smaller than the value 1.2 used in our study. So far, there are many studies on the estimation of forest fine root productions and turnover rates. However, few are about the carbon allocation ratio between new fine root and new leaf, especially for northeast forests of China. Therefore, further field measurements focused on this issue would be beneficial to forest carbon cycle research. The normalized NPP-age curves under mean site condition index (SCI = 14) as shown in Fig 7(b) are useful for regional carbon cycle modeling. Some process-based models, such as the Integrated Terrestrial Ecosystem Carbon (InTEC) model [55], use normalized NPP-age relationships but not the absolute NPP values. These normalized curves accentuate the NPPage variation patterns of different species, and a curve can be used to reconstruct historical stand dynamics based on NPP at a time with a known stand age [12,27]. From the perspective of ecosystem carbon uptake, the capacity of plantations is higher than natural forests since older forests in protected natural areas tend to reach carbon neutrality at older ages. However, pure plantations may encounter the problems of pest, fire and wind disturbances easily. It is therefore more advisable to consider planting coniferous-broadleaf mixed forests in northeast area of China.
Uncertainty analysis
The model for total living biomass has the highest accuracy at about 90%, followed by the model for stem biomass with an accuracy of 87.5% [37]. The accuracies of the biomass models for coarse roots and foliage are relatively low, but are still greater than 80% for all species. The method to obtain the sum of three components (biomass increment, litter-fall and fine-root turnovers) for NPP estimation is similar to Thomas et al. [47]. The magnitude of NPP estimated in our study differs from other studies. For example, the simulated annual NPP of northeast forests in 2008 by CASA (Carnegie-Ames-Stanford Approach) model ranges from 6.4 gC m -2 yr -1 to 933.5 gC m -2 yr -1 with the mean value of 451.6 gC m -2 yr -1 [22]. The NPP of broad-leaf forests based on field measurements [55] is a little higher than our results. The underestimation of NPP by our method may be due to the exclusion of mortality and understory NPP. But for coniferous, our results are consistent with NPP based on field measurements [55]. Wang [14] showed that mean annual NPP of deciduous needle-leaf forests was 424.7 (gC m -2 yr -1 ) in Heilongjiang province. In this study, due to the lack of data, we assume that carbon allocated to standing dead trees is a very small proportion of whole stand NPP and that understory NPP does not affect the normalized NPP-age curves. Zhou et al. [56] found that the understory carbon contents for three old-growth forests on Changbai Mountains accounted for only less than 2% of total carbon storage was relatively stable over time. These results appear to differ from those of Chen et al. [36] who showed that understory NPP exponentially declines with ages.
In addition to biomass model errors, there are several other sources of errors: (1) the average diameter at breast height and the average tree height used in biomass models give a smooth biomass curve under mean environmental conditions. The accuracy in these diameter and height data is affected by uncertainties from yield tables; (2) specie-specific leaf or fine-root turnover rates and the ratio of new fine root carbon to new leaf carbon allocation are found from literature for the same genus elsewhere, and these literature values may not be representative of the species best in our study area; (3) the assumption that LAI does not vary with age after reaching its maximum may not always be true and may have an effect on NPP at old ages; and (4) stand ages in this study are limited by ranges of stand ages in the yield table, which produced errors on biomass and NPP fittings. The total uncertainty in the NPP-age curve in consideration of all these factors is estimated about 25%, and it increases with age. However, since the bias in NPP estimation caused by a factor is mostly uniform across the age spectrum, we expect that the normalized curves (Fig 6(a)) would be most reliable for carbon cycle modeling.
Conclusions
Based on yield tables from the Heilongjiang Province in northeast China, we derived forest type-dependent NPP-age curves under different SCIs. The following species are considered in our study: Pinus koraiensis, Pinus sylvestrisvar, Larix gmelinii, Picea koraiensis, Populus davidiana, Quercus mongolica, Betula davuria, Tilia amurensis, and Betula platyphylla. We found that NPP increases quickly in young forests, and then decreases at various rates after reaching its maximum at mid ages. Trees in better site conditions are more productive at young ages and their NPP decreases more sharply at old ages. The age at which NPP reaches its maximum value is earlier in better site conditions. The relative variations of NPP with age in poor site conditions are smaller than those in good site conditions. NPP of old forests (> 100 years) maintains about 30-60% of the maximum NPP occurred at mid-ages (15-40 years). The normalized NPP-age relationships are much less sensitive to SCIs, within 10% for most species.
These findings are achieved in China, and confirm the researches previously conducted in USA and Canada. The results produced in this paper will be useful for carbon cycle modelers to decide specific NPP-age curves based on SCIs and also will be helpful for foresters to decide the most suitable curve for use in a particular area.
Supporting information S1 File. Yield table was used in this study. (DOCX) | 2018-04-03T02:32:49.874Z | 2017-05-11T00:00:00.000 | {
"year": 2017,
"sha1": "c80855c2ed912da4ee46157cb198e540782c630c",
"oa_license": "CCBY",
"oa_url": "https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0177084&type=printable",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "c80855c2ed912da4ee46157cb198e540782c630c",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": [
"Environmental Science",
"Medicine"
]
} |
259521058 | pes2o/s2orc | v3-fos-license | An assessment of Arctic diurnal water-vapour cycles in Canada’s weather forecast model and ERA5
The diurnal water-vapour cycle is a critical component of the hydrological cycle, yet it is one of the hardest components to accurately reproduce in forecast and climate models. Previous studies have shown that both forecast and climate models underrepresent the diurnal water-vapour cycle, which leads to errors in precipitation, cloud, and radiative transfer parameters. Most diurnal cycle studies were conducted in the Tropics, and very few model evaluations of this process exist for the Arctic. Additionally, the majority of studies focus on total column water-vapour cycles; almost none use height-resolved measurements. In this study, we evaluate the diurnal water-vapour cycles in Environment and Climate Change Canada’s Global Environmental Multiscale–High Resolution Deterministic Prediction System (GEM–HRDPS) numerical weather forecast model and the European Centre for Medium-Range Weather Forecasts Reanaly-sis v5 (ERA5) using a Vaisala preproduction differential absorption lidar (DIAL) and a co-located Global Positioning System (GPS) located in Iqaluit, Nunavut (63.75 ◦ N, 68.55 ◦ W). Both numerical products reproduce the phase of the diurnal cycle well below 1 km year-round
INTRODUCTION
The diurnal cycle of water vapour is one of the most critical components of the hydrological cycle. It is an integral component of convection, precipitation, and other physical processes in weather and climate numerical models. As such, it is an excellent diagnostic variable for model evaluation, since it encompasses and mirrors several aspects of model performance. The diurnal cycle has been, and largely still is, one of the most difficult components to reproduce in atmospheric numerical models (for climate and weather forecasting) (Dai and Trenberth, 2004;Bock et al., 2007;Bechtold et al., 2008).
Though climate and weather forecasting models can reproduce climatological, seasonal, subseasonal, and mesoscale processes quite well, the diurnal cycle remains a challenge. Typical problems have included convection starting too early during warm seasons by 2-4 hr, extended daytime convection over land, incorrect cloud type formation resulting in errors in radiation transfer, and low integrated water vapour (IWV) diurnal amplitudes (Dai and Trenberth, 2004;Bechtold et al., 2008;Stratton and Stirling, 2012). For example, Bechtold et al. (2008) found that the early initiation of convection and extended convection throughout the day delayed the onset of convective precipitation in many regions, thereby resulting in the underrepresentation of extreme precipitation events and less precipitation than observed. Thus, key improvements to modelling the diurnal cycle of convection were made (e.g., (Stratton and Stirling, 2012;Bechtold et al., 2014)), which has resulted in the improvement of the precipitation diurnal cycle across much of the globe, particularly in regions of deep convection and in midlatitudes over land. Improvements to the convection diurnal cycle have direct impacts on precipitation, cloud, and radiative diurnal cycles. Thus, we should also expect to observe improvements in the water-vapour diurnal cycle.
Accurately modelling the diurnal water-vapour cycle has been hampered by the lack of instrumentation and observations capable of resolving the diurnal scale. Radiosondes have been the instrument of choice for humidity profile measurements; however, as they are usually only launched twice per day, they cannot accurately resolve the diurnal component. Additionally, the humidity sensors are prone to internal diurnal cycles that must be corrected (Miloshevich et al., 2009;Dirksen et al., 2014). Bevis et al. (1992) demonstrated that Global Positioning System (GPS) satellites could be used to obtain total column water vapour or IWV measurements on the scale of minutes; thus, it became one of the choice instruments for detecting diurnal cycles. Bouma and Stoew (2001) presented some of the first measurements of the diurnal water-vapour cycle in the Baltic region and examined the peak-to-peak amplitude of the 24 hr cycle. Dai et al. (2002) presented IWV diurnal cycles using GPS receivers for North America and further found that the IWV diurnal cycle has two primary components: the 24 hr diurnal and 12 hr semidiurnal. Since then, GPS measurements have become the preferred technique for deriving IWV diurnal cycles (e.g., Bock et al., 2007;Jakobson et al., 2009;Galisteo et al., 2011;Lees et al., 2021). GPS receivers lack the ability to distinguish height-resolved cycles and thus cannot resolve how the process evolves with altitude. There remains a lack of height-resolved diurnal water-vapour cycle studies, particularly in the boundary layer, due to the continued lack of high-frequency boundary-layer profiling instruments. Satellite-based imagers have been used to study upper tropospheric humidity diurnal cycles (Chung et al., 2013;Xue et al., 2020), as have ground-based microwave radiometers for the stratosphere and mesosphere (Haefele et al., 2008;Hocke et al., 2017). Louf et al. (2015) and Wang et al. (2002) both used a novel microwave radiometer to measure diurnal mixing-ratio cycles in the first 4 km as well at 100 m vertical resolution in Africa and Oklahoma respectively. Finally, Chepfer et al. (2019) used satellite-based microwave radiometer measurements to get 1-4 km vertical resolution height-resolved diurnal relative humidity cycles from 900 hPa to 200 hPa in the Tropics.
Thus, the bulk of diurnal water-vapour cycle research has focused on the Tropics and midlatitudes, been limited to either total-column or upper tropospheric and mesospheric altitude ranges, and most studies have relied on relatively low vertical resolution observations. Though the Tropics make an excellent region for diurnal cycle comparisons, higher latitudes continue to lack comparison studies due to the difficulty in maintaining ground-based observations for sufficient time-scales. High frequency and metre-scale weather (vertically) occur in the boundary layer; therefore, higher vertical resolution measurements can help resolve boundary-layer processes. Higher vertical water-vapour measurements have also been identified as a priority variable by the World Meteorological Organization (WMO) to improve our understanding of boundary-layer processes and forecasting (World Meteorological Organization, 2016;Montmerle, 2020).
In this study, we address both the need for higher vertical resolution and higher latitude observations of water vapour profiles. In 2018, Environment and Climate Change Canada (ECCC) deployed a preproduction Vaisala lidar Newsom et al., 2020;Gaffard et al., 2021;Mariani et al., 2021) and installed it at the Iqaluit, Nunavut supersite (Joe et al., 2020). The lidar is a differential absorption lidar (DIAL) capable of producing specific humidity profiles from 90 m to 3 km altitude above ground level (depending on the meteorological conditions) at 100-300 m vertical resolution (in the first 1.5 km altitude) at a temporal frequency of 20 min. Its high temporal and vertical resolution make it an ideal instrument for studying boundary-layer processes and the diurnal cycle. The Vaisala DIAL is also not assimilated into forecasting or climate models, and is therefore an ideal tool as an independent source for model verification. Hicks-Jalali et al. (2021) presented the first height-resolved diurnal water-vapour cycles using lidar measurements in the Arctic. They showed that the diurnal water-vapour cycles measured by the DIAL agreed well with the co-located WMO surface station mixing ratio and temperature cycles, as well as with the co-located GPS IWV diurnal cycles. The DIAL results were also verified against available literature Jakobson et al., 2009;Jakobson et al., 2014) for Arctic diurnal water-vapour cycles. As such, in this follow-up study we use the Hicks-Jalali et al. (2021) results to evaluate the diurnal specific humidity and IWV cycles from the European Centre for Medium-Range Weather Forecasts Reanalysis v5 (ERA5; Hersbach et al. (2020)) and ECCC's Global Environment Multiscale (GEM)-High Resolution Deterministic Prediction System (HRDPS) numerical weather prediction (NWP) model . In this article we will refer to both ERA5 and GEM-HRDPS as "numerical products". Both numerical products have recently implemented new convection and microphysics schemes and have made updates to their assimilated observations. Additionally, the diurnal water-vapour cycles in GEM have never been evaluated for the Arctic. The goal of this study is to demonstrate the value in using lidar measurements for NWP validation, as well as to evaluate the numerical products' current diurnal water-vapour cycles to help guide future improvements.
The instrumentation, NWP model, and reanalysis used in this study, and the preparation of their measurements or output, are described in Section 2. Section 3 presents the method used to calculate the diurnal water-vapour cycle. The results and their discussion are in Sections 4 and 5 respectively. Finally, the summary and conclusions are presented in Section 6.
2
DATA AND DATA PROCESSING
Vaisala differential absorption lidar
The Vaisala DIAL is a preproduction water vapour broad-band lidar designed for operational meteorological measurements. DIALs are ideally suited for operational meteorology owing to their high temporal and vertical resolution in the boundary layer. The DIAL provides water vapour profiles from 90 m above ground level to approximately 3 km depending on the meteorological conditions. Unlike the Raman method, in which the return signal measured by the lidar is proportional to the Raman scattering of a single laser signal, the DIAL method uses the difference in atmospheric absorption between two laser signals. The difference in atmospheric absorption can be calculated by taking the ratio of the "online" (911.0 nm, water sensitive) and "offline" (910.6 nm, water insensitive) signals (P on , P off ): where S on,off is the normalized laser spectrum for each frequency , T WV ( , z). Equation (2) is the one-way atmospheric transmission due to water vapour absorption at each frequency with altitude z: . (2) The water vapour transmission is a function of the water vapour number density N(z) and the water vapour absorption cross-section ( , z). The derivation of Equation (1) can be found in Newsom et al. (2020). The DIAL uses an inverse method similar to that in South et al. (1998) to solve for the water vapour number density profile. The water vapour profiles are calculated with respect to ground level; however, for simplicity we just use "metres" to represent "metres above ground level". Note that the Iqaluit supersite is only approximately 10 m above sea level; therefore, metres above ground level and metres above sea level are almost the same. The water vapour mixing ratio profiles are then calculated using a weather sensor present on the DIAL and extrapolating the temperature and pressure profiles using the International Organization for Standardization (ISO) 2533:1975 standard atmospheric model with a vertical temperature gradient of −0.0065 K⋅m −1 . A more detailed description of the DIAL's retrieval algorithm can be found in Newsom et al. (2020). Updated parameters for the preproduction model retrieval and design are presented in tab. 2 of Hicks-Jalali et al. (2021) and Mariani et al. (2021).
The DIAL mixing ratio profiles have been evaluated over several validation campaigns in various climates and meteorological conditions. The initial prototype was presented in Newsom et al. (2020), which validated the prototype DIAL against radiosondes, a Raman water vapour lidar, and an atmospheric emitted radiance interferometer (Knuteson et al., (2004a); Knuteson et al., (2004b)) at the Great Plains Atmospheric Radiation Measurement research site in Oklahoma. They found that the DIAL was effectively unbiased against the radiosondes and the Raman lidar (−0.01 g⋅kg −1 ), but had a slight dry bias of −0.22 g⋅kg −1 against the atmospheric emitted radiance interferometer. ECCC deployed the new preproduction model in early 2018 and validated it against a Raman lidar at the ECCC site in Toronto . It was found to work well in urban conditions with a minimal overall bias of 0.17 ± 0.14 g⋅kg −1 . It was then validated a second time for Arctic conditions against coincident radiosondes and the Canadian autonomous Arctic aerosol lidar (a Raman lidar) over a year-long period. Mariani et al. (2021) found a small systematic wet bias in the DIAL with respect to the radiosonde (0.13 ± 0.01 g⋅kg −1 ) and the Canadian autonomous Arctic aerosol lidar (0.18 ± 0.02 g⋅kg −1 ). The wet bias between the two instruments is due to an overestimation of the laser spectral width in the DIAL's water vapour retrieval algorithm and will be corrected by Vaisala in future algorithm updates. In addition to the small systematic bias, Mariani et al. (2021) found a diurnal cycle in the water vapour retrieval; the current retrieval is sensitive to solar background noise in the overlap region between the low-altitude and high-altitude channels (250-450 m). This sensitivity will be resolved by Vaisala in future models. Consequently, we have removed this altitude region from our diurnal analysis. Most recently, Gaffard et al. (2021) conducted a 1 month validation of a second preproduction version at the UK Met Office testing site in Cardington. They found biases of similar range and magnitude to the previous studies.
The height-resolved diurnal analysis in this study uses the same profiles and data processing as in Hicks-Jalali et al. (2021). We use DIAL water-vapour mixing ratio profiles from September 2018 to June 2020. The DIAL has a temporal resolution of 20 min but produces a rolling average profile every minute. For the purposes of this study, we used one profile per hour at the start of the hour for consistency with both numerical products. The profile is an average of the 20 min prior to the hour mark; for example, the average of 2040-2100 UTC is the profile used for 2100 UTC. Though the DIAL reports measurements every 4.8 m, Newsom et al. (2020) calculated the vertical resolution to range from 100 to 300 m, increasing with altitude. We vertically average the mixing ratio profiles to 100 m for the entire profile, noting that above 1 km the vertical resolution increases to 200 m. Therefore, the bins above 1 km may have some correlation.
GPS
The Iqaluit ground-based GPS receiver is managed by National Resources Canada and the International Global Navigation Satellite System Service (IGS). The station was installed in late 2009 and has been operating continuously since then with less than 2% downtime. The total column water vapour amount, or IWV, can be calculated by determining the amount a satellite signal is delayed when it is sent from the satellite to the ground receiver. Bevis et al. (1992), Bevis et al. (1994), Emardson et al. (1998), andJones et al. (2020) provide detailed summaries on the GPS IWV retrieval. The National Geodetic Laboratory at the University of Nevada has recently reprocessed the IGS database with an updated IWV retrieval (Blewitt et al., 2018). The new IGS product now provides an IWV calculation. Hicks-Jalali et al. (2021) validated the National Geodetic Laboratory IWV measurements for Iqaluit against the twice-daily radiosondes launched from the Iqaluit supersite. The radiosonde and GPS IWV average values were well within the standard deviation of their respective measurements for all seasons, with the exception of a 1 mm wet bias for the GPS in the winter months. The wet bias is likely due to the radiosonde humidity sensor's sensitivity to cold temperatures (Miloshevich et al., 2009;Dirksen et al., 2014). Regardless, a systematic bias has little impact on the diurnal cycle solutions as the diurnal cycle is a relative calculation. This study uses the same GPS diurnal cycle analysis conducted in Hicks-Jalali et al. (2021) from September 2009 through December 2019. Hicks-Jalali et al. (2021) found a discontinuity in the diurnal GPS measurements at 0000 UTC/1900 h local solar time. The IGS GPS IWV solutions are calculated in 24 hr batches, with no correlation between each 24 hr period. Small errors in total zenith delay from the satellite can result in discontinuities at the 24 hr mark (Brockman, 2020), which are unavoidable. To mitigate their effect on the diurnal solutions, the GPS measurements are averaged to a temporal resolution of 1 hr and filtered as in Hicks-Jalali et al. (2021). For this study, the solutions are shifted to UTC to facilitate comparisons with the models. GPS zenith total delay (ZTD) measurements are assimilated into the GEM-HRDPS forecast model, but not ERA5. As IWV measurements are derived from ZTD, the GPS and GEM results are not entirely independent. However, the GPS results are used to complement the DIAL results and illustrate the degree to which the numerical products agree or disagree with both instruments. We find that the disagreement between numerical products and GPS is similar to the disagreement between numerical products and DIAL, which leads us to conclude that the GEM-HRDPS's dependence on the GPS results is minimal.
GEM-HRDPS
The HRDPS is a high-resolution NWP forecasting model developed by ECCC based on the GEM atmospheric model (Côté et al., 1998). Its domain covers the northern portion of the United States and the majority of Canada and has a horizontal resolution of approximately 2.5 km, or roughly 0.0225 • on its rotated grid. At Iqaluit, the grid resolution is 2.42 km. It uses 62 vertical levels, with an approximate resolution of 100 m in the first 2 km of altitude (with 11 levels gradually stretched). The kilometre-scale resolution of HRDPS was shown to improve forecasts of 1-2 days, particularly in cases of squall lines and convective storms. The system has been operational at the Meteorological Service of Canada since 2015, with regular updates. Of importance for the period used in this study, the predicted particle properties microphysics scheme (Morrison and Milbrandt, 2015;Milbrandt and Morrison, 2016;Milbrandt et al., 2018) was introduced with minor subsequent modifications. GEM-HRDPS is run four times per day starting at 0000, 0600, 1200, and 1800 UTC for 48 hr. Finally, despite the fact that HRDPS falls into the convection-permitting model category, the Kain and Fritsch (Kain and Fritsch, 1990) convection parametrization scheme was introduced to HRDPS and reduced the bias of the precipitation diurnal cycle for summertime precipitation . Initial conditions for HRDPS come from different sources. Hydrometeor fields are created from the previous run's first 6 hr . This allows the HRDPS to start a forecast with fully developed clouds and microphysics and additional spin-up time is not required. The Canadian Land Data Analysis System (Carrera et al. (2015)) is a 2.5 km coupled assimilation system that provides a soil and vegetation scheme with the mean surface temperature and soil moisture variables to drive the atmospheric components. The Canadian Land Data Analysis System assimilates observations such as snow depth, temperature, dew-point, soil moisture, snow albedo, precipitation, and several other snow parameters. The kilometre-scale assimilation of soil moisture and precipitation can be important for convection due to the increased ability to model moisture gradients . It is worth mentioning that detailed initialization of sea-ice cover and surface temperature is not provided in the region of Iqaluit but only in the south for the Gulf of St Lawrence.
The HRDPS is based on and is driven by the ECCC's lower resolution (10 km) Regional Deterministic Prediction System (RDPS) model for initialization of other variables and for the lateral boundary conditions. In addition to the standard synoptic and satellite measurements, the RDPS assimilates ground-based GPS measurements from the NOAA Global Systems Division GPS network every 2 hr (Macpherson et al., 2008;Buehner et al., 2015;Caron et al., 2015). Details on the GPS assimilation procedure can be found in Macpherson et al. (2008). RDPS assimilates both the ZTD and the surface pressure measurements from each station, including the Iqaluit station. Assimilating ZTD from GPS improved the precipitation forecasts for certain regions and lead times . Hydrometeor recycling is also performed in RDPS .
We use GEM-HRDPS specific humidity and IWV variable output over the same time frame as the DIAL measurements (September 2018-June 2020). The GEM-HRDPS (henceforth just GEM) variables are retrieved on the model native grid and then interpolated (by using the nearest grid point) to the Iqaluit supersite's latitude and longitude. To minimize dependence from radiosondes, which are assimilated at 0000 and 1200 UTC, we chose to use the forecasts starting at 0600 UTC. Avoiding dependence on the GPS assimilation was not possible since the GPS measurements are assimilated for every forecast run. To further mitigate any artefact from spin-up, we start the time-series calculation from the sixth hour after the 0600 UTC initialization time (sixth hour "lead time"), such that the time series is calculated in a 24 hr window from 1200 UTC to 1200 UTC the following day. The DIAL, GPS, and ERA5 time series begin at the same time to avoid differences due to the end points of the series.
In addition to the specific humidity and IWV variables, we also use the geopotential height variable provided for each model level to linearly interpolate the GEM specific humidity profiles to a standard altitude grid. The standard grid is a 100 m resolution grid chosen to coincide with the DIAL altitude grid. The standard model pressure-level grid has a vertical resolution of 25 hPa in the first 2 km; therefore, interpolation error is minimal. Note that the IWV variable in GEM is not a prognostic variable but is derived (and then integrated) from the specific humidity profile. The IWV is provided as a standard output variable; therefore, we have not calculated it separately.
ERA5
ERA5 is the most recent version of the European Centre for Medium-Range Weather Forecasts reanalysis models (Hersbach et al., 2020). ERA5 profiles are provided on 37 standard pressure levels from 1,000 hPa up to 1 hPa with a vertical resolution of 25 hPa up to 100 hPa. ERA5's horizontal resolution was increased to 0.25 • × 0.25 • (approximately 28 km north-south and 13 km east-west at Iqaluit) and uses a four-dimensional variational data assimilation scheme with the IFS Cycle 41r2 forecast model. The assimilation system uses 12-hourly windows from 0900 to 2100 UTC and then 2100 to 0900 UTC the next day. ERA5's land and ocean components are coupled to the atmosphere. The technical details of the ERA5 physics and assimilation scheme are discussed thoroughly in Hersbach et al. (2020), in addition to a detailed list of the observations used in the assimilation scheme. Since ERA5's release there have been several validations of the humidity diurnal cycles. Most studies have focused on the Tropics, where convection is the strongest and a clear diurnal cycle can be observed (Chepfer et al., 2019;Xue et al., 2020;Lees et al., 2021). Xue et al. (2020) and Lees et al. (2021) conducted comparisons between ERA5 and IWV and upper tropospheric humidity respectively. They found that ERA5 exhibited some discrepancies between the model and observations with regard to the IWV diurnal cycle and upper tropsospheric humidity. The ERA5 model IWV diurnal cycle amplitudes were underestimated with respect to the local GPS observations, and ERA5 had difficulty capturing night-time maximums over the open ocean. Lees et al. (2021) observed that ERA5 underestimated the diurnal amplitude of upper tropospheric humidity.
In this study we use the hourly profile product on pressure levels (Hersbach et al., 2018a) for the specific humidity and geopotential profiles and the single-level product for the IWV variable (Hersbach et al., 2018b). The ERA5 diurnal cycles are calculated over the same time periods as the DIAL cycles. We use the grid point closest to the DIAL for comparison (63.5 • N, 68.75 • W). ERA5 specific-humidity profiles from the hourly product are used for the height-resolved cycles, whereas the IWV cycles use the total column water vapour ("tcwv") variable. The relatively coarse vertical resolution of ERA5 in the boundary layer corresponds roughly to an altitude resolution of 200 m. Hourly ERA5 geopotential profiles were used to interpolate the specific humidity measurements onto a standard altitude grid. The standard grid was defined to minimize interpolation error across the seasons and centred on the same bins as the DIAL and GEM altitude grids.
THE DIURNAL CYCLE
A diurnal cycle can be calculated via two methods. The first is by fitting a sinusoid to a time series via least-squares fitting (e.g., Dai et al., 2002;Dai and Trenberth, 2004;Galisteo et al., 2011;Jakobson et al., 2014). The second method is to use fast-Fourier transforms to determine the phase and amplitude of each harmonic or component in the cycle (Soden, 2000;Tian et al., 2004;Hocke et al., 2017). The second method either requires a continuous time series with no breaks, or constructing a "composite day". A composite day method is necessary when measurements of a specific location are inconsistently spaced (such as with satellite measurements). The time series in this study are sampled with high enough frequency that creating a composite day is not necessary. The first methodology enables direct comparisons with previous Arctic diurnal cycle work, such as Dai et al. (2002), Jakobson et al. (2014), and Hicks-Jalali et al. (2021). Hence, in this study we use the first method and follow the least-squares fitting procedure as in Dai et al. (2002). The diurnal cycle is represented as a sinusoidal function of the form S n (t ′ ) = a n cos(nt ′ ) + b n sin(nt ′ ).
The water vapour time series I t is the summation of the daily average I 0 , residuals R, and the diurnal cycle components S n (t ′ ). Dai et al. (2002) found that two major components were present in their IWV time series across North America, and subsequent studies have corroborated their results. Hicks-Jalali et al. (2021) also found that the third and fourth harmonics in the cycle were negligible at Iqaluit. Therefore, n = 1, 2, where S 1 is the 24-hr component and S 2 is the 12-hr component. The amplitude A n and phase n for each harmonic are found via a n and b n , using the following relationships: All diurnal cycles presented here are relative diurnal cycles, or diurnal anomalies, where I 0 has been subtracted from the cycle. Subtracting the average removes any systematic biases between the models and the observations, highlighting the diurnal behaviour.
RESULTS
As Iqaluit is relatively far north, we did not use conventional seasonal definitions. We follow the same seasonal definitions as described in Hicks-Jalali et al. (2021), where summer is April-August (AMJJA), fall is in September through October (SO), winter is November through January (NDJ), and the spring is February and March (FM). These seasons correspond to the amount of short-wave radiation received at Iqaluit throughout the year, which directly influences the corresponding water-vapour cycle (Semmler et al., 2005).
Seasonal comparisons
The seasonally averaged water vapour profiles for ERA5, the DIAL, and GEM are presented in Figure 1 to provide environmental context and an evaluation of the models on a seasonal scale. Seasonal averages are equivalent to I 0 (z), Equation (3). ERA5 and GEM individual hourly profiles are limited to the same altitude range as the DIAL before averaging and then linearly interpolated to a standard altitude grid of 100 m for GEM and 200 m resolution for ERA5. Note that the spike between 250 and 350 m in the DIAL average is due to the interference from the solar background in the overlap region between the low-altitude channel and the high-altitude channel (Mariani et al., 2021). Both models agree well with the DIAL and are within the standard deviation of each-other's measurements for all seasons. There are small biases between the models and the DIAL; however, they are all well within the 1 variation of the water vapour over each season. Above 2 km there are not enough measurements to accurately calculate an average; therefore, we have removed these data from the analysis. Summer and fall exhibit similar average profiles of 2.0-3.0 g⋅kg −1 due to the summer profile including the transition months (April and May) and a lag in water vapour concentration decrease from August to September. The standard deviation of the measurements decreases with altitude from approximately 1.0 g⋅kg −1 at the surface to 0.8 g⋅kg −1 at 1,800 m. During both seasons, there is a slight dry bias in the DIAL with respect to the numerical products of 0.25 g⋅kg −1 between 850 m and 1050 m, possibly due to the rapid drop in coincident measurements. In the summer above 1100 m, the DIAL is wet by 0.25 g⋅kg −1 with respect to both numerical products. This is a known issue, due to an incorrect assumption in the spectral width of the laser, which produces a systematic wet bias of approximately 0.2 g⋅kg −1 (Mariani et al., 2021). The wet bias appears higher in altitude in the fall at 1,750 m. Water vapour concentrations drop significantly in the winter to 0.8-0.9 g⋅kg −1 , with standard deviations between 0.7 and 0.3 g⋅kg −1 (decreasing with altitude). Similar to summer and fall, ERA5 is drier than the DIAL by approximately 0.1-0.2 g⋅kg −1 above 1 km. In the winter, GEM shows no dry bias relative to the DIAL until 1,750 m and has a smaller bias than ERA5 in the spring. Both numerical products are dry by 0.05 g⋅kg −1 with respect to the DIAL in the first 500 m in winter and spring.
ERA5 and GEM have excellent agreement with the GPS IWV measurements in summer and fall (Table 1). In the summer, the GPS measures a seasonal average of 10.84 mm and the ERA5 and GEM averages are within −0.44 mm and +0.09 mm respectively. All datasets had a standard deviation around 5 mm. The standard deviation is largest in the summer because April and May have smaller water vapour values than June through August, thereby increasing the variability over the entire season. In the fall, ERA5 again has a small dry bias with respect to the GPS of −0.41 mm, whereas GEM has almost no bias at −0.04 g⋅kg −1 . ERA5 has a dry bias with respect to the DIAL between 250 to 800 m (Figure 1), which is likely the cause for the dry bias in IWV as GEM has no dry bias in that region and almost no bias in IWV. Both models and the GPS have standard deviations on the order of 4 mm.
In the winter and spring, both models have around a 1 mm dry bias with respect to the GPS. Hicks-Jalali et al. (2021) found that in the winter months the radiosondes launched from Iqaluit have a similar dry bias with respect to the GPS, and radiosondes are known to have a dry bias at colder temperatures (Miloshevich et al., 2009;Dirksen et al., 2014). It is likely that the dry bias in the radiosondes affects the numerical products, as radiosondes are one of the primary data sources assimilated for their humidity profiles. Though we attempt to mitigate the affect of the radiosonde in the GEM results by choosing a lead time where they are not assimilated, there is some memory in the model that may contribute; this is explored in more detail in Section 4.4.
Height-resolved diurnal cycles
The height-resolved diurnal cycles were calculated by fitting Equation (3) to the DIAL, ERA5, and GEM time series for each altitude bin and season. The amplitude and phase of each solution was then calculated using Equation (5) and Equation (6) respectively. Figure 2 shows the total diurnal cycle (S 1 + S 2 ) for each model and the DIAL, as well as the R 2 values for each fit as a function of altitude. Since the daily average I 0 (z) has been subtracted from the fit, this figure provides the relative change in the water vapour concentration. Using relative cycles removes the systematic biases discussed in Section 4.1. The grey/ masked region in the DIAL solutions represents the region where there is a known internal instrumental diurnal cycle Mariani et al., 2021); therefore, we do not discuss the results from these altitudes. The time axis starts at 1200 UTC due to the GEM cycle time series starting at the 0600 UTC + 6 hr of lead time afterwards (1200 UTC). The DIAL and ERA5 cycles were also calculated starting at 1200 UTC to eliminate any possible discrepancy due to starting the diurnal fit at different times. Note that Iqaluit is located at UTC−0500. For all seasons, DIAL amplitudes corresponding to lower R 2 values (generally above 1 km) must be treated with caution. Though some of the decrease in the fit may be due to random effects in the time series, comparisons with Jakobson et al. (2014) demonstrated that the summer cycle amplitudes and phases above 1 km (as below 1 km) are representative for the DIAL's latitude (Hicks-Jalali et al., 2021), increasing confidence in these observations despite their lower R 2 values.
As Iqaluit is still below the Arctic Circle (66 • 33 ′ 48.9 ′′ ), there are still 3 hr of twilight around the summer solstice and around 6-10 hr of twilight and night-time in April, May, and August. Therefore, a clear diurnal cycle is visible in both models and the DIAL during the summer. The DIAL observes a peak in water vapour occurring at 2000 UTC (1500 h local time) in the first few hundred metres that then shifts to 25/0100 UTC above 600 m altitude as it is transported vertically. ERA5 and GEM also reproduce maximums around 2000 UTC in the first few hundred metres, which then gradually shift to later hours at higher altitudes. GEM reproduces larger maxima than the DIAL in the first few hundred meters by between 0.05 F I G U R E 2 Diurnal water-vapour cycle solutions and accompanying R 2 values for the European Centre for Medium-Range Weather Forecasts Reanalysis v5 (ERA5; column 1), differential absorption lidar (DIAL; column 2), and Global Environmental Multiscale-High Resolution Deterministic Prediction System (GEM-HRDPS; column 3) for each season. Red colours (solid contours) represent increasing water-vapour mixing ratios (MRs) and blue colours (dashed contours) represent decreasing water-vapour MRs. Summer cycles (April-August, AMJJA) are shown in row 1, fall (September-October, SO) in row 2, winter (November-January, NDJ) in row 3, and spring (February-March, FM) in row 4. The daily average water vapour profile I 0 (z) has been subtracted from each solution to show the relative change in the MR (or diurnal anomaly). The DIAL's grey region between 250 and 450 m corresponds to the region of unreliable fits due to the known instrumental diurnal cycle and should be treated with caution. [Colour figure can be viewed at wileyonlinelibrary.com] and 0.08 g⋅kg −1 . GEM is able to reproduce the increase in amplitude that the DIAL observes at 700 m of 0.15 g⋅kg −1 ; however, it does not reproduce the increase in amplitude above 1 km. Although ERA5 reproduces the overall phase of the cycle, its amplitudes are generally smaller by 0.1 g⋅kg −1 for all altitudes, with larger biases above 1 km.
The maximum of the DIAL diurnal cycle shifts to earlier in the day for the fall cycles for the first kilometre from the surface, with similar amplitudes as the summer. At 1 km there is an abrupt shift in the maximum's phase to 28/0400 UTC and a large increase in amplitude up to 0.2 g⋅kg −1 . ERA5 is not able to reproduce the fall cycle with the correct phase or magnitude until above 1 km. It has a maximum of 0.05 g⋅kg −1 at 0000 UTC, almost out of phase with the DIAL. GEM is able to reproduce a diurnal cycle that agrees with the first 800 m of the DIAL observations, albeit with a slightly larger amplitude of 0.15 g⋅kg −1 . However, whereas the DIAL observes an abrupt shift in phase at 1 km, the GEM maximum gradually shifts with altitude to 0000 UTC at 1,400 m. At 1,400 m the DIAL and GEM have maxima occurring at 1600-1700 UTC.
We note that numerical products tend to represent a smoother vertical transition of the diurnal cycle than the DIAL (in all seasons). This is expected due to their coarser resolution and need to preserve physical coherence/numerical stability through vertical levels. The abrupt shift at 1,000-1,200 m in the DIAL vertical profile of the diurnal cycle in the fall is possibly associated with the shift from the boundary layer to free troposphere (Hicks-Jalali et al., 2021). This physically driven shift is, however, difficult to be depicted by the DIAL diurnal cycle fit (as shown by lower values of the goodness-of-fit measure, R 2 ), given its variability (e.g., in altitude) during transition seasons (such as fall). Overall more noisy data, with less performing fit, are expected from the DIAL (or any observation source) than in numerical products, but still they might better reflect real physical phenomena.
The winter and spring are extremely dry in Iqaluit, making the diurnal cycles difficult to resolve and thus creating low R 2 values, particularly for the DIAL. Similar to the fall, the winter maxima below 200 m for both models and the DIAL occurs around 1800 UTC. The time of the maximum stays relatively constant with altitude, with the exception of a slight shift to 2000 UTC between 400 and 1,250 m. Amplitudes are generally low, with a maximum of 0.02 g⋅kg −1 up to 1,250 m, where the DIAL observes a marked increase in amplitude up to 0.08 g⋅kg −1 . There is almost no diurnal cycle present in ERA5, which is also 12 hr out of phase with respect to the DIAL above 1 km. GEM produces a slightly larger diurnal cycle than ERA5 and is slightly more in phase with the DIAL (in the first kilometre). GEM's maxima in the first 200 m are also larger than the DIAL's by 0.02 g⋅kg −1 . GEM produces an increase in amplitude above 1 km similar to the DIAL; however, GEM and ERA5 phases are shifted by 12 hr (with respect to the DIAL) at that altitude.
As the amount of daylight increases in the spring, the DIAL's lower-altitude maximum increases to 0.05 g⋅kg −1 and shifts to later in the day closer to 1900 UTC. However, above 400 m the DIAL observes a later maximum at 0000 UTC, with a second maximum appearing at 1700 UTC above 600 m. The more complicated structure in the cycle is due to the shift in phase of the S 2 component (as will be shown). ERA5 and GEM's cycle are shifted later by 2 hr compared with the DIAL near the surface, and neither model agrees with the DIAL's cycle above 500 m (both for phase and amplitude). Whereas ERA5's cycle is of comparable amplitude to the DIAL between 150 and 300 m, GEM's amplitude is 40% more than that of the DIAL's. Figures 3 and 4 separate the total diurnal cycle into their S 1 and S 2 components to further understand the behaviour of the total cycle. Figure 3 shows the amplitudes of the S 1 (row 1) and S 2 (row 2) components of ERA5, the DIAL, and GEM for each season, and Figure 4 provides the phases. The shaded region around each amplitude and phase profile is the standard deviation of the amplitude and phase solutions. The standard deviation is calculated by shifting the starting point of the time series from 0000 UTC to 2400 UTC to provide a distribution of solutions and examine the stability of the solutions. For ERA5 and the DIAL, this represents the sensitivity of the solutions to the edges of the time series. For GEM, it represents the stability of the solution with forecast lead time. We will discuss the standard deviations for GEM in more detail in Section 4.4.
The DIAL exhibits a clear correlation between amplitude and the seasonal cycle in the S 1 component, with the largest amplitudes in the summer (0.15-0.25 g⋅kg −1 ) and smallest in the winter (0.02-0.1 g⋅kg −1 ). The exception being between 450 m and 1 km, where the spring amplitudes are smaller than the winter amplitudes. The DIAL S 2 component exhibits a weaker correlation with the solar cycle, although the summer and fall cycle amplitudes are generally larger than the winter and spring cycles above 500 m, and the winter cycle has the lowest amplitudes for most altitudes. The S 2 amplitudes are smaller than the S 1 amplitudes. The DIAL amplitudes in both S 1 and S 2 components increase gradually with altitude, with the S 2 increasing faster than the S 1 . The standard deviations of the DIAL amplitudes are small for all altitudes and seasons; therefore, the DIAL solutions are stable in amplitude.
Unlike the DIAL, ERA5 and GEM's amplitudes do not increase with altitude and are instead fairly constant or exhibit a decrease with altitude (e.g., for GEM's S 1 amplitudes, for fall winter and spring). GEM's S 1 amplitudes exhibit a clear seasonal cycle with summer amplitudes between 0.1 and 0.2 g⋅kg −1 and winter amplitudes between 0.003 and 0.02 g⋅kg −1 . The standard deviation of the GEM amplitudes is larger than the DIAL's, particularly in the winter and spring when the water vapour concentrations are low. The winter S 1 and S 2 , as well as the spring S 2 , shading continues out of bounds of the figure due to the log scaling. The shading encompasses values close to zero, which are not shown here. ERA5 S 1 amplitudes are also correlated with the season; however, the spring amplitudes are larger than the fall amplitudes in the first 600 m. The standard deviations of the ERA5 amplitudes are also small, except in the fall. The larger standard deviation in amplitude in the fall may explain why ERA5's diurnal cycle does not resemble the DIAL or GEM's cycle in Figure 2. Both numerical products' S 2 amplitudes are an order of magnitude smaller than their S 1 amplitudes. Unlike the DIAL S 2 components, the ERA5 and GEM S 2 components are more correlated with season, with the exception of the first 400 m.
The phases of the individual diurnal cycle components are shown in Figure 4. Here, we show the phases as the hour (UTC) at which the first maximum occurs. For visualization purposes, the S 1 phases are continued past 24 hr. The GEM fall S 2 phase is also wrapped around 12 hr. The phase of the DIAL S 1 component is subject to a solar cycle in the first 200 m. In the summer, the maximum occurs at 2000 UTC and gradually moves towards earlier in the day at 1600 UTC in the winter. However, the S 2 component does not seem to be subject to a solar cycle at those altitudes and is, for the most part, constant with altitude. Above 500 m, almost all seasons shift by 5-10 hr from their surface phases. Above 1 km, all seasons either reach a constant phase with altitude (summer and fall) or shift back towards their surface values (winter and spring). The standard deviation of the DIAL phases are less than an hour in both S 1 and S 2 , suggesting that the DIAL solutions are also stable in phase.
ERA5 phases are shifted with respect to the DIAL's. The summer, winter, and spring S 1 phases are 2-4 hr later than the DIAL phases, and the ERA5 fall is completely out of phase with the DIAL by 15 hr, resulting in the disagreement seen in Figure 2. Unlike the DIAL S 2 components, which are constant with altitude, ERA5 S 2 phases start earlier than the DIAL at lower altitudes and gradually shift to later in the day. Although they are biased with respect to the DIAL, the solutions are relatively stable with standard deviations around 1 hr.
GEM produces S 1 phases that are close to the DIAL's in the first few hundred metres in the summer and fall, but the winter and spring maxima occur later by 3-4 hr. Although the GEM S 1 phases do shift towards later hours at higher altitudes, as the DIAL does, the GEM solutions shift more smoothly, and sometimes they reach the same phase as the DIAL at a higher altitude. For example, in the summer, the DIAL abruptly changes phase from 1700 UTC to 0200 UTC at 500 m, but the GEM S 1 cycle gradually shifts the phase to later in the day/early morning and reaches the 0200 UTC phase at 1,000 m. The GEM S 1 is out of phase with the DIAL in two altitude regions: above 1 km in the winter and between 800 and 1,200 m in the spring. The sizes of the GEM phase standard deviations are sensitive to low water vapour concentrations. In the spring, summer, and fall (larger water-vapour concentrations) the standard deviation values are small, but they gradually increase with altitude as the water vapour concentrations decrease. In the winter, when concentrations are small, the standard deviation values are larger (3-5 hr) throughout the profile.
The GEM S 2 phases shift smoothly with altitude and are not constant like the DIAL S 2 phases, with the exception of the winter. Whereas the summer GEM S 2 phase agrees well with the DIAL, the winter is shifted by roughly 3 hr for almost the entire altitude range. The GEM spring S 2 agrees well above 750 m but is shifted below by 2 hr. Finally, the fall S 2 agrees well below 500 m but quickly shifts out of phase above 500 m.
Total column diurnal cycles
The total column diurnal cycles in ERA5 and GEM were examined and evaluated against the GPS IWV measurements. As with the height-resolved cycles, the GPS, ERA5, and GEM IWV diurnal cycles are calculated by fitting Equation (3) to their respective IWV time series. Figure 5 shows the diurnal cycles for ERA5, the GPS, and GEM, by season, and Tables 2 and 3 list the corresponding amplitudes and phases of the S 1 and S 2 fitted components. As with the height-resolved cycles, the comparison starts at 1200 UTC. In some seasons, a discontinuity can be seen in each of the models (2400 UTC for ERA5, 36/1200 UTC for GEM) and the GPS time series (29/0500 UTC). The mechanism of the GPS discontinuity is discussed in Section 2. ERA5 is also known to have discontinuities in the surface temperature and humidity diurnal cycles (Hersbach et al., 2018a), which may also contribute to the ERA5 discontinuity in IWV. The different timing of the discontinuities is due to averaging the hourly bins and the times that were chosen to complete each dataset's analysis. The GPS measurement solutions are calculated over 24 hr in local time (Hicks-Jalali et al., 2021) and then shifted back to UTC, producing a discontinuity at 29/0500 UTC (since Iqaluit is at UTC−0500). The discontinuity in the GEM cycles at 36/1200 UTC is due to starting the time series hourly averaging at 1200 UTC. The ERA5 hourly average is calculated from 0000 to 2400 UTC, hence the discontinuity occurs at 2400 UTC. Though these discontinuities are unfortunate, and they can be mitigated with different averaging techniques (in the case of the GPS), they are known artefacts of the datasets and cannot be completely removed.
When comparing the GPS and GEM IWV cycles, we must also consider the slight interdependence of the results. As GEM assimilates GPS ZTDs and the GPS IWV values are derived from the same measurements, the GEM and GPS IWV values may have some dependence. However, the degree to which the GPS and GEM cycles disagree across all seasons suggests that the dependence on each other is minimal and does not affect the results significantly. Additionally, the GEM and ERA5 (where no GPS values are assimilated) IWV results show very similar behaviour, suggesting that the GPS influence on the GEM IWV cycle is minimal.
During the summer, all three datasets' diurnal cycles are primarily driven by the S 1 component. Though the GPS cycle is primarily driven by the S 1 component, it has a significant contribution from the S 2 component as well. The extended peak is due to the constructive interference between the two components and their slight offset in phase. The ERA5 summer IWV total cycle is almost entirely driven by the S 1 component, which is slightly larger in amplitude than the GPS by 0.08 mm ( Table 2). The S 1 component also peaks 2 hr later than the GPS S 1 component (Table 3). GEM also produces a larger S 1 component (0.31 mm) relative to the S 2 , although its S 2 cycle is of similar magnitude to the GPS's (0.06 mm). GEM produces both S 1 and S 2 maxima roughly 1 hr later than the GPS observations. The fall cycles have a larger contribution from the S 2 component in both numerical products and in the GPS. The GPS S 2 amplitude increases to the same magnitude as the S 1 , and the superposition of the two components creates the largest maximum at 31/0700 UTC. The GPS and ERA5 measurements have larger discontinuities in the fall due to the rapid decrease in water vapour per day (Hicks-Jalali et al., 2021). ERA5 has almost no S 1 component during the fall, and though the GEM S 1 amplitude is larger than ERA5's it is still smaller than the GPS amplitude. The ERA5 and GPS S 1 components peak earlier than the GEM S 1 maximum by 14 hr. The ERA5 and GEM S 2 components are half that of the GPS S 2 amplitude and they are both off-phase by more than 5 hr with respect to the GPS.
Winter IWV cycles are equally influenced by both components. Both ERA5 and GEM have similar cycles in amplitude (0.02 mm for S 1 and S 2 ) and phase (0400-0500 UTC for S 1 and 0800-0900 UTC for S 2 ). On the other hand, the GPS cycles are slightly larger, as was observed in the DIAL's height-resolved cycles. The ERA5 S 1 maxima occurs 3 hr later than the GPS maxima, whereas the GEM S 1 maxima occurs 1 hr later than the GPS maxima. The GPS S 2 maxima occur earlier than either ERA5 or GEM by about 2-3 hr. In the spring, both numerical products reproduce the increase in the S 1 component; however, only GEM reproduces it at the same magnitude as the GPS (0.05 mm). ERA5 produces the S 1 peak 3 hr earlier than the GPS, and 2 hr earlier than GEM. The most significant difference between the numerical products and the GPS in this season is that the magnitude of the S 2 component is five times larger for the GPS than the models. The ERA5 S 2 component is 5 hr 30 min earlier than the GPS component, whereas GEM is 1 hr 30 min earlier.
Lead-time dependence
In this section we present the stability of the diurnal water-vapour cycle with respect to the starting point of the time series for the ERA5 and DIAL measurements. For the GEM model, we analyse the sensitivity of the water-vapour cycle as the lead time of the forecast solution is gradually increased (whereas ERA5 and the DIAL do not have lead times). To examine the stability of the diurnal cycle solutions, we shifted the starting points of ERA5's and DIAL's time series for each hour such that the time series were calculated by taking the 24 hr from 0000 UTC, 0100 UTC, 0200 UTC, up to 2400 UTC and recalculating the diurnal cycle solution for each shifted time series. In GEM's case we always consider forecasts with initialization time at 0600 UTC; the first calculation will use lead times from initialization through the 24 hr forecast, then the 1 hr forecast lead time through the 25 hr forecast lead time, and so on until the 24 hr forecast through the 48 hr forecast. When discussing GEM we will use the term "lead time", whereas we will use "starting point of the time series" with ERA5, the DIAL, and the GPS, as they do not have lead times. We analyse the effects of changing the starting point of the time series (for ERA5 and the DIAL) and gradually increasing the lead time (for GEM) on the biases of the amplitude and phase of the water vapour diurnal cycle, as fitted by Equation (3). We found no systematic features in the bias for a specific lead time, indicating that no particular lead time in GEM was better than the others for the GEM S 1 amplitudes are consistently larger than the DIAL's in the first 200 m by 0.02-0.05 g⋅kg −1 for every season. As the altitude increases, GEM S 1 bias is increasingly negative compared to the DIAL's S 1 amplitude. This is because the DIAL's S 1 amplitude tends to increase with altitude, while the GEM's either decrease or remain the same. The standard deviation of the GEM − DIAL bias is small, between 0.02 and 0.003 g⋅kg −1 (5-45% of the corresponding seasonal diurnal amplitude) depending on the season and altitude. ERA5 − DIAL S 1 amplitude biases vary by season in the first 200 m (blue solid line in Figure 6) but overall agree well with the DIAL. ERA5 − DIAL S 1 amplitude bias behaves similarly to the GEM − DIAL bias (decreasing with altitude) and is more negative than the GEM − DIAL bias in summer and spring. The standard deviation of the ERA5 − DIAL bias is also small, on the order of 0.01-0.02 g⋅kg −1 . The region between 450 m and 1 km is where the best agreement of the numerical products with the DIAL occurs, particularly for GEM. There is worse agreement close to the surface and above the planetary boundary layer.
In the summer and spring, GEM S 1 amplitudes are consistently larger than ERA5's by 0.05 g⋅kg −1 , whereas amplitudes are similar in the winter. In the fall, GEM − ERA5 amplitude bias is positive for lower altitudes and then it becomes neutral above 700 m. The standard deviation of the bias is generally larger between GEM and ERA5 than between the numerical products and the DIAL.
Whereas the S 1 amplitude biases between the numerical products and the DIAL varied with season, the S 2 amplitude biases did not. As seen in Figure 3, the ERA5 and GEM S 2 amplitudes are smaller than the DIAL S 2 amplitudes, and their difference increases with altitude. The best agreement between the DIAL and the models is in the first 200 m, where the amplitudes are similar. However, this is because the S 2 amplitude is almost negligible at F I G U R E 7 Same as Figure 6, except for diurnal cycle phase. [Colour figure can be viewed at wileyonlinelibrary.com] those altitudes. As the S 2 amplitude increases with altitude and becomes more prominent above 1 km, the discrepancy between the numerical products and the DIAL worsens and the amplitude differences between them increases. The standard deviation of the S 2 amplitude biases is fairly constant with season and altitude, around 0.005 g⋅kg −1 . The GEM − ERA5 S 2 amplitude bias is almost neutral (slightly positive), for all seasons and altitudes, and again exhibits a larger standard deviation than the bias of the numerical products against the DIAL. The larger standard deviation of the GEM − ERA5 bias is due to the larger standard deviation of the GEM solutions compared with ERA5. Their almost zero bias between the two numerical products suggests that the models both have a difficult time reproducing the S 2 component. Figure 7 is calculated in the same manner as Figure 6, but for the phases. The phase bias is calculated in hours. Similar to the amplitude biases, for S 1 we see the best agreement between the DIAL and GEM in the first 200 m, with biases ranging from 0 to 5 hr and standard deviations of the bias on the order of 0.5 to 1 hr. The GEM − DIAL S 1 phase solutions are most stable in the summer and fall, with very small standard deviations, whereas they exhibit a large spread in winter (same as the GEM − ERA5 S 1 bias). The GEM − DIAL biases switch with altitude between positive and negative values in summer and fall, whereas spring is dominated by a negative bias (GEM maximum occurs earlier than in the DIAL). Winter S 1 GEM phase solutions were highly sensitive to lead time. This was likely due to the low concentrations of water vapour, which made fitting difficult. Additionally, the models have difficulty replicating the diurnal cycle phase above 1 km in the winter (as seen in Figure 7), which likely contributes to the higher standard deviations between the GEM and DIAL above 1 km. The ERA5 − DIAL winter phase biases, on the other hand, were not sensitive to the time-series shift.
The S 2 phases show the best agreement between DIAL and the numerical products in summer and the worst agreement in winter. In the shoulder seasons (spring and fall) the S 2 phase bias is positive for all comparisons, indicating that the ERA5 peak occurs later than the DIAL (blue), that the GEM peak occurs later than in ERA5 (green), and that the GEM peak occurs later than in the DIAL (red). The standard deviations of the S 2 GEM − DIAL and GEM − ERA5 biases increase with altitude, except in the winter, likely due to the smaller water vapour concentrations above the boundary layer.
Similar to the height-resolved cycles, we evaluated model IWV diurnal cycles biases for their dependence on lead time and starting hour of the time series. Results for the amplitude biases are presented in Table 4 for each season. In the summer, the GEM model has a significant positive bias with respect to both the GPS and ERA5, whereas the ERA5 − GPS bias is small. The larger positive bias from TA B L E 4 Average Integrated water vapour (IWV) amplitude bias and standard deviation with changing lead time. GEM is likely due to an assimilation effect, which also contributes to the larger standard deviations and sensitivity to lead time. In the fall, the GEM bias with respect to GPS improves, but the standard deviation remains slightly high. In the fall, the ERA5 and GEM biases with respect to the GPS are similar in magnitude and indicate an underprediction of the amplitude of the numerical products.
In the winter, both ERA5 and GEM had average S 1 amplitude biases lower than the GPS and both biases were stable with respect to the diurnal cycle initial hour. GEM and ERA5 solutions were similar with respect to each other, with a small average bias of −0.004 mm. In the spring, ERA5 had an overall negative bias with respect to the GPS, and the solution had little dependence on the starting point of the time series. GEM agreed well with the GPS with an equally small standard deviation in the bias. As expected, GEM − ERA5 S 1 amplitude exhibits a positive bias of 0.02 mm.
The S 2 IWV amplitude biases for the summer behave differently than in the other seasons. In the summer, ERA5 underestimate the amplitude with respect to GPS (by 0.033 mm), whereas GEM slightly overestimates it (by 0.006 mm). As expected, the difference between GEM and ERA is of 0.039 mm. The spread of the biases when comparing GEM with either GPS or ERA5 is significantly larger than when comparing GPS versus ERA5, suggesting once again (as for S 1 ) a stronger sensitivity to the GEM lead time than to the GPS and ERA5 initial hour for the time-series fit. In fall, winter, and spring, both GEM and ERA5 underestimate the S 2 amplitude (by 0.05 mm, 0.03 mm, and 0.04 mm respectively) with respect to the GPS. As expected, the GEM − ERA5 bias is almost null in these seasons. The spread for these seasons shows stability in the results.
The average IWV biases in phase and their standard deviations are presented in Table 5. With the exception of the fall GEM solution, ERA5 and GEM in general reproduce the S 1 phase within 1-3 hr of the GPS S 1 phase. The standard deviation of the biases varies somewhat more with the season, being stable in the summer (0.15-0.58 hr) and less stable in the fall and winter. Though the ERA5 and the GPS S 1 phases agree well in the fall, the GEM S 1 arrives almost 9 hr earlier. The high standard deviation in the fall biases is likely caused by the larger discontinuity in the ERA5 and GPS solutions and the steep decrease in water vapour over the season. In the winter, ERA5 and GEM exhibit similar average phase biases, but GEM's variation in the bias is much larger due to the large variability in the specific humidity height-resolved winter solutions. However, the spring GEM bias (−0.9 hr) and variability in the bias (−0.96 hr) improve as the water vapour concentrations increase.
The S 2 phase biases and their stability vary more with season compared with the S 1 phase solutions. In the summer, both ERA5 and GEM have small average phase biases with respect to the GPS, but their standard deviations are large (4 hr). The S 2 IWV amplitudes are at least half the size of the S 1 amplitudes in the summer. Their effect on the overall cycle is minimal; therefore, they are more sensitive to changes in the starting point of the time series. The opposite occurs in the fall, where both the GEM and ERA5 S 2 components have a large bias (−6 hr) yet small standard deviation (0.4 hr), suggesting that their solutions are stable with respect to the GPS and consistent across all lead times. In the winter, both ERA5 and GEM have similar S 2 phase biases with respect to the GPS, but the standard deviation of their biases are relatively small. Lastly, the spring average S 2 phase bias improved for GEM (−0.711 hr) but increased for ERA5 (−5.856 hr). However, the GEM − GPS and GEM − ERA5 biases are less stable, likely due to the larger standard deviation in the higher altitudes of the GEM height-resolved cycles.
Connecting the height-resolved and IWV cycles
Both ERA5 and GEM numerical products agreed well with the DIAL and GPS seasonally. As expected, the summer months had the largest average water vapour concentrations (around 3 g⋅kg −1 ), and spring was the driest season (between 0.5 and 0.7 g⋅kg −1 ). Both models also accurately reproduced (with respect to the DIAL, Figure 1) the variability in the water-vapour profiles at all altitudes. The GEM and ERA5 seasonal IWV averages all agreed with the GPS averages within their respective standard deviations, except for a small 1 mm bias in the winter and spring. This bias is likely caused by the assimilation of radiosondes, which are known to have a dry bias in cold temperatures (Miloshevich et al., 2009;Dirksen et al., 2014). The average specific humidity profiles in this study are consistent with others conducted at similar latitudes (Serreze et al., 1995;Jakobson and Vihma, 2010;Graham et al., 2019). Graham et al. (2019) compared ERA5 with 27 independent radiosonde measurements over the Greenland Sea and showed that ERA5 was consistently within ±0.15 g⋅kg −1 of the radiosondes, with the sign depending on the altitude. We observed slight biases of less than 0.25 g⋅kg −1 between the DIAL and both numerical products, mostly due to sampling differences. Mariani et al. (2021) evaluated GEM at Iqaluit using measurements from September 2018 to August 2019 and found a mean bias of −0.16±0.02 g⋅kg −1 (GEM − DIAL) for the entire year. In this study, we expanded the analysis in Mariani et al. (2021) to include measurements from January to June 2020 and found a similar mean bias of −0.13 ± 0.44 g⋅kg −1 . Note that in our study we use the average bias plus/minus the standard deviation of the bias, whereas Mariani et al. (2021) used the standard deviation of the mean. Mariani et al. (2021) also found that GEM has a slight moist bias around 150 m.
ERA5 and GEM produced mixed results when evaluated at the diurnal time-scale (Figure 2), varying with season and altitude. In general, the DIAL and ERA5 diurnal-cycle solutions were stable and showed very little variation when the starting point of their time series was changed (Figures 3,4,6, and 7). The GEM model solutions were more sensitive to decreasing water vapour concentrations, with low standard deviations in their solutions below 1 km and in the summer and fall, but the variability increased in the winter and spring and above 1 km for all seasons.
In the summer, ERA5 is able to reproduce the overall phase of the diurnal specific-humidity cycle, but it lacks the structure and amplitudes observed by the DIAL measurements. The structure observed in the total diurnal cycle in the DIAL is due to the influence of the S 2 component (Hicks-Jalali et al., 2021), which is an order of magnitude smaller in ERA5 than in the DIAL. The ERA5 summer S 1 amplitudes are smaller by 0.1 g⋅kg −1 for all altitudes. Thus, the combination of smaller diurnal amplitudes in both S 1 and S 2 and the lack of variability in phase in both components creates a smaller ERA5 IWV diurnal cycle with a broader maximum compared with the DIAL. Similar to ERA5, GEM lacks the structure and variability of the DIAL cycle in the summer, but it is better able to reproduce the magnitude of the diurnal height-resolved cycle than ERA5 is. However, GEM produces larger S 1 amplitudes than DIAL by 0.06 g⋅kg −1 at 150 m. Additionally, above 1 km the GEM S 2 component is an order of magnitude smaller than the DIAL S 2 amplitude and is out of phase; consequently, it does not reproduce the interaction between the S 1 and S 2 components to create the increase in total diurnal amplitude observed by the DIAL. Hicks-Jalali et al. (2021) found that the DIAL's height-resolved and GPS's IWV diurnal cycles corroborated each other in phase, which we also find to be true in ERA5 and GEM and their corresponding IWV cycles. The summer ERA5 IWV cycle has a prominent S 1 component and almost no S 2 component. The ERA5 S 1 IWV component is also larger than the GPS S 1 component by 0.08 mm. The GEM total IWV diurnal cycle is also dominated by the S 1 component, although the S 2 component is of similar magnitude to the GPS S 2 amplitude. The lack of S 2 component in the ERA5 IWV diurnal cycle is consistent with the very small S 2 component present in the height-resolved cycles.
The larger S 1 amplitude in ERA5 suggests that ERA5 expects a larger diurnal cycle than what the GPS measures; however, the height-resolved diurnal cycles were smaller in amplitude than in the DIAL. One possibility for the discrepancy is the unfortunate coincidence that the ERA5 IWV cycle maximum occurs at the same time as the discontinuity in the ERA5 cycle; thus, this maximum may have been artificially increased. It is also possible that the DIAL amplitudes above 1 km are overestimated. The R 2 values in that region are low; however, Hicks-Jalali et al. (2021) found that the summer amplitudes at those altitudes were in agreement with Jakobson et al. (2014). Hicks-Jalali et al. (2021) also compared the summer GPS cycles with the Jakobson et al. (2009) GPS diurnal cycles and found that although the Iqaluit GPS diurnal cycles were on the low end of the distribution they still agreed with the Jakobson et al. (2009) observations. The bias in ERA5's IWV amplitude could be related to an overestimation of evapotranspiration at Iqaluit.
The GEM IWV S 1 amplitude also exhibited a large average bias (0.2 mm) with respect to the GPS S 1 component (Table 4). This positive bias in the IWV S 1 component in GEM is corroborated by the corresponding positive bias in the height-resolved S 1 component in the first 200 m in the DIAL observations and is likely due to the assimilation of the GPS ZTD measurements at the 0600 UTC initialization time. During the summer, the assimilation effect was larger than the other seasons and persisted to the sixth hour lead time. At later lead times the bias between GEM and the GPS decreased (later lead times exhibited decreased bias between GEM and GPS; not shown). The large change in the bias with lead time was apparent in the larger standard deviation of the GEM − GPS bias shown in Table 4. The larger amplitudes in the first few hundred metres of the height-resolved cycles are likely affected by the ZTD assimilation as well, because the hydrometric component of the ZTD is dominated by boundary-layer water vapour. Another possibility is that the GEM model also overestimates the evapotranspiration rate for Iqaluit.
In the fall, ERA5 has almost no diurnal cycle present due to extremely small amplitudes in both S 1 and S 2 components at all altitudes. The DIAL total diurnal cycle is driven by the S 1 component for the first 1 km. Above 1 km the DIAL has a large diurnal maximum that is created via the superposition of the S 1 and S 2 components. Though GEM correctly reproduces the S 1 phase below 1 km, the S 2 component phase is too early and is an order of magnitude smaller than the DIAL S 2 amplitude, such that it cannot create the same intense maximum above 1 km.
The fall GPS IWV cycles are clearly influenced by the S 2 component; however, the DIAL height-resolved cycles are primarily driven by the S 1 component up to 1 km. We can examine how the height-resolved cycles influence the total IWV cycle by looking at the similarities in phase between the two cycles. The GPS IWV maxima correspond to the maximum from 100 m to 1 km in the DIAL and the 30/0600 UTC maximum above 1 km. Both the GEM S 1 and S 2 IWV amplitudes are smaller in magnitude than the GPS IWV amplitudes, thus creating a smaller total cycle than the GPS cycle. The GEM IWV total cycle is mostly characterized by the increased S 2 amplitude (0.05 mm). As with the DIAL and GPS cycles, the GEM S 2 IWV phase corresponds with the GEM S 2 phase at 1,400 m, with a slight offset of 1 hr. The phase differences between the GEM components and the DIAL components above 1 km appear to have a significant effect on their respective total IWV diurnal cycles. The GEM IWV maxima of the total diurnal cycle are shifted by 3 hr with respect to the GPS due to the phase shift of the GEM S 2 component above 1 km, and the GEM IWV cycle has a deep minimum at 32/0800 UTC that is not present in the GPS cycle. Thus, the phases of the height-resolved cycle components impact the behaviour of the IWV cycle.
It is difficult to resolve a diurnal cycle in the DIAL and numerical products in the winter and spring months due to the lower water vapour content. Nevertheless, as discussed in Hicks-Jalali et al. (2021), the agreement in phase between the DIAL and GPS and in the surface amplitudes from the WMO surface station provide confidence in the DIAL results. In the winter, ERA5 has almost no diurnal cycle present for the entire altitude range, and GEM has a small amplitude of 0.03 g⋅kg −1 at 150 m (a small positive bias of 0.02 g⋅kg −1 with the DIAL). GEM and ERA5 are not able to reproduce phase shifts and the S 2 component above 500 m that are observed by the DIAL. Most notably, the DIAL observes a large maximum at 1700 UTC above 1 km created by the superposition of the two components and increased amplitude of the S 2 component. Though both GEM and ERA5 reproduce increases in magnitude at those altitudes, they are 12-14 hr out of phase with the DIAL. The GEM and ERA5 S 1 and S 2 components are shifted with respect to each other, resulting in a total extended maximum instead of a defined peak.
As observed in the fall, the winter IWV cycles seem to reflect the height-resolved cycles around or above 1 km. The GPS cycle is clearly defined by the S 2 component, with maxima that correspond to the DIAL S 2 phases between 900 and 1,250 m. ERA5's IWV S 2 phase is consistent with its height-resolved S 2 phase around 1,200 m, as are the GEM cycles. The superpositioning of the DIAL height-resolved cycles above 1 km results in stronger maximums in the GPS IWV cycles. In GEM and ERA5, the S 1 and S 2 components are out of phase and have smaller S 2 amplitudes, such that they get more of an extended IWV maximum with smaller S 2 peaks.
In the spring, the ERA5 and GEM S 1 components at 150 m are in phase with the surface cycle and agree well in magnitude; however, the GEM S 1 amplitude has a slight positive bias of 0.025 g⋅kg −1 around 50 m. Though the DIAL S 2 amplitudes increase in altitude, the GEM and ERA5 height-resolved mixing ratio amplitudes decrease, thus limiting the IWV S 2 component. Both ERA5 and GEM do not produce the S 2 component during the spring or the winter. The lack of the S 2 component in spring and winter, and in the summer as well, suggests that there is a process (or several) that is not being well represented in the models.
The results of the fall, winter, and spring cycles indicate that the IWV cycles are heavily influenced by the water vapour above 1 km, and less so by the water vapour close to the surface. Jakobson et al. (2014) also found that 60% of the diurnal variability over land was driven by the layers between 800 and 900 hPa, which corresponds to approximately 500-2,000 m at Iqaluit. The behaviour observed here agrees with their results and is easily observed in the fall and winter when the S 1 amplitude is small enough to allow the S 2 component to dominate. Louf et al. (2015) examined height-resolved diurnal water vapour cycles in Africa using a microwave radiometer. Their fig. 6 depicts the height-resolved cycles and IWV cycles of the total column, a column below 1,400 m and another above 1,400 m. The column above 1,400 m has more water vapour, and their fig. 6 indicates that the total cycle can be determined by the summation of the two layers. In their fig. 6b, the total cycle is a summation of the two layers but retains the overall behaviour of the layer above 1,400 m. This would suggest that the diurnal IWV cycle may strongly depend on the 800-900 hPa region; however, more investigation is required on a global scale to verify if this is true elsewhere.
Possible causes of the differences between the model and observation diurnal cycles
This is the first time that the GEM diurnal water-vapour cycles have been evaluated, whereas a few recent papers have evaluated the IWV diurnal cycles in ERA5 in the Tropics. Lees et al. (2021) evaluated ERA5 diurnal IWV cycles around the south Indian Ocean basin. They also found that that the ERA5 IWV diurnal amplitudes were underestimated with respect to the GPS measurements. Xue et al. (2020) analysed the diurnal cycles of upper tropospheric humidity between ±60 • latitude using brightness temperatures from satellite-based imagers. They found that all five reanalysis datasets (including ERA5) had smaller diurnal amplitudes than the observations and that the phase of the maximum humidity lagged behind the observations by 3 hr. Our results corroborate the findings from both papers, that both ERA5 IWV and height-resolved amplitudes are smaller than those observed.
There are several factors that may be influencing the differences between the two models themselves and between the models and observations. The first major difference is their resolutions. ERA5's horizontal resolution is significantly coarser than the resolution of GEM. Other studies have shown that models with higher resolutions are more successful at reproducing diurnal cycles of precipitation and convection. Sato et al. (2009) tested three different resolutions of the Nonhydrostatic Icosahedral Atmospheric Model between 14 km, 7 km, and 3 km across the globe between ±30 • latitude to evaluate the model's precipitation diurnal cycle. They found that the 3 km model outperformed the lower resolution models over both land and the ocean. suggests that at higher model resolutions there is less need for deep convection parametrization; however, more attention must be paid to the microphysics at kilometre scales. GEM does have a deep convection scheme (Kain and Fritsch, 1990) in addition to the predicted particle properties microsphysics scheme. ERA5 also has a deep convection parametrization scheme that did improve the diurnal cycle of precipitation and convection in the Tropics (Bechtold et al., 2014). However, the Bechtold et al. (2014) model is dependent on the level of convective available potential energy, which may not work as well for sites at higher latitudes where there is little to no convection. We used vertical wind measurements from the wind lidar to examine convection in Iqaluit during the course of this study to verify this; however, we could not detect a diurnal signal in the vertical wind during any season, suggesting that convection is not very strong at Iqaluit. ERA5 also made improvements to their microsphysics scheme (Ahlgrimm and Forbes, 2014), and though it did improve top-of-atmosphere short-wave radiation bias in the model globally, it did not improve the negative bias in top-of-atmosphere short-wave radiation around Iqaluit. Future work is required to evaluate how well Arctic clouds are reproduced in GEM and ERA5 at Iqaluit.
The ratio of land to water in each grid point is another significant difference between the two models. The Iqaluit ERA5 pixel is 50% split between land and water, whereas the GEM grid point is 80% land and 20% water. Iqaluit is far enough north that sea breezes are not much of an influence on the diurnal cycle and can only occur during July and August Mariani et al. (2018). However, a lack of sea-surface temperature information and the influence of the ice in Frobisher Bay may not be completely accounted for in ERA5, resulting in low diurnal amplitudes in the fall and winter.
GEM exhibited a small but consistent positive bias in the near-surface bins with respect to the DIAL, particularly in the summer months. Land-surface processes (such as evapotranspiration) and other complex processes (such as water vapour flux from snow) may need to be examined in further detail to explain this bias. The type of vegetation assumed for Iqaluit and the seasonal soil moisture are also variables that would contribute to the evapotranspiration estimates and should be evaluated in future analyses. There have been fewer opportunities to evaluate the evapotranspiration process for vegetation found in Iqaluit and the tundra environment in GEM in comparison with lower latitude Canada, and as such it may not be accurately represented. Both GEM and ERA5 assimilate soil moisture measurements, either from satellites or from ground-based in-situ measurements (Albergel et al., 2012;De Rosnay et al., 2013;Carrera et al., 2015). However, in GEM's case, the soil moisture measurements are only assimilated in grid points with a land/water ratio higher than 90% and are not assimilated if snow, frozen soil, or precipitation is present in the nature run (Carrera et al., 2015). Therefore, soil moisture is likely not assimilated for the Iqaluit grid point in this study, which could be causing some of the bias in the near-surface cycles. GEM and ERA5 may also be assuming that the Iqaluit soil is not frozen in the summer, and though the top layers of soil will melt in July and August, the lower layers will remain frozen, which would impact the amount of moisture available to evaporate in the summer.
Given Iqaluit's location and the lack of observations available to assimilate in this region, models are generally forced to rely on assimilating satellite measurements. However, satellite-based instrument measurements are generally limited in temporal coverage of a specific location. Additionally, satellite-based instrument humidity measurements usually have low vertical resolution (several hundred metres to kilometres) and either have no measurements or have high uncertainties in the boundary layer. ERA5 significantly increased its assimilated observations worldwide; however, Hersbach et al. (2020) did not differentiate the number of measurements by latitude. GEM is able to assimilate more radiosondes and aircraft measurements than previous versions , but since those measurements do not have high temporal resolution they would have less of an effect on the diurnal cycle. Additionally, ERA5 does not assimilate ZTD, whereas GEM does for every run. Assimilating the GPS measurements improved the IWV forecasts over North America (Macpherson et al., 2008;Buehner et al., 2015). Assimilating the ZTDs, which are mostly influenced by the water vapour in the first few kilometres, may help constrain the boundary-layer humidity in GEM. Nevertheless, it is clear that the models would benefit from additional higher temporal and vertical resolution humidity measurements in the Arctic.
CONCLUSIONS
Diurnal cycles are one of the most challenging components of numerical weather prediction and climate models to reproduce due to their complexity and dependence on numerous processes. High-frequency thermodynamic measurements in the Arctic are difficult to maintain for extended periods of time due to the harsh environment and large expense, making evaluations of numerical weather and climate models difficult. In this article, we have evaluated the seasonal and diurnal water vapour behaviour in the ECMWF ERA5 model and the ECCC GEM NWP model against a preproduction DIAL system and ground-based GPS observations taken at the ECCC Iqaluit supersite. Both ERA5 and GEM were able to reproduce the seasonal averages and the standard deviation of the water vapour observed by the DIAL and the GPS. However, although the numerical products perform well on the seasonal scale, the diurnal cycle remains challenging, particularly when resolved by height. Near-surface, summer height-resolved, and summer IWV cycles showed the best agreement between both numerical products and the observations. Winter and spring exhibited the worst agreement in the height-resolved cycles due to low water vapour concentrations. Though GEM was able to reproduce the fall cycle below 1 km, it struggled above; ERA5 was unable to resolve a height-resolved or IWV cycle in the fall.
Both numerical products have trouble reproducing the diurnal cycle above 1 km, particularly the S 2 component. Other studies have shown that ERA5, although it has improved over ERA-Interim, still produces diurnal amplitudes that are smaller than those observed. We showed that, in Iqaluit, ERA5 S 1 amplitudes agreed with the DIAL in the summer and spring close to the surface but were smaller above 500 m by 0.1 g⋅kg −1 . The ERA5 S 2 amplitudes were significantly smaller than those of the DIAL and GPS at all altitudes and were generally out of phase. The lack of an S 2 component significantly affected the numerical product IWV cycles, such that their amplitudes were also smaller than the GPS IWV amplitudes except in the summer, when their amplitudes were larger than the GPS'. ERA5's bias in phase and amplitude was constant with shifting the time series starting point, so we can conclude that the solution was mostly unaffected by changing the starting point of the time series.
The GEM diurnal cycles were closer to the DIAL and GPS than the ERA5 cycles. This was likely primarily due to GEM's higher resolution (2.5 km), which has been demonstrated to improve diurnal cycles in previous studies, but also potentially due to the grid-point tile water/land composition. GEM also assimilates GPS measurements, which provide additional water vapour information in the lower troposphere and boundary layer, whereas ERA5 does not. GEM is also designed specifically for Canada. The GEM S 1 component agreed well with the DIAL below 1 km and was mostly in phase with the DIAL results below 1 km. However, above 1 km the GEM S 1 amplitudes were typically smaller than the DIAL S 1 amplitudes. Like ERA5, GEM had trouble reproducing the S 2 component in both the height-resolved profiles and consequently in the IWV cycles as well. The DIAL measurements revealed an S 2 amplitude that increases with altitude, which neither ERA5 nor the GEM could reproduce. The GEM bias with respect to the DIAL and the GPS was fairly constant, with minimal dependence on lead time. Higher altitudes were subject to larger variation, presumably because the amplitudes were smaller at those altitudes and the average water vapour content was much smaller (typically 1 g⋅kg −1 less than at the surface).
This article has shown that the S 2 component is a critical process in Iqaluit's diurnal cycle. The inability to reproduce the S 2 component in the diurnal cycle profiles has important consequences on the accuracy (magnitude and phase) of the models' net (sum) diurnal cycle, possibly affecting other atmospheric processes, such as precipitation. We found that the IWV diurnal cycle is highly dependent on the diurnal water vapour cycle between 500 m and 2,000 m. In this region, the influence of the S 2 component becomes visible due to its increase in amplitude and its constructive interference with the 24 hr component. If the S 2 component is incorrect, it affects the IWV cycles such that their amplitudes are either too small or they are shifted in phase with respect to the GPS. Total column and humidity profiles are crucial components of the radiative transfer equations that govern NWP. Incorrect timing of the diurnal component can have consequences on the timing of precipitation, cloud formation, or convective processes. Finding the drivers of the S 2 component is an important next step in improving diurnal cycles in NWP. Future work will involve investigating measurements of pressure, precipitation, soil moisture, cloud, and surface turbulent flux diurnal cycles at Iqaluit and comparing them with ERA5 and GEM to hopefully determine which processes are the source of the discrepancies. | 2023-07-11T15:06:28.151Z | 2023-07-08T00:00:00.000 | {
"year": 2023,
"sha1": "ae3a087908c0d8f7c2d4ae26bb49962ce18fcef8",
"oa_license": "CCBYNC",
"oa_url": "https://doi.org/10.1002/qj.4520",
"oa_status": "HYBRID",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "9c5b8b0ae4108425a633ab0c677ff86fc0a1a0d9",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": []
} |
248689439 | pes2o/s2orc | v3-fos-license | Pedestrian Detection Using Integrated Aggregate Channel Features and Multitask Cascaded Convolutional Neural-Network-Based Face Detectors
Pedestrian detection is a challenging task, mainly owing to the numerous appearances of human bodies. Modern detectors extract representative features via the deep neural network; however, they usually require a large training set and high-performance GPUs. For these cases, we propose a novel human detection approach that integrates a pretrained face detector based on multitask cascaded convolutional neural networks and a traditional pedestrian detector based on aggregate channel features via a score combination module. The proposed detector is a promising approach that can be used to handle pedestrian detection with limited datasets and computational resources. The proposed detector is investigated comprehensively in terms of parameter choices to optimize its performance. The robustness of the proposed detector in terms of the training set, test set, and threshold is observed via tests and cross dataset validations on various pedestrian datasets, including the INRIA, part of the ETHZ, and the Caltech and Citypersons datasets. Experiments have proved that this integrated detector yields a significant increase in recall and a decrease in the log average miss rate compared with sole use of the traditional pedestrian detector. At the same time, the proposed method achieves a comparable performance to FRCNN on the INRIA test set compared with sole use of the Aggregated Channel Features detector.
Introduction
Pedestrian detection, a fundamental task in computer vision, can assist in self-driving, the monitoring of crowded environments, and other activities by automatically detecting and localizing human bodies from images. This type of task is challenging, mainly because human bodies are depicted in numerous ways, hindering the overall description of their features. Firstly, human bodies distinguish themselves from each other in appearance, for example, the color of the skin, clothes, hairstyle, body structure, and pose. Secondly, human bodies often suffer from occlusion in crowded environments, such as train stations, shopping malls, and others. Occlusion reduces the area of a body exposed to smaller irregular-shaped areas. Lastly, images are sensitive to the image acquisition setup, such as the varying illumination conditions, view angles, and resolution. For instance, poor illumination or overexposure result in low-contrast images, which blur the human bodies. Furthermore, the human body shape appears different from various viewing angles, for example, from frontal and profile views. Finally, low-resolution human bodies cannot be easily identified, even by the human eye. To handle the diversity of human bodies, it is imperative to describe their appearances effectively using a robust and rich feature set that allows human bodies to be discriminated effectively from the background.
Feature extractors based on deep learning have drawn extensive attention in recent years due to their outstanding performance in extracting high-level semantic information and its large number of features, which has improved the detection performance dramatically. Those methods require training on extremely large datasets to learn representative and robust features; otherwise, the extracted features may be not generalized enough. Because of this, advanced deep learning methods often require massive volumes of annotated data and computational resources such as high-performance computers with GPU facilities. Considering this dilemma, many researchers have focused on the improvement of widely used traditional handcrafted features or the combination of Convolutional Neural Networks (CNNs) with manually extracted features for various computer vision tasks, including pedestrian detection [1][2][3][4][5][6].
One of the earliest popular families of handcrafted features for describing object silhouettes is the histogram of edge orientations, which was initially used for hand gesture recognition [7]. These features are signals that represent occurrences of gradient orientations in localized portions of an image. Another popular contour-based model is the so-called Shape Context [8], designed for measuring shape similarity for the purpose of shape matching. It is defined as a histogram of the relative coordinates of a reference point to a predefined set of neighboring points. This feature is suitable for matching objects whose contours (edges) are corrupted by weak noise or are noiseless and are placed on simple or ideally uniform backgrounds. Considering the complexity of real-life crowded environments, this feature is not recommended for use in human detection. In contrast to those two features, Haar-like features have been used for human detection [9,10] since the early stages. An image is filtered by several Haar wavelets with different predesigned patterns to extract edge features, line features, and center-surround contrast features. The Haar-like feature method is more suitable for objects with simple structures, for example, facial features, which have relatively simple structures. Otherwise, highly textured structures, like grass and trees, which challenge the sufficient representation of predesigned limited Haar wavelets can cause false positives [10]. Another of the most well-known features is the Scale-Invariant Feature Transform (SIFT) model, which consists of the position, scale, and orientations at selected key points, which are "interesting" points of the image signal, for example, contour corners [11]. An advanced version of SIFT is the Principal Components Analysis SIFT (PCA-SIFT) which uses PCA to represent the normalized gradient key point patches, proposed by Ke and Sukthakar in [12]. The authors of this work demonstrated that this method considerably improves both the accuracy and speed compared with the standard SIFT.
In [13], Dalal and Triggs compared the PCA-SIFT, wavelets, and shape context features with the Histogram of Oriented Gradients (HOG) for pedestrian detection and demonstrated that the HOG features greatly outperform the other features on both the MIT and INRIA pedestrian datasets. Owing to [13], HOG has become one of the most widely used features to extract silhouette information for human and other types of object detection. The HOG calculates occurrences (histograms) of gradient orientation in localized portions of an image. Gradient magnitudes serve as weights (votes) to strengthen or weaken the contributions of orientations. This technique differs from the methods mentioned above in that it is computed on a dense grid of uniformly spaced cells and uses overlapping local contrast normalization for improved accuracy. The HOG feature is the concatenation of histograms in overlapping blocks.
To further improve the detection results produced by HOG, some research has focused on enriching the HOG feature sets by combining them with additional cues. The celebrated Aggregated Channel Features (ACF) [14], variants of the Integral Channel Features (ICF) [15], were proposed. The ACF detector begins with the computation of a color channel, such as RGB, LUV or HSV, the magnitude channel, and the HOG channel. These channels are aggregated and vectorized into an enhanced feature vector before being sent to the classifier. As mentioned previously, the ACF detector is characterized by taking both the color feature and the silhouette feature into consideration. Such rich features • A novel pedestrian detector integrating the multitask cascaded CNN and ACF is proposed; • Improved detection performance for significantly occluded pedestrians and beyond is achieved; • Robustness of the proposed detector in terms of datasets and beyond is achieved.
The paper is organized as follows: Section 2 presents the methodology used to construct the integrated detector, including the technical procedures involved and detailed explanations regarding the function and design principles of each module in the proposed detector. In Section 3, we investigate the choices of parameters and test the proposed detector on various datasets. The discussion is presented in Section 4.
The Proposed Detector
Traditionally, the ACF human body detector takes the features of a sliding window b x , b y , b w , b h as the input. The two-dimensional vector (b x , b y ) contains the x and y Cartesian coordinates of the top left corner of the sliding window, which has a width of b w and a height of b h , as denoted in Figure 1. Our integrated detector differs from the ACF human body detector, mainly by introducing a face detector module and score combination, the output of which is then fed into the ACF detector together with the features, as shown in Figure 2.
The paper is organized as follows: Section 2 presents the methodology used to construct the integrated detector, including the technical procedures involved and detailed explanations regarding the function and design principles of each module in the proposed detector. In Section 3, we investigate the choices of parameters and test the proposed detector on various datasets. The discussion is presented in Section 4.
The Proposed Detector
Traditionally, the ACF human body detector takes the features of a sliding window ( , , , ℎ ) as the input. The two-dimensional vector ( , ) contains the x and y Cartesian coordinates of the top left corner of the sliding window, which has a width of and a height of ℎ , as denoted in Figure 1. Our integrated detector differs from the ACF human body detector, mainly by introducing a face detector module and score combination, the output of which is then fed into the ACF detector together with the features, as shown in Figure 2. The proposed integrated detector starts with feeding an image into a face detector which outputs predicted face bounding boxes ( , , , ℎ ) with centers located at The paper is organized as follows: Section 2 presents the methodology used to construct the integrated detector, including the technical procedures involved and detailed explanations regarding the function and design principles of each module in the proposed detector. In Section 3, we investigate the choices of parameters and test the proposed detector on various datasets. The discussion is presented in Section 4.
The Proposed Detector
Traditionally, the ACF human body detector takes the features of a sliding window ( , , , ℎ ) as the input. The two-dimensional vector ( , ) contains the x and y Cartesian coordinates of the top left corner of the sliding window, which has a width of and a height of ℎ , as denoted in Figure 1. Our integrated detector differs from the ACF human body detector, mainly by introducing a face detector module and score combination, the output of which is then fed into the ACF detector together with the features, as shown in Figure 2. The proposed integrated detector starts with feeding an image into a face detector which outputs predicted face bounding boxes ( , , , ℎ ) with centers located at The proposed integrated detector starts with feeding an image into a face detector which outputs N predicted face bounding boxes f n x , f n y , f n w , f n h with centers located at ( f n x + f n w /2, f n y + f n h /2) and corresponding face scores of s n , n = 1, 2, . . . N. The vector ( f n x , f n y ) stands for the x and y coordinates of the top left corner of the n-th face bounding box with a width of f n w and a height of f n h , as shown in Figure 1. Considering that the (1) to keep consistent with the ACF scores as the first step of the score combination. Secondly, face bounding boxes are filtered following two rules. One rule is to eliminate face bounding boxes that are beyond the enlarged sliding window area shown as the dotted box in Figure 1. The height and the width of the enlarged sliding window are increased by 2 × o f f set compared with those of the original sliding window. The other rule is to eliminate relatively large face bounding boxes determined by the sliding window and two predesigned parameters, namely r w and r h . The remaining M face bounding boxes meet the following conditions: In the last step of the score combination, the scores (s m f , m = 1, 2, . . . M) of the remaining face bounding boxes are scaled to s m sc according to where f ss is the face score scale, and d m is the Euclidian distance from the center of the m-th face bounding box to the anchor, which is located at b where d hr is the ratio of the Euclidian distance of the zeroscaled score to b h . Furthermore, f ss , f wr , f hr , and d hr are predesigned parameters, which are illustrated in Figure 1. The maximum scaled score, if it exists, is assigned to the initial overall score s all = max( s m sc ); otherwise, the initial overall score is assigned as 0. The initial overall score is fed into the ACF detector module to produce the final overall score, which is sent to the threshold module. In this module, the final overall score is considered in two cases according to its initial score. The first case is that, if the initial overall score is assigned by the scaled face score, the sliding window with s all > s thr will be the output as a nominal body bounding box. Otherwise, the sliding window with s all > −1 will be the output. Note that −1 is the default threshold of the ACF detector, and s thr is a predesigned parameter that should be larger than −1.
Modules of the Proposed Integrated Detector
In this part, the functionality and design principles of each module shown in Figure 2 are explained in detail.
Sliding Window
This module (Figure 3) outputs the location, size, and ACF features of a cropped area, namely the sliding window, of the image. As explained previously, the location and size are expressed as b x , b y , b w , b h on the original scale. This means that if the sliding window is in a subsampled layer in a feature pyramid, its size and location must be expanded according to the specific subsampling rate. The ACF features consist of a feature vector, which is the concatenation of the LUV color channel, the gradient magnitude channel, and the HOG channel. This module (Figure 3) outputs the location, size, and ACF features of a cropped area, namely the sliding window, of the image. As explained previously, the location and size are expressed as ( , , , ℎ ) on the original scale. This means that if the sliding window is in a subsampled layer in a feature pyramid, its size and location must be expanded according to the specific subsampling rate. The ACF features consist of a feature vector, which is the concatenation of the LUV color channel, the gradient magnitude channel, and the HOG channel.
Face Detector
This module ( Figure 4) takes an image as the input and outputs the detected face bounding boxes jointly with the face scores. Face detection plays a significant role in the overall integrated detection, because the larger the face score, the more likely it is for the corresponding human body to be detected. To correctly identify human bodies, the face detector should have high precision and yield a small number of false positives. We tested the pretrained Viola Jones (VJ) face detector [20], the fast face detector [21], and the MTCNN detector [22] on the INRIA pedestrian test dataset [13]. According to the results shown in the left column of Figure 5, the VJ detector tends to miss faces that are not presented in the frontal view and those that are occluded. False positives appear when the background is relatively complicated or when there are structures that resemble human faces, such as wheels. The fast face detector was designed based on the work of [23]. This new modified version of the detector extends the ACF used in [23] by adding an integral image channel, in which every pixel is the summation of all of the pixels above and to the left of it. As shown in the middle column of Figure 5, some multiview faces are identified at the expense of a dramatically increasing number of false positives. Compared with these two face detectors, the MTCNN detector gives more accurate detection results with lower false positives rates. As shown in the right column of Figure 5, no false positives occur in these four sample images, and only faces that are largely occluded or presented from mainly the back view are missed, which is expected. This level of performance is due to the collaboration of three convolutional neural networks: the Propose-
Face Detector
This module ( Figure 4) takes an image as the input and outputs the detected face bounding boxes jointly with the face scores. Face detection plays a significant role in the overall integrated detection, because the larger the face score, the more likely it is for the corresponding human body to be detected. To correctly identify human bodies, the face detector should have high precision and yield a small number of false positives.
namely the sliding window, of the image. As explained previously, the location and size are expressed as ( , , , ℎ ) on the original scale. This means that if the sliding window is in a subsampled layer in a feature pyramid, its size and location must be expanded according to the specific subsampling rate. The ACF features consist of a feature vector, which is the concatenation of the LUV color channel, the gradient magnitude channel, and the HOG channel.
Face Detector
This module ( Figure 4) takes an image as the input and outputs the detected face bounding boxes jointly with the face scores. Face detection plays a significant role in the overall integrated detection, because the larger the face score, the more likely it is for the corresponding human body to be detected. To correctly identify human bodies, the face detector should have high precision and yield a small number of false positives. We tested the pretrained Viola Jones (VJ) face detector [20], the fast face detector [21], and the MTCNN detector [22] on the INRIA pedestrian test dataset [13]. According to the results shown in the left column of Figure 5, the VJ detector tends to miss faces that are not presented in the frontal view and those that are occluded. False positives appear when the background is relatively complicated or when there are structures that resemble human faces, such as wheels. The fast face detector was designed based on the work of [23]. This new modified version of the detector extends the ACF used in [23] by adding an integral image channel, in which every pixel is the summation of all of the pixels above and to the left of it. As shown in the middle column of Figure 5, some multiview faces are identified at the expense of a dramatically increasing number of false positives. Compared with these two face detectors, the MTCNN detector gives more accurate detection results with lower false positives rates. As shown in the right column of Figure 5, no false positives occur in these four sample images, and only faces that are largely occluded or presented from mainly the back view are missed, which is expected. This level of performance is due to the collaboration of three convolutional neural networks: the Propose- We tested the pretrained Viola Jones (VJ) face detector [20], the fast face detector [21], and the MTCNN detector [22] on the INRIA pedestrian test dataset [13]. According to the results shown in the left column of Figure 5, the VJ detector tends to miss faces that are not presented in the frontal view and those that are occluded. False positives appear when the background is relatively complicated or when there are structures that resemble human faces, such as wheels. The fast face detector was designed based on the work of [23]. This new modified version of the detector extends the ACF used in [23] by adding an integral image channel, in which every pixel is the summation of all of the pixels above and to the left of it. As shown in the middle column of Figure 5, some multiview faces are identified at the expense of a dramatically increasing number of false positives. Compared with these two face detectors, the MTCNN detector gives more accurate detection results with lower false positives rates. As shown in the right column of Figure 5, no false positives occur in these four sample images, and only faces that are largely occluded or presented from mainly the back view are missed, which is expected. This level of performance is due to the collaboration of three convolutional neural networks: the Propose-Network (P-Net), Refine-Network (R-Net), and Output-Network (O-Net). An image is first fed into the fully convolutional neural network P-Net to quickly yield a large number of candidate detections, which are subsequently refined by the R-Net by further correction of the regression vector of the face candidate frame and nonmaximum suppression. The final face regression boxes and facial landmarks (contour key points) are output after correcting and filtering the detections produced earlier with the landmarks, corrections, and probabilities output by the O-Net. The MTCNN detector finally outputs the adjusted face bounding boxes, the facial landmarks, and face scores in the range [0, 1]. Note that more faces are detected by the MTCNN detector without the use of the O-Net, as shown in Figure 6. However, as expected, this structure also results in more false faces, as shown in Figure 7. To achieve the best and most robust face detection, the overall MTCNN detector, therefore, was chosen as our integrated detector.
of the regression vector of the face candidate frame and nonmaximum suppression. The final face regression boxes and facial landmarks (contour key points) are output after correcting and filtering the detections produced earlier with the landmarks, corrections, and probabilities output by the O-Net. The MTCNN detector finally outputs the adjusted face bounding boxes, the facial landmarks, and face scores in the range [0, 1]. Note that more faces are detected by the MTCNN detector without the use of the O-Net, as shown in Figure 6. However, as expected, this structure also results in more false faces, as shown in Figure 7. To achieve the best and most robust face detection, the overall MTCNN detector, therefore, was chosen as our integrated detector. We observed that these are associated with particular structures, for example, car wheels.
Score Combination
The score combination consists of three modules, namely the face score transformation module, the face bounding box filtering module, and the score scaling and assigning module.
Face Score Transformation
This module takes the face scores generated by the MTCNN detector as inputs and outputs the corresponding weights, which are the transformed face scores according to Equation (1). This is to remain consistent with the weights produced by the cascading decision trees in the ACF detector. After the transformation, the face detector can be regarded as a single decision tree. The transformed face scores are used in the score scaling and assigning module.
Face Bounding Box Filtering
This filtering module ( Figure 8) takes ( , , , ℎ ) and ( , , , ℎ ), i.e., the sizes and locations of the face bounding boxes and the sliding windows respectively as inputs and decides which face bounding boxes should be sent to the score scaling and assigning module by checking whether they are potential faces of this sliding window according to the two rules explained below. We observed that these are associated with particular structures, for example, car wheels.
Score Combination
The score combination consists of three modules, namely the face score transformation module, the face bounding box filtering module, and the score scaling and assigning module.
Face Score Transformation
This module takes the face scores generated by the MTCNN detector as inputs and outputs the corresponding weights, which are the transformed face scores according to Equation (1). This is to remain consistent with the weights produced by the cascading decision trees in the ACF detector. After the transformation, the face detector can be regarded as a single decision tree. The transformed face scores are used in the score scaling and assigning module.
Face Bounding Box Filtering
This filtering module ( Figure 8) takes ( , , , ℎ ) and ( , , , ℎ ), i.e., the sizes and locations of the face bounding boxes and the sliding windows respectively as inputs and decides which face bounding boxes should be sent to the score scaling and assigning module by checking whether they are potential faces of this sliding window according to the two rules explained below. We observe that the MTCNN detector produces fewer false faces (a), although some faces with occluded facial features are missed. In contrast, a large number of false faces appear in the complex backgrounds (b) when the O-Net is removed from the MTCNN detector. We observed that these are associated with particular structures, for example, car wheels.
Score Combination
The score combination consists of three modules, namely the face score transformation module, the face bounding box filtering module, and the score scaling and assigning module.
Face Score Transformation
This module takes the face scores generated by the MTCNN detector as inputs and outputs the corresponding weights, which are the transformed face scores according to Equation (1). This is to remain consistent with the weights produced by the cascading decision trees in the ACF detector. After the transformation, the face detector can be regarded as a single decision tree. The transformed face scores are used in the score scaling and assigning module.
Face Bounding Box Filtering
This filtering module ( Figure 8) takes f n x , f n y , f n w , f n h and b x , b y , b w , b h , i.e., the sizes and locations of the face bounding boxes and the sliding windows respectively as inputs and decides which face bounding boxes should be sent to the score scaling and assigning module by checking whether they are potential faces of this sliding window according to the two rules explained below. The first rule is that the true face must be within the potential body bounding box, so the potential faces should be inside the sliding window. There are also face bounding boxes that are only partly inside the sliding box, for example, face bounding box D in Figure 1. To leave some flexibility for such bounding boxes, the four window edges are enlarged by a predesigned , shown as the dotted box in Figure 1. This rule is mathematically expressed as (2)-(5). Figure 9 shows an example from the INRIA pedestrian dataset. The overall scores of the ACF detector (a) and the integrated detector (b) are both 57.6. This is because the face (d) should have contributed to the overall score but is filtered out, as it is located at the edge of the sliding window (b). By setting the as5 pixels, the true positive (c) has a higher score of 62.55, as the face score is successfully included. The second rule is that the face bounding box should not be too large compared with the sliding window. According to the ground truth of the INRIA pedestrian test dataset, only the bounding boxes that contain most parts of a body where the person's height is larger than 100 pixels are labelled as true positives. Therefore, only the largest bounding boxes shown in Figure 10a-c are true positives, while other smaller bounding boxes are false positives, even though some of them do contain human body parts, such as those shown in Figure 10a. This phenomenon is exacerbated by the incorporation of the face detector in the system, as shown in Figure 10b. This is because the new face scores are large enough to alter the detection results of the ACF detector by introducing false positives when the sliding window is too small to contain sufficient features. To eliminate such false positives, the second rule is adopted by setting the minimum width ratio and height ratio ℎ . The width ratio is the ratio of the width of the sliding window to that of the face bounding box, and the height ratio is the ratio of the height of the sliding window to that of the face bounding box. Face bounding boxes with any ratio larger than or ℎ are filtered out, as expressed in (6) The first rule is that the true face must be within the potential body bounding box, so the potential faces should be inside the sliding window. There are also face bounding boxes that are only partly inside the sliding box, for example, face bounding box D in Figure 1. To leave some flexibility for such bounding boxes, the four window edges are enlarged by a predesigned o f f set, shown as the dotted box in Figure 1. This rule is mathematically expressed as (2)-(5). Figure 9 shows an example from the INRIA pedestrian dataset. The overall scores of the ACF detector (a) and the integrated detector (b) are both 57.6. This is because the face (d) should have contributed to the overall score but is filtered out, as it is located at the edge of the sliding window (b). By setting the o f f set as 5 pixels, the true positive (c) has a higher score of 62.55, as the face score is successfully included. The first rule is that the true face must be within the potential body bounding box, so the potential faces should be inside the sliding window. There are also face bounding boxes that are only partly inside the sliding box, for example, face bounding box D in Figure 1. To leave some flexibility for such bounding boxes, the four window edges are enlarged by a predesigned , shown as the dotted box in Figure 1. This rule is mathematically expressed as (2)-(5). Figure 9 shows an example from the INRIA pedestrian dataset. The overall scores of the ACF detector (a) and the integrated detector (b) are both 57.6. This is because the face (d) should have contributed to the overall score but is filtered out, as it is located at the edge of the sliding window (b). By setting the as5 pixels, the true positive (c) has a higher score of 62.55, as the face score is successfully included. The second rule is that the face bounding box should not be too large compared with the sliding window. According to the ground truth of the INRIA pedestrian test dataset, only the bounding boxes that contain most parts of a body where the person's height is larger than 100 pixels are labelled as true positives. Therefore, only the largest bounding boxes shown in Figure 10a-c are true positives, while other smaller bounding boxes are false positives, even though some of them do contain human body parts, such as those shown in Figure 10a. This phenomenon is exacerbated by the incorporation of the face detector in the system, as shown in Figure 10b. This is because the new face scores are large enough to alter the detection results of the ACF detector by introducing false positives when the sliding window is too small to contain sufficient features. To eliminate such false positives, the second rule is adopted by setting the minimum width ratio and height ratio ℎ . The width ratio is the ratio of the width of the sliding window to that of the face bounding box, and the height ratio is the ratio of the height of the sliding window to that of the face bounding box. Face bounding boxes with any ratio larger than or ℎ are filtered out, as expressed in (6) The second rule is that the face bounding box should not be too large compared with the sliding window. According to the ground truth of the INRIA pedestrian test dataset, only the bounding boxes that contain most parts of a body where the person's height is larger than 100 pixels are labelled as true positives. Therefore, only the largest bounding boxes shown in Figure 10a-c are true positives, while other smaller bounding boxes are false positives, even though some of them do contain human body parts, such as those shown in Figure 10a. This phenomenon is exacerbated by the incorporation of the face detector in the system, as shown in Figure 10b. This is because the new face scores are large enough to alter the detection results of the ACF detector by introducing false positives when the sliding window is too small to contain sufficient features. To eliminate such false positives, the second rule is adopted by setting the minimum width ratio r w and height ratio r h . The width ratio is the ratio of the width of the sliding window to that of the face bounding box, and the height ratio is the ratio of the height of the sliding window to that of the face bounding box. Face bounding boxes with any ratio larger than r w or r h are filtered out, as expressed in (6)
Score Scaling and Assigning
This last module (Figure 11) of the score combination takes the filtered face scores and the locations and sizes ( , , , ℎ )asinputs. They are used to compute the scaled score, which is assigned to the initial overall score and fed into the subsequent ACF detection module. Figure 11. Diagram of the score projection and assigning module. Note that the filtered face information is fed into this module. The output initial overall score is assigned to the corresponding input sliding window.
The basic rule of score scaling is that the nearer the face bounding box is to the anchor of a sliding window, the higher the initial overall score is. The anchor is located at the most likely position that a face of an upright human body will appear at. According to this rule, the scaled score is computed by (8), as illustrated in Figure 12 below. The highest scaled score appears at = 0, which means that the face bounding box is located exactly at the anchor, and the sliding window is temporarily considered to be the most likely to contain a human body. The scaled score decreases as the face bounding box moves away from the anchor and reaches 0 when = 0 . The parameter 0 is the zero-scaled score distance, as marked in Figure 1. 0 should be linearly related to the size of the sliding window to adapt to the changes in the size of the area in which the faces are likely to appear, as introduced by the varying sliding window sizes, so we set 0 = ℎ ⋅ ℎ . The scaled score becomes negative when > 0 and reaches its minimum value at the maximum value of before the face bounding box is filtered out. Such negative scores can help to eliminate false human body bounding boxes. According to (8), the scaled scores of the five face bounding boxes shown in Figure 1 are ranked as > > = 0 > > . They are used to compute the scaled score, which is assigned to the initial overall score and fed into the subsequent ACF detection module.
Score Scaling and Assigning
This last module (Figure 11) of the score combination takes the filtered face scores and the locations and sizes ( , , , ℎ )asinputs. They are used to compute the scaled score, which is assigned to the initial overall score and fed into the subsequent ACF detection module. Figure 11. Diagram of the score projection and assigning module. Note that the filtered face information is fed into this module. The output initial overall score is assigned to the corresponding input sliding window.
The basic rule of score scaling is that the nearer the face bounding box is to the anchor of a sliding window, the higher the initial overall score is. The anchor is located at the most likely position that a face of an upright human body will appear at. According to this rule, the scaled score is computed by (8), as illustrated in Figure 12 below. The highest scaled score appears at = 0, which means that the face bounding box is located exactly at the anchor, and the sliding window is temporarily considered to be the most likely to contain a human body. The scaled score decreases as the face bounding box moves away from the anchor and reaches 0 when = 0 . The parameter 0 is the zero-scaled score distance, as marked in Figure 1. 0 should be linearly related to the size of the sliding window to adapt to the changes in the size of the area in which the faces are likely to appear, as introduced by the varying sliding window sizes, so we set 0 = ℎ ⋅ ℎ . The scaled score becomes negative when > 0 and reaches its minimum value at the maximum value of before the face bounding box is filtered out. Such negative scores can help to eliminate false human body bounding boxes. According to (8), the scaled scores of the five face bounding boxes shown in Figure 1 are ranked as > > = 0 > > . Figure 11. Diagram of the score projection and assigning module. Note that the filtered face information is fed into this module. The output initial overall score is assigned to the corresponding input sliding window.
The basic rule of score scaling is that the nearer the face bounding box is to the anchor of a sliding window, the higher the initial overall score is. The anchor is located at the most likely position that a face of an upright human body will appear at. According to this rule, the scaled score is computed by (8), as illustrated in Figure 12 below. The highest scaled score appears at d m = 0, which means that the face bounding box is located exactly at the anchor, and the sliding window is temporarily considered to be the most likely to contain a human body. The scaled score decreases as the face bounding box moves away from the anchor and reaches 0 when d m = d 0 . The parameter d 0 is the zero-scaled score distance, as marked in Figure 1. d 0 should be linearly related to the size of the sliding window to adapt to the changes in the size of the area in which the faces are likely to appear, as introduced by the varying sliding window sizes, so we set d 0 = d hr · b h . The scaled score becomes negative when d m > d 0 and reaches its minimum value at the maximum value of d m before the face bounding box is filtered out. Such negative scores can help to eliminate false human body bounding boxes. According to (8), the scaled scores of the five face bounding boxes shown in Figure 1 are ranked as A > B > C = 0 > E > D.
Sensors 2022, 22, x FOR PEER REVIEW 11 of 23 Figure 12. The scaled score divided by versus the distance from the -th face center to the anchor is drawn according to (8). The higher the scaled score is, the more likely the sliding window is to contain a human body.
To increase the weight of the MTCNN detector, the predesigned face score scale is introduced in (8), so that the scaled score of the MTCNN detector is equivalent to the sum score of decision trees. The larger is, the greater the face detector's influence on the result of the integrated human body detector is. In Figure 13, an example is illustrated, where the human body is missed by the ACF detector (left) but is detected by the integrated detector with = 1 (right). In another example illustrated in the top sequence of Figure 14, multiple human bodies are missed, although their faces are correctly detected. This is because the face score is overly small compared with the summed score of up to 2048 decision trees in the ACF body detector. As a result, even with the face score included, the overall score of a bounding box associated with a missed body is still too small to reach the threshold. However, we observe in Figure 15 that when is increased to 5, 8, or 10, the previously missed human body is now detected. This means that as the weight of the MTCNN detector increases, the accuracy of a body bounding box increases. (a) (b) (c) Figure 12. The scaled score divided by f ss versus the distance from the m -th face center to the anchor is drawn according to (8). The higher the scaled score is, the more likely the sliding window is to contain a human body.
To increase the weight of the MTCNN detector, the predesigned face score scale f ss is introduced in (8), so that the scaled score s m sc of the MTCNN detector is equivalent to the sum score of f ss decision trees. The larger f ss is, the greater the face detector's influence on the result of the integrated human body detector is. In Figure 13, an example is illustrated, where the human body is missed by the ACF detector (left) but is detected by the integrated detector with f ss = 1 (right). In another example illustrated in the top sequence of Figure 14, multiple human bodies are missed, although their faces are correctly detected. This is because the face score is overly small compared with the summed score of up to 2048 decision trees in the ACF body detector. As a result, even with the face score included, the overall score of a bounding box associated with a missed body is still too small to reach the threshold. However, we observe in Figure 15 that when f ss is increased to 5, 8, or 10, the previously missed human body is now detected. This means that as the weight of the MTCNN detector increases, the accuracy of a body bounding box increases. Figure 12. The scaled score divided by versus the distance from the -th face center to the anchor is drawn according to (8). The higher the scaled score is, the more likely the sliding window is to contain a human body.
To increase the weight of the MTCNN detector, the predesigned face score scale is introduced in (8), so that the scaled score of the MTCNN detector is equivalent to the sum score of decision trees. The larger is, the greater the face detector's influence on the result of the integrated human body detector is. In Figure 13, an example is illustrated, where the human body is missed by the ACF detector (left) but is detected by the integrated detector with = 1 (right). In another example illustrated in the top sequence of Figure 14, multiple human bodies are missed, although their faces are correctly detected. This is because the face score is overly small compared with the summed score of up to 2048 decision trees in the ACF body detector. As a result, even with the face score included, the overall score of a bounding box associated with a missed body is still too small to reach the threshold. However, we observe in Figure 15 that when is increased to 5, 8, or 10, the previously missed human body is now detected. This means that as the weight of the MTCNN detector increases, the accuracy of a body bounding box increases. After score scaling, the maximum scaled score max( ) is assigned to the overall score of the sliding window for initialization. Sliding windows with positive initial scores are more likely to survive the ACF detector, while those with negative initial scores are more likely to be eliminated.
ACF Detector
The ACF detector ( Figure 16) that is employed to detect the human body in the proposed framework is mainly based on the pretrained cascading decision trees [24]. It takes the aggregated color channel, the gradient magnitude channel, and the HOG channel of the sliding window as the input feature vector and outputs the final overall score and the location of the sliding window as the nominal human body bounding box. Note that the final overall score is obtained through the addition of the scaled face score, considered the initial overall score, and the body score. One modification is that the initial overall score of the ACF detector is set as max( ), as mentioned in the score scaling and assigning module. After score scaling, the maximum scaled score max( ) is assigned to the overall score of the sliding window for initialization. Sliding windows with positive initial scores are more likely to survive the ACF detector, while those with negative initial scores are more likely to be eliminated.
ACF Detector
The ACF detector ( Figure 16) that is employed to detect the human body in the proposed framework is mainly based on the pretrained cascading decision trees [24]. It takes the aggregated color channel, the gradient magnitude channel, and the HOG channel of the sliding window as the input feature vector and outputs the final overall score and the location of the sliding window as the nominal human body bounding box. Note that the final overall score is obtained through the addition of the scaled face score, considered the initial overall score, and the body score. One modification is that the initial overall score of the ACF detector is set as max( ), as mentioned in the score scaling and assigning module. After score scaling, the maximum scaled score max(s m sc ) is assigned to the overall score of the sliding window for initialization. Sliding windows with positive initial scores are more likely to survive the ACF detector, while those with negative initial scores are more likely to be eliminated.
ACF Detector
The ACF detector ( Figure 16) that is employed to detect the human body in the proposed framework is mainly based on the pretrained cascading decision trees [24]. It takes the aggregated color channel, the gradient magnitude channel, and the HOG channel of the sliding window as the input feature vector and outputs the final overall score and the location of the sliding window as the nominal human body bounding box. Note that the final overall score is obtained through the addition of the scaled face score, considered the initial overall score, and the body score. One modification is that the initial overall score of the ACF detector is set as max(s m sc ), as mentioned in the score scaling and assigning module.
the aggregated color channel, the gradient magnitude channel, and the HOG channel of the sliding window as the input feature vector and outputs the final overall score and the location of the sliding window as the nominal human body bounding box. Note that the final overall score is obtained through the addition of the scaled face score, considered the initial overall score, and the body score. One modification is that the initial overall score of the ACF detector is set as max( ), as mentioned in the score scaling and assigning module. Figure 17. The human bodies in these samples have already been correctly marked by the bounding boxes with high body scores, as shown in the bottom set of images in Figure 17, and the bounding boxes may still have relatively high body scores if they are misplaced by only a short distance. After including the scaled face scores, these misplaced boxes are easily identified as false positives. This also explains why each false positive shown in Figure 17 contains a human face. Though the final overall scores of such false positives are higher than the default threshold, they are much lower than those of the true positives, as shown in Figure 17. Considering this phenomenon, a face threshold s thr , higher than the default threshold and lower than the final overall scores of the true positives, is set to filter out false positives containing faces. This face threshold must be higher than the default threshold, because the final overall scores are increased by the scaled face scores, whereas the default threshold is designed without considering face scores. The face threshold s thr helps with the elimination of false positives introduced by the incorporation of face scores and the implementation of f ss . Note that the default threshold is used if the sliding window does not contain any faces. Figure 17. The human bodies in these samples have already been cor marked by the bounding boxes with high body scores, as shown in the bottom set ages in Figure 17, and the bounding boxes may still have relatively high body sco they are misplaced by only a short distance. After including the scaled face scores, misplaced boxes are easily identified as false positives. This also explains why each positive shown in Figure 17 contains a human face. Though the final overall scores o false positives are higher than the default threshold, they are much lower than th the true positives, as shown in Figure 17. Considering this phenomenon, a face thre ℎ , higher than the default threshold and lower than the final overall scores of th positives, is set to filter out false positives containing faces. This face threshold mu higher than the default threshold, because the final overall scores are increased b scaled face scores, whereas the default threshold is designed without considering scores. The face threshold ℎ helps with the elimination of false positives introduc the incorporation of face scores and the implementation of . Note that the d threshold is used if the sliding window does not contain any faces. The procedure of the proposed integrated detector is summarized in Algorithm 1.
Output
Qualified Step 1 For each n: If Equations (2)-(7) are fulfilled: Send s all to the cascading decision trees.
Step 3 If s all = max{s m sc } is implemented: If s all > s thr :
Experiments Analysis and Results
In this section, we first investigate the influences of eight predesigned parameters on the performance of the proposed detector to obtain the optimal parameters. Afterwards, the tuned detector is compared with state-of-the-art methods, and its robustness in is evaluated on various datasets.
In the following experiments, the INRIA pedestrian dataset, Caltech pedestrian dataset, Citypersons dataset, and the ETHZ dataset were used.
•
INRIA pedestrian dataset [13]: The test dataset contains 288 positive color images with 589 labeled human bodies. These images were shot at around eye-level. Most of these human bodies have an upright orientation with some extent of occlusion. • Caltech pedestrian dataset [25]: This dataset consists of approximately 250,000 frames, 640 × 480 in size, and a total of 350,000 annotated bounding boxes. The standard test set with 4024 images and corresponding new annotations [26] were used in subsequent experiments. Each image contains about 1.4 persons.
•
Citypersons dataset [27]: The validation set contains 500 high-resolution images, 1024 × 2048 in size, and a total of 3938 persons. Each validation image contains about 7.9 persons. • ETHZ dataset [28]: This dataset is a collection of 8 video sequences from busy inner-city locations with annotated human bodies. We assessed two representative sequences from this dataset, namely the BAHNHOF sequence and the Sunny Day sequence. As the pretrained ACF detector cannot classify human bodies with very small sizes, ground truths with widths and heights smaller than 32 and 80 pixels, respectively, were filtered out from the image sequences. After this, the BAHNHOF sequence had 999 images with 3341 ground truths and the Sunny Day sequence had 354 images with 1560 ground truths.
For Caltech and Citypersons, pedestrians were allocated to the reasonable subset, heavily occluded subset, and all subset. The reasonable subset is a collection of pedestrians with heights greater than 50 pixels and visibility levels greater than 0.65. For the heavily occluded subset, the visibility lies in the range [0.2, 0.65]. The all subset consists of pedestrians with heights greater than 20 pixels and a visibility level greater than 0.2.
The MTCNN detector utilized in the experiments is based on the convolutional neural network, which not only detects human faces but also locates facial landmarks. It is available online and is well-trained, and therefore, it was directly applied to our detector. Note that, facial landmark locations were discarded, as only the face bounding boxes and probabilities were used in our detector.
The ACF detector is available in the Piotr's MATLAB toolbox version 3.40. It is an Adaboost classifier based on cascading binary decision trees. The ACF detector was pretrained on both the INRIA and Caltech pedestrian datasets, respectively.
Parameter Design
The integrated detector has eight predesigned parameters, namely o f f set, r w , r h in the face bounding boxes filtering module, f wr , f hr , d hr , f ss in the score scaling and assigning module, and s thr in the ACF detector. To fully exploit the power of the integrated detector, we investigated the influences of these parameters on the detection results via control variates in the following experiments. The MTCNN face detector available at [22] and the pretrained ACF detector provided by [24] were utilized, and the pre-designed parameters were initially set as o f f set = 0, r w = 3, r h = 7, f wr = 1/2, f hr = 1/8, d hr = 1/4, f ss = 8, s thr = 25. The integrated detector was tested on the INRIA pedestrian test dataset, which has 589 ground truths. The results were evaluated quantitively by calculating the recall and the log-average miss rate. The recall was calculated as the number of true positives divided by the number of groundtruths. The log-average miss rate refers to the average miss rate over the False Positives Per Image (FPPI) in the range 10 −2 , 10 0 , which can be calculated automatically using Piotr's MATLAB toolbox. The miss rate is defined as (1 − Recall).
Tables 1-6 list the detection results from the integrated detector for various choices of parameters. Their corresponding Receiver Operating Characteristic (ROC) curves are shown in Figure 18. Considering that s thr and f ss are correlated parameters because the face threshold should adapt to the final overall score, their 3D histograms are drawn in Figure 19, instead of using 2D curves and tables. Note that some abbreviations are used in these figures and tables, namely TP (the number of True Positives), FP (the number of False Positives), R (Recall), and AMR (the Average Miss Rate). According to Table 1 and Figure 18a, the same recall and 3 more false positives are produced when the increases from 0 to 6. This shows that the contribution of the is not obvious in our setup. A value of = 0 is recommended when one wishes to minimize false positives. Tables 2 and 3 and Figure 18b,c show that = 1-4 and ℎ = 7, 8 can produce the maximum number of true positives. Increasing and ℎ leads to less false positives, because relatively small sliding windows are filtered out. However, some true positives are also eliminated when and ℎ are too large. To maintain as many true positives as possible, = 1 and ℎ = 7 are recommended. As for the score scaling and assigning, the anchor best locates at the middle of the width ( = 0.5) and 1/8th of the height ( ℎ = 1/8), as presented in Tables 4 and 5 and Figure 18d,e. This location is in line with the face positions of the most upright adult human bodies. Table 6 and Figure 18f show that the best ℎ is 1/4. As shown in Figure 19a-c, ℎ is inversely proportional to the number of true positives, false positives, and recall, whereas is directly proportional to them. This is because a higher threshold brings in fewer bounding According to Table 1 and Figure 18a Table 6 and Figure 18f show that the best d hr is 1/4. As shown in Figure 19a-c, s thr is inversely proportional to the number of true positives, false positives, and recall, whereas f ss is directly proportional to them. This is because a higher threshold brings in fewer bounding boxes, but more are obtained when the final overall scores are augmented by f ss . To strike a balance between these two parameters, parameters of s thr = 25, f ss = 8 are suggested, which produces the smallest average miss rate, as shown in Figure 18d.
According to these experimental results, the predesigned parameters can be allocated into four types according to their functions. First, r w , r h , and s thr can be increased to eliminate false positives but with the sacrifice of some true positives. In contrast, f ss can be increased to bring in both additional true and false positives. Thirdly, choosing appropriate f wr , f hr , and d hr values can increase the true positives and decrease the false positives at the same time. Finally, the o f f set has a negligible influence on the detection results.
Evaluation
Following an extensive experimental validation study, we can claim that, for the dataset considered, the integrated detector is finely tuned and produces the best detection results when we choose o f f set = 0, r w = 3, r h = 7, f wr = 0.5, f hr = 0.125, d hr = 0.25, f ss = 8, s thr = 25. Furthermore, for the dataset considered, as shown in Figure 20, the integrated detector produces a better performance than the traditional ACF detector with the average miss rate decreasing from 16.85% to 14.29%. By fusing the MTCNN face detector with the body detector, the number of true positives increases from 543 to 549, while the number of false positives decreases from 328 to 319, as shown in Table 7. Some image samples that depict the increased true positives and the eliminated false positives are shown in Figures 21 and 22, respectively. 8, ℎ = 25. Furthermore, for the dataset considered, as shown in Figure 20, the integrated detector produces a better performance than the traditional ACF detector with the average miss rate decreasing from 16.85% to 14.29%. By fusing the MTCNN face detector with the body detector, the number of true positives increases from 543 to 549, while the number of false positives decreases from 328 to 319, as shown in Table 7. Some image samples that depict the increased true positives and the eliminated false positives are shown in Figures 21 and 22, respectively. , ℎ , and ℎ values can increase the true positives and decrease the false positives at the same time. Finally, the has a negligible influence on the detection results.
Evaluation
Following an extensive experimental validation study, we can claim that, for the dataset considered, the integrated detector is finely tuned and produces the best detection results when we choose = 0, = 3, ℎ = 7, = 0.5, ℎ = 0.125, ℎ = 0.25, = 8, ℎ = 25. Furthermore, for the dataset considered, as shown in Figure 20, the integrated detector produces a better performance than the traditional ACF detector with the average miss rate decreasing from 16.85% to 14.29%. By fusing the MTCNN face detector with the body detector, the number of true positives increases from 543 to 549, while the number of false positives decreases from 328 to 319, as shown in Table 7. Some image samples that depict the increased true positives and the eliminated false positives are shown in Figures 21 and 22, respectively.
Comparison with the State-of-the-Art Detectors
We also compared the proposed detector with state-of-the-art detectors, including the handcrafted feature based detectors HOG + SVM [13], DPM [16], ACF [14], and ACF + HSC [1]; learning based detectors, such as ConvNet [29]; and the deep models YOLOv3 [18], FRCNN [19], FRCNN + BN [19], SAR R-CNN [30], and RPN-BF [31]. The AMRs shown in Table 8 are cited from [1], except for the proposed detector, ACF, and YOLOv3. As shown in the table, the proposed detector produced the lowest AMR of the listed nondeep model-based detectors. It even outperformed YOLOv3 and achieved a performance comparable to that of FRCNN for the INRIA test dataset. Other deep-modelbased detectors produced much lower AMRs by taking advantage of extracting a large number of high-level features and training many epochs on high-end GPUs. We also compared the proposed detector with state-of-the-art detectors, including the handcrafted feature based detectors HOG + SVM [13], DPM [16], ACF [14], and ACF + HSC [1]; learning based detectors, such as ConvNet [29]; and the deep models YOLOv3 [18], FRCNN [19], FRCNN + BN [19], SAR R-CNN [30], and RPN-BF [31]. The AMRs shown in Table 8 are cited from [1], except for the proposed detector, ACF, and YOLOv3. As shown in the table, the proposed detector produced the lowest AMR of the listed nondeep model-based detectors. It even outperformed YOLOv3 and achieved a performance comparable to that of FRCNN for the INRIA test dataset. Other deep-model-based detectors produced much lower AMRs by taking advantage of extracting a large number of high-level features and training many epochs on high-end GPUs.
Evaluation of Robustness
To evaluate the robustness of the integrated detector, it was tested on the ETHZ dataset, Caltech test set, and Citypersons validation set in the following experiments. The parameters of the integrated detector were still o f f set = 0, r w = 3, r h = 7, f wr = 0.5, f hr = 0.125, d hr = 0.25, and f ss = 8, s thr = 25. The embedded ACF detector was pretrained on the INRIA dataset or the Caltech dataset, if specified.
The results for the two sequences of the ETHZ dataset (Tables 9 and 10) show that the proposed integrated detector produced more true positives and fewer false positives, leading to an increased recall and decreased AMR positives. For the Caltech test set, a decrease in AMRs (Table 11) was also observed. A cross dataset evaluation on the Citypersons dataset was performed. The detectors were first pretrained on the INRIA and Caltech datasets and were then tested on Citypersons validation set. It was observed that the integrated detector significantly decreased the AMR (Figures 23a-c and 24a-c) and increased the recall (Figures 23d-f and 24d-f) for the reasonable, heavily occluded, and all subsets with both pretrained datasets. This, together with the information shown in Tables 9 and 10, indicates that the proposed integrated detector as well as the parameter design are robust when used with the training set and are applicable for use on an unseen dataset. the proposed integrated detector produced more true positives and fewer false positives, leading to an increased recall and decreased AMR positives. For the Caltech test set, a decrease in AMRs (Table 11) was also observed. A cross dataset evaluation on the Citypersons dataset was performed. The detectors were first pretrained on the INRIA and Caltech datasets and were then tested on Citypersons validation set. It was observed that the integrated detector significantly decreased the AMR (Figures 23a-c and 24a-c) and increased the recall (Figures 23d-f and 24d-f) for the reasonable, heavily occluded, and all subsets with both pretrained datasets. This, together with the information shown in Tables 9 and 10, indicates that the proposed integrated detector as well as the parameter design are robust when used with the training set and are applicable for use on an unseen dataset. We also investigated the robustness of the proposed detector under different thresholds. The results (Table 11, Figures 23 and 24) show that the integrated detector improved the performance under varying thresholds, and the improvement was more significant under higher thresholds. This means that the proposed method is not only robust to thresholds but performs better when fewer false positives are required.
The influence of the minimum size (denoted as ms) of faces on the integrated detector was studied. The smaller the parameter ms, the more faces MTCNN was able to detect. According to Table 11 and Figures 23 and 24, more noticeable improvements were observed in most cases, except for the reasonable and all subsets pretrained on INRIA when ms was decreased to 13 pixels. This means that the performance of the face detector influences the improvements of the whole integrated detector. We also investigated the robustness of the proposed detector under different thresholds. The results (Table 11, Figures 23 and 24) show that the integrated detector improved the performance under varying thresholds, and the improvement was more significant under higher thresholds. This means that the proposed method is not only robust to thresholds but performs better when fewer false positives are required.
The influence of the minimum size (denoted as ms) of faces on the integrated detector was studied. The smaller the parameter ms, the more faces MTCNN was able to detect. According to Table 11 and Figures 23 and 24, more noticeable improvements were observed in most cases, except for the reasonable and all subsets pretrained on INRIA when ms was decreased to 13 pixels. This means that the performance of the face detector influences the improvements of the whole integrated detector.
Detection Speed
With regard to the detection speed, the time costs of detecting 288 INRIA pedestrian test images for different detectors via Intel(R) Core (TM) i7-8565U CPU @ 1.80GHz are compared in Table Figure 25. To increase the true positive rate while maintaining the least false positive rate, the complete MTCNN was utilized in our integrated detector. shown in Figure 25. To increase the true positive rate while maintaining the least false positive rate, the complete MTCNN was utilized in our integrated detector.
Conclusions
In conclusion, we presented an integrated pedestrian detector, namely the MTCNN+ACF detector, for small pedestrian datasets. It detects human bodies considering not only color and edge information but also facial features. The integrated detector aggregates multiple detection tasks uniformly to produce a final overall score, which is the sum of the scaled face score and the body score. The fusion rules and parameter choices were investigated in depth. The idea is simple and easy to implement, but the proposed detector can effectively and robustly improve the detection performance compared to the sole use of ACF detector on various pedestrian datasets. The proposed detector only utilizes the CPU device and does not require any further training; however, it achieves a performance level (14.29%) comparable to deep models such as FRCNN (14%) and YOLOv3 (14.75%) on the small pedestrian dataset. The recall and average miss rate were observed to have a steady increase and decrease, respectively, on the Citypersons, ETHZ, and Caltech datasets. Therefore, the proposed detector is an effective paradigm of multitask collaboration, and it serves as a cost-effective choice for pedestrian detection in the case of limited data and computational resources.
Conclusions
In conclusion, we presented an integrated pedestrian detector, namely the MTCNN + ACF detector, for small pedestrian datasets. It detects human bodies considering not only color and edge information but also facial features. The integrated detector aggregates multiple detection tasks uniformly to produce a final overall score, which is the sum of the scaled face score and the body score. The fusion rules and parameter choices were investigated in depth. The idea is simple and easy to implement, but the proposed detector can effectively and robustly improve the detection performance compared to the sole use of ACF detector on various pedestrian datasets. The proposed detector only utilizes the CPU device and does not require any further training; however, it achieves a performance level (14.29%) comparable to deep models such as FRCNN (14%) and YOLOv3 (14.75%) on the small pedestrian dataset. The recall and average miss rate were observed to have a steady increase and decrease, respectively, on the Citypersons, ETHZ, and Caltech datasets. Therefore, the proposed detector is an effective paradigm of multitask collaboration, and it serves as a cost-effective choice for pedestrian detection in the case of limited data and computational resources. | 2022-05-11T15:08:05.154Z | 2022-05-01T00:00:00.000 | {
"year": 2022,
"sha1": "ab567210d8d6347eb05162be1772d92cf26b8e99",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1424-8220/22/9/3568/pdf?version=1651919902",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "7bdd6957f2db2b6c773a6d0be382399458bc8986",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science",
"Medicine"
]
} |
218973884 | pes2o/s2orc | v3-fos-license | Interannotator Agreement for Lexico-Semantic Annotation of a Corpus
This paper examines the procedure for lexico-semantic annotation of the Basic Corpus of Polish Metaphors that is the first step for annotating metaphoric expressions occurring in it. The procedure involves correcting the morphosyntactic annotation of part of the corpus that is automatically annotated on the morphosyntactic level. The main procedure concerns annotation of adjectives, adverbs, nouns and verbs (including gerunds and participles), including abbreviations of the words that belong to the above classes. It is composed of three steps: deciding whether a particular occurrence of a word is asemantic (e.g. anaphoric or strictly grammatical), whether we are dealing with a multi-word expression, reciprocal usages of the się marker and pluralia tantum, which may involve annotation with two lexical units (having two different lemmas) for a single token. We propose an interannotator agreement statistics adequate for this procedure. Finally, we discuss the preliminary results of annotation of a fragment of the corpus.
Introduction
In this paper we want to describe the procedure for lexicosemantic annotation of the Basic Corpus of Polish Metaphor (BCPM), which is a first step for annotation of metaphoric expressions. This task is part of the Cognitive and sociocultural analysis of metaphoric expressions in Polish texts project, aimed at, among other things, automatic detection of metaphoric expressions in Polish texts.
Annotated corpora form a basis for natural language processing. Several corpora annotated on various levels of linguistic information exist: morphosyntactic, syntactic (shallow or deep), multi-word expression (MWEs), and finally semantic, including lexico-semantic, and for various languages.
The usual procedure is to manually annotate a small corpus, use it to train NLP tools and annotate the whole corpus (a substantially larger one) by means of those tools. The quality of the manual annotation is crucial here. To ensure this happens, at least two linguists usually annotate each text sample in the corpus and the conflicts are resolved by a superannotator. Furthermore, a so-called interannotator agreement is calculated to show the difficulty of the task and the quality of its performance.
Several statistics are used to calculate the interannotator agreement. The simplest calculate the percentage of identical annotation. The most popular coefficients are Cohen's (Cohen, 1960) κ, Scott's (Scott, 1955) π and Bennett's S (Bennett et al., 1954) which take into account the possibility of chance agreement. All the statistics satisfy the equation (1), but they differ in calculating P (E). S is the simplest in this respect, as it considers only the cardinality of the set of classes, whereas the two other coefficients take into account the distribution of decisions of annotators in the cor-pusP a1 (k),P a1 (k), where a 1 , a 2 are annotators.
where t ∈ T is a particular token and k, t are cardinalities of the sets K, T , respectively. For S coefficient we have which can be alternatively calculated as: where p e t = 1 k .
κ considers annotators' choices independently, i.e. P κ ai (k) = n i k t , where n i k indicates how often an annotator i chooses category k. Contrarily, π averages these value, i.e. P π a1 (k) =P π a2 (k) =P π (k) = n 1 k +n 2 k 2t = n k 2t . When each sample is annotated by two annotators, but there are several of them involved in the whole procedure,P κ a1 (k) and P κ a2 (k) values depend on the way particular annotators are assigned as the first or the second.P π ai (k) is not sensitive for such partitions, hence it is more suitable in such cases.
In contrast to Bennett's S, reformulating P π (E) in a way calculating it for each token separately is not straightforward. However, putting we obtain The formulae are calculated with the assumption that the set of classes K is the same for all tokens. However, this is not the case in practical applications, henceP a (k) should be calculated w.r.t. tokens for which a category k could be chosen. Unfortunately, in the case of lexico-semantic annotations, the set of categories (LUs) is different for each lexeme. Since both distributions of lexemes and their senses are Zipfian, most LUs are chosen by an annotator once of twice. Therefore, Bennett's S seems to be the only reliable coefficient, with p e t = 1 kt , where k t is the cardinality of the set of classes K t appropriate for a token t. All the statistics are based on the assumption that annotators choose the value from predefined lists (potentially different for different tokens). In this paper we want to show, analysing a particular procedure applied for lexico-semantic annotation of BCPM, that in practice such a single choice may consist of a chain of interdependent decisions, and annotators can agree or disagree at every step. Thus, we have to value each such decision separately and then combine the result 1 .
In what follows, we present other lexico-semantically annotated corpora (cf. section 2.). Section 3. includes the main information about the corpus being lexico-semantically annotated and about the Polish wordnet used in this annotation. The entire procedure for annotation is discussed in section 4., whereas a method of adapting standard interannotator agreement statistics to this particular procedure is proposed in section 5.. Finally, in section 6. we discuss the preliminary interannotator agreement results of two phases of annotation calculated for a small fragment of the corpus that is already annotated by two linguists.
Related works
The most famous semantically annotated corpus is Sem-Cor (Miller et al., 1993), a subcorpus of the Brown Corpus (Francis andKucera, 1964 revised andamplified 1979) containing 250 000 words semantically annotated by means of Princeton WordNet 2 (PWN) (Miller et al., 1990;Fellbaum, 1998;Miller and Fellbaum, 2007) synset identifiers. Annotation was performed by means of a dedicated interface called ConText (Leacock, 1993). The corpus was preprocessed in order to find proper names and collocations (the ones present in PWN). The collocations were joined into single units by concatenating them with underscores (e.g., took_place). ConText performs a corpus word by word (only open-class words). Annotators choose an appropriate sense from a list. They also have the possibility to add comments, when no available sense is considered appropriate. A 1.7 mln. subcorpus of the British National Corpus was semantically annotated manually as a part of the Hector lexicographic project (Atkins, 1991). All occurrences of 300 word types that have between 300 and 1000 occurrences in this subcorpus were tagged, resulting in 220 000 tagged tokens. As for Slavic languages, most words in the balanced subcorpus of the Russian National Corpus (RNC) (Grishina and Rakhilina, 2005) were semantically annotated. The semantic annotation (Apresjan et al., 2006;Lashevskaja, 2006;Kustova et al., 2007) is based on a hierarchical taxonomic classification of a Russian lexicon Lexicograph 3 (Filipenko et al., 1992). The texts were semantically tagged with the Semmarkup program (created by A. Polyakov). For Polish, lexico-semantic annotation was performed for the sake of experiments in word sense disambiguation (WSD), and was limited to small sets of highly polysemous words, e.g. Broda and Piasecki (2011) annotated 13 nouns with the number of P W senses varying from 3 to 15 (only 72 of the total number of differentiated senses were encountered in the resulting corpus). All occurrences of 106 selected lexemes (50 nouns, 48 verbs and 8 adjectives) in NKJP 1M were annotated with coarse senses, cf. ch. 7 of (Przepiórkowski et al., 2012). P W -based 4 annotation of all open-class tokens was performed for the Składnica part of the NKJP 1M corpus (Hajnicz, 2014b;Hajnicz, 2014a). Unfortunately, only parsed sentences were considered.
Resources
BCPM is composed of two parts: • 700 samples of the Polish Coreference Corpus (PCC), randomly selected in a way that balances various registers of texts accordingly to NKJP assumptions, cf. ch. 3 of (Przepiórkowski et al., 2012), • 2000 samples of a fragment of NKJP 1M considered in the Składnica treebank, selected in way that maximises its size and the number of sentences that have parses in Składnica, but preserving the balance of registers.
NKJP 1M is a subcorpus of the Polish National Corpus (Polish acronym NKJP) manually annotated on the morphosyntactic level, cf. ch. 6 of (Przepiórkowski et al., 2012). The PCC (Ogrodniczuk et al., 2015), in turn, is randomly selected from the whole NKJP corpus. Therefore, BCPM as a whole is part of NKJP. The size of the whole BCPM corpus is 344,118 tokens. It is worth noting that the PCC annotation has not been manually corrected on the morphosyntactic level.
In contrast to NKJP, we decided to lexico-semantically annotate tokens with very fine-grained semantic types represented by wordnet lexical units. We use P W (Piasecki et al., 2009), in particular its 4.0 version (Dziob and Piasecki, 2018;Piasecki et al., 2016). It includes 288243 lexical units for 190648 lemmas, 54791 (52709) of them being multi-word 5 . P W is a network of lexico-semantic relations, an electronic thesaurus with a structure modelled on that of the Princeton WordNet and those constructed in the EuroWord-Net project. Polish WordNet describes the meaning of a lexical unit by placing it in a network representing relations such as synonymy, hypernymy, meronymy, etc. A lexical unit (LU) is a string which has its morphosyntactic characteristics and a meaning as a whole. Therefore, it may be an idiom or even a collocation, but not a productive syntactic structure (Derwojedowa et al., 2008). An LU is represented as a pair lemma, meaning , the last being a natural number. Technically, any LU also has its unique numeric identifier. Each lexical unit belongs to a synset, which is a set of synonyms.
The procedure for annotation
The annotation is performed independently by two linguists, and conflicts are resolved by a third. The whole procedure, together with the annotation of metaphoric expressions, is performed by means of the WebAnno tool (de Castilho et al., 2016) by means of a web browser. In what follows we present the lexico-semantic part of this procedure. The lexico-semantic annotation is based on the morphosyntactic level of annotation. Since the PCC is automatically annotated on that level, annotators have to deal with erroneously annotated tokens. To make the lexico-semantic annotation more comprehensive, we ask annotators to correct errors on the basic level, part of speech (POS) and lemma, required for the lexico-semantic annotation. This correction includes spelling errors, segmentation errors (tech nicz ny instead of techniczny 'technical'), hyphened tokens (naprawdę 'really' for n-a-p-r-a-w-d-ę), resolving abbreviations (w. is used for wiek 'century', wiersz 'line', wieś 'village', wyspa 'island' and wielki 'large'), etc. There are 6 possible error codes (lemma_error, pos_error, tag_error, spell_error, hyphen and case being a subcase of lemma_error limited to the differences in case) and 14 POSs 6 . Next, an annotator has to decide whether a particular token should undergo annotation. Only adjectives, adverbs, nouns and verbs (including gerunds and participles) are annotated as they appear in P W . Typically, the annotation consists in choosing the corresponding P W lexical unit or stating that no such LU exists. Nevertheless, there are some asemantic usages of words that need not to be annotated. They include: • grammatical usage of a word, mainly the verb 'to be' in future and passive constructions or a correlate 'this', • interrogative or anaphoric usage of a pronoun, • the nominal element of a compound preposition, e.g.
Personal pronouns are not represented in P W , with one exception: 'I' meaning 'ego'. The case of interrogatives is a bit more complicated. There are strict instructions in the annotation manual how to deal with pronouns. What is important here is that they cannot be annotated fully automatically. A somewhat different situation appears for named entities (NEs). Several are included in P W , mainly geographical names, but most of them are not. Therefore, a detailed annotation of such tokens, besides the tag name, is optional. Unfortunately, our annotation rules are still more complicated. P W contains several multi-words expressions. The simplest are composed of a verb and the reflexive marker się, e.g.
'to fear'. Usually such annotations exclude each other, e.g.
'to run away' and 'to resort'. However, P W includes reciprocal usages of the się marker, e.g.
'attack each other' for 'attack', which makes both meanings adequate. Typical MWEs may be compositional (e.g. dawka śmiertelna 'deadly dose' is a 'dose') or not (e.g. there is not a meaning for 'centre' for centrum handlowe 'shopping centre', 'mall'). Therefore, we decided to allow linguists to optionally annotate elements of MWEs. This will be especially important, if the corpus is used to train WSD methods, as usually they do not consider MWEs. Technically, we assume that a corresponding LU for a MWE is assigned to its head while annotating its other elements turns to be optional. The procedure for pluralia tantum is similar. Most of them are contemporary used only in the plural (e.g. 'scruple') or have another meaning in plural (e.g. 'efforts' vs. 'treatment'). However, some are distinguished solely for conventional, cultural reasons, e.g. święta 'holidays' (e.g. Christmas, Easter, not 'vacations') is connected with 'holiday', and we want to preserve this connection in annotation. Last but not least, meanings in P W are distinguished in a very detailed, fine-grained way. Hence, sometimes it is hard to decide which sense is adequate in a particular context. We decided to demand assigning a single sense for every annotated item. Nevertheless, annotators are allowed to point out a list of senses that are very close to a chosen one and seem adequate as well. These senses are supposed to be used in the procedure of updating the annotation to the new versions of P W , if a chosen LU is deleted.
Quality of annotation and interannotator agreement
In the previous section we have shown that the procedure for lexico-semantic annotation of BCPM (especially its PCC part) is a complicated, hierarchical process. As a consequence, evaluation of its quality and calculating the interannotator agreement in particular are very important tasks.
Interannotator agreement on correction of morphosyntactic annotation
The first annotator's decision -stating whether the lemma and POS of a particular token is correct and correcting them if needed -influences the whole annotation procedure. Lexico-semantic annotations of two different lemma, POS pairs cannot be compared. On the other hand, the correction itself is not a genuine part of the lexico-semantic annotation per se. Therefore, we decided to evaluate this step separately. There are four possibilities: 1. Only one annotator makes corrections, 2. Both annotators make them, but their corrections differ, 3. Both annotators agree on their corrections, 4. Both annotators accept the original annotation. We decided to calculate two various statistics: taking into account all tokens (i.e. including the case 4.) or considering only potentially improper tokens, i.e. the ones for which at least one annotator intervene. This means considering a set T C ⊆ T of tokens with morphosyntactic annotation changed.
if annotators fully agree, 0.8 if annotators agree on lemma and POS, 0.6 if annotators agree on error code and lemma or POS, 0.4 if annotators agree on lemma or POS, 0.1 if annotators agree on error code only, 0 if annotator fully disagree.
In (6) values of p a t w.r.t. particular annotators' decisions are proposed, which enables us to calculate P (A) in (1). P (E) is calculated accordingly to Scott's π assumptions.
Interannotator agreement on lexico-semantic annotation
The lexico-semantic annotation itself consists of three steps: S1. deciding, whether a token should undergo annotation, S2. deciding, whether it is a case of pluralia tantum, the reciprocal się marker or MWE, S3. performing actual annotation, potentially on two levels, including a decision, whether there are senses "close" to the chosen one adequate in the context. There are 6 tags (grammatical, anaphora, interrogative, prep_element, brev:phrase and rhetoric) used for an asemantic occurrence of a word 7 . 7 A neologism is actually a kind of lack tag, indicating that a word is not supposed to be considered in P W .
For simplicity, elements of MWEs or NEs that are not annotated are treated as asemantic as well. However, they are used for different POSs and cases, in a completely different context, so this is very unlikely to confuse them. Therefore, we decided to treat this decision as a binary one. The next decision concerns how "additional" annotation (pluralia, reciprocal się, MWEs) should be treated. The simplest way is to treat them as independent annotations, a sort of "added tokens". However, this is not the case. Annotating MWEs and pluralia tantum is more important than "basic" single-token annotation as a more precise one (hence its weight is 0.6), whereas the role of LUs with reciprocal sięis auxiliary (hence its weight is 0.4). Instead, we have made a simplifying assumption that a particular token can be annotated only in two ways. It is a bit controversial: a MWE or a NE can include a plurale tantum or a reciprocal verb phrase. The above weight will be referred to in the overall formulae as ω b , ω a for "basic" and "additional" annotation, respectively. To simplify the evaluation procedure, we decided that performing "additional" annotation only is equivalent to performing two levels of annotation with the lack value assigned for "basic" annotation. Nevertheless, asemantic annotations, on the one hand, and "additional" annotation, on the other hand, are proportionally rare: the most typical case is a "basic" annotation.
Since this is a case of a close set of classes, the same for all tokens, we can calculate interannotator agreement for these two phases in a spirit of Scott's π. We will refer to it in the overall formula as β l , l = 2+0, 2+1, 2+2, 1+0, 1+1, depending on whether annotators choose 0 (asemantic), one (only basic) or two levels of annotation of a token. The chance of agreeing or not on the type of "additional" annotation will be denoted as α 0 , α 1 . Furthermore, even if the annotators agree on the number and types of annotation, they can choose a different LU (including a lack decision). The agreement of "close" values is not calculated. Instead, they are used to evaluate the degree of disagreement of annotation of a particular token, i.e. C1. choices of both annotators are the same (100% of agreement despite "close" values); C2. choices of both annotators are included in the partners "close" lists (60% of agreement); C3. a choice of one annotator is included in the partner "close" list (20% of agreement); C4. neither choice is included in the partner "close" list (0% of agreement). This weight will be referred to in the overall formulae as γ. The combinations of these decisions results in values of the annotator's agreement p a t and expected agreement p e t for a token t gathered in table 1.
Preliminary results
The annotation of the BCPM corpus is an ongoing task. Till now, only 162 samples of the PCC part of the corpus composed of 46,350 tokens was annotated twice, which enables us to calculate the interannotator agreement 8 . The fre- One annotator assigns two LUs and the second none 0 β 2+0 One annotator assigns two LUs and the second one Both annotators assign two LUs with a different type of the "additional" annotation Both annotators assign two LUs with the same type of the "additional" annotation One annotator assigns one LU and the second none 0 β 1+0 Both annotators assign one LU Both annotators assign none LU of a different type 0 β 0+0 Both annotators assign none LU of the same type 1 β 0+0
Correction of morphosyntactic annotation
Correction of errors of the automatic morphosyntactic annotation turned to be marginal, it concerns 3.34% of tokens. What is much more surprising, only 15% of them is corrected by two annotators. The reason is that some annotators ignored this step and focused on choosing senses accordingly to the text level. On the other hand, some other were too thorough, correcting conjunctions, prepositions etc. not supposed to be annotated. Because of that, we decided to calculate agreement for three sets of tokens: all annotated tokens, tokens corrected by at least one annotator and tokens corrected by two annotators. In all cases, we present basic agreement P (A), expected agreement P (E) and resulting agreement ϕ. The numbers of particular types of corrections used to calculate P (E) are shown in table 3. The results presented in table 4 show that the most influential is decision whether to correct morphosyntactic annotation of a token or not. Nevertheless, its impact on the whole procedure is weak and can be ignored.
Actual semantic annotation
Semantic annotation was performed for more than 55% of tokens. The numbers do not include tokens for which the corrected lemma or POS differ. The frequency of levels of annotation used to calculate P (E) as in table 1 is presented in table 5. This is not sur- prising that more than 80% of semantically annotated tokens have single, "basic" interpretation given by both annotators. Asemantic usages range between 6% and 15%, but annotators do not agree on that. Two level annotation is marginal. As one may expect, if both annotators decide to assign an "additional" annotation, they agree on its type, only 4 (0.98%) of 189 such annotations are inconsistent w.r.t. the type. The results, presented in table 6, are poor. The reason is that the task is complicated and some annotators seem to misunderstand the instructions. The best results are obtained for the most frequent class, namely nouns. Nevertheless, the
Conclusions
In this paper, we have examined the procedure for lexicosemantic annotation of a particular corpus, namely the Basic Corpus of Polish Metaphors, by means of a particular repository of senses, namely P W . Furthermore, we have shown that it is hard to apply any of the standard interannotator agreement statistics directly and propose a method for adapting them to this very procedure. Most authors declare using Cohen's κ for this. However, they usually do not analyse why they chose this particular statistics and whether it is appropriate for their decision. According to Artstein and Poesio (2008), there are several terminological inconsistencies concerning interannotator agreement statistics in the literature. The tools used for corpora annotation offer to calculate some interannotator agreement statistics. In particular, the WebAnno we are using, makes it possible to calculate Kohen's Kappa, Fleiss' Kappa and Krippendorff's Alpha. However, in order to do it properly, a tool needs to know all interdependencies among categories of tags. This is not the case in WebAnno. de Castilho et al. (2016) report possibility of constrains concerning applicability of one category (or its set of values) w.r.t. another category only for categories with close set of values (selected from a list). What is more, inserting a sense number "manually" or choosing it from the list is merely a technical difference, but it can influence the calculation, the first being a choice from the opened set of values, the second being a choice from a close set of values. The first can be interpreted as binary classification (agree/disagree) or limiting the set of values to the introduced ones (close word assumption). None of these interpretations is correct, as such a set of numbers that represent senses are understood as uniform for all tokens. To sum up, the most important conclusion is that the choice of an interannotator agreement statistics that is appropriate for a particular annotation task is not obvious and it needs reasonable consideration every time. As for the results of annotation, large number of spelling errors etc. shows that traditional linguists do not understood and disregard computational requirements of the annotation procedure. Further analysis reveals that some annotators have misunderstood the instructions. We are aware that the procedure for annotation is complicated and some its cases turned to be controversial. Nevertheless, this emphasises the sense of evaluating the results of annotation in such a preliminary stage. It is a good moment to establish weak point of the procedure (the great role of superannotators), change the instructions and train the team of annotators. The improved version of the guidelines of the annotation procedure (in Polish) is available at http://zil.ipipan.waw.pl/CORMETAN? action=AttachFile&do=view&target= instrukcja_sem-web.pdf. | 2020-05-29T13:12:14.546Z | 2020-05-01T00:00:00.000 | {
"year": 2020,
"sha1": "5b13e19bf6d3cf8f5d43fe8b09ea41702585ae3b",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "ACL",
"pdf_hash": "30cc7a9eec819a86de915ea06db939c9beef1f6b",
"s2fieldsofstudy": [
"Linguistics"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
118852925 | pes2o/s2orc | v3-fos-license | Discrimination of correlated and entangling quantum channels with selective process tomography
The accurate and reliable characterization of quantum dynamical processes underlies efforts to validate quantum technologies, where discrimination between competing models of observed behaviors inform efforts to fabricate and operate qubit devices. We present a novel protocol for quantum channel discrimination that leverages advances in direct characterization of quantum dynamics (DCQD) codes. We demonstrate that DCQD codes enable selective process tomography to improve discrimination between entangling and correlated quantum dynamics. Numerical simulations show selective process tomography requires only a few measurement configurations to achieve a low false alarm rate and that the DCQD encoding improves the resilience of the protocol to hidden sources of noise. Our results show that selective process tomography with DCQD codes is useful for efficiently distinguishing sources of correlated crosstalk from uncorrelated noise in current and future experimental platforms.
The accurate and reliable characterization of quantum dynamical processes underlies efforts to validate quantum technologies, where discrimination between competing models of observed behaviors inform efforts to fabricate and operate qubit devices. We present a novel protocol for quantum channel discrimination that leverages advances in direct characterization of quantum dynamics (DCQD) codes. We demonstrate that DCQD codes enable selective process tomography to improve discrimination between entangling and correlated quantum dynamics. Numerical simulations show selective process tomography requires only a few measurement configurations to achieve a low false alarm rate and that the DCQD encoding improves the resilience of the protocol to hidden sources of noise. Our results show that selective process tomography with DCQD codes is useful for efficiently distinguishing sources of correlated crosstalk from uncorrelated noise in current and future experimental platforms.
I. INTRODUCTION
Recent multi-qubit experiments have reinforced the need to precisely characterize the dynamical processes governing emerging quantum computing devices [1][2][3][4] . The convenient assumption that a qubit experiences only independent noise is rarely valid and capabilities to accurately differentiate between separable and correlated quantum dynamics is needed 5,6 . The task of differentiating between two possible models for a quantum process can be cast as a decision problem in the context of statistical hypothesis testing [7][8][9][10] . In general, channel discrimination selects a model for an underlying dynamical process by inferring the completely-positive, tracepreserving map that takes a set of known input states to a set of measured output states. Measurements of the output state indirectly reveal characteristics of the CPTP map that can be used to discriminate between different potential models.
Previously, quantum process tomography (QPT) has been used to completely characterize and, therefore, discriminate one-and two-qubit channels using full reconstruction of the governing quantum process [12][13][14][15][16][17] . However, this complete form of QPT quickly becomes intractable for higher dimensional systems because the number of required measurements scales exponentially with the system size. Several alternative characterization methods have emerged to address the outstanding challenges of QPT such as state preparation and measurement (SPAM) errors. This includes randomized benchmarking, which reports an averaged fidelity for known gates 18,19 , and gate-set tomography, which requires even more measurements than standard QPT 20,21 . While these methods are operationally more robust for channel characterization, they are not intended for efficient channel discrimination.
Complete reconstruction of a quantum process is not generally necessary for purposes of channel discrimination. Indeed, relatively few measurements may suffice to decide between different models for a set of observed behaviors. This is especially relevant for multi-qubit models, where exhaustive measurements are intractable but a priori information about the expected dynamics may be available. We address channel discrimination in this context by using selective process tomography. In particular, we leverage recent advances in the direct characterization of the quantum dynamics (DCQD) to impose the structure of quantum error detection codes on the task of discriminating between different channel models. As shown previously, DCQD allows for selective retrieval of tomographic information characterizing a quantum process [22][23][24][25][26] . Initially, DCQD was shown to enable piece-wise reconstruction of a channel by directly measuring elements of the underlying process matrix. This idea method was later extended to simultaneously encode logical qubits while performing tomographic measurements 27,28 . More recently, the inclusion of error detection techniques was shown to further improve estimation of the process matrix elements in the presence of quantum noise 29 . A similar idea has been put forward by Unden et al. for metrological measurements 11 .
We show that selective characterization of an unknown process matrix using DCQD is sufficient to perform channel discrimination. In addition, we show that DCQD codes afford a natural and transparent framework for this task while also increasing the resiliency to unknown (i.e., not modeled) channel noise. We illustrate these points using numerical simulations of multi-qubit dynamical processes under the influence of correlated and uncorrelated noise models. In particular, we consider the case of discriminating between a coherent entangling channel and its noisy equivalent. We also treat the case of selecting between incoherent correlated dynamics and an identical, independent noise model. For both examples, we confirm that selective process tomography is sufficient to discriminate the correct channel with very high probability at low false alarm rate. The latter quantities relate the performance of the protocol to operational goals.
The remainder of this work is organized as follows: in Sec. II we introduce the direct characterization methods employed to perform selective process tomography. Next we formulate statistical estimation and inference tests taking the select tomographic data as input. Sec. III introduces a specific example of a CNOT gate parameterized by an angle θ quantifying the amount of entanglement which can be generated by the gate. We numerically simulate the estimation protocol and detail the statistical process underpinning the hypothesis testing protocol. We further explore the efficiency of the protocol in the presence of noisy quantum sources and investigate noise filtering using quantum error detection protocols. Next, arXiv:1607.05292v2 [quant-ph] 28 The initial composite state ρ (alternatively a pure state ψ0) for the pure and ancilla systems evolves under an unknown channel E. Channel discrimination selects between possible models for the channels, e.g., Ei = (E0, E1). Afterwards a unitary or projection operator may be applied to the principal system and this is followed by a syndrome readout. Using the syndrome measurement results as an input, the selected elements of the process matrix χnm are reconstructed. [4,0,2]] code error syndromes ei partition the Hilbert space into the direct sum of states |i = Ei|0 indexed by the integer i for the group of 16 located errors E l = P2/{±1, ±i}. The oneto-one correspondence between the located error operators and the syndromes means that the code is non-degenerate with respect to the set of located errors. The [ [4,0,2]] qubit syndromes are generalized to [[6,0,2]] qubit syndromes by the addition of the parenthesis terms as described in Ref. 29. Unlike the errors in this table, the syndromes for weight one operators with support on the ancilla system (Eu ∈ E) begin with either 01, 10 or 11.
in Sec. IV we explicitly address another example, namely that of estimating the degree of correlation present in a two-qubit incoherent noise source. We develop a model for the correlated noise and show that a two-qubit correlated channel can be detected, even in the presence of single qubit noise sources. Finally, we discuss the results in a broader context and make a few concluding remarks.
II. DIRECT CHARACTERIZATION OF QUANTUM DYNAMICS
We outline the basic theory underpinning the tomography based channel discrimination scheme. The central concepts presented include A) DCQD-style selective process tomography, B) model-specific parameter estimation, and C) statistical hypothesis testing.
A. Selective Process Tomography
The DCQD protocol enables direct experimental characterization of unknown process matrix elements 22,29 . Underlying this idea is the Choi-Jamiołkowski isomorphism between ddimensional channels and d 2 dimensional states, which allows one to directly associate experimental measurement probabilities with process matrix elements. Consequently, performing DCQD tomographic measurements on an n qubit system requires a minimum of 2n qubits. The first n qubits form the principal system P to be characterized while the remaining qubits represent an ancilla system A used for measurements. Focusing on a two-qubit channel, consider the composite system to consist of four qubits. An example of the corresponding DCQD protocol for the two-qubit channel is presented in Fig. 1.
All of the qubits are initialized into a maximally entangled state with respect to the principal-ancilla bipartition, i.e., |Ψ 0 = 1/ √ n j |j ⊗ |j , where |j runs over the n qubit basis set. We assume that the initial state preparation is ideal and the density operator representation of the initial state is ρ 0 = |Ψ 0 Ψ 0 |. It is important to note that the state |Ψ 0 represents the one-dimensional codespace for a stabilizer code. For example, the state |Ψ 0 = 1/2(|0000 = XIXI, IXIX, ZIZI, IZIZ , where X, Z denote the single-qubit Pauli bit and phase flip operators and |Ψ 0 is expressed in the computational basis. More generally, the 2n-qubit DQCD code is generated by a group of stabilizer elements with each element having a matching support on one qubit from both the principal and ancilla systems. To simplify notation we perform a unitary rotation to the stabilizer basis which is indexed by an integer representation of the error syndrome (i.e. |i = E i |0 for E i in the Pauli group supported by the principal system, see also Tab. II). In the stabilizer basis the code state is simply ρ 0 ≡ |0 0|. The prepared code state is next sent through the channel E i as shown in Fig. 1. Expressing the output state in the stabilizer basis, E i (ρ 0 ) = j,k χ j,k |j k| ≡ j,k χ j,k E j |0 0|E † k , highlights the connection between the process matrix representation and the DCQD code space. Stabilizer measurements project the composite system into the subspaces indexed by the stabilizer error syndromes e = {e 1 , e 2 , ..., e 2n } where e i = 1, 0 refers to a state belonging to the ±1 eigenspaces of the generator g i . Stabilizer measurements that correspond to the state |j occur with probability χ j,j and characterize the diagonal of χ in the stabilizer basis. By comparison, the off-diagonal elements of the process matrix may be recovered by applying a (two-local) unitary or projection operator to the output state E i (ρ 0 ) before the stabilizer measurements. This extra rotation or projective measurement maps the offdiagonal χ elements onto the diagonal, which may then be extracted by direct measurement in the stabilizer basis. Additional details regarding these DCQD measurements have been presented previously in Refs. 22,27,29. B. Model-specific Parameter Estimation We use the DCQD framework as a tool for discriminating between a pair of quantum channels. The first step in comparing two models is to expand each candidate channel in terms of its process matrix representation. We assume the models are parameterized by a quantity of interest which is to be estimated by model-specific process tomography. The representation of the process matrix for a channel can have many terms that vanish (or are independent of the channel being identified) with the parameter of interest appearing in relatively few terms. Knowledge assumed for the model of the channel reduces the resources needed to perform channel discrimination, in contrast to the exponential resources necessary for complete tomography of an unspecified model. Unlike arbitrary assumptions, knowledge about the expected channel behavior may be inferred from composing constituent parts or from indirect characterization of the system.
In general there are many ways to probe a specific parameter. Furthermore, the various measurement probabilities (i.e. stabilizer measurements after possible unitary rotations or projectors are applied) will depend differently on the parameter. This leads us to emphasize that it is important to maximize the sensitivity of the measurements depending on the parameter being estimated. We minimize the estimator variance, calculated using the Cramer-Rao lower bound, by picking a measurement set which maximizes the Fisher information.
Given a measurement scheme it remains to collect data from syndrome measurements and estimate the parameter. There is freedom in how to estimate the model parameter(s), and we compare two different estimation techniques for the different examples presented below. We use the maximum likelihood (ML) estimator to find the parameter that maximizes the likelihood function of the observed data in parameter space for the case of the noisy entangling channel. For the incoherent noise model, we develop analytic expressions that relate syndrome frequency directly to the parameter of interest, which we denote as a direct estimator. While the latter analytic estimation avoids the use of maximization searching (which may fail due to local maxima), we find that finite sampling can also lead to an unphysical parameter estimate, e.g. a negative probability. We resolve this inconsistency by simply mapping all unphysical estimates to the physical estimates in an ad-hoc manner. We find this approach works quite well in the context of correlated channel discrimination.
C. Model Selection
Parameterization by a continuous variable generates an infinite set of channels which we partition into either null or alternative classes of channels. For example, in Sec. IV we ask if a two-local gate induces correlated bit-flip errors on the qubits supporting the gate. The answer must either be i) yesthe gate induces correlated errors -or ii) no -the noise is of a local form. From the point of view of channel discrimination, the exact magnitude of the errors is not the main quantity of interest. This leads to the question of given a parameter estimate, how should one decide to accept or reject the alternative hypothesis in favor of the null case? The answer is that the decision should be statistically motivated as explained below.
Statistical inference from experimental data is common in channel discrimination and in general statistical decision theory. The Wald test is a parametric statistical test applicable to continuous variables (our unknown parameters will be come from a continuous space) which we will use to test the true value of a parameter based on a sample data. The univariate Wald statistic is given as 31 whereθ is our estimator, θ 0 is the parameter value under the null hypothesis, and the denominator is the variance of the estimator. We use the Cramer Rao lower bound for the variance in Eq. 1 which is itself found using the Fisher information where the expectation value E [ ] is taken over the stabilizer outcomes (X) conditional on a given value of θ.
The binary decision for which quantum channel should be chosen is performed by comparing the Wald statistic to a threshold value λ * . If W > λ * then the alternative hypothesis is selected while the null is selected only when W ≤ λ * . We use the fact that the Wald statistic is χ 2 distributed under the null model to set the value of the threshold λ * . This allows us to pick a critical threshold which bounds the probability of a 'false alarm' event, i.e., selection of the alternative hypothesis when the null model is true. The probability of detection p D ≡ Pr(W > λ * ) approaches unity in the asymptotic limit for any θ = θ 0 . However, given a finite set of measurements, we may still bound the range of possible θ and perform nontrivial channel discrimination.
III. COHERENT NOISE DISCRIMINATION
As a first example illustrating the channel discrimination, consider an imperfect CNOT that partially entangles two qubits. This may correspond in practice to a scenario in which an ideal CNOT gate was intended to be implemented, but evidence suggests that the resulting qubits were only partially entangled. We model the imperfect gate as a controlled rotation about the angle θ with the latter related directly to the degree of entanglement.
Circuit representation of the (a) maximal and (b) noisy entangling channels denoted as E0, E1 respectively. In Sec. III we refer to E 0(1) as the null (alternative) channel. The process matrix decompositions of the entangling gates are given in Eqs. 3,4. We write the CNOT operator as CX 12 where qubit 1 is the control and qubit 2 is the target and the final useful expansion is a summation over operator elements belonging to the two qubit Pauli basis. A common way to realize the CNOT gate is via a controlled-Z (phase) gate conjugated by Hadamard gates on the target qubit as shown in Fig. 2(a). Now consider a variant of the CZ gate where the |1, 1 state acquires an arbitrary phase e iθ instead of −1. This controlled phase gate interpolates between the maximally entangling CZ (θ = π) and the trivial identity gate (θ = 0, 2π). Likewise, the gate generated by the controlled phase gate conjugated by Hadamard operators continuously interpolates between the identity and a CNOT as a function of θ as is represented by Fig. 2 (b).
The CNOT process acting on a density operator ρ is E CX (ρ) = CXρCX † and may also be expressed in the process matrix representation as for which The above matrix is represented using the partial basis {F ↑ m } = {II, ZI, IX, ZX}, whose elements correspond to E 0 , E 8 , E 1 , E 9 , respectively, in Tab. II). The noisy entangling gate CX(θ) is similarly expressed as CX 12 which reduces to Eq. 3 for the null model at θ = π and the identity gate for θ = 0, 2π. With the rotated CNOT model in mind, let us now probe the entangling nature of the gate by performing a set of measurements designed to estimate θ.
Our goal now is to take the [[4,0,2]] code state as our resource state and perform stabilizer measurements to determine the true θ after passing through the CX(θ) channel. The probability for each stabilizer result (denoted by p i ) is given by the diagonal elements χ i,i (θ) of Eq. 4. Relating the observed syndrome frequencies with the diagonal χ elements we can then obtain a ML estimate for the over rotation parameterθ M L . We can then use the ML estimate to represent the alternative hypothesis and use the perfect CNOT as the null hypothesis.
The simplest measurement scheme is to make stabilizer measurements following the CX(θ) channel. The four nonvanishing syndrome probabilities would be p 0 = (3 cos(θ) + 5)/8 and p 1 = p 8 = p 9 = (1 − cos(θ))/8. Unsurprisingly, this is not the best measurement basis for estimating θ. This is because all the stabilizer probabilities depend only on cos(θ) which is a flat function near θ ≈ π. Thus, a small change in the parameter θ leads to a very small change in the measurement outcomes, in contrast to the high sensitivity one would like to achieve. To make this observation more quantitative we calculate the Fisher information dθ 2 log (Pr(X|θ)) for this set of syndrome probabilities where the expectation value E [ ] is taken over the stabilizer outcomes conditional on a given value of θ. Note that we have verified the ML estimator is unbiased since the score of the log-likelihood vanishes for all values of θ. Our intuition is confirmed by the Cramer-Rao lower bound CRLB(θ) = I(θ) −1 = 1 N ( 2 3(cos(θ)+1) + 1) which fundamentally lower bounds the variance Var(θ). Because CRLB(θ) diverges at θ = π the initial diagonal scheme is not optimal and we look for another measurement basis.
We verify that the stabilizer measurements saturates the Cramer-Rao lower bound by numerically sampling N stabilizer results generating the data set X = {x 0 , x 1 , ...x 15 } where each x i indicates the number of times the ith stabilizer outcome is measured (see Tab. II). Given X we determine theθ M L which maximizes the likelihood function Pr(X|θ). We repeat this calculation M times to numerically calculate the variance of the distribution from whichθ M L is drawn. In Fig. 3 panel (b) we plot the numerical variances as a function of N for ture underlying values of θ = (π, 1.1π) and compare to the theoretical solid lines. We that the lower bound is achieved for the O = U 1 ⊗ U 2 pre-processed measurements and we use the current measurement scheme for the remainder of this section.
Given a set of stabilizer data X and the estimatorθ M L what can be said about the true value of the parameter θ? We answer this question by returning to the the Wald statistical test defined in Eq. 1. We know that under the null hypothesis the test statistic is central χ 2 distributed. We can use this fact to bound the probability p F A of a false alarm detection (i.e. the probability to reject θ = π when in fact the null hypothesis is true). In Fig. 4(b), we numerically calculate the probability p F A as a function of the decision value λ * . In order to determine the probability that W > λ * we numerically create the χ 2 distribution which contains 2 × 10 6 samples. Choosing to constrain p F A < 0.02 (shaded region) we pick a critical test statistical value of λ * = 4. To understand the range of detectable θ's given the critical threshold chosen we turn to the receiver operating characteristic (ROC) curves plotted in Fig. 4(c) 32 . The ROC curve shows the dependence of the false alarm vs. detection probabilities as λ * is varied. Setting λ * = 4 simultaneously sets the false alarm probability at p F A = 0.02 (see red dashed line panel c) and restricts the range of θ one can detect given a finite number of measurements. Specifically panel (c) illustrates how given our choice of λ * modifies the probability of detecting a δθ = (0.005, 0.01)π (where δθ ≡ θ − π) as a function of the number of measurements performed. Given 10 4.25 measurements we'll detect a 1% over rotation with unity probability and a 0.5% over rotation with probability p D ≈ 0.6. In the asymptotic limit N → ∞ any θ = π is detected. Given the λ * bound for pF A, pD approaches unity for small δθ when the sample size N is large.
A. Effects of Noise
We presented in the last section how statistical inference testing can determine the character of a channel. However, employing these methods in realistic systems means accounting for the presence of noisy quantum channels and imperfect measurements in the inference test performance. We now investigate two different noisy quantum channels and analyze how the corresponding inference results change when using our protocol. We expect that incorrect inferences arise because the noisy quantum channels modify the stabilizer probabilities in an unforeseen manner. Therefore, in an effort to mitigate the harmful effects of the noisy channels, we also consider the case where an extra pair of ancillas is used for error detection purposes as introduced in Ref. 29. FIG. 5. Syndrome probabilities (p0, p1, p8, p9) as a function of the entangling rotation angle θ and noisy channel parameter magnitude. Panels (a-d) correspond to the independent and identical amplitude damping on each qubit while panels (e-h) correspond to the the depolarizing channel. Vertical slices at γAD(pDP ) = 0 correspond to the probabilities given in Eq. 5 for noiseless measurements. Starting from the point θ = π, γAD(pDP ) = 0 notice the qualitatively similar behavior of the probabilities under an increase or decrease in θ while keeping γAD(pDP ) = 0 versus an increase in γAD(pDP ) while keeping θ = π. The probability shift seen above coupled with the noiseless limit probability distributions Eq. 5 bias the estimatorθ.
Consider the scheme outlined in Fig. 1, where a noisy quantum channel acts identically on each physical qubit. We assume the noisy channel acts only once and allow the magnitude of noise to vary. Specifically, we act independently and identically with amplitude damping (AD) or depolarizing (DP) channels on each qubit. An operator sum representation for the amplitude damping channel on qubit i is E AD i (ρ) = a E a,i ρE † a,i using the Kraus operators The noisy syndrome probabilities can still be calculated according to p i = Tr [Π i ρ] but these expressions are rather complicated, so instead, to develop a qualitative understanding of how noise affects our inference decisions, we plot the probability for each of the four syndromes used for θ estimation (e 0 , e 1 , e 8 , e 9 ) as a function of the CNOT rotation angle θ and the strength of the noise in Fig. 5. To see how the inference is affected we focus on the syndrome probabilities in the relevant regime of small deviations δθ away from θ = π and weak noise γ AD , p DP 1. By increasing θ beyond π, the probabilities p 1 , p 8 increase from zero to a finite value while p 0 increases and p 9 decreases. The behavior of the syndromes is symmetric under the exchange of syndromes and reflection of θ about π with p 0 (−θ) = p 9 (θ) and similarly for syndromes 1 and 8. The behavior under the increase in θ is qualitatively similar to that of the probabilities starting at θ = π and turning on the noisy channels. If we do not account for the noise sources in our model of the channel, the noise biases the esti-matorθ.
Biasing away from θ = π now occurs in the presence of noise since we are actually probing an effective CX(θ) with noise channel. The effects of this bias are illustrated by the solid lines in Fig. 6 (a,b) where the solid data points correspond to the estimate bias B[θ] = E[θ] − θ as a function of noise strength. In order to reduce the effect of this biasing, we look at a similar protocol with error detection where we utilize two ancilla qubits to generate a larger resource state known as the [[6,0,2]] code 29 capable of detecting weight one errors on the two ancilla system qubits not involved in the CNOT rotation. The [ [6,0,2]] code has 2 6 syndrome outcomes, compared with 2 4 in the [ [4,0,2]] code, and is structured so that errors with support on the qubits not involved in the CNOT map the system to the newly extended syndrome space (see syndromes in parenthesis in Tab. II). The overall effect is that the measured syndrome probabilities can more accurately reconstruct the probabilities given in Eq. 5. Indeed, this error reduction can be seen by comparing the dashed and solid data points in Fig. 6 (a,b).
Returning to channel discrimination, let us recall that the estimateθ is used to make a hard decision concerning if an over/under rotation or perfect rotation has occurred. The decision is performed by comparing the Wald statistic, given in Eq. 1 as a function the estimated parameter, to a critical threshold. The overall probability for an incorrect inference is Pr(θ = π)p F A + Pr(θ = π)(1 − p D ) where p F A and p D are the false alarm and detection probabilities. Fig. 6 shows these quantities of interest (p F A , p D ) as function of noise strength using the λ * = 4 threshold decision from the noiseless case we'd like to relate the matrix elements of Eq. 7 to syndrome measurements, thus directly estimating each bit flip probability.
B. Direct estimation of bit flip rate estimation
We estimate the parameters for the alternative model from stabilizer measurements that relate the measured process matrix elements to the bit-flip rates. The idea is that the χ matrix elements in Eq. 7 are to be experimentally probed via a DCQD procedure discussed in Sec. II and the matrix elements are functions of the bit flip rates via the functions α, β, γ, δ defined in Eq. 8. Recalling the normalization condition α + β + γ + δ = 1 we have three equations for the three unknown parameters p X 1 , p X 2 , p X 12 . Defining for we find simple expressions for the bit flip parameters which are implicitly functions of the χ matrix elements. Given stabilizer measurements, the Eq. 10 provide us with the alternative hypothesis estimators (i.e., with a non-zero p X 12 ). However, in the case of the null hypothesis we set p X 12 = 0 which simplifies the null hypothesis bit flip expressions So unlike the maximum likelihood parameter estimation presented in the last section, we are now using the structure of the correlated bit flip model to directly estimate the hypothesis parameters through the closed form expressions given in Eqs. 9-11. Now lets return to the direct estimation of the bit flip parameters.
In order to do this, we'll solve Eqs. 10,11, in terms of Eq. 9 which will determined by finding specific elements of Eq. 7 as per the Sec. II procedure.
If we perform stabilizer measurements on the final state E X1X2 (E X1 (E X2 (E CX (ρ 0 )))) the probabilities for each syndrome outcome are p = {a, a, 0, 0, b, b, 0, 0, a, a, 0, 0, b, b, 0, 0} where a = α + β, b = γ + δ where α, β are defined in Eq. 8 and p = (p 0 , ...p 15 ). These probabilities are simply the diagonal elements of Eq. 6. Bit flip rate estimation involves the three unknowns p X 1 , p X 2 , p X 12 -or alternatively any three of α, β, γ, δ -while O j = I (i.e. diagonal) stabilizer measurements only reveal the linear combinations α + β and γ + δ. At least one additional measurement with an appropriate O j pre-measurement operator must is needed in order to estimate the full set of bit flip probabilities.
Let's now find the appropriate O j operators needed to extract information about the full set of bit flip probabilities from the set of stabilizer measurements. Since all the off diagonal elements of Eq. 7 are real (as opposed to the imaginary terms of interest in Eq.4) we'll need the projective operators P ± j = I±Ej √ 2 where the operators E j refer to the Pauli operators listed in Tab. II (i.e. with support on the principal system). The set of coherence maintaining projective operators } will work for the task (see Ref. 29 and appendix B for details). Intuitively, the projection operators prepare a new quantum state whose syndrome probability distributions are functions of the off diagonal χ elements. Together with the diagonal measurement data, we can completely specify χ ↑ , χ ↓ . The projection operators P ± 1,8,9 can easily be implemented by measuring the single qubit eigenvalues of the Z 1 , X 2 , and Z 1 X 2 operators respectively. Given the state CX 12 |Ψ 0 (i.e. the state before bit flips) both ±1 outcomes (for all three single qubit Pauli projectors) occur with probability 1/2. Implementing a projective operator and then measuring stabilizers, one will come up with two data sets denoted X ± , with ± denoting the measured eigenvalue of the E i used in the projector. The differences between the ±1 syndrome frequencies yield the process matrix elements as given by Eq. B2 in the appendix. Since only a subset of off-diagonal χ elements is needed to fully determine the bit flip probabilities we choose to use P ± 8 = I±Z1 √ 2 in our simulation with no loss of generality.
As a final comment on the direct estimation method before presenting our simulation results, we have seen that one drawback of direct parameter estimation is that, given finite samples, the measured χ ij can sometimes lead to unphysical parameters (with p X i < 0). This issue only occurs in the regime when the true underlying model has a small p X 12 ≈ 0. In the event of an unphysical parameter we simply truncate the estimated value of p X 12 to 0, thus returning the correlated bit flip mapping back to a CPTP map. Of course, maximum likelihood and other techniques could by construction always yield physical mappings 30 , but the point of this section is to contrast a direct estimation technique to the ML estimation used in the previous section so we shall simply use a truncation scheme to enforce physically on the estimated dynamical mappings.
C. Results and Effects of Noise
We now perform a Monte-Carlo simulation where syndrome results are sampled from a true underlying probability distribution given by p i = Tr [Π i ρ]. To perform hypothesis testing, we'll again use the Wald test which takes the estimated valuep X 12 and Fisher information as input and yields a test statistic which can be compared to a pre-determined threshold value. Since we are estimating all three bit flip probabilities, the Fisher information is now a matrix I(θ) ij = −E d 2 dθidθj log (Pr(X|θ)) (which we evaluate analytically from Eq. 7). The variance for each estimator is now lower (a,b) show the probability of the Wald statistic being greater than the critical thresholdp F A(D) in the case that the null (alternative) hypothesis is true -as a function of noisy amplitude damping (with strength γAD ) and depolarizing channels (pDP ) strength. True underlying values of p X 12 are given by the color code in panel (b) and p X 1 = 0.01, p X 2 = 0.02 are the uncorrelated parameters. A total of N = 1000 measurements were used to generate a single estimation event, and a Wald statistic, and M = 5000 instances of the simulation were used to determine the probabilities that W > λ * . bounded by the appropriate elements of the inverse Fisher information matrix. In Fig. 8 panel (a) we plot the Cramer Rao lower bound on p X 12 for various p X 1 , p X 2 values. We note that the variance lower bound vanishes in the limit {p X 12 , p X 1 , p X 2 } → 0 and that the functional dependence on p X 12 for the CRLB resembles the CRLB for a classical biased coin.
Using the CRLB as a variance lower bound and an estimatorp X 12 we can calculate the Wald statistic (Eq. 1) as before. We simulate syndrome results and calculate the distribution of W given both hypothesis and various single qubit bit flip probabilities. The results are plotted in Fig. 8 panel (b) and a critical threshold of λ * = 4.25 is chosen (for blue circled data) which sets the false alarm probability approximately to p F A = 0.02. In general the threshold value depends on the independent bit flip rates as well as seen in panel (b) for three different individual bit flip rates. In what follows we proceed assuming the independent values chosen are p X 1 = 0.01, p X 2 = 0.02. We have so far numerically confirmed that statistical testing can be successfully implemented by directly estimating (as opposed to ML estimation) model parameters. Further, we have upper bounded the Wald test false alarm probability at p F A = 0.02 by our judicious choice for the critical statistic threshold λ * . As before, in the event that the alternative model is true one can detect any arbitrary non-zero parameter p X 12 in the asymptotic limit. To understand how our protocol performs in a realistic setting, we subject the protocol to noisy channels following the logic of Sec. III. Specifically, we apply a single round of amplitude damping and depolarizing channels parameterized by γ AD and p DP respectively. We will use the probably of false alarm (p F A ) and detection probabilities (p D ) as metrics for quantifying the degree to which the protocol effectiveness is reduced. Recall that the p F A is the probability that W > λ * = 4.25 in the event that the null hypothesis is true (p X 12 = 0) and that p F A was explicitly set to 0.02 as seen in Fig. 8(b). Fig. 9 (blue filled in circles) shows that p F A increases as the noise strength increases, eventually saturating to unity in the large noise regime. In general, the noisy binary hypothesis probabilities p F A and p D for the correlated channel behave qualitatively similar to the binary hypothesis probabilities seen in Fig. 6. However, a notable difference between the two is that for coherent rotation estimation the amplitude was more detrimental discrimination (p F increased more quickly for γ AD ) while for the bit flip example the situation is reversed with the depolarizing channel leading to a greater failure rate. The explanation for this behavior is that the depolarizing channel increases the individual bit flip rates across all qubits which directly biases the correlated bit flip parameter rate. On the other hand the amplitude damping channel affects other elements of the system density matrix which are not directly probed by the measurements we consider. Also similarly to the last section, the intuition behind the increase of p F A is that noisy channels modify the underlying stabilizer probabilities from the known null hypothesis probability distribution to one which resembles that of a finite p X 12 thus biasing the estimator and increasing p F A . After seeing that noise biases p X 12 estimates upwards, it is also interesting to see how simple error detection mitigates the effects of noise. The empty circles in Fig. 9 plot the false alarm and detection probabilities when a [[6,0,2]] 29 initial codeword and syndromes are used. The results of this encoding are qualitatively similar to those in the last section, however the bias reduction in this case is clearly more pronounced especially in the case of the noisy AD channels as seen in panel (a).
V. CONCLUSIONS AND DISCUSSION
We have presented a general and intuitive theory for quantum channel discrimination based on selectice process tomography within the framework of direct characterization of quantum dynamics. This approach avoids the complete characterize of the quantum channel required by QPT for purposes of discrimination. The main steps in the protocol include postulating (parameterized) models for the quantum channel and then selecting between these choices by comparing estimates for the model parameters. The estimates are efficient as they do not require complete tomographic reconstruction of the channel. Moreover, we have shown how unknown source of noise may bias these estimates and how the ability for DCQD codes to improve estimation lead to improved channel discrimination.
We have explicitly validated these ideas for imperfect entangling operations and for a crosstalk model involving correlated bit flips noise induced by entangling operations. Numerical simulations showed that accurate channel discrimination and parameter estimation are indeed possible, given the test statistic distribution for a detection event lies above a critical threshold value. This is always guaranteed in the asymptotic limit of infinite measurements, but we have also shown the capabilities for discrimination using finite numbers of measurements. While our simulations were made for a two-qubit principal systems, the theory itself applies to the discrimination of general multi-qubit channels.
The effects of unknown sources of quantum noise on discrimination have also been studied. We used amplitude damping and depolarizing channels to quantify the bias in model parameter estimates as well as the probability of an incorrect decision. In general, the addition of unknown noise is harmful to the protocol as the probability of selecting the wrong underlying model increased as a function of the noise magnitude. However, for the case of correlated bit-flip errors a single layer of DCQD error detection was shown to reduce the false alarm rate at low noise magnitudes. This result suggests that some DCQD parameter estimators are more sensitive to channel noise. Further, the development of an entirely selfconsistent procedure to estimate specific elements of a process matrix (but not the full map) in the presence of unknown noise is deferred to future work.
Throughout this work we have assumed the ability to reliably initialize states as well the ability to perform noiseless measurements. However, SPAM errors are present in most physical systems and therefore limit the ultimate effectiveness of tomographic protocols. In light of these issues, we conclude by discussing the impact of SPAM on the discrimination protocol as well as proposed how to potentially detect and reduce the overall effects of SPAM errors. It is also worth noting that systematic preparation and measurement errors were also partially addressed in the context of DCQD in an earlier work 24 .
Our earlier results indirectly investigated the effect of measurement errors by adding sources of noise after the application of the channels of interest. This correspondence arises from the fact that a measurement error can be simulated by choosing the opposite syndrome measurement value with some probability. Such an error model alters the final measurement outcome probability distributions, which is exactly what our sources of error did -albeit with a specific structure corresponding to amplitude damping or depolarizing physical processes as seen in Fig. 5. State preparation measurements can analogously be included by the composition of additional noisy channels with the parameterized channel of interest. Therefore in the presence of SPAM one can expect our approach to be limited in much the same way that is summa-rized by the noisy results presented in Figs. 6 and 9 where the protocol is unable to distinguish between channels in the regime of large error rates, but partially retains the discriminatory capabilities for small error rates.
Using these results, one can therefore propose a simple scheme to detect the presence of SPAM. Suppose that one has control over the parameter θ in the case of the CX(θ) gate or can adjust the time between measurements to vary the depolarizing noise in the second example. One can then in principle vary the relevant parameters, traversing between the alternative and null channels, and such a change should be detected in the SPAM-less limit by the changing functional behavior of the Wald statistic on the estimator parameters as seen in Fig. 4 (a). Alternatively, in the presence of large SPAM errors one would not observe any change in the value of W , since W would be constant as seen in Figs. 6 and 9 for the large error rate regime.
Besides the above approach for indicating the presence of SPAM errors, we would also like to develop a protocol that isolated the channel dynamics from the SPAM. While the best approach to this task is an open question, we draw inspiration from current protocols that are designed to be robust against SPAM errors, namely randomized benchmarking and gate set tomography [18][19][20][21] . The main idea is to repeatedly apply the channel multiple times and perform the direct characterization measurements as the number of channel applications is varied. Repeated channel application then results in an amplification of the noisy signal. For example, N applications of CX(θ) results in an over rotation angle N θ while the effect of SPAM errors is constant since there is only a single preparation and a single measurement step. Indeed this scheme actually shares some similarities to the parameter sweeping discussed above since large numbers of repeated channels would produce a single large effective parameter which could be detected despite systematic preparation and measurement errors. We are hopeful that these future modifications of the protocol will lead to a more general and robust framework of efficient channel discrimination.
ACKNOWLEDGMENTS
This work was supported by a grant from the Intelligence Community Postdoctoral Research Fellowship program to the University of Tennessee. This manuscript has been authored by UT-Battelle, LLC, under Contract No. DE-AC0500OR22725 with the U.S. Department of Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for the United States Government purposes. The Department of Energy will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan.
where the index runs over the located operators listed in Table. II. | 2016-10-28T17:07:41.000Z | 2016-07-18T00:00:00.000 | {
"year": 2016,
"sha1": "2326d10dbc0ffffbd73ea294943a13a7979a3cb8",
"oa_license": "publisher-specific, author manuscript",
"oa_url": "https://link.aps.org/accepted/10.1103/PhysRevA.94.042107",
"oa_status": "HYBRID",
"pdf_src": "Arxiv",
"pdf_hash": "2326d10dbc0ffffbd73ea294943a13a7979a3cb8",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
221739010 | pes2o/s2orc | v3-fos-license | Galaxy properties in the cosmic web of EAGLE simulation
We investigate the dependence of the galaxy properties on cosmic web environments using the most up-to-date hydrodynamic simulation: Evolution and Assembly of Galaxies and their Environments (EAGLE). The baryon fractions in haloes and the amplitudes of the galaxy luminosity function decrease going from knots to filaments to sheets to voids. Interestingly, the value of L$^*$ varies dramatically in different cosmic web environments. At z = 0, we find a characteristic halo mass of $10^{12} h^{-1}\rm M_{\odot}$, below which the stellar-to-halo mass ratio is higher in knots while above which it reverses. This particular halo mass corresponds to a characteristic stellar mass of $1.8\times 10^{10} h^{-1}\rm M_{\odot}$. Below the characteristic stellar mass central galaxies have redder colors, lower sSFRs and higher metallicities in knots than those in filaments, sheets and voids, while above this characteristic stellar mass, the cosmic web environmental dependences either reverse or vanish. Such dependences can be attributed to the fact that the active galaxy fraction decreases along voids, sheets, filaments and knots. The cosmic web dependences get weaker towards higher redshifts for most of the explored galaxy properties and scaling relations, except for the gas metallicity vs. stellar mass relation.
INTRODUCTION
The large scale structure (LSS) of the Universe exhibits a web-like structure, which is usually categorized into four components: voids, sheets, filaments and knots. It originates from small perturbations in the very early Universe and is shaped gradually by the large scale gravitational fields. Different components correspond to different stages of gravitational collapse. Large sheets of matter form via gravitational collapse along one principal direction, filaments form via gravitational collapse along two principal axes and knots form via gravitational collapse along three principal axes. The relatively empty regions of the Universe between knots, filaments and sheets are referred to as cosmic voids, whose density is well below the cosmic mean value. Most mass is contained in knots and filaments, while most volume is filled by voids. The relative mass fractions and filling factors of each component could evolve significantly from high to low redshifts (Cui et al. 2019;Zhu & Feng 2017a). galaxy properties primarily through the strong dependence of the galaxy properties on the mass of their host halo (e.g. Gay et al. 2010;Goh et al. 2019;Yan et al. 2013). This is also supported by Eardley et al. (2015) who found in Galaxy And Mass Assembly survey (GAMA) that the strong variation of galaxy properties in different cosmic web structures vanishes when comparing at the same local density. However, by using SDSS data, Poudel et al. (2017) found that at a fixed halo mass, central galaxies in filaments have redder colors, higher stellar mass, lower specific star formation rates and higher abundances of elliptical galaxies compare to those outside the filaments. Observationally it is hard to measure the halo mass for an individual galaxy while stacking methods could smear out the cosmic web dependence if it is not strong enough. Theoretical work using cosmological hydro-dynamical simulations can provide more clues to such studies. Using IllustrisTNG, Martizzi et al. (2020) found at a given halo mass, galaxies with stellar masses lower than the median value are more likely to be found in voids and sheets, whereas galaxies with stellar masses higher than the median are more likely to be found in filaments and knots. Liao & Gao (2019) found that haloes in filament have higher baryon fractions and stellar mass fractions compare to those in the field.
In recent years, cosmological hydro-dynamical simulations have gained great success in reproducing many observed galaxy properties (e.g. Vogelsberger et al. 2014;Schaye et al. 2015). In such simulations the cosmic web effect is taken into account automatically. Here we use the state-ofart cosmological hydrodynamical simulation, Evolution and Assembly of Galaxies and their Environments (EAGLE) to disentangle the connections between galaxies, dark matter haloes and the cosmic web and revisit the relation between galaxy properties and geometric cosmic web structures. This paper is organized as follows. We introduce the EAGLE simulation and describe the environmental classification method in Section 2. The cosmic-web dependence of the various baryonic components and the scaling relations are presented in Section 3 and Section 4, respectively. In Section 5, we summarize our results.
EAGLE simulation
The EAGLE simulation consists of a series of cosmological simulations performed with a modified version of the N-body Tree-PM smoothed particle hydrodynamics (SPH) code GADGET-3 (Springel 2005). In this paper we use the largest volume EAGLE simulation, labelled as L100N1504, which was carried out in a box of 100Mpc each side, tracing 1504 3 dark matter particles and an equal number of baryonic particles. The initial mass of the gas particles is 1.81×10 6 M and the dark matter particles mass is 9.70 × 10 6 M . The EAGLE simulations adopts the cosmological parameters taken from the Planck results (Planck Collaboration et al. 2014): Ω m = 0.307, Ω Λ = 0.693, Ω b = 0.04825, h = 0.6777, σ 8 = 0.8288, n s = 0.9611 and Y = 0.248.
The simulations used state-of-art numerical techniques and subgrid physics including radiative cooling and photoheating (Wiersma et al. 2009a), star formation law (Schaye & Dalla Vecchia 2008), stellar evolution and enrichment (Wiersma et al. 2009b), stellar feedback (Dalla Vecchia & Schaye 2012), and black hole seeding and growth Rosas-Guevara et al. 2015). These models were proven successful in reproducing many observed galaxy properties including the galaxy stellar mass function, galaxy sizes and the amplitude of the galaxy-central black hole mass relation, etc. (Schaye et al. 2015). The friends-offriends(FOF) method (Davis et al. 1985) was performed on the particle data to generate FOF groups by linking particles separated by 0.2 times the average particle separation. In each FOF group, the SUBFIND (Springel et al. 2001;Dolag et al. 2009) algorithm was applied to identify the self-bound particles as subhalos/substructures. M 200 is adopted to refer to the virial mass, which is the total mass within R 200 within which the average density is 200 times the critical density. Galaxies reside in the center of each substructure. The stellar mass is defined as the total mass of stellar particles within 30 pkpc radii of the centre of each subhalo. The luminosity and color are calculated using a stellar population synthesis model taking into account the SFR history and metallicity of each star particle (Trayford et al. 2015). In the public EAGLE simulation catalog, it provides magnitudes in the five rest-frame SDSS bands and 3 UKIRT bands. This sample contains absolute rest-frame magnitudes for all galaxies with M * > 10 8.5 M . No dust attenuation has been included.
Environmental classification
The cosmic web present itself in large-scale surveys, as well as in cosmological simulations. Many different approaches to classify the cosmic web elements have been developed in the literature. The tidal tensor and the velocity shear tensor are two of the most popular quantities to identify the cosmic web (see Libeskind et al. (2018) for a comparison of different classifications of the cosmic web). Here we follow Hahn et al. (2007) to use the tidal tensor to classify the four cosmic web elements.
A region is identified as void, sheet, filament or knot according to the number of dimensions that this particular patch collapses along. The tidal tensor T i j is given by the Hessian matrix of the gravitational potential φ: The gravitational potential φ from the matter density field is obtained using the Poisson's equation: where ρ denotes the mean mass density of the universe and δ = ρ ρ − 1 denotes the density contrast. In practice, we split the simulation box into 256 3 cartesian cells and estimate the density and density contrast by assigning the particles to each cell using the cloud-in-cell method (Sefusatti et al. 2016). We further smooth the discrete density field with a Gaussian filter of width R s = 1.25Mpc/h.
There are three eigenvalues of the tidal tensor T i, j , λ 1 ≥ λ 2 ≥ λ 3 . Hahn et al. (2007) used the number of positive eigenvalues of T i j (λ th = 0) to classify the four possible environments. However, in reality, such criteria lead to an under-estimate of the filling factor of the voids. We follow Forero-Romero et al. threshold λ th , as a free parameter and define the cosmic web elements as following: (i) Void: all eigenvalues λ i below the threshold λ th (λ th ≥ λ 1 ≥ λ 2 ≥ λ 3 ).
Here we adopt a fixed λ th = 0.25 at z = 0, 1, 3.01, 5.97. We test our results using different λ th at different redshifts and find the results are in qualitative agreement with those with the fixed λ th . This is consistent with the conclusions reached by Zhu & Feng (2017b). In Fig. 1, we show the map of the overdensity and the corresponding cosmic web in a slice 0.265Mpc/h thick at z = 0. The cosmic structures are well captured by the tidal tensor method. is consistent with the results using Illustris simulation that most matter resides either in haloes or in filaments but very little in voids, though the voids contribute a significant fraction of the total volume (Haider et al. 2016). Both the volume fraction and mass fraction of voids increase with redshift, whilst they decrease in knots and filaments.
Total baryons
For the baryon component, as expected, it is a good tracer of the total matter at large scales across cosmic time (Cen & Ostriker 1999;Davé et al. 2001;Eckert et al. 2015). We find that the baryon fraction in knots and filaments decreases dramatically with redshifts. Cui et al. (2019) reached similar conclusions as ours, though their classification of the cosmic web adopted a different cell size and eigenvalue threshold, indicating our results are qualitatively robust against the method of classification of the cosmic web. In addition, Cui et al. (2019) demonstrated that the baryon fraction in different components of the web and its evolution depend only very weakly on the different physics implemented in the simulations.
We further investigate the normalized baryon fraction ( f bar /f = (M bar /M 200crit )/(Ω b /Ω 0 )) as a function of halo mass in different cosmic web environments (Fig. 2), wheref is the universal baryonic fraction 15.7%, f bar is the baryon fraction measured within R 200 . It shows that the baryonic fraction is an increasing function of the halo mass in all web environments. The strong halo mass dependence of the baryonic fraction presents itself over all the redshifts from 0 to 6. At z =0, only in massive clusters (∼ 10 14 h −1 M ), the baryonic fraction does reach the universal value, while this fraction drops dramatically towards lower masses. At low masses, the overall baryonic fraction increases significantly with redshifts. For example, for haloes of M 200crit ∼ 10 10 h −1 M the baryon fraction increases from 0.1 at z ∼ 0 to 0.4 at z ∼ 5.97.
In the MassiveBlack-II simulation, Khandai et al. (2015) found an even stronger redshift evolution of the baryon fraction. Peirani et al. (2012) found a similar trend of the mean baryonic fraction in groups over cosmic time using hydrodynamical zoom-in simulations. They claimed that it is due to the lower accretion rate of dissipative gas onto the haloes compared to that of dark matter at low redshifts. Another reason for the high baryon fraction at high redshifts is that the potential well is deeper and it is thus harder to expel baryons out even for haloes of relatively low masses. On the other hand, for the massive systems (more massive than 10 13 h −1 M ) their baryonic fractions hardly change with redshifts. Haloes of such masses have potential wells deep enough to keep most of their baryons both at low and at high redshifts.
The baryonic fraction increases from voids, sheets, filaments towards knots. For haloes of 10 11 h −1 M , the baryonic fraction in voids is lower than that in knots by 0.8 dex at z=0. It is much stronger dependence than that found by Metuki et al. (2015) who used simulations with different subgrid physics. This cosmic web dependence gets weaker towards lower masses and almost vanishes at 10 10 h −1 M . Fig. 2 also shows that in general the cosmic web dependence gets stronger with redshifts up to z =3, especially at low masses.
At the highest redshift, z∼6, the cosmic web dependence gets weaker again.
Stellar mass-to-halo mass ratios
In this section, we focus on the stellar mass and its relation to the cosmic web. The cosmic web dependences of the galaxy properties vs. stellar masses relations are presented in Sec. 4.
We show the stellar mass-to-halo mass ratio for central galaxies (M ,central /M 200crit ) at various redshifts in Fig. 3. As presented by Schaye et al. (2015), the trend of the full sample is consistent with that inferred by the subhalo abundance matching methods (e.g. Guo et al. 2011;Moster et al. 2013) at z = 0. A similar analysis is applied to the total stellar mass fraction (M ,group /M 200crit , dotted curves), where the total stellar mass referred to is the total stellar masses within R 200 . It peaks at ∼ 10 12 h −1 M and drops fast both towards the low mass and high mass ends. Different from the stellar-to-halo mass ratio for central galaxies, the slope at high masses is much flatter for the total stellar masses. At low masses, the central and total stellar mass to halo mass ratios are almost identical, indicating that satellite galaxies and intra-cluster lights contribute more to massive systems than to low mass systems. These trends are found at redshifts up to z ∼ 3. At z = 6 no high mass systems are formed due to the limited box size. The amplitudes of the central and total stellar-to-halo mass ratios decrease towards higher redshifts, consistent with the finding from the subhalo abun- dance matching method (Moster et al. 2013) that the galaxy formation efficiency decreases slightly with increasing redshifts. At z = 0 the central and total stellar mass-to-halo mass ratios decrease from knots to filaments to sheets and to voids for haloes below ∼ 10 12 h −1 M . In haloes of 10 11 h −1 M , where most of the dwarf galaxies reside, the ratios drop by a factor of ∼ 2. At high masses, it is at the opposite. In knots the ratios are the lowest while in voids the ratios are the highest. At the peak location (∼ 10 12 h −1 M ), about the mass of the Milky Way's halo, the central and total stellar mass-to-halo mass ratios do not vary between different cosmic web environments. Such environmental dependencies are also found at z=1, though somehow weaker than that at z=0. At even higher redshifts(z > 3), the cosmic web dependence vanishes at low masses, and no statistical results can be obtained at high masses due to the limited number of high mass systems.
Stellar mass functions
Most stars are locked in galaxies. The fraction of galaxies as a function of stellar mass in different cosmic web environments are shown in Fig. 4. In a given stellar mass bin, the fraction of galaxies in any cosmic web environment is calculated by dividing the number of galaxies in the corresponding web environment by the total number of galaxies. In the top left panel, it shows that at z = 0 most of the massive galaxies reside in knots, consistent with previous results (Metuki et al. 2015;Eardley et al. 2015). For galaxies of the Milky Way mass and dwarf galaxies, most reside in filaments, though a comparable, yet slightly lower fraction reside in knots. This is also found by Eardley et al. (2015) using GAMA data and Metuki et al. (2015) using simulations. Only a very small fraction of galaxies reside in voids at all masses considered here. The fractions in different cosmic web environments are similar at z = 1. At higher redshifts, more galaxies are found in voids and sheets than at lower redshifts, while less is found in knots and filaments, especially at low masses.
The dependence on the cosmic web is more significant when converting the total number of galaxies to the volume number density (the stellar mass functions) as shown in Fig. 5. At z =0, the number density drops by two orders of magnitude from knots to voids. Such strong dependences are also found by theoretical work (Metuki et al. 2015) and observational work (Alpaslan et al. 2015). Interestingly, we find the mass at the knee (corresponding to L * ) decreases by a factor of ∼ 8, with the value of 10 10.8 h −1 M in the knots and 10 10.0 h −1 M in voids. This change is dramatic given that the corresponding mass for the global stellar mass function barely changes since z ∼1 (e.g. Beare et al. 2019). As for the number fractions, the cosmic web environmental dependence of the stellar mass functions gets weaker at high redshifts. The evolution is stronger at low masses than that at high masses.
SCALING RELATIONS IN THE COSMIC WEB
The observed scaling relations are very important in revealing the underlying physics of galaxy formation. In this section, we show how cosmic web environments affect the color vs. stellar mass relation, specific star formation rate (sSFR) vs. stellar mass relation and stellar/gas metallicity vs. stellar mass relation. Since the environmental dependence of satellite galaxies are very much different from that of central galaxies (e.g. Peng et al. 2010), in this section we focus on central galaxies only. Previous work found that the cosmic web dependence and the halo mass dependence are highly degenerate (e.g. Brouwer et al. 2016). Large scale web environments could shape the galaxy properties through the halo mass for galaxies properties depend strongly on halo mass (e.g. Metuki et al. 2015) and large scale structures affect halo masses significantly. We show in Fig. 6 that for a given stellar mass, the typical halo mass varies in different cosmic environments. For galaxies with stellar mass below 1.8×10 10 h −1 M , the host haloes are less massive in knots, while for those with stellar mass above 1.8×10 10 h −1 M , the host haloes are more massive in knots. This is consistent with what we found in Fig. 3. In order to disentangle the connections between the cosmic web, dark matter haloes and galaxies, we measure the pure cosmic environmental effects by removing the dark halo effects. In practice, for any quantity of interest, X, we calculate R X as a function of the cosmic web environment, where R X is defined as: For each M 200crit at a given redshift, we calculate the average value of X, X|M 200crit , in advance. R X quantifies the off-set of the property X that deviates from the expected values for given halo masses.
For any quantity X that depends directly only on halo mass, the mean of R X will equal 0 for any subsample, including one defined by stellar mass and environment, as below. By using R X , we can test whether an apparent dependence of X on environment at a given stellar mass is entirely driven by the dependence of halo mass on environment at given stellar mass, or whether there is some remaining dependence on environment at fixed halo mass.
Color vs. stellar mass
Color is one of the most important observables for it is key to understand galaxies' star formation history. Previous studies found that galaxies on the color vs. stellar mass diagram can be grouped into three categories, blue cloud (active), red sequence (passive) and green valley (those in between). Galaxies in low density environments (e.g. filaments, sheets and voids) tend to be bluer, while those in high density environments (knots) tend to be redder (Poudel et al. 2017). However, this could be caused by the fact that in low density regions galaxies and their host haloes are smaller, while smaller galaxies are in general bluer (Cucciati et al. 2011;Baldry et al. 2006). Fig. 7 shows the g − r color vs. stellar mass relation as a function of the environments and stellar mass. Observed colors can be influenced by dust extinction and we thus use the intrinsic color. For the color is a dimensionless quantity and can be 0 in many cases, instead of using Eq. (3) we adopt a slightly different quantity to remove the halo dependence. For each galaxy, we calculate R g−r = (g-r) -(g − r)|M 200crit . At z=0, for galaxies less massive than 1.8×10 10 h −1 M , it shows a clear dependence on environments, especially for those in knots. Galaxies in knots are redder than those in voids by 0.08 mag (1.8 σ). This is related to the fraction of active galaxies in different environments which we will discuss in more detail in the next subsection. At stellar masses above 1.8×10 10 h −1 M , there is no clear cosmic web environmental dependence. The difference between voids and sheets is weak across all the stellar mass ranges.
The dependence on environments vanishes for galaxies at high redshifts: z =1,3, and 6 at all masses. This is consistent with Darvish et al. (2017b) who found no cosmic web dependence of galaxy color in the COSMOS fields at z = 1.
Specific star formation rate vs. stellar mass
Colors are influenced by the star formation history, especially the current star formation rate. Fig. 8 shows the specific star formation rate (sSFR) as a function of stellar mass and cosmic environments. Similar to the color vs. stellar mass relation, at z=0, the cosmic environmental dependence is the strongest for galaxies with stellar mass less than 1.8 × 10 10 h −1 M . The sSFR is lower by 50% in knots than that in voids. At higher masses, given the large scatters, we can not find clear environmental dependence. This is consistent with the observational results by Poudel et al. (2017) who found central galaxies in low-density environments with higher sSFR compared to those in high-density environments at a fixed group mass using group catalogs sSFR < 10 −11 [yr −1 ] z = 0 Figure 9. The deviation of the sSFR from the expected values as a function of stellar mass in different environments for active (sSFR>10 −11 /yr, left panel) and passive (sSFR<10 −11 /yr, right panel) galaxies at z=0. Red, cyan, blue and green curves donate the median value of the deviation in knots, filaments, sheets and voids, respectively. Error bars are generated using the bootstrap method. (Tempel et al. 2014) extracted from the SDSS DR10. The cosmic environmental dependence is much weaker at high redshifts, broadly consistent with the results of Scoville et al. (2013) and Darvish et al. (2016). Different from the color vs. stellar mass relation, at z = 3 the sSFR is slightly higher in knots compared to that in other environments.
Corresponding to the red sequence and blue cloud, galaxies can be separated into passive and active subpopulations using their specific star formation rates. Unlike galaxy colors, sSFR is not affected by metallicity and star formation history. Here we adopt log (sSFR[/yr])= -11 as the threshold: galaxies with log (sSFR[/yr])< -11 are referred to as passive galaxies, while those with log (sSFR[/yr])> -11 are taken to be active (star-forming) galaxies. The threshold value is fixed over cosmic time as its evolution is not very strong (Matthee & Schaye 2019).
The left panel in Fig. 9 shows the environmental dependence of the sSFR for active galaxies at z=0. For active galaxies, no environmental dependences present themselves except for those with masses ∼ 2.8 × 10 9 h −1 M where the sSFR of central galaxies is lower in voids than in other cosmic web environments. At higher masses, such dependence vanishes. For passive galaxies, as shown in the right panel, there is no environmental dependence over all the stellar mass ranges considered.
The rather low sSFR in knots in the first panel of Fig. 8 might be explained by their low fraction of active galaxies. The average active galaxy fraction increases from knots, filaments to sheets and voids, ranging from 0.58, 0.81, 0.87, 0.88, respectively. When taking into account the expected active galaxy fraction for any given halo mass and the number of haloes of the given mass in each cosmic web environment, the derived expected active galaxy fractions are 0.66, 0.78, 0.87, 0.91 in knots, filaments, sheets, voids, respectively. To make it more clear, we redo this analysis in each stellar mass bin and subtract the direct measurement of the active galaxy fraction from the correspondingly expected active galaxy fraction. The results are presented in Fig. 10. It shows that at masses below 1.8 × 10 10 h −1 M , there are less active galaxies in knots than in filaments, sheets and voids. It is the low fraction of active galaxies in knots that leads to the low sSFR in knots. It also explains the redder color in knots as shown in Fig. 7. Sobral et al. (2011) have also argued that the environment is responsible for star-formation quenching in dense environments. In other words, denser environments increase the possibility of galaxies to become quenched. At high masses, on the other hand, there is instead not much difference in the active fraction in different environments. As a consequence, the cosmic web dependences of the color and sSFR also vanish. At high redshifts, most galaxies are star-forming and there is almost no environmental dependence of the active Knot Filament Sheet Void Figure 11. The deviation of the stellar metallicity from the expected values as a function of stellar mass for central galaxies in different environments. Red, cyan, blue and green curves donate the median values of the deviation in knots, filaments, sheets and voids, respectively. Scatters are for each galaxy with the same color coding as the curves. Errors are generated using the bootstrap method.
Redshifts are indicated at the top right corner of each panel.
fraction: at z > 1, the active fraction approaches 1 in all cosmic web environments. The environmental dependence of the sSFR thus vanishes.
Metallicity vs. Stellar mass
Metal enrichment is one of the most important processes in galaxy evolution which involves the gas cooling, star formation, supernova feedback, etc. Schaye et al. (2015) found the gas and stellar metallicity vs. stellar mass relations in EA-GLE are in broad agreement with observations for galaxies more massive than 10 9 M , though at the low mass end the relations are not as steep as the observed ones. Fig. 11 shows the stellar metallicity vs. stellar mass relation in the different cosmic web environment. At low masses (M ∼ 1.8 × 10 10 h −1 M ) the metallicity is higher in knots than in other environments. This is related to the relation between stellar mass, metallicity and star formation rate, as discovered in previous works (e.g. Ellison et al. 2008;Mannucci et al. 2010;Yates et al. 2012;De Rossi et al. 2015). Specifically, De Rossi et al. (2017) found that in EAGLE simulations the metallicity of low-mass systems decreases with SFR for a given stellar mass. The high metallicity in knots is thus consistent with their low star formation rates as shown in Fig. 8. Such environmental dependence is much weaker at high redshifts.
Different from the relation with sSFR and color, the stellar metallicity vs. stellar mass relation shows a strong dependence on cosmic environments at high masses. The metallicity increases from knots, filaments, sheets towards voids. This is consistent with the increasing central stellar mass-to-halo mass ratios along knots, filaments, sheets and voids, i.e. more metals are generated if more stars are formed. Such dependence persists up to z=3. At even higher redshifts, there are not enough samples to make solid conclusions.
As for the sSFR, we split the galaxies into active and passive subsamples at z = 0. The transition mass at M ∼ 1.8 × 10 10 h −1 M presents itself both for the active and passive galaxies, below which the metallicity is higher in knots while above which the metallicity is lower in knots. The environmental dependence is stronger for passive galaxies than for active galaxies.
The cosmic web dependences of the gas metallicity as a function of stellar mass and redshift are presented in Fig. 13. Since there is very little gas in passive galaxies, here we focus on active galaxies. At z=0, at stellar masses below 1.8 × 10 10 h −1 M the gas metallicity is higher in knots, while at high masses the difference between different web environments disappears. Different from other scaling relations, this cosmic web dependence of gas metallicity gets stronger at higher redshifts for massive galaxies. The gas metallicity is the lowest in knots and gets higher along filaments, sheets and voids. Knot Filament Sheet Void Figure 13. Similar to Fig. 11 but for gas metallicity of active galaxies.
Combined effects of cosmic web and dark matter haloes
We remove the halo effect from the cosmic web dependence in previous sections. However, it is difficult to obtain halo mass observationally. Here we present the apparent cosmic web dependence (including halo effects, hereafter we refer to it as combined dependence) on the scaling relations. Since the cosmic environmental effect is weak at high redshifts for most of the scaling relations, we focus on results at z=0.
The top left panel of Fig. 14 shows the g-r color vs. stellar mass relations for galaxies in different environments. Galaxies are redder in knots compared to other environments at all masses. This is different from the results shown in Fig. 7 that when removing halo effects at stellar mass above 1.8 × 10 10 h −1 M galaxies in knots have similar colors as those in filaments, sheets and voids. Such dependences are also found in the sSFR vs. stellar mass relations (middle left panel). These can be explained by the fact that there are more massive haloes in high density regions and galaxies in Active Fraction Figure 14. The g − r color, sSFR, active fraction, stellar metallicity and star forming gas metallicity as a function of stellar mass for central galaxies in different environments. Red, cyan, blue and green curves donate the median values in knots, filaments, sheets and voids, respectively. Each dot represent an individual galaxy with the same color coding as the curves. Errors are generated using the bootstrap method. massive haloes are usually redder/with lower sSFR. The active fraction increases with stellar mass up to ∼ 3×10 9 h −1 M and decreases towards higher masses. This variations with stellar mass are similar to each other between voids, sheets and filaments. For those in knots, the amplitude is lower, the turn-over mass is higher, and at low masses the slope is steeper.
For a given stellar mass, the halo mass is lower in knots for those with stellar mass below 1.8 × 10 10 h −1 M (Fig. 6). In low mass haloes the feedback is usually more effective and it is thus easier for new formed heavy elements to escape. This compensate with the cosmic web dependence of stellar metallicity as shown in Fig. 11, resulting in the absence of variance between different environments as shown in the right top panel of Fig. 14. At high masses, the stellar metallicity decreases with environmental densities, similar to those without halo effects (Fig. 11). For the gas metallically vs. stellar mass relation (right bottom panel of Fig. 14), there is almost no difference between voids, sheets and filaments. On the other hand, the scaling relation in knots is very different. It increases rapidly with stellar masses below 1.8 × 10 10 h −1 M and also decreases towards high masses.
The absolute values are lower in knots both at low and high masses compared to other environments, but is higher at the turn over mass.
In summary, for the color vs. stellar mass relation and sSFR vs. stellar mass relation, the combined dependence follows the pure cosmic web dependence at low masses quanti-tatively, while at high masses, the combined dependence is stronger. For the stellar metallicity vs. stellar mass, the combined dependence mimic the pure cosmic web dependence at high masses, but vanish at low masses. The gas metallically vs. stellar mass relation behaves significantly different in knots compared to other environments, and the combined dependence deviate from the pure cosmic web dependence at all masses.
CONCLUSIONS
In this paper, we investigate the dependence of galaxy properties on different cosmic web environments and their evolution, using the EAGLE cosmological hydrodynamical simulations. We split the simulation box into 256 3 cells and generate the web elements adopting the web classification method of Hahn et al. (2007). Here we summarize our results as follows.
We find the baryon fraction increases with halo mass in all environments, and the fraction is higher in denser regions, i.e. increasing along the sequence voids, sheets, filaments and knots. This environmental dependence persists up to redshift 6. The cosmic web dependence becomes slightly stronger at higher redshifts up to z ∼ 3 and then becomes weaker towards even higher redshifts. The central and total stellar mass-to-halo mass ratios both peak at halo masses ∼ 10 12 h −1 M . At low masses, more stars are formed in knots than in other web environments, while at high masses, less stars are formed in knots. The cosmic web dependence of the galaxy stellar mass functions is very strong at all redshifts, with the amplitude decreasing along the sequence knots, filaments, sheets and voids. Interestingly, though the average characteristic stellar mass corresponding to L * does not evolve much since z=1, it changes by an order of magnitude going from knots to voids at z =0.
We remove the halo mass dependence and investigate the relation between the cosmic web and various scaling relations for central galaxies. We find a characteristic stellar mass of 1.8 × 10 10 h −1 M , below and above which the cosmic web dependence behaves oppositely. Galaxies with stellar mass below the characteristic mass are redder, with lower active fraction, lower sSFR, and higher stellar metallicity in knots than in voids, while at stellar masses above the characteristic mass the dependences on the cosmic web either reverse or vanish. At low masses, the relatively strong web dependences of the color vs. stellar mass relation and of the sSFR vs. stellar mass relation can be attributed to the cosmic web dependence of the active galaxy fraction, i.e. the active fraction is higher in voids than in knots. For active galaxies, the stellar metallicity is higher in knots compared to other web environments for those with stellar mass below the characteristic mass.
The cosmic web dependences are weaker at higher redshifts for almost all the galaxy properties and scaling relations we explored, including the central (total) stellar-tohalo mass ratio, color vs. stellar mass relation, sSFR vs. stellar mass relation and stellar metallicity vs. stellar mass relation. But this is not the case for the gas metallicity vs. stellar mass relation. For galaxies above the characteristic stellar mass, this cosmic web dependence gets even stronger at high redshifts, decreasing along the sequence voids, sheets, filaments and knots.
The combined halo + cosmic web dependence follow the cosmic web dependence at low masses for the color/sSFR vs. stellar mass relations, but is stronger at high masses. It reverses for the stellar metallicity at high masses, the combined dependence mimic the cosmic web dependence, while at low masses, the combined dependence almost vanish. One can not use the combined dependence of gas metallicity relation vs. stellar mass relation to explore the pure cosmic web dependence at all for they behave different at all masses.
DATA AVAILABILITY
The data presented in this article are available at the EA-GLE simulations public database (http://icc.dur.ac.uk/ Eagle/database.php). | 2020-08-20T10:12:39.688Z | 2020-08-20T00:00:00.000 | {
"year": 2020,
"sha1": "575809c940ce915a845d198f3bd6a9983bda95c1",
"oa_license": null,
"oa_url": "http://dro.dur.ac.uk/31738/1/31738.pdf",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "e4f393e12ac8b9d38bf654856554c19f56dedad5",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
24646514 | pes2o/s2orc | v3-fos-license | Identifying Speakers Using Their Emotion Cues
This paper addresses the formulation of a new speaker identification approach which employs knowledge of emotional content of speaker information. Our proposed approach in this work is based on a two-stage recognizer that combines and integrates both emotion recognizer and speaker recognizer into one recognizer. The proposed approach employs both Hidden Markov Models (HMMs) and Suprasegmental Hidden Markov Models (SPHMMs) as classifiers. In the experiments, six emotions are considered including neutral, angry, sad, happy, disgust and fear. Our results show that average speaker identification performance based on the proposed two-stage recognizer is 79.92% with a significant improvement over a one-stage recognizer with an identification performance of 71.58%. The results obtained based on the proposed approach are close to those achieved in subjective evaluation by human listeners.
Introduction
Speaker recognition focuses on extracting, characterizing and recognizing the information in speech signals conveying speaker identity. There are two kinds of speaker recognition: speaker identification and speaker verification (authentication). Speaker identification decides who is speaking from a set of known speakers, whereas speaker authentication determines whether a speaker belongs to a particular known voice or to some other unknown voice. Speaker recognition is divided into two sets: "open set" and "closed set". In the "open set", a reference model for the unknown speaker may not exist. However, in the "closed set", a reference model for the unknown speaker should be available to the system. Speaker recognition typically functions in one of two styles: textdependent (fixed-text) style or text-independent (free-text) style. Text-dependent requires a user to regenerate utterances containing the same text. In the textindependent, there is no prior knowledge of the text to be spoken.
It is well known that speaker recognition performance is almost ideal in neutral talking environments [1], [2], [3], [4]; on the other hand, the performance is sharply degraded in emotional talking environments [5], [6], [7], [8], [9]. In this work, we address the issue of enhancing the degraded speaker identification performance in emotional talking environments by proposing, implementing and testing a new approach. This approach is based on identifying the unknown speaker using his/her emotion cues.
Motivation
In literature, there are some studies that focus on the field of speaker recognition in emotional talking environments. Bao et al. focused in one of their studies on emotion attribute projection for speaker recognition on emotional speech [5]. Wu et al. investigated the rules based feature modification for robust speaker recognition with emotional speech [6]. Li et al. proposed an approach of speech emotion-state conversion to enhance speaker recognition performance in emotional talking environments [7]. Shahin focused in two of his earlier studies on using emotions to identify speakers (emotion-dependent speaker identification) [8] and on speaker identification in emotional talking environments [9]. In the first study, he achieved an average speaker identification performance of 78.8% (in a closed set with forty speakers and six emotions) [8]. In the second study, he obtained an average speaker identification performance of 61.4%, 66.4% and 69.1% based on, respectively, hidden Markov models, second-order circular hidden Markov models and suprasegmental hidden Markov models using forty speakers and five emotions [9].
The contribution of this work is positioned on proposing, implementing and evaluating a new approach based on a two-stage recognizer in emotional talking environments. The two-stage recognizer combines and integrates both emotion recognizer and speaker recognizer into one recognizer in a closed set of textindependent using both HMMs and SPHMMs as classifiers. The aim of the new proposed approach is to alleviate the deteriorated speaker identification performance in such talking environments. The emotional talking environments in this work consist of six emotions including the neutral state. These emotions are neutral, angry, sad, happy, disgust and fear. The applications of speaker identification in emotional talking environments appear in criminal investigations to identify the suspected persons who produced voice in such talking environments and in Text-To-Speech (TTS) communication-aid that can help expressing the emotions of the speaker. Traditional techniques that convert text to speech result in a dry message. However, adding emotion (e.g. happy emotion or angry emotion) to the read text would result in a more realistic and live message.
The structure of the paper is as follows. The overviews of SPHMMs are given in the next section. Section 4 describes the speech database used to assess the proposed approach. Section 5 is committed to discussing the proposed approach and the experiments. Section 6 discusses the results achieved in this work.
Concluding remarks are given in Section 7.
Suprasegmental hidden Markov models have been developed, implemented and tested for speaker identification in shouted talking condition [15] and in emotional talking environments [9]. SPHMMs have proven to be superior models over HMMs for speaker recognition systems in each of the emotional and shouted talking environments [9], [15].
Several states of HMMs can be encapsulated into a new state called suprasegmental state. Suprasegmental state can look at the observation sequence through a larger window. Such a state permits observations at proper rates for the situation of modeling. As an example, prosodic information can not be observed at a rate that is used for acoustic modeling. The most important acoustic parameters that express prosody are fundamental frequency, intensity and duration of speech signals [16]. The prosodic features of a unit of speech are labeled suprasegmental features because they have impact on all the segments of a speech signal unit. Therefore, prosodic events at the stages of phone, syllable, 6 word and utterance are represented using suprasegmental states; on the other hand, acoustic events are represented using conventional hidden Markov states.
Prosodic information can be combined with acoustic information within HMMs [17]. The following formula shows how to perform this combination, where is a weighting factor. When: v : is the acoustic model for the vth emotion.
v : is the suprasegmental model for the vth emotion.
O: is the observation vector or sequence of an utterance.
: is the probability of the vth HMM emotion model given the : is the probability of the vth SPHMM emotion model given the observation vector O. Refs. [9] and [15] have more information about suprasegmental hidden Markov models.
Speech Database
A total of twenty five male speakers and twenty five female speakers were asked to generate the speech database used to evaluate the new proposed approach. All the speakers were healthy adult native speakers of American English. The speakers were asked to portray eight sentences nine times each under each of the neutral, angry, sad, happy, disgust and fear emotions. The first four sentences were used in the training stage, while the last four sentences were used in the test stage (text-independent experiment). The total number of utterances was 21600 (50 speakers times 8 sentences times 9 utterances/sentence times 6 emotions). The eight sentences were unbiased towards any emotion (no correlation between any sentence and any emotion). The eight sentences are:
1) He works five days a week.
2) The sun is shining.
3) The weather is fair. 4) The students study hard.
5) Assistant professors are looking for promotion. 6) University of Sharjah. 7) Electrical and Computer Engineering Department. 8) He has two sons and two daughters.
A speech acquisition board with a 16-bit linear coding A/D converter and a sampling rate of 16 kHz was used to capture the speech database in an uncontaminated environment. The database was a 16-bit per sample linear data.
The speech signals were applied every 5 ms to a 30 ms Hamming window.
In this work, Mel-Frequency Cepstral Coefficients (MFCCs) have been adopted as the features to represent the phonetic content of speech signals. MFCCs have been employed in the areas of speech recognition in stressful talking environments and speaker recognition in stressful talking environments because such coefficients 8 outperform other features in the two areas and because they offer a high-level approximation of human auditory perception [5], [18], [19]. In each of HMMs and SPHMMs, a 16-dimension feature analysis of MFCC was used to construct the observation vectors. The number of conventional states, N, was nine and the number of suprasegmental states was three (each three conventional states were combined into one suprasegmental state) in SPHMMs and a continuous mixture observation density was selected for each model. The number of mixture components, M, was ten per state.
In the last four decades, the majority of research carried out in the fields of speech recognition and speaker recognition on HMMs have been done using left-to-right hidden Markov models (LTRHMMs) because phonemes follow strictly the left to right sequence [11], [20], [21]. In this work, left-to-right suprasegmental hidden Markov models (LTRSPHHMs) have been derived from LTRHMMs. Figure 1 shows an example of a basic structure of LTRSPHMMs that has been derived from LTRHMMs. In this figure, q 1 , q 2 ,…, q 6 are hidden Markov states. p 1 is a suprasegmental state that consists of q 1 , q 2 and q 3 . p 2 is a suprasegmental state that is made up of q 4 , q 5 and q 6 . p 3 is a suprasegmental state that is composed of p 1 and
Emotion-Dependent Speaker Identification Approach and the Experiments
Given n speakers talking in m emotions, the proposed architecture is composed of two cascaded stages as illustrated in Figure 2. This figure shows that emotiondependent speaker identification recognizer is nothing but a two-stage recognizer that integrates and combines both emotion recognizer and speaker recognizer into one system. The two stages are:
Stage a: Emotion Recognizer
First, the emotion of the unknown speaker is identified (emotion identification problem). In this stage, m probabilities are computed based on SPHMMs and the maximum probability is chosen as the identified emotion as given in the following formula, where, E * : is the index of the identified emotion.
O: is the observation sequence of the unknown emotion that belongs to the unknown speaker.
: is the probability of the observation sequence O of the unknown emotion that belongs to the unknown speaker given the eth SPHMM emotion model.
In this stage, the eth SPHMM emotion model has been obtained in the training session for every emotion using the fifty speakers uttering all the first four sentences of the database (text-independent) with a repetition of nine utterances/sentence. The total number of utterances used to derive each SPHMM emotion model in this session is 1800 (50 speakers times 4 sentences times 9 utterances/sentence). The training session of SPHMMs is very similar to the training session of the conventional HMMs. In the training session of SPHMMs, suprasegmental models are trained on top of acoustic models of HMMs. A block diagram of this stage is shown in Figure 3. Derivation of the eth SPHMM emotion model in this training session is illustrated in Figure 4.
Stage b: Speaker Recognizer
Given that the emotion of the unknown speaker was identified, the next stage is to identify the unknown speaker (emotion-specific speaker identification problem).
In this stage, n probabilities per emotion are computed based on HMMs and the maximum probability is chosen as the identified speaker for the recognized emotion as given in the following formula, where, S * : is the index of the identified speaker. The sth HMM speaker model has been derived using nine utterances per sentence (the first four sentences of the database). The total number of utterances used to build each emotion-dependent HMM speaker model is 36 (4 sentences times 9 utterances/sentence). Derivation of the sth HMM speaker model for every emotion in this training session is shown in Figure 5.
In the test or identification session (completely separate from the training session), each one of the fifty speakers used nine utterances per sentence (the last four sentences of the database) under each emotion including the neutral state.
The total number of utterances used in this session is 10800 (50 speakers times 4 sentences times 9 utterances/sentence times 6 emotions). This stage can be shown in the block diagram of Figure 6.
Results and Discussion
In this work, emotion cues have been used to identify the unknown speaker in emotional talking environments in order to improve the deteriorated speaker identification performance. Table 1 a) The most easily recognizable emotion is neutral (94%). Therefore, the expected highest speaker identification performance will occur when speakers speak in a neutral state.
b) The least easily recognizable emotion is angry (78%). Consequently, the predicted least speaker identification performance will happen when speakers speak in an angry emotion. c) Column 3 (angry column), for example, shows that 4% of the utterances that were portrayed in an angry emotion were evaluated as generated in a neutral state, 5% of the utterances that were uttered in an angry emotion were identified as produced in a sad emotion. This column shows that angry emotion has the highest confusion percentage with disgust emotion (10%). Therefore, angry emotion is highly confusable with disgust emotion. This column also shows that angry emotion has no confusion with happy emotion (0%). Table 2 shows speaker identification performance based on emotion-dependent speaker identification approach (two-stage recognizer) using SPHMMs when = 0.5, while Table 3 gives speaker identification performance based on emotionindependent speaker identification approach (one-stage recognizer) using the same models. The average speaker identification performance based on the twostage recognizer using SPHMMs is 79.92%, while the average speaker identification performance based on the one-stage recognizer using the same models is 71.58%.
A statistical significance test has been carried out to investigate whether speaker identification performance differences (speaker identification performance based on the two-stage recognizer and that based on the one-stage recognizer in emotional talking environments) are real or simply due to statistical fluctuations. 13 The statistical significance test has been performed based on the Student t Distribution test as given by the following formula, SD two-stage : is the standard deviation of the second sample (two-stage recognizer) of the same size.
Based on Table 2 and Table 3 . Based on these values, the calculated t value is t two-stage, one-stage = 6.093. This calculated t value is much greater than the tabulated critical value at 0.05 significant level t 0.05 = 1.645. Therefore, it is evident from Table 2 and Table 3 that the two-stage recognizer is superior to the one-stage recognizer for speaker identification in emotional talking environments. 14 The achieved speaker identification performance based on the proposed approach has been compared with the results obtained in some previous studies. The proposed approach in the current work yields better speaker identification performance than that reported in some prevoius studies: 1) Emotion-independent speaker identification performance attained by Shahin.
Shahin obtained in one of his studies an average speaker identification performance of 69.1% in emotional talking environments based on HMMs [9]. Hence, it is evident that inserting an emotion identification stage into emotion-independent speaker identification system significantly enhances speaker identification performance in such talking environments. iii. Emotion-state conversion approach. This approach was proposed to enhance speaker identification performance in emotional talking conditions. They achieved 70.22% as an average speaker identification performance based on their proposed approach.
Speaker identification performance using his/her emotions based on the proposed approach is limited as shown in Table 2. This table is the resultant of both stage a and stage b. The reasons of limitations are: i. The unknown emotion that belongs to the unknown speaker in stage a is not perfectly identified. The emotion identification performance of this stage as calculated from Table 1 is 83.83%.
ii. The unknown speaker in stage b is not 100% correctly identified. Table 4 yields speaker identification performance based on using HMMs in both stages of the proposed recognizer. The average speaker identification performance based on Table 4 is 75.92% (with a standard deviation of 6.44). To make a comparison between a two-stage speaker identification performance based on SPHMMs and that based on HMMs, the t two-stage (SPHMMs), two-stage (HMMs) has been calculated. The calculated t value is t two-stage (SPHMMs), two-stage (HMMs) = 3.206. This calculated t value is greater than the tabulated critical value t 0.05 = 1.645. Therefore, the conclusion that can be drawn in this experiment shows that suprasegmental hidden Markov models outperform hidden Markov models for speaker identification in emotional talking environments based on the two-stage recognizer. Table 5.
Speaker identification performance based on the proposed two-stage approach using SPHMMs when = 0.5 and using Emotional Prosody database is given in Table 6. Based on this table, the average speaker identification performance is 78.92%. It is evident from Table 2 and Table 6 that the two average speaker identification performances are close to each other.
3) Experiment 3: The proposed two-stage recognizer has been assessed for different values of the weighting factor (. Figure 7
Concluding Remarks
In the present work, we proposed, applied and tested a new approach based on using emotion cues to enhance speaker identification performance in emotional talking environments. This work showed that the significant improvement of speaker identification performance using the proposed two-stage recognizer over Table 4 Speaker identification performance based on the two-stage recognizer using Table 6 Average speaker identification performance based on the proposed two-stage approach using SPHMMs when = 0. 5 | 2018-01-22T11:33:50.000Z | 2011-06-01T00:00:00.000 | {
"year": 2018,
"sha1": "28d62a9e91c00ee1973851f8f967261ce6fe49c9",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "28d62a9e91c00ee1973851f8f967261ce6fe49c9",
"s2fieldsofstudy": [
"Business"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
221200452 | pes2o/s2orc | v3-fos-license | Association between a DJ-1 polymorphism and the risk of Parkinson's disease: a PRISMA-compliant systematic review and meta-analysis
Objective In recent years, a number of case–control studies have focused on the association between the DJ-1 g.168_185del polymorphism and the risk of Parkinson's disease (PD). However, the results have been conflicting. To estimate the relationship between the DJ-1 g.168_185del polymorphism and PD susceptibility, a comprehensive meta-analysis was performed. Methods Eligible studies concerning the DJ-1 g.168_185del polymorphism and PD susceptibility were searched for in the PubMed, Web of Science, Embase, Wanfang, CNKI, and VIP databases. Odds ratios and 95% confidence intervals were calculated to estimate the strength of the associations. In total, 11 studies were included in this meta-analysis, including 13 case–control studies with 2890 cases and 3043 controls. Results This meta-analysis revealed that DJ-1 g.168_185del variants are associated with PD susceptibility in the non-Asian population, but not in the Asian population. Conclusions Our meta-analysis suggests that DJ-1 gene variants are not associated with the risk of PD in the overall population.
Introduction
Parkinson's disease (PD) is one of the most common neurodegenerative movement disorders worldwide, affecting more than 1% of the population over 65 years of age. 1 It is characterized by variable combinations of bradykinesia, rigidity, resting tremors, and postural abnormalities. 2 In the human brain, PD is pathologically typified by the degeneration of dopaminergic neurons and the presence of Lewy bodies. 3,4 The etiology of PD remains unclear. However, genetic factors such as PARK16, 5 SNCA, 6 or VPS13C variants, 7 as well as certain environmental factors, have been shown to contribute to the increased risk of PD.
In humans, the DJ-1 (PARK7) gene is located on chromosome 1p36. It contains eight exons, spanning 24 kb, and encodes a protein consisting of 189 amino acids that belongs to the ThiJ/PfpI superfamily. 8,9 Oxidative stress and mitochondrial damage reportedly play important roles in the pathology of PD. 10 Notably, DJ-1 is considered to play a key role in protecting neurons from oxidative stress and mitochondrial damage. 11,12 DJ-1 is also regarded as a chaperone, protease, and oncogene in glial cells and neurons of the substantia nigra and striatum. [13][14][15][16][17] With regard to the DJ-1 gene g.168_ 185del polymorphism, numerous casecontrol studies have estimated the association between this polymorphism and PD susceptibility. [18][19][20][21][22][23][24][25][26][27][28] However, results have been conflicting. We therefore performed a comprehensive meta-analysis in the present study, to clarify the relationship between the DJ-1 gene g.168_185del polymorphism and PD risk.
Materials and methods
The present meta-analysis was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. 29 Literature and search strategy Two investigators (J Liu and WS Deng) identified all studies with a focus on the association between the DJ-1 g.168_185del polymorphism and PD risk using the PubMed, Web of Science, Embase, Wanfang, CNKI, and VIP databases, dating to March 15 2019. The keywords "(DJ-1) and (polymorphism) and (Parkinson's disease or PD or parkinsonism)" were used to search within the English electronic databases, and the Chinese electronic databases were searched using the corresponding Chinese characters. Only studies published in English or Chinese were included in this metaanalysis. The two investigators also reviewed the references of the relevant and included studies to identify any additional studies.
Inclusion and exclusion criteria
The published studies adhered to the following inclusion criteria: (1) focused on the association between the DJ-1 g.168_185del polymorphism and PD susceptibility; (2) case-control or cohort studies; (3) provided the genotype distributions of the cases and controls, so that odds ratios (ORs) and 95% confidence intervals (CIs) could be calculated; (3) published in English or Chinese; (4) the genotype distribution of the controls was consistent with the Hardy-Weinberg equilibrium (HWE). Reviews, abstracts, case reports, and duplicate reports were excluded.
Data extraction and quality assessment
For each eligible study, the following information was independently extracted by two investigators: first author, publication year, region or country, HWE in controls, sample size, and the numbers of case and controls. Any disagreements between the two investigators were resolved through discussion. In addition, the Newcastle-Ottawa Scale (NOS) was used to estimate the quality of eligible studies; 30 a score above 5 was considered to be of moderate-to-high quality.
Statistical analysis
The strength of the association between the DJ-1 gene g.168_185del polymorphism and PD susceptibility was assessed using ORs and their 95% CIs under the five genetic models (allelic model: D vs. I, dominant model: DDþDI vs. II, homozygous model: DD vs. II, heterozygous model: DI vs. II, and recessive model: DD vs. DIþII). The degree of heterogeneity among the included studies was determined using the Q test and inconsistency index (I 2 ) statistics (no heterogeneity: I 2 < 25%, moderate heterogeneity: I 2 ¼ 25% to 50%, significant heterogeneity: I 2 > 50%). 31 If I 2 > 50%, the random-effect model was used to calculate the OR and 95% CI; otherwise, the fixed-effect model was adopted. 32 To assess the reliability of the present metaanalysis, a sensitivity analysis was performed according to the leave-one-out method. 33 Begg's funnel plots and the Begg's test were conducted to evaluate potential publication bias. Here, an asymmetrical funnel plot and a Begg's test P-value of < 0.05 implied potential publication bias. 34 The Hardy-Weinberg equilibrium (HWE) was estimated using the chi-squared test in the genotype distributions of the control groups. 35 In addition, a subgroup analysis was performed based on ethnicity (Asian and non-Asian) within the overall population. All statistical analyses in this study were performed using Stata software, version 12.0 (Stata Corporation, College Station, TX, USA).
Association of the DJ-1 gene polymorphism with PD susceptibility
All eligible studies (including 2890 cases and 3043 controls) were used to estimate the association between the DJ-1 gene g.168_185del polymorphism and PD susceptibility. The pooled ORs and their 95% CI are summarized in Table 2. There were no significant associations in the overall population in any of the five models ( Figure 2 and Table 2). The fixed-effect model was used in all genetic models ( Table 2). In the subgroup analysis by ethnicity (Asian and non-Asian), the DJ-1 gene g.168_185del polymorphism was associated with a significantly increased risk of PD in the non-Asian population, but not in the Asian population (Table 2).
Sensitivity analysis and publication bias
A sensitivity analysis was performed to assess the influence of each study on the pooled ORs and 95% CIs by omitting each study in turn. There were no significant changes in the pooled ORs or 95% CIs in the dominant model (Figure 3), indicating the stability of the present meta-analysis. Begg's funnel plot and Begg's test were used to assess the publication bias of the included case-control studies. The shapes of the funnel plots were roughly symmetrical (Figure 4), and Begg's test revealed no significant publication bias in this metaanalysis (DDþDI vs. II: P ¼ 0.161).
Discussion
PD is a common neurodegenerative movement disorder in individuals over 65 years of age. In recent years, numerous case-control studies have focused on the relationship between the DJ-1 gene g.168_185del polymorphism and PD susceptibility. However, the results of these studies have been inconsistent. To assess if any such association exists, we performed a comprehensive meta-analysis of 11 studies, including 13 case-control studies with 2890 cases and 3043 controls. Variants in promoter regions are involved in gene transcription activity because of the DNA-binding ability of transcription factors. Siegel et al. 36 reported that the Ins allele of g.168_185del variants might affect the transcriptional activity of DJ-1 by binding to nuclear factors. The DJ-1 protein, which was initially identified almost a century ago, is expressed in many different tissue types. 14,37,38 DJ-1 is considered to play a key role in protecting neurons from oxidative stress and mitochondrial damage. 11,14 In addition, DJ-1 gene polymorphisms are closely associated with autosomal recessive early-onset PD. 9,39 Moreover, DJ-1 can eliminate hydrogen peroxide by undergoing selfoxidation. In doing so, reactive oxygen species are decreased. 12 DJ-1-related oxidative damage is reportedly evident within the brains of sporadic PD patients. 40,41 In addition, Waragai et al. 42 reported that DJ-1 levels in the cerebrospinal fluid of sporadic PD patients are significantly higher than in healthy controls. Furthermore, DJ-1 was found to promote the expression of antiapoptotic genes and suppress apoptosisassociated pathways. [43][44][45][46] Overall, it seems that DJ-1 is associated with PD, although one study reported no significant difference between PD patients and controls in serum levels of DJ-1 protein. 47 In the present study, the DJ-1 gene g.168_185del polymorphism was associated with an increased risk of PD under both the allelic and dominant models, based on data collected from 13 case-control studies (2890 cases and 3043 controls in the overall population). No significant heterogeneity was present in the five models that were used (Table 2). Additionally, in the subgroup analysis, we revealed that DJ-1 gene g.168_185del variants were not associated with PD susceptibility among the Asian population (Table 2). However, the DJ-1 gene g.168_185del polymorphism was associated with an increased risk of PD among the non-Asian population (Table 2). Furthermore, no obvious publication bias was detected in this meta-analysis (DDþDI vs. II: P ¼ 0.161). In summary, this comprehensive meta-analysis indicated that the DJ-1 gene g.168_185del polymorphism is associated with an increased risk of PD in the non-Asian population, but not in the Asian population.
Several potential limitations exist in this meta-analysis. First, the included studies were published only in Chinese or English. Thus, potential publication bias may exist. Second, only four studies focused on the associations between the DJ-1 gene g.168_185del polymorphism and PD risk among Caucasians, while one study 25 targeted mixed and black populations. Hence, a sub-group analysis was conducted based on ethnicity (Asian and non-Asian). Third, we were unable to test environmental factors and gene-gene or gene-environment interactions because insufficient information was collected.
Conclusion
This comprehensive meta-analysis of 13 case-control studies demonstrated that the DJ-1 gene g.168_185del polymorphism is associated with PD susceptibility among the non-Asian population, but not among the Asian population. However, considering the limitations of this study, the present results should be interpreted with caution. Further studies with larger sample sizes and diverse ethnic groups should be conducted to validate the resulting associations.
Declaration of conflicting interest
The authors declare that there is no conflict of interest.
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. | 2020-08-21T13:01:51.536Z | 2020-08-01T00:00:00.000 | {
"year": 2020,
"sha1": "0d09f81a11d32f5967276b7d601f748f121d60d4",
"oa_license": "CCBYNC",
"oa_url": "https://journals.sagepub.com/doi/pdf/10.1177/0300060520947943",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "0b18db2437f29364011e41084d7c9e5cff8baa53",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
83143944 | pes2o/s2orc | v3-fos-license | Cotton fertilization using PGPR Bacillus amyloliquefaciens FZB42 and compost: Impact on insect density and cotton yield in North Benin, West Africa
Abstract This work has compared the effects of the biofertilizer Bacillus amyloliquefaciens FZB42 with that of compost for cotton production. The population dynamics of pests and predators have been studied in order to check whether the use of both fertilization materials can contribute to pest management in cotton. Three treatments were considered: (i) dressing of seeds in rhizobacteria suspension, (ii) introduction of rhizobacterial suspension directly in the pocket, same time with the seeds, and (iii) fertilization with compost. The study was carried out in northwest Benin (West Africa). Results showed that cotton aphids, Aphis gossypii, pink bollworm, Pectinophora gossypiella, leaf roller, Sylepta derogata, and cotton bugs, Dysdercus sp. are the major insect pests encountered in the experimental plots. Cotton bollworm, Helicoverpa armigera, was present but under the economic threshold. The coccinellid predators, Cheilomenes spp., occurred in the experimental plots and almost suppressed aphid proliferation. Other natural enemies such as chrysopids and ant species also occurred and probably contributed to maintain the cotton bollworm under the economic threshold. The treatment with seeds dressed with the rhizobacteria suspension yielded 39% more cotton compared to the compost fertilization. The use of both fertilization materials without application of chemicals can contribute to pest management in cotton.
PUBLIC INTEREST STATEMENT
Intensive use of mineral fertilizers provokes soil erosion, water pollution as well as plant susceptibility to insect pests. Cotton production loss attributable to pests in West Africa averages about 40% of the potential yield and may reach 70% in some areas. It is therefore essential to look for an improved cotton production system that favors the activity of beneficial insects and limits the proliferation of insect pests. The present article has evaluated whether the use of compost combined with the plant growthpromoting rhizobacterium (PGPR) Bacillus amyloliquefaciens FZB42 can improve the cotton yield and simultaneously contribute to insect pest management in cotton.
Introduction
In West Africa, plants are cultivated on tropical soils that are very poor in nutritious mineral elements such as nitrogen, potassium, and phosphorus. Growers should use up to 250-kg/hectare mineral fertilizers every year. Despite intensive mineral fertilization, crops' yield remains very low (Bationo et al., 2006). Moreover, this intensive use of mineral fertilizers provokes soil erosion, water pollution as well as plants' susceptibility to insect pests (Maltais, 1951;Strong, Lawton, & Southwood, 1984). Cotton production loss attributable to pests in West Africa averages about 40% of the potential yield and may reach 70% in some areas. Four groups of pests are found on cotton in West Africa: bollworms, leaf-eating caterpillars, mites, and sucking insects (Silvie, Deguine, Nibouche, Michel, & Vaissayre, 2001;Vaissayre & Cauquil, 2000;Vaissayre & Deguine, 1996).
Management guidelines for the cotton pest complex in West Africa have been designed to provide growers with simple and inexpensive calendar-based intensive chemical spraying recommendations. This intensive application of chemical insecticides has led to the development of resistance to the major chemical families of insecticides (Martin, Ochou, Hala-N'Klo, Vassal, & Vaissayre, 2000). It is therefore essential to look for an improved cotton production system that favors the activity of natural enemies and limits the proliferation of insect pest populations.
The influence of the PGPR FZB42 seed dressing combined with a low and full mineral nitrogen plant fertilization on the cotton growth and yield was tested in field experiments. Results showed that the cotton yield in the combined application of PGPR and low nitrogen amount increased by 75% (Monir, Bochow, & Junge, 2012).
The present investigation has compared the effects of PGPR B. amyloliquefaciens FZB42 with that of compost for organic cotton production. The population dynamics of insect pests and predators have been studied to determine whether the use of both fertilization materials can contribute to pest management in cotton.
Application of the PGPR B. amyloliquefaciens FZB42 and compost
The experiment was carried out in the district of Tanguiéta, near National Pendjari Parc, northwest Benin, where only organic cotton is produced.
Cotton seed variety H279-1 was used. Three treatments were tested: T1-seed dressing with rhizobacteria suspension (without compost), T2-pouring rhizobacteria suspension directly into the pocket, simultaneously with the seeds (without compost), and T3-fertilization with compost alone (according to the current recommendation of organic cotton fertilization). Each treatment was replicated five times on plots measuring 115 m 2 .
The liquid formulation of PGPR FZB42 (RhizoVital®, ABiTEP GmbH, Germany) was used for cotton seed dressing as follows: 20-ml product (bacterial concentration of about 10 10 cells per ml) was diluted with 10-L water to get a bacterial suspension with 2 × 10 7 cells/ml in which 2-kg cotton seeds were dipped and stirred for 15 min. Thereafter, treated seeds were removed, spread in a thin layer on paper, air dried, and sown (Treatment T1). For Treatment T2, 1.5-ml bacterial suspension (2 × 10 7 cells/ ml) was poured immediately into the pocket where non-treated cotton seeds were sown. Regarding treatment T3, compost was applied, twice during the experiment (on day 30 and 76 after sowing), respectively, 0.4 and 0.2-kg compost was applied per m² (corresponding to four and two tons of compost per hectare, respectively). The compost has been applied according to the organic cotton production guidelines for West Africa (Ouédraogo, Yombi, Doumbia, Eyhorn, & Dischl, 2008).
Assessment of insect population dynamics and pest damages to cotton bolls in tested treatments
The diversity and populations dynamics of all cotton insect pests and predators were studied. We counted the number of insects on 10 plants chosen in the 2 diagonals of each plot, every 3 days. The observations started on day 30 after sowing and continued until the end of the season.
Damage of cotton bollworms and other pests was assessed by counting injured as well as safe bolls on 20 plants chosen in the 2 diagonals of each plot, every week.
Impact of the fertilization materials on cotton yield
To assess the impact of tested treatments on the cotton yield, we harvested cotton on two lines located in the center of each experimental plot and the yield per hectare was calculated.
Statistical analyses
We performed analysis of variance (one-way ANOVA) for data that met the ANOVA hypotheses and non-parametric tests (Mann-Whitney test) for the remaining data. SPSS, Statistics Package version 16.0, was used for these analyses.
At beginning of the season, from 15 August to 12 September, aphid populations were abundant in all experimental plots. After this time period, aphid population density significantly decreased and was about only one individual per plant by 2 November. A slight increase of population density was, however, noticed in November, but this increase did not reach the initial density level (Figure 1). There was no significant difference between the tested treatments (p > 0.05).
The aphid predator, Cheilomenes spp., has been encountered in all experimental plots; their population density was also abundant at the beginning of the season. At this time, we counted about 4 individuals of coccinellid predators for 10 plants in all experimental plots. These beneficial organisms maintained themselves in all plots until the end of the season, but their density was less abundant compared to the beginning of the season (Figure 2). There was no significant difference between the tested treatments (p > 0.05).
Cotton leaf roller, S. derogata, appeared in all experimental plots, but this pest was more abundant in plots that received seeds soaked in PGPR, from 24 September to 29 October. In this plot, the number of leaf roller reached 12 for 10 plants against only 2 individuals for both "compost" and "PGPR in pockets" treatments ( Figure 3).
Pink bollworm, P. gossypiella, appeared in the plots by 16 October; their population density slowly increased and reached, respectively, 17, 21, and 15 individuals for 10 plants, in "seeds dressed in PGPR," "PGPR in pockets," and "compost" treatments. Its density decreased, however, starting from 24 November (Figure 4). There was no significant difference between the tested treatments (p > 0.05).
Regarding the cotton bug, Dysdercus völkeri, it appeared in the experimental plots by 13 October and was abundant until 15 November, after which its density decreased ( Figure 5). There was no significant difference between the tested treatments (p > 0.05).
Discussion
The cotton aphid, A. gossypii, had infested plants at the beginning of the season, from 15 August to 12 September; aphid populations were abundant in all experimental plots. After this time period, aphid population density significantly decreased and was about only one individual per plant by 2 November. The decrease in aphid populations coincided with the occurrence of the coccinellid predator Cheilomenes sulfurea, in the experimental plots. We think that this spectacular decrease in the aphid population density is attributable to the occurrence of the coccinellid species. In fact, insects belonging to Cheilomenes spp. are known as major predators of A. gossypii. Laboratory studies conducted to find out the consumption rate and biology of coccinellid predators Cheilomenes spp. revealed that their per day predation rate (number of aphids) ranged from 35 to 57 (Pandi, Bishwajeet, Shah, & Shankarganesh, 2012;Prabhakar & Roy, 2010). Numerous studies have also demonstrated the predation ability of Cheilomenes spp. against aphids (Hodek, Chakrabarti, & Rejmanek, 1984;Ofuya, 1986Ofuya, , 1995Omkar & Bind, 2004). We therefore conclude that the coccinellid predators significantly contributed to the protection of cotton plants against aphids since it was not necessary to apply any other control measure. Other factors may explain the limited aphid proliferation in our experimental plots. In general, nitrogen concentration in plant tissues increases with mineral N-fertilization (Carrow & Betts, 1973). Nitrogen is the principal component of plant tissues which is vital for phytophagous insects such as aphids (Kytö, Niemelä, & Larsson, 1996); their development and proliferation is strictly correlated to nitrogen concentration in plant tissues (Mattson, 1980). Therefore, the fact that no mineral fertilizer has been used in the present study may also contribute to the reduced abundance of aphids in the experimental plots.
The cotton bollworm, Helicoverpa armigera, (Lepidoptera: Noctuidae), the cotton leafworm Spodoptera littoralis (Lepidoptera: Noctuidae), the leaf roller S. derogata (Lepidoptera: Pyralidae), and the cotton spiny bollworm Earias spp. (Lepidoptera: Noctuidae) are major lepidopteran pests in Northern Benin (Alavo, Yarou, & Atachi, 2010). All these insect pests have been encountered in our experimental plots; however, only a limited number of S. littoralis, Earias spp. and H. armigera occurred. This may be explained by the presence of natural enemies of these pests in the experimental plots. For instance, Chrysopids and ant species, which are known as the most important predators of H. armigera (Alavo, 2006;Mansfield, Elias, & Lytton-Hitchins, 2003;Romeis & Shanower, 1996;van den Berg, Cock, & Oduor, 1997;van den Berg & Cock, 1993, 1995, were occasionally encountered in the plots. We considered that the presence of these beneficial organisms in the plots was due to the fact that these plots have not been treated with any chemicals. We think that the beneficial organisms maintained the population density of H. armigera (the most damaging cotton pest in the region) under the economic threshold (one individual for 40 plants).
Other lepidopteran pests and cotton bugs (whose population density was much more important) have certainly induced important crop damage. S. derogata rolls the leaves into a cone in which it stays until pupation; the cotton strainers Dysdercus spp. attacked green bolls. Nevertheless, the most important damages caused to the bolls came from the pink bollworm, P. gossypiella, which was particularly abundant in our experimental plots during the period of bolls formation. Percent of injured bolls was, respectively, 18, 25, and 15%, for plots receiving seeds dressed in PGPR, PGPR in pockets, and compost. The substantial damage to the bolls is certainly the reason why we obtained, in general, a weak cotton yield in our plots in comparison to the yield obtained usually when insecticides are applied against exocarpic and endocarpic bollworms.
On the other hand, plots receiving seeds dressed with the PGPR yielded 39% more cotton yield in comparison to the compost treatment. This yield increase may be attributed to the beneficial properties of the PGPR which is known to promote plant and root growth. The stronger root system induced by the bacterial-born auxin in bacterized plants leads to an improved uptake of water and nutrients, and hence to faster plant growth and higher stress tolerance against pathogen attacks and dryness. The PGPR Bacillus spp. produces phytase that mobilizes phytate, the organic-bound phosphorus for the plant. Because plants do not form this enzyme, only bacterized plants profit from increased phosphorus nutrition (Bochow et al., 2001;Idris et al., 2002Idris et al., , 2004Kilian et al., 2000;Kloepper, 2004;Koumoutsi et al., 2004). In field experiments, the influence of the PGPR FZB42 (B. amyloliquefaciens) seed dressing combined with different levels of mineral nitrogen plant fertilization on the cotton growth and yield was tested in Egypt. Results showed that the PGPR promoted the plant growth and the cotton yield (in the combined application of PGPR and low nitrogen amount) increased by 75% (Monir et al., 2012). These data confirm that the yield increase obtained in our experimental plot is really the effect of seed dressing with the PGPR B. amyloliquefaciens FZB42. | 2018-12-07T08:16:57.989Z | 2015-07-13T00:00:00.000 | {
"year": 2015,
"sha1": "953dfdc2319eb70933fcab0c583670ccc0eca51d",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.1080/23311932.2015.1063829",
"oa_status": "GOLD",
"pdf_src": "TaylorAndFrancis",
"pdf_hash": "953dfdc2319eb70933fcab0c583670ccc0eca51d",
"s2fieldsofstudy": [
"Agricultural and Food Sciences"
],
"extfieldsofstudy": [
"Biology"
]
} |
25208768 | pes2o/s2orc | v3-fos-license | Novel insights into proteolytic cleavage of influenza virus hemagglutinin
Abstract The influenza virus hemagglutinin (HA) mediates the first essential step in the viral life cycle, virus entry into target cells. Influenza virus HA is synthesised as a precursor protein in infected cells and requires cleavage by host cell proteases to transit into an active form. Cleavage is essential for influenza virus infectivity and the HA‐processing proteases are attractive targets for therapeutic intervention. It is well established that cleavage by ubiquitously expressed subtilisin‐like proteases is a hallmark of highly pathogenic avian influenza viruses (HPAIV). In contrast, the nature of the proteases responsible for cleavage of HA of human influenza viruses and low pathogenic avian influenza viruses (LPAIV) is not well understood. Recent studies suggest that cleavage of HA of human influenza viruses might be a cell‐associated event and might be facilitated by the type II transmembrane serine proteases (TTSPs) TMPRSS2, TMPRSS4 and human airway trypsin‐like protease (HAT). Here, we will introduce the different concepts established for proteolytic activation of influenza virus HA, with a particular focus on the role of TTSPs, and we will discuss their implications for viral tropism, pathogenicity and antiviral intervention. Copyright © 2010 John Wiley & Sons, Ltd.
INTRODUCTION
Infection with influenza viruses, enveloped viruses with a negative stranded, segmented RNA genome, is a significant source of global morbidity and mortality [1]. The ever-changing nature of influenza viruses allows these agents to continuously evade host immunity and to pose a constant threat to human and animal health [1,2]. The acquisition of subtle alterations (antigenic drift), mainly amino acid changes in the viral surface protein hemagglutinin (HA), is responsible for the yearly occurring influenza epidemics (seasonal influenza), which claim 36 000 lives per year in the Unites States alone [3]. The exchange of entire genomic segments (reassortment) between influenza viruses of different subtypes in a co-infected cell can result in the emergence of new viruses (antigenic shift), which, upon introduction into an immunological naïve population, can initiate a pandemic spread (pandemic influenza) with potentially severe medical and social consequences [1,4,5]. It is estimated that up to 50 million people, or about 2% of the world's population, succumbed to the influenza virus circulating in 1918, the causative agent of the disastrous Spanish influenza [6,7]. A distantly related virus, which emerged in Mexico in February 2009 [8,9], is responsible for the first influenza pandemic of the 21st century, with 258 698 laboratory confirmed cases by 5 December 2009 (the number of actual cases is expected to be much higher) [10]. Albeit infection by the 2009 H1N1 virus usually does not cause severe disease in patients without underlying illness, it is notable that the virus shows augmented pathogenicity in animal models compared to contemporary seasonal influenza viruses [11,12], raising concerns that viral variants with
Reviews in Medical Virology
DOI: 10.1002/rmv.657 increased pathogenicity might emerge in the course of the pandemic. Vaccines and antivirals target the viral surface proteins HA (16 subtypes: H1-H16), neuraminidase (NA, 9 subtypes N1-N9) and the ion channel M2; these proteins normally facilitate entry (HA), release (NA) and viral uncoating (M2) in target cells [13][14][15][16]. However, the variability of influenza viruses compromises the antiviral defences directed against the surface proteins. Thus, as a result of antigenic drift, the vaccine needs to be reformulated on an annual basis to provide protection against the viral strains expected to be circulating during the next 'influenza season', and seasonal vaccination will not be effective against newly emerged viruses, like the 2009 H1N1 virus. In addition, viruses frequently acquire mutations, which confer resistance in the molecular targets of M2 and NA inhibitors [17]. Therefore, novel strategies to combat influenza are required, which should be based on a thorough understanding of viral and host cell factors essential for influenza virus spread and pathogenesis. Proteolytic activation of the influenza virus HA by cellular enzymes is indispensable for the virus to transform into an infectious form [18], and the responsible enzymes are potential targets for intervention. However, the nature of the proteases, which activate human influenza viruses, is unclear. Recently, evidence was reported that type II transmembrane serine proteases (TTSPs) might play a major role [19,20]. Here, we will review our knowledge on proteolytic activation of human influenza viruses, with a particular focus on the role of TTSPs, and we will discuss the implications of different modes of HA activation for viral tropism, pathogenicity and antiviral intervention.
INFLUENZA VIRUS HEMAGGLUTININ: THE VIRAL KEY TO THE HOST CELL
The HA mediates influenza virus binding to host cells and fusion of the viral membrane with the limiting membrane of host cell endosomes, thereby providing the virus with access to the host cell interior. These processes are essential for infection, are tightly regulated and involve the formation of several structural HA intermediates, which have been characterised at the atomic level. Here, we will describe how HA brings about membrane fusion and why HA cleavage is essential for its activity.
Hemagglutinin drives fusion of the viral membrane with the membrane of host cell endosomes
Mature HA is composed of two subunits, the globular surface unit HA1 and the stalk-like transmembrane unit HA2, which are covalently linked by a disulfide bond ( Figure 1) [21,22]. The subunits function differently in the entry process: HA1 facilitates binding to receptors on the host cell surface while HA2 drives membrane fusion. Important receptor determinants for influenza viruses are alpha 2-6 (preferred by human viruses) and alpha 2-3 (preferred by avian viruses) linked sialic acids, which are bound by a pocket located in the membrane distal part of HA1 [21,22]. Upon uptake of bound particles, which can proceed via multiple endocytic pathways [23], the low pH environment in endosomal vesicles induces conformational changes in HA2, which facilitate membrane fusion [24]. The successful membrane merger requires the concerted action of several functional elements in HA2, including an Nterminal fusion peptide (FP), the transmembrane domain and a central coiled coil, which is located between the FP and the transmembrane domain [21,22]. The membrane fusion reaction commences with a pH-induced loop-to-helix transition of a sequence located between the FP and the coiled coil. As a consequence of this transition, the hydrophobic FP is propelled towards the target cell and inserts into the cellular membrane [21,22]. Subsequently, the Cterminal extracellular portion of HA2 folds onto the central coiled coil in an antiparallel fashion, forming a stable post-fusion structure in which the FP and the membrane anchor of HA are located in close proximity. As a result of this conformational change, the viral and the target cell membrane are pulled into close contact and ultimately fuse [21,22].
Proteolytic cleavage is essential for hemagglutinin-driven membrane fusion cleaved into HA1 and HA2 by host cell proteases [18,28,29]. Cleavage occurs at a linker sequence connecting the HA1 and HA2 subunits, which is located on a partially surface exposed loop, and proceeds in two steps. First, an endoprotease cleaves HA at the carboxyl terminus of an arginine located at the border between HA1 and HA2 and thereby generates the N-terminus of mature HA2 [28,29]. Subsequently, a carboxypeptidase removes the arginine and thereby creates the C-terminus of mature HA1 [30]. Cleavage is associated with a structural change in HA, during which the FP, probably due to electrostatic interactions, inserts into a conserved cavity formed by residues within HA1 and HA2 [31]. This configuration is essential for fusion, since only the liberated (separated from the C-terminus of HA1) FP is able to insert into the target cell membrane upon a low pH stimulus. Amino acids within the cavity and the FP are essential for membrane fusion and determine at which pH fusion is triggered [27,32], suggesting that the cavity and the FP are potential targets for therapeutic intervention [28].
The sequence of the cleavage site in hemagglutinin is an important determinant of pathogenicity of avian influenza viruses
The nature of the linker sequence critically determines access to proteases, which in turn impacts viral tropism and pathogenicity, at least in the context of avian viruses. Thus, for low pathogenic avian influenza viruses (LPAIV) the linker consists of a single arginine (in some cases lysine), which is accessible to only a limited number of trypsin-like proteases [28,29]. Tissue expression of these proteases is believed to be restricted to the respiratory tract (terrestrial birds) and gastrointestinal tract (water fowl, terrestrial birds). Consequently, viral replication is confined to these compartments. In contrast, the linker sequence in highly pathogenic avian influenza viruses (HPAIV) contains several arginines, with R-X-R/K-R as consensus motif [33][34][35]. This sequence is surface-exposed and readily recognised by furin and PC5/6, members of the proprotein convertase family of eukaryotic subtilisin-like serine endoproteases, which are ubiquitously expressed [36,37]. Therefore, HPAIVs can replicate systemically and cause severe disease. In this context, it needs to be noted that some HPAIVs do not harbour a furin consensus sequence. However, cleavability of these Figure 1. Structural rearrangements associated with influenza virus hemagglutinin cleavage by host cell proteases. The HA precursor HA0 is cleaved into two subunits HA1 (red) and HA2 (blue) by cellular proteases which recognise either a multibasic or monobasic cleavage site located in a loop between HA1 and HA2. Multibasic cleavage sites harbour multiple arginines and/or lysines and are found in the HA-proteins of HPAIV. Monobasic cleavage sites consist of a single arginine (Arg 344 (purple) in case of the 1918 influenza virus) or lysine and are found in the HAproteins of human influenza viruses and LPAIV. Upon proteolytic cleavage, the fusion peptide (green) in HA2 is liberated from HA1 and the HA is primed for activation by low pH, which involves insertion of the fusion peptide into a negatively charged pocket located adjacent to the cleavage site-bearing loop in HA0. The subunits HA1 and HA2 remain covalently linked by a single disulfide bond (yellow), the newly formed C-terminus of HA1 and N-terminus of HA2 are labelled in purple. viruses might be increased due to the absence of a carbohydrate side chain, which would otherwise limit recognition by proteases [38][39][40][41], or due to amino acid substitutions close to the cleavage site, which might increase surface exposure of this sequence and thus accessibility to proteases [21].
Considering the clear correlation between cleavability, viral tropism and pathogenicity observed for avian viruses, a similar interdependence could also be expected for human viruses. However, none of the influenza virus subtypes previously pandemic in humans (H1N1, H2N2 and H3N2) contained a multi-basic cleavage site, including the highly pathogenic 1918 influenza virus [42]. The implications of the cleavage site for spread and pathogenicity of human viruses are thus not overt, and can only be elucidated once the HA-activating proteases have been defined.
Collectively, HA exhibits a particular domain organisation, which reflects the molecular mechanism by which it drives membrane fusion. A similar architecture and membrane fusion mechanism have been identified for other viral glycoproteins, like the HIV envelope protein, which are termed class I fusion proteins [22]. Most class I fusion proteins, including influenza virus HA, require cleavage by host cell proteases to transit into a fusion-competent state. Proteolytic activation of influenza viruses can occur in the Golgi apparatus or at the plasma membrane of infected cells, as well as in the extracellular space and in target cell vesicles, so the nature of the cleavage site and the respective activating proteases have important implications for the biological properties of influenza virus as well as for therapeutic intervention, as discussed below.
PROTEOLYTIC ACTIVATION OF INFLUENZA VIRUS HEMAGGLUTININ: PROTEASES AND CELLULAR COMPARTMENTS HPAIV are activated by furin in the trans-Golgi apparatus
Early, groundbreaking studies on the activation of influenza viruses noted that viral infectivity correlated with the status of HA cleavage and that activation of viruses by trypsin correlated with HA cleavage [43][44][45]. Within these studies it was also observed that some viruses contained cleaved HA and were highly infectious irrespective of the cellular systems used, while HA cleavage and infectivity of others were strongly dependent on the host cell [43][44][45]. Viruses of the former phenotype were subsequently shown to harbour several arginine and lysine residues at the cleavage site [33], with an R-X-R/K-R consensus sequence being indispensable for efficient cleavage [46]. In addition, evidence was obtained that cleavage of HA might occur in the trans-Golgi network (TGN) [47]. Stieneke-Gröber and colleagues then demonstrated that these viruses are activated by furin ( Figure 2) and that peptide derivatives spanning the R-X-R/K-R consensus sequence were able to suppress viral spread [36]. Furin, which is mainly expressed in the Golgi apparatus but also found on the cell surface, processes a host of cellular pro-proteins in the secretory pathway and is indispensable for normal embryonic development [37]. Increase of influenza virus pathogenicity in poultry is associated with acquisition of a multibasic cleavage site [1,33,48] and this effect is reversed when the cleavage site is mutated [49], underlining that the protease recognition site in HA of avian influenza viruses is an important determinant of pathogenicity. Nevertheless, introduction of a multibasic cleavage site is not necessarily sufficient to immediately transform a LPAIV into a HPAIV [50], suggesting that determinants in HA other than the cleavage site might contribute to viral pathogenicity. Besides influenza virus, several other viruses, including HIV [51], Respiratory Syncytial Virus [52] and Chikungunya virus [53], misuse furin and potentially related proteases (seven pro-protein convertases were identified in humans: furin, PC2, PC1/3, PC4, PACE4, PC5/6 and PC7) to facilitate their activation by cleavage. In addition, furin activates several bacterial toxins, including the anthrax toxin [54,55] and Clostridium septicum a-toxin [56], making furin and related enzymes attractive targets for therapeutic intervention [37]. Identification of the pro-protein convertases with the highest activity for the pathogen protein in question [57] and generation of specific inhibitors are therefore important tasks [58].
Soluble trypsin-like proteases can activate influenza viruses in the extracellular space
Activation of human viruses and LPAIV is generally believed to be mediated by soluble proteases, which are secreted by lung epithelial cells. Several such enzymes have been identified, including, among others, tryptase Clara [59], mini-plasmin [60] and ectopic anionic trypsin I [61] (Figure 2). The role of these proteases in influenza virus infection has recently been reviewed [62], and will only be briefly discussed here. For some of the soluble, HAactivating proteases distinct expression patterns in the respiratory epithelium have been identified, suggesting that different enzymes might activate influenza viruses in different sections of the respiratory tract [62]. In addition, endogenous inhibitors of these enzymes, like secretory leukoprotease inhibitor and pulmonary surfactant, have been identified and were shown to inhibit viral replication in vitro and in experimentally infected rats [62]. Although these observations suggest a role of soluble cellular proteases in influenza virus spread in the infected host, direct evidence to support this hypothesis, for instance reduced viral spread in knockout animals, is largely missing.
Soluble bacterial proteases generated by Staphylococcus aureus and Aerococcus viridans strains can also activate influenza viruses [63,64]. Since bacterial superinfection by i.e. Staphylococcus aureus is a frequent complication in influenza [65], it is conceivable that bacterial proteases might promote viral spread and pathogenesis in a substantial fraction of influenza patients. Indeed, co-infection of mice with a Staphylococcus aureus strain secreting an HA-activating protease and an influenza virus with a monobasic cleavage site caused severe disease, while infection with bacteria or virus alone did not induce overt symptoms [64]. Severe disease was also not observed when bacteria were inoculated, which did not secrete the HA-activating protease, or when a virus was employed [64], which was not activated by the protease in question, indicating that the HAactivation by the Staphylococcus aureus protease was the pathogenic mechanism underlying the above described observations. In summary, soluble proteases produced by both host cells and bacteria [63,64,66] can activate influenza virus HA. While an important contribution of the former enzymes to influenza virus spread remains to be demonstrated, experimental infection of mice indicates that the latter might contribute to pneumonia development in influenza patients with bacterial superinfection.
Activation of influenza viruses at the plasma membrane by TTSPs
Many studies assessing influenza virus HA cleavage were conducted with permanent cell lines. The establishment of culture systems for primary human lung epithelial cells allowed the field to revisit earlier findings on HA activation in a more relevant experimental model. Zhirnov and colleagues used primary human adenoid epithelial cells (HAEC) to study proteolytic activation of an influenza virus harbouring a monobasic cleavage site [67]. HAEC contain differentiated cells, which exhibit ciliary motion, secrete mucins and express markers of Clara cells, and are thus considered an adequate model for the epithelium of the upper respiratory tract [67,68]. Notably, analysis in this system revealed that HA cleavage was largely cell associated, and occurred either during HA transport in the secretory pathway of productively infected cells or during entry of non-activated viruses into uninfected cells [67]. Cleavage was reduced by serine protease inhibitors [67], suggesting that the enzymes responsible for cleavage of influenza viruses in human respiratory epithelium might be cell-associated serine protease(s).
TMPRSS2, TMPRSS4 and HAT cleave the hemagglutinin of human influenza viruses
A milestone study by Bö ttcher and colleagues provided evidence that the elusive proteases responsible for cell-associated cleavage of influenza viruses with a monobasic cleavage site might be members of the TTSP family, TMPRSS2 and human airway trypsin-like protease (HAT) [19] ( Figure 2). Both proteases are expressed in the human lung, and engineered expression of either of the two enzymes in MDCK cells was sufficient to support trypsin-independent spread of influenza viruses representing the subtypes previously pandemic in humans (H1N1, H2N2 and H3N2) [19]. A subsequent study by Wang et al. confirmed cleavage-activation of HA (subtypes H1, H3 and H5) by TMPRSS2 and HAT [69], and Chaipan et al. showed that TMPRSS2 and a related TTSP, TMPRSS4, activate the HA of the 1918 influenza virus by cleavage [20]. In addition, evidence was presented that mosaic serine protease large form (MSPL) and a splice variant thereof, TMPRSS13, can cleave an influenza virus HA-derived peptide with a particular multibasic cleavage site (M-R-N-V-P-Q-K-K-K-R#-G-L-F-G from A/chick/Penn/ 1370/83 (H5N2)), which is not efficiently recognised by furin [70]. Although information on the contribution of TTSPs to influenza virus spread is missing at present, it is conceivable that these enzymes play a major role. We will therefore introduce the relevant TTSPs in more detail.
TTSPs in health and disease
The members of the TTSP family, 17 and 19 of which have been identified in humans and mice, respectively, play important roles in homeostasis and development [71]. For instance, mutations in tmprss3 are associated with deafness [72] and matriptase is required for survival, epidermal barrier function, hair follicle development and thymic homeostasis in mice [73]. In addition, deregulation of TTSPs is associated with several cancers [74]. TTSPs are synthesised as precursor proteins, zymogens, which undergo autoproteolytic activation. Although a disulfide bond links the catalytic domain to the membrane-inserted remainder of the molecule, release of the protease domain into the extracellular space has been described for several TTSPs, including HAT [75]. TTSPs exhibit a distinct domain organisation: The N-terminal domain of TTSPs is localized in the cytoplasm, and is followed by a hydrophobic transmembrane domain, a stem region and a protease domain, containing a catalytic AA triad composed of H, D and S, which is essential for proteolytic activity. The intracellular domain might interact with components of the cellular cytoskeleton as well as signalling molecules and might be required for correct intracellular trafficking of TTSPs, while the stem region might mediate protein-protein interactions and binding to macromolecules [76]. The catalytic domain cleaves cell-membrane receptors, growth factors, cytokines and components of the extracellular matrix [71,74]. In summary, the TTSP family comprises several enzymes which play an important role in health and disease and has therefore received increasing attention in the recent years.
Domain organisation, tissue expression and cellular localisation of TMPRSS2, TMPRSS4 and HAT
Two of the TTSPs, which cleave influenza virus HA with a monobasic cleavage site, TMPRSS2 and TMPRSS4, exhibit an identical domain organisation ( Figure 3). The low-density lipoprotein receptor domain class A (LDLA) and scavenger receptor cysteine-rich domain (SR) present in the stem region of these enzymes are not found in the third influenza virus HA-activating TTSP, HAT, which instead harbours a single sea urchin sperm protein (SEA) domain ( Figure 3). The role of these domains in TTSP biology and HA recognition is not well understood. The protease domains of TMPRSS2, TMPRSS4 and HAT exhibit high sequence identity (43-44%), in agreement with their shared specificity for influenza virus HA. However, it is at present unclear if lack of HA cleavage by other TTSPs, for instance TMPRSS3 (S.B., I.G. and S.P., unpublished observations), is due to differences in the protease domain or in the spatial presentation of the protease domain by the stem region. Localisation at the cell surface has been reported for many TTSPs, including TMPRSS2 [77], HAT [78] and TMPRSS4 [79], suggesting that HA cleavage by these proteases occurs upon insertion of HA into the plasma membrane. However, a recent study indicates that there might be differences in the localisation of the HA cleavage event facilitated by TMPRSS2 and HAT [78]. Thus, cleavage by TMPRSS2 but not HAT was resistant to certain protease inhibitors and resistance seemed to correlate with cleavage of HA inside the cell compared to cleavage at the plasma membrane [78] (Figure 2). This finding has implications for inhibitor development (since membrane permeable inhibitors will be required to block TMPRSS2 but not HAT) and for our understanding of HA cleavage in lung epithelial cells. Furthermore, the nature of the cellular compartment(s) where TTSPs mediate HA cleavage deserves further characterisation. Expression of all known influenza virus-activating TTSPs in lung tissue has been detected. TMPRSS2 is expressed in an epithelial-cell-specific fashion and expression was found in lung [80,81]. However, the nature of the TMPRSS2-positive lung cells awaits further characterisation. Initial studies with human tracheo-bronchial epithelium indicate expression in a subset of non-ciliated cells [82]. Expression of HAT protein was detected on ciliated cells in bronchial epithelium but was absent from basal and goblet cells [83]. Messenger RNA for TMPRSS4 has been detected in a variety of tissues, but data on protein expression are largely missing [20,79,84]. Messenger RNA for MSPL has been detected in several tissues, including lung, and TMPRSS13 mRNA seems to be ubiquitously expressed [70,85]. It remains to be determined if TMPRSS2, TMPRSS4 and HAT are co-expressed with alpha 2-6 and alpha 2-3 linked sialic acids, respectively, and if TTSPs-positive cells are targeted by influenza virus in the lung of infected individuals.
Elucidating the role of TTSPs in influenza virus infection
The intriguing data on influenza virus HA cleavage by TTSPs in cell culture should stimulate efforts to determine the relevance of these proteins for influenza virus spread in vivo. Expression and knock-down studies with cultures of human respiratory epithelium should be an important part of these endeavours. It will also be interesting to determine if HA-activating TTSPs are expressed in cell lines, which support trypsin-independent influenza virus spread, most prominently Caco-2 cells [86]. Analysis of knockout mice should then allow the contribution of specific TTSPs to influenza virus spread in the infected host to be defined. Finally, cloning and functional analysis of the respective TTSPs from animals susceptible to influenza virus infection, particularly birds (water fowl is the natural reservoir) and swine (believed to favour reassortment due, at least in part, to susceptibility to infection by viruses with different receptor specificities) [1], should provide further insights into the importance of these enzymes for viral spread in and between species.
Hemagglutinin cleavage activation in target cell vesicles
The different modes of cleavage-activation discussed so far involved proteolytic cleavage of HA in the secretory pathway of productively infected cells and cleavage of HA on progeny particles released into the extracellular space. Evidence from other viruses indicates that proteolytic activation can occur even after attachment of progeny virions to target cells. The envelope proteins of Ebola virus (EBOV) and SARS-coronavirus are activated by the endosomal cysteine proteases cathepsin B and L [87,88], and these enzymes constitute potential therapeutic targets. Cathepsins require low pH for their activity, and the previously reported pH-dependence of EBOV and SARScoronavirus entry is due to the requirement for cathepsin activity rather than direct glycoprotein activation by low pH [87,88]. Notably, activation in target cell vesicles has also been described for influenza virus A/WSN/33 [89] (Figure 2). This process was observed with MDBK but not CV-1 cells and was sensitive to serine-but not cysteine protease inhibitors [89], suggesting that influenza virus activation in target cells might occur in a celltype dependent fashion and is not facilitated by cathepsins. The protease(s) responsible for influenza virus activation in MDBK cells remains to be identified. However, the above described observation by Zhirnov and colleagues that HA activation can occur upon influenza virus uptake into cultures of respiratory epithelium [67] suggests that this mode of HA cleavage might be important in the infected host and thus deserves further investigation.
NEURAMINIDASE PROMOTED CLEAVAGE OF THE HEMAGGLUTININ OF A/WSN/33 AND THE 1918 INFLUENZA VIRUS
Despite the ample possibilities for HA cleavage discussed above, two prominent influenza viruses evolved additional mechanisms, which depend on the presence of the viral NA. One is the frequently studied, neurovirulent [90] laboratory strain A/ WSN/33 (H1N1), which was obtained by extensive passaging of the parental virus, WS/33, in different animals. The spread of WSN/33 in infected mice is not limited to brain and lung tissue but is rather systemic [91], suggesting that the virus evolved a mechanism to ensure efficient HA cleavage in a broad spectrum of cell types. Studies dating back to the 1970s demonstrated that (i) WSN/33 was activated by the serum protease plasmin [92] [93] and that (iii) NA [90], specifically adequate glycosylation of NA [94], was required for neurovirulence in mice. Work by Goto and Kawaoka linked these observations by demonstrating that WSN/33 NA recruits serum plasminogen, which upon conversion to plasmin, activates HA [95][96][97]. One of several amino acid exchanges present in WSN/33 NA but not the parental WS/33 NA was found to be crucial for HA activation: Mutation N146R (N2 numbering), which inactivated a signal for N-linked glycosylation, conferred plasminogen binding to WS/33 while the reverse exchange abrogated plasminogen recruitment by WSN/33. In addition, the carboxy-terminal lysine residue 453, which is conserved among NAs of the N1 subtype, was essential for plasminogen recruitment [95][96][97], suggesting that certain influenza virus NAs might have the intrinsic capacity to bind plasminogen, which is masked by posttranslational modification of N146 by a glycan side chain. The recruitment of plasminogen by WSN/33 NA and the resulting expansion of the viral tropism raised the intriguing question, whether a similar mechanism might contribute to spread of other influenza viruses, most importantly the highly pathogenic 1918 influenza virus. Analysis of the 1918 NA revealed the presence of the lysine residue crucial for plasminogen binding but also showed the presence of the glycosylation signal incompatible with plasminogen recruitment by WSN/33 [98]. Notably, the 1918 virus replicated in the dog kidney cell line MDCK with high efficiency in the absence of trypsin, and trypsinindependent spread was dependent on the presence of the 1918 NA [99]. These results suggest that the 1918 NA also facilitates HA cleavage, at least in some cellular systems (MDCK cells), and might do so by recruiting a protease. A subsequent study demonstrated that WSN/33 NA but not 1918 NA was able to sequester plasminogen and to activate WSN/33 HA [20]. Conversely, and unexpectedly, the 1918 HA but not the WSN/33 HA bound plasminogen and the biological significance of the plasminogen capture is at present unclear [20]. These results argue against 1918 NA facilitating HA cleavage in a manner directly comparable to WSN/33 NA. It can be speculated, however, that 1918 NA might recruit a factor which promotes plasminogen conversion into plasmin, like annexin II, which was shown to be incorporated into influenza virus particles and to support viral replication by activating plasminogen [100]. Regardless of the, at present unclear, mechanism underlying NA-dependent 1918 HA cleavage, it needs to be noted that this process was not detected in the human hepatoma cell line Huh-7 or the human kidney-derived cell line 293T [20], and is thus of uncertain relevance to 1918 virus spread in the human host. Finally, it is noteworthy that the 1918 influenza virus and a seasonal influenza virus replicated in the human lung cell line Calu-3 with high efficiency in the absence of trypsin [99]. Cleavage of the HA proteins by TTSPs, which does not depend on NA, might have been responsible. Expression analysis and knock-down studies are needed to test this hypothesis.
PROTEOLYTIC CLEAVAGE OF HEMAGGLUTININ AS A TARGET FOR THERAPEUTIC INTERVENTION
As to be expected for a highly variable RNA virus, usage of NA and M2 inhibitors as monotherapy for influenza triggers emergence of resistant viruses, and a substantial proportion of the currently circulating viruses are insensitive against at least one or more of the drugs available. The potential of combination therapy should therefore be explored, and the development of new drugs is imperative. The cellular proteases which activate HA are attractive targets for several reasons. For one, they are essential for viral spread and, as host cellencoded factors, of invariable nature. As a consequence of the latter property, viral resistance due to mutation of the drug target can be excluded. The alteration of the cleavage site in response to protease inhibition is conceivable. However, it is important to note that evolution of seasonal influenza viruses with a multibasic cleavage site, which would confer resistance to inhibitors directed against proteases recognising monobasic cleavage sites, has never been observed in humans. Another advantage of protease inhibitors is the mere fact that ample proof of concept is already available: protease inhibitors have been documented to suppress influenza virus spread in cell culture [78], in experimentally infected animals [101,102] and, most importantly, in humans [102]. Major unwanted side effects were not observed in these studies but are of concern when considering clinical development of protease inhibitors for influenza therapy. However, identifying which of the so far described HA-processing proteases indeed support(s) spread of human influenza virus in infected individuals would allow the generation of highly specific inhibitors, which can be expected to be both safe and effective. The discovery of TTSPs as HA-activating enzymes combined with the observation that knockout of TMPRSS2 has no phenotype in mice [103], might be the first step in this direction.
CONCLUSIONS
The cleavage of HA by cellular proteases is essential for virus spread and pathogenesis, so the responsible proteases are attractive targets for intervention. For avian influenza viruses, a correlation between the cleavage site sequence, the nature of the HA-processing proteases and the degree of viral pathogenicity has been firmly established. For human influenza viruses, a similar correlation does not exist and the paradigm on HA cleavage shifted from cleavage by soluble proteases to cleavage by cell-associated proteases. Recent studies suggests that members of the TTSP family, specifically TMPRSS2, TMPRSS4 and HAT, mediate cell-associated HA-cleavage, and future studies must define the importance of the respective enzymes. Protease inhibitors are effective as treatment for influenza [102], and TTSPs might emerge as prime targets. Finally, inhibitors of TTSPs might be of therapeutic value for conditions others than influenza, since recent evidence suggests that TTSPs are also involved in the proteolytic activation of human metapneumovirus [104] and SARS-coronavirus [105], and an important role of TTSPs in several cancers is well documented [71,74]. | 2018-04-03T02:10:58.325Z | 2010-05-07T00:00:00.000 | {
"year": 2010,
"sha1": "ccef8386210aaf7fbc6e73e7e98c9a276befea1b",
"oa_license": "implied-oa",
"oa_url": "https://europepmc.org/articles/pmc7169116?pdf=render",
"oa_status": "GREEN",
"pdf_src": "PubMedCentral",
"pdf_hash": "31d68502107054251b66b501919ae1d903d101c7",
"s2fieldsofstudy": [],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
250677522 | pes2o/s2orc | v3-fos-license | Experimental investigation of aerosol formation in laser fusion reactor chamber by discharge method
Liquid wall chambers are one of the most critical technologies in inertial fusion energy (IFE) plants. The critical problem for the liquid wall IFE chamber is the chamber clearance after laser shots. This paper presents preliminary results from an experimental simulation of ablation processes induced by alpha-particle heating.
Introduction
A liquid LiPb flow, whose flow rate is stabilized using a cascade scheme, is being considered for the first wall of the fast-ignition laser-fusion power plant, KOYO-Fast. The first wall is irradiated by pulses of X-rays, neutrons, and charged particles generated by fusion burning [1]. The heating effect of alpha particles needs to be considered, especially when determining the ablation of the first surface, since alpha particles deposit their energy in a short range of about 10 μm [2]. Alpha particles release their energy when they are completely stopped inside the liquid wall. The energy density of alpha particles at the surface is estimated to be 0.35 MJ/m 2 . Ablation of the liquid wall occurs because the temperature inside the wall becomes higher than that at its surface. It resembles a 10-μm-thick membrane peeling off from its surface. In the case of KOYO-F, 10 kg of LiPb evaporates after each laser shot [3]. In this experiment, we aimed to simulate the formation and distribution of aerosols from a plane source and also the hydrodynamic phenomena. Some pioneering studies on the formation and behaviour of aerosols are described in Ref. [4]. In previous studies, formation of aerosols was discussed in terms of laser heating and only surface heating by thermal conduction. In this present study, we used a discharge method to simulate volumetric heating.
Discharge method
A monolayer film was used to investigate the spatial distribution of aerosols from a plane source. In addition, a multilayer film consisting of layers having different thermal conductivities was used to investigate formation of aerosols in the presence of hydrodynamic instabilities. For simplicity, the influence of Li was ignored in this experiment. A Pb membrane was evaporated and, after expansion and thermal radiation cooling, the Pb vapour in the slow component formed aerosols. A schematic diagram of the experimental set up is shown in Fig. 1. The specifications of the discharge circuit used in this experiment were: a capacitor of C=150 μF, an applied voltage of V=1 kV, a Pb membrane having dimensions 5.0 mm×10.0 mm×12.2 μm. The available energy stored in the capacitor was 75 J.
The discharge current was monitored using the potential drop between the ignitron and the capacitor and a high-voltage probe was connected across the anode and cathode electrodes to detect the discharge voltage.
Temperature measurement
The Pb membrane was converted from a solid to a plasma by pulse heating. Plasma emission was detected using a charge-coupled device (CCD) camera equipped with bandpass filters which were inserted in front of two of the pinholes to disperse the plasma emission at 460 nm and 660 nm in order to determine the colour temperature. An amplified photodetector was used to measure the temporal variation of the emission. The following relational expression for the colour temperature was used [5], is the relative spectral distribution, which is subjected to the Planck radiation law and is given as follows:
Capture of Aerosols
The slow components of the ablated vapour form aerosols due to condensation of the Pb vapour. We measured the particle size and distribution of these aerosols. The distances between the Pb membrane and the glass used to capture the aerosols were L=5, 10, 12, 20 mm. The internal pressure in the experimental chamber was P=12.5 Pa.
Heating power and emission from plasmas
The goal of this analysis was to determine the heating rate and the temperature of the vapour source. We simultaneously detected the pulse current, discharge voltage and intensity of visible light after discharge. The discharge power and energy of the Pb membrane are plotted as a function of time in Fig. 2. where, V is the discharge voltage and I is the pulse current. The current pattern was a dumping oscillation due to the low resistance of the discharge circuit. The discharge power has two large peaks. 10% of the energy stored in the condenser was expended heating the Pb membrane. To calibrate the colour temperature measurement, we measured the CCD response to 10-ms exposure to the standard halogen lamp and to plasma emission. The average plasma was determined using this calibration data and Eqn. (1), and found to be = t D T , 4050±53 K. Figure 3-(a) shows an AFM image (Atomic Force Microscope) of the aerosols deposited on the glass plate and figure 3-(b) shows a SEM image (Scanning Electron Microscope) of a cross sectional view of captured aerosols that were deposited at 20 mm front of the ablation source and the chamber pressure at discharge was 12.5 Pa. The diameters of the clusters of aerosol particles ranged from 20 to 40 nm and spherical particles are deposited on a continuous membrane. Figure 3-(b) indicates that leading hot vapour was captured as the membrane and trailing slow components, which were cooledaerosols, were deposited as the particles. Table 1 shows the experimental results for diameter of captured aerosol particles and average mass of deposited aerosols. The areal mass density was determined by the intensity of X-ray-stimulated fluorescence. When L is shorter, the diameter is larger. This indicates that either particles are grown-up larger in the stagnated vapour between the source and the capture plate or particles are grown-up on the capture plate.
Summary
A monolayer membrane was used to compare the results of a numerical simulation of aerosol formation with experimental results. Preliminary results indicate that growth of aerosols strongly depends on the geometrical effect formed by the source plate and capture plate. Further investigations are necessary to determine whether captured particles are grown in flight or on the capture plate. | 2022-06-27T23:26:15.968Z | 2008-01-01T00:00:00.000 | {
"year": 2008,
"sha1": "8287b4c4083927adcbdeaa2ad15cdf7ac31f5dd7",
"oa_license": null,
"oa_url": "https://doi.org/10.1088/1742-6596/112/3/032040",
"oa_status": "GOLD",
"pdf_src": "IOP",
"pdf_hash": "8287b4c4083927adcbdeaa2ad15cdf7ac31f5dd7",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
90697884 | pes2o/s2orc | v3-fos-license | Relationship between bird-of-prey decals and bird-window collisions on a Brazilian university campus
Bird-window collisions are a dramatic cause of bird mortality globally. In Latin America, statistics are generally very scarce and/or inaccessible so the frequency of such incidents is still poorly understood. Nevertheless, civilians have applied preventive methods (e.g. adhesive bird-of-prey decals) sparsely but, to our knowledge, no study has evaluated their effectiveness in Brazil. Here, we estimated the mortality rate of bird-window collisions and tested the effectiveness of bird-of-prey decals at preventing such accidents. We undertook daily searches for bird carcasses, presumably resulting from window collisions, near all buildings on a university campus over seven months. Adhesive bird-of-prey decals were then applied to the two buildings with the highest mortality rates and surveys continued for over 12 more months. The mortality rates before and after the application of decals and between seasons were then compared using Friedman test. We recorded 36 collisions, 29 around the two buildings with the highest collision rates 19 prior and 10 after our intervention with associated collision rates of 0.08 and 0.04 collisions/day. Although mortality was reduced by almost half, this difference was not statistically significant. The Blue-black grassquit, Volatinia jacarina (Linnaeus, 1766), and Ruddy ground dove, Columbina talpacoti (Temminck, 1810) suffered the highest number of collisions, followed by the Rufous-collared sparrow, Zonotrichia capensis (P. L. Statius Müller, 1776). Our bird-of-prey decals and efforts were insufficient to prevent or dramatically reduce the number of bird-window collisions. Therefore, we recommend that different interventions be used and additional longterm studies undertaken on their efficacy.
INTRODUCTION
Bird collisions with human structures (eg.vehicles, aircrafts, communication towers, wind turbines, power lines and buildings) are among the major anthropogenic causes of bird mortality in the world (Erickson et al. 2005, Jenkins et al. 2010, Rytwinski and Fahrig 2012, Calverti et al. 2013, Loss et al. 2014, DeVault 2015, Washburn et al. 2015, Santos et al. 2016).In North America alone, up to one billion birds are estimated to die due to collision with buildings (Loss et al. 2014, Klem 2015).However, what estimates exist are limited and they are largely extrapolations of data collected from a handful of locations in the Northern Hemisphere (Machtans et al. 2013), often without considering the spatial variation of the urban landscape (Hager et al. 2013).Furthermore, the frequency of collisions is typically underestimated as predators or scavengers often consume carcasses before they can be counted (Klem 2009a, Hager et al. 2012).
Human constructions with reflective glass are especially lethal, and have been considered one of the greatest causes of declines in global bird populations (but see Arnold and Zink 2011), second only to habitat loss (Klem 2006(Klem , 2009a)).Several studies in North America reported bird collisions with tall window-covered buildings numbering in the thousands over both the short-and long-term, even in a single day (Erickson et al. 2005).
They occur because birds do not perceive windows as obstacles (Klem and Saenger 2013), especially when they can see the sky and surroundings reflected in the glass (Menacho-Odio 2015).It is estimated that about half of all collisions result in death (Klem 1990), but it may not be instantaneous, and many victims die as a result of subsequent shock, injury, or being more vulnerable to predation while recovering (Klem 1990, Parkins et al. 2015).
Virtually all flying birds are faced with the threat of window collision with occurrences having been reported for ~6% of all bird species (Klem 2006(Klem , 2009b)).Vulnerability assessments to determine which species are at higher risk of collisions are important to understand the real impact of human infrastructure on birds and to inform practical management decision-making (Loss et al. 2014).In North America, upwards of one quarter of bird species have been listed as potential victims, regardless of sex, age or residency status (Klem 1989).
In Latin America, statistics about bird collisions with windows are generally very scarce and/or inaccessible, but a few publications are available.In Costa Rica, Menacho-Odio (2015) listed 131 species for which window collisions were reported based on museum specimens and published reports.Little is known about such incidents in Brazil, and there have been few studies published in the broader scientific literature (but see Barros 2010, Stolk et al. 2015).This underscores both the surprising lack of information about the impacts of bird-window collisions on bird populations and the absence of evaluations of the best management practices for preventing or reducing these accidents.
Several methods have been proposed in the Northern Hemisphere in an effort to reduce bird-window collisions, including both 2D adhesive decals and 3D mobiles simulating birds of prey, wind bells, flashing lights, UV markings and paintings, and stripes (Klem 1990, Klem and Saenger 2013, Oviedo and Menacho-Odio 2015, Rössler et al. 2015).In Brazil, adhesive bird-of-prey decals are sparsely applied to glass windows in cities and parks throughout the country but without standardization and unknown effectiveness.Indeed, to our knowledge, the effectiveness of these decals has never been properly tested in Latin America.
Given the widespread use of bird-of-prey decals as a preventative measure for bird-window collisions but the lack of empirical support, we aim to test their efficacy at reducing bird-window collisions at a local scale in Brazil.However, based on previous studies in the northern hemisphere (e.g.Klem 1990, Rössler et al. 2015) we predict that bird-of-prey shaped decals will be ineffective at reducing the number of bird-window collisions and that it is independent of season.We also estimate which species are prone to window collisions.
MATERIAL AND METHODS
This study was conducted on the campus of the Federal University of São Carlos in Sorocaba, São Paulo, Brazil (47°31'28"W, 23°34'53"S) (Fig. 1).It is approximately 70 ha in extent and consists of patches of abandoned pastures, tropical savanna ('cerrado'), secondary seasonal Atlantic forest, and small waterbodies (Fig. S1 -Suppl.material 1) (Viviani et al. 2010).Buildings cover 46,402 m 2 or nearly 6.5% of the total campus area (University City Hall, unpublished data).The dry season in the region is from April to September, and the rainy season from October to March (CEPAGRI 2016).Breeding in most tropical birds peeks during the rainy season (Wikelski et al. 2000).
We first estimated the mortality rate of bird-window collisions before any intervention was applied.We carried out daily surveys for bird carcasses, presumably resulting from fatal bird-window collisions, around eight buildings (Fig. S2 -Suppl.material 2), covering an area of approximately 18,000 m 2 and up to five meters away from each windowpane from 11 August 2014 to 8 March 2015 (262 days), usually between 1 and 3 pm.Our data was complemented by finds by reliable third parties, who were made aware of this study and assisted in collecting carcasses.
The two buildings with the highest collision rates were Aulas Teóricas e Laboratórios (hereafter ATLAB), and Gestão Administrativa (hereafter GAD).ATLAB covers 8,320.96m 2 is 7 m high and has approximately 260 m 2 (3.1% of the total surface area) of translucent glass windows.It is mainly purple in color with several overhangs and awnings and is surrounded (to within 5 m) by lawns and scattered trees.GAD covers 1,067.53m 2 , 7 m high and has approximately 354 m 2 (33.2% of the total surface area) of reflective glass windows that received a dark film for filtering the incoming light.It is orange in color and surrounded by over a dozen trees and several shrubs, which provide shade at certain times of the day (see Fig. S2A, H -Suppl. material 2).We constructed 154 generalized bird-of-prey decals, 20 × 40 cm (0.08 s.m.), based on the photo (http://www.ideiasedicas.com/aulas-sobre-aves-falcao/falcao-voando).It was converted to silhouette-form using Gimp 2.8 to serve as a template that was traced onto adhesive white paper (Fig. S3 -Suppl.material 3).We chose white to be conspicuous against the dark windows.We attached 96 decals to ATLAB and 58 to GAD between March and May 2015 (while continuing to collect data), distributing them evenly by alternating between one or two windows (Figs 2-4).We distributed decals across as much of the exposed area of each building as possible but concentrated on places where the most carcasses were found, whilst access to other areas was restricted.
We continued surveys around GAD and ATLAB until 8 March 2016 (236 days).The total number of mortalities without and with decals and between the dry and wet seasons was then estimated and compared using the Friedman test and post hoc analysis using RStudio (R Development Core Team 2015).
RESULTS
We recorded 36 fatalities resulting from collision with windows (0.07 fatalities/day) from twelve species (Fig. S4 -Suppl.material 4, Table S1 -Suppl.material 5).Volatinia jacarina (Linnaeus, 1766) and Columbina talpacoti (Temminck, 1810) were the most commonly recorded (n = 8 for both), followed by Zonotrichia capensis (P.L. Statius Müller, 1776) (n = 6).Four individuals could not be identified as a result of third-party information and/ or carcasses that were in advanced states of decomposition.The buildings with the highest number of incidents were GAD (n = 21) and ATLAB (n = 8).Seven fatal collisions occurred at other buildings, but these were not included in our analyses.The total number of collisions in windows to which decals were applied (Fig. 5) was nearly half of that prior to intervention (n = 10; 0.04 collisions/day following intervention, compared to n = 19; 0.08 collisions/day prior to intervention).However, there was no statistically significant difference in the number of mortalities between windows without and with decals or between season (MaxT = 1.53; p = 0.42); so collisions were neither influenced by decals or season (Table 1).
DISCUSSION
Our data suggests that representatives of the Columbidae (e.g., C. talpacoti) and certain passerines (e.g., V. jacarina and Z. capensis) may be more prone to collision with windows.Birds having rapid flight and heavy bodies with small wings are less able to react swiftly to unexpected obstacles (Bevanger 1998).This is also in keeping with Rayner (1988) who determined that the relatively small thin wings of the Columbidae make them among the bird groups most susceptible to collisions.In North and Central America, nocturnal migrants, grouses, falcons, hummingbirds, and passerines such as warblers, manakins, sparrows and thrushes were the most likely victims (Gelb and Delacretaz 2009, Breithaupt et al. 2013, Klem 2014, Loss et al. 2014, Menacho-Odio 2015).
Overall, the collision rate is at least partially explained by the density of birds in the surrounding area (Sabo et al. 2016) as well as factors including age, sex, and behavior (Hager andCraig 2014, Kahle et al. 2016).Still, some species that occur at high densities in urban areas, such as the Rock Dove (Columba livia) and House sparrow, Passer domesticus (Linnaeus, 1758), may learn from experience to avoid windows (Klem 2014).In this study, the highest number of mortalities resulting from bird-window collisions was recorded for the building (GAD) with the greatest total and percentage area of windows and number of surrounding trees.This supports the findings of Gelb and Delacretaz (2009), Loss et al. (2014), Cusa et al. (2015), Ocampo-Peñuela et al. (2016a) and Bracey et al. (2016).They stressed that, apart from bird ecology and behavior, the number of windows and the percentage area of a built structure covered by windows influences the number of collisions.Collisions are also strongly dependent upon the surrounding landscape and its potential to attract birds (Hager et al. 2008, Hager andCraig 2014).In particular, Dunn (1993) related high collision rates in US winter homes to the fact that residents placed bird feeders next to their windows.
Although the daily collision rate decreased by half (0.08 to 0.04) following application of decals, this was not enough to generate a statistically significant difference.Given the lack of similar studies in Brazil, comparisons of the efficiency of our method are restricted to elsewhere in the world.Klem (1990) tested various methods in the USA, including similar bird-ofprey decals and 3D mobiles, wind chimes and flashing lights, but found no significant reduction in the number of collisions.He also found the largest reductions were associated with the application of horizontal or vertical stripes in a rectangular mesh.In Europe, Rössler et al. (2015) tested several methods using achromatic patterns and found that 2mm-wide stripes covering < 7% of a window were as effective as stripes of 13mm covering half of pane at reducing collisions.He also found that 10 cm long vertical stripes were more effective than horizontal stripes of the same size.In Costa Rica, people living in urban areas preferred mitigation structures that have a low cost, high aesthetic quality, and do not impair the passage of light or obscure views (Oviedo and Menacho-Odio 2015).Patterned glass, UV-reflective films and objects separated by 5-10cm have also proven effective (Klem 2009a).UV-reflecting adhesives (0.41 decals/m 2 ) applied to a residence in Colombia resulted in an Table 1.Bird-window collisions for two buildings (ATLAB and GAD) on the campus of the Federal University of São Carlos, Sorocaba, Brazil, during the wet (October to March) and dry (April to September) seasons, in windows without and with bird-of-prey decals.DS = dry season; WS = wet season.In parentheses, values expressed as (number of fatalities/day) x100*.However, UV reflection would only be efficient for birds with strong ultraviolet perception (including most passerines).For other groups, it would only be optimized under certain light conditions (Hastad and Ödeen 2014).
Species
We also found no difference between seasons, but this requires verification by long-term studies.In the northern hemisphere, collisions tend to increase during the spring and autumn migrations (Gelb andDelacretaz 2009, Kahle et al. 2016).However, in Colombia, collisions occurred year-round but peaked in August and September (Ocampo-Peñuela et al. 2016b).In Bogota the number of accidents was higher in October, coinciding with the arrival of boreal migrants (Agudelo-Álvarez et al. 2010).
Unfortunately, the small scale and relatively low number of collisions recorded during our study limited the scope of analyses, even after 19 months of data collection across seasons.It is also possible that the stickered area was insufficient but this is a consequence of the method, which repeats the measures taken by the Brazil public.Thus, while we caution against making sweeping generalizations we conclude that our bird-of-prey decals and efforts were insufficient to reduce the number of bird-window collisions.
Houses, public buildings and facilities in protected areas commonly use similar birds-of-prey shaped decals in an effort to reduce bird collisions.However, our study suggests that their use is merely aesthetic and does not reduce the number of bird-window collisions, as desired.Therefore, we suggest that this tradition be abandoned and efforts should rather be encouraged to identify the factors most frequently associated with bird-window collisions (Gelb and Delacretaz 2009).Based on this, an effective protocol could be developed to incorporate the testing of a various interventions.This would include decals separated by up to 10cm (Klem 2009b, Ocampo-Peñuela et al. 2016b), 10 cm x 2 mm vertical stripes (Rössler et al. 2015) and avoiding "ecological traps", such as bird feeders near glass windows (Klem 1990, Klem et al. 2004, Krummer and Bayne 2015) in an effort to reduce collisions.The main goal would be to create a signal of any kind that allows birds to detect and avoid windows (US Fish and Wildlife Service 2016).
In North America, citizen scientists have proven to be strong allies in estimating the number of bird-window collisions (Machtans and Thogmartin 2014, Loss et al. 2015, Kummer et al. 2016a, b).We therefore recommend undertaking similar efforts in Latin America and that nationwide data on bird-window collisions be gathered in Brazil to develop effective public policies in urban management.Nevertheless, it is still necessary to produce robust local data on bird-window collisions and introduce preventive measures across Brazil.
Figure 1 .
Figure 1.Location of the study area in Sorocaba, São Paulo, Brazil.
Figure 5 .
Figure 5.Total number of bird fatalities resulting from bird-window collisions in windows without and with bird-of-prey decals on the campus of the Federal University of São Carlos, in Sorocaba, Brazil. | 2019-01-02T01:03:02.522Z | 2017-05-06T00:00:00.000 | {
"year": 2017,
"sha1": "f3eecb8740545336d941a0d64d2b3f12e80fd800",
"oa_license": "CCBY",
"oa_url": "https://zoologia.pensoft.net/article/13729/download/pdf/",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "f3eecb8740545336d941a0d64d2b3f12e80fd800",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": [
"Biology"
]
} |
243815590 | pes2o/s2orc | v3-fos-license | Animal protein consumer's perception on the welfare of production animals in Belém, Pará State, Brazil
The objective was to identify the level of knowledge of animal protein consumers about the welfare of production animals, in Belém city, Pará State, Brazil. The survey was carried out in two supermarkets, using 401 structured questionnaires, with 10 closed questions, with a “Yes” or “No” answer, taking into consideration, for the formation of a socioeconomic profile, questions related to gender, age group, educational level and family income of the interviewees. Most consumers of animal protein from the municipality of Belém, Pará State, Brazil, recognize that these products are part of their daily diet, however, the interviewees have no knowledge about the topic, and are not willing to pay more for the product with a seal of quality, as well as showing no interest or concern with the way animals are reared or slaughtered. The age group of consumers influences the level of knowledge about the welfare of production animals, observing that consumers aged 40 to 50 years have more knowledge on the subject. Education was a decisive factor in determining the level of knowledge of animal protein consumers about the welfare of production animals, as consumers from high school had more knowledge about the subject.
Introduction
The human being, probably, started activities of maintenance of animals for production about ten thousand years ago (Zeder & Hesse, 2000). Among the forms of interaction between humans and animals, the relationship between them and their creators, perhaps is the one that has undergone a number of changes more pronounced throughout history. According to Fraser and Broom (2002), at the beginning of the 20 th century, the use of animals for production increased in association with the expansion of human needs. Then, in response to such needs, there were changes to a production system model, where animals were subjected to high stocking densities to meet commercial pressures.
In Brazil, concerns about animal welfare grow in parallel with socioeconomic development (Rocha, Lara, & Baião, 2008). In addition, it is noted that Latin American consumers have been concerned with animal welfare and rearing systems, considering welfare as a guarantee of good quality meat (Vargas-Bello-Pérez, Riveros, Köbrich, Álvarez-Melo, & Lensink, 2017;Miranda de la Lama et al., 2017).
The increase and distribution of income in Brazil caused changes in the consumption behavior of the population. The classes with greater purchasing power are directed to the consumption of products of animal origin, prioritizing quality, certification and food safety. This part of the population tends to seek products with higher added value. Classes with lower purchasing power and the rising classes, prioritize the price, practicality, convenience and safety of the food (Abreu et al., 2021). However, in general, consumers of animal origin have been increasingly interested in knowing how farm animals are raised and whether there is suffering during production and slaughter (Queiroz, Barbosa-Filho, Albiero, Brasil, & Melo, 2014).
In Rio Verde, Goiás State, Brazil, for example, consumers of products of animal origin do not have contact with farm animals, although they believe they have reasonable knowledge of the way farm animals are raised. These consumers are willing to pay up to 3% more for a product from production systems that respect animal welfare (Schaly, Oliveira, Salviano, & Abreu, 2010). In Fortaleza, Ceará State, Brazil, on the other hand, most consumers demonstrate knowledge about the benefits of animal welfare and, as a result, would be willing to pay more for meat with a certificate of animal welfare (Queiroz et al., 2014).
On the other hand, it is clear that there is a need for greater divulgation regarding the welfare of production animals, as there is a large portion of the population that still does not care about this issue, such as, for example, consumers in Santarém, west Pará State, Brazil, that although animal protein is part of their diet, they do not have knowledge about animal welfare, and as a result, they are not willing to overpay for products with a seal and are not concerned with the ways of raising the animals (Silva, Silva, & Silva, 2020).
In Belém, Pará State, Brazil, however, there are no results regarding what the consumer thinks about the welfare of production animals and how important this issue is when choosing a product. Thus, this research aimed to identify the level of knowledge of consumers of products of animal origin about the welfare of production animals, in Belém, Pará State, Brazil.
Material and methods
The research was carried out with the aid of a questionnaire with ten questions, with answers "Yes" or "No", considering gender, age group (between 18 and 30 years old, 30 and 40 years old, 40 and 50 years old; 50 and 60 years old, and over 60 years old), education (incomplete elementary, elementary, incomplete high school, high school, incomplete and higher education) and family income (between 1 and 3 minimum wages -MW; 3 and 6 MW; 6 and 9 MW; 9 and 12 MW, and over 12 MW, whose interviewees were organized into six educational levels and five levels of family income. The questionnaire contained the questions described in Table 1. Have you ever heard of animal welfare? 5 Do you think animals raised under welfare standards will lead to higher quality products? 6 Would you pay more for a product to ensure that animals were raised under welfare conditions? 7 Do you think that production animals are subjected to some kind of suffering during their creation? 8 Do you know that there are laws that ensure animal welfare? 9 Would you choose products certified (seals) by animal welfare agencies? 10 Would you like supermarket chains to offer products originated in accordance with welfare standards?
The interviews were conducted in two stores of a supermarket chain, in two different neighborhoods in the metropolitan region of Belém city, in Pará State, Brazil, in order to reach an audience of different socioeconomic levels. Socioeconomic classification was made, based on data from IBGE (2010) ( Table 2), ordered according to the number of monthly minimum wages. Questionnaires were applied to different consumers of products of animal origin, in two supermarkets, one from each chain (chain 1 -C1 and chain 2 -C2), in a total of 401 individuals, chosen at random. Two supermarket chains were used to allow repetition among respondents of different income levels. The answers obtained were analyzed individually, in percentage, using an Excel ® 2014 electronic spreadsheet. For statistical analysis, the questions were grouped as follows: 1 and 2, 3 and 4, 5 and 6, 7 and 8, 9 and 10. The questions in pairs were analyzed with a non-parametric test of significance, using the chi-square (X 2 ), as described by Levin (1987). The R program was used. The contingency matrix used contained two rows and two columns (2 x 2), with a degree of freedom equal to 1. The significance level was set at 5%, with a critical value of 3.84. The experiment hypothesis was considered accepted when it had a value equal to or greater than 3.84.
Results and discussion
It was found that consumers answered "Yes" in all age groups (G1, G2, G3, G4, and G5) for questions 2, 5, 6, 7, 9, and 10 ( Figure 1), demonstrating that most of the interviewees say they have knowledge about the way the animals they consume are raised, highlighting that the younger consumers, members of the G1 group (18 to 30 years old), had a higher rate of conviction, corroborating with Silva et al. (2020), who found that 60% of respondents have no knowledge of how production animals are raised. 3,4,5,6,7,8,9, and 10, who answered "Yes".
Consumers in Belém, Pará State, Brazil agree that animals raised according to animal welfare standards tend to have good quality meat, as observed in the different age groups G1 (94.4%), G2 (93.7%), G3 (92.9%), G4 (90.1%), and G5 (86.6%), and as a result, would be willing to pay more for the product, to guarantee animal welfare, totaling G1 (100%), G2 (92.4%), G3 (93.9%), G4 (92.3%), and G5 (91.8%). In addition, all consumers of different age groups believe that animals go through some type of suffering (question 7) during their creation or slaughter, with a higher rate of this statement being observed in young people in G1. In a study developed in Rio Verde city, Goiás State, Brazil, Schaly et al. (2010) found that 66.9% of consumers were willing to pay up to 3% more to have meat from a livestock system that would favor animal welfare. In Santarém, in the northern region of Pará, Silva et al. (2020) found that consumers in classes A and B would be willing to pay more for meat with an animal welfare quality seal. In contrast, the authors also describe that classes C and D would not pay (Silva, Silva, & Silva, 2020).
However, it was found that in questions 3, 4, and 8, there was a predominance of negative responses (No), thus, it is noted that consumers in groups G2 (64.6%), G3 (66.3%), G4 (60.4%), and G5 (53.6%) are not concerned with knowledge about the methods of rearing or slaughtering production animals, with the exception of G1 (27.8%), which corresponds to respondents aged 18 to 30, who totaled "Yes" answers in 72.2%, therefore, it is evident that younger consumers show greater concern with the methodologies adopted during the animal's life. In Rio Verde, Goiás State, Brazil, 49.01% of consumers' lack of knowledge about farming methods was also evidenced (Schaly et al., 2010).
Regarding the laws, it was found that consumers aged G1, G2, G3, G4, and G5) were unaware of the legislation that protects animals (55.6, 69.6, 23.1, 84.6, and 83.5%, respectively). Consumers in the G1 (44.4%) and G2 (30.4%) groups demonstrated to know more about this issue when comparing with the other groups. Different results were obtained in Porto Alegre, Rio Grande do Sul State, Brazil, where there was a greater demand on the part of consumers, with a higher index of knowledge on matters related to their health (Barcellos, 2004;Francisco, Nascimento, Loguercio, & Camargo, 2007).
Regarding the different levels of education, it is noted that consumers answered positively (Yes) to questions 2, 5, 6, and 7 (Figure 2), which demonstrates that the majority of respondents (E -71%, IE -79.4%, HS -75.2%, IHS -90.9%, HE -70.6, and IHE-64.3%) say they have knowledge about the way of raising the animals they consume, highlighting that consumers of incomplete high school (IHS) had a higher rate of conviction (90.9%). Hotzel and Machado Filho (2004) describe that most of the information is passed on to the public in a simplistic way and strongly charged with emotions, which makes it difficult for consumers to understand. Consumers of all educational levels claim that animals bred in accordance with animal welfare standards have good quality meat, showing that individuals with higher education (96.3%) were more convinced about the issue (E -93.5 %, IE -73.5%, HS -91.6%, IHS -72.7%, HE -96.3%, and IHE -92.9%). In this survey, consumers with all levels of education would be willing to pay more per product with a guarantee that the production animals were raised on animal welfare standards, the group with incomplete high school (100%) was the more convinced in the answer (E -96.8%, IE -79.4%, HS -94.1%, IHS -100%, HE -94.5%, and IHE -92.9%). According to Souza, Casot, and Lemme (2013), a seal guaranteeing the absence of mistreatment in the meat production process could have good acceptance on the part of the consumer market, if the public were correctly communicated about the differences in management standards in animal welfare.
In this survey, all consumers at different levels of education believe that production animals go through suffering during creation or slaughter (E -74.2.%, IE -70.6%, HS -75.7%, IHS -72.7%, HE -74.3 %, and IHE -85.7%). However, consumers who had incomplete higher education demonstrated greater certainty about this issue.
It was found that most consumers of different levels of education answered "No" in questions 3, 4, 8, 9 and 10, thus, people with incomplete high school (81.8%) demonstrated not to be concerned with the methods of breeding of the production animals, as well as the other levels of education (E -61.3.%, IE -64.7%, HS -57.4%, HE -57.8%, and IHE -28.6%), with the exception of consumers of incomplete higher education who demonstrated greater level of concern, answering "Yes" at 71.4%. Similar results were found in Curitiba, Paraná State, Brazil, by Bonamigo, Bonamigo, and Molento (2012) and in Fortaleza, Ceará State, Brazil, by Queiroz et al. (2014), who report the lack of knowledge of the majority population about the systems of creation of animal production. The lack of information is one of the main bottlenecks that makes it difficult to purchase qualified products with an animal welfare seal (Raineri et al., 2012).
Most consumers of different educational levels in Belém, Pará State, Brazil, had not heard of animal welfare (E -48.4.%, IE -44.1%, HS -63.4%, IHS -54.5%, HE -72.5%, and IHE -57.1%). People with incomplete high school were the ones who most demonstrated knowledge about the theme, with total answers "Yes" in 55.9%. The majority of consumers from all schools answered "No" -not to know about the existence of animal protection laws (E -51.6%, IE -85.3%, HS -72.8%, IHS -81.8%, HE -82.6%, and IHE -85.7%), so there is no distinction of knowledge, that is, having higher education does not mean knowing about every topic. Although there is currently great ease in obtaining information, through the media and the internet, it is noted that there is a need for greater interest from institutions on this important topic. It was also observed that many people prefer not to delve into this knowledge for psychological and moral reasons, as they state that when they learn about the suffering of production animals, they will feel "guilty" for consuming products of animal origin, however, many of them will not have financial means to replace these products on their tables, since it is very difficult to acquire animal protein from other foods, which they consider "more expensive".
For this reason, it was noted that respondents from all schools, in Belém, Pará State, Brazil, are not willing to pay more to obtain a product with a seal of animal welfare breeding (E -96.8%, IE -97.1%, HS -94.6%, IHS -100%, HE -98.2%, and IHE -100%) and would not like supermarkets to supply products, according to animal welfare standards (E -100%, IE -94.1%, HS -99.5%, IHS -100%, HE -100%, and IHE -100%). It is worth mentioning that only a minority of consumers of incomplete elementary education (5.9%) would like supermarket chains to supply products in accordance with animal welfare standards. In Porto Alegre, Rio Grande do Sul State, Brazil, however, only 17% of consumers were not willing to overpay for certification, as they understand that it is the country's obligation to ensure the rights linked to the quality seal and signal the high value of meat as an obstacle (Velho, Barcellos, Lengler, Elias, & Oliveira, 2009).
When assessing consumers' family income, a positive response was seen (Yes) from all respondents (1-3 MW -81.9%, 3-6 MW -56.6%, 6-9 MW -62.5%, 9-12 MW -60.9%, and above 12 MW -85%) only in question 2 (Figure 3), where most of the interviewees reported having knowledge related to the way of raising the production animals that they consume, showing superiority in consumers with higher income to 12 MW, with 85% of positive responses. Similar results were found in Fortaleza, Ceará State, Brazil, where class A dominated, that is, with a family income exceeding 12 minimum wages (Queiroz et al. (2014).
In questions 3, 4, 5, 6, 7, 8, 9, and 10 there were negative answers (No), thus, it can be observed in this research, that regardless of family income, consumers in Belém, Pará State, Brazil, are not concerned with the methods of raising production animals. However, consumers with an income between 6 and 9 minimum wages had a higher level of concern, with a total of "Yes" answers of 46.9%. Similarly, in Roraima, most consumers (63.78%) do not have knowledge about the creation of production animals (Pinheiro, Gomes, & Lopes, 2008). In Santarém, Pará State, Barzil, Silva et al. (2020) described that the majority of individuals interviewed have no interest or concern in knowing which method is used during the slaughter of the animals.
Consumers in the different salary ranges had not heard of animal welfare (1-3 wages -57.6%, 3-6 wages -77.1%, 6-9 wages -65.6%, 9-12 wages -56.5%, and above 12 wages -65%), however consumers who had between 9 and 12 minimum wages showed to have more knowledge on the subject. It was identified that consumers with a family income between 9 and 12 wages are the ones who heard most about the theme (43.5%) and consumers with a family income above 12 wages do not think that animals raised under animal welfare standards can generate good quality meat (100%), in agreement with the other salary ranges (1-3 wages 88.1%, 3-6 wages 95.2%, 6-9 wages 93.8%, and 9-12 wages 95.7%).
Most consumers, regardless of family income, would not be willing to pay for a product with a guarantee that animals were raised on the condition of animal welfare (1-3 wages -92.6%, 3-6 wages -94%, 6-9 wages -93.8%, 9-12 wages -100%, and over 12 wages -90%). However, it was noted that consumers with an income higher than 12 salaries showed more interest in this question, answering "Yes" with a total of 10%. It was found that the majority of consumers do not believe that production animals go through some kind of suffering during their breeding, affecting meat quality (1-3 wages -75.3%, 3-6 wages -79.5%, 6-9 wages -68.8%, 9-12 wages -65.2%, and over 12 wages -75%). However, consumers 34.8% with an income between 9 and 12 wages answered that they believe that animals go through some suffering during creation. In Niterói, Rio de Janeiro State, Brazil, Andrade, Cecchin, Pinto, Nepomuceno, and Silva (2019) found that 83.6% of consumers think that production animals go through suffering during slaughter.
It was found that, regardless of family income, most consumers were aware of animal welfare laws (1-3 MW -72%, 3-6 MW -77.1%, 6-9 MW -87.5%, 9 -12 MW -82.6%, and over 12 MW -85%), however, as they would not be willing to buy products with animal welfare certificate (1-3 MW -96.3%, 3-6 MW -94%, 6-9 MW -96.9%, 9-12 MW -100%, and over 12 MW -100%), and would not like supermarkets to supply products produced in accordance with welfare standards (1-3 SW -98.8%, 3-6 SW -100%, 6-9 SW -100%, 9-12 SW -100%, and over 12 SW -100%). Respondents with an income between 1 and 3 minimum wages (28%) are the most knowledgeable about welfare laws, and consumers with an income between 6 and 12 wages (100%) would be willing to pay a little more for products with animal welfare seal and were interested in the selling of these products in supermarkets (100%), a fact also found with consumers with an income over 12 wages (100%). Similar results were observed in Porto Alegre, Rio Grande do Sul State, Brazil, identifying that 68% of consumers would be willing to overpay, around 10% for quality meat and well-kept animals (Velho et al., 2009). Table 3 shows the result of the chi-square test (X 2 ) in the different age groups for each question. It was found that there was a difference (p < 0.05) between the groups in questions 2, 3, and 8. In question 2, it was identified that the G3 group aged between 40 and 50 years old had the highest percentage (74a, b consumers) of negative responses (No), demonstrating to have less knowledge about the animal welfare theme, differing from the groups G1 (18 to 30 years old) (33a consumers) and G2 (30 to 40 years old) (52b consumers), but with results similar to the group G4 (70a, b consumers) and G5 (68a, b consumers). Young consumers (18 to 30 years old) show a lower rate of negative (No) responses about knowledge of the subject, when comparing to other groups. Table 3. Distribution of frequencies obtained for questions 2, 3, 4, 5, 6, 7, 8, 9, and 10 according to the age group of consumers. No 0a 0a 0a 2a 1a Yes 36a 79a 98a 89a When evaluating the positive answers (Yes) on question 2, it was found that the group G5 (29a, b consumers) has more sufficient knowledge about how animals are raised, diverging (p < 0.05) from the group G1 (3a consumers) and G2 (27b consumers), and presenting similar results to the group G3 (24a, b consumers) and G4 (21a, b consumers), that is, younger consumers are those who have less knowledge about the subject. Acta Scientiarum. Animal Sciences, v. 43, e53784, 2021 Corroborating the study by Queiroz et al. (2014), who in Fortaleza, Ceará State, Brazil, identified greater knowledge about the theme among consumers aged 40 to 50 years, stating that about 60% of the interviewees did not hear about the theme.
Consumers over the age of 60 belonging to the G5 group (45a, b consumers) showed that they were not concerned (p < 0.05) with the methods of raising or slaughtering the animals. Young consumers G1 (26a consumers) had a lower level of concern, in addition, there were no differences between groups G2 (28b consumers), G3 (33b consumers), and G4 (36b consumers). Bonamigo et al. (2012) describe similar results, where they showed that the population of Curitiba, Paraná State, Brazil, does not know the ways of raising animals.
Analyzing the positive answers (Yes) on question 3, it was found that the G3 group (65b consumers) are the ones that are most concerned with the methods of raising or slaughtering farm animals (p < 0.05), diverging from the group G1 (10a consumers), and presenting similar results to the group G2 (51b consumers), G4 (55b consumers) and G5 (52a, b consumers), which indicates that younger consumers are the least knowledgeable about the thematic. This can be explained by the fact that the purchase of meat products is carried out mainly by their parents, who tend to obtain more information about the products that the family consumes.
The G3 group (28a, b consumers), aged between 40 and 50 years, demonstrated knowledge about animal welfare laws (p < 0.05), differing from the other groups G1 (16a consumers), G4 (14b consumers) and G5 (16b consumers), however did not differ from the G2 group (24a, b consumers), having similar level of responses. Consumers in the G4 group (50 to 60 years old) had a lower response rate when compared to the other groups. It was observed that questions 4, 5, 6, 7, 9, and 10 showed no differences (p > 0.05). Evaluating the "Yes" answers, in question 8, it was found that the G5 group (81b consumers), aged over 60 years, are those who have the most knowledge about animal welfare laws (p < 0.05), differing from the G1 group (20a consumers) and presenting similar results to the G4 group (77b consumers). Consumers in the G2 group (55a, b consumers) and G3 (70a, b consumers) had similar levels of knowledge, not diverging from each other, that is, younger consumers are the least knowledgeable about the laws that ensure animal welfare in Belém, Pará State, Brazil.
Consumers who had high school education (74a, b consumers) (p < 0.05) did not hear about animal welfare and do not believe that animals raised on animal welfare generate good meat quality (17a, b consumers), in addition to not being aware of animal welfare laws (55a, b consumers), when compared to the other groups (Table 4). Consumers with incomplete high school (5a, b consumers) were the least responders to have heard about the issue and have no knowledge of the laws that ensure animal welfare (2a, b consumers), as well as consumers of incomplete higher education (2a, b consumers). Respondents who had incomplete higher education (1a, b, c, and d consumer) are the ones who least answered "No" to believe that animals raised on animal welfare generate good quality meat. For the other questions and educational levels, there were no differences (p < 0.05) between them. When analyzing the indexes of "Yes" answers in the questioned questions, it was found that the interviewed consumers who had a high school education (128a, b consumers), said (p < 0.05) that they had heard more about the subject of animal welfare, as well as they believe that animals raised on welfare generate good meat quality (185a, b consumers), in addition to being aware of animal welfare laws (147a, b consumers). Consumers with incomplete high school (6a, b consumers), on the other hand, were those who least heard about the topic and are those who, to a lesser extent, were aware of the laws that ensure animal welfare (9a, b consumers), in addition to being the ones that least answered "Yes" when asking whether animals raised under animal welfare provide good quality meat. For the other questions and educational levels, no differences were observed (p < 0.05) between the averages.
In question 2, it was identified that consumers with an income between 1 and 3 minimum wages (199a, b consumers) had the highest number of negative responses "No" (p < 0.05), thus demonstrating less knowledge about the animal welfare theme and heard little about it, diverging widely from the other salary ranges (3-6 -64a, b, 6-9 -20a, b, 9-12 -14a, b, and over 12 -17a, b ) ( Table 5). On the other hand, consumers with incomes between 3 and 6 wages (47b consumers) differed (p < 0.05) from salary ranges 6 and 9 (20a, b consumers), 9 and 12 (14a, b consumers), and over 12 (17a, b consumers). However, among these groups there were no differences, so consumers with salary ranges 6 and 9, 9, and 12 and above 12 have similar knowledge. When evaluating the "Yes" answers obtained in question 4, it was found that consumers with an income between 1 and 3 minimum wages (140a, b consumers) are the ones who least heard about animal welfare. However, when looking at the total number of "Yes" answers, it was noted that this same salary range is higher in the positive answers on the question. This was due to the greater number of respondents having an income between 1 and 3 minimum wages. Among the other questions and age and wage groups, there were no differences (p > 0.05). Consumers with an income between 1 and 3 minimum wages (44a consumers) answered "Yes" in question 2, therefore they have greater knowledge about the welfare of production animals (p < 0.05), differing from the other salary ranges.
Conclusion
Most consumers of products of animal origin in Belém city, Pará State, Brazil, recognize that these products are part of their daily diet, but do not have knowledge on the subject of animal welfare, and are not willing to pay more for the product with a seal of quality, as well as, do not show interest or concern with the way in which livestock are bred or slaughtered. In addition, consumers in Belém are not aware of animal welfare laws and are not concerned with how animals are raised or slaughtered, a fact that is mainly seen among young people. There is influence of the age group, however, education is a decisive factor in determining the level of knowledge of consumers of products of animal origin, in Belém, Pará State, Brazil, about the welfare of production animals. It is clear, therefore, the need for more research, disciplines in | 2021-11-07T16:04:40.601Z | 2021-10-11T00:00:00.000 | {
"year": 2021,
"sha1": "cf2eea325d21637174ad30c018e66f08345290d5",
"oa_license": "CCBY",
"oa_url": "https://periodicos.uem.br/ojs/index.php/ActaSciAnimSci/article/download/53784/751375152840",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "b5fe16f6c1ed69cf5e6975243ff1bdcab78ab0a0",
"s2fieldsofstudy": [
"Agricultural and Food Sciences"
],
"extfieldsofstudy": []
} |
247776685 | pes2o/s2orc | v3-fos-license | Acute haemodynamic impact of transcatheter aortic valve implantation in patients with severe aortic stenosis
Abstract Aims There are limited data about the intraprocedural haemodynamic study performed immediately before and after transcatheter aortic valve implantation (TAVI) in patients with severe aortic stenosis (AS). We aimed to evaluate the acute haemodynamic impact of TAVI in patients with severe AS and to investigate invasive and non‐invasive parameters predicting all‐cause mortality. Methods and results A total of 245 consecutive AS patients (43% male, mean age 80.3 ± 7.3 years) undergoing TAVI were enrolled. Intraprocedural left heart catheterization (LHC) and echocardiogram before and after TAVI were performed. The clinical endpoint was the death for any cause. LHC after TAVI revealed significant changes in aortic and left ventricular (LV) pressures, including indexes of intrinsic myocardial contractility and diastolic function such as positive dP/dT (1128.9 ± 398.7 vs. 806.3 ± 247.2 mmHg/s, P ˂ 0.001) and negative dP/dT (1310.7 ± 431.1 vs. 1075.1 ± 440.8 mmHg/s, P ˂ 0.001). Post‐TAVI echo showed a significant reduction in LV end‐diastolic (P = 0.036) and end‐systolic (P ˂ 0.001) diameters, improvement in LV ejection fraction (from 55 ± 12% to 57.2 ± 10.5%, P ˂ 0.001), and pulmonary artery systolic pressure (42.1 ± 14.2 vs. 33.1 ± 10.7 mmHg, P < 0.001). After a mean follow‐up time interval of 24 months, 47 patients died. Post‐TAVI significant aortic regurgitation at echocardiography was the only independent predictor of mortality (hazard ratio 5.592, confidence interval 1.932–16.184, P = 0.002). Conclusions Left heart catheterization performed immediately before and after prosthesis release offers a unique insight in the assessment of LV adaptation to severe AS and the impact of TAVI on LV, catching changes in indexes of intrinsic contractility and myocardial relaxation. Aortic regurgitation assessed by echocardiography was the only independent predictor of mortality in patients undergoing TAVI.
Introduction
Aortic stenosis (AS) is the most common heart valve lesion, with an incidence rate of severe AS of~4% to 7% per year among people ≥65 years of age. 1 For decades, the only available treatment for severe AS has been surgical aortic valve replacement (SAVR), which significantly improves survival, symptoms, and quality of life. 2 However, as much as 30-40% of patients are deemed unsuitable for SAVR. 3 Lately, the traditional treatment paradigm for patients with severe AS has been challenged by the development of transcatheter aortic valve implantation (TAVI) procedure. [4][5][6][7][8][9] TAVI profoundly changed the management of patients judged at prohibitive surgical risk, including those with severe heart failure (HF). Patients with AS and HF benefit from this procedure in terms of symptom alleviation, and a post-TAVI favourable prognosis has been observed, particularly in patients with HF with preserved or mildly reduced ejection fraction (EF) compared with patients with HF and reduced EF. 10 TAVI leads to cardiac reverse remodelling, although there are conflicting results about the regression of left ventricular (LV) hypertrophy as well as about the improvement of LV systolic and diastolic function. [11][12][13][14] There are limited data regarding the invasive haemodynamic evaluation of patients undergoing TAVI and even less about the prognostic role of the haemodynamic study performed immediately before and after prosthesis deployment. [15][16][17][18][19][20] Most of the available invasive studies focused on right heart catheterization, [17][18][19][20] and the few studies performed with left heart catheterization (LHC) focused on the role of residual aortic regurgitation (AR) as predictor of long-term mortality. 21,22 The aims of our study were (i) to assess the acute effect of TAVI on LV haemodynamics as well as the early impact on echocardiographic parameters and (ii) to identify the stronger predictor of all-cause mortality after TAVI.
Study design
The present study is a clinical retrospective study that reports data prospectively collected in a large registry that includes all consecutive patients undergoing TAVI in our centre.
We retrospectively selected consecutive patients who underwent TAVI between January 2016 and March 2019 at our institution. The study population included symptomatic patients with severe AS defined according to guidelines. 8 The exclusion criteria were TAVI performed for pure AR and valve-in-valve procedures.
Transcatheter aortic valve implantation procedure
The management of patients with AS in our centre is routinely performed according to internal guidelines (in the institutional clinical pathway dedicated to patients with heart valve diseases: https://www.policlinicogemelli.it/servizipaziente/percorsi-assistenziali/valvulopatia/). 23 All TAVIs were conducted according to a procedural plan performed on the basis of computed tomography (CT) scan. For each patient, the CT scan was revised by at least two operators to assess the potential suitability for TAVI implantation.
Transfemoral approach was considered the preferred option, other accesses being considered in the case of absence of suitable aorto-iliac-femoral anatomy due to insufficient lumen size, extreme tortuosity, and/or severe atherothrombosis. All TAVIs were performed following the manufacturer's best practice recommendations. Details of the transfemoral TAVI technique have been recently published. 24
Pre-transcatheter aortic valve implantation and post-transcatheter aortic valve implantation evaluation
Each patient underwent a thorough clinical and echocardiographic evaluation before the procedure according to the standard practice of our centre, and all patients were referred for TAVI on the basis of formal, multidisciplinary, Heart Team discussion. Clinical data and procedure details were prospectively entered into a dedicated database that allowed previously to assess the impact of EuroSCORE on coronary interventions 25 and the safety of trans-radial procedures. 26 Patients' surgical risk was graded according to the Society of Thoracic Surgeons (STS)-predicted operative mortality at the time of Heart Team consultation. TAVI risk was graded according to the STS/American College of Cardiology Transcatheter Valve Therapy 27 using the online TAVI in-hospital mortality risk calculator (https://tools.acc.org/tavrrisk/ #!/content/evaluate/). In-hospital clinical outcomes were prospectively recorded because the continuous monitoring of in-hospital clinical outcomes for TAVI is part of our institutional clinical pathway dedicated to patients with heart valve diseases. The study was in compliance with the Declaration of Helsinki, and all patients signed a dedicated informed consent to the study procedure, which included the authorization to database insertion and clinical follow-up assessment.
In the pre-TAVI evaluation, the following data were collected: • scheduled procedure and a second evaluation within 2-5 days after TAVI. • Computed tomography: all patients underwent CT for work-up and pre-procedural planning before TAVI for the selection of prosthesis type sizing as well as for the evaluation of vascular access site. Clinical follow-up was available for all the patients enrolled, and data were obtained by telephone or in an outpatient visit to assess vital status.
All the echocardiograms were performed by experienced physicians (F. G., G. L., and A. L.) from our heart valve clinic. Images were acquired with patients at rest in the left decubitus position using commercially available ultrasound systems (Toshiba Artida, Toshiba, Tokyo, Japan; Philips Epiq 7, Philips Medical Systems, Andover, Massachusetts, USA) equipped with 3.5 MHz or M5S transducers. Twodimensional, colour, pulsed-wave (PW), and continuous-wave Doppler data were obtained in parasternal and apical views. Apical two-chamber and four-chamber views were used for quantification of LV end-diastolic and end-systolic volumes, and LVEF was calculated using Simpson's biplane method. LV dimensions and wall thicknesses were measured on M-mode recordings of the parasternal long-axis view. All acoustic windows were explored in order to obtain the best ultrasound beam alignment for Doppler recording of the aortic valve (AV): aortic mean and peak gradients were evaluated using continuous-wave Doppler with the simplified Bernoulli equation. AS was assessed using a multi-parametric approach as suggested by the European Association of Cardiovascular Imaging (EACVI)/American Society of Echocardiography (ASE) guidelines. 29,30 Heart chamber dimension classification, as well as LV mass index (LVMi) for men and women, and the grading of diastolic dysfunction and associated valvular heart disease were all carried out in accordance with the most recent EACVI/ASE guidelines. 29,30 Regarding diastolic function assessment, pertinent echocardiographic variables were obtained as follows: early flow peak velocity (E), E-wave deceleration time (DT), and peak velocity during atrial systole (A) were measured from trans-mitral flow patterns obtained between the mitral leaflets using PW Doppler in the apical four-chamber view. Early diastolic annular velocity (e0) of the lateral and septal LV wall was measured by PW tissue Doppler imaging. Left atrial volume was calculated using Simpson's rule. Pre-procedural LV diastolic dysfunction was classified into three grades (I, II, and III) based on 2016 ASE/ EACVI Recommendations for Evaluation of Left Ventricular Diastolic Function by Echocardiography. 31 Finally, we applied the staging classification recently proposed by Généreux et al. 32 for the characterization of the extent of cardiac damage among patients with severe AS.
Aortography and LHC were performed before and after valve deployment. LHC was performed using two 5F or 6F diagnostic catheters positioned in the left ventricle (through the main TAVI access site) and in the ascending aorta (through the radial or femoral ancillary arterial access). LV and aortic pressure-time curves were simultaneously recorded and stored on our Hemodynamic Recording System (Mac-Lab, GE Medical Systems Informa Technologies, Milwaukee, Wisconsin, USA). The values of systolic and diastolic pressures were collected, and, after data processing, systolic and diastolic intervals and pressure gradients between the LV and the aorta were calculated. Figure 1 shows an example of LHC (with the evaluated parameters before and after TAVI). AR index was calculated as previously described 33 before and after TAVI and the AR index ratio 34 derived for each patient.
Statistical analysis
Categorical variables were expressed as percentages and analysed by χ 2 or Fisher's exact test as appropriate. Continuous variables were expressed as mean ± standard deviation and/or median [inter-quartile range] and compared using the paired t-test or the non-parametric Wilcoxon test, as appropriate. Normality of distribution of continuous variables was assessed by Kolmogorov-Smirnov test. Kaplan-Meier curves were constructed for mortality. Comparison of the variables between patients who survived and those who died was performed. To assess the independent predictors of survival, a Cox regression univariate and multivariate analysis model was realized. Variables with a P value lower than 0.1 in the univariate analysis were included in the backward stepwise multivariate analysis. Differences were considered significant with P < 0.05. Data were analysed using IBM SPSS Statistics Version 24.0 software (IBM, Armonk, NY, USA).
Results
Out of the 267 patients who underwent TAVI during the period considered, 245 patients undergoing TAVI for severe AS of the native valve were selected and constituted the study population.
The TAVI procedure was successful in 244/245 patients (99.6%). In one case, only the procedure was not completed because of the occurrence of ascending aortic rupture so that conversion to urgent SAVR was successfully performed. After TAVI, 26 (10.6%) patients needed a permanent pacemaker implantation.
Haemodynamic parameters obtained before and after the procedure are reported in Table 2. LHC data were incomplete/missing in eight patients due to insufficient quality traces.
As expected, AV peak and mean gradient significantly decreased after TAVI (P < 0.001), as well as AV closure pressure (P = 0.001) and time to aortic peak pressure (P < 0.001), while aortic peak pressure increased (P < 0.001). The relief of abnormal afterload after the deployment of the prosthesis was followed by a significant reduction of LV peak pressure (P < 0.001) and LV positive dP/dT (P < 0.001, Figure 2A). While end-diastolic pressure did not significantly change after the procedure (P = 0.107), negative dP/dT decreased after TAVI (P < 0.001, Figure 2B) and LV proto-diastolic pressure increased (P = 0.001).
After TAVI, as expected, mean aortic gradient decreased (from 52 ± 14.4 to 8.9 ± 4.8 mmHg, P < 0.001), but, more interestingly, there was a significant reduction in LV diameters and volumes (respectively: LV end-diastolic diameter from 48.4 ± 7.5 to 47.4 ± 6.9 mm, P = 0.036; LV end-systolic diameter from 33 ± 8.6 to 30.1 ± 8.1 mm, P < 0.001; and LV end-diastolic volume index from 54.6 ± 18.4 to 51.7 ± 17.5 mL/m 2 , P = 0.017; Figure 3A) and a significant improvement in LV systolic function with increase in post-procedural LVEF (from 55 ± 12% to 57.2 ± 10.5%, P < 0.001; Figure 3B). In particular, patients with worse pre-TAVI systolic function showed significant improvement in LVEF compared with patients with preserved pre-procedural LVEF (ΔLVEF respectively 6% vs. 1.6%, P < 0.001). Moreover, pulmonary artery systolic pressure decreased significantly [pulmonary artery systolic pressure (PASP) from 42.1 ± 14.2 to 33.1 ± 10.7 mmHg, P < 0.001; Figure 3C]. Acute impact of TAVI in aortic stenosis Interestingly, TAVI procedure was found to significantly impact on Généreux's classification: as shown in Table 3, after TAVI, the Généreux classification stages significantly changed (P < 0.001), and this change was mainly driven by a major reduction of Stage 3 patients. We found no major differences in diastolic function indexes before and after TAVI; none of the patients with Grade III diastolic dysfunction before TAVI showed an improvement after the procedure. A borderline statistical significance was reached for the decrease of patients with moderate-to-severe mitral regurgitation (before vs. after TAVI: 6.3% vs. 3.5%, P = 0.05). Neither of the patients with atrial fibrillation before TAVI showed spontaneous restoration of the sinus rhythm nor anyone underwent cardioversion.
After a mean follow-up time interval of 24 months (range 3-50 months), 47 patients died. Table 4 shows the results of survival Cox regression analyses for clinical, echocardiographic, and haemodynamic variables, in order to investigate possible predictors of mortality.
Discussion
The present study analysed invasive data obtained from LHC before and soon after TAVI as well as early echocardiographic changes after the procedure, offering a unique insight in the pathophysiology of severe AS and the impact of TAVI. Moreover, we investigated the prognostic significance of haemodynamic and echocardiographic parameters before and after TAVI.
Our relevant findings are as follows: (i) global haemodynamics, afterload, and contractility change immediately after TAVI; (ii) an early LV remodelling is appreciable at echocardiography after TAVI; and (iii) post-TAVI significant AR at pre-discharge echocardiography is the only independent predictor of mortality after an interval of 24 months.
Despite several studies aiming to compare non-invasive (echocardiographic and cardiac magnetic resonance) pre-TAVI and post-TAVI data have been performed, poor data are available on the role of the haemodynamic assessment performed immediately before and after prosthesis deployment in patients with severe AS. 15,16,35,36 The few studies available on this topic have mainly focused on the variations of LV end-diastolic pressure (LVEDP). 36 In our study, LVEDP did not show any significant change after TAVI, while LV proto-diastolic pressure increased. Previous studies showed that, after deployment of the prosthesis, the LVEDP tends to reduce significantly (>5 mmHg). 36 However, other studies reported that immediately after TAVI, LVEDP may even increase. 15 This may be explained by the fact that during TAVI procedure, rapid ventricular pacing (RVP) is required to produce a transient decrease in cardiac and catheter motion. The transient cycle of hypotension and low cardiac output induced by repeated RVP may cause LVEDP elevation; moreover, deterioration in coronary perfusion induced by RVP 15 may have been partly responsible for the temporary myocardial stunning and the non-reduction in LVEDP observed in our study. Finally, the administration of intravenous fluid infusion and iodinated contrast agent during procedure could have had a role in the lack of reduction of LVEDP. Even if LVEDP did not change significantly, we observed a decrease of negative dP/dT, an intrinsic index of diastolic function.
We performed a comprehensive assessment of haemodynamic changes after TAVI, including LV systolic function, so far poorly understood. Beyond the expected reduction of AV peak and mean gradient, driving the reduction of LV peak pressure, several parameters significantly changed after the procedure. Interestingly, positive dP/dT, an index of myocardial contractility, decreases after TAVI. We can speculate that patients with AS and preserved LVEF use efficiently the contractile reserve in order to overcome the increased afterload; once the key pathophysiological element of the AS is fixed, that is, the increased valvular load, indexes of intrinsic contractility decrease as well as those of myocardial relaxation (i.e. negative dP/dT). Our results are in line with the recent findings of Seppelt et al. 16 They performed (in a pilot study on eight patients) an invasive LV pressure volume loop analysis and found impaired systolic and diastolic function in the early phase after TAVI, even if ventricular-arterial coupling Acute impact of TAVI in aortic stenosis data suggested an early improvement of global cardiovascular energy efficiency. Moreover, at multivariate analysis, we found that higher values of pre-TAVI aortic peak pressure and post-TAVI LV peak systolic pressure were weakly associated with reduced mortality. Several previous studies demonstrated that the acute relief of the increased afterload by TAVI leads to immediate improvement of cardiac function assessed with echocardiography, followed by reverse heart remodelling with regression of LV volume and mass as well as improvement of LV systolic and diastolic function. 11,[37][38][39] In line with previous findings, we found major LV remodelling in the early post-TAVI echo, with significant reduction in LV diameters and volumes together with significant improvement in LV systolic function, expressed as an increase in post-procedural LVEF. In particular, patients with worse pre-TAVI systolic function are those who benefit the most in terms of LVEF recovery. We observed an immediate reduction of PASP after TAVI, together with a slight reduction of tricuspid annular plane systolic excursion, whose values however stayed within normal range. Interestingly, Généreux staging classification, a well-recognized tool to objectively characterize the extent of cardiac damage associated with AS, significantly changed after TAVI, reflecting the early cardiac remodelling and pulmonary pressure reduction soon after the procedure. When investigating the prognostic significance of LHC and echocardiographic data before and after TAVI, we found that the only significant independent predictor of mortality was significant AR assessed at pre-discharge echocardiography. This finding is in line with previous studies. 21,22,40 Indeed, in a multicentre registry by Zahn et al. 40 on 1444 patients treated with TAVI, one of the strongest predictors identified as modifiable was residual AR ≥ II°. However, other variables independently associated with mortality were identified, among which atrial fibrillation and prior decompensation. The differences found with Zahn et al. could rely on the different size of the study population; the larger sample size likely allowed them to identify more than one independent predictor of prognosis. However, we confirmed the prognostic role of post-TAVI AR, and in light of these considerations, patients with moderate-to-severe AR (3+ and 4+) pre-discharge must be considered high-risk group, who would probably benefit of a closer follow-up and personalized therapeutic management.
Limitations of the study
Some limitations of the study should be acknowledged: first, its retrospective observational single-centre design with wide follow-up range. This might have had an impact on results, which need to be confirmed in larger studies. Second, despite the collection of systematic haemodynamic data, some variables were not collected (such as pharmacological treatment including hypertensive drugs and diuretics), and we cannot exclude that they might have had an impact. Moreover, right heart catheterization was not performed.
Conclusions
Left heart catheterization performed immediately before and after prosthesis release offers a unique pathophysiology insight in the assessment of LV adaptation to severe AS and the acute impact of TAVI.
Our data suggest that patients with AS and preserved LVEF use efficiently the contractile reserve in order to overcome the increased afterload; once the key pathophysiological element of the AS is fixed, that is, the increased valvular load, indexes of intrinsic contractility, as well as those of myocardial relaxation, decrease.
Aortic regurgitation is the only independent predictor of mortality in patients undergoing TAVI. | 2022-03-30T06:17:48.442Z | 2022-03-29T00:00:00.000 | {
"year": 2022,
"sha1": "94475eb1b55717c3b963379d6ba5d45e16a5fe4f",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Wiley",
"pdf_hash": "211e37320ff3e1b3a5fd913fc5685bb174ad3473",
"s2fieldsofstudy": [
"Medicine",
"Biology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
10060417 | pes2o/s2orc | v3-fos-license | An optimal search filter for retrieving systematic reviews and meta-analyses
Background Health-evidence.ca is an online registry of systematic reviews evaluating the effectiveness of public health interventions. Extensive searching of bibliographic databases is required to keep the registry up to date. However, search filters have been developed to assist in searching the extensive amount of published literature indexed. Search filters can be designed to find literature related to a certain subject (i.e. content-specific filter) or particular study designs (i.e. methodological filter). The objective of this paper is to describe the development and validation of the health-evidence.ca Systematic Review search filter and to compare its performance to other available systematic review filters. Methods This analysis of search filters was conducted in MEDLINE, EMBASE, and CINAHL. The performance of thirty-one search filters in total was assessed. A validation data set of 219 articles indexed between January 2004 and December 2005 was used to evaluate performance on sensitivity, specificity, precision and the number needed to read for each filter. Results Nineteen of 31 search filters were effective in retrieving a high level of relevant articles (sensitivity scores greater than 85%). The majority achieved a high degree of sensitivity at the expense of precision and yielded large result sets. The main advantage of the health-evidence.ca Systematic Review search filter in comparison to the other filters was that it maintained the same level of sensitivity while reducing the number of articles that needed to be screened. Conclusions The health-evidence.ca Systematic Review search filter is a useful tool for identifying published systematic reviews, with further screening to identify those evaluating the effectiveness of public health interventions. The filter that narrows the focus saves considerable time and resources during updates of this online resource, without sacrificing sensitivity.
Conclusions:
The health-evidence.ca Systematic Review search filter is a useful tool for identifying published systematic reviews, with further screening to identify those evaluating the effectiveness of public health interventions. The filter that narrows the focus saves considerable time and resources during updates of this online resource, without sacrificing sensitivity.
Background
Systematic reviews have been integral to the evidenceinformed practice movement [1][2][3][4][5] in the field of public health [6][7][8][9]. A systematic review consists of an examination of all of the primary studies on a topic, which includes searching for, collating, and assessing the studies, to establish conclusive evidence about a topic [10]. Systematic reviews present a more consistent and conservative estimate of the effect of interventions across a body of literature and as such, can have an important impact on program planning decisions in public health.
However, public health decision makers state that finding and accessing systematic reviews related to public health continues to be a barrier to evidence-informed public health practice [11][12][13][14][15][16]. The field of public health can be defined as a combination of sciences, skills, and values that function through collective societal, legislative, and political activities. It involves both public and private programs, services, and institutions aimed at protecting and improving the health of all people, including preventing disease, promoting health and wellbeing, and prolonging life. When necessary, public health also engages in restoring the health of individuals, specified groups, populations or communities through mobilizing and engaging local, state, national, and international resources to assure the conditions in which people can be healthy [17][18][19]. In short, the field of public health is broad, and decision makers wear many hats, requiring evidence on a wide range of topics.
Public health practitioners have expressed a need for a single place where they can access reviews evaluating the effectiveness of interventions, have confidence in the methodological quality of the evidence, and access plain language review summaries with corresponding implications for policy and practice [20]. Health-evidence.ca is a free, searchable online registry of systematic reviews and meta-analyses evaluating the effectiveness of public health and health promotion interventions. This registry represents one component of a larger knowledge translation and exchange (KTE) [21] strategy that supports users in accessing and interpreting research evidence. KTE is a two-way process involving dialogue, interaction, and the sharing of knowledge and evidence between and among the producers and users of knowledge and research evidence. It is a broad term that is often used to include knowledge transfer, exchange, translation, dissemination, and diffusion. The target audience for health-evidence.ca is decision makers working in public health and health promotion at all levels (front line practitioners to senior management and policy makers in government). Public health decision makers need to find, assess and interpret research evidence quickly and easily if it is to inform program and policy decisions. Health-evidence.ca provides decision makers with easy access to public health-relevant, qualityappraised systematic reviews evaluating the effectiveness of public health interventions. The site is freely accessible and can be searched by selecting common public health indexing terms. Search results include links to published review abstracts and a rating of the methodological quality of each review. In addition, healthevidence.ca team members write evidence summaries for reviews of good methodological quality to summarize key findings and provide recommendations for policy and practice. A more complete description of this online resource has been published and is accessible at http://www. biomedcentral.com/1471-2458/10/496. Health-evidence.ca was updated quarterly until 2012 and is now updated on a monthly basis. Updates consist of conducting monthly searches of relevant electronic databases, importing results into a bibliographic database management program, screening titles to identify relevant articles, retrieving potentially relevant articles and screening full document versions for inclusion. Included reviews must meet relevance criteria and must be systematic reviews that focus on public health, provide outcome data on the effectiveness of interventions, and include a documented search strategy.
As of February 2012, over 1,017,500 titles had been screened, yielding 2,450 relevant reviews. The large number of titles screened to reach the final, relevant set reflects the challenges of searching bibliographic databases for public health and health promotion literature. These challenges stem from the lack of a single database dedicated exclusively to public health and health promotion literature, requiring searches in multiple health (MEDLINE, EMBASE, CINAHL), science, and social science databases (BIOSIS, PsycINFO, SPORTDiscus, Sociological Abstracts). There are also several limitations inherent in searching these databases. For example, 33-44% of the journals identified by experts in the field as public health journals are not indexed in MEDLINE. These challenges are not limited to public health as others have encountered similar difficulties in searching for mental health content [23] and health services research literature [24]. A further challenge is identifying what is relevant to public health and health promotion practitioners, given that it is a dynamic field characterized by a wide scope of practice, defined regionally and changing constantly.
Along with the challenges of searching for public health and health promotion content, review literature, though rapidly growing, remains limited in volume when compared to primary studies. For example, over 700,000 articles were indexed in MEDLINE in 2010, of which approximately 2500 (0.36%) were health-related systematic reviews [25]. Currently, there is no single MEDLINE subject heading term for 'systematic review'; this lack of an indexing term requires the end user to employ a Clinical Query developed to locate systematic reviews, or to screen very large sets of irrelevant articles in order to retrieve systematic reviews. MEDLINE does have an indexing term for 'review' however its application is very broad. Of the 19,430,768 articles currently indexed in MEDLINE as of February 13, 2012, 8.5% (1,656,583) [26] were indexed as reviews. Upon screening a small portion of this results set, it was evident that the majority were not systematic reviews, but rather literature reviews and overviews. While the MEDLINE indexing term 'metaanalysis' is useful for identifying systematic reviews, it only captures systematic reviews that use statistical software to combine the results of the included primary studies in a single pooled estimate of effect. However, meta-analyses represent a small portion of all reviews evaluating the effectiveness of public health interventions. For example, fewer than half of public health intervention reviews indexed on health-evidence.ca are meta-analyses, thus reliance on this text word to identify reviews is not sufficient. A combination of indexing terms is required to detect relevant reviews that can be captured in online databases such as MEDLINE. Thus, although it has been time-consuming, screening a high number of irrelevant articles has been necessary. Search filters, also referred to as "search hedges", are "collections of search terms intended to capture frequently sought research methods such as randomized controlled trials, or other aspects of health care" [27]. While search filters for the retrieval of systematic reviews were being used by others for searching MEDLINE [19][20][21][22][23][24][25][26][27][28][29][30][31], EMBASE [32], and CINAHL [33], none had been used and tested for locating public health and health promotion reviews that we were aware of at the time of this project. These filters, including those targeting contentspecific literature relevant to the subject of interest [24,25], provided guidance as we developed a systematic review filter for health-evidence.ca.
Prior to 2008, we used a Public Health (PH) search filter that was developed in collaboration with health science librarians at McMaster University. The Head of Public Services worked with one of the authors (KD) to systematically run and informally evaluate the results of various search strategies for retrieving systematic reviews and meta-analyses evaluating the effectiveness of public interventions in MEDLINE, EMBASE, CINAHL, PsycINFO, and Sociological Abstracts. Search strategies were assessed and improvements made based on findings. The resulting PH search filter consisted of two distinct components: 1) indexing terms and keywords referring to systematic review methods, combined with the Boolean 'OR' operator (systematic, meta analysis, review); and 2) indexing terms and keywords referring to public health content areas, combined with the Boolean 'OR' operator (community health services, education, health education, health promotion, prevention, preventive). The content and methods components were then combined using the Boolean ' AND' operator. Seventeen topic areas were included in the content component: addiction, adult health, chronic diseases, communicable disease and infection, community health, dental health, environmental health, food safety and inspection, injury prevention and safety, mental health, nutrition, parenting, physical activity, pregnancy, sexual education, sexually transmitted infections, and women's health. This search strategy also made it more likely that we would capture articles for which established indexing terms did not exist such as social determinants of health and healthy communities.
Our PH search filter typically yielded a very high volume of results with very low precision. For example, between January 2006 and December 2007, of the 136,427 titles screened, 409 were relevant for the healthevidence.ca registry, or in other words, precision was 0.3%. In addition to using the PH search filter, more than 40 public health-relevant journals were hand searched annually, as well as the reference lists of all relevant reviews. Given this systematic search of the published review literature, we were reasonably confident that our retrieval methods were capturing a near complete set of relevant articles. We considered this set (the electronic database searches plus additional search strategies), the 'gold standard' for health-evidence.ca. A gold standard is "a set of relevant records against which a new search filter is tested and validated to determine how effective it is at retrieving particular types of records" [34]. While it is impossible to prove that the gold standard for health-evidence.ca identified all public health relevant systematic reviews, we are confident that this approach captured the vast majority of relevant reviews.
Given that the precision of the PH search filter was so low, we began to create an effective search filter that would decrease the total number of results retrieved, while maximizing the number of relevant results. The health-evidence.ca Systematic Review (SR) search filter we developed in 2008 was adapted from a previouslyvalidated filter [30], which included the terms: MED-LINE.tw, systematic review.tw, meta-analysis.pt, combined with the Boolean OR operator. While this filter was highly specific, it captured less than 82% of articles identified by our gold standard set. To customize this filter to retrieve only those systematic reviews of interventions, the term 'intervention' was added as an indexing term. This is referred to as the development data set.
The MEDLINE version of our health-evidence.ca SR search filter included the following indexing terms, combined with the Boolean 'OR' operator: MEDLINE.tw, systematic review.tw, meta-analysis.pt, intervention$.ti. We slightly modified the filter for use in EMBASE and CINAHL due to differences in indexing terms between the various databases. The indexing terms systematic review.tw and intervention$.ti are viable in both EMBASE and CINAHL, therefore these terms were consistent across all three databases. However, in both EMBASE and CINAHL, meta-analysis was not an indexed publication type, and therefore the term meta-analysis was included as a keyword in the search filter for these two databases. Each database employs a unique controlled vocabulary, thus the search strategy is tailored to the database. For example, MEDLINE does not have a preferred search term for systematic review so that concept must be searched as a text word. EMBASE and CINAHL, however, do have a specific indexing term for systematic review, so that term is used when tailoring the search to those databases.
The objective of this paper is to report the results of our efforts to evaluate and validate the health-evidence.ca SR search filter for retrieving systematic reviews and meta-analyses that evaluate the effectiveness of interventions. First, we compared the performance of the healthevidence.ca SR search filter to the PH search filter. We then compared the health-evidence.ca SR search filter to other known search filters targeted at capturing systematic reviews in existence at the time (Tables 1, 2 and 3).
Our intent was to identify a search filter that resulted in the optimal use of time and resources in updating the health-evidence.ca registry. Specifically, this paper reports the performance of each filter with respect to sensitivity, specificity, precision, and the number needed to read. The best option for our purposes is one that achieves high precision while not compromising sensitivity.
Methods
The health-evidence.ca SR search filter was evaluated and validated in two distinct ways.
Health-evidence.ca SR search filter vs. PH search filter
We compared the retrieval performance of the healthevidence.ca SR search filter in MEDLINE, EMBASE, and CINAHL with what we had retrieved using the gold standard, for both our development and validation data sets. The results are reported in Table 4. To test our health-evidence.ca SR search filter, we selected sub-sets from our gold Four indices were used to evaluate filter performance: sensitivity, specificity, precision and "number needed to read (NNR)". Sensitivity is a measure of the proportion of actual positives which are correctly identified. We defined sensitivity as the proportion of systematic reviews identified by the gold standard that were also identified by each search filter. Sensitivity was calculated as: number of systematic reviews retrieved by a search filter relevant number of articles in the gold standard Â100 The higher the sensitivity, the more successful the search filter was in capturing a large number of the articles, in comparison to the gold standard, with 100% meaning there was perfect agreement between the search filter and the gold standard.
Specificity is a measure of the proportion of negatives which are correctly identified. We defined specificity as the proportion of irrelevant articles not retrieved by the search filters. Specificity was calculated as: number of non À relevant articles not retrieved by a search filter total number of records that are not relevant systematic reviews  100 Specificity is a reflection of how well a search filter omits non-relevant articles from the retrieved set, which in this case were articles that were not systematic reviews. The specificity score declines if a search filter retrieves an article that it deems to be relevant when, in fact, it is not (a false positive). A specificity of 100% means that the filter recognized all actual non-relevant articles; no articles were retrieved that were not relevant systematic reviews.
Precision (or positive predictive value) is the proportion of retrieved articles that represent relevant articles and can be calculated as: If a search filter has a high degree of precision, it can locate a high number of relevant articles while keeping the number of non-relevant articles retrieved low. A good precision score (N = 1.0) indicates that a high proportion of all articles retrieved for a particular search were actually relevant. In other words, if a search identified 10,000 articles of which 100 were relevant, the precision score would be 0.01, which would be low precision.
Finally, the NNR represents the number of articles that must be read before a relevant article is identified.
Number needed to read ¼ 1 precision For example, if the NNR was 16, then for every 16 articles identified by the search filter and read, one would be deemed relevant.
Results
Fifty-three relevant articles were identified in the development data set between January 1 and December 31, 2001. Of those 53 relevant reviews, all 53 were published in MEDLINE, 33 in EMBASE and 36 in CINAHL (see Table 4), with some overlap of the same articles being published in more than one of the databases. The initial set of 53 results (development data set) used to test and develop the search strategy was used to explore the sensitivity, specificity, precision, and NNR for both the PH and health-evidence.ca SR search filters.
The second set of 219 results (validation data set), represented a sub-set of the gold standard and was made up of relevant articles indexed in each of the 3 databases of interest between January 1, 2004 and December 31, 2005. Of the 219 articles, 207 were indexed in MED-LINE, 107 in EMBASE, and 129 in CINAHL, again with overlap of the same articles being published in more than one of the databases. During that same time period, a total of 1,174,817 records were indexed in MEDLINE, 990,862 records in EMBASE, and 272,264 records in CINAHL (see Table 4).
Specificity
In addition to being sensitive, the health-evidence.ca SR search filter demonstrated a slightly higher degree of [30,31,[35][36][37] offered relatively high sensitivity (85.5%-88.9%) combined with good performance on specificity (98.5%-99.2%), precision (1.1-1.9), and number needed to read (52.0-94.9). Table 2 describes the results of the health-evidence.ca SR search filter in comparison to the seven other search filters tested in EMBASE. The health-evidence.ca SR and Scottish Intercollegiate Guidelines Network [37] search filters performed the best overall in terms of the combination of outcomes for sensitivity (87.9% and 81.3%), specificity (98.2% and 99.0%), precision (0.5 and 0.8) and NNR (186.0 and 118.6). The health-evidence.ca SR search filter, while having greater sensitivity, resulted in an additional 67 articles having to be read in comparison to the Scottish Intercollegiate Guidelines Network filter.
EMBASE
Sensitivity The health-evidence.ca SR search filter's sensitivity of 87.9% was slightly lower than that of the two top performing search filters which both obtained sensitivity scores of 96.3% (Wilcynski and Haynes, Sensitive query; Wilcynski and Haynes, Best optimization query).
Specificity All but the Wilcynski and Haynes (2007) search filter (sensitive query) achieved a level of specificity above 85%, with the health-evidence.ca SR search filter achieving 98.2%. The health-evidence.ca SR search filter was outperformed by the two Wilcynski and Haynes filters (99.3% for the 'Small drop in specificity, substantive gain in sensitivity' query, and 99.5% for the specific query), the BMJ Best Clinical Evidence filter (98.5%), and the Scottish Intercollegiate Guidelines Network filter (99.0%).
Precision The most precise filter had a score of 1.1 (Wilcynski and Haynes, 'Small drop in specificity, substantive gain in sensitivity' query) while retaining a high level of sensitivity (75.7%). The health-evidence.ca SR search filter offered moderate precision (0.5) in comparison.
Number needed to read The best performing filters for NNR were SIGN, BMJ Clinical Evidence filter, and the health-evidence.ca SR search filter at 118, 167.9, and 186 respectfully. Although the Wilcynski and Haynes ('Small drop in specificity, substantive gain in sensitivity' query) filter offered an NNR of 88.2, its sensitivity was much lower than that of other filters at 75.7%. Table 3 presents the results of the health-evidence.ca SR search filter along with the six other search filters tested in CINAHL. Although not performing with the best result on any single outcome, the health-evidence.ca SR search filter appeared to offer the best overall combination of sensitivity (89.9%), specificity (97.6%), precision (1.8), and NNR (57.2).
CINAHL
Sensitivity Two search strategies achieved a sensitivity of greater than 95% (Wong, Best sensitivity; Centre for Reviews and Dissemination (CRD) [38] filters), with the health-evidence.ca SR search filter achieving 89.9% sensitivity.
Specificity The Wong Best sensitivity query scored highest on specificity (99.4%), matched by the Wong Best optimization (sensitivity > specificity) query (99.4%). The Wong queries were followed closely in specificity by the McKibbon (1998) filter (98.9%) and the health-evidence.ca SR search filter (97.6%).
Precision The most precise search filter was Wong's, Best optimization query at 3.8, followed by the Best Specificity Query [33]
Discussion
The objective of health-evidence.ca is to contribute to evidence-informed decision making in public health by facilitating access to published systematic reviews evaluating the effectiveness of public health and health promotion interventions. An optimal search filter for healthevidence.ca is one that has high sensitivity, specificity, and precision and a relatively low NNR. However, any reduction in NNR was desirable. A filter such as this allows us to have confidence that all relevant articles will be identified (sensitivity), fewer non-relevant articles will be retrieved (specificity), most of the identified articles will be relevant (precision), and the NNR will be reduced. Reducing the NNR is of great importance since screening is a resource-and time-intensive process.
Although a search filter may perform exceptionally well on any single outcome, it is the balance of performance across these four domainssensitivity, specificity, precision, NNRthat distinguishes the best filter for our purposes. By replacing the PH search filter with the health-evidence.ca SR search filter, the overall number of articles retrieved from health-evidence.ca electronic searches was greatly reduced without losing relevant content. The balance struck by the SR search filter means that this filter would be useful to those wishing to retrieve systematic reviews related to health care, with wider application than that of our own database of reviews on the effectiveness of interventions. The desired benefit of filters is that they save time both in search strategy development and screening. One study demonstrated how filters reduce the number of results needed to screen [37], while another found that saving time both in search strategy development and screening of results was the most common benefit reported by librarians [38]. For our purposes, the health-evidence.ca SR search filter offered overall improvements in specificity and precision, with the associated decrease in the NNR, substantially decreasing screening time. The desired improvement in precision was feasible while only minimally impacting the sensitivity of the search strategy. The results of this study illustrate that for the most part, the health-evidence.ca SR search filter outperformed the PH search filter with respect to sensitivity, specificity, precision and NNR in all three databases. However, it was the overall balance among these variables and the fact that high precision could be combined with high sensitivity that made the health-evidence.ca SR search filter the optimal choice for identifying systematic reviews evaluating the effectiveness of interventions.
When compared to other filters in MEDLINE, EMBASE and CINAHL, overall, the health-evidence.ca SR search filter offered the right balance of sensitivity, specificity, precision, and NNR. Although other filters had higher sensitivity scores than the health-evidence.ca SR search filter in MEDLINE, these higher sensitivity scores were generally accompanied by poorer precision and NNR performance. In EMBASE, the health-evidence.ca SR and Scottish Intercollegiate Guidelines Network search filters performed the best overall and were comparable in terms of performance across all of the outcome measures. Likewise in CINAHL, though the health-evidence.ca SR search filter did not outperform other filters on any single outcome, it offered the most robust overall result of high sensitivity and specificity with a reasonably low NNR in comparison to other filters.
The health-evidence.ca SR search filter streamlines the process of locating and screening relevant reviews by allowing us to effectively search health databases with a simpler strategy that maintains a high level of both sensitivity and precision. The task of searching the health databases for every relevant systematic review evaluating effectiveness of public health interventions is a challenging one that requires balance. Because of the growth of the literature in the area of systematic reviews, highly sensitive searches often come up with result sets that are unmanageably large. However, if a search is too specific, then it has the risk of missing relevant articles. It is important to establish the right balance in the trade-off between sensitivity and specificity depending on what will best serve the purpose at hand [39,40]. Using the healthevidence.ca SR search filter has allowed us to achieve the right balance in our searches by retaining greater than 85% sensitivity across all three databases, while reducing the NNR by two thirds. We estimate that this has translated into a savings of 384 hours of staff time per quarterly update of health-evidence.ca by reducing the hours required to execute database searches, screen results, retrieve full-text versions of potentially relevant reviews, and test reviews for relevance. The reduction has meant that resources are available for the exploration and development of new protocols for searching other relevant but previously unexplored electronic databases covering areas such as environmental health, social welfare, and veterinary sciences for relevant public health content.
The health-evidence.ca SR search filter is an easy-touse tool. It can be entered into the OVID interface for searching in MEDLINE and EMBASE. Compared to other more complex filters, the health-evidence.ca SR search filter is easily entered. A survey of librarians revealed that users find search strings too long [38,40]. The SR search filter used by health-evidence.ca is a relatively short search filter, with other authors also finding that the brief search filters work well. Our results, which are similar to those of others [39,38], indicate that methodological search filters can be as or more effective than content filters for retrieving relevant systematic reviews [27][28][29][30][31][32][33][34][35]39]. Using a methodological filter allows us to circumvent the need to generate an accurate and all encompassing definition of public health that can be translated and applied across indexing systems within different databases. However, if desired, the search strategy can be combined (using Boolean logic, e.g. AND) with topic-specific search terms to reduce the number of articles retrieved, if applied for a specific topic area (e.g. influenza).
Limitations
Searching was conducted in OVID's search interface for all three databases; other search interfaces for these databases (e.g. PubMed) may handle the searches somewhat differently. As of August 30, 2008, CINAHL moved from OVID Technologies to be hosted by EBSCO, exclusively. Unfortunately, this change to EBSCO renders the CINAHL filters included in this paper, including our filters, out of date. The performance of these filters would require reevaluating them in the EBSCO platform before their application. This brings light to a key limitation of search filterscreation dates must always be considered before using a filter as changes to indexing terms and hosting platforms can impact filter function.
The sensitivity scores calculated for each search filter can be applied to broader searches for systematic reviews evaluating various interventions and are not necessarily applicable only to public health interventions. However, precision and NNR scores were calculated specifically for public health content and cannot be generalized to topic areas outside of public health. The low precision scores yielded across all search filters were expected, since precision is generally low when searching large databases [39,40]. Lastly, our group's own manual screening set was used as the gold standard. Although a consistent set of relevance criteria were applied to generate this results set, screening was shared between two authors (MD, KD), and several other members of the health-evidence.ca team. Although either MD or KD acted as second reviewer on each article, there was still potential for reviewer bias through the involvement of a small number of reviewers. Additionally, having a combination of both systematic review methodology indexing terms and public health indexing terms in our PH search filter dually limited our results sets, retrieving only content which met all requirements for both methodology and public health content. | 2017-06-21T07:17:14.579Z | 2012-04-18T00:00:00.000 | {
"year": 2012,
"sha1": "2cf762cf061be9b3dbb1c4cb5a1d2b8b11536610",
"oa_license": "CCBY",
"oa_url": "https://bmcmedresmethodol.biomedcentral.com/track/pdf/10.1186/1471-2288-12-51",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "e9b859ec75213904e54bf72c45febc88cd978738",
"s2fieldsofstudy": [
"Medicine",
"Psychology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
231756847 | pes2o/s2orc | v3-fos-license | Evaluation of the geroprotective effects of withaferin A in Drosophila melanogaster
Withanolides are a class of compounds usually found in plant extracts which are an attractive geroprotective drug design starting point. We evaluated the geroprotective properties of Withaferin A (WA) in vivo using the Drosophila model. Flies were supplemented by nutrient medium with WA (at a concentration of 1, 10, or 100 μM dissolved in ethanol) for the experiment group and 30 μM of ethanol for the control group. WA treatment at 10 and 100 μM concentrations prolong the median life span of D. melanogaster’s male by 7.7, 9.6% (respectively) and the maximum life span (the age of death 90% of individuals) by 11.1% both. Also WA treatment at 1, 10 and 100 μM improved the intestinal barrier permeability in older flies and affected an expression of genes involved in antioxidant defense (PrxV), recognition of DNA damage (Gadd45), heat shock proteins (Hsp68, Hsp83), and repair of double-strand breaks (Ku80). WA was also shown to have a multidirectional effect on the resistance of flies to the prooxidant paraquat (oxidative stress) and 33° C hyperthermia (heat shock). WA treatment increased the resistance to oxidative stress in males at 4 and 7 week old and decreased it at 6 weeks old. It increased the male’s resistance to hyperthermia at 2, 4 and 7 weeks old and decreased it at 3, 5 and 8 weeks old. WA treatment decreased the resistance to hyperthermia in females at 1, 2 and 3 weeks old and not affected on their resistance to oxidative stress.
INTRODUCTION
One of the key challenges within life sciences is the search for the substances that can increase the resistance of living systems to various stress factors and contribute to their active longevity. The most promising direction of research in this aspect is the identification of such substances among the plant metabolites. Therefore, the properties of plant extracts are currently actively studied to find the optimal approach to include them in new pharmacological preparations. Among these compounds withanolides are considered as a promising class of candidates for the design of new drugs. Indeed, withanolides display a wide range of relevant pharmacological activities, good bio accessibility and a low risk of side effects. Currently, the preparations containing withanolides from Withania somnifera are used in the Ashwagandha composition as a sedative, hypnotic and antiseptic drug [1].
Withanolides are widely studied worldwide. For instance, PubMed contains more than 300 publications with the keyword "withanolides". Withanolides attract a lot of interest for their potential use as inhibitors of apoptosis. They are also considered as therapeutic candidates for the treatment of neurodegenerative, autoimmune and inflammatory diseases. Their antitumor properties have also attracted a lot of interest for the development of novel cancer therapies. It is common knowledge that Drosophila melanogaster has notable advantages as a AGING model system for studying the effects of pharmacological interventions on aging [2]. In our study we hypothesized that the addition of a Withaferin A (WA) supplement to the diet of Drosophila melanogaster wild type Canton-S (CS) could have a beneficial effect on their health status, especially when they get older.
The first withanolide, "withaferin," was found in the leaves of the Withania somnifera (Solanaceae) in 1962 [3]. This metabolite was a new type of steroid containing alpha, beta-unsaturated lactone linked to the C-17 of the sterane skeleton [4,5]. However, this "withaferin" turned out to be 2,3-dihydro-3-methoxywithaferin A, which occurs in mixture with WA [5]. Independently, in 1965 Kupchan et al. found WA in the leaves of Acnistus arborescens (Solanaceae) [6]. Later, other representatives of this class of compounds were discovered in the plants of the Solanaceae family. Withanolides have been found in some Tacca species from the Dioscoreaceae family (taccanolides) and Ajuga sp. from the Lamiaceae family, as well as in some marine organisms [7,8].
Today, the class of withanolides contains more than 400 chemical compounds. This includes closely related congeners that are found in the plants of Solanaceae [9][10][11][12][13][14]. They consist of C-28 steroidal lactones built on a sometimes modified framework of ergostane, which can form a six-membered lactone ring formed by the oxidation of C-22 -C-26 ( Figure 1) [15].
The term "withanolide" is commonly used for 22hydroxyergostane-26-acid-22,26-olide [4]. Its structural diversity is due to the modifications of the carbocyclic backbone or side chain. The other typical substitutions and modifications of the naturally occurring metabolites are as follows [16]: oxo group at C-1; instead, less commonly a hydroxyl group; double bond C-2 -> C-3; instead, less often the hydroxyl group at C-3; deltalactone (26 -> 22O), often unsaturated (24,25); a fragment of gamma-lactone (26 -> 23O) instead of deltalactone, often also unsaturated; lactol part instead of lactone residue; high oxidation state in many positions of the entire molecule (for example, oxo groups, hydroxyl groups, epoxy substructures, hemicetals); oxidative degradation and or new cyclization of the molecule.
The stereochemistry of WA was determined in 1966 [25]. Its structure has five functional groups: an unsaturated ketone ring A, 2 hydroxyl groups, an AGING epoxide in ring B, a 6-carbon lactone ring, and an unsaturated carbonyl group (Figure 1). The double bond in ring A and the epoxy ring are responsible for the cytotoxicity of the compound. NMR spectral analysis identified C3 as a major nucleophilic target site for ethyl mercaptan, thiophenol, and ethyl L-cysteine in vitro [26]. These five functional groups allow WA to interact with multiple molecular targets leading to a wide range of biological activities.
Previous in vitro and in vivo studies showed that WA displays anti-tumor activity. It is well established that WA induces apoptosis in cancer cells via different mechanisms [27][28][29][30]. In most cancer cell lines, WA inhibits tumor cell proliferation by stopping the cell cycle during the G2/M checkpoint [31,32] and inhibits nuclear factor kappa B (NF-κB) activation by interacting with the IKKγ subunit, which prevents IκB phosphorylation [33,34]. A decrease in NF-κB activity leads to a decrease in the production of proinflammatory and stress response mediators [35]. Antitumor activity is also linked to ability of WA to promote oxidative stress. WA decreases the mitochondrial membrane potential and activates various caspases and proteases, which trigger the degradation of various substrates, such as cytoskeletal proteins and poly (ADPribose) polymerase [36,37]. Also WA regulates the activity of antioxidant enzymes (such as superoxide dismutase) [38] and mRNA expression of antioxidant genes: erythroid 2-like 2 (NFE2L2), heme oxygenase 1 (HMOX1), glutathione-disulfide reductase (GSR), and NAD(P)H quinone dehydrogenase 1 (NQO1)) in cancer cells [39]. Also tumor activity of WA involves induction of heat shock response via Akt / mTOR and MAPK signaling pathways [40].
The anti-inflammatory and anti-fibrotic effects of WA have been demonstrated in an in vivo model of bleomycin-induced scleroderma. Daily intraperitoneal injections of WA over the span of 28 days cause reduced dorsal skin thickness in this model. The study has shown that WA suppresses the pro-inflammatory phase of fibrosis regulated by the TGF-β/Smad signaling cascade, and also significantly reduces the proportion of fibroblasts that turn into myofibroblasts. The authors have associated the antifibrotic effect with the inhibition of the FoxO3a-Akt-dependent NF-κβ/IKK-mediated cascade, which is involved in the process of the fibrotic tissue transformation [41].
Due to its wide positive properties and availability, WA can be considered as a promising substance for improving health span and life span. In the present study, our hypothesis is that the addition of WA to the Drosophila's feed would have a beneficial effect on its vitality, especially with age.
Effect of WA on the life span of Drosophila melanogaster wild type Canton-S
The effect of WA at concentrations of 1, 10, 100 μM on the life span of male and female Drosophila melanogaster of the wild type Canton-S was studied. WA at 10 and 100 µM concentrations increased the median life span in male by 7.7, 9.6% (respectively) (p<0.0001) and the maximum life span (the age of death 90% of individuals) by 11.1% both ( Figure 2B). While we sighted significant shift of these group curves to the right relative to the control curve ( Figure 2A). The 1 µM of WA treatment not affected on studied life span parameters. Also WA treatment not affected on studied lifespan parameters in Drosophila's females ( Figure 2C, 2D).
Effect of WA on stress resistance of Drosophila melanogaster wild type Canton-S
The effect of WA at 1, 10, 100 μM concentrations on the resistance of Drosophila to the action of paraquat (20 mM solution in 5% sucrose, oxidative stress) and hyperthermia (33° C, heat shock) at the age of 1 to 8 weeks (from young to old individuals) demonstrated in Tables 1, 2 WA treatment had a different effect in response to studied stressors in males. It increased the resistance to paraquat only at the age of 4 and 7 week at 1 μM (by 16.7, 42.9 %% increased 25 percentiles respectively). At 6 weeks decreased 90 percentiles by 14.7% at 10 μM of WA. Also we observed shift in mortality distribution curves at 1 µМ of WA in 1 and 8 weeks relative to the control curve. WA treatment increased the resistance of male's to hyperthermia at 2 weeks after 100 µМ of WA (by 9.7% increased 50 percentiles), at 4 weeks after 1 µМ of WA (by 11.1% -90 percentiles) and at 7 weeks after 100 μM of WA (by 26.9% -90 percentiles). While it decreased the resistance to hyperthermia at 3 weeks after 1 and 10 μM (by 44.4, 52.0, 46.2%%; 66.7, 72.0, 61.5 %% decreased 25, 50, 75 percentiles respectively). At the age 5 weeks we found decreased by 42.9% in 25 percentiles after 1, 100 μM and by 6.9% in 90 percentiles after 10 μM WA treatments. And at the age 8 weeks decreased by 13.3, 15.0, 21.7 %% in 50, 75, 90 percentiles after 1 μM of WA. These data confirm by mortality distribution. We observed shift in mortality curves at 1 µМ of WA in 3, 4 and 8 weeks, at 10 µМ in 3 and 5 weeks, and at 100 µМ at 7 and 8 weeks relative to the control curve. In other variants of the experiment WA not affected on studied survival parameters (Table 2 and Supplementary Figures 1, 2). WA treatment not affected on female's resistance to paraquat and reduced resistance to hyperthermia. At the age of 1 week at 10 μM (by 18.8, 16.4 %% decreased 25 and 50 percentiles respectively), of 2 weeks at 100 μM (by 26.3 %-50 percentiles) and of 3 weeks for 1 and 10 μM (40.0, 43.3,42.9; 56.0, 56.7, 54.3%% -25, 50 and 75 percentiles respectively). Also these data confirm mortality distribution. We observed shift 100 µМ at 2 weeks and 1 μM and 10 μM at 3 weeks curves to the right relative to the control curve. In other variants of the experiment WA not affected on studied survival parameters ( Table 2 and Supplementary Figures 3, 4).
Effect of WA on the intestinal barrier permeability of Drosophila melanogaster wild type Canton-S
Changes in the permeability of the intestinal barrier in Drosophila melanogaster wild type Canton-S at the age of 4, 6, 8 weeks were studied against the background of treatment of WA at concentrations of 1, 10, 100 µM.
In males at the age of 8 weeks, a decrease in the proportion of flies with the «smurfs» phenotype by 67, 61 and 89% was observed relative to the control when taking WA at concentrations of 1, 10, 100 µM, respectively ( Figure 3C). Females also showed a decrease in the «smurfs» rate only at the age of 8 weeks.
The proportion of flies with the «smurfs» phenotype was lower by 73, 42 and 61 %% relative to the control after WA at 1, 10, 100 µM concentrations, respectively ( Figure 3F). Other group of the experiment showed no significant differences ( Figure 3A, 3B, 3D, 3E).
of double-strand breaks (Ku80) studied. The combined results are presented in Figure 4 and considered. In males 1 μM WA effected on expression of Gadd45
DISCUSSION
Currently, the identification of geroprotectors which could be used as a cure against aging constitutes an important area of research. Although there are around 400 compounds known to extend the life span of model organisms, only few of them meet the criteria to be used as potential geroprotectors [42]. Moreover, there is a lack of clinical studies which have been conducted to analyze the effects of these potential geroprotectors on humans. It should be emphasized that an ideal geroprotector should not only increase the average but also the maximum life span. Furthermore geroprotectors should contribute to shifting the entire mortality curve to the right so that extended life span would be associated with the extension of the active period of life.
Balanced nutrition is one of the most important factors promoting increased life span. Currently, rapamycin, metformin, Skulachev ions (SkQ), and some other compounds are known to be promising geroprotective substances [43][44][45][46]. The data available indicate that these geroprotective substances prolong the life in model organisms (Caenorhabditis elegans, D. melanogaster, Mus musculus, Rattus norvegicus, etc.). In many cases they also reduce the likelihood of agingassociated diseases. Metformin and rapamycin are two FDA-approved mTOR inhibitors. However, the use of metformin and rapamycin has various side effects. A. Aliper et al. [47] applied several bioinformatic approaches and deep learning techniques to a dataset from the Library of Integrated Networkbased Cellular Signatures (LINCS) to find the substances that could emulate the genetic response to metformin and rapamycin. Using this approach, the authors predicted the safety of each selected compound. As a result of the analysis, many new candidate mimetics of metformin and rapamycin were identified, WA being one of them.
Effect of WA on the life span of Drosophila melanogaster wild type Canton-S
Life span is regulated by multiple interrelated phenotypic and genotypic factors and is a temporal characteristic of the damage-restoration process in the body, leading to old age and death [48][49][50]. In gerontology, it is important to make a distinction between chronological and biological age. Chronological age is measured as the period passed since the time of birth. In humans, chronological age is not a sufficient metric to evaluate the health and performance of an aging person. A more appropriate measure to this end is the biological age which aims to estimate how the progressive degradation occurring within the aging organism affect a combination of metabolic, structural, functional, regulatory features, and adaptive capabilities. These alterations affect vital functions of the organism, leading to the onset of age related diseases, an increase in the probability of death, or a decrease in life span [51][52][53]. It is well known that it is possible to delay aging and the onset of agerelated diseases to prolong the period of active life. In our study, we have shown that WA at concentrations of 10, 100 µM increases the median and maximum life span of male CSs (Figure 2A, 2B). While WA treatment not affected on life span parameters in Drosophila's females ( Figure 2C, 2D).
Effect of WA on stress resistance of Drosophila melanogaster wild type Canton-S
There are a lot of articles in which plant materials have a positive influence on stress resistance of model organisms. For example it was shown in Drosophila melanogaster that the apple phlorizin [54], cloudberry extract [55], oil from Caryocar coriaceum (Pequi) [56] increase resistance to oxidative stress. Blueberry extract [57], Lonicera japonica extract [58] promote stress tolerance in Caenorhabditis elegans. Styphnolobium japonicum fruits [59], Rhodiola rosea root extract [60], Cotinus coggygria extract [61] increase stress resistance and exert antioxidant properties in mouse models. Therefore, we decided to check the effect of WA on the resistance of Drosophila's flies to oxidative stress and hyperthermia.
Oxidative stress is caused by elevated intracellular levels of reactive oxygen species (ROS), which damage lipids in cell membranes, oxidize proteins, and damage DNA [62]. In our experiment we use paraquat as inductor of ROS [63]. It is known that WA can suppress oxidative stress [64][65][66][67]. In most variants of our experiment, WA not led to significant changes on studied survival parameters. WA treatment at 1 μM increased the resistance of males only at the age of 4 and 7 week. Negative effect was found after 10 μM treatment at the age 6 weeks. Also we observed shift in male's survival curves at 1 µМ of WA in 1 and 8 weeks relative to the control curve (Table 1 and Supplementary Figure 1). While WA treatment not affected on female's resistance to paraquat (Table 2 and Supplementary Figure 3).
Prolonged or intense heat shock causes numerous changes in cell metabolism and disrupts the state of its structural units [68,69]. Protein damage is the main type of damage during heat shock. Its downstream effects, higher metabolic rate and free radical production, lead to consequent DNA damage [70][71][72]. In our experiment, WA treatment had different effect in response to hyperthermia. Thus, it increased the male's resistance to hyperthermia at 2 and 7 weeks after 100 µМ of WA and at 4 weeks after 1 µМ of WA. It decreased its resistance to hyperthermia at 3 weeks after 1 and 10 μM, at AGING 5 weeks after 1, 10 μM, 100 μM and at 8 weeks after 1 μM of WA. Also we observed shift in male's mortality curves at 1 µМ of WA in 3, 4 and 8 weeks, at 10 µМ in 3 and 5 weeks, and at 100 µМ at 7 and 8 weeks relative to the control (Table 1 and Supplementary Figure 2). WA treatment reduced female's resistance to hyperthermia at the age of 1 week after 10 μM, of 2 weeks after 100 μM and of 3 weeks after 1 and 10 μM. Also we observed shift 100 µМ at 2 weeks and 1 μM and 10 μM at 3 weeks curves to the left relative to the control curve (Table 2 and Supplementary Figure 4). Thus, it has been shown that WA has a multidirectional effect on the resistance of CSs to the stress factors under study.
Effect of WA on the intestinal barrier permeability of Drosophila melanogaster wild type Canton-S
The gastrointestinal tract has a barrier function that prevents the penetration of food antigens, bacterial toxins, viruses and microorganisms into circulation [73,74]. There are a lot of articles that have been written about positive role of plant extracts on intestinal microflora and intestinal epithelial barrier [75][76][77][78][79]. The disadvantage of these studies is that they were performed in vitro and do not take into account the effect of aging. The deregulation of the barrier function, which typically occurs in the elderly, can cause the development of pathological conditions [65,[80][81][82]. In order to prevent the development of such conditions, methods for diagnosing the violations of the permeability of the intestinal barrier are being intensively developed. Therefore, an analysis of the permeability of the intestinal barrier was performed. We have shown that WA at all studied concentration increases the strength of the intestinal barrier in old CSs ( Figure 3C, 3F). The rate of flies with the «smurfs» phenotype was lower by 67, 61 and 89% relative to the control after WA at 1, 10, 100 µM concentrations, respectively in males, and by 73, 42 and 61 %% respectively in females. WA's effects on the strength of the intestinal barrier have not been found in literature. But there is study in which authors did not succeed to increase the strength of intestinal barter using plant materials in aging aspect: ursolic acid (triterpenoid) does not affect gut integrity in male D. melanogaster at the age 30 days [83]. And pectin supplementation was not affected by the intestinal barrier function in healthy young adults and in healthy elderly [83].
Effect of WA on the expression of stress response genes in Drosophila melanogaster wild type Canton-S
Genetic and epigenetic mechanisms and genes that are involved in the regulation of life span are highly interconnected and related to stress response [50]. Moreover, the overexpression of longevity genes listed in [84] as stress response genes almost exclusively resulted in life span extension. A wide-scale comparative analysis of the 1805 known longevity-associated genes across 205 species disclosed that these genes are consistently overrepresented across diverse taxa, compared with the orthologs of other genes, and this conservation is highly. Also in that study it was shown that longevity-associated genes were enriched in translational processes, energy metabolism and DNA repair genes [84]. The genes analyzed in our study play important roles in the following molecular and biological processes: antioxidant defense (Cat, PrxV), metal detoxification (fh), heat shock response (Hsp68, Hsp83), DNA damage recognition (Gadd45) and double-strand break repair (Ku80). More detailed information of these genes can be found in Supplementary Table 1.
Here in Drosophila's male WA decreased heat shock proteins (Hsp68 or Hsp83) genes expression at all concentration and increased Gadd45 and PrxV genes expression at 1 μM of treatment. In Drosophila's female found decreased expression of Gadd45, Hsp68 and Hsp83 genes after 1 μМ of WA. The 100 μМ of WA treatment decreased expression of only Gadd45 gene.
The effects of plant materials on gene regulation have been shown in numerous experiments on model organisms and cancer cell lines. It was shown that licorice and orange extract provoke enhancement of catalase activity and also extend Caenorhabditis elegans life span [85,86]. Citrus and apple pectin's have induced the expression of genes involved in DNA repair (D-Gadd, mei-9, spn-B), apoptosis (wrinkled/hid) and heat shock response (hsp70Aa) in Drosophila [87].
Overexpression of PrxV gene can abrogate shikonininduced cell apoptosis in HT29 colon cancer cells [88]. Modulation of HSP 90 and HSP 70 genes expressions is a possible mechanism by which the Flueggea leucopyrus (Willd) decoction mediates cytotoxic effects in breast cell lines [89]. Anticancer property has also been studied in Glycyrrhiza glabra which inhibited proliferation in HT-29 cell line due to down-regulation of HSP90 gene expression which implied an ability to induce apoptosis [90]. Crude phenolic extracts from extra virgin olive oil directly up-regulated the expression of the Gadd45 gene family in JIMT-1 human breast cancer cell line that circumvent breast cancer resistance to HER1/HER2targeting drug [91]. Protective role of Podophyllum hexandrum rhizomes and Myrtus communis leaves against DNA damage proved. Shown significant upregulation of DNA-PKcs and Ku80 and downregulation of ATM and 53BP1 gene expressions in cell lines which were pre-treated with mixture of three active derivatives isolated from the rhizomes of Podophyllum hexandrum, and then irradiated [92]. Myricetin-3-o-galactoside and myricetin-3-o-rhamnoside, isolated from the leaves of AGING Myrtus communis, modulated the expression patterns of cellular genes involved in DNA damaging repair (XPC, LIG4, RPA3, PCNA, DDIT3, POLD1, XRCC5, MPG) [93].
We have repeatedly observed gender specific reactions to WA treatment. Individuals of different genders can response differently to dietary restriction and distorted activity of nutrient-sensing pathways [94]. The main pathways and interventions that lead to sex-specific life span responses, include the growth-hormone/insulinlike growth factor 1 (GH-IGF1) axis, mechanistic target of rapamycin (mTOR) signaling, and nutritional and pharmacological interventions [95].
Thus, WA at concentrations of 10, 100 μM increases the median and maximum life span and shifts the curve to the right side in Drosophila's male. Together with WA at all concentration decreased expression of genes involved in heat shock response (Hsp68 or Hsp83). The 1 μM of WA increased expression of DNA damage recognition (Gadd45) and antioxidant (PrxV) genes. WA treatment had no effect on life span parameters in Drosophila's females. While 1 μM and 100 μM of WA decreased the expression of Gadd45 gene. And 1 μM of WA also decreased the expression of Hsp68 and Ku80 (double-strand breaks repair) genes. WA has also a multidirectional effect on the stress resistance of flies. The 1 μM of WA treatment increased the male's resistance to oxidative stress only at 4 and 7 week old. Negative effects were found after 10 μM treatment in males at the age 6 weeks, while WA treatment did not affect the female's resistance to oxidative stress. WA increased the male's resistance to hyperthermia at 2 and 7 weeks after 100 µМ of WA and at 4 weeks after 1 µМ of WA. The 1 μM of WA decreased male's resistance at 3, 5 and 8 weeks old. The 10 μM of WA decreased it resistance at 3 and 5 weeks old. Also 100 μM of WA decreased male's resistance at the age 5 weeks. WA treatment reduced female's resistance to hyperthermia at the age of 1 week after 10 μM, of 2 weeks after 100 μM and of 3 weeks after 1 and 10 μM. In contrast to this WA increases the permeability of the intestinal barrier of old flies both sexes.
MATERIALS AND METHODS
In our study, we used the wild type strain Canton-S (Bloomington, USA) Drosophila melanogaster (CS).
To obtain the experimental CS flies, their parents were pre-planted in jars of nutrient medium in the amount of 10 pairs per tube and left for 24 hours to lay eggs. After the appearance of imago, flies were anesthetized using CO2 anesthesia (Genese Scientific, USA), were separated by sex and were placed in test tubes with nutrient with WA and without for further experiments. Non-virgin females were used. Males and females lived separately with 30 animals per tube. From day 1 of life, 30 μl of of 1, 10 or 100 μM WA ethanol solution on top of the flies' nutrient medium instilled. As a control, we used flies fed with a medium supplemented with 30 μl of ethanol. The final concentration of the drug in the media was 1, 10 or 100 μM. This concentration has shown its ability to increase life span in human fibroblast (internal preliminary tests) and represents a suitable concentration range for pro-longevity effects in invertebrates. To maintain these concentrations flies were transferred to a fresh nutrient medium twice a week [98][99][100][101][102].
Life span assay
To assess life span, 150 flies were selected for each experiment in a single repetition. Two biological repeats were made. Combined data are presented. Flies were placed in test tubes with nutrient medium (30 animals per tube). The counting of the number of dead flies was performed daily. The data were used to compute survival curves and the median, maximum life span, 90% death time were calculated. The Kolmogorov-Smirnov test was used to compare the distribution of mortality in survival curves and the Gehan-Breslow-Wilcoxon test was used to compare the differences in median life span. The significance of the differences in maximum life span was evaluated using the Wang-Allison test. In order to apply this method, animals in each variant of the experiment were divided into two groups: with a life span above the age of 90% mortality, or below the age of 90% mortality. Data were recorded in a 2x2 contingency table and compared using the chi-squared test. According to Bonferroni correction were considered significant differences at p less than 0.017. Analyses were performed using Statistica 6.1 (Stat Soft), and online application for survival analysis «Oasis2» (Structural Bioinformatics Lab).
Stress resistance analysis
The stress resistance of the flies was assessed every week up to 8 weeks of age. The DAM (Drosophila Activity Monitor) system (TriKinetics, USA) was used to look into stress resistance. For analyzing the resistance to oxidative stress, the flies were placed in glass tubes 5 mm in diameter with 20 mM paraquat (Methyl Viologen, Sigma) in 5% sucrose and kept at 25° C until the complete death of flies' cohort. To assess the resistance AGING to hyperthermia, the flies were seated in glass tubes 5 mm in diameter with a standard medium and kept at 33° C until the complete death of flies' cohort. The data were used to compute differences in survival distribution with age and in percentiles (25,50,75,90) of death.
Fisher's exact test was used to calculate the statistical differences in percentiles of death. Log-rank criteria were used to assess the statistical significance in survival function distribution. Data were computed using «Oasis2» (Structural Bioinformatics Lab). According to Bonferroni correction were considered significant differences at p less than 0.017. The experiment was performed in three biological repetitions (32 flies in each).
Smurf test
We used 100 flies per variant of experiment. The smurf test was performed at 4, 6 and 8 weeks of age. For this, the test tubes were prepared with a nutrient medium stained with 2.5% (w / v) blue dye No. 1 (Sigma Aldrich, USA). The flies were moved to this medium for 9 hours. Then the number of «smurfs» and «nonsmurfs» was counted. Flies were considered «smurfs» if they were blue outside the digestive system [103]. The data obtained were used to construct histograms of the distribution of «smurf» proportion in samples. Fisher's exact test was used to assess the statistical significance of differences at p less than 0.017 using «Oasis2» (Structural Bioinformatics Lab).
Analysis of stress response gene expression
For each variant of the experiment, 60 flies were selected, separated into two groups of 30 and kept under standard conditions, (Genesee Scientific, USA). For each point of the experiment, 10 females and 20 males were used. Expression analysis was carried out 10 days after the emergence of adults. The experiment was carried out in two biological and three analytical replicates.
Gene expression was measured by "real-time" quantitative PCR with a reverse transcription step (RT-qPCR). RNA was isolated using an Aurum Total RNA mini kit (Bio-Rad, USA) according to the manufacturer's instructions. The concentration of the resulting RNA was measured using a Quant-iT RNA Assay Kit (Invitrogen, USA). Next, cDNA was synthesized according to the iScript cDNA Synthesis Kit (Bio-Rad, USA). The reaction mixture for carrying out the PCR reaction was prepared according to the manufacturer's instructions iTaq Universal SYBR Green Supermix (Bio-Rad, USA) and primers (Lumiprobe, USA) ( Table 3). The polymerase chain reaction was carried out in a CFX96 amplifier (Bio-Rad), with a DNA melting step using the following program: 1) 95° C for 30 s, 2) 95° C for 10 s, 3) 60° C for 30 s, 4) steps 2-3 were repeated 40 times, 5) DNA melting cycles.
The expression of studied genes calculated relatively to the expression of the housekeeping genes Tubulin, AGING eEF1α2, RpL32 using the CFX Manager 3.1 software (Bio-Rad, USA).
ACKNOWLEDGMENTS
We are grateful to the Bloomington stock center (Indiana University, USA) for providing the Drosophila melanogaster strains. We thank the Institute of Biology of the Komi Science Center for assistance in the experiments with Drosophila melanogaster and Insilico Medicine for the help with data analysis. | 2021-02-02T03:14:58.713Z | 2021-01-26T00:00:00.000 | {
"year": 2021,
"sha1": "95f407be07cef438a66310910f86798cae671dea",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.18632/aging.202572",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "95f407be07cef438a66310910f86798cae671dea",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
103219480 | pes2o/s2orc | v3-fos-license | Removal of zinc (II) ion from aqueous solution by adsorption onto activated palm midrib bio-sorbent
In this paper, palm midrib that was activated with mixed citric acid and tartaric acid as biosorbent was used to remove Zn (II) ion from aqueous solution. The aim of this research is to activate palm midrib by using a mixed citric acid and tartaric acid and to determine adsorption capacity of activated palm midrib biosorbent on Zn (II) ion uptake from aqueous solution. The effect of several parameters such as contact time, initial Zn (II) ion concentration and activator concentration on the degree of Zn (II) ion removal was examined. Atomic Absorption Spectroscopy method was performed to determine adsorbed amount of Zn (II) ion into activated biosorbent. The result showed that the adsorption process was relatively not so fast and equilibrium was reached after contact time of 120 min. The adsorption capacity of biosorbent reached a maximum when the concentration of mixed citric acid and tartaric acid was 1.6 M. The optimum adsorption capacity was 5.72 mg/g. The result was obtained on initial Zn (II) ion concentration of 80 ppm for 120-min contact time. Langmuir isotherm was found as the best fit for the equilibrium data indicating homogeneous adsorption of metal ions onto the biosorbent surface.
Introduction
Recently, the growth of chemical industry moves rapidly so that the disposal of heavy metals into the environment increases accordingly. The heavy metals aqueous wastes have induced many problems for human being and its environment. Electroplating and metal surface treatment processes produce large quantity of wastewater containing heavy metals, which include zinc (II) ion [1].
The presence of zinc (II) ions in aquatic environment, even at low concentrations, is potentially harmful to human health. Zinc may cause some diseases including high blood pressure, infantilism, destruction of red blood cells, impaired wound healing, and kidney damage when it accumulates in the human body [2]. Considering the toxicity and harmful effects of these metals, therefore, before discharging to environment heavy metal has to be treated. There are a lot of treatment processes for heavy metal removal from wastewater such as reverse osmosis, electrocoagulation, ultrafiltration, dialysis/electrodialysis, solvent extraction, precipitation and ion exchange [3,4,5,6].
These separation processes are not entirely economically feasible due to their relatively high costs. Therefore, it is important to study new processes, which are more efficient and economical. Adsorption is one of the most widely used methods for the removal of metal ions from contaminated water [7].
Many researchers utilized activated carbons as adsorbent for wastewater treatment industries. Because activated carbon is an expensive adsorbent. Therefore, the searching of cheap adsorbents from agricultural solid waste could be an alternative method [8]. Agriculture, forestry and fisheries have been generating large quantities of various biomass wastes, and some of them contain various natural materials with interesting functional groups, such as carboxyl, hydroxy and so on which are as the active site on the adsorbent.
Palm midrib is one of the valuable biomass wastes but it has been partly utilized as cattle feed while the remaining is mostly wasted into the nature. For application of adsorption process, palm midrib that have variety of pore size inside its particles with a proper activation process, those pores size can be enlarged and adapted to the size of molecules of proper liquid waste of heavy metal so that adsorbate can be adsorbed easily into pore of adsorbent.
So far, some researchers have used other biomaterials to sorb heavy metals including coffee waste, orange peel, banana peel, mangosteen peel, hardwood, corn straw and other adsorbents [9,10,11,12,13,14]. The adsorption capacity of biomass-based biosorbent on heavy metals can be enhanced by modifying the physicochemical properties of its surface [15,16,17]. The modification has been carried out using various reagents such as some acids such as citric acid, tartaric acid and phosphoric acid can increase the adsorption capacity of heavy metal ions [15,16,17,18,19,20,21]. With this activation, some acid groups can occur and form ester groups [17].
As far as we know, there are no previous studies that deal with the chemical modification of palm midrib by using a mixed citric acid and tartaric acid. Thus, the preparation of biosorbent from this agricultural by-product was here proposed. The purpose of this study is to prepare activated palm midrib biosorbent by using mixed citric acid and tartaric acid as activator, and to investigate the adsorption of zinc (II) ions from aqueous solution toward adsorption capacity. The effect of mixed activator concentration at similar ratio (citric acid and tartaric acid) was studied at neutral pH.
Materials and Equipment
Palm midribs, a waste product of palm tree, were used for the removal of Zn (II) ions from aqueous solution. These palm midribs were collected from area of Aceh Besar District, milled, powdered and sieved to get uniform particle size of 40-60 mesh. Some chemicals that used in this research like sodium hydroxide, citric acid 99.9 %, tartaric acid 97.5% and a solution of Zn(NO3)2 were purchased from chemical supplier with brand of Merck and distilled water was collected from our laboratory. Some equipment that used in this research such as hot plate, glasswares, desiccator, digital scale, oven dryer and sieve shaker. The analysis apparatus was Atomic Absorption Spectrometry (AAS) type AA 7000 (Shimadzu) and SEM TM-3000 (Hitachi).
Research Procedure
Raw material preparation of palm midrib was performed as following: at the beginning, about 10 mm x 10 mm size of raw palm midrib was washed several times with clean water in order to remove impurities on raw materials and after that was dried under sunlight for 1-2 days. Further washing was performed by soaking of dried palm midrib in hot water for 1 hr, then it was dried again under sunlight for a day and finally was dried in oven dryer at 110 °C for 1 hr. A dried palm midrib was then milled and sieved on the particle size of 40-60 mesh. Lastly, powdered palm midrib was dried again at 110 °C for 4 hr and then was stored in a desiccator for further application.
For preliminary activation process, 10 gr of dried palm midrib was mixed with solution of 0.5 M KOH. The mixing was performed at speed of 80 rpm for 30 min at room temperature. After completion of mixing and filtering, palm midrib biosorbent was neutralized several times using a distilled water. Finally a neutral of palm midrib biosorbent was dried at 50 °C for 2 hr. The purpose of base solution addition is to create some new active sites on the surface of palm midrib so that later would increase adsorption capacity [22]. About 2 gr of netralized palm midrib was mixed with 200 ml of citric acid and tartaric acid (on the same ratio of 0.4 M) at speed of 80 rpm for 2 hr, then filtered and activated palm midrib biosorbent was then washed with distilled water to remove the activator excess until reached a pH of ± 7, followed by re-drying at 50 °C for 24 hr [16]. The similar procedure was repeated for citric acid/tartaric acid concentrations of 0.8; 1.2, and 1.6 M. For adsorption process, 1 gr of activated palm midrib biosorbent was mixed with 100 ml of 20 ppm of Zn(NO3)2 at speed of 80 rpm for different time (0 to 160 min) at room temperature. After the adsorption process was completed, the filtered solution was analyzed by using AAS. The above treatment was repeated for initial concentrations of Zn (II) ion of 40, 60, and 80 ppm (15,44,54 and 73 ppm of AAS actual results).
Effect of Contact Time Toward Adsorption Capacity
Based on the literature study it is known that the longer of contact time between biosorbent and adsorbate then the concentration of metal ion that adsorbed will increase until the equilibrium occurs where the largest adsorption capacity occurs at the beginning of the adsorption process. So in order to know the effect of the contacting time on the adsorption capacity of Zn (II) ion by the activated biosorbent, the preliminary experiment has been carried out at different contact time until 160 min. The effect of contact time on the adsorption capacity of Zn (II) ion can be seen in Figure 1. Based on Figure 1 we can see that at the beginning of the adsorption process, the adsorption rate increases rapidly until 30 min and later the increasing of adsorption tends to stabilize from the 120 min until reaching equilibrium at the 160 min. The condition of this research was performed by using biosorbent that was activated by citric acid and tartaric acid concentration of 0.4 M while for adsorbate concentration was at initial concentration of 20 ppm Zn (II) ion. The initial rapid uptake of metal ions may be ascribed to the presence of large number of vacant sites available for metal ions, and afterwards the remaining free metal ions are difficult to be occupied, because of repulsive forces between the free and adsorbed metal ions. The adsorption capacity reached 1.38; 1.48; 1.48, 1.50; 1.49; 1.50 mg/g at different contact time of 0, 30, 60, 90, 120, 140 and 160 min, respectively. Equilibrium time occured after contact time of 120 min, where the adsorption capacity of Zn (II) ion reached 1.5 mg/g on the initial adsorbate concentration of 20 ppm. With increasing of contact time, the capacity of biosorbent to bind to surface of biosorbent will be higher that is enabling more and more bonds to be formed between surface of biosorbent particles and heavy metal ions in solution of adsorbate.
Citric acid and tartaric acid that used as activator during activation process plays an important role on modification of biosorbent surface. The presence of citric acid and tartaric acid during the activation process resulted in degradation of materials on the surface of biosorbents that will form the more pore on particles that later can be also seen on SEM graph in Figure 4. By increasing the concentration of citric acid and tartaric acid, it might be proportional to the increase the pore distribution and surface area of the biosorbent causing a number of binding metal ions of zinc increases and as well as adsorption ability also increases during adsorption process.
Other researcher [23] found that the adsorbed amount of Zn (II) ion on the first 40 min of adsorption process increased significantly and after contact time of 60 min it still remains increasing but reached equilibrium after contact time of 120 min as our finding. The quite fast adsorption of metal ions in the early minutes of adsorption process may be due to at the beginning of the adsorption process many sites and pore of biosorbent is still empty so that the tendency of the metal ion solution to be adsorbed into the biosorbent is much higher with increasing contact time and slow the rate of adsorption began to decline until they reach equilibrium time. Similar to biosorbent that activated by activator of citric acid and tartaric acid in this study, [24] suggested that some natural biosorbent contains cellulose and lignin which is a material with huge potential to be used as biosorbent.
Effect of Initial Zn (II) Ion Concentration on Adsorption Capacity
To study the effect of initial concentration of Zn (II) ion on the adsorption capacity, the research was done by varying the Zn (II) concentration of 20, 40, 60 and 80 ppm and the adsorption process was done at contact time of 120 min. The effect of the initial concentration of adsorbate on adsorption capacity is shown in Figure 2. Figure 2 showed that the adsorption capacity increases with increase of initial adsorbate concentration. The highest adsorption capacity was 5.72 mg/g that obtained after adsorption process was performed at initial adsorbate concentration of 80 ppm. On the other hand, the adsorption capacity on others initial adsorbate concentration of 20, 40, 60 and 80 ppm using biosorbent that activated by mixed citric acid and tartaric acid of 1.6 M were 1.56,; 3.99; 4.69; 5.72 mg/g, respectively. The utilization of mixed citric acid and tartaric acid as chemical activator to activate the palm midrib biosorbent has improved the adsorption capacity of zinc (II) ion. As comparison, from Figure 2 it can be seen that the adsorption capacity of zinc (II) ion on activated palm midrid by using mixed citric acid and tartaric acid was higher than that of by using only citric acid. Based on the previous study [25,26,27], the presence of hydroxyl groups in the biosorbent caused the metal ion interact easily with the hydroxyl group (O-H) from both acids and improving adsorption capacity. The adsorption capacity on the biosorbent increases linearly with increasing adsorbate concentration. This is because when the initial concentration of the solution was increased, the mass transfer driving force became larger, resulting greater adsorption capacity. It can be clarified also that the more concentration, the more adsorbate molecules and adsorbents interact in the adsorption process. This causes adsorption tends to increase its capacity.
Effect of Activator Concentration on Adsorption Capacity
The addition of different concentration citric acid and tartaric acid during activation process would affect the adsorption capacity of the adsorbate solution on palm midrib biosorbent. This effect was illustrated in Figure 3. The increase of citric and tartaric acid concentration that used to activate biosorbent affected adsorption capacity on each samples with different uptake capacity of Zn (II) ion. The highest adsorption capacity of 5.72 mg/g was obtained when we used biosorbent that activated by using citric acid and tartaric acid concentration of 1.6 M at initial concentration of 80 ppm Zn (II) ion. While the lowest adsorption capacity was obtained at citric acid and tartaric acid concentration of 0.4 M with the uptake capacity of 1.52 mg/g. The higher concentration of the activator would be the higher the adsorption capacity value of the Zn (II) ion. The increase of citric acid and tartaric acid concentration increases proportionally pore numbers and surface area of the biosorbent, causing the number of Zn ion-binding sites to increase and their adsorption capacity increased as well. Therefore as observed in SEM images analysis, the presence of citric acid and tartaric acid during activation results in the cleaning of impurities materials on the surface of biosorbent to form more pores and active sites.
Morphological Analysis on Biosorbent Surfaces
To know the morphology on the surface of the biosorbent, SEM analysis of biosorbent has been done and the results can be seen in Figure 4 below. Figure 4 (a) showed clearly that the pores and cavities present in the unactivated biosorbent are still covered by many impurities and the pores and cavities are not clearly visible on the surface of the biosorbent. However, after activation by citric acid and tartaric acid as seen in Figure 4 (b), the surface of the biosorbent has been cleaned and broaden and we can see clearly there are many pores and cavities appeared that are some parts look quite regular.
Adsorpstion Isotherm
Determination of Langmuir adsorption isotherms was obtained by creating a relationship curve between the equilibrium Zn (II) ion concentration in aqueous solution (Ce) with Ce/qe (qe, represents the equilibrium adsorption capacity). In Figure 5 it can be seen that the correlation value of R 2 obtained is 0.998. The determination of Freundlich adsorption isotherms was obtained by creating a relationship curve between the log Ce and the log qe. Freundlich adsorption isotherm graphs can be seen in Figure 6 where the correlation value obtained is 0.9934. The determination of Langmuir or Freundlich adsorption isotherms is known from the value of R 2 obtained. The adsorption isotherms of zinc (II) ions using palm midrib biosorbent followed the equation having R 2 value close to 1. Based on Figure 5 and Figure 6 it can be seen that Langmuir's isotherm equation has R 2 value close to 1 that is 0.9989, so it can be concluded that adsorption of zinc (II) ions on activated palm midrib follows Langmuir's isotherm equation. The adsorption isotherms show that the initial Zn (II) ion concentration increase is followed by the increase in the amount of adsorbed so as to achieve a balanced state.
The Langmuir equation is used to obtain the maximum adsorption capacity value (qm) and the Langmuir adsorption constant (kL) where by obtaining this value, we can know the amount of adsorbate adsorbed on 1 gr of biosorbent. For this biosorbent the qm value obtained is 4.23 mg/g and the value of kL is 0.62.
Conclusions
The adsorption capacity of Zn (II) ion on activated palm midrib reached the equilibrium time at 120 min. The highest adsorption capacity of 5.72 mg/g by using palm midrib that was activated using mixed citric and tartaric acid concentration of 1.6 M was obtained at adsorption condition of initial concentration of 80 ppm zinc (II) ion.
Citric acid and tartaric acid that used as activator during activation process plays an important role on modification of biosorbent surface. The presence of citric acid and tartaric during the activation process resulted in degradation of materials on the surface of materias. The adsorption capacity of biosorbent was influenced significantly by addition of citric and tartaric acid as mixed activator during activation process. The cleaness and opening of pore and cavities would also increase the adsorption capacity significantly. Adsorption of zinc (II) ion on activated palm midrib follows Langmuir's isotherm equation. | 2019-04-09T13:09:07.362Z | 2018-04-06T00:00:00.000 | {
"year": 2018,
"sha1": "a89f4b2fb5fa3b4b7764982901ecd630d5558ee7",
"oa_license": null,
"oa_url": "https://doi.org/10.1088/1757-899x/334/1/012027",
"oa_status": "GOLD",
"pdf_src": "IOP",
"pdf_hash": "8a29b158dd859202a6fbcf9320c05fe7addfa8bc",
"s2fieldsofstudy": [
"Engineering"
],
"extfieldsofstudy": [
"Chemistry"
]
} |
4795884 | pes2o/s2orc | v3-fos-license | Reconfigurable soliton networks optically-induced by arrays of nondiffracting Bessel beams
We address the propagation of solitons in reconfigurable two-dimensional networks induced optically by arrays of nondiffracting Bessel beams in Kerr-type nonlinear media. We show that broad soliton beams can move across networks having different topologies almost without radiation losses, opening new prospects for all-optical soliton manipulation. We also discuss various switching scenarios for solitons launched into multi-core directional couplers optically-induced by suitable arrays of Bessel beams.
Introduction
Since their early theoretical prediction [1] and recent experimental observation [2,3], discrete optical solitons in waveguide arrays have attracted large attention because of their potential for all-optical switching and power-and angle-controlled steering (for a recent review see Ref. [4]). In particular, the possibility of construction of twodimensional soliton networks of nonlinear waveguide arrays was established [5,6]. In landmark recent experiments [7][8][9][10][11][12] it was shown that periodic nonlinear lattices with flexibly controlled refractive index modulation depth and period can be induced alloptically, in particular, in photorefractive media. Such lattices constituted by continuous nonlinear media with an optically imprinted modulation of refractive index offer a number of new opportunities for the all-optical manipulation of light as well [13][14][15][16], since they can operate in both weak-and strong-coupling regimes, depending on the depth of refractive index modulation.
The technique of optical lattice induction opens a wealth of opportunities for creation of waveguiding configurations with various nondiffracting light beams. An important example is set by Bessel beams, which under ideal conditions do not diffract upon propagation. Single-mode Bessel beams allow creation of opticallyinduced lattices with radial symmetry, where solitons can be set into controllable rotary motion [17,18]. Combination of several incoherent Bessel beams can be used to build couplers and switching junctions that can trap and steer solitons, while junction properties can be tuned by intensity, intersection angles, and width of the central cores of Bessel beams. Notice that complex configurations of Bessel beams can be created, e.g., by computer-generated holograms [19][20][21].
In this paper we address different types of two-dimensional networks created with arrays of mutually incoherent parallel Bessel beams in Kerr-type nonlinear media. Each of the Bessel beams forming the network induces a well-pronounced guiding channel that overlaps with its neighbors through slowly decaying tails. We show that broad solitons launched across the network can jump between neighboring channels almost without radiation losses and thus can follow network bends, phenomena that open a wealth of opportunities for managing the soliton propagation trajectories. The networks suggested here are advantageous in comparison with their technologically prefabricated counterparts, because of the tunability afforded by the optical induction.
Model
We consider the propagation of light along the ξ axis of a focusing Kerr-type nonlinear media with an optically-induced modulation of the refractive index in the transverse direction. The evolution of the complex light field amplitudeq is described by the reduced equation where the transverse η ζ and the longitudinal ξ coordinates are scaled to the characteristic beam width and diffraction length, respectively. We suppose here that the refractive index modulation is induced optically by the multiple incoherent zeroorder Bessel beams, so that the refractive index profile features the total intensity of the interference pattern, as it occurs in photorefractive crystals. The guiding parameter is proportional to the refractive index modulation depth, which is assumed to be comparable to the nonlinear contribution of the refractive index. The function describing the refractive index modulation profile is given by is the total number of beams in the array, the scaling parameter b defines the radii of rings of Bessel beams and here is taken small enough to ensure that the width of the central core of Bessel beams largely exceeds the wavelength; finally, and ζ are the coordinates of the Bessel beam centers. Note, that Eq. (1) admits several conserved quantities including the power or energy flow The stationary solutions of Eq. (1) that propagate along the guiding channels of the network have the form q w , where w is a real function and is the real propagation constant. General families of soliton solutions are defined by the propagation constant b , and by guiding and scaling parameters p and Since scaling transformation q p can be applied to obtain various soliton families from a given one (notice that this also implies variation of the lattice), below we set b and vary and . Families of stationary solutions were obtained by solving Eq. (1) with a relaxation algorithm. 2 2 ( , , , ) Although here we focus on the optical context, we stress that the concept we put forward here is expected to hold for Bose-Einstein condensates trapped in Bessel optical lattices.
Discussion
First, we address the simplest uniform line network created with an array of Bessel beams equally spaced along η -axis ( Fig. 1(a)). Because of the mutual incoherence of Bessel beams in the network, the interference pattern between pronounced guiding channels is suppressed. Such network can support stationary solitons whose profiles are elongated along the η direction. This is especially evident for low-power solitons that extend over several network channels ( Fig. 1(b)). With increase of the propagation constant (or power) the soliton width decreases (Fig. 1(c)), so that in the limit b soliton occupies only one channel and its power approaches the critical value of the unstable soliton supported by uniform cubic media. Nevertheless, for the considered set of parameters the soliton power was found to be a monotonically growing function of propagation constant that implies stability of stationary soliton solutions in the entire domain of their existence according to the Vakhitov-Kolokolov criterion. Below we consider broad low-power solitons that can be set in motion across the network by imposing the initial linear phase tilt (or angle) onto the input field distribution. Such tilted beams are no more stationary solitons, but they can travel across the uniform network along the η direction almost without the radiation losses ( Fig. 1(d)), that otherwise unavoidably appear upon crossing of the guiding channels of the structure and that then lead to a fast trapping of high-power narrow solitons.
exp( ) iαη
One of the interesting results found in this work is shown in Fig. 1(d): Removal of one of the Bessel beams from the array can cause deflection of soliton in the direction opposite to the input tilt. This process is accompanied by a strong beam reshaping in the deflection point, but the resulting beam shape and the modulus of the propagation angle are typically very close to the input ones. The soliton deflection at the network defect is a simplest example of operation accessible with opticallyinduced networks whose structure (open and closed paths) can be easily changed by blocking or switching of the individual Bessel beams from the array. The potential of such structures for creation of all-optical photonic circuits is readily apparent. [5,6]; however, the tunability afforded by the technique of optical-induction offer the additional advantage of the reconfigurability. Another interesting phenomenon occurs in the circular network, as the one depicted in Fig. 2(c). In this case soliton can perform rotary motion across the ring, with its power remaining almost constant because of small radiation losses. Switching off elements of the circular networks causes reversal of the soliton rotation direction. The similar results were obtained in a variety of cases, in terms of input light conditions and lattice-creating Bessel beams. These examples illustrate that the networks induced by the Bessel beam behave like soliton wires and thus they can be effectively used to manage soliton paths. Here we focus on the results obtained by simultaneous excitation of several coupler channels. We launch into neighboring or opposite channels two identical solitons whose profiles were found from Eq. (1) by means of relaxation method, under the assumption that only one Bessel beam is present. In the presence of other guiding channels (with the second beam launched into one of them) the propagation process is accompanied by periodic transfer of the energy between different channels that are coupled via their tails. The energy exchange process is very sensitive to the relative phase of input beams. We show the output intensity distributions at the distance corresponding to the maximal energy transfer into two channels that were not initially occupied. Thus, in Fig. 3, where solitons are launched into neighboring channels, one finds that the energy transfer is negligible for in-phase solitons and almost 100% efficient for out-of-phase solitons. The outcome changes drastically when input solitons are launched into the opposite coupler channels. Therefore, the important result is that by varying the refractive index modulation depth (which is proportional to the intensity of nondiffracting Bessel beams inducing the coupler), the separation between Bessel channels, the power and phases of input solitons, it is possible to achieve a variety of switching scenarios.
Concluding remarks
We conclude stressing the potential of reconfigurable soliton networks optically induced by multiple incoherent nondiffracting Bessel beams in Kerr-type nonlinear media. We showed that dynamics of soliton beams propagating in such opticallyinduced networks can be used for all-optical manipulation of light [5,6], with the important additional advantage of the easy reconfigurability afforded the opticalinduction concept. The scheme holds for light signals propagating in focusing cubic nonlinear media and for Bose-Einstein condensates trapped in optical lattices induced by Bessel beams. | 2015-03-21T21:52:17.000Z | 2005-01-21T00:00:00.000 | {
"year": 2005,
"sha1": "3d230739780e059043033acfa220962118c1d6a1",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.1364/opex.13.001774",
"oa_status": "GOLD",
"pdf_src": "Arxiv",
"pdf_hash": "6e880143f76f0cf1db1c6436261427ff817e8c4f",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics",
"Medicine"
]
} |
228907143 | pes2o/s2orc | v3-fos-license | Feelings perceived by students during the phases of accounting dissertation guidance
ABSTRACT The aim of this study was to understand how students graduating from accounting master’s courses perceived the feelings experienced in the different dissertation guidance phases. This investigation enables us to identify which stages generate negative feelings and to thus propose actions to improve the affective relationship between the advisor and student, considering that the feelings substantially affect the potential of the scientific works originating from that relationship. The advisor-student relationship and the feelings involved in it are among the main factors that positively or negatively affect the conclusion of the academic work, and understanding it is important in order to improve this interaction process. One of the implications of the research relates to the importance of stricto sensu programs monitoring the guidance process and somehow searching for mechanisms that promote harmony in the guidance process, as the experiences during the dissertation guidance process can affect the student’s likelihood of continuing on their academic journey. For the data collection, a survey was conducted in which the participants used emojis to represent the feelings experienced in their relationship with the advisor during the construction of their master’s dissertation. In addition, semi-structured interviews were carried out, which enabled the elaboration of a collective subject discourse regarding the feelings experienced during the dissertation guidance process. The 88 responses and eight interviews enabled the identification of feelings such as anxiety, fear, and insecurity in the first phases of the dissertation building. Abandonment, confusion, and frustration were felt in the last phases of the relationship with the advisor. Regarding the positive feelings, there are indications of joy, satisfaction, and calm, presenting greater frequency in the last stages of the dissertation. The collective subject discourses also indicated that the relationship with the advisor affected the quality of the scientific research and the student’s academic trajectory.
INTRODUCTION
Teachers' behavior and the affective way in which they treat their students arouse motivation and interest in the teaching-learning process, leading to greater student participation and, consequently, better assimilation of the knowledge presented in the classroom. Moreover, teachers who express their affectivity and reach their students in a positive way improve social skills, help to develop the students, and promote unity and cooperation (Reis et al., 2012).
The attention paid to cognitive aspects instead of affective relationships has been a focus of discussions in the university environment (Veras & Ferreira, 2010). Studies indicate that the relationship between teachers and students have been an influencer of the quality of the activities developed in post-graduation, especially regarding the guidance process for scientific works (Falaster et al., 2017). The advisor performs the role of guide in the student's academic journey; part of the student's success or lack of it can be attributed to their interactions with the advisor, transforming this relationship into a form of partnership in which the final product is the master's dissertation or doctoral thesis (Nóbrega, 2018).
The guidance process and the relationship between the advisor and student are investigated in the literature from various perspectives. Nóbrega (2018) addressed the relationship in academic guidance processes and the formal role of the advisor; Falaster et al. (2017) discussed the impact of the guidance on the future publications of PhD students; Massi and Giordan (2017) addressed the advisor's training; and Leite and Martins (2006) analyzed the influences of the advisor-student relationship on the production of theses and dissertations in business and accounting courses.
The evidence found in those studies revealed a lack of professionalization and formalization of the advisor figure (Massi & Giordan, 2017;Nóbrega, 2018), it indicated the positive influence of the production of advisors on the quality of students' publications (Falaster et al., 2017), and it indicated students' lack of commitment and respect for deadlines and toward the stricto sensu activities, compromising the quality of the scientific research. In addition, these papers shed light on the existence of an autocratic and verticalized environment in which the guidance occurs asystematically and where the students experience feelings of anguish and insecurity.
The feelings originating from the guidance process form the core of this study, given that this element may somehow affect the behavior and activities carried out by the students, such as their relationship with the content addressed in the teaching environment (Mahoney & Almeida, 2005;Tassoni & Leite, 2013). Thus, this research specifically employs the concept of feeling as a constituent part of the affective dimension. Consequently, the theoretical scope of the research is based on Henri Wallon's (2007) Theory of Development, which places feelings within the affective domain, allowing for expression of the sensations felt based on certain experiences (Gratiot-Alfandéry, 2010).
The student's academic performance and their cognitive and affective development are influenced by a vast set of conditions that, in most cases, can be measured by the quality of the interpersonal relationships that occur in the school environment; among these is that of the teacher with the student, which is affected by variables such as beliefs, feelings, motivations, and abilities (Del Prette et al., 2005).
In a correlated study, Leite and Martins (2006) addressed the influence of the advisor-student relationship on the process of elaborating theses and dissertations in accounting post-graduate courses in the municipality of São Paulo. The reports highlighted the students' insecurity and solitude, in relation to the guidance process, caused primarily by a lack of interaction with and distance from their advisors. They discovered that advisors value students' technical characteristics, while students give a high level of importance to the personal and affective dimensions in their relationship with advisors.
Therefore, the discussions on the importance of the advisor-student relationship for the academic activities developed in stricto sensu and the importance of addressing the feelings originating from that process lead to the concern that guides this research: what are the feelings perceived by students graduating from accounting master's courses in the different dissertation guidance phases? Consequently, the aim of this research consisted of understanding how students finishing accounting master's courses perceived feelings experienced in the different dissertation guidance phases. Graudating students make up the body of analysis of the study as their feelings are consolidated over time, making this group more suitable for the analysis than students who are still in the middle of master's or PhD post-graduate courses.
In stricto sensu, the restructuring of the criteria of the institutional evaluation model used by the Coordination for the Improvement of Higher Education Personnel (Capes), while strengthening observation of the scientific production in quantitative terms, downplays the attention given to qualitative aspects (Nascimento, 2010). The interaction, partnership, comradeship, and friendship between the advisor and student, resulting from the longer time needed to build the work, take second place to a heteronomous, productivist, and superficial relationship due to the excess work given to the advisor and the demand for productivity indicators (Alves et al., 2012).
This misalignment has worried researchers, given that these relationships embody "the most concrete face of post-graduate programs, that is, it is the work of the teachers and students that results in the research and publications" (Nóbrega, 2018(Nóbrega, , p. 1056. Thus, this research is warranted as it enables an understanding of one of the aspects involved in the dynamism of the relationship that exists in guidance processes: the feelings developed among individuals who strive to build the scientific knowledge in the country and to obtain a prominent position within the universe of the most productive. Observation of core elements of the knowledgebuilding process enables the construction of a social representation in order to outline the feelings originating from the guidance process and, consequently, to identify aspects that can be improved in this interaction. It is known that the quality of the advisor-student relationship is seen as one of the main factors that positively or negatively affect the conclusion of academic work (Alves et al., 2012). Nonetheless, there is a lack of investigations that address the guidance process in Brazil and abroad (Nóbrega, 2018). Levecque et al. (2017) also highlight the relevance of considering psychological aspects in the post-graduate environment, as the pressures that permeate it can affect students psychologically.
Specifically in business area courses, exploring the feelings experienced during dissertation guidance processes can contribute to an evaluation of the guidance conditions in which the scientific works are developed and explore these conditions to solve the interpersonal relationship conflicts that occur in this environment (Leite & Martins, 2006). Dias et al. (2011) warn that a lack of pedagogical training of advisors can mean their guidance style is affected by the successful or unsuccessful experiences they have had with students, with feelings occupying a key role in this relationship.
FEELINGS AND THE ADVISOR-STUDENT RELATIONSHIP
Universities, previously seen as places of exclusive cognitive training, have sought to include affectivity in their educational practices, encouraging reflections on their place in the individual's entire human education (Ferreira & Acioly-Régnier, 2010). Given the amplitude of the concepts that permeate affectivity, researchers with different interests have examined the theme from different perspectives, such as Henri Wallon (1941Wallon ( -2007, who highlighted affectivity as an essential dimension in knowledge building and of people over the course of their intellectual development (Veras & Ferreira 2010).
Henri Wallon's (2007) Theory of Development unites affectivity and cognition so that they are dialectically in a constant dynamic, altering in the different stages of knowledge that the individual acquires during their existence (Gratiot-Alfandéry, 2010). Mahoney and Almeida (2005) explain that affectivity involves the individual's propensity to be affected through the external or internal environment by sensations that may be good or bad. In the Wallonian theory, there are three distinct moments that are consequences of organic and social elements in the evolution of affectivity: emotion, feeling, and passion.
Emotion is displayed through bodily expression and represented by the person's instantaneous attitudes.
Feeling, as the representational expression of affectivity, is displayed through mimicry or language. Passion is revealed as a form of behavioral self-control used to silence emotion, emerging with the evolution of mental representations (Mahoney & Almeida, 2005;Veras & Ferreira 2010).
The distinction between the concepts of emotion, feeling, and passion and their representations is one of the main contributions of Wallon's theory, which enabled affectivity to be defined as a functional domain composed of different moments that evolve over the course of development and that originate from the organic until reaching the cognitive, as can be observed in feelings (Ferreira & Acioly-Régnier, 2010).
Specifically regarding feelings, Wallon argues that these do not imply sudden and momentaneous behaviors, like emotions do, as they are the opposite of these and tend to contain them (Veras & Ferreira, 2010). In feelings, people have more mechanisms of representational expression, given that they analyze before assuming an attitude, they wait for the right moment to express themselves, and they intellectually portray their reasons and context in a more evolved way (Mahoney & Almeida, 2005).
According to Tassoni and Santos (2013), the classroom environment is permeated by a dynamic of knowledge R. Cont. Fin. -USP, São Paulo, v. 32, n. 85, p. 158-173, Jan./Apr. 2021 and feelings; this combination demonstrates that aspects of the affective domain are important for the teacher's training and for the development of effective pedagogical practices. For Pereira et al. (2016), feelings are assignable, representational, and consolidated over time; they correspond to the affectivity function and can be denominated and detailed more assertively in a study. Silva (2005) used the Wallonian theory to study the relationship between the feelings and emotions of students and teachers in both theory and practical classes and the teaching-learning process. A large portion of the participants revealed that they believed that feelings and emotions influenced the teaching-learning process. The teachers reported feelings of pleasure, accomplishment, calm, and motivation, especially in the practical classes. The students described negative feelings associated with discouragement, displeasure, insecurity, anxiety, and fear in both the practical and theory classes. However, feelings related to motivation, pleasure, and accomplishment were reported more frequently in the practical classes.
With regard to the journey toward obtaining the master's or PhD title, the feelings originating from the advisor-student relationship are a constituent part of this affective dimension and are also present in the various stages of the student's education. Over the course of this journey, the student is led by a teacher-advisor (Nóbrega, 2018). Competence to guide a student goes beyond specific knowledge, techniques, and scientific methodology, also implying didactic abilities that primarily require the capacity of the advisor to establish a significant coherent learning relationship with their student (Dias et al., 2011).
For Viana (2008), the advisor and the student are two figures who are linked by a choice or by an obligation to share the knowledge-building journey. In this journey, both are people with feelings that fluctuate from one extreme to another, as in all human relationships.
The guidance process for academic studies is one of the most important teaching activities, although in Brazil it is neglected and lacks teacher training to develop the competences needed for it (Costa et al., 2015). For Nóbrega (2018), it is possible to perceive that in postgraduate programs the advisor-student relationship is unexplored and treated as an unimportant element for the student's education and conclusion of the dissertation or thesis. However, the author notes that the production of knowledge and research occurs in the actual contact between those involved, as the affectivity is interlinked with the scientific work, as in any practice that involves people.
During the development of theses or dissertations, it is essential for the advisors and students to know their attributions and establish a productive relationship that fosters the knowledge building. When the advisor does not perform their role, the relationship with the student becomes compromised and has negative effects on the knowledge-building process and on the quality of the scientific research (Leite & Martins, 2006).
In Henri Wallon's (2007) Theory of Development, the teaching-learning process is considered a unit in which the interpersonal teacher-student relationship plays a definitive role, given that the affective-cognitive-motor dimensions work in an integrated way. Seeing the student in their entirety and reality is a challenge for the teacher, who is not normally trained in these joint dimensions (Mahoney & Almeida, 2005).
Thus, the teacher's role of teaching has become, over the years, more comprehensive and complex, as they have gone from only fullfilling the role of conveyer of information and knowledge to being the student's guide or mediator during the knowledge-building process. This role requires training composed of diverse knowledge and attitudes that enable the student to assimilate the cognitive, affective, and behavioral concepts during the process of learning the curricular content, as these are linked to the student's performance in the learning environment (Ribeiro, 2010). Leite (2012) emphasizes that the teacher-student relationship is also affective in nature and, as it develops, it causes positive or negative affective results in the relationship between the students and the curricular content developed, such as affinity or distance. Within this context, the teacher performs a relevant role in the knowledge building and has been the object of studies at different educational levels (Camargo, 2017;Pereira et al., 2016;Ribeiro et al., 2005;Veras & Ferreira, 2010). Specifically in stricto sensu, positive and negative results are also found, originating from the dissertation and thesis guidance process. For example, Leite and Martins (2006) state that the building of knowledge during the guidance process is not a solitary action and that the relationship between the advisor and student is essential, as it is from the particular interaction with advisors that research will be produced that is capable of contributing to the growth and solidification of the science in the area studied. Viana and Veiga (2010) highlight that the success of the relationship and of the scientific production requires the advisor and student to establish interaction rooted in dialogue, dedication, interest, discipline, responsibility, and commitment. These can inhibit the difficulties found both by advisors, such as a lack of time for their students to finish the activities within a certain timeframe, and by R. Cont. Fin. -USP, São Paulo, v. 32, n. 85, p. 158-173, Jan./Apr. 2021 students, such as the unavailability of advisors to attend to them. Costa et al. (2015) highlight that it is up to the advisor to guide the student during their education, as the student's development process is experiential and can occur based on their experiences in the classroom, in their relationship with teachers, course mates, and, especially, with the advisor themself. Falaster et al. (2017) argue that there is a relationship between the quality of the guidance and that of the student's academic development. The results of their study indicated that more experienced advisors with higher quality publications influence their students to produce quality publications instead of a high quantity. Sheldon et al. (2015) evaluated the relevant qualities of advisors inherent to the student's knowledge, to their willingness, and to supporting their autonomy. The evidence indicated that supporting their autonomy was the most relevant factor in students' satisfaction and performance in relation to their advisors. The authors also revealed the relationship between the time the advisor makes available in the meetings with their students and the autonomy-supporting factor, given that this factor considers not only the way the student can acquire knowledge from their advisor, but also how the advisor helps them to develop interpersonal skills.
Within this context, the findings and the discussions that exist in the literature shed light on the relationship between the advisor and student, recognizing that the feelings experienced during the different phases of the guidance process can affect the quality of the activities developed and indicate elements that can be improved in that process.
METHODOLOGY
The descriptive research uses as its field of study students graduating from nine master's courses in the South Region of the country, which are affiliated with the following higher education institutions: the Regional University Foundation of Blumenau (FURB), the Federal University of Rio Grande (FURG), the State University of Maringá (UEM), the Federal University of Paraná (UFPR), the Federal University of Rio Grande do Sul (UFRGS), the Federal University of Santa Catarina (UFSC), the State University of the West of Paraná (Unioeste), the University of Vale do Rio dos Sinos (Unisinos), and the Communitary University of the Chapecó Region (Unochapecó). The sample is characterized as nonprobabilistic and was defined according to accessibility.
The South Region of Brazil was chosen to make the study economically and operationally viable. The data collection was operationalized in two stages: a survey and semistructured interviews. In the first stage, the online survey was carried out on the SurveyMonkey ® platform with the aim of identifying the feelings experienced during the master's dissertation guidance process based on a questionnaire composed of two sections.
Section I contained the 10 phases that form the dissertation-building period and one question summarizing the whole process; with this, it was possible to capture the feelings experienced in each stage of building the dissertation. Table 1 shows the guidance phases that composed the questions in the research instrument. For each phase of the dissertation, the respondents were asked to indicate between one and three emojis with connotations of positive or negative feelings experienced in the respective phase of the research. Each emoji was given a number and the participant had to indicate the box corresponding to the emoji number. Ten emojis with negative connotations were shown and 10 with positive connotations. After indicating the emojis, in each phase an open text field was included for the respondent to indicate between three and five feelings experienced during that stage of the dissertation. Novak et al. (2015) mention that emojis enable feelings to be expressed based on the maximization of non-verbal elements that draw attention and allow for greater understanding of the message by means of images. Thus, the use of emojis aims to share different affective representations in two different groups: positive feelings and negative feelings. The study does not intend to discuss whether the respondents have unanimously labeled feelings for each emoji, but one of the aims of the interviews was to pluralize the representation of the feelings motivated by the use of the emojis indicated as the most frequent ones in this study. The selection of 20 emojis (10 positive and 10 negative) resulted from a search in published materials about those most used on a daily basis.
To validate the research survey externally, two research specialists (both affiliated with the post-graduate programs of the authors of the article) were contacted. It was certified that these specialists frequently use emojis to represent feelings in their day-to-day and that they are researchers from the area of accounting education. Their suggestions were accepted in full and resulted in the final version used in this study and presented in Figure 1. Section II aimed to characterize the research participants. Based on the answers obtained, it was possible to outline the profile of the 88 respondents according to Table 2. Most of the respondents were of the female gender (59.09%), 61.36% are between 26 and 35 years-old, 86.39% live in the South Region, 60.23% did not change city to take their master's, and 45.45% received financial assistance during this period. The data obtained in the first stage of the study were analyzed using descriptive statistics. To identify and count the feelings described in the first stage of the research, the Text Analyzer ® online tool was used.
In the second stage, semi-structured interviews were held with eight research participants, with the aim of understanding the experiences and feelings experienced during the guidance process, as well as identifying what the negative and positive emojis most indicated in the first stage represent in terms of feelings for the students finishing the master's. In the survey applied in the first stage, a field was included for the respondents to provide their email for participation in the second stage, which consisted of the interviews. The interview participants were chosen with the aim of covering the plurality of the courses. Thus, interviewees were chosen from different programs in order to minimize the risks of self-selection of opinions. The interview period occurred in January and February of 2019 and each one lasted approximately 30 minutes. The data saturation method was employed, in which the interviews were concluded when adding new participations did not add fresh and crucial evidence to the reports already gathered (Sampieri et al., 2013). Table 3 presents the script used in the semi-structured interviews. Table 3 Semi-structured interview script What types of feelings do the most indicated negative/positive emojis convey? Do you think that during the guidance process positive or negative feelings prevailed in relation to the advisor? Could you describe why you had that sensation?
If you could go back in time, would you stick with the same master's advisor?
If the program presented a list of advisor names for continuing the studies into PhD level, would you stick with the same advisor?
Would you recommend your advisor to friends you wish well?
Source: Elaborated by the authors.
It is observed that the interviews focused on identifying feelings in the advisor-student relationship and not on the phases for elaborating the work (data collection, tabulation, and analysis).
Besides the pre-defined script, secondary questionnaires were carried out to obtain more details of the experiences and feelings experienced during the dissertation process. The evidence gathered were analyzed, as well as the descriptive statistics, using the collective subject discourse technique, which is "a way of retrieving and presenting the social representations obtained in empirical research" (Lefevre & Lefevre, 2014, p. 503). Thus, the key expressions from the interviewees' speeches were identified, referring to the positive and negative feelings of the students graduating from the accounting master's and grouping the main ideas to build the collective subject discourse derived from these first-person reports, given that the content of the opinions with a similar meaning present in different accounts are associated with each category, in order to use this content to form a synthesized account, drafted in the first-person singular, as if it were a collective speaking as an individual person. (Lefevre & Lefevre, 2014, p. 503) Afterwards, collective discourses were built to represent the positive and negative feelings of the interview participants.
RESULTS ANALYSIS
The concepts proposed in the Frenchman Henri Wallon's Theory of Development contribute to the understanding of the affective dimension that feelings form part of. For Mahoney and Almeida (2005), understanding these concepts unveils important aspects to be considered for a satisfactory and constructive teaching-learning process. Table 4 shows the five feelings described and the three negative and positive emojis most indicated in each phase of the relationship between the advisor and student over the course of the dissertation. To identify the total feelings described, the Text Analyzer ® online tool was used in the open answer field. In the case of a tie, all the feelings and/or emojis within the same percentage of occurrence are listed. Through the feelings and emojis indicated by the respondents, it was noted that the relationship between the student and advisor in the topic discussion and research question phase was characterized by negative feelings of anxiety, fear, and doubts originating from the beginning of their interactions for building the research proposal. Similarly, during the discussion of the objectives and contributions of the dissertation stage, there was a predominance of insecurity, solitude, and anguish regarding the construction of the work and the established relationship. In the initial stages of the scientific research in which the study proposal is outlined, the advisor's attention is important, as the knowledge building does not occur in isolation and the interaction between the advisor and student is fundamental to reduce the scientific insecurity of the students and avoid damage in the following stages of building the dissertation (Leite & Martins, 2006).
The relationship experienced in the period for discussing the structure of the theoretical framework was considered laborious and a generator of anxiety, tiredness, and solitude. In contrast, reports of feelings of satisfaction, confidence, support, and assistance from the advisor were also noted. The findings of Leite and Martins (2006) show that the student-advisor relationship is characterized by meetings that are asystematic and insufficient to meet the student's needs, which aligns with the feelings of solitude listed by the respondents. Nóbrega (2018) mentions that the advisor should specify, in their meetings with the students, readings and studies that enable the scientific research to be built, this being one of the aspects that can help in overcoming the difficulties in academic writing. The methodology discussion phase was shown to be permeated primarily by fear, anxiety, and confusion. Nóbrega (2018) recommends that, in this period, the advisors encourage the students to carry out scientific readings and improve their academic writing, given that this ability is essential for the methodological development of the research (Nóbrega, 2018). Despite the evidence indicating that, when choosing students, advisors prefer those who have previous knowledge in the methodological field (Leite & Martins, 2006), negative feelings permeate the relationship in this phase of the dissertation, seen as an essential stage for completing the work, as it involves the operationalization of the research.
In the phase for discussing the adjustments suggested by the qualifying panel, there was a predominance of positive feelings of calmness, joy, and satisfaction. This period marks the start of the academic research, a pre-requisite for subsequently defending the work (Nóbrega, 2018). It is understood that, despite being a stage that generates fear, anxiety, insecurity, and nervousness, the qualifying panel's approval tends to provide a more harmonious climate that directs the discussion of adjustments with the advisor, as the effects generated by the qualifying panel's approval promote calmness in the student.
Based on the qualification phase, an increase was perceived in the indication of positive feelings. In this context, the student's maturity collaborated so that, in the discussion phase regarding the results analysis, there was a sensation of satisfaction, joy, confidence, enthusiasm, and determination. Also worth highlighting are the feelings of tiredness and exhaustion in the relationship between the advisor and student. In the phase for discussions and conclusions of the dissertation, there was a predominance of feelings of relief, satisfaction, joy, confidence, peace, and pride.
In the phase for discussing the adjustments suggested by the defense panel, positive feelings of calmness, relief, happiness, satisfaction, and joy were observed. Nonetheless, there are reports of negative feelings of anger, solitude, anguish, ignorance, frustration, tiredness, and anxiety. In the research stage, the feelings of satisfaction and frustration originating from the relationship with the advisor may be mirrored in the quality of the dissertation perceived by the student. The reports shown in the research of Leite and Martins (2006) converge with this aspect, as the quality of the scientific research depends on the dialogue between the student and advisor, engagement, empathy originating from this relationship, and indications by the advisor of paths to be followed by the student.
The discussions on the article clippings generated by the master's dissertation were permeated by positive feelings of joy, satisfaction, accomplishment, calmness, and enthusiasm. In a lower percentage, there was an indication of difficulties and abandonment during the process of discussions inherent to the articles originating from the dissertation.
With relation to the discussions on the continuity of studies in the program, feelings of joy, calmness, incentive, commitment, satisfaction, support, enthusiasm, and pride were indicated, which are aspects that can promote the student's desire to continue with their academic training in the same post-graduate program as the master's. However, there were feelings of tiredness, anger, abandonment, frustration, and absence of the advisor regarding these discussions, leading to non-continuity of academic training in the post-graduate program.
Actions formed and displayed based on antagonistic feelings can generate conflicts in the advisor-student relationship and make the teaching-learning process more difficult, as it involves a hierarchical relationship (Silva, 2005). The Wallonian theory helps to recognize the students' affective displays as signs of situations experienced in a positive or negative way.
Thus, understanding how affectivity manifests through feelings enables the definition of strategies suited to the specificities of each student (Gratiot-Alfandéry, 2010). In order to broaden the understanding of these feelings, semi-structured interviews were conducted and collective subject discourses were built based on the speeches of the eight graduating students interviewed. First, the three negative and positive connotation emojis most indicated in the survey were chosen and the interviewees were asked to describe the feelings that those emojis represent in the relationship between the student and advisor. Table 5 shows the percentage of indications of each emoji with a negative connotation, highlighting the most frequent ones in each phase. The most indicated emoji with a negative connotation was 7, which was indicated 159 times during the dissertation-building phases, followed by emoji 2, with 147 indications, and by 6, with 114 indications. By means of the interviewees' reports, it was possible to build a collective subject discourse that explains why these emojis were the most indicated in the interviewees' perception, as presented in Table 6.
Table 6
Representation of the most-indicated emojis involving negative feelings Represents dynamism, negativity, and disagreement with the way the guidance was given. Dissatisfaction and disappointment with the guidance gave rise to feelings of disapproval and conflict, meaning the guidance was marked as a negative experience.
Represents those students who had a misalignment of ideas with the advisor. The attitudes and ideas did not converge, generating confusion. There were conflicts in the relationship, generating stress, nervousness, anguish, desperation, and mental exhaustion.
There were disappointments during the course of the master's, having an impact in the form of dissatisfaction, anger, and frustration.
Source: Elaborated by the authors.
The negative emojis represent feelings generated by turbulences in the guidance process. The feeling of abandonment and differences in attitude between the advisor and students lead to feelings that affect the construction of the research. Falaster et al. (2017) observed that the teacher's availability and support of the student's autonomy interfere in the result of the scientific research. It is important for course coordinations to observe these aspects throughout guidance processes.
Similarly, the most indicated positive connotation emojis were identified through the construction of Table 7. The most indicated positive connotation emoji was 11, with 300 indications, followed by 19, with 194 indications, and by 12, with 175 indications. The collective subject discourses that represent the interviewees' perceptions are shown in Table 8 Table 8 Representation of the most indicated emojis involving positive feelings Represents praise for the dedication, relief about the research results, overcoming the obstacles and reworking, alignment and conclusion of the research in a positive way and with good feelings about accomplishing the work.
Represents feelings of gratitude for the knowledge built with the advisor, happiness, joy, and pride about the work in partnership with the advisor, recognition by the teacher of the student's effort in developing the research, and the establishment of a harmonious relationship.
Represents overcoming obstacles, enthusiasm about the guidance, conclusion of stages of the dissertation, and a feeling of pride.
Source: Elaborated by the authors.
Next, collective subject discourses were built for questions that composed the script of the semi-structured interviews. Table 9 shows the discourses for the feelings that prevailed in the relationship. Table 9 Feelings that prevailed Do you think that during the guidance process positive or negative feelings in relation to the advisor prevailed?
Could you describe why you had this sensation?
Negative connotation
My guidance process was very painful and stressful; I felt abandoned, neglected, alone, and with no support from my advisor.
It was very frustrating in terms of the human relationship. There was a lack of good judgement. I think that they lacked sensitivity in relation to me as a student, they were never available and that was infuriating; I had to seek help from course mates and other professors on the program and, if it weren't for them, I wouldn't have coped.
Positive connotation
We had a good relationship, they were very receptive, I always had access to them and all the decisions were shared, I always had their help, I didn't feel alone, they gave me the freedom to grow in my own time.
I have good memories and feelings of gratitude, partnership, friendship, collectivity, and comradeship.
Source: Elaborated by the authors.
The collective discourse with a negative connotation draw attention to the attitude of abandonment and negligence taken by the advisor regarding the development of the student's research. These reports explain the presence of feelings of abandonment, absence, and insecurity, such as those indicated in the survey. This lack of monitoring generates psychological consequences in the student and affects the quality of the academic work developed, where the support to overcome difficulties comes from friends and other professors who accompany them in their academic journey. The discourse with a positive connotation indicates that the harmonious relationship between the advisor and student implies the student gradually maturing academically, with this relationship being strengthened through feelings of gratitude and comradeship.
The findings contribute to the existing evidence in the literature, for example in the paper by Viana and Veiga (2010), who reinforce the idea that a good relationship with the advisor is essential for the conclusion of the work, in which the advisors' availability, care, and receptiveness favor the completion of the research. In contrast, feelings of solitude and abandonment, as well as the advisor's unavailability, cause damage to the advisor-student relationship and, consequently, to the quality of the work.
Wallon's theory values the teacher-student relationship and the educational environment as essential elements in the person's full development (Ferreira & Acioly-Régnier, 2010). In this theory, the educational environment is formed of the human relationships, physical objects, and knowledge, and committed to the student's evolution in a practice that involves the social and individual dimension (Galvão, 2000). Within this context, the guidance is a joint process in which the development of the work occurs in the communication established between the parties involved, by means of the analytical and constructive vision of the advisor, who performs the role of stimulating the student's desire to carry out the research and seek knowledge and their intellectual autonomy (Viana, 2008). Table 10 shows the discourses relating to the exchanges with the advisor, to the continuity of guidance, and to recommending the advisor to friends. The reports denote that those who had negative experiences during the master's guidance process would choose to change guidance and would not continue with the guidance into PhD level. Thus, the results are able to direct the focus toward the organization and management of the guidance of the accounting post-graduate programs investigated. The allocation of guidance should not be seen as something that is fixed, but should instead allow the student or advisor to choose to interrupt the guidance if there are any ambiguous situations. Misalignment of work styles and personality between the student and advisor can cause friction and weaken the academic development process.
The discourses with a positive connotation reinforce valued aspects that even transcend the advisors' technical qualities. The reports indicate that greater affective proximity with the advisors fosters synchronism in the development of the scientific research and maximizes the desire to continue the guidance process into PhD level, as well as to recommend the advisor to others. Within this scope, Leite and Martins (2006) show that students tend to value the affective characteristics of advisors, while advisors prioritize students' technical skills. It is believed that the master's adaptation period collaborates in affective qualities being valued by the students, given that over the course of the master's they face various challenges that require the support of others. Table 11 shows the discourses built regarding other aspects in which the advisor was more important in the process and influenced the quality of the academic work. It is noted that the initial phase of the dissertation building is characterized by greater dependence on the advisor, when the student anxiously waits for the initial contact with their advisor to discuss the work. The student's expectation is to engage in a dialogue and, from then on, have a direction. The advisor, in turn, expects the student to have an interesting research idea that fits their research line. This is a decisive phase in building a relationship that will strengthen throughout the process of elaborating the dissertation, positively or negatively affecting the quality of the work (Viana, 2008).
The reports are aligned with the feelings indicated in the survey, given that in the initial stages there is a predominance of fear, insecurity, doubt, and anguish originating from the relationship with the advisor and from the dissertation building. In this sense, the advisor was described as the student's guide, where their role was to indicate the paths to be taken and help in defining the overall scope of the research. The collective subject discourse reveals that their essentiality directly affects the quality of the work developed, which is also argued by Sheldon et al. (2015), as well as the emotional state of the students throughout the dissertation-building process.
Thus, numerous feelings are experienced during the dissertation guidance process. It is perceived that these feelings provide drive or cause obstacles over the course of building the research, being factors that interfere in the quality of the academic work and that affect the student's psychological state. Distance between the advisor and the student is a barrier in the guidance process and to the success of dissertations, besides being a determinant of the feelings of anguish and solitude displayed by the students. Thus, it is essential for the guidance process to be based on respect and democracy with the aim of driving the advisor-student relationship in a harmonious way, promoting dialogue, and reducing the pressure and autocracy over the students (Alves et al., 2012).
The relationship between affectivity and learning has been considered an essential element for discussing different themes, such as the advisor-student relationship, academic performance, and motivation, among others (Tassoni & Santos, 2013). Within this context, the concept of affectivity can be observed in the discussions about feelings, as addressed in this study, in which their manifestation is perceived as a response to the events experienced during the dissertation guidance period.
CONCLUDING REMARKS
The relationship between the advisor and student has been indicated as a relevant element in the construction of academic research. Focusing on master's dissertations, the advisor figure plays a predominant role in the student's adaptation to the academic demands and in overcoming the adversities that can occur along the way. The evidence in the literature (Alves et al., 2012;Falaster et al., 2017;Leite & Martins, 2006) shows that the alignment, interaction, and feelings experienced during the guidance process are able to affect both the quality of the scientific research and the student's academic development.
The findings of this research indicated the predominance of negative feelings at the start of the guidance process, when anxiety, fear, anguish, and insecurity characterize the first phases of the relationship with the advisor. Over the course of the dissertation, other feelings were reported, such as abandonment, confusion, and frustration, highlighting that the accumulation of stressful elements originating from the dissertation building and from the relationship with the advisor marks the stages of the dissertation and causes obstacles to development in stricto sensu. In contrast, there were reports of a maximization of positive feelings, such as joy, satisfaction, happiness, and calmness, which collaborate with the establishment of an affective relationship based on comradeship, support, and an understanding of the student's difficulties.
The relationship established during the guidance is able to affect the likelihood of the student continuing their education, as well as being linked to the quality of the activities developed in stricto sensu. The discordances found by the students regarding their advisors may be due to a lack of professionalization and clear guidance concerning the advisor's role in the Brazilian university environment (Massi & Giordan, 2017). It is suggested that post-graduate programs monitor and evaluate the quality of their guidance by consulting the students. Similarly, advisors should be consulted with the aim of verifying the development of the student and their research.
The definition of rules and the possibility of changing guidance could also be addressed as a formal process in stricto sensu courses that is carried out without obstacles. Misalignments in work styles and personality could thus be addressed as something that is common, with their being structures in the programs for overcoming these adversities.
The representations obtained through the emojis and the indications made by the participants express the plurality of feelings that permeate all the phases of dissertation building. As proposed in Henri Wallon's (2007) Theory of Development, the social representations obtained in the study indicate that the feelings are perpetuated and consolidated over time, and the concerns weigh on the impact that these feelings cause in the development of the students' academic journey, as well as on the perpetuation of these practices when they themselves carry out teaching activities.
Thus, periodic monitoring of the students' emotional state and the availability of channels for helping them to overcome difficulties that can arise throughout the research are presented as practical implications provided by the evidence. In the theoretical field, building social representations based on feelings represented by emojis presents an advance for the topic, as emojis are visual representations that are widely used on an everyday basis and are able to convey a plurality of feelings that it would be hard to show through other forms of communication.
The limitations of this study consist of the absence of research that uses emojis to represent feelings that would enable a possible comparative analysis. On the other hand, this limitation can also be understood as one of the methodological contributions of the research. For future research, we suggest analyzing the formal mechanisms of post-graduate programs, in order to verify how conflicts in the relationship between advisors and students are addressed by advisors and course coordinators. Moreover, covering and comparing the different regions of the country could outline the profile of the relationships produced in stricto sensu accounting courses and enable benchmarking to be practiced among the programs. | 2020-11-19T09:07:43.694Z | 2021-04-01T00:00:00.000 | {
"year": 2021,
"sha1": "ccd6ae0bf0d5ac6a940a1069656dee17fe2f22b7",
"oa_license": "CCBY",
"oa_url": "https://www.scielo.br/j/rcf/a/RjpSvx7YJbrvpjH8vfR5gVP/?format=pdf&lang=en",
"oa_status": "GOLD",
"pdf_src": "Adhoc",
"pdf_hash": "8095fed262368e40975f6080e159465120cc2ad7",
"s2fieldsofstudy": [],
"extfieldsofstudy": [
"Psychology"
]
} |
252090078 | pes2o/s2orc | v3-fos-license | Understanding the Quality Factor of Mass-Loaded Tensioned Resonators
Mechanical resonators featuring large tensile stress have enabled a range of experiments in quantum optomechanics and novel sensing. Many sensing applications require functionalizing tensioned resonators by appending additional mass to them. However, this may dramatically change the resonator mode quality factor, and hence its sensitivity. In this work, we investigate the effect of the crossover from no mass load to a large mass load on the mode shape and quality factor of a tensioned resonator. We show through an analytical model and finite element analysis that as the load mass increases, surprisingly, the resonator mode shape becomes independent of the exact load mass, and therefore, its quality factor saturates. We validate this saturation effect experimentally by measuring the quality factor of a silicon nitride trampoline resonator while varying the load mass in a controlled manner.
An adequate sensor requires both appreciable coupling to the physical parameter it is designed to sense and as low as possible environmental noise. An important figure of merit that quantifies the sensitivity of a resonator normal mode is the quality factor, denoted hereafter as Q. The Q depends on environmental isolation and is a measure of the number of coherent oscillations a resonator mode can provide before losing a significant portion of its energy to the environment [1,21]. Therefore, resonator modes with utmost Q values are ideal for sensing.
To couple a sensor to a quantity of interest, it is often necessary to functionalize it with a coupling agent, resulting in an additional resonator mass, e.g. magnetic particles for magnetic force sensing [11,13] or test masses for acceleration detection [16,17]. Certain applications, such as those that aim to measure gravitational forces, require appending a mass that is much larger than the resonator mass [22][23][24]. In general, specific interest is devoted to modes in which the mass moves appreciably. For a large mass, typically, a single mode has a significant motion in the load, and it is usually the lowest frequency or fundamental mode. Although essential for sensing, the addition of a load mass might lower the resonator's Q, and degrade its performance as a sensor. It is therefore of interest to understand how the Q of a resonator mode is affected by mass loading.
The Q of a resonator mode is defined as Q = 2π W ∆W , where W is the energy stored in the mode, and ∆W is the energy loss per oscillatory cycle [21]. In general, ∆W is a sum of the contributions from multiple individual loss mechanisms [25]. One way to improve the Q of a resonator is by increasing the stored energy without a similar increase in the dissipation. This can be achieved by using a tensioned or high-stress film as the resonator and it is the result of a phenomenon called dissipation dilution [26][27][28]. In addition to increased Q, high-stress also extends the mode spectrum of a resonator to higher frequencies for a given resonator length scale. High frequency is of interest for certain sensing protocols, while being out of reach for similar-dimension non-tensioned sensor geometries such as cantilevers.
Low optical absorption combined with high-stress obtained in fabrication, make silicon nitride (SiN) a natural material for the design of high Q resonators [29][30][31]. In tensioned SiN resonators, typically, the Q of modes are limited by two loss mechanisms -bending loss, which is the energy lost due to the bending of the mode, and radiation loss, which is the energy lost from the mode via acoustic radiation into the surroundings. This allows us to define individual quality factors, Q bend and Q rad , which are associated with bending loss and radiation loss respectively. Therefore, we can write that 1 Q = 1 Q bend + 1 Q rad . Q bend can be significantly improved by tailoring the mode-shape to reduce bending loss at the mode edge, commonly referred to as clamping loss [27,28,[32][33][34][35][36][37]. Q rad can be improved by resonator patterning, substrate engineering, and strategic device mounting [27,32,[37][38][39][40][41][42][43][44]. These techniques have led to resonators with modes limited by Q bend .
In this work, we study the effect of a localized mass load on the Q bend of highly tensioned resonators. We show through analytical calculations and finite element analysis (FEA) simulations that as the load mass increases, the modes of the resonator change in frequency and displacement shape. Further, we show that for a large enough mass, each mode displacement function becomes independent of the mass, which we claim leads to mass-independent Q bend . We refer to this phenomenon as Q mass saturation. We validate this saturation experimentally by measuring the Q of a tensioned SiN trampoline resonator [34,36] as a function of the load mass. We use magnetic grains for the load mass, and we vary the mass by sequentially stacking the grains using their arXiv:2209.01488v1 [cond-mat.mes-hall] 3 Sep 2022 mutual magnetic attraction to avoid varying the amount of lossy adhesive as mass was added. This allows us to compare Q measurements for different load masses on a single device.
In order to obtain an expression for the Q bend of a mode, we assume pure out-of-plane mode displacement u (x, y), where x and y are in-plane resonator coordinates, as well as small displacement, |u (x, y)| h, where h is the resonator thickness. In addition, we assume large inplane tensile stress σ 0 , such that the speed of sound along the resonator is approximately proportional to √ σ 0 .
Then we can write W ≈ V , where E 2 is the imaginary part of the Young modulus, ν is the Poisson ratio, z is the resonator coordinate along its thickness and V is the volume of the resonator [27,45]. This leads to: Eq. 1 affirms that if the mode displacement function u(x, y) stops changing, then Q bend saturates. We can elucidate this Q saturation and its origin through an example model. We study the fundamental mode of a highly pre-tensioned suspended beam, which is referred to as a string. (A second example for a tensioned, mass loaded circular membrane can be found in the supplementary material.) The string resonator is fixed at both ends, with displacement u (x) in the vertical axis, where x is the coordinate along the resonator (Fig. 1a).
Here u (x) satisfies the string equation [25]: for 0 ≤ x ≤ L, where L is the string length. Here, E, h, σ 0 , are the string's Young modulus, thickness and the tensile stress respectively, and f is the mode frequency. The string width w is absent from the equation because the string has a uniform width. The position-dependent density ρ (x) accounts for possible increased density at some region a ≤ x ≤ b, for 0 < a < b < L. We denote the density at 0 ≤ x < a and b < x ≤ L, subsequently referred to as the outer regions, by ρ m , and the density at a ≤ x ≤ b, the inner region, by ρ m + ρ M , such that ρ M is the added load density. The pre-stress σ 0 is assumed to be large, satisfying the length scale inequality l = 2π Eh 2 12σ0 L.
Given this high pre-stress condition the frequency of the string's fundamental mode can be well approximated by neglecting the fourth derivative term in Eq. 2 [25]. The solution then reads: where we introduced the density ratio χ = ρ M ρm and the inner region wavenumber α = 2πf ρm+ρ M σ0 , and imposed the fixed boundary conditions u (0) = u (L) = 0. An equation for α can be found by using the continuity of u (x) and its first derivative at x = a and x = b (supplementary material). We see excellent agreement between FEA simulations of α and f for the fundamental mode of a string, and the corresponding numericallyevaluated analytical calculation, for different values of R the resonator mass ratio (Fig. 1c). We define the resonator mass ratio as R = M load M unloaded using the quantities M load = hw (b − a) (ρ M ) and M unloaded = hwLρ m corresponding to the load mass and the total unloaded resonator mass, respectively. This parameter was chosen to emphasize that we are investigating the crossover between no load (R = 0) and a load that is significantly more massive than the entire resonator mass (R 1). In addition to α and f , we used the results from the FEA simulations for the fundamental mode along with Eq. 1 to calculate Q bend . Fig. 1d shows the dimensionless geometric parameter Q bend /Q 0 , where Q 0 is the Q bend without a load mass, as a function of R. While calculating Q bend , in order to focus specifically on the effect of the load mass, we chose to exclude the mechanical clamping loss contribution from the edges of the mode. From the plot, we see that Q bend /Q 0 becomes independent of the exact load mass for R 1 and it saturates to a value that is lower than when R is close to zero. (Note the plot points corresponding to R = 0 have been omitted from Fig. 1c and Fig. 1d because the horizontal axis is on a logarithmic scale.) Fig. 1b evinces that the fundamental mode inner region wavenumber α saturates to a finite value, α lim , as the mass grows larger (supplementary material). Simultaneously, the outer region wavenumber diminishes, and in the large mass limit the mass loaded string equation can be approximated as where Because this equation is independent of the load mass, so is its solution (supplementary material).
In the high mass load limit, the mode shape of the unloaded regions can be understood in two equivalent ways. The first is temporal: for an unloaded resonator, the fundamental mode period is given by ρm σ0 2L, which can be interpreted as the time it takes an out-of-plane mechanical perturbation to make a roundtrip across the length L of the resonator, with speed of sound σ0 ρm . As the density of the loaded region increases to the high mass load limit (ρ m → ρ m + ρ M ) the period of the fundamental mode increases, meaning T loaded is now much longer than the roundtrip time a mechanical perturbation propagates in the unloaded region. As a result, the mode shape in the unloaded regime approximates a quasi-static displacement, with the mode shape limit being a static displacement at any moment. The second is spatial: the mode wavelength T loaded σ0 ρm is much longer than the unloaded region length scale. As a result, the mode shape converges to a displacement function that doesn't curve in the unloaded region. For a string, the resulting mode shape limit is linear at the outer region, and since the inner region has to match with the outer regions at the boundaries, it implies that both the outer and inner regions approach a limit shape, which approximates a triangular shape. (Fig. 1b). This leads to the splitting of the string equation as is described in Eq. 4 and Eq. 5, and as explained, its solution is independent of the load mass which leads to Q saturation.
Although our analysis thus far has focused on the case of the mass loaded string, the key points of the explanation above apply more generally to highly tensioned mass loaded resonators of arbitrary geometry and imply Q saturation for a high enough mass load. Specifically, as an example we have also carried out an analysis for a circular membrane (supplementary material).
To validate the model-predicted Q saturation effect experimentally, we measured the fundamental mode Q, denoted hereafter as Q fund , of a high-stress SiN trampoline resonator for different load masses [34,36]. A trampoline resonator features a wide pad for low-imprecision optical detection and the geometry is designed to reduce mechanical clamping losses, allowing access to high Q modes [34,36]. The device we used was microfabricated with 110 nm thick SiN with a side length of 1 mm, a tether width of 5 µm, and a pad area of approximately 86 µm 2 (Fig. 2b). Stoichiometric SiN with a film prepatterning tensile stress of roughly 1 GPa was used and an unloaded fundamental mode frequency of ≈ 143 kHz was measured. The resonator was suspended from a 385 µm thick square silicon chip with a side length of 4 mm (supplementary material).
In order to vary the load mass while keeping the other variables constant, we employed a "magnetic stacking" technique. We used magnetic particles (Neo Magnequech MQFP-B+ (D50=25m)) as the load mass. The first par-ticle was affixed to our device using ultra-violet (UV) cured epoxy (NEA 123SHGA) and is shown in Fig. 2a. We then magnetized the first particle by placing the device inside a strong magnetic field. Next, a second particle was brought near the first, causing the first particle's magnetic field to induce a magnetic moment in the second particle, resulting in magnetic attraction between the particles. The device was then put under a strong magnetic field once again, this time to magnetize the second particle, and effectively make both particles a single, inseparable mass load without using additional epoxy. This procedure was repeated for multiple magnetic particles (Fig. 2c) and Q fund was measured between each addition of mass. Although we expect the epoxy to have high mechanical loss compared to SiN, by keeping its volume and three dimensional orientation constant through magnetic stacking, we were able to study how Q fund is affected by mass loading in a systematic fashion, avoiding errors arising from multiple epoxy applications on separate devices.
We note that the Q fund of trampoline resonators is sensitive to mounting due to radiation loss [34,36]. To minimize changes in Q fund due to varying radiation loss, we kept the mounting the same between Q fund measurements. This was achieved by affixing our devices at one corner onto a custom-made silicon base by using silver epoxy (EPO-TEK H20E). We found that this mounting scheme consistently allowed for Q fund ≈ 33 × 10 6 without a mass load. This matches the bending loss limited Q fund obtained from FEA simulations (supplementary material), thus implying Q rad Q bend , and therefore, Q fund ≈ Q bend for trampoline resonators without mass loading when mounted in this way.
The key results of our experiment are shown in Fig. 3a along with relevant FEA simulations. The open black circle and solid black circles correspond to the experimentally measured Q fund values for a trampoline resonator at 300 K, without a load mass and with a variable load mass, respectively. They depict how Q fund decreases as the load mass increases, but becomes distinctively insensitive at higher load mass values. This is manifested as a plateau for masses larger than ∼ 10 ng and is indicative of Q saturation. The plateau value amounts to approximately four orders of magnitude reduction in Q fund compared to the unloaded case. To model these results, we used FEA simulations at 300 K which took into account bending loss in both the SiN and the epoxy, while disregarding radiation loss (black dotted line). In the simulations, we scanned the epoxy length scale and loss tangent, and we chose realistic model parameters to align with our data while also setting an upper bound on Q fund (supplementary material). We see reasonable agreement between our measured data and simulations. However, the simulated model does not fit all the data points to within the measurement noise, which we believe is because, irrespective of device mounting, at higher mass, and therefore lower frequency, the resonator is more susceptible to residual radiation loss.
In order to further validate that the large reduction in Q fund originates mainly from loss in the epoxy, rather than residual radiation loss which manifests when the load mass is large, we performed additional measurements at temperatures of 300 K and also 8 K (light pink and sky blue markers respectively), on two different devices (triangle and square). Each device was measured at both temperatures. The results show that the Q fund rises by about two orders of magnitude as the resonator is cooled to cryogenic temperatures. Although SiN resonators are expected to have reduced bending loss at 8 K compared to 300 K, empirically it is by a small factor of ∼ 3, which cannot explain the measured difference in Q fund [32]. We therefore conclude that when cold, the epoxy's bending loss reduces significantly, which affirms that the reduction in Q fund seen is primarily due to the lossy epoxy.
A fourth device (sky blue diamond) measured at 8 K provides evidence that a fairly high quality factor Q fund ≈ 3.7 × 10 5 can be achieved far in the saturated regime. Previous studies have shown that using better epoxy can lead to higher quality factors [46].
In order to set a theoretical limit on the saturated Q fund at 300 K, we performed FEA simulations which only consider bending loss in the SiN and disregard all other forms of loss (green dotted line). The results indicate it is possible to obtain a Q fund decrease of only a factor of ∼ 1.5 (Q as high as 10 7 ) in the limit of a large load mass. The saturated Q fund for trampolines can be further optimized by changing the location of local mass loading (supplementary material), and possibly by using trampoline resonators with carefully engineered geometries [34,36,47].
In conclusion, we have analyzed the problem of tensioned mechanical resonators with a local mass load. We have shown that the added mass changes the shape of the resonator fundamental mode, and therefore, the Q fund of the resonator. Further, we have demonstrated that for a large load mass, the mode shape becomes independent of the load mass and converges to a limit shape. This implies that Q fund saturates in the limit of a large mass load. We have validated our model by measuring the Q fund of a SiN trampoline resonator for different load masses. We believe that this work provides important guiding principles for the design of sensors where mass loading is needed, such as in magnetic field sensors [11][12][13], accelerometers [16,17], mass balances [18][19][20], and gravity detectors [22][23][24]48]. Further, the ideas presented could be extended to other high Q tensioned two dimensional devices such as phononic crystals [27,32,38], hierarchical structures [35,49], and machine-learning optimized resonators [47,50]. | 2022-09-07T01:15:55.548Z | 2022-09-03T00:00:00.000 | {
"year": 2022,
"sha1": "97eabacb4ac2a5c73655b4e4dd2e09a718859ad3",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "97eabacb4ac2a5c73655b4e4dd2e09a718859ad3",
"s2fieldsofstudy": [
"Physics",
"Engineering"
],
"extfieldsofstudy": [
"Physics"
]
} |
219676168 | pes2o/s2orc | v3-fos-license | Late Presentation of Congenital Diaphragmatic Hernia with Malrotation of Midgut in Adult
Case Report 18 years old female admitted with complaints of abdominal pain, breathlessness and constipation for 2 days. On examination patient had tachycardia, tachypnea, abdomen symmetrically distended with tenderness on left side. Heart sounds S1 and S2 was heard on right side of the chest, decreased air entry on left side. X ray and CT chest showed diaphragmatic hernia with bowel obstruction. Patient was taken for emergency procedure content reduced and defect was closed, respiratory support given. Clinical and radiological investigations helps in diagnosing the congenital diaphragmatic hernia early and intervention before respiratory failure increases the survival rate in both neonates and adults.
INTRODUCTION
The aim of this case report is to provide information on the clinical symptoms and signs, diagnosis and management of the late presentation of congenital diaphragmatic hernia in adult.Congenital diaphragmatic hernia is a major malformation occasionally found in newborns and babies. Congenital diaphragmatic hernia is defined by the presence of an orifice in the diaphragm, more often to the left and posterolateral, which permits the herniation of abdominal contents into the thorax. In our case the patient presented with small bowel obstruction due to diaphragmatic hernia Bochdalek type with malrotation of midgut.
CASE REPORT
The subject is an 18 years old female admitted with complaints of abdominal pain and breathlessness for 2 days with history of constipation and has not passed flatus for 2 days. Patient was first born child and neonatal period was uneventful with no previous significant medical and surgical history. On general examination patient is conscious, afebrile. Tachycardia and tachypnea present, abdomen symmetrically distended with tenderness in left hypochondrium, lumbar and iliac fossa region. On Cardiovascular system examination S1 and S2 heard on right side of the chest, Respiratory system examination revealed decreased air entry and bowel sounds heard on left side. X ray chest and abdomen showed bowel shadows in left chest, on CT chest showed mediastinal shift to right side, collapse of left lung, bowel loop in left chest. CT abdomen revealed Ileum, transverse colon proximal descending colon herniating into the thorax. Jejunum, Transverse & ascending colon appeared dilated, Distal descending colon is collapsed and recto sigmoid on right side. Due to small bowel obstruction and respiratory distresspatient was taken for emergency procedure through left subcostal incision peritoneal cavity entered -jejunum was dilated, ileum, cecum with appendix, transverse colon was found herniating into the thorax, content was reduced and defect of size 6.5*4 cm noted in posterolateral surface of diaphragm which was closed by using non absorbable suture material prolene in two layers. Intercostal drainage tube was inserted on the left side. Patient was on mechanical ventilation for 4 days after that she was weaned gradually and oral feeds started.
DISCUSSION
Incidence is 1 in 2000 to 5000 live births and common in left side around 80% and bilateral in < 5% of the cases. Hernia sac is present in 20% and overall survival rate 70 to 90 %. Incidence in adult is 0.1 to 6 % and most cases are sporadic.
Pathophysiology
CDH is due to failure of closure of pleuroperitoneal canal in developing fetus. Usually the pleuroperitonealcavity become separated by the developing membrane during 8 to 10 weeks of gestation. When this process fails the pleuoperitoneal canal does not close and a posterolateral defect in diaphragm results.It impairs growth of ipsilateral lung and causes pulmonary hypertension
Clinical features
Respiratory distress at birth (tachypnea, grunting, use of accessory musclescardinal sign), Dyspnea, Cyanosis, Scaphoidabdomen, Increased chest wall diameter, Bowel sound heard in chest, Decreased/absent breath sound on side of hernia, Shifting of apex beat to the opposite side
Surgical treatment
Mostly surgery was done 48 hrs of birth after stabilization. Preferred approach is sub costal abdominal incision both laparoscopic and thoracoscopic repair can be done. Content reduced and defect closed by interrupted non absorbable sutures, if present the hernia sac should be excised. If primary closure is not possible -abdominal /thoracic muscle flaps can be used. For tension free repair native tissue / Prosthetic material Synthetic patch (GORE TEX) and biological mesh are used.
Prognosis
Overall survival 67% in newborns and poor prognosis in major anomalies, severe pulmonary hypoplasia -need of ECMO. In adults prognosis is good if surgery is done before developing respiratory complications.
CONCLUSION
Congenital diaphragmatic hernias are usually diagnosed in neonatal period it may present as late in adults with incidence of 0.17 to 6%. Clinical and radiological investigations help in diagnosing the condition, early intervention before respiratory failure increases the survival rate in both neonates and adults. In our case early diagnosis and early intervention saved the patient before developing complications. | 2020-05-21T00:12:57.887Z | 2020-02-20T00:00:00.000 | {
"year": 2020,
"sha1": "f4cc7458a05e87df7250841430c9a4d0108c6e68",
"oa_license": null,
"oa_url": "https://doi.org/10.36347/sasjs.2020.v06i02.003",
"oa_status": "GOLD",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "104ffaf90fbc2354eeb68a432785b3647338a4dc",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
221950293 | pes2o/s2orc | v3-fos-license | Controlling for presentation effects in choice
Experimenters make theoretically irrelevant decisions concerning user interfaces and ordering or labeling of options. Reanalyzing dictator games, I first show that such decisions may drastically affect comparative statics and cause results to appear contradictory across experiments. This obstructs model testing, preference analyses, and policy predictions. I then propose a simple model of choice incorporating both presentation effects and stochastic errors, and test the model by reanalyzing the dictator game experiments. Controlling for presentation effects, preference estimates become consistent across experiments and predictive outof-sample. This highlights both the necessity and the possibility to control for presentation in economic experiments.
Introduction
Many economic experiments analyze individual behavior to understand preferences, as understanding preferences is required for applied theoretical analyses and quantitative policy recommendations. Preferences appear to be contradictory across studies, however, which mitigates the reliability of counterfactual predictions and policy recommendations. Such contradictory results appear to be particularly prevalent in studies of social preferences. As I show here, a major reason for inconsistency between experiments appears to be that "presentation" of options affects choice. Presentation matters due to default effects (McKenzie, Liersch, and Finkelstein (2006), Dinner, Johnson,Goldstein, and Liu (2011)), left-digit effects (Poltrock and Schwartz (1984), Lacetera, Pope, and Sydnor (2012)), round-number effects (Heitjan and Rubin (1991), Manski and Molinari (2010)), and positioning effects (Dean (1980), Krosnick (1998), Feenberg, Ganguli, Gaule, andGruber (2017)). Absent a model of such presentation effects that would allow researchers to control for them, they induce biased and inconsistent utility estimates. This implies incompatibility of observations across experiments and may also be a major reason for the failure to reach a consensus on (social) preference theories.
To fix ideas, consider an experimenter designing a dictator game experiment to test a model of social preferences. In a dictator game, player 1 ("dictator") chooses how many tokens, x ∈ {0 B}, to transfer to player 2 ("recipient"). Whatever model the experimenter seeks to test, some theoretically irrelevant decisions have to be made to run the experiment. I refer to them as presentation decisions. For example, the value of the cake to be redistributed may be fixed and the experimenter has to set the total number of "slices" B to run the experiment. By all received theories, the choice of B is theoretically irrelevant in that the budget share transferred by player 1 is independent of B (aside from discreteness). Yet, if B = 20, then the equal split results from a transfer of 10 tokens, and if say B = 25, then the equal split is not attainable at all, let alone by choosing a round number.
In actual experiments, this seemingly trivial decision, cutting the cake into either 20 or 25 slices, has drastic behavioral implications. As this effect seems to have escaped attention so far, Figures 1 and 2 illustrate this in some detail. Table 1 below reviews the experiments I am reanalyzing here. A detailed discussion follows, but briefly, I focus on dictator game experiments that are behaviorally equivalent in the sense that if behavior depends solely on outcome-based preferences, then the observed choice patterns should be statistically indistinguishable. In each dictator game, each of the B tokens to be allocated by the dictator is worth ρ 1 monetary units to the dictator and ρ 2 monetary units to the recipient. I refer to ρ 2 /ρ 1 as the transfer ratio, where ρ 2 /ρ 1 > 1 indicates subsidized transfers and ρ 2 /ρ 1 < 1 indicates taxed transfers. Such dictator games are widely used to study social preferences, but specific experimental designs differ in theoretically irrelevant ways.
First, Andreoni and Miller (2002, AM02) allow for budgets up to B = 100 where the equal split is generally a multiple of 10 or 25, and choices mainly are multiples of 10 ( Figure 1(a)). Similarly, Cappelen, Hole, Sørensen, and Tungodden (2007, CHST07) allow for budgets up to B = 1600 with round-numbered equal splits, and subjects primarily choose multiples of 100 (Figure 1(b)). Second, Harrison and Johnson (2006, HJ06) allow for budgets up to B = 100, but the equal split is often a plain integer or not attainable at all; as a result, it is chosen less often and choices overall are more dispersed (Figure 1(c)). Third, Fisman, Kariv, and Markovits (2007, FKM07) implemented a graphical user interface preventing round-number effects (Figure 1(d)). There, choices vary continuously and the equal split is not chosen more frequently than other options. The choice of either of these three types of experimental designs thus affects the relative frequency of the equal split, the dispersion of choices-and as Figure 2 shows, even the comparative statics. Figure 2 displays the share of tokens transferred as a function of the transfer ratio (tax or subsidy rate) for the three experiments actually varying ρ 1 and ρ 2 . In HJ06, transfers are generally high and independent of ρ 2 /ρ 1 , in AM02 transfers start low but are increasing in ρ 2 /ρ 1 , and in FKM07 transfers are low and again independent of ρ 2 /ρ 1 .
To summarize, even in theoretically equivalent experiments, choice patterns differ widely and elicited comparative statics end up being contradictory-all of which results in utility estimates that are inconsistent across studies (shown below). The implications are substantial. Since comparative statics depend on presentation, experimental studies Figure 1. Choice patterns across dictator game experiments. Note: In AM02, the equal split is a round number (transferring 20 tokens), in CHST07 as well (300 tokens), in HJ06 it is not an integer and thus unavailable (transferring 33.3 tokens), and FKM07 use a graphical user interface. As above, B indicates the number of tokens available, and "transfer" (e.g., "1 : 2") indicates the inverse transfer ratio ρ 1 : ρ 2 as defined in the text. cannot measure comparative statics (without controlling for presentation) and experimental results cannot be taken at face value. Since utilities are inconsistently estimated, social preference models are not predictive, evaluating preference models by comparing results from different experiments is futile, and convergence of social preference theory is put out of reach. Such concerns regularly surface in critiques of experimental and behavioral economics, and to address them, we need to control for presentation effects.
The present paper proposes a simple model of presentation effects, applies it to standard data sets (including those in Figure 1), and shows that it allows to effectively factor out and control for presentation effects in analyses. These results are fairly positive in nature. While behavior and social, time, and risk preferences appear to differ a lot across experiments, this paper provides both an analytical framework and econometric evi- Figure 2. Comparative statics of transfer in tax/subsidy rate across experiments. Note: The plot captures the budget share transferred as a function of the transfer ratio in AM02, FKM07, and HJ06. The transfer ratio ρ 2 /ρ 1 ranges from 1/4 (highly taxed transfers) to 4/1 (highly subsidized transfers). dence suggesting that neither the preference theories nor experimental measurement as such are necessarily inadequate. Instead, current measures are being confounded by presentation effects, which we need to filter out, and by doing so, we can get much further in analyzing experimental studies than we previously thought.
Section 2 introduces the focal choice adjusted logit (Focal) model. Section 3 presents the framework for the econometric test of the model. Section 4 presents the results and Section 5 concludes. The Appendix contains relegated proofs and definitions, the Online Supplemental Material is available in the replication file (Breitmoser (2021)) and contains results on a number of robustness checks.
The model
The purpose of the model proposed here is to provide a framework for the estimation of utility parameters given an experimenter's hypothesis about the utility function. To obtain data, the experimenter runs an experiment where each decision maker DM chooses an option x ∈ B from a finite budget set B ⊂ X. Each potential option x ∈ X is associated with a list of attributes, and according to the experimenter's model, a subset of those attributes affect DM's utility. The remaining attributes do not affect DM's utility (theoretically) but may otherwise affect choice. In the existing literature, such utility-irrelevant yet choice-relevant attributes have been associated with temptation (Gul and Pesendorfer (2001)), salience (Bordalo, Gennaioli, and Shleifer (2012)) and focality (Schelling (1960)) of options, as discussed below.
Considering my focus on presentation effects, I will say that the utility-irrelevant attributes affect the focality of options, which refers to the "focal points" discussed by Schelling (1960) that reflect the "prominence or conspicuousness" (p. 57) of options. To be clear, Schelling discussed focal points in coordination games, where actions may gain prominence or conspicuousness from common pieces of knowledge or cultural background, but since presentation equally affects prominence or conspicuousness of options, it seems appropriate to borrow the term for this paper.
The set of possible budgets in the experiment is the set of finite subsets of X, denoted as P(X). Further, π : X → R m denotes the mapping from options to values of theoretically utility-relevant attributes and φ : X → R n denotes the mapping from options to values of other attributes such as ordering, positioning, labeling, and roundedness, which typically are considered utility irrelevant. 1 I denote the respective attribute values of option x as π x and φ x . Example 1. DM chooses an option x ∈ {1 2 3}, the payoff of which depends on a coin toss (with unknown probabilities). Option 1 yields payoffs of 10 and 0 in the cases of heads and tails, respectively, option 2 yields 4 and 4, and option 3 yields 0 and 10. The options are presented in a column, with 1 being the top option (arbitrarily labeled "1") and 3 being at the bottom (arbitrarily labeled "0"). Thus: The images of π and φ are denoted π[X] and φ[X], respectively. We assume that DM has complete, transitive, and reflexive orderings over both sets of attribute values, π[X] and φ[X], intuitively interpreted as "preferred to" and "more focal than." The orderings are represented by a utility index u : π[X] → R and a focality index v : φ[X] → R, respectively.
Our task will be to infer utilities u[π [X]] and focalities v[φ[X]] of options from DM's choice probabilities. In this task, the numeric labels attached to "Top," "Middle," and "Bottom" in the example are indeed arbitrary.
The pair of attribute mappings (π φ) is called context of DM's choice and may vary between decision tasks, for example, by experimenters changing outcomes π or permuting the order φ. The set of possible contexts (π φ) is denoted as C. The probability that DM chooses x ∈ B from budget B ∈ P(X) in context (π φ) ∈ C is denoted as Pr(x|π φ B).
As discussed in detail next, DM seeks to maximize u + v subject to random perturbations of both utilities and focalities. This yields the following model. Definition 1. The choice profile Pr has a focal choice adjusted logit (Focal) representation if there exist λ κ ∈ R, u : π[X] → R, and v : φ[X] → R such that for all contexts 1 Presentation is obviously more complex than just comprising ordering, positioning, labeling, and roundedness of numbers. Take, for example, the way questions are asked in surveys. For my model to be applicable, we would have to be able to categorize (or quantify) the way questions are asked, so we can define a function mapping each category to focalities of different options. To the extent that such categories can be defined objectively, the model seems applicable, but I leave this for future research.
(π φ) ∈ C and all budgets B ∈ P(X), If C is a singleton set, that is, comprises just a single context, then Pr has a Focal representation if and only if it has a logit representation. The indices u and v may therefore be identified only if we have observations from multiple contexts. Next, I briefly review related literature and then discuss identification. The Appendix contains a simple axiomatic characterization of the model.
Related literature
The idea to distinguish utility-relevant and utility-irrelevant (but still choice-relevant) attributes of options when modeling choice has a long tradition in behavioral economics, presumably starting with Schelling (1960) who discussed focality of options as a choice relevant input in coordination games. Schelling's ideas have been picked up frequently again (for a discussion see Kreps (1990)), though with little attention to formalizing the notion of focality (but see Sugden (1995)). The idea of focal points, that is, prominence or conspicuousness of options, naturally extends to individual decision making, as evidenced by the well-documented effects due to ordering, labeling, or positioning of options or roundedness of numbers. The above approach, to model the interaction of utility and focality based on two indices u and v, that affect decision making in an additively separable manner (u + v) follows three strings of results in the literature on individual decision making.
First, additive separability of two such indices was proposed in Gul and Pesendorfer (2001) based on an axiomatic characterization of a preference for commitment. Formally, GP01 model a decision maker who, in a first stage, selects the menu of options to choose from, and in a second stage, an option from this menu. GP01's preference for commitment is a weak preference of DM in stage 1 for removing certain options from her eventual menu in stage 2. GP01 have phenomena typically related to time inconsistency in mind, for example, removing the option to have a burger at tomorrow's lunch, but the model itself is general and speaks to any phenomenon related to decision making that involves option attributes that are choice relevant but not utility relevant. To see this, note that if there exist option attributes that do not affect utility but are relevant to choice, then GP01's preference for commitment naturally arises. Such attributes may include attributes affecting Gul and Pesendorfer's notion of temptation, but critically for our discussion, they may also include attributes affecting (Schelling's) focality of options-formally, temptation as in GP01 and focality as in Schelling equally imply a preference for commitment. The key result of Gul and Pesendorfer is that, under seemingly standard assumptions, the preference for commitment implies that DM is influenced by two indices u and v, seeking to maximize u + v when making the decision, as we assumed above-with the only difference that we call v "focality" index rather than "temptation" index.
Here, u represents the binary preference relation over singleton sets of options (hence, "utility" of options) and is revealed by the choice of a menu in the first stage.
Further, u + v represents the preferences over options for any given menu of options and is revealed by the choice of an option in the second stage. Thus, anticipating that she will be tempted to choose options that do not maximize her utility u (due to the influence of v when making the decision), DM may prefer to eliminate some options from the menu prior to making her actual choice. In typical experimental designs, DM has no control over the menu of options and we get to observe DM only in Gul and Pesendorfer's second stage, where she seeks to maximize u + v. Not observing the menu preferences complicates identification, but in general, even knowing both u and u + v (up to ordinal transformation) is not sufficient to infer v up to ordinal transformation (Gul and Pesendorfer (2006)).
As a second approach, Matejka and McKay (2015) modeled a DM with rational inattention and show that with attention costs following Sims (2003), choice probabilities obey a generalized logit form, for some λ ∈ R, some utility index u and a "prior index" α : X → R. That is, based on a conceptually independent analysis of stochastic choice, Matejka and McKay also obtain the additive separability earlier derived by GP01. In their model, α captures the choice propensities if DM has no knowledge of utility-relevant attributes, and thus, α again captures utility-irrelevant but choice-relevant attributes in choice, potentially related to temptation, salience, or focality of options, for example. Their results suggest that we may identify u and α (or, v) if we can present the decision problem without revealing utility relevant attributes, as the elicited choice probabilities then reveal α. This approach is feasible if the utility-relevant option attributes are known and have an empty intersection with the attributes affecting α, but may in general be hard to implement. Third, in an analysis of the axiomatic foundation of conditional logit, Breitmoser (2020) observed that the same generalized logit representation, including the additive separability of utilities and "prior propensities," obtains if choice probabilities obey positivity, independence of irrelevant alternatives, and invariance of choice probabilities to translation of utilities. Based on this observation, an axiomatic foundation of the Focal model is straightforward, as demonstrated in the Appendix. As in Matejka and McKay, the prior propensities α are shown to be independent of utilities, that is, α remains unchanged if we permute or transform utilities in any way. Since utilities as such are not observable, such permutations are not straightforward, but in any model-based analysis, there are theoretically utility-relevant attributes and we may be able to permute them without affecting focality-relevant attributes. This would enable identification, as demonstrated next.
Identification As indicated, I envision an analyst seeking to estimate the parameters of a specific utility function controlling for specific presentation effects due to ordering, positioning or round numbers. I illustrate a prototypical example of such an application below by re-analyzing a set of dictator games that form the backbone of many studies of social preferences and exhibit the widely documented round-number effects. In these cases, the attributes affecting utility are specified prior to the analysis, but secondary attributes such as ordering of options (Example 1) need to be set to run the experiment, and their impact is supposed to be factored out. Then it will be possible to change the secondary attributes (say, by reordering options) without affecting the theoretically utility-relevant attributes, and equally, circular permutations of utility-relevant attributes are possible.
Definition 2.π is a circular permutation of π with respect to B ∈ P(X) if there exists a function σ : B → B such that: (i)π σ(x) = π x for all x ∈ B, and (ii) σ k (x) = x for all x ∈ B and all k < |B|.
As indicated, I propose to think of this assumption as clarifying that we are able to reposition, reorder or relabel options without affecting their utilities, which is true by all received models of utility in the behavioral literature. Obviously, such an assumption may become inadequate as our understanding of utility and presentation evolves, and in this sense, the present paper merely makes a first step in formally capturing presentation effects, but as we shall see, with respect to the experiments reanalyzed here, this assumption seems to be adequate. Circular permutations are sufficient for identification, as the following result demonstrates.
Proposition 1. Fix a budget B ∈ P(X) and consider two contexts (π φ) (π φ) ∈ C, assuming DM's choice profiles have Focal representations with the same λ κ ∈ R in both. Then utility index u • π and focality index v • φ over options x ∈ B are identified up to affine transformation ifπ is a circular permutation of π with respect to B.
That is, we can disentangle u • π and v • φ, that is, the mappings from options to utilities and from options to focalities, using appropriate experimental designs. Since π and φ are induced by the experimenter, this can be used for inference about functional forms and parameters of u and v. At this point, I am not aware of experiments actually implementing circular permutations, as opposed to say reversions of lists on which I comment below, and thus, I am not able to illustrate the approach based on existing data. The proposition is therefore a "possibility result," that is, a demonstration that it can be straightforward to disentangle u and v using experiments that vary u and v in a sufficiently informative way. Indeed, the experiments I reanalyze below do not seem to permit circular permutations in an obvious manner, as they require subjects to enter numbers rather than picking items from lists, but they generally provide observations from more than two contexts per subject, which at least for the experiments I am reanalyzing seems to provide sufficiently informative variation of π and φ (based on inspection of the gradients and Hessians of the likelihood function).
Briefly, let me illustrate how we can disentangle u and v using circular permutation of options. Assume that we do not know u and v from Example 1, but that we observe choice probabilities in the context (π φ) defined in Example 1 and in the context (π φ) defined next.
Example 3. Equivalent to Example 1, but utility-relevant attributes are given byπ as follows:π 1 = (0 10) π 2 = (10 0) π 3 = (4 4) Note thatπ in Example 3 is a circular permutation of π as defined in Example 1. By Proposition 1, observing choice probabilities in the two contexts, (π φ) from Example 1 and (π φ) from Example 3, suffices to identify the three utilities u(π 1 ) u(π 2 ) u(π 3 ), and the three focalities v(φ 1 ) v(φ 2 ) v(φ 3 ) up to affine transformation. The basic idea is as follows. Knowing that the choice probabilities have a Focal representation with constant λ κ in both contexts, we can demonstrate that the log-probabilities satisfy for all x ∈ {1 2 3} = B and both contexts, (π φ) and (π φ), with (λ κ) as in the Focal representation and some a ∈ R. In total, we have 2 · |B| = 6 such equations and 2 · |B| + 4 unknowns: the three utilities, the three focalities, λ and κ, and the two constants labeled a above (one per context). By fixing levels and scales of both utilities and focalities, we eliminate four unknowns, implying that the number of observations equates with the number of unknowns. The resulting equation system is regular ifπ is a circular permutation of π. After fixing levels and scales of utilities and focalities, they are thus identified up to affine transformation. Circular permutations are feasible in many cases. For example, many experimental analyses assume that utilities are functions of payoff profiles over states of the world or over players, and they allow subjects to choose one of n options. These options are arranged arbitrarily on a screen, or are associated arbitrarily with labels, and the analyst is worried that positioning or labeling of options biases choice. Any such arrangement permits a circular permutation, where say option 1 assumes the position (or, label) of option 2, option 2 that of 3, . . . , and option n assumes that of option 1. If the analyst observes choice probabilities for two such arrangements, where one is a circular permutation of the other, then identification up to affine transformation is possible.
There are two related points worth noting. On one hand, not all permutations of π yield a regular equation system. Most prominently, experimental studies involving lists of options tend to revert the list in order to nullify ordering effects by pooling observations from the original list and the reverse list. I am not aware of a possible model of choice justifying this approach, and as an example, assume both first and last option have high focality and all other options have low focality-then reversion has no effect on choice probabilities and neither utility nor focality is identified. Using the Focal model, one may define an equation system as above, but the resulting equation system is in general singular. In this case, identification of utilities using the Focal model can be achieved only if the ordering effect (i.e., v) is linear in the position of the option (φ), while circular permutation works in general.
On the other hand, we do not require assumptions about the functional form relating utility-relevant attributes and utilities of options in order to infer utilities of options, or about the functional form relating focality-relevant attributes and focalities in order (2007)) 50 500-1000 76 × 50 4 : 1,. . . , 1 : 3 Note: For each of the experiments reanalyzed here, the table lists the number of decisions per subject (#Treatments), the number of options per choice task (#Options), the number of observations overall (#Observations, which is the number of subjects times the number of treatments), and the range of transfer ratios implemented across treatments in the experiment.
to infer focalities of options. That is, circular permutation permits nonparametric estimation of cardinal utilities controlling for stochastic choice and presentation effects if we can observe choice probabilities. An analyst interested in estimating parameters may still do so, of course, but it is not necessary to estimate cardinal utilities.
Framework for model test and application
One approach to test the Focal model would be to test underlying axioms (see the Appendix) in isolation, that is, in specifically designed choice tasks. Such a test would be of limited informativeness here, as violations of axioms appear to be context dependent, suggesting that an applied model such as Focal is tested ideally based on data representative of those it is supposed to be applied to. I therefore test Focal on data sets representative of some of the behavioral literature, and attempt to answer three questions that are particularly relevant in applied work: Does Focal enable more accurate counterfactual predictions than existing models? Does it provide consistent and reliable estimates across studies? Do other models fail in this respect, and is the relevance of presentation effects economically substantial? The general idea, to test behavioral models by evaluating validity in-sample and out-of-sample using typical data sets, follows a literature comprising analyses of decision under risk (Harless and Camerer (1994), Wilcox (2008), Hey, Lotito, and Maffioletti (2010)), learning (Camerer and Ho (1999)), strategic choice in normal-form games (Camerer, Ho, and Chong (2004)), and social preferences (De Bruyn and Bolton (2008)).
The selected data sets are from the four well-known experiments on generalized dictator games discussed in the Introduction, reviewed in Table 1.
Dictator games enable an analysis of utility and focality in a context where utility and focality individually appear to be understood reasonably well and additional factors can be ruled out thanks to the simplicity of dictator games.
On one hand, most analyses of dictator games that estimate utility parameters, including three of the four studies reanalyzed here, assume that preferences have a CES utility representation. 2 In this sense, assuming CES utilities fits my objective of reanalyzing standard data sets under standard assumptions, with the sole novelty being the attempt to control for presentation effects using the Focal model. Definition 4 (Utility). The utility of option x, and given the associated payoff profile On the other hand, there are unambiguous presentation effects (round-number effects), the choice task does not involve confounds due to, for example, risk, probability weighting, or (strategic) uncertainty, and there exist four well-known experimental analyses designed to estimate utility functions and differing only in presentational aspects. Further, there is consensus on the roundedness of numbers from analyses of surveys, which consistently find that 100 50 10 5 1 0 5 0 1 exhibit decreasing levels of roundedness (Battistin, Miniaci, and Weber (2003), Whynes, Philips, and Frew (2005), Covey and Smith (2006)). This eliminates another unknown and allows me to define the focality of option x as the level of the highest number in this sequence that divides x. The focality of multiples of 100 is 4, other multiples of 50 have level 3, and so on, down to multiples of 0 1, which have focality −2 and represent the minimum here. 3 Definition 5 (Focality). Let φ x denote the number that is to be entered in order to Finally, one of these experiments (Fisman, Kariv, and Markovits (2007)) avoids presentation effects due to round numbers by using a graphical user interface, which provides "counterfactual" information when analyzing consistency and reliability. 4 In conjunction, these four large-scale experiments therefore appear particularly suitable for testing Focal, as they provide a fairly comprehensive picture of choice in a context that is otherwise well understood. cial norms. Relatedly, these studies also found that individual choices are frequently consistent with GARP, potentially suggesting that modeling stochastic choice is not needed in analyses of dictator games. All four of the studies reanalyzed here allow for stochastic choice, however, since the choices of the majority of subjects, sometimes the vast majority as in Fisman, Kariv, and Markovits (2007), are not perfectly consistent with any particular family of utility functions and since the standard approaches toward capturing population heterogeneity, such as finite mixture or mixed logit modeling, require a notion of stochastic choice to evaluate log-likelihoods in the first place. Note that the necessity for capturing population heterogeneity does not arise in studies analyzing behavior at the individual level (see, e.g., Halevy, Persitz, and Zrill (2018)).
3 Note that negativity of focality has no particular meaning in the Focal model, as level shifts cancel out. 4 Since the graphical user interface prevents round-number effects, on which I focus here, my analysis will assume that focality is constant at zero for all options in Fisman, Kariv, and Markovits (2007). Further work may investigate if the graphical user interface induces prominence effects other than via round numbers, at for example midpoints, which is possible by evaluating the significance of parameters of corresponding focality functions.
Specification Each subject is characterized by precision λ, altruism α, efficiency concerns β, and focality weight κ. Subjects may be heterogeneous, which we model by allowing for all parameters to be distributed randomly across subjects. Using p = (λ κ α β) to describe the parameter profile of a given subject and f (·|d) to describe its joint density in the population given distribution parameters d, the likelihood that the model d describes the choices o s of subject s ∈ S is As usual, the underlying distributions are variations of normal distributions: altruism α is normal truncated to the interval [−0 5 0 5] to avoid excessive altruism or spite, efficiency β is normal without truncation, precision λ and choice κ are log-normal to avoid negativity. From a wider perspective, this specification closely relates to the widely-used family of mixed-logit models and the distributional assumptions reflect standard practice in analyses of social preferences (Cappelen et al. (2007), Bellemare, Kröger, and van Soest (2008)) and risk preferences (Harrison, List, and Towe (2007), Andersen, Harrison, Lau, and Rutström (2008), Wilcox (2008)). For each of the random variables, both mean and variance are considered free parameters of the model. Overall, the models thus have (up to) eight free parameters, which is conservative in relation to regression models used in experimental analyses and in relation to the more progressive structural models (Harrison, List, and Towe (2007), Bellemare, Kröger, and van Soest (2008)). Regardless, identifiability of the parameters is verified explicitly by analyzing reliability and consistency across experiments. Parameters are estimated by maximum likelihood. Likelihoods are computed by numerical integration using quasi-random numbers (Train (2003)) and maximized in a three-step approach, using first a robust gradient-free approach (NEWUOA, Powell (2006)), second, a Newton-Raphson method to ensure convergence, and finally extensive cross-testing of estimates across data sets and models to ensure that global maxima are found. Models are evaluated using the likelihood ratio test of Schennach and Wilhelm (2016), which is robust to both misspecification and arbitrary nesting of models. I will always indicate the actual p-values, but let me note that a level of 0 005 roughly implements the Bonferroni correction given the simultaneous tests of four models on three data sets (in relation to an original level of 0 05).
Benchmark models Throughout the analysis, I relate the results for Focal to those for three key benchmark models for the lack of comparable earlier studies. An obvious benchmark is "multinomial logit," which is the model used in most current analyses. Accordingly, DM with utility u chooses x with probability In addition, I consider a cross-nested logit model allowing for similarity effects between proximate options (Ordered GEV, Small (1987)) and a model of limited attention following Echenique, Saito, and Tserenjigmid (2018), which provides an alternative explanation for the focus on round numbers.
Similarity Choice violates IIA in the presence of "similarity" effects, and intuitively proximate numbers are more similar than distant numbers. Such similarity effects can be expressed by nested logit (McFadden (1976)) where DM first chooses a "nest" of similar options and secondly makes a final choice from this nest. Small (1987) In the analysis, I use a large bandwidth w equal to half the available options, weights following a standard normal distribution, and free parameters λ to capture precision and κ to capture the extent of the choice bias due to similarity.
Limited attention Round-number effects can be interpreted two ways: subjects either focus on some options (as in Focal) or neglect other options, possibly due to limited attention. Masatlioglu, Nakajima, and Ozbay (2012) generalize revealed preference to account for DMs not considering all their options, Manzini and Mariotti (2014) generalize this idea to stochastic choice, and Echenique, Saito, and Tserenjigmid (2018) generalize the model further by allowing for a weak "perception ordering": first all options at the highest perception level are considered, second the options at the next-highest level, and so on. This Perception Adjusted Luce Model (PALM) straightforwardly applies to focality effects, first the most focal options are considered, next the second layer, and so on, and hence it constitutes a natural benchmark for Focal. Formally, DM with utility u, focality φ, precision λ, and choice bias κ ∈ [0 1], chooses x ∈ B with PALM : where μ(x X) = Logit(x) = exp{λu x }/ x ∈X exp{λu x }. The focality index φ used here will (of course) be equivalent to the one used in Focal. While Echenique, Saito, and Tserenjigmid define and axiomatize PALM only for κ = 1, I allow for the whole spectrum down to κ = 0 (which is logit). Further, I rescale the choice probabilities so they add up to 1, following Manzini and Mariotti's suggestion for cases without "outside options."
Results
For brevity, I pool the results for AM02 and HJ06 due to their similarity, both entailing choice from up to B = 100 options with 8 or 10 observations per subject, reducing the data to three types of experiments: manual choice with many observations per subject ("large manual", AM02 and HJ06), with few observations per subject ("small manual," CHST07), and graphical choice ("graphical," FKM07).
In-sample accuracy First, I analyze descriptive adequacy, that is, in-sample fit. It will be convenient to have the standard formal definition as a reference when discussing other measures of adequacy below. Using d = (d α d β d λ d κ ) as the vector of distribution parameters with, for example, d α = (μ α σ α ) characterizing the distribution of α in the population, the descriptive log-likelihood with respect to data set o is for all data sets o ∈ O := {AM02 HJ06 CHST07 FKM07}, using the log-likelihood function defined in equation (4). Based on this, we define the descriptive Bayes information criterion (BIC, Schwarz (1978)) and the pseudo-R 2 (Nagelkerke (1991)) as usual, where LL max o denotes the log-likelihood of a (hypothetical) "clairvoyant" model predicting the choice distributions as they have been observed, and LL min o denotes the loglikelihood of a naive model predicting uniform randomization in all conditions. The first panel of Table 2 summarizes the results on in-sample accuracy. As indicated, I pool the (similar) "large manual" experiments AM02 and HJ06 in terms of their BIC, reporting the sum BIC descr AM02 + BIC descr HJ06 , based on which the pooled pseudo-R 2 is computed. In these "large manual" experiments (AM02 and HJ06), Focal captures 89% of observed variance in terms of pseudo-R 2 , logit captures around 65% of variance, and PALM and OGEV improve slightly but statistically significantly on logit, explaining around 69% of observed variance. In the "small manual" experiment (CHST07), Focal's adequacy is similar, at 83% in terms of the pseudo-R 2 , but the benchmark models drop to 35% in terms of the pseudo-R 2 . In the graphical experiment, all models explain around 62% of observed variance. The differences in the manual DGs are statistically highly significant, always at the 0 005 level surviving the Bonferroni correction.
Result 1 (Model adequacy). Focal captures manual choice of numbers in-sample, explaining 88% of the observed variance. All models capture graphical choice equally well.
A drop of R 2 by 20 percentage points between Focal and the benchmark models, from 89% to 69% as in AM02 and HJ06, is substantial, and a drop from 83% to 35% as in CHST07 is drastic. Figure 3 below illustrates this by plotting the models' predictions in representative treatments, showing that the benchmark models do not capture choice "reasonably" well. Notably, the benchmark models are equally adequate in the (large) experiments with and without round-number effects (AM02 and HJ06 vs. FKM07), suggesting that they simply fail to capture the round-number clustering. Focal captures round-number clustering, and thus capitalizes on the fact that choice in manual DGs is more predictable. The remainder examines if this actually yields more accurate or reliable utility estimates.
Counterfactual predictions As for counterfactual predictions, there is a vast set of applications ranging from policy recommendations (where we may have little ex ante information) to hypotheses for experiments building on earlier experiments (where we may have much prior information about say precision and choice patterns). In order to account for this diversity, I distinguish predictions to three degrees: there is no information at all (third degree), there is information about the extent of the choice bias κ (second degree), or there is information about both precision and choice bias (first degree). In turn, the first degree thus evaluates the predictiveness of preference estimates in isolation, taking the distributions of λ and κ in the target population as known.
The counterfactuals to be predicted in my analysis are the choices in other experiments. I estimate parameters on a given experiment, predict observations in the other experiments, and rotate such that each experiment is predicted based on estimates from each other experiment. As outlined above, in predictions to the first degree ("Prediction of preferences" in Table 2), the utility parameters are out-of-sample and precision λ as well as choice bias κ are assumed to be known for the sample to be predicted-to measure the reliability of preference estimates in isolation. I implement this measure by taking the utility parameters from the other sample as given and estimating the distributions of λ and κ for the sample to be "predicted." Formally, when "predicting preferences" in data set (
IIA (Logit)
Accuracy in-sample (BIC and corresponding pseudo-R 2 based on LL descr o , see equation (8) Note: Explanation: The table provides results for "large manual" experiments (AM02 and HJ06 in aggregation), the "small manual" experiment (CHST07), and the "graphical" experiment is FKM07. It provides measures for both BIC and pseudo-R 2 (see text). We distinguish fit in-sample (both BIC and pseudo-R 2 ), out-of-sample (reporting pseudo-R 2 ), and consistency (difference of out-of-sample BIC and in-sample BIC). All details are provided in the text. The six columns list the "clairvoyant" LL max o of a hypothetical model (with zero-free parameters) predicting the choice distributions as observed, the minimal LL min o of a hypothetical model predicting uniform randomization in each treatment, and the results for the four models defined in Sections 2 and 3. Relation signs indicate the direction of differences and the p-values evaluated by the robust Schennach and Wilhelm (2016) LR test are provided in parentheses underneath the relation signs. Regarding the consistency measure BIC, the superscripts indicate the p-values of testing the null that the in-sample BIC equals the out-of-sample BIC (or equivalently, that BIC is zero).
In predictions to the second degree ("Prediction of preferences and precision" in Table 2), also the distribution parameters of λ are taken out-of-sample, and in predictions to the third degree ("Prediction of preferences, precision, and choice bias" in Table 2), predictions are fully out-of-sample, Counterfactual predictions to the third degree entail prediction of choice bias κ, that is, predicting the extent of round-number effects, which is not meaningful between manual and graphical experiments. There I focus on predictions between the manual experiments. Using these log-likelihoods, I compute BIC (e.g., BIC pred-1 o and BIC pred-1 o|o ) and pseudo-R 2 as above, noting that the penalty term in BIC is a function of the number of free parameters actually fitted to the data set that is predicted (e.g., zero free parameters in predictions to the third degree). The second panel of Table 2 reports the results. For brevity, I focus on the pseudo-R 2 , that is, observed variance captured by the predictions, which is proportional to the BICs but numerically easier to interpret. The classes of experiments are labeled as above, large manual, small manual, and graphical. Overall, the adequacy of the models for predictions is very similar to their adequacy in capturing behavior in-sample, but the differences between models increase further. Predicting manual choice, Focal maintains 85% accuracy, showing that Focal did not overfit in-sample, while the accuracy of the benchmark models drops to 40%-50%, suggesting they did overfit. Predicting graphical choice, where all models capture 62% of variance in-sample, Focal stays at 58%, while the benchmark models drop to 45%. These differences in predicting graphical choice are perhaps most informative about the quality of preference estimates, as graphical choice does not exhibit round-number effects and all models are equally adequate in-sample. Hence, all differences in predictions stem from inaccurate measurement of preferences in the original data sets.
Result 2. Focal's counterfactual predictions are highly reliable in absolute terms, and much more so than those of the benchmark models. Focal maintains almost in-sample accuracy, showing that the in-sample accuracy was not an artefact of overfitting.
Note that the reliability of counterfactual predictions is robust to limiting the knowledge about the target environment, that is, even precision (degree 2) and choice bias (degree 3) are predicted accurately, and thus represent robust facets of behavior. In applications, neither precision nor choice bias therefore need to be known for the target environment. Figure 3. Impact of controlling for round-number effects in capturing manual choice. Note: These plots illustrate the predictions according to the four choice models, next to the empirical distribution, in representative treatments after fitting the model parameters to all treatments from the respective experiments. The full list of plots is provided as supplementary material (as are the underlying parameter estimates). FKM07 is left out here, as Treatments as such are not defined (budget sets and transfer rates are individually randomized) and the differences between models are very minor in any case.
Consistency of estimates I evaluate the consistency of estimates between experiments in likelihood ratio tests, by comparing likelihoods in-sample and out-of-sample. Consistency is violated if in-sample and out-of-sample differ significantly. Obviously, estimates could be inconsistent simply due to differences in the subject pools, but in this case, all models will detect inconsistency. Note that consistency has been argued to be negatively correlated with in-sample accuracy, since a model that does not fit well in-sample may still be particularly robust due to being "simpler" (Hey, Lotito, and Maffioletti (2010)). This may put Focal at a disadvantage. Again, I distinguish the three degrees. Consistency to the first degree evaluates the consistency of preference estimates in isolation, by relating the in-sample BIC to the out-of-sample BICs obtained from predictions to the first degree, That is, I use preference estimates (distributions of α β) of data set o, and given these parameters I estimate precision and choice bias (distributions of λ κ) on any other data set o , in order to obtain a full set of predictions to the first degree based on preference estimates from o. Next, I take the differences of the resulting out-of-sample BICs to the respective in-sample BIC for each o and aggregate, thus evaluating the consistency of the preference estimates (in terms of log-likelihood differences) obtained from data set o across other data sets. Consistency to the second degree uses out-of-sample preference and precision estimates, that is, the BICs of predictions to the second degree, and thus evaluates the joint consistency of preference and precision estimates across samples. Consistency to the third degree uses the out-of-sample estimates for all behavioral parameters in the prediction stage, Note that for each data set we only have two sets of predictions to the third degree, namely the two sets based on estimates from the other two manual DG experiments, since we are focusing on manual experiments when evaluating predictions fully out-ofsample. The third panel of Table 2 provides the results, listing the differences BIC between in-sample and out-of-sample and using asterisks to indicate significance of differences. In total, Focal violates consistency at the Bonferroni level of 0 005 in 1/8 cases, while logit, PALM and OGEV yield estimates that violate consistency in 7/8 cases each. I also evaluate "relative consistency" by means of the relation signs in Table 2. A model is significantly "more consistent" than another model if its inconsistency (difference out-ofsample and in-sample) is significantly lower than the inconsistencies of the other models, evaluated again in Schennach-Wilhelm LR tests. Universally across the three groups of experiments and the three degrees of knowledge about the target environment, Focal's estimates are significantly more consistent than those of the other models at the robust 0 005 level. 5 Result 3. Preference estimation using the Focal model is consistent across large experiments (at least eight observations per subject). Estimates from the small experiment (CHST07) differ weakly significantly, indicating weak identification. Estimates obtained using the benchmark models are inconsistent across all experiments.
Combined, the counterfactual predictions show that estimates from other experiments allow to predict behavior in CHST07 accurately, close to achieving in-sample accuracy, while estimate consistency shows that the estimates from CHST07 are not suitable to predict behavior in the other experiments. Jointly, this attests weak identification of preferences based on the CHST07 data (with only two observations per subject), as opposed to attesting differences between subject pools. That is, reliable preference measurement is possible with adequate choice models (such as Focal) and a sufficiently large number of observations per subject (at least 8-10, as in AM02 and HJ06).
Comparative relevance Finally, let me evaluate the economic relevance of controlling for presentation effects, both in absolute terms and in relative terms compared to the relevance of controlling for subject heterogeneity. To this end, I aggregate the BIC measures defined above across all data sets, in order to get overall assessments of descriptive and predictive adequacy, and using these aggregate BIC I compute the corresponding pseudo-R 2 as before.
These aggregate pseudo-R 2 are reported in Table 3. I focus on the results for logit and Focal, as the other benchmark models are similar to logit. As for modeling subject heterogeneity, I distinguish four model families: (i) representative agent models (called "Repr Agent" in Table 3) where all subjects have identical parameters; models with heterogeneous preferences ("Het Prefs") allowing for heterogeneity in preferences (random α and β with distributions as above) while subjects are assumed to have homogeneous precision λ and choice bias κ; models allowing for heterogeneity in preferences and precision (random α β λ as above, called "Het Pref and Prec" in Table 3) while subjects are assumed to have homogeneous choice bias κ, and models allowing for full heterogeneity (called "Full Het") used in the previous subsection. As indicated, I further distinguish the aggregate pseudo-R 2 in-sample ("Accuracy in-sample"), and of predictions to the first and second degree ("Out-of-sample degree 1" and "Out-of-sample degree 2"). Table 3. Relevance of controlling for round-number effects compared to relevance of controlling for subject heterogeneity (Pseudo-R 2 , less is better).
Repr Agent Het Prefs Het Pref and Prec Full Het
Accuracy in-sample (pseudo-R 2 based on BIC descr , see equation (15) Note: As discussed in the text, each panel provides pseudo-R 2 for overall 4 × 2 models: four models of subject heterogeneity (starting with the "representative agent" model, i.e., homogeneity) and two models of choice (multinomial logit and Focal). The underlying likelihood ratio tests follow Schennach and Wilhelm (2016), the relation signs indicate the direction of differences, and the p-values of the LR tests are printed underneath the relation signs.
In-sample, the representative agent logit model captures 21 2% of observed variance, and controlling for heterogeneity allows us to explain up to 59 8%, that is, an additional 38 percentage points. Controlling for focality has the same impact in isolation, raising the pseudo-R 2 to 60 3% in the representative agent model, and in addition to controlling for heterogeneity, it improves the R 2 still by 30 percentage points (up to a pseudo-R 2 of 88 5% in total). That is, focality and subject heterogeneity are quantitatively of similar relevance and fairly complementary in nature.
The results out-of-sample are qualitatively similar but quantitatively even stronger in demonstrating the relevance of controlling for both heterogeneity and focality. In predictions to both the first and the second degree, the representative agent logit model has negative validity (pseudo-R 2 of −9 1% and −4 8% out-of-sample), implying that representative-agent predictions provide at best uninformed guesses. This is intuitive, as the comparative statics depend on presentation, and if we calibrate a representative agent based on one experiment in Figure 2, then the predictions for other experiments, with differing contribution levels and comparative statics, must be off. Representative agent modeling seems to be fairly limited as far as giving (as in dictator games) is concerned.
The reliability of preference estimation (prediction to the first degree) improves by around 49 percentage points in terms of R 2 by controlling for heterogeneity, up to 40 1% in total, and by another 43 percentage points on top controlling for focality (up to 83 8%). The reliability of joint preferences and precision estimation (prediction to the second degree), that is, the distributions of (α β λ) across subjects, improves by 45 percentage points controlling for heterogeneity (up to again 40 1%) and controlling for focality adds another 44 percentage points on top (up to 84%).
Result 4.
Controlling for focality and controlling for heterogeneity are both highly relevant and largely complementary in reliable preference measurement.
Concluding remarks
This paper introduces and tests a simple model allowing to capture presentation effects in stochastic choice. The idea that presentation affects choice seems widely recognized, as a number of reasons for presentation effects such as ordering, round-number, and left-digit effects are well documented. Considering this, surprisingly few studies explicitly analyze presentation effects in choice, and the few examples I am aware of, for example, Bernheim and Rangel (2007) and Salant and Rubinstein (2008), all focus on rational (nonstochastic) choice, and thus are of limited help in econometric analyses. This paper contributes to this literature by showing the striking importance, comparable to controlling for heterogeneity, and the actual possibility of controlling for presentation effects in experimental data. The focal choice adjust logit (Focal) model developed here is tested by reanalyzing seminal experiments on dictator games, the most widely analyzed experimental games, and the results confirm the impression of the introductory comments. Choice patterns and comparative statics strongly depend on presentation, and inadvertently, counterfactual predictions and utility analyses neglecting presentation are unreliable (Results 2 and 3). This implies that falsifying predictions or detecting behavioral differences, by analyzing experiments without controlling for presentation, risks being uninformative or misleading.
The econometric results explicitly show that we can and need to control for presentation when making ex ante predictions for experiments or policy interventions, and equally when evaluating behavioral models. The latter requires rich data sets, with multiple observations per subject to disentangle preferences, noise and presentation bias. Without such experimental designs, estimates are inconsistent (for all models) and convergence of say preference modeling appears unrealistic. In the econometric analysis, I assumed that the focality of options is linear in the "level of roundedness" observed in statistical analysis of survey responses. This can be generalized straightforwardly, but linearity ex post seems sufficient when analyzing dictator choice. Further, the existing experimental literature recognizes potential ordering effects in choice from lists and routinely reverses the ordering to nullify such effects. As discussed in Rubinstein and Salant (2006), the extreme options in lists may well have relatively high or relatively low focality, implying that simple reversions do not nullify ordering effects. Proposition 1 suggests an alternative approach, namely to arrange the options along a circle, which can be rotated, and the resulting equation system identifies utility and focality indices up to affine transformation. Finally, in two of the experiments considered here, the payoff-equalizing "Leontief" choices had been round numbers in all tasks (AM02 and CHST07), and these two experiments happened to yield estimates of utility parameters deviating the most from the graphical choice benchmark FKM07. It thus seems advisable to vary the roundedness of predictions associated with particular models (as in HJ06). Alternatively, the graphical interface of FKM07 mitigates round-number effects, but this "number free" choice elicitation is not applicable in experiments on strategic choice and difficult to use outside the laboratory.
To conclude, the Focal model allows to control for the "focality" of options arising from the presentation of choice tasks. This is as critical for external validity as controlling for heterogeneity, as presentation affects even comparative statics in dictator games. The Focal model is widely applicable, as it directly generalizes logit, which is the workhorse model in analyses of both individual choice (as analyzed here) and strategic choice, including quantal response equilibrium (McKelvey and Palfrey (1995), McKelvey and Palfrey (1998)), cognitive hierarchy models (Camerer, Ho, and Chong (2004), Rogers, Palfrey, and Camerer (2009)), level-k models (Stahl and Wilson (1994), Costa-Gomes, Crawford, and Broseta (2001), Costa-Gomes and Weizsäcker (2008)) and noisy introspection (Goeree and Holt (2004)). In this sense, Focal is a versatile and promising model for behavioral analyses, besides being a rigorous framework to study presentation effects including nudging interventions. This opens up a wide range of interesting opportunities for analyses of behavior across experiments, as controlling for presentation enables joint analyses of multiple experiments, which in turn allows us to condense the information contained therein.
Appendix: Proof and technical definitions
A.1 Proof of Proposition 1 Fix (π φ) ∈ C, B ⊂ X, and define p x := log Pr(x|π φ X) for all x ∈ B. Further, define u x = u(π x ) and v x = v(φ x ), for all x.
Step 1: There exists a ∈ R such that p x = λu x + κφ x + a for all x. If Pr has a Focal representation, then by p x = log Pr(x|π φ X) that Now define a : X → R as a(x) = p x − λu x − κv x for all x ∈ X. Hence, for all x ∈ B ∈ P(X), and by transitivity, we obtain exp λu x + κv x + a x for all x x ∈ X implying a(x)/a(x ) = 1 for all x x ∈ X. Thus, there exists a ∈ R such that p x = λu x + κv x + a for all x.
Step 3: Ifπ is any circular permutation of π, then relabeling options transforms the circular permutation into a simple rotation as defined in Step 2. Hence, the claim holds true for all circular permutations.
A.2 A simple axiomatic foundation
Based on Breitmoser (2020), it is straightforward to provide a simple axiomatic foundation of the Focal model. The starting point that I use in this analysis is Gul and Pesendorfer (2001), who establish that a "rational" DM (with a preferences for commitment) maximizes u + v. The question I ask is, if we take the existence of u and v as given, in which conditions does "injecting" stochastic choice imply the Focal model. Formally, I inject stochastic choice by requiring positivity of choice probabilities, then translate max u + v into invariance statements about choice probabilities, and finally solve for the functional form that choice probabilities may take if they are compatible with these invariance statements. In order to reflect that we now take u and v as given, and simply ask about the functional form relating choice probabilities with u and v, I now denote choice probabilities as Pr(x|u v B). Relatedly, each pair (u v) characterizes a DM, and the set of all conceivable DM is denoted D. I write u x = u(x) and v x = v(x) for all x ∈ X.
The invariance statements are that both dimensions u and v matter (essentialness, otherwise the functional form is not unique), that choice satisfies independence of irrelevant alternatives (IIA), that choice depends solely on u and v ("completeness"), and that it is invariant to translation of u and v ("narrow bracketing" and "relative focality"), that is, that DMs with u or v differing only by translation choose options with equal probabilities. Note that all of these statements are compatible with max u + v, and in this sense, they (partially) characterize max u + v. As indicated, in addition, we assume positivity of choice probabilities.
Assumption 1 (Axioms). For all (u v) ∈ D, all B ∈ P(X), all x ∈ B and all r ∈ R, We solve for the functional form that choice probabilities may take, given they are compatible with these statements, assuming the following assumptions are satisfied. First, we need to make sure that we can manipulate u and v independently (Richness). Second, for any conceivable DM (u v), we need to ensure that DMs for all affine transformations of u and v are equally conceivable, based on which I show that DMs differing only by affine transformations actually are indistinguishable. 6 Finally, to guarantee uniqueness, the set of possible options X as well as the images of u and v are convex subsets of R.
Assumption 2 (Richness). 1. Choice tasks: For all (u v) ∈ D, (a u + b u · u a v + b v · v) ∈ D for all (a u b u ) (a v b v ) ∈ R 2 .
2. Richness: there exist (u v) and (x y) such that u x = u y and v x = v y , and there exist (u v) and (x y) such that u x = u y and v x = v y .
The following result establishes that the above invariance statements are satisfied if and only if choice probabilities have a Focal representation. Considering that the invariance statements simply characterize max u + v, and that the only other assumption is positivity, this suggests that the Focal model is a natural formulation of max u + v if we seek to allow for stochastic choice.
Theorem 1. Given Assumption 2, the following two statements are equivalent: 1. Choice profile Pr satisfies Axioms 1-6 2. Choice profile Pr has a Focal choice adjusted logit (Focal) representation: there exist unique (λ κ) ∈ R 2 such that, for all x ∈ B ∈ P(X) and all (u v) ∈ D, Since a formal proof of a slightly more general statement allowing for an arbitrary number of dimensions n is provided in Breitmoser (2020), I will resort to outline the key intuition here. Positivity and IIA imply that choice probabilities are functions of "choice propensities" (Luce (1959)). Without further information, the choice propensity of x ∈ X is defined only in relation to a benchmark option y ∈ X (McFadden (1974)), and as IIA applies to each DM (u v) separately, propensities may be DM dependent, implying Pr(x|u v B) = V (x y|u v) x ∈B V x y|u v with V (x y|u v) = f u v (u x u y v x v y x y) 6 I identify all real numbers as constant functions such that addition and multiplication of a function with a real are well-defined. Thus, for any u : X → R and any a b ∈ R, u = a + bu is equivalent to u x = a + bu x for all x ∈ X.
Any y ∈ X may serve as benchmark option, and by convexity of X, the reference to the benchmark can be dropped, that is, V (x y|u v) =f u v (u x v x x) for some functionf . By Axiom 6, completeness, options that are equivalent in terms of both utility and focality have equal choice propensities in any choice task, implying that choice propensities can be represented independently of x. Formally, a family of functionsṼ u v exists such that V u v (u x v x ) =f u v (u x v x x) for all x ∈ X, and all (u v). Given this characterization of propensities, narrow bracketing and focality implỹ for all x x ∈ X and r u r v ∈ R which in turn impliesṼ u+r u v+r v (u x + r u v x + r v ) =Ṽ u v (u x v x ) · g(r u r v ) for some function g. Due to the DM dependence ofṼ ,Ṽ u+r u v+r v =Ṽ u v is possible, but "completeness" restricts DM dependence by allowing the functional equation to be expressed as h(u x + r u v x + r v ) = h(u x v x ) · g(r u r v ) for some function h : R 2 → R. The main technical difficulty is that h is not necessarily differentiable. 7 By positivity, the logarithmic transformationh = log h andg = log g is admissible, which yields the Pexider functional equationh(u x + r u v x + r v ) =h(u x v x ) +g(r u r v ), and by their relation to probabilities,h is bounded from above for all values in the images of u and v (for any DM), 8 which each have positive length by "convexity" (Assumption 2). This implies that all solutions ofh are linear in u x and v x , and the essentialness of u and v implies that the respective coefficients (λ κ) are unique. Thus,h(u x v x ) = λu x + κv x + c x for all x, with unique λ κ ∈ R and c : X → R (Aczél and Dhombres (1989)). UsingṼ u v = exph, Completeness implies that c x is constant in x and cancels out, yielding the Focal representation. The detailed proof is given in Breitmoser (2020). | 2020-09-26T14:57:04.206Z | 2021-01-01T00:00:00.000 | {
"year": 2021,
"sha1": "67d69dfbc9e7e671ad61d8efecd9d1e70c188c9c",
"oa_license": "CCBYNC",
"oa_url": "https://qeconomics.org/ojs/index.php/qe/article/view/1564/1385",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "2275b8af3d2047543f337f933a72b999d53d3c4e",
"s2fieldsofstudy": [
"Economics"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
118355424 | pes2o/s2orc | v3-fos-license | Chiral color symmetry and $G'$-boson mass limit from Tevatron data on $t \bar t$-production
A gauge model with chiral color symmetry of quarks is considered and possible effects of the color $G'$-boson octet predicted by this symmetry are investigated. The contributions of the $G'$-boson to the cross section $\sigma_{t\bar{t}}$ and to the forward-backward asymmetry $A_{\rm FB}^{p \bar p}$ of $t\bar{t}$ production at the Tevatron are calculated and analysed in dependence on two free parameters of the model, the mixing angle $\theta_G$ and $G'$ mass $m_{G'}$. The $G'$-boson contributions to $\sigma_{t\bar{t}}$ and $A_{\rm FB}^{p \bar p}$ are shown to be consistent with the Tevatron data on $\sigma_{t\bar{t}}$ and $A_{\rm FB}^{p \bar p}$, the allowed region in the $m_{G'} - \theta_G$ plane is discussed and around $m_{G'}=1.2 \, TeV, \; \theta_G=14^\circ $ the region of $1 \sigma$ consistency is found.
The search for a new physics beyond the Standard Model (SM) is now one of the aims of the high energy physics. The simplest extentions of the SM (such as two Higgs models, models based on supersymmetry, left-right symmetry, four color quark-lepton symmetry or models implying the four fermion generation, etc.) predicting the new physics effects at one or a few TeV energies are most interesting now in anticipation of the new results from the LHC which will allow the investigations of new physics effects at the TeV energy scale with very large statistics [1].
One of the simplest extentions of the SM can be based on the idea of the originally chiral character of SU c (3) color symmstry of quarks. i.e on the gauge group of the chiral color symmetry which is assumed to be valid at high energies and is broken to usual QCD SU c (3) at low energy scale. The immediate consequence of the chiral color symmstry of quarks is the prediction of new color-octet gauge particle: the axigluon G A µ in the case of g L = g R [2][3][4][5] or the G ′ -boson in general case of g L = g R [6][7][8].
The G ′ -boson is the octet-colored gauge particle with vector and axial vector coupling constants to quarks of order g st which are defined by gauge coupling constants g L , g R . Some features of the axigluon (including its phenomenology at the Tevatron) were investigated in ref. [9][10][11][12] and the massive color octet with arbitrary vector and axial vector coupling constants to quarks has been considered phenomenologicaly in ref. [13].
Since it is the colored gauge particle with vector and axial vector coupling to quarks, the G ′ -boson should give rise the increase of the cross section as well as the appearance of a forwardbackward asymmetry in QQ production.
The main goal of my talk is to clean up if the gauge chiral color symmetry (1) is consistent with the CDF data (2), (3) and what bounds on the mass of G ′ -boson are imposed by these data.
In the case of the gauge chiral color symmetry (1) the 3 × 3 matrices of the usual gluon fields G µ and of the G ′ -boson fields G ′ µ are constructed from the basic gauge fields G L µ and G R µ as To reproduce the usual quark-gluon interaction of QCD the gauge coupling constants g L , g R of the gauge group G c must satisfy the relation where M chc is the mass scale of the chiral color symmetry breaking and g st (M chc ) is the strong coupling constant taken at this mass scale. The interaction of the G ′ -boson with quarks in this case takes the form where v and a are the vector and axial-vector coupling constants for the which the gauge chiral color symmetry group G c gives the expressions As a result of the chiral color symmetry breaking the G ′ -boson picks up the mass where η is the VEV of the (3 L ,3 R ) scalar field Φ αβ of the group G c , which breaks the chiral color symmetry, Φ αβ = δ αβ η/(2 √ 3), α, β = 1, 2, 3 are the SU L (3) and SU R (3) indices. So, the gauge chiral color symmetry model has two free parameters, the G ′ -boson mass m G ′ and the G L − G R mixing angle θ G , tg θ G = g R /g L , which gives the possibility to study the phenomenology of the G ′ -boson in more detail in dependence on these two parameters.
The differential cross section of the process qq g, G ′ → QQ in tree approximation with account of the G ′ -boson interaction (6) and of the gluon contributions has the form [8] dσ(qq where f (±) = (1 + β 2 c 2 ± 4m 2 Q /ŝ), c = cosθ,θ is the scattering angle of Q-quark in the parton center of mass frame,ŝ is the squared invariant mass of QQ system, β = 1 − 4m 2 Q /ŝ, M chc is the mass scale of the chiral color symmetry breaking and µ is a typical scale of the process.
The corresponding to (7) total cross section takes the form [8] σ(qq The enterring into (7), (8) hadronic width of the G ′ -boson is known [7,13] and can be written as is the width of G ′ -boson decay into QQ-pair. At M chc = 1.2 T eV , for example, we obtain the next estimations for the relative width of G ′ -boson Γ G ′ /m G ′ = 0.08, 0.14, 0.33, 0.60, 1.37 for θ G = 45 • , 30 • , 20 • , 15 • , 10 • respectively. As concerns the process of QQ production in gluon fusion gg → QQ the G ′ -boson does not contribute, in tree approximation, to this process.
The differential and total partonic cross sections of the process of QQ production in gluon fusion gg → QQ in tree approximation of the SM are well known and have the form The G ′ -boson can generate, at tree-level, a forward-backward asymmetry in QQ-pair production due to the forward-backward difference in the qq → QQ partonic cross section [8] which can give rise to the corresponding forward-backward asymmetry A pp FB of tt-pair production in pp collisions at the Tevatron. Figure 1: The m G ′ − θ G regions consistent with CDF data on cross section σ tt and forwardbackward asymmetry A pp FB in tt production within 1σ (dark region), 2σ (grey region) and 3σ (light-grey region).
We have calculated the cross section σ(pp → tt) of tt-pair production in pp-collisions at the Tevatron energy using the total parton cross section of quark-antiquark annihilation (8), the total SM parton cross section (10) of the gluon fusion gg → QQ and the parton densities AL'03 [35] (NLO, fixed-flavor-number, Q 2 = m 2 t ) with the appropriate K-factor K = 1.24 [36]. Here and below we beleive µ 2 = Q 2 , M chc = m G ′ .
With the same parton densities we have calculated the forward-backward asymmetry A pp FB in tt-pair production at the Tevatron in the form where A G ′ FB is the corresponding G ′ boson contribution which has been calculated using the differential parton cross section (7) (one can use also the expression (11)) and A SM FB is the SM prediction for A pp FB for which we have used the value (5) of ref. [12]. We have analysed the cross section σ(pp → tt) and the forward-backward asymmetry A pp FB in dependence on two free parameters of the model, the mixing angle θ G and G ′ mass m G ′ , in comparision with the Tevatron data (2), (3) on σ tt and A pp FB . The result of this analysis is shown in m G ′ − θ G plane in Fig.1.
The Fig.1 shows the regions in the m G ′ − θ G plane which are simultaneusly consistent with the data (2) and (3) within 1σ (dark region), 2σ (grey region) and 3σ (light-grey region). As seen from the Fig.1 for which are consistent with the CDF data (2), (3) on σ(pp → tt) and A pp FB within 1σ. In parentheses in (14)-(16) we show for comparision the G ′ -boson contributions in A pp FB defined by (11), without the SM contribution (5). As seen, the G ′ -boson can give in the forward-backward asymmetry A pp FB the contribution of about 10 % . So, the G ′ -boson induced by the chiral color symmetry (1) in general case of g L = g R is consistent with the data (2), (3) and can reduce the difference between the experimental and SM values (3), (5) of the forward-backward asymmetry A pp FB in the tt production at the Tevatron.
Summary
• The contributions of G ′ -boson predicted by the chiral color symmetry of quarks to the cross section σ tt and to the forward-backward asymmetry A pp FB of tt production at the Tevatron are calculated and analysed in dependence on two free parameters of the model, the G ′ mass m G ′ and mixing angle θ G .
• The G ′ -boson contributions to σ tt and A pp FB are shown to be consistent with the Tevatron data on σ tt and A pp FB and the allowed region in the m G ′ −θ G plane is discussed, in particular, it is shown that for m G ′ > 1.02 T eV in the m G ′ − θ G plane there is the region with 1σ consistency.
• So, the G ′ -boson induced by the chiral color symmetry of quarks in general case of g L = g R is consistent with the Tevatron data on σ tt and A pp FB and can reduce the difference between the experimental and predicted by SM values of the forward-backward asymmetry A pp | 2011-06-11T15:07:30.000Z | 2010-10-27T00:00:00.000 | {
"year": 2010,
"sha1": "389eb60c1c2440cf4db9d6bc8207932cecd2dcdd",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "389eb60c1c2440cf4db9d6bc8207932cecd2dcdd",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
248496752 | pes2o/s2orc | v3-fos-license | Trisecting the 9-vertex complex projective plane
In this paper we will give a short and direct proof that Wolfgang Kuehnel's 9-vertex triangulation of the complex projective plane really is the complex projective plane. The idea of our proof is to recall the trisection of the complex projective plane into 3 bi-disks and then to see this trisection inside a symmetry-breaking subdivision of the triangulation. Following the basic proof, we will elaborate on the construction.
Introduction
A k-simplex is a k-dimensional convex polytope with k + 1 vertices. For k = 0, 1, 2, 3 respectively, a k-simplex is usually called a vertex , edge, triangle, tetrahedron. When k is not important, a k-simplex is just called a simplex .
A simplicial complex is a finite collection C of simplices, all in an ambient Euclidean space, such that • If S ∈ C and S ′ is a sub-simplex of S then S ′ ∈ C.
• If S, T ∈ C then S ∩ T is either empty or in C.
Informally, the simplices in a simplicial complex fit together cleanly, without crashing through each other. The support |C| of C is the union of all the simplices in C. Often we blur the distinction between C and |C| and think of a simplicial complex as a union of simplices.
A simplicial complex may be described with no mention of the ambient space containing it, but there is always the understanding that in principle one can find an isomorphic complex in some Euclidean space. To give a pertinent example, let RP 2 6 be the quotient of the regular icosahedron by the antipodal map. This simplicial complex has 6 vertices, 15 edges, and 10 faces. One can reconstruct RP 2 6 in R 5 by fixing some 5-simplex Σ ⊂ R 5 , the convex hull of vertices v 1 , ..., v 6 , then mapping vertex k of RP 2 6 to v k and extending linearly. Figure 1 shows another incarnation of RP 2 6 . In this picture, the outer edges of the hexagon are supposed to be identified according to the labels. The complex RP 2 6 is called a 6-vertex triangulation of the real projective plane RP 2 because its support is homeomorphic to RP 2 . This triangulation has the fewest number of vertices amongst triangulations of RP 2 , so it is called a minimal triangulation of RP 2 . It is in fact the unique minimal triangulation of RP 2 . (Smaller examples like the quotient of the regular octahedron by the antipodal map fail to be simplicial complexes.) Here are some other examples related to minimal triangulations.
• The boundary of a tetrahedron is the unique 4-vertex minimal triangulation of the 2-sphere. More generally, the boundary of a (k + 1) simplex is the unique minimal triangulation of the k-sphere.
• If you identify the opposite sides of the big hexagon in Figure 4 below, you get the unique minimal triangulation T 2 7 of the 2-torus. T 2 7 has 14 triangles, 21 edges, and 7 vertices.
• In 1980, W. Kühnel discovered CP 2 9 , the unique 9-vertex minimal triangulation of the complex projective plane CP 2 . This triangulation has 36 4-simplices and a symmetry group of order 54.
• So far it an open question as to whether there is a 27-vertex triangulation of OP 2 , the octonionic (a.k.a. Cayley) projective plane.
• The minimal triangulations of RP 3 and RP 4 respectively have 11 and 16 vertices. See [D], • In 2021, K. Adiprasito, S. Avvakumov, R. Karasev [AAK] proved that real projective space can be triangulated using a sub-exponential number of simplices.
The survey article by B. Datta [D] has a wealth of information about minimal triangulations up to the year 2007 and a large number of references. The subject of this paper is CP 9 2 . In [KB], Kühnel and T. Banchoff establish many interesting properties of CP 2 9 and give a rather intricate proof that CP 2 9 really is homeomorphic to CP 2 . Since [KB], there has been a lot of work done trying to understand CP 2 9 from various points of view. In particular, there are a number of proofs that CP 2 9 ∼ = CP 2 , and also a number of proofs that CP 2 9 is the only minimal triangulation of CP 2 . See the article by B. Morin and M. Yoshida [MY] for a survey of these proofs. See also the paper by B. Bagchi and B. Datta [BD].
The purpose of this paper is to give a new and very nice proof that CP 2 9 ∼ = CP 2 . The basic idea of the proof here is to recall the trisection of CP 2 into 3 bi-disks, and then to see this trisection inside a symmetrybreaking subdivision of CP 2 9 . The construction is perfectly compatible with an easier version that works for RP 2 6 , so I will explain that as well. The picture developed here is related to the 10-vertex triangulation CP 2 10 of CP 2 that in [BK] is constructed by building outward from T 2 7 . Indeed Denis Gorodkov, in a private communication, explained to me how one can find a "path" from CP 2 9 to CP 2 10 using the subdivision idea and then something akin to bi-stellar flips. (I'll let Denis tell this story elsewhere if he wants to, but see the end of §2 for a hint.) My proof also has a close kinship with the "red-white-blue discussion" in §1.3 of the M.P.I.M. preprint by Morin and Yoshida that is the precursor to [MY] (and has the same title). This discussion is, in turn, related to Figure 8 in [KB]. Morin and Yoshida describe the red-white-blue discussion as a "topological insight" but they don't really push it forward into a proof. I think that my picture is very similar, but clarified by the special subdivision.
The approach here possibly could shed light on Gorodkov's result that HP 2 15 ∼ = HP 2 . The same subdivision and trisection ideas go through for HP 2 15 almost verbatim, and I can see computationally that each of the 3 subcomplexes is shellable and therefore PL homeomorphic to an 8-ball. However, the high dimensional topology involved in analyzing HP 2 15 makes a direct topological analysis of the whole complex formidable. For instance, the sub-complex that plays the role of T 2 7 has 288 6-simplices. A key step in extending the proof here to HP 2 15 would be showing that this 288-monster is homeomorphic to (S 3 × S 3 × S 3 )/S 3 in a 3-fold symmetric way.
Here is an outline of the paper.
• In §2 I will give the analogous version of my proof for RP 2 6 . This case is quite concrete and one can see the whole idea at a glance.
• In §3 I will recall the trisection of CP 2 and discuss a few key properties of the central torus in this decomposition.
• In §4 I will describe CP 2 9 and then explain my symmetry-breaking subdivision. The construction parallels the real case.
• In §5 I will find the trisection inside the subdivision and construct a homeomorphism h : CP 2 9 → CP 2 which respects the trisections. • In §6 I will explain how one can see the real case of the construction inside the complex case. This analysis leads to a refinement of h and gives the full power of our main result, Theorem 5.1.
• In §7 I will sketch how to make h completely explicit.
I thank Tom Banchoff, Kenny Blakey, Thomas Goodwillie, Denis Gorodkov, Joe Hlavinka, Wolfgang Kühnel, Tyler Lane, Dennis Sullivan, and Oleg Viro for helpful discussions. (Many of these discussions were about issues related to HP 2 15 .) I also thank the anonymous referee for a number of helpful comments, especially those pertaining to the real case of the construction. These comments from the referee inspired §6-7.
The Real Case
RP 2 is the space of scale equivalence classes of nonzero vectors in R 3 . We denote the equivalence class of ( We have the trisection Points in β 1 may be written uniquely in the form [1 : x 2 : x 3 ], with |x 2 |, |x 3 | ≤ 1. Thus β 1 is a square. So are β 2 and β 3 . Each intersection β i ∩ β j is a pair of opposite edges, and the triple intersection is a union of the 4 points [±1 : ±1 : ±1]. If we interpret RP 2 as the quotient of a cube by the antipodal map, then the 3 quotient faces are β 1 , β 2 , β 3 . The trisection has 3-fold symmetry. The map Σ : (x 1 , x 2 , x 3 ) → (x 2 , x 3 , x 1 ) permutes the sets β 1 , β 2 , β 3 . In terms of the cube, Σ rotates around the appropriate long diagonal. RP 2 6 has a very similar 3-fold symmetry: The permutation S = (123)(456) 6 into 18 triangles. Using the new vertices, we divide the central triangle of RP 2 6 into 6 triangles and we subdivide each of the adjacent triangles in half. The subdivision has 18 = 3 × 6 triangles, with each having exactly one vertex from the set {1, 2, 3}. For j = 1, 2, 3 we let B k be the subset of 6 new triangles having k for a vertex. The sets B 1 , B 2 , B 3 are colored green, red, blue in Figure 2.
This is now the trisection, and there is a clear homeomorphism from this subdivided complex to RP 2 which maps B j to β j and conjugates S to Σ.
The Smooth Trisection: Complex Case
The complex projective plane CP 2 is defined just as RP 2 but with respect to the field C of complex numbers. We denote points in CP 2 by [z 1 : z 4 : z 7 ]. The variable names will line up with the notation for CP 2 9 . We have the trisection CP 2 = β 1 ∪ β 4 ∪ β 7 , where β j is defined just as in the real case, using the complex norm in place of the absolute value. This time, β j is the product of 2 unit disks. The bi-disks β 1 , β 4 , β 7 have disjoint interiors and are permuted by the same map Σ as defined in the real case.
The boundary ∂β 1 is a 3-sphere, and it decomposes into the solid tori β 14 and β 17 . Here β ij = β i ∩ β j . To see that β 14 is a solid torus, note that β 14 consists of points of the form [1 : u : z] with |z| ≤ |u| = 1 and is therefore the product of the unit disk and the unit circle. The central torus β 147 = β 14 ∩ β 17 = β 1 ∩ β 4 ∩ β 7 consists of points where |z 1 | = |z 4 | = |z 7 |. We discuss β 147 in more detail, with a view towards seeing it inside CP 2 9 .
Hexagonal Structure: Let R 3 0 ⊂ R 3 denote the plane of points whose coordinates sum to 0. Let H = Let [ij] be the midpoint of the edge i ↔ j. Let [ijk] be the center of the triangle ijk. Let the rank of a simplex be the number of vertices which belong to the set {1, 4, 7}. Our list above goes by rank. Parallel to the real case, we divide each rank k simplex into k! smaller simplices, as follows: The rank 1 simplices are untouched. The rank 2 simplex 14abc divides into 1 [14]abc 4 [14]abc, and likewise with the indices 1, 4, 7 permuted. The rank 3 simplex 147ab divides into We replace our original 36 simplices with the subdivided simplices. Since there are respectively 18, 12, 6 simplices of rank 1, 2, 3 we get a total of (1, 2, 6) · (18, 12, 6) = 78 = 3 × 26 new simplices. (The rank 1 simplices count as "new".) Each new simplex has exactly one vertex from the set {1, 4, 7}.
The Combinatorial Trisection
We have CP 2 9 = B 1 ∪ B 4 ∪ B 7 , where B j is the union of the 26 new simplices having j ∈ {1, 4, 7} as a vertex. Each B j is the cone to vertex j of ∂B j . Hence B i and B j have disjoint interiors for i = j. Here is our main result. 2. h maps B j to β j for j = 1, 4, 7.
h conjugates T to T .
In this section I will construct a non-explicit homeomorphism h which has the first 3 properties but not necessarily the fourth. This should satisfy a reader who just wants to see why CP 2 9 ∼ = CP 2 . In §6, I will give a more refined version of h which has the fourth property. In §7 I will sketch how to make h explicit.
The first thing we do is list the tetrahedra in B 14 = B 1 ∩B 4 . We will derive this tetrahedron list from the simplex list above. The reader might want to check that this actually works, so for convenience we repeat the simplex list here: • 15289 12389 13689 45289 42389 43689 • 14256 14356 14259 14368 • 14726 14768 (14783 14735 14759 14792) Now for the derivation. We get 13 tetrahedra contained in B 14 by subdividing the simplices on our list above and omitting 1 or 4. The tetrahedra are listed in a way that corresponds to the simplices.
Lemma 5.2 B 14 is a solid torus.
is the union of the first 3 tetrahedra above and B ′′ 14 is the union of the last ten. B ′ 14 is a 3-ball because it is the join of the path 5236 with the segment 89, and B ′′ 14 is a 3-ball because it is the cone to vertex [14] of ∂B ′′ 14 , a 10-triangle triangulation of the 2-sphere. Figure 3 below shows ∂B ′ 14 and ∂B ′′ 14 . Each one is drawn as the union of 2 combinatorial hexagons glued along their boundaries according to the labels. B ′ 14 ∩ B ′′ 14 is the union of the 2 disjoint grey triangles 259 and 368. Topologically, we get B 14 by gluing two 3-balls together along a pair of disjoint disks in their boundaries. The orientations of the gluings are such that the result is a solid torus (as opposed to the so-called solid Klein bottle, a nontrivial disk bundle over the circle). ♠ We get the triangulation of B 147 = ∂B 14 by gluing the two triangulations from Figure 3 along the grey triangles. Figure 4 shows the universal cover of the triangulation. We get back to B 147 by gluing the opposite sides of the big hexagon by translations. This triangulation of B 147 is exactly T 2 7 . Note that From all this structure we see that (after suitably scaling) there is an isometry h 147 : B 147 → β 147 which conjugates S to Σ and which maps the green loop a 14 to α 14 . The labels of the hexagon vertices, such as (1, −1, 0), indicate precisely how the hexagon here lines up with the one described in connection with the central torus of CP 2 . Note that a 14 is contractible in B 14 because a 14 ⊂ B ′′ 14 , and is a homeomorphism which respects the individual pieces and their intersections. Since B j and β j are cones over ∂B j and ∂β j we can extend h, by coning, to a homeomorphism from CP 2 9 = B 1 ∪B 4 ∪B 7 to CP 2 = β 1 ∪β 4 ∪β 7 .
The Extra Symmetry
When we use the top label of each pair, Figure 5 shows Y 14 . (Think about cutting a pink-frosted grey donut in half.) The pink boundary is half of B 147 . The left grey disk is D ′ 14 and the right grey disk is D ′′ 14 . The map T acts as rotation by 180 degrees about the purple line bisecting the grey disks. The intersection of the purple line with the grey disks is the part lying in our copy of RP 2 6 . When we use the bottom labels, Figure 5 shows the same things for Υ 14 .
By construction h 147 (∂D ′ 14 ) = ∂∆ ′ 14 and h 147 (∂D ′′ 14 ) = ∂∆ ′′ 14 . We define h 14 on each of D ′ 14 and D ′′ 14 by coning over the boundaries. By symmetry this extension conjugates T to T and is defined in particular on ∂Y 14 . Our extension maps the (pink and grey) sphere ∂Y 14 to the (pink and grey) sphere ∂Υ 14 . We now extend to a homeomorphism from the ball Y 14 to the ball Υ 14 and use the action of T and T to extend the homeomorphism to all of B 14 .
Our improved h 14 conjugates T to T .
The rest of the construction is as above. The union map h 14 ∪ h 17 ∪ h 47 , defined on ∂B 1 ∪ ∂B 4 ∪ ∂B 7 , conjugates T to T because the pairs (S, T ) and (Σ, T ) commute. The final coning process respects T and T , so the final Here is where h sends the vertices: • The remaining images can be readily deduced from the action of S, T, Σ, T . The last two entries require some explanation. The coordinates of the point p 2 ∈ R 2 0 corresponding to vertex 2 are (−1, 5, −4)/7. We found this by solving the equation 2p 2 − Σ 2 (p 2 ) = (−1, 2, −1). The choice of where to send [259] is not determined by the construction above, but we might as well make it. The explicit construction below makes this choice, and so it is convenient to list it here.
Making the Homeomorphism Explicit
The only non-explicit part of our construction is the extenson of the sphere map h 147 : ∂Y 14 → ∂Υ 14 to the ball map h 14 : Y 14 → Υ 14 . In this section we sketch an explicit extension. Gluing the opposite blue sides of the parallelogram in Figure 6 gives the cylinder ∂Y 14 − D ′ 14 − D ′′ 14 , drawn pink in Figure 5. The green loops a 14 and a 14 ′ are on the boundary. Figure 6 suggests an explicit foliation of ∂Y 14 by polygonal loops. Intrinsically these are geodesic bigons in B 147 . The intersections with the blue edges of the triangulation move linearly.
Let the core of Y 14 be the path with vertices [14], [259], [[23][89]]. Figure 6 indicates a piecewise linear correspondence between the loops in the foliation and the points on the core. We cone each loop in the foliation to the corresponding point on the core. Now (after some checking of disjointness) we have a disk foliation of Y 14 which interpolates between D ′ 14 and D ′′ 14 and respects the partition of Y 14 into Y 14 ∩ B ′ 14 and Y 14 ∩ B ′′ 14 . We use h 147 to transfer our foliation to ∂Υ 14 − ∆ ′ 14 − ∆ ′′ 14 . We change coordinates: [1 : u : z] → (arg u, z).
In these coordinates, Υ 14 = [0, π] × D 2 . Let s be the blue line segment in Figure 6 that runs all the way across the cylinder and contains vertex 2. The | 2022-05-03T06:47:28.826Z | 2022-05-02T00:00:00.000 | {
"year": 2022,
"sha1": "83eeee4490a3b2f52196ba69e8ecaacfef6214f5",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "871b1bd5ac0fcc21a760d1c0198f33c0411c7a02",
"s2fieldsofstudy": [
"Mathematics"
],
"extfieldsofstudy": [
"Mathematics"
]
} |
240796025 | pes2o/s2orc | v3-fos-license | READING
; guess from the context; analyze vocabulary; distinguish between literal and implied meanings; activate linguosociocultural and linguistic knowledge; practice asking and answering questions on the text; practice narrow reading
Formulation of the problem. Reading is one of the four language skills and presupposes getting information from different types of texts. It is extremely important for future philologists to master reading skills on an appropriate level to perform professional duties properly.
Analysis of recent researches and publications. The problem of teaching reading English academic texts were studied by W.Grabe, L. Miller, F. Stollar, I. Lytvynenko, L. Kozhedub.
The objective of the article is to analyze the peculiarities of the process of teaching reading English academic texts to future teachers-philologists.
Presentation of the main research material. M. Wolf defines reading as a neuronally and intellectually circuitous act, enriched as much by the unpredictable indirections of a reader's inferences and thoughts, as by the direct message to the eye from the text [17, p. 16].
K. S. Goodman [8] offered the distinction of the bottom-up and the top-down reading.
In bottom-up processing, readers first recognize a variety of linguistic signals (letters, morphemes, syllables, words, phrases, grammatical cues, discourse markers) and use their linguistic dataprocessing mechanisms to impose some sort of order on these signals [13, p. 106].
Bottom-up reading starts at the "bottom" level of text structure, and moves «upward» to phrases, sentences, paragraphs, and chunks of written discourse [11, p. 17].
A top-down strategy presupposes using previous knowledge and assumptions as the reader goes down from the general meanings to the specific ones of the text.
The analysis of the scientific literature [13; 14; 15] showed that the most typical are the following reading strategies: identifying the purpose of reading and the type of reading; for rapid comprehension use silent reading techniques; skim the text to understand main ideas, then read it more in depth; scan the text to look for specific information; to understand the text better, follow the line of argument or memorize the information; guess from the context (a meaning of a word, discourse relationship, implied meaning, cultural reference and content messages); analyze vocabulary (prefixes, suffixes, familiar roots, grammatical and semantic context); distinguish between literal and implied meanings; activate linguosociocultural and linguistic knowledge; practice asking and answering questions on the text; practice narrow reading.
The analysis of scientific literature showed that scholars [6] recognize the exceptional role of the reader who brings a number of different things into the reading process, including purposes for reading (information/learning, or pleasure), background (the influences of family, school, and culture or subculture of origin), attitudes toward reading and literacy in general (which may be shaped both by purpose and background), and prior knowledge (linguistic, content, and rhetorical) related to the information in the text. In addition, readers bring individual differences in personality, learning styles and strategies, reading strategies, and life experience. Thus, a group of readers encountering the same text at the same point in time will not have identical experiences with reading the text or the same reactions to or interpretations of it.
E. B. Bernhardt claimed that at the higher levels of language learning the role of first-language-based knowledge is more powerful than at lower levels because the nature of the upper-level knowledge is much more dense, complex, and complicated [6].
To develop the methodology of teaching philology students English academic reading it is necessary to analyse the components of reading competence among which scientists distinguish skills and knowledge.
Skills are defined as: an acquired ability to perform well [7]; essential academic habits [9]; an acquired ability that comprises interrelated yet separable subskills [6, p. 36]; an ability to solve communuicative tasks in new situations.
It is necessary to analyse reading skills and determine which ones are the most important for effective academic reading. W. Grabe and F. L. Stoller think that reading skills represent linguistic processing abilities that are relatively automatic in their use and their combinations [9].
According to W. Grabe and F. L. Stoller, fluent reading is a rapid, efficient, interactive, strategic, flexible, evaluating, purposeful, comprehending, learning and linguistic process [9]. Word-reading efficiency, vocabulary development, text reading ease, comprehension, use of strategies -all these are constituents of fluent reading [9].
Reading is interactive as different processes are carried out simultaneously (readers recognise words rapidly and keep them active for some time in working memory [5], analyse the structure of sentences to understand their meanings, make conclusions about main ideas etc); information from the text interacts with information (linguistic and background) activated by the reader from long-term memory.
The reading process must be strategic as the reader needs to deal with different skills, encounter difficulties, monitor comprehension, seek solutions, correct imbalance between text information and reader knowledge, shift goals. Changing purposes and monitoring comprehension demands the ability to read flexibly. The reader has to evaluate if the information being read is useful, interesting, meets the reader's goals. Reading is always purposeful as, firstly, students usually have some purpose set internally or externally and, secondly, they have to read differently depending on the reading purposes. In academic settings, where students usually learn from texts, reading is always a learning process [9, p. 12] from which philology students learn some professional information and develop useful skills. As for philology students academic reading is always a linguistic process as readers have to understand different words, structures etc., analyse them.
The basic requirement for fluent reading comprehension is automatic word recognition. Following W. Grabe and F. L. Stoller's view on similar first and second language reading abilities at advanced levels [10] the given data seem to be of equal importance for fluent second language readers. For such readers word recognition is very fast and automatic which requires a lot of ptactice in reading. It also proves the necessity of expanding student vocabulary and forming strong lexical skills which mean that the words are well represented on orthographic, phonological, semantic and syntactic levels [9].
A fluent reader extracts grammatical information from words taken together, defines the meanings of words that have multiple meanings out of context [10, p. 16]. However, rapid and automatic syntactic processing is less obvious for second language readers in comparison with word recognition. W. Grabe and F. L. Stoller define two main reasons for it: 1) most students master second language grammatical structures before they become fluent readers; 2) to develop automaticity in using information from grammatical structures second language students need, first of all, countless hours of practice [10, p. 18].
Working memory plays an exceptional role on the level as it supports processing for word recognition, stores activated words, carries out syntactic processing and stores information, deletes unnecessary one, combines information to build comprehension [9, p. 36].
There are also higher-level processes that refer to reading comprehension. The most fundamental of them is the coordination of main and supporting ideas of the text to build the text model of comprehension. As new meaning units are added, the ideas that are used repeatedly and are logically connected to other information are considered to be the main ideas of the text. Under the influence of background knowledge, goals, motivation, task, text peculiarities, reader attitudes toward the text, feelings, expectations etc. the reader starts to interpret the information from the text, design a possible direction of the reading [10].
The purpose of reading defines which reading process will be greater emphasized. For example, reading for general comprehension entails coordination of main and supporting ideas as well as text interpretation; finding specific information focuses on word recognition and background knowledge anticipation of what to look for; reading to learn presupposes creating an accurate text model of comprehension and logical interpretation which integrates new knowledge with existing one [10, p. 23].
Higher-level processes are also closely connected with working memory, the major component of which is executive control recognized as central to comprehension processing. It carries out main attentional processes, stores information during reading and suppresses irrelevant information or the information that is not needed any more, controls shifting attention.
T. Hudson identified [12] four reading skill categories: decoding skills; fluency skills; comprehension skills; critical reading skills), we have come to the conclusion that the first two are typical of the lower-level processes while the other twoof the upper-level ones.
B. V. Rosenshine [16] made a conclusion about seven subskill areas which are included into the above-mentioned four categories: information sequence recognition; recognition of words in context; identification of main ideas; decoding of details; inferencing; cause and effect recognition; comparing and contrasting.
According to W. Grabe and F. L. Stoller texts are read with the following purposes: to search for information, learn from texts, integrate information, write, search for information needed for writing, analyse and critique texts, for general comprehension. All these purposes are typical of academic and professional contexts. Students usually have to learn a considerable amount of information from a text which they have to read at a rather slow speed to comprehend it well, remember main ideas and important details, connect them to students' background knowledge. Reading to integrate information requires critical evaluation of the information from the text and taking decision about its importance and how to integrate and compose the information from different texts and with what students already know [10].
There are different types of reading. When a text is read just for specific information, it is called scanning which presupposes hurrying over most of the text until the necessary information is found. Students read with the speed of 400-500 (up to 600 words per minute [11]) and understand 40-50 % of information.
Reading for gist or skimming is reading quickly through the text to get its general idea (for example, when it is necessary to decide whether to read an article or not) with the speed of 180-190 words per minute and understanding 70-75 % of the text. Reading for a detail students try to get the meaning out of every word and fully understand the text. The speed of reading is 50-60 words per minute [2].
In academic settings usually six academic purposes for reading are usually distinguished. Apart from generally recognized three aims 1) searching for information (scanning), 2) reading for general information (skimming), 3) reading for a detail. W. Grabe suggests three more targets to be attained by academic reading: 4) reading to learn some information, 5) reading to integrate information from different sources and build a general frame, 6) reading to analyse, critique and use information [9]. Two more purposes of reading academic texts may be added: 7) to write (so the article serves as a model for completing some written task) and 8) to search for the information needed for writing.
On the basis of modern literature analysis [10; 11] a list of reading comprehension skills for academic reading is drawn up and the types of reading are classified: -Scanning: identify the topic, contents and significance of academic texts (articles, theses, reports etc.) and decide whether the text is worth reading more in detail; find the necessary information in academic texts; select and understand evaluative judgements in academic texts; assess the relative importance of the information.
-Skimming: understand the gist and main details of academic texts (articles, theses, reports etc.); define quickly the contents and expediency of a more detailed reading of academic texts; understand the gist and major details of long reviews, analytical and polemical articles which present different views; evaluate importance, novelty, certainty and persuasiveness of the information presented; understand the gist and major details of complicated academic texts which presuppose different treatment and be able to evaluate certainty and truthfulness of events and facts; infer the main idea using patterns and other clues; use context clues to understand meaning; classify ideas into major ones and details; anticipate the contents of the text on the basis of the title and key words; identify tone or emotion in a text; evaluate the accuracy of a text judging from what the reader already knows; integrate text information with existing schematic knowledge; assess the relative importance of the information; recognize and repair miscomprehension.
-Reading for a detail: understand argument in an academic text; follow the development of argument in academic texts; classify ideas into major ones and details; anticipate the contents of the text on the basis of the title and key words; distinguish facts, ideas and opinions in complicated academic texts expressed explicitly; distinguish author's opinion expressed implicitly and explicitly; differentiate information about real facts and assumptions; evaluate the accuracy of a text judging from what the reader already knows; make conclusions about the author's choice of lexical items; make conclusions about the author's syntax; understand linking words; ask questions in an inner dialog with the author; categorize words and ideas into general and specific; identify the relationships between ideas; understand the structure of the text; be able to see connections between ideas; recognize and repair miscomprehension; integrate text information with existing schematic knowledge; identify tone or emotion in a text.
-Critical reading: evaluate the accuracy of a text with respect to what the reader already knows; identify the relationships between ideas; understand the structure of the text; be able to see connections between ideas; evaluate the accuracy of a text with respect to what the reader already knows; evaluate author's argument and its strength; analyse and evaluate author's meaning stated explicitly and implicitly; integrate the new ideas into the known ones; prioritize writer's ideas; develop own ideas using writer's.
Conclusions. On the basis of scientific literature analysis [10; 11] we have specified that to read academic texts efficiently students should know: types of reading; academic texts structure; academic style characteristics; linking words; genre characteristics; explicit and implicit meanings in the text; reading strategies; how to analyse the information critically; how to use strategies efficiently; general purposes; sources of information; peculiarities of one's own learning style; preferred strategies in accordance with one's own learning style.
The components of academic reading competencereading skills, reading comprehension skills and study skills as well as knowledge -have been determined, analysed and specified in the section. The abovementioned skills and knowledge cannot and should not be separated, but learned in support of each other. Academic reading competence development is influenced by the following factors: learning goals, motivation, practice, subject knowledge, strategy use, social interaction, selfreflection, autonomy support, correctly organized instruction. | 2020-09-10T10:07:55.455Z | 2020-01-01T00:00:00.000 | {
"year": 2020,
"sha1": "57d00178caa5393d82049773398853f2e6acb2bf",
"oa_license": null,
"oa_url": "https://doi.org/10.32840/1992-5786.2020.70-3.23",
"oa_status": "GOLD",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "39c68d22a666af49f7a2573cf40a3399279b38f6",
"s2fieldsofstudy": [
"Linguistics",
"Education"
],
"extfieldsofstudy": [
"History"
]
} |
31720235 | pes2o/s2orc | v3-fos-license | Neutrophil 'connectivity': key to neutrophil-mediated tissue injury?
Neutrophils use cell surface molecules to communicate with their external environment. These molecules are markers reflecting neutrophil development, activation status and cell function. They are also critically important in controlling neutrophil behavior. Targeting these cell surface molecules is an attractive approach in the treatment of neutrophil-mediated conditions.
Neutrophils are professional phagocytic cells that provide the host with a first line of defense against acute bacterial and fungal diseases. They sense the focus of infection, they adhere to the endothelium of capillaries and venules adjacent to the inflammatory locus, they migrate through the vessel wall and the interstitial tissues to the infectious site and they phagocytose, kill and digest the invading microorganisms using a large number of proinflammatory mediators and proteolytic enzymes. During the inflammatory process, neutrophils produce factors to ensure their survival in the hostile inflammatory milieu, they recruit additional phagocytes, they inactivate their own toxic products and they induce their own death pathway to prevent damage to normal host tissue. Neutrophils may damage normal tissue, and a number of clinical conditions such as acute respiratory distress syndrome, septicemia with multiorgan failure, ischemia reperfusion injury and rheumatoid arthritis have all been linked to inappropriate neutrophil-mediated tissue damage.
In the present issue of Critical Care, Seely and colleagues delineate the important association of neutrophil cell membrane molecules with specific neutrophil function [1]. They argue that neutrophil membrane molecules mediate the processes integral to neutrophil delivery, function and clearance. These surface molecules connect neutrophils to their external environment (connectivity), and alterations in these molecules reflect changes in cell function and behavior during every stage of the neutrophil's lifespan. Cell surface molecules alter with neutrophil proliferation, differentiation, maturation and the release of cells from the marrow into the circulation. Surface adhesion molecules and chemoattractant receptors are pivotal in determining tissue localization and recruitment of neutrophils, and when in the tissues, cell surface receptors are critical in cell activation and recognition of foreign pathogens. Surface molecules promote phagocytosis of pathogens and foreign material, initiate exocytosis of granules and participate in their final removal from the inflammatory site. Unraveling the relationship between neutrophil surface molecules and their behavior could impact on our understanding of the pathogenesis of neutrophil-mediated diseases and could spearhead potential therapeutic approaches [2].
Inflammation and stress accelerate neutrophil production, shorten their maturation time in the marrow and allow immature neutrophils to enter the circulation. This marrow response results in heterogeneity in the expression of surface receptors on circulating neutrophils such as the lipopolysaccharide receptor CD14, Toll-like receptors, CD16 and CD11/CD18 [2][3][4]. The CD11/CD18, Toll-like receptors and CD14 receptors have been implicated in the pathophysiology of septicemia and septic shock [5]. Interaction of lipopolysaccharide with these transmembrane receptors transduces intracellular activation signal through a signaling complex comprising heat-shock protein 70 and Commentary Neutrophil 'connectivity': key to neutrophil-mediated tissue injury?
Furthermore, we have shown that neutrophils newly released from the bone marrow express higher levels of L-selectin, a molecule that contributes to the recruitment of neutrophils to a site of inflammation [7]. L-selectin expression on neutrophils decreases as they age in the circulation with neutrophils in the circulation [8]. This unique L-selectin heterogeneity of expression determines which neutrophils participate in the inflammatory response and which neutrophils are destined to be removed from the circulating pool. Bone marrow stimulation induced by infection or smoking cigarettes causes a skip in cell division and a rapid transit of neutrophils through the mitotic stage. This leads to the production of neutrophils with higher granule numbers and greater destructive capabilities [9,10].
The primary granules in neutrophils are formed at an early stage (promyelocytic) in the mitotic pool in the marrow, and the number of granules is reduced by mitosis as the cells pass through the mitotic stage. These granules contain proteolytic enzymes such as myeloperoxidase, which is critical for neutrophil reactive oxygen radical production. Other granule proteolytic enzymes such as elastase, proteinase 3, cathepsin G and metalloproteinases are also formed early during the mitotic phase of development in the marrow. We suspect that neutrophils released during marrow stimulation contain higher levels of these potentially destructive proteolytic enzymes that play a pivotal role in inappropriate neutrophil-mediated tissue injury associated with infection and sepsis.
Targeting these potentially damaging cells using surface molecules (such as L-selectin, CD11/CD18) may have therapeutic benefits to minimize tissue damage in conditions such as sepsis or with ischemia-reperfusion injury. Blocking antibodies against the β 2 -integrin prevented the ischemiainduced renal infiltration of granulocytes and reduced infarct size in experimental models, with human studies still controversial and ongoing [11,12]. Immune complex diseases such as glomerulonephritis, immune vasculitis, arthritis and systemic lupus [13] are similarly characterized by neutrophilic inflammation, and targeting surface molecules such as Fcγ receptors or surface myeloperoxidase may potentially provide novel therapeutic strategies in the treatment of autoantibody-triggered inflammation.
Recently discovered molecules such as lipid rafts and tetraspanins have generated renewed interest in the studies of the cell membrane. It is thought that signaling events taking place in immune cells including neutrophils occur in specialized membrane domains called lipid rafts. Lipid rafts function as platforms for the formation of multicomponent Confocal laser scanning microscopy images of immunostained peripheral blood neutrophils of asthmatic subjects. Representative images of human neutrophils stained with Alexa-conjugated secondary antibody (red) to detect elastase, and BODIPY-FL-conjugated secondary antibody (green) to detect immunoreactivity against CD63. transduction complexes and they may represent clinically relevant potential targets for immune regulation [14]. It has been shown in human neutrophils that localization of FcγRII to lipid rafts is important for the activation of Src family protein tyrosine kinases to initiate the tyrosine phosphorylation cascade leading to superoxide generation [14].
The tetraspanin superfamily also seems to be an important component of cell surface molecules. Although their precise function is not known, data from knockout mice suggest that they play a major role in membrane biology. One of the welldocumented properties is their ability to facilitate the formation of multimolecular complexes (tetraspanin web) in which a number of molecules such as integrins are included [15]. Studies from our laboratory showed that the tetraspanin CD63 may be involved in neutrophil exocytosis (Fig. 1).
The study of neutrophil surface molecule functions and kinetics has advanced our understanding of the pivotal role of neutrophil 'connectivity' to their behavior and their ability to induce tissue damage through the release of damaging mediators. Future studies should focus on understanding the role of surface molecules in the multistep regulatory mechanisms in neutrophil behavior to limit the harm and tissue injury caused by neutrophil-derived mediators. | 2014-10-01T00:00:00.000Z | 2003-02-24T00:00:00.000 | {
"year": 2003,
"sha1": "5eed7c3d8c07fd3c65c5feeb7c8ab5c69c1931e6",
"oa_license": "CCBY",
"oa_url": "https://ccforum.biomedcentral.com/track/pdf/10.1186/cc1884",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "5eed7c3d8c07fd3c65c5feeb7c8ab5c69c1931e6",
"s2fieldsofstudy": [
"Biology",
"Medicine"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
56208259 | pes2o/s2orc | v3-fos-license | Study of Thickness Effect on Fracture Toughness of High Grade Pipeline Steel
The critical fracture toughness decreases when thickness of specimens increases and stress-strain field in crack tip start changing to plane strain state from plane stress state. In this paper, fracture toughness tests were carried. Based on the analysis of stress-strain field in crack tip and fracture toughness test results, a fracture toughness-thickness empirical model was established and the plane strain fracture toughness and critical thickness of X80/X100 were calculated. Then the validity of the empirical model was discussed and verified. The analytical results indicate that the safety of thick wall pipelines is worth of attention.
Introduction
In fracture mechanics when stress intensity factor in crack tip is equal to the material fracture toughness, crack will start to grow.That means fracture toughness of material can forecast the remaining strength of a component with an initial crack.Although fracture toughness is material inherent attribute, for the same material, different fracture toughness values was determined in different tests as test conditions (temperature, loading rate, et al) and specimen size are different.Of all the influence factors for fracture toughness test, specimen thickness is a most important factor.
Because there are amounts of oil and gas pipeline in the world, so cracks existing in some pipes body are almost inevitable.Stress in crack tip varies as pipe thickness is different.As thickness increases, stress-strain field in crack tip start changing to plane strain state from plane stress state, which means crack tip is in tension state in all three directions and the plastic zone will be limited in a small scope.So critical fracture toughness will decreases when pipe thickness increases in plane strain state, and brittle fracture is prone to happen, which is much more dangerous compared with ductile fracture.
In Chinese, pipe with high steel grade, large diameter and thick wall have been developing all the time.For pipes with high steel grade and thick wall which have been widely used, studies are focusing on whether fracture toughness has reached plane strain state or not.Fracture toughness tests of X80 pipeline steel that has been widely used and industrial trial -produced X100 pipeline steel with different wall thickness-were conducted and carefully analyzed in this paper.Based on analysis of stress field in crack tip, plane strain fracture toughness forecasting model was built, which is used to forecast whether the pipe thickness meets the plane strain condition or not.
Thickness Effect of Fracture Toughness
For component with a crack and external load, stress intensity factor K I (take I mode crack as example) is a mechanics parameter to describe stress field in crack tip.K I would increase if external loading increases.When K I increases to a critical value, the crack in component begins to grow.This critical value is called fracture toughness K C or K IC , which represents the material's ability to resist unstable propagation of crack.Fracture toughness K C or K IC is inherent attribute of a material.The difference is that K C is fracture toughness under plane stress state, which would be influenced by thickness of plate or test specimen.When specimen thickness increased, fracture toughness tends to be a stable and lowest value, which would not be influence by thickness.This value is called K IC or plane strain fracture toughness.K IC is the real material constant, which reflects the material's ability to prevent crack extension.
Test Materials
X80 and X100 specimens were separately sampled from Ø1219 26.4mm and Ø1219 23.5mm longitudinal submerged arc welding pipe.Mechanical properties of the X80 and X100 pipes were shown in Table 1.
Test Results
In order to work out the thickest test specimen, CT(compact tensile ) specimens were used.For pipelines, axial crack is most dangerous because the largest stress on pipe locate in hoop direction.So the direction of precast crack in specimen was parallel to pipe axial direction.For X80 pipe material, specimen thickness was 25, 22, 20, 18, 16 and 14, 12, 10, 8, 6mm respectively.The thickest X100 specimen was 22mm.There are 3 specimens for each set (grouped by thickness) of sample.Tests were carried out in accordance with the provisions of GB/T 21143-2007 [2] .The results are shown in Overall, fracture toughness of X100 is bigger than X80.Fracture toughness reaches to the biggest value when the specimen thickness is 18mm for X80 and 20mm for X100.After peak value, fracture toughness begins to gradually decrease with increase of wall thickness.Before reaching to the biggest value, fracture toughness value has a platform which is measured with smaller thickness specimens.
Fractographic Features Analysis
Fracture SEM photographs of X80 and X100 sample with different thickness are shown in Fig. 1 and Fig. 2. It is found that all fractures are ductile fracture.There are many dimples with different size and non-uniform distribution on the fracture section.So in the thickness range of all the tests, material has not yet to embody the characteristics of embrittlement, and the crack propagation region presents plastic state.A tongue shape crack propagation region existed in fracture section, which is shown in Fig. 3.In the figure, B is the thickness of the original sample, and A is the width of crack propagation in experiment process.Due to constraint effect in thickness direction, central sample material is in three-dimensional tensile stress state.So middle part of the sample begun to crack firstly.With the increase of external load, the region reaching to critical stress intensity factor in thickness direction became larger, but the crack speed in central part of specimen was the biggest.So the fracture section shows a tongue shape.
Fracture Toughness-thickness Empirical Model
The shape of crack tip plastic zone is shown in Fig. 4. For stress state in crack tip, the ratio of plastic zone to thickness is an important coefficient.If the values of plastic zone size and specimen thickness are in same order of magnitudes, which means B r p / tends to be 1, the plane stress state is dominant.In order to make sure most of the material in thickness direction is in stress state of plane strain, B r p / must be significantly less than 1, which means material near surface in plane stress state in thickness direction just be a small part through thickness section.Tests confirmed that when B r p / was about 0.025, then stress state when crack happens was typical plane strain [3] .Where r p is the radius of plastic zone size, ı ys is yield strength, K I is stress intensity factor.An obvious problem in this formula is that when crack tip stress exceeds the material yield strength, the exceeded load must be bore by material outside the hypothesis boundary.So the plastic zone will larger than the expression of Eq.( 1).In spite of great progress of working in determining true size and shape of plastic zone, but there is still no perfect theory which can give a satisfactory description to the shape of plastic zone.The biggest radius of plastic zone locates in specimen surface.From surface to specimen center, as crack tip stress field is changing to plane strain state from plane stress state, and the size of plastic zone is also gradually decreasing.For the specimen, the overall plastic zone is smaller than that decided by surface plastic zone radius.So if surface plastic zone is used as specimen plastic zone, it has already amended the size by a correction factor more than 1.
On other hand, the theory of K IC was established on the basis of elastic mechanics.So in situation of fitness for linear elastic fracture mechanics, which means plastic zone is much less than crack size, then plastic correction is not necessary.If plastic zone is not small compared with crack size, then expression of K based on elastic mechanics should be limited to apply.Plastic zone radius determined by linear elastic mechanics and von Mises yield criterion is used to characterize the plastic zone in plane stress state in this paper.As mentioned above, when 025 .0 B / p r , then the specimen is in stress state of plane strain.
That means the plastic zone size radius is only 1/40 of the specimen thickness.In this case, the plastic zone is very small compared with specimen size and Eq.( 1) can be used.
In case of 025 .0 B / p r , plane strain fracture toughness K IC is substituted for K I in Eq.( 1), then the critical thickness of plane strain can be expressed as Where B IC is the critical thickness of plane strain state.The effect of thickness on fracture toughness was analyzed by Anderson [4] based on existing experimental data.In his study, it was reasonable that K IC linearly decreased with thickness increase.In other studies [5] , the relationship between K C and thickness was expressed as Where [ and k were material constants, t was thickness.This equation is applicable to material of TC4 titanium alloy.
Through calculation of Eq.( 3), point of fracture toughness-thickness can be got as shown in Fig. 5. Linear pattern was used to fit the points between maximum value of K C to value close to plane strain fracture toughness, and the fitting curve is shown in Fig. 5.It shows that linear fitting is suitable for this curve segment as the value of correlation coefficient is 0.998.
But according to the Eq. ( 3), K C would increase from 0 with thickness increase before reaching to the maximum value.While according to the test data shown in Table 1, the fracture toughness of pipeline steel K C has not decreased with wall thickness increase before reaching to the maximum value, and there is a platform in the curve.Some material would accord with this law mentioned in the literature [3].So for pipeline steel, the error is large when fitting the relationship between fracture toughness and thickness.
Experiment data of X80 pipeline steel after fracture toughness reached to the maximum value (data of thickness larger than 18mm) were used and fitted by linear pattern, the linear relationship between fracture toughness and specimen thickness can be expressed as As shown in Fig. 6, it can be seen that when thickness is bigger than 14mm, the linear relationship can be obtained between ratio of crack width to thickness and thickness itself.This ratio increases as thickness increases.When thickness is less than 14mm, no obvious relationship can be found.The possible reason is that in thin specimen test, proportion of plane strain crack is small.In thin specimen crack, Plane stress crack dominates the test, so the crack extension law is different from that of plane strain.
Test results of thickness more than 16mm were used, the linear relationships for both X80 and X100 were obtained between crack width to thickness ratio and thickness.Taking the width-to-thickness ratio up to 100% as criterion of plane strain, it is found that the critical thickness of X80 is 45mm, and that of X100 is 38.7mm.For X80 it can be seen that the difference of plane strain critical thickness calculated from our model and from width-to-thickness ratio model in literature [7] is very little(4.7mm).But for X100, the difference is very large.The possible reason is that in our tests, only one point was obtained after peak point in fracture toughness-thickness curve.So the error will be big if the test data is limited when using our model to estimate the plane stress fracture toughness or critical thickness.
Linear fitting was adopted for both the above two kinds of material.But in case of thickness increase, especially the thickness is nearly plane strain condition, the fracture toughness and specimen thickness has not shown a simple linear relationship.Because even if the specimen is thick enough, it cannot completely reach ideal plane strain state, so the ratio of A/B can only be infinitely close to 1, but not to be 1.So calculated critical thickness will be less than actual value, but it can infinitely close to it.From the test results, it can be confirmed that this representation is reasonable, and the model proposed in this paper has been validated.
Conclusions and Suggestions
In the second west to east gas pipeline project, the thickness of bends and fittings is more than 50mm.From analysis above, the possibility of the brittle fracture does exist.So the safety of thick wall pipelines is worth of attention.On the other hand, because wall thickness of general X80 pipe is usually not so thick, and the critical thickness of plane strain fracture toughness is nearly 50mm.So characteristics of the plane strain can't appear in actual X80 pipeline and the crack resistance is not determined by K IC .In such cases, the actual fracture toughness is usually higher than K IC , which means that if K IC is used to estimate critical crack
Fig. 3
Fig.3 Tongue shape in fracture section of fracture toughness specimen.
Fig. 4 T
Fig.4 Plastic zone size in crack tip.Based on the elastic solution of plane stress state, crack tip plastic zone radius 0 T
Fig. 6
Fig.6Relationship of crack width-to-thickness ratio and specimen thickness.
Table 2 ,
in which test results is average value of three specimens. | 2018-12-19T19:32:39.775Z | 2016-01-01T00:00:00.000 | {
"year": 2016,
"sha1": "eef00541b17a67860f57e10cbb7c76749a6bc497",
"oa_license": "CCBY",
"oa_url": "https://www.matec-conferences.org/articles/matecconf/pdf/2016/30/matecconf_smae2016_03016.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "eef00541b17a67860f57e10cbb7c76749a6bc497",
"s2fieldsofstudy": [
"Engineering",
"Materials Science"
],
"extfieldsofstudy": [
"Materials Science"
]
} |
73563891 | pes2o/s2orc | v3-fos-license | Relationship between Abundance of Risk Factors and Prevalence of Asthma and Allergic Disease at South of Cairo Publication History :
Background: Many studies have focused on environmental exposures among inner city subject with asthma, allergen exposure potentially leading to a higher prevalence of asthma and it has been suggested as among causes. Objectives: In this study we investigate the relation between risk factors (exposure to air pollutions, family history and exposure to family pets or far animals), and asthma and other allergic disease exacerbation, and indicate the severity of like this diseases. Methods: It is a cross-sectional study on children was conducted in 2014 in Al Oba zone in public and private schools. A sample of 315 children participated, where standardized questionnaires of ISAAC protocol phase 1 written were distributed among 11-14 years old children completed the questionnaires. Results: Sensitization to exposure allergens was associated to exacerbation to asthma symptoms and provokes allergic disease. Family history, nasal allergy, and exposure or living nearby factories, coffee shops and animal were a significantly difference (level of significance was taken at p value < 0.05). Conclusion: Environmental and host factors may influence the development and expression of asthma and allergic diseases. Relationship between Abundance of Risk Factors and Prevalence of Asthma and Allergic Disease at South of Cairo Publication History: Received: June 16, 2015 Accepted: August 29, 2015 Published: Septmber 01, 2015
Introduction
Allergic disease become noticeable to be rise worldwide, and although an allergic family history is one of the strongest and important risk factors for childhood allergy [1], many international studies [2].It focus on geographical differences in allergy prevalence, and suggest that environmental influences also play a causal role.Although pets are known to aggravate asthma, allergic rhinitis, and eczema in sensitized individuals [3], controversy remains about whether early life pet exposure is a risk factor or a protective factor in their development.Current guidelines issued in Australia [4], the United States [5], the United Kingdom, and by the Global Initiative for Asthma [6] all agree that there is currently insufficient evidence to provide any recommendations in relation to pet-keeping in early life, the development of asthma and allergic disease because systematic reviews [7], and a meta-analysis have reached different conclusions.Early reviews found pet-keeping increased the risk of sensitization and allergic disease [7] with later reviews finding no effect.A recent meta-analysis [8] reported less risk of childhood asthma associated with cats, but increased risk with dogs.Environmental exposures can either increase or decrease the risk of subsequent allergic disease development [9].
The higher prevalence of food sensitization and food allergy risk categories in participants with asthma, hay fever, and eczema is consistent with other national studies [13].The link between food allergy and asthma appeared especially strong.There was increased prevalence of all food sensitization and food allergy risk categories in those with diagnosed asthma, as well as increased prevalence and likelihood of FA, those with current asthma [14].
The relationship between FA and asthma severity may or may not be causal in nature.In such as, severe asthma may be associated with greater atopy in general [14].Moreover, the relationship between FA and asthma severity could be causal, noting that symptoms are induced by foods in people with food allergy up to 30% of the time, although usually with other allergy symptoms [15].However, FA has been recently found to be a major risk factor for severe asthma and life threatening asthma episodes [14].
Several epidemiologic studies have shown that exposure to environmental factors like tobacco smoking or living in urban areas of industrialized regions, particularly among children living along busy roads and trucking routes are associated with poor respiratory health in children, adversely affect lung function and may increase the risk factor of development of bronchial asthma [16,17].
Design and setting:
A cross-sectional study conducted in a random sample.ISAAC phase one (International Study of Asthma and Allergies in Childhood) was used in this study on students in the most polluted area at South of Cairo (Al Oba zone).
•
Number of students with asthma like symptoms = 142 • Number of students without asthma like symptoms = 166 Inclusion criteria: All children are eligible to be included in this study.
Exclusion criteria:
No exclusion criteria accept a questionnaire that is not completed or if a questionnaire contains double answer for example the students answer by yes and no in the same question or refused to participate accordingly.ISAAC questionnaire was adapted, translated into Arabic and it used to be included socio-demographic data and environmental exposure factors.
Statistics
All data in this study are collected, coded, tabulated and statistically analyzed using SPSS program for windows version 14 (SPSS Inc, Chicago II).Comparative studies were done for numerical data using student's t-test and were done for categorical data by Chi-square test for comparison between variable, and it was done to estimate the relation between risk factors and development of asthma.
The level of significance was taken at p value < 0.05.
Asthma symptoms:
In this study we estimated asthma like symptoms on basis of positive answers to written question (have you had wheezing or whistling in the chest in last 12 months?).
Allergic Rhinitis: In this study, allergic rhinitis was considered on the basis of positive answer to written question (have you had repeated attacks of sneezing, runny or blocked or itching in the nose when you did not have a cold of flu?).
Atopic Dermatitis: In the study, atopic dermatitis was considered on the basis of positive answers to written question (have you had repeated attacks of itchy rash ?).
Results
315 of children with mean age of 12.5 years were evaluated for asthma, allergic rhinitis and atopic eczema prevalence and their associated factors.Frequency distribution of children according to presence or absence of asthma symptoms were 142 and 166 respectively (Table 1).
:it shows the distribution of children with and without asthma symptoms as regard to presence and absence of allergy, where this table showed a statistical significant higher % of children with allergy in children with asthma symptoms compared to those with no asthma symptoms.Nasal allergy was the most predominant form of allergy ( P value = 0.05).
According to environmental exposure of children, higher percentage of children with asthma symptoms lives nearby factories, nearby coffee shop and higher percentage of them living nearby or exposed to pets as (table 2) showing below ( P value = 0.05).
There is a statistical significant different between children with and without asthma symptoms as regard family history.As higher percent of children with asthma symptoms has positive family history (Figure 1).
There was a significant association between risk factors like environmental factors or host factors and development of asthma symptoms and other allergic diseases like eczema or allergic rhinitis.
Discussion
This study was carried out to assess the risk factors and prevalence of asthma like symptoms and other allergic diseases like allergic rhinitis, and atopic dermatitis among Egyptian school children in the south of Cairo (Al Oba zone) on aged group range 11-14 years.
Exposure to allergens were significantly associated with sensitization of children with asthma symptoms rather than children without asthma symptoms (63.4% and 22.3% respectively, P value = < 0.001*).Nasal allergy was significant between them, whereas, there are not a statistical significant as regard to skin and food allergy (P value = 0.232 and 0.953 respectively).
This study found that the relation among exposure to nearby factories, nearby coffee shops and development of asthma symptoms were 26.1% and 65.5% respectively, and these percentage were a significant difference of children with asthma symptoms compared to others without asthma symptoms ( P value = 0.002* and 0.006*).Moreover, pets exposure are a significant difference between children with or without asthma symptoms (P value= < 0.001*).[18] found the prevalence of asthma symptoms was significantly more common in families with animal in their home compared to others with no exposure to animal.
In this study we followed the family history of the studied sample as regards atopy, where we found that 47.9% of children with asthma symptoms had positive a family history of atopic diseases compared to 52.1% with had negative a family history.[19], found that the presence of asthma cases aggregation among related individual may indicate either shared genes or a common household environment.
Figure 1 :
Figure 1 : comparison between children with and without asthma symptoms as regards family history.
Table 1 :
Al Dhduh MAA, Mohamed Sabri NA, Fouda EM (2015) Relationship between Abundance of Risk Factors and Prevalence of Asthma and Allergic disease at South of Cairo .Int J Pharma Sci Res 2: 110.doi: http://dx.doi.org/10.15344/2394-1502/2015/110Frequency distribution of children with asthma symptoms according to presence or absence of allergy. Citation:
Table 2 :
Frequency distribution of children with and without asthma symptoms as regards environmental pollution. | 2018-12-29T13:15:02.142Z | 2015-09-01T00:00:00.000 | {
"year": 2015,
"sha1": "dd798886877b3b026bbdf5e3eda216c4484c03a6",
"oa_license": "CCBY",
"oa_url": "https://www.graphyonline.com/archives/IJPSR/2015/IJPSR-110/article.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "dd798886877b3b026bbdf5e3eda216c4484c03a6",
"s2fieldsofstudy": [
"Environmental Science",
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
15552165 | pes2o/s2orc | v3-fos-license | Automated classification of coronary artery disease using discrete wavelet transform and back propagation neural network
An automated classification of coronary artery disease using discrete wavelet watershed transform and back propagation neural network has been proposed which basically segments the blood vessels of the coronary angiogram image as a first step, which in turn involves various stages such as preprocessing, image enhancement, and segmentation using discrete wavelet transform and watershed transform along with morphological operations. Pre-processing is done to remove the noise using the bicubic interpolation method followed by Daubechies 4 discrete wavelet transform and Weiner filtering. Further, image enhancement is done to improve the quality of the image using the histogram equalization technique. Auto thresholding is done to segment the edges of the blood vessel accurately and efficiently using distance and watershed transforms followed by normalization and median filtering. Finally, morphological operations are performed to remove the noise due to segmentation. Features such as area, mean, standard deviation, variance, brightness, diameter, smoothness, compactness, skewness, kurtosis, eccentricity and circularity are extracted from the segmented coronary blood vessel to train the neural network using back propagation network. Thus, the system is able to achieve 93.75% normal classification and 83.33% abnormal classification. Also, 90% efficiency is achieved in classifying Type 1 and 92% efficiency is achieved in classifying Type 2 stenosis at a learning rate of 0.7 and Type 1 classification efficiency of 85% and Type 2 classification of 89% has been achieved for 50 hidden units of the neural network.
INTRODUCTION
Medical images such as coronary angiogram images account for a large portion of noise.There is a real challenge to segment such blood vessels.Angiography is a procedure to observe the blood vessels of a human being and further investigation is carried with the help of angiograms which detects the edges of the blood vessel.
Blood vessel detection is an important diagnostic concept of image segmentation, which detects the edges of the angiographic blood vessels (Albert, 2006;Rice et al., 2000;Bouchet et al., 2007;Cemil and Francis, 2004;Nain et al., 2004;Lacoste et al., 2006;Brieva et al., 2005;Suri et al., 2002;Nyual et al., 2000;Lorigo et al., 2001;Lo et al., 2006;Sarry et al., 2001;Hassouna et al., 2003;Wink et al., 2004;Saha et al., 2000;Rafael et al., 2002;Chan et al., 2000;Espin et al., 2004;Zhou et al., 2006).Jean et al. (2008) have used the application of minimal surfaces and Markov random fields as models and applied to the region adjacency graph of the watershed transform to segment the liver tumors.The researcher attempted to segment the relevant tumors from the liver image.Although the researcher did not segment the blood vessels of the angiogram image, but it is understood that unsupervised watershed transform along with Markov model is applied.Jayadevappa et al. (2009) developed a hybrid segmentation model based on Watershed and Gradient Vector flow (GVF) for the detection of the brain tumor.Generally the GVF suffers from a very high computational requirements and sensitive to noise.These are overcome by the integrated method which makes use of the watershed algorithm.Ning et al. (2007) proposed a new algorithm which combines the watershed transform (Nassir, 2005) and level set method in order to extract the accurate boundary of the vessels.The researcher demonstrated the cost time which mainly depends on the number of pre segmented regions.However, performance of the system could be improved only if some prior information and distinguishable features are included and the researcher did not carry any performance evaluation.Pinaki and Dibyendu (2012) proposed an easy and simple method to overcome over segmentation by using the distance transforms and image smoothing of the morphological techniques (Sun et al., 2007;Deng and Heijmans, 2002;Tsair et al., 2009Tsair et al., , 2003;;Wong et al., 2005;Weili et al., 2010) along with watershed for segmentation.Plot with the number of pixels before and after smoothing was done to prove the performance of the system and testing was performed only on color images and not on medical images which is one of the limitations.Ishita and Monisha (2012) proposed a new method for brain tumor segmentation using watershed (Jean et al., 2008) and edge detection algorithm in HSV color model.Initially the RGB image is converted into HSV color image and watershed algorithm is applied to each region after contrast enhancement which is then followed by canny edge detection.Finally all the three images are combined to get the segmented image.However, the performance of the system was not evaluated using any parameter.Yugander and Sheshagiri (2012) developed an improved watershed algorithm and modified level set method.The over and under segmentation problems were overcome by using dual tree complex wavelets and modified watershed based on Wasserstein distance.This method was used to extract the finger print from the original image and not for segmentation of the blood vessels and the performance of the system was not evaluated.Also, the system involves complex wavelets which itself is complex.Qing et al. (2004) and Li et al. (2006) evaluated the performance of the watershed algorithm by analyzing the binary images and comparing them against different distance transforms.The researcher concluded that the Watershed algorithm is very effective for grey level segmentation of medical images.The researcher concluded that the chessboard distance transform performed well as against the Euclidean distance.Ghassan et al. (2009) pointed out that although watershed transformation is generally used for segmentation it is limited due to over segmentation and sensitivity to noise.However, these drawbacks were overcome by enhancing the prior shape and appearance knowledge.The problem of over segmentation was overcome by using clustering namely k-means segmentation and noise is removed or suppressed by computing the mean intensity of each segment.However, the researcher did not evaluate the performance of the system proposed and it involves complex algorithm.Nassir et al. (2006) has combined K-means clustering, watershed and difference in strength (DIS) for segmentation.Although the proposed method overcomes the over segmentation there is no valid performance evaluation done.Zulong and Kaiqiong (2010) used contour based segmentation which in turn uses the morphological operations (Qing et al., 2004;Chan et al., 2000;Shoujun et al., 2009;Sidahmed et al., 2013) were used to detect the edges rather than using the gradient or gray level intensity, whereby the problems due to inhomogeneity is overcome.However, the researcher did not carry any coronary quantitative or statistical analysis to prove the efficiency of the system (Hernandez et al., 2000;Morteza et al., 2011;Schrjver et al., 2002;Pascal et al., 2006;Yan et al., 2005;Kobashi et al., 2000;Mukhopadhyay et al., 2003;Zhou et al., 2008;Chan and Vese 2001;Law et al., 2001;Yu, 2002).
Chih-Yang and Yu-Tai (2005) introduced the extraction of coronary angiogram blood vessels from the digital angiographic images.The researchers used three important steps to segment or extract the coronary angiogram blood vessel, they are namely, background elimination and noise removal, blood vessel enhancement and blood vessel segmentation from the coronary angiogram image.In the first stage, the researcher applied a temporal fourier transform followed by a high pass temporal filtering, which allows only the low-frequency terms, followed by inverse fourier transform.The transformed image so formed will not have any unwanted background, however, contains some spikes or noise, which may be due to isolated noise which cannot be detected and removed by using the fourier transform as it is based on the pure frequency domain.Hence to remove the isolated noise Discrete Wavelet Transform (DWT) (Ali et al., 2009;Elly, 2011;Jorge et al., 2003;Yu, 2002;Zulong et al., 2010;Tang et al., 2006) is applied.A three level decomposition was employed and the LLL sub-band region will consist of the features pertaining to the blood vessel.Thresholding is applied to preserve the edge information and finally Inverse Discrete Wavelet Transform (IDWT) is applied.
In the second stage, blood vessel enhancement is done using 72 matched filters (Hoover et al., 2000) and is projected onto the xy plane, followed by blood vessel segmentation, based on clustering analysis using a stencil mask.Further, histogram analysis is carried with thresholding and 18-adjacency clustering to segment the coronary angiogram blood vessel.The method used is very complicated involving two transforms, 72 matched filters, masking and thresholding.Also, the overall execution time was less than 3 min.However, the researcher did not carry any coronary quantitative or statistical analysis to prove the efficiency of the system.Wenwei et al. (2010) proposed new segmentation method which is based on the transition region extraction (Liang et al., 2001).The researcher used 6 Gaussian matched templates to basically enhance the input coronary angiogram image followed by local complexity method.Segmented image was obtained by applying thresholding to the histogram of the transition region (Yao et al., 2008;Rivest, 2004;Socher et al., 2008;Shoujun et al., 2009) which outperformed the top-hat method.However, the researcher neither did not carry any coronary quantitative or statistical analysis to prove the efficiency of the system nor computed the overall execution time.Santhiyakumari and Madheswaran (2010) have proposed a method to categorize the carotid artery subjects into normal and diseased subjects namely, cerebrovascular and cardiovascular diseases.For each and every pre-processed ultrasound carotid artery image, contours are extracted using contour extraction techniques.Multilayer Back Propagation Network (MBPN) system has been developed for categorizing the carotid artery subjects.The obtained results showed that MBPN system provides higher classification efficiency, with minimum training and testing time.Sidahmed et al. (2013) have proposed an algorithm to produce a 85.5% classification accuracy in the diagnosis of Coronary Artery Disease (CAD), in which Genetic Algoirthm (GA) generates in each iteration a subset of attributes that will be evaluated using the Bayes Naïve (BN) based feature selection in the second step of the selection procedure.Thus, the assest of the algorithm is then compared with the Support Vector Machine (SVM), Multi-Layer Perceptron (MLP) and C4.5 decision tree algorithm.The results of classification accuracy for those algorithms are 83.5, 83.16 and 80.85% respectively.Brieva et al. (2005) evaluated four segmentation algorithms for coronary angiogram images.The four algorithms were namely wavelets, snakes, level sets and dynamic threshold (Masoomeh et al., 2009;Mohammed et al., 2011).Here, a multi-resolution wavelet method was employed which consists of filter banks in turn consisting of one dimensional wavelet functions.A set of five filters are applied and segmented based on thresholding.The researcher used mean specificity and mean sensitivity for evaluating the performance of the system.However, the researcher did not carry any other coronary quantitative or statistical analysis to prove the efficiency of the system nor computed the overall execution time.Thus, the system looks complicated with filter banks and set of five filters.
PROPOSED METHOD
From the literature review discussed in introduction, it is clearly evident that the limitations of the watershed algorithm are overcome by enhancing various other techniques from the literature.Also, it is found that the watershed algorithm along with some other techniques is used for the segmentation of medical images such as liver, retina and brain, but is not done for the angiographic images of the heart.In the literature, the researchers have used wavelet discrete transform for suppressing noise and to get more information on the edges of the medical image for segmentation.Thus, the proposed method integrates the watershed algorithm and discrete wavelet transform (Tsair et al., 2009(Tsair et al., , 2003) ) which overcomes the limitations of the watershed algorithm along with the morphological operations and drawbacks mentioned in the introduction.Furthermore, it automatically classifies the segmented image into stenosis Type 1 or 2 CAD using the back propagation neural network.
Figure 1 In this step, first the given coronary angiogram image is resized to 256 x 256.Then, pre-processing is done using the following steps as shown in Figure 2 to produce sharper images of the input coronary angiogram image.Also, it removes noise acquired in the angiogram image.
Bicubic interpolation method
In order to produce sharper input coronary angiogram images, the bicubic interpolation method is adopted in the pre-processing stage.In the bicubic interpolation method the output pixel is obtained as a weighted average of the pixels in the nearest 4-by-4 neighbourhood followed by noise removal and background elimination.
Discrete wavelet transform
Wavelet transform is taken to extract the features initially.After pre processing, the coronary angiogram images do contain some noise or spikes, which are due to the presence of the isolated noise.These noises are usually not detected and removed by the frequency domain filters or fourier transforms.Hence a wavelet transform is applied after pre-processing which is done to remove the noise.A wavelet transform normally consists of a pair of high pass and low pass filters as shown in Figure 3, which are used to decompose the given coronary angiogram image into the low frequency and high frequency components respectively, according to the different number of levels.The decomposed coronary angiogram image is further down (D) sampled by 2 to get the approximation coefficients (cA) and detailed coefficients (cD), which are said to be called as the DWT wavelet coefficients.However, the image can be reconstructed again using the Inverse Discrete Wavelet Transform (IDWT), which is so called as the reconstruction process.Normally, the input coronary angiogram image is divided into four regions, namely: (a) LL -This sub band is obtained by using two low-pass filters, (b) LH and HL -These two sub bands are obtained by using one low-pass filter and one high-pass filter, and (c) HH -This sub band is obtained by using two high-pass filters.
After the first level of decomposition, the LL sub band is decomposed again using the same pair of low-pass filter and highpass filter to perform n-stage discrete wavelet transform.Thus scale based decomposition is obtained by the wavelet transform in which the noise is represented by the finer scaled wavelet coefficients.The LL sub band does not contain any noise while all other sub bands do contain noise.However, the coefficients of such scales represent the edge information, which must be maintained by selecting a threshold to remove the noise.Thus, the unwanted noise is removed while the important local features of the coronary angiogram blood vessel is retained that is, the edge information is preserved.
In the proposed integrated method, the Daubechies 4 discrete wavelet transform (Elly, 2011) is applied with periodization to the coronary angiogram image.It produces the smallest length wavelet decomposition and the same mode of periodization is applied for IDWT to ensure perfect reconstruction.The Daubechies 4 wavelet transform is comparatively distinct from Haar wavelets because the scaling signals and wavelets produced are due to average's and differences from the signals.It conserves the signals energy and redistributes in a compact form.
Weiner filtering
Usually interpolation tends to increase the mean square error and hence in order to minimize it, the Weiner filtering approach is employed.It performs very well on the edges of the given coronary angiogram images, which is applied before taking the IDWT and after the DWT.Thus Weiner filtering is used to detect the edges of the given coronary angiogram blood vessel.This produces the edge enhanced image, which is sharper than the original input coronary angiogram image in which more details can be seen clearly.Thus, the Daubechies 4 wavelet transform is used to preserve the edge and detailed visibility information, which is the
(iii) Image segmentation
Image segmentation is done after image enhancement followed by applying watershed transform and auto thresholding.The detailed algorithm implemented is shown in Figure 5.
Auto thresholding algorithm
Step 1: Histogram is performed to determine the threshold value in order to segregate the coronary angiogram blood vessel from the background.
Step 2: Obtain the texture colored image.
Step 3: Compute the true Euclidean distance transform of the binary image.The distance transform assigns a number to each pixel of the given image which is the distance between that pixel and the nearest non-zero pixel of the given image.
The Euclidean distance (Qing et al., 2004) between two pixels (x 1 , y1) and (x 2 , y2) is defined as in Equation (1), Step 4: Next compute the distance transform of the complement of the binary image.
Step 5: Compute the distance transform and force the pixels which do not belong to the edges to be at infinity.
Step 6: Compute the watershed transform and display the resulting label matrix as an RGB image.
Assuming that the image f is an element of the space C(D) of a connected domain D, then the topographical distance between points x and y is given as in Equation ( 2), Where, 'inf ᵞ ' is the overall paths inside D. Thus, let f ε C(D) have a minima {m k } kεI , for some index set I. The catchment basin CB(m i ) of a minimum m i is defined as the set of points C ε D, which are topographically closer to m i than to any other regional minimum m j and is given by Equation (3).
Thus, the watershed (Javadevappa et al., 2009) is the set of points which do not belong to any of the catchment basin as shown in Equation ( 4 Step 7: Perform normalization and median filtering.Preform image subtraction to eliminate the background of the given image.This is done by subtracting each element in the array y from the corresponding element in an array x of the given array(x,y) and then the difference between them is returned to the output array z.
Step 8: Perform intensity adjustment.This performs a mapping between the input image and output image based on intensity.Intensity values below Lin are mapped to Lout, intensity values above Hin are mapped to Hout, and values between Lin and Hin are mapped to values between Lout and Hout.
Step 9: Finally, auto thresholding is performed using the ISODATA method.
ISODATA method is used to perform auto thresholding which uses iterative technique.In this method, a global threshold is computed which is used to convert the intensity image to a binary image and is a normalized intensity value that lies in the range [0, 1].This iterative technique was developed by Ridler and Calvard (1978).The histogram is initially segmented into two parts using a starting threshold value such as zero equal to half the maximum dynamic range.The sample mean (mf, 0) of the gray values associated with the foreground pixels and the sample mean (mb, 0) of the gray values associated with the background pixels are computed.Thus, a new threshold value 1 is now computed as the average of these two sample means.This process is repeated based upon the new threshold until the threshold value does not change any more.
Isodata algorithm
Step 1: Compute the mean intensity of the given image from the histogram H.
Step 2: Compute the Mean above H (MAH) and Mean below H (MBH) using H from step 1.
Morphological operations
Finally morphological operations are performed to segment the detected edges of the coronary angiogram image.First remove the interior pixels by setting a pixel to '0', if all its 4-connected neighbours are 1, thus leaving only the boundary pixels ON.Then remove the isolated pixels such as the individual 1's surrounded by 0's and also remove the H-connected pixels as these do not contribute to the edges of the coronary angiogram blood vessel.Generally a median filter is employed when the goal is to simultaneously reduce noise and preserve edges.Here each output pixel contains the median value in the m by n neighbourhood around the corresponding pixel in the input image.Thus, the blood vessels of the given coronary angiogram image is segmented using the integrated method of Discrete Wavelet transform and Watershed Transform (DWWSHD) along with the morphological operations.
(iv) Feature extraction
After segmentation feature extraction is done.From the segmented coronary angiogram image, features such as area, mean, standard deviation, variance, brightness, diameter, smoothness, compactness, skewness, kurtosis, eccentricity and circularity are extracted to train the neural network
(v) Classification of CAD using BPN
The classification of CAD using neural network basically consists of the following steps: Step 1: Extracted feature transformation Step 2: Network architecture definition Step 3: Learning algorithm Step 4: Validation In the first step, the extracted features of section D are given to the neural network for recognition of particular patterns.In the second step, the number of neurons in each layer, and the number of hidden layers are defined.Also, the connectivity between each layer is defined.Usually, the number of input neurons, hidden neurons and output neurons depends on the problem studied.The network may not have required degrees of freedom to learn the process correctly, if the numbers of hidden neurons are small and vice versa.
In the third step, a leaning algorithm is used to train the network to respond correctly to a given set of inputs.Normally, the neural network is said to be well trained when there are more input data's available.It is possible to determine the weights through training, in which initially the weights are assigned to be random or based on experience.Thus, in the process of the learning algorithm, the network is said to change the weights systematically in order to perform the desired input output relation properly.In the fourth step, the validation is done in order to evaluate the performance of the system trained.The back propagation algorithm is given as below for the classification of the type of coronary artery disease.
Step 1: Read the number of layers and number of nodes in each layer Step 2: Initialize the weights Step 3: Read the pattern Step 4: Calculate the number of output nodes in each successive layer Step 5: Calculate the mean square error for a pattern Step 6: Calculate the momentum of a node in the output node Step 7: Update the weights between layers Step 8: If the weights are not updated for all the layers, then calculate the momentum of a node in the hidden layer and repeat step 7 Step 9: Then calculate the mean square error for all the patterns Step 10: If not go to step 3 The learning algorithm depends on three main parameters, namely, the learning rate, momentum, and the mean square error.The learning rate usually determines the change in the size of the weight, while momentum causes the weight changes to be dependent on more than one input pattern.Mean square error is the value at which the network error dips below a particular error threshold.
The two types of the CAD diseases are defined as Type 1 and 2 diseases.Type 1 disease is called eccentricity, in which the atherosclerotic plaque is not distributed along the entire circumference, leaving a variable arc of disease free wall.Type 2 disease is called concentric, in which the atherosclerotic plaque is distributed along the entire circumference of the internal elastic membrane.
EXPERIMENTAL RESULTS
The proposed algorithm was tested on the images acquired from 50 patients.The proposed algorithm was implemented and tested in MATLAB 7.0 on a Pentium IV Personal Computer (with Central Processing Unit 2.8G and 512 memory).The performance of the automated classification of CAD using back propagation network is evaluated using the following parameters: (a) Confusion matrix (b) Learning rate (c) Hidden units Testing of the back propagation neural network is done after the training phase.The data that are not trained by the netwrok, is applied to the neural network to evaluate its performance.Table 1 shows the confusion matrix for the result classification between normal and abnormal for this neural network.It is observed that two normal samples are classified incorrectly by the neural network as the subjects do not have either Type 1 or 2 coronary artery disease.Also, three abnormal samples are incorrectly classified as normal although the subjects do have coroanry artery disease.The confusion matrix can also be expressed as percentage of correctness as shown in Table 2.
From Table 2, it could is observed that the according to the percentage of correctness, the normal samples are classified correctly by the neural network with 93.75% of correctness and incorrectness by 6.26%.Persons who are having coronary artery disease are correctly classified with 83.33% while incorrectly classified with 16.67%.Thus, accuracy of classifying normal persons and abnormal persons are 93.75 and 83.33%.The performance of the back propagation neural network for various learning rates ranging between 0.1 to 1.0 is observed as shown in Table 3.The training is done with 50 coronary angiogram images.From Table 3, it is observed that the average training time ranges from 2.36 to 1.98 s and the average testing time ranges from 0.0142 to 0.0088 s for the learning rate varying from 0.1 to 1.0 in steps of 0.1.Also, the Type 1 classification efficiency is 90% and Type 2 classification efficiency is 92% at a learning rate of 0.7 which is said to be the optimum learning rate.The corresponding average traininng time is 2.18 s and average testing time is 0.0105 s.
From Table 4, it is observed that the objective of the network is not met upto hidden units of 30, beyond which it is achieved.The maximum classification efficiency for Type 1 and type 2 are 85 and 89% at 50 hidden units with an average training time of 1.93 s and average testing time of 0.0156 s.Thus, the application is said to have one hidden layer with fifty hidden units.From Table 5 it can be seen that the accuracy of classifying the CAD using DWWT and BPN automatically by the proposed method is 89%, which is predominant as compared to the other methods specified in the literature.
SIMULATED RESULTS
The simulated results are shown in Figures 6 to 7 for both the proposed automated classification of CAD using BPN.Only two input images and its superimposed output coronary angiogram images along with its classification of CAD type are shown for simplicity.Figure 6 shows the input coronary angiogram image which is tested for the Type 1 CAD. Figure 7 shows the region properties stages, Figures 8 and 9 shows the features extracted and the extracted blood vessel.Figure 10 shows the training of the neural network and Figure 11 shows the testing phase, while Figure 12 shows the superimposed image and classified Type 1 CAD by Figure 13. Figure 14 shows the input coronary angiogram image which is tested for the Type 2 CAD along with the extracted blood vessel.
Figure 15 shows the training of the neural network and Figure 16 shows the superimposed image and classified Type 2 CAD by Figure 17.
Conclusion
It is concluded that the automated classification of coronary artery disease using the back propagation neural network has yielded a 93.75% normal classification and a 83.33% abnormal classification by integrating the discrete wavelet transform and watershed algorithm along with morphological operations.Also, Type 1 classification efficiency of 90% and Type 2 classification of 92% has been achieved at a learning rate of 0.7 and Type 1 classification efficiency of 85% and Type 2 classification of 89% has been achieved for 50 hidden units of the neural network.An accuracy of 89% in classifying the CAD using DWWT and BPN automatically is achieved, which is predominant as compared to the other methods specified in the literature.
shows the flow diagram of the automated classification of the integrated segmented blood vessel.The proposed automated classification and integrated coronary angiogram image segmentation algorithm using the Discrete Wavelet transform and Watershed Transform (DWWSHD) consists of five major steps, namely: Step 1: Image pre-processing Step 2: Image enhancement Step 3: Image segmentation Step 4: Feature extraction Step 5: Classification of CAD (i) Image pre-processing
The DWT is applied to divide the given Coronary Angiogram Image (CAI) into four bands, the Low Low [LL], Low High [LH] , High Low [HL] and High High [HH] band.The DWT decomposes the given image based on the scale wavelet coefficients, which represent the noisy signals of the image contained in all the bands except the LL, which retains the features or the edge information of the coronary angiogram image.
(a) Perform the DWT of the given input coronary angiogram image.(b) Preserve the edge information of the LL band by applying Weiner filtering.(c) Perform IDWT of the modified image from step b.
Table 1 .
Confusion matrix of the back propagation neural network.
Table 2 .
Percentage of Correctness for the back propagation neural network.
Table 3 .
Average training time, average testing time and classification effeiciency for different learning rate.
Table 4 .
Average training time, average testing time and classification effeiciency for different hidden units.
Table 5 .
Percentage of correctness for the back propagation neural network. | 2016-01-15T18:20:01.362Z | 2014-05-30T00:00:00.000 | {
"year": 2014,
"sha1": "09bd95fa02bb81e6cbc169045a13d8317d1c5454",
"oa_license": "CCBY",
"oa_url": "https://academicjournals.org/journal/SRE/article-full-text-pdf/6E2A6B844935.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "09bd95fa02bb81e6cbc169045a13d8317d1c5454",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
55689084 | pes2o/s2orc | v3-fos-license | The Role of Procalcitonin as a Biomarker in Sepsis
Sepsis and its complications are one of the leading causes of mortality. Timely diagnosis and treatment is highly important in reducing the morbidity and mortality. Serum biomarkers may aid in the early diagnosis of sepsis and therapeutic intervention. Procalcitonin (PCT) is a peptide precursor of the hormone calcitonin and its primary trigger is infection. PCT is identified as part of the complex pro-inflammatory response of the innate immune system. PCT is widely reported as a useful biochemical marker to differentiate sepsis from other non-infectious causes. Serum PCT levels are elevated in patients with bacterial infections. The diagnosis of infection in critically sick patients is challenging as the current biomarkers are non specific. Our review showed that PCT is a more accurate diagnostic parameter for sepsis and a better predictor of mortality. PCT is a more reliable marker than other biomarkers including C-reactive protein, Interleukins and lactate levels. PCT has been proved to be superior biomarker, however its use still has to be interpreted in the context of clinical presentation. Further study on the role of PCT is needed for more effective and targeted approach in sepsis.
general, sepsis occurs in approximately 2% of all hospitalizations in developed countries [3].Severe sepsis is a common, frequently fatal, and expensive condition.Epidemiological studies indicate an incidence of severe sepsis approximately 751,000 sepsis cases per year in the United States (3.0 cases per 1,000 population and 2.26 cases per 100 hospital discharges) [4].
Mortality rate in sepsis remains high despite the current advances in medical science, technology and practice.Timely diagnosis and treatment is highly important in reducing the morbidity and mortality associated with sepsis.At-times the diagnostic uncertainty still remains high despite the available clinical information.Thus, a laboratory test with more specificity is essential.Serum biomarkers like procalcitonin may aid in the early diagnosis of sepsis and therapeutic intervention.
Procalcitonin
Procalcitonin (PCT) is a peptide precursor of the hormone calcitonin, the latter being involved with calcium homeostasis.This 116-amino acid prohormone is comprised of three constituent peptides: a 57-amino acid sequence at the amino terminus (NProCT); the centrally positioned immature CT that contains a terminal glycine; and a 21-amino acid CT carboxyterminus peptide I (CCP-I) [5].All these peptides are found in the serum of normal persons.In contrast to the short half-life of calcitonin (10 minutes), procalcitonin has a long half-life of (25-30 hours) [6].
PCT is produced by parafollicular cells (C cells) of the thyroid and by the neuroendocrine cells of the lung and the intestine.PCT originate from the calcitonin I (CALC-1) gene on chromosome 11.Expression of the PCT-producing calcitonin 1 (CALC-1) gene is increasing in multiple extrathyroid tissues throughout the body in the bacterial infections.In the absence of infection, the extrathyroidal transcription of CALC1 gene is suppressed [7,8].Therefore PCT detectable in the plasma during infection is not produced by C-cells
Introduction Sepsis
Sepsis is a systemic immune response to infection by microbial organisms.Sepsis is defined as the presence (probable or documented) of infection together with systemic manifestations of infection.The most common primary source of infection resulting in sepsis is the lungs, accounting for about half of all cases, followed by the abdomen, and the urinary tract.No definitive source is found in one third of cases [1].Sepsis encompasses a spectrum of illness that ranges from minor signs and symptoms through to organ dysfunction (severe sepsis) and shock.Increasing severity correlates with increasing mortality, which rises from 25-30% for severe sepsis up to 40-70% for septic shock [2].
Sepsis and its complications have a significant and increasing impact on health sector, and are one of the leading causes of mortality.The incidence of sepsis is increasing in all areas of the world.In and the inflammation-related functions of its propeptides [10].It may follow either a classical hormonal expression or, alternatively, a cytokine-like expression pathway.The production of hormokines is mediated by as yet unknown factors and may be induced either directly via microbial toxins or indirectly via a humoral or cellmediated host response [11].
Procalcitonin and Pathogenesis of Sepsis
Cytokines have been implicated in the pathogenesis of sepsis.Macrophages phagocytose bacteria and produce a range of proinflammatory cytokines, which initiate the innate immune system's response to the bacterial pathogen.This result in the production of interleukin (IL)-1β, tumor necrosis factor (TNF), and IL-6.The surge of proinflammatory cytokines during the innate immune response is a clinically visible and widely studied aspect of the pathophysiology of sepsis [12].
As the paradigm of sepsis pathogenesis has evolved over time and as different therapeutic approaches to sepsis have been tried, different biomarkers have been used for diagnosis of sepsis and monitoring of treatment [13].TNF, IL-1β and IL-6, as well as C-reactive protein (CRP), were all investigated as potential biomarkers.In the 1990s, investigators discovered that the levels of PCT were elevated in patients with bacterial infection, and it emerged as another potential biomarker [6].Increased plasma PCT was suggested to be added to the updated definition of sepsis in 2003, as one of the diagnostic criteria for sepsis [14].
The primary pathophysiological trigger for elevated level of PCT is infection.Iinvestigations identified PCT as part of the complex proinflammatory response of the innate immune system [15].A marked increase in serum PCT often indicates an exacerbation of the disease, and a decreasing level is a sign of improvement.
During Sepsis, there is an increase in CALC-1 gene expression which causes a release of PCT and, more importantly, their levels persist for relatively long periods of time and correlate with sepsis severity and mortality.However, researchers have not determined the exact role that PCT plays in the pathogenesis of sepsis, further pathophysiological studies recommended regarding the biological role of calcitonin precursors during sepsis [16].
A recent experiment showed that PCT is a potent amplifier of the inflammatory cascade.It has shown that it induces pro-inflammatory like effects on leukocytes (increased the expression of surface markers on neutrophils and lymphocytes), increases leukocyte-derived cytokines and also augments nitric oxide (increases level correlates with severity of inflammation) [17].Advance researches that are designed in clarifying the role of PCT in sepsis would aid in better understanding the pathogenesis of sepsis.The PCT test is relatively new, but its utilization is increasing.Multiple studies have shown that it has promise in helping to evaluate the risk of developing sepsis.
Procalcitonin as a Diagnosis Marker for Sepsis
A number of the inflammatory markers, such as leukocyte cell count, CRP, and cytokines (TNF-α, IL-1β, or IL-6), have been applied in the diagnosis of inflammation and infection, but their lack of specificity has generated a continued interest to develop more specific clinical laboratory tests [2].PCT is widely reported as a useful biochemical marker to differentiate sepsis from other non-infectious causes.The PCT test has been approved by the U.S. Food and Drug Administration (FDA) for use in conjunction with other laboratory findings and clinical assessments to assist in the risk assessment of critically sick people for progression to severe sepsis and septic shock.
Serum PCT levels are elevated in patients with bacterial infections, but are below the detection limit in healthy individuals and in patients with viral infections.This indicates that PCT level is useful for the diagnosis of systemic bacterial infections [18].PCT secretion reflects the severity of the inflammatory insult, with higher levels associated with more severe disease and declining levels with resolution of illness.Normal reference value is < or =0.15 ng/mL.PCT level between 0.15 and 2.0 ng/mL do not exclude an infection, because localized infections (without systemic signs) may be associated with such low levels.Levels > 2.0 ng/mL are highly suggestive of systemic bacterial infection/sepsis or severe localized bacterial infection.Although increased PCT levels may not always be related to systemic bacterial infection [19].
The diagnosis of infection in critically sick patients is challenging because traditional markers of infection are often non-specific and misleading.PCT was found to be a more accurate diagnostic parameter for sepsis, and therefore daily determinations of PCT may be helpful in the follow up of critically sick patients [20,21].A systemic review and meta-analysis done on 30 studies (2013) revealed that PCT has a mean sensitivity of 0.77 (95% CI 0.72-0.81)and specificity of 0.79 (95% CI 0.74-0.84)[22].The authors concluded that PCT is a helpful marker for diagnosis of sepsis in critically sick patients.However, it cannot be recommended as the single definitive test for sepsis diagnosis but rather it must be interpreted in context with information from clinical data.
The application of procalcitonin has to be considered along with clinical diagnosis of sepsis and septic shock.A description of the specific clinical diagnostic criteria is used.Systemic inflammatory response syndrome (SIRS) can be readily diagnosed at the bedside by the presence of at least two of the following four signs: body temperature variations (hyperthermia or hypothermia), tachycardia, tachypnea, and variations in white blood cell count (leukocytosis or leukopenia).The clinical diagnosis of sepsis requires the presence of SIRS resulting from presumed or known site of infection.Septic shock, a subset of sepsis, can be clinically diagnosed as a persistently low blood pressure despite adequate fluid resuscitation.
A literature review on the usefulness of PCT test among patients with chronic renal insufficiency found that PCT can be a useful test in identifying systemic infections among patients with renal dysfunction [23].It is often challenging to distinguish between bacterial infections and disease-flares in patients with systemic autoimmune diseases, because of similar clinical presentation.A high PCT level is an excellent marker of bacterial infection in patients with systemic autoimmune diseases, even when they are being treated with corticosteroids and immunosuppressive agents [24].
Procalcitonin as a Prognosis Marker for Sepsis
A comprehensive knowledge of the biology, advantages and limitations of biomarkers is essential before applying them as a routine clinical tool for prognostic value.Elevated level of PCT at admission to the ICU was found to be a better predictor of mortality that helps in the stratification of patients and to identify patients at higher risk of adverse outcomes [25].
Many studies have demonstrated that serum PCT levels are increased in patients with sepsis, and the high levels of PCT correlate with the outcome of the disease.PCT can be used for differential diagnosis, prognosis, and follow-up of critically sick patients [26].Serum PCT levels have been noted to increase with increasing severity of sepsis.In addition, a rising PCT level might be used as an indicator that an infectious process is not under control and that better source control is required [27].
A large cohort study (2006) done among critical care patients with daily PCT measurements showed that a high maximum PCT level and a PCT increase for 1 day are both independent predictors of 90 day mortality [28].The study also revealed that the relative risk for mortality increased with every day rise of the PCT value.PCT levels should be determined serially, it is used for monitoring the host response to the infection and the antibiotic treatment.Few studies suggest that if PCT levels decrease more than 30% of the initial value after the first 24 hours of antibacterial treatment, the infection can be considered under control and the treatment is favorable [29].
A systemic review of these trials revealed that measurement of PCT levels for antibiotic decisions in patients with sepsis appears to reduce antibiotic exposure without worsening the mortality rate [30].
Procalcitonin and other Bio-Markers of Sepsis
Biomarkers should provide a more reliable tool in ascertaining the presence of a relevant bacterial infection, its severity and treatment response.An ideal biomarker should allow, with high diagnostic accuracy, for an early and rapid recognition of sepsis.PCT is a biomarker that fulfills many of these requirements, especially in comparison to other commonly used biomarkers, and that has demonstrated superior diagnostic accuracy for sepsis [31].
PCT is found to be superior to CRP in terms of accuracy at identifying sepsis and assessing the severity of sepsis [32][33][34][35].PCT was a more reliable marker in the diagnosis of sepsis than other biomarkers including TNF-α, IL-2, IL-6 and IL-8 [20].Serum PCT concentration is more sensitive and specific marker of sepsis as compared with serum IL-6 and lactate levels [21].
There is still a need for new sepsis biomarkers that can aid in therapeutic decision making and add information about screening, diagnosis, risk stratification, and monitoring of the response to therapy.Novel approaches to sepsis promise to transform sepsis from a physiologic syndrome into a group of distinct biochemical disorders and help in the development of better diagnostic tools and effective adjunctive sepsis therapies [36].
Procalcitonin, Sepsis and Emergency Room
Biomarkers may play an important role in the management of patients with sepsis in emergency rooms.PCT is considered a relatively innovative and highly specific biomarker for the diagnosis of clinically relevant bacterial infections and sepsis; therefore it is increasingly recognized as an important diagnostic tool in clinical practice of emergency room.[37].In addition, a multimarker panel approach performed by rapid and accurate assays (PCT being one of the important biomarkers) is useful for emergency physicians to promptly identify sepsis thus managing through better diagnosis, treatment and risk stratification.[37,38].Therefore PCT could be a safe and effective tool to guide clinical and therapeutic decisions in emergency rooms.
An observational study conducted in Europe and Asia involving 340 patients who were enrolled from the emergency department investigated the diagnostic and prognostic utilities of PCT [39].The study evidently showed that PCT concentration-based sepsis diagnosis is more reliable than the clinical diagnosis.Therefore assessment of patients with sepsis in emergency room should include appropriate use of PCT for improved diagnosis and management of patients.
Procalcitonin and Sepsis Treatment
PCT provides important information in early stages of sepsis as well as during antimicrobial treatment.In fact, PCT can be useful for antimicrobial stewardship and its utilization may safely lead to significant reduction of unnecessary antimicrobial therapy [31].
A systemic review and meta-analysis (2013) on PCT-guided antibiotic therapy indicated that PCT guidance can safely reduce antibiotic usage when used to discontinue antibiotic therapy in critically sick patients [40].PCT-protocols to guide antibiotic treatment in severe infections are known to be effective.These protocols are based on daily measurement of PCT and clinical signs & symptoms of infection.Implementation of a PCT-protocol in a real-life clinical setting was also associated with a reduced duration of antibiotic therapy in septic patients without compromising clinical or economical outcomes [41].
Conclusion
PCT has been proved to be superior biomarker, however its use still has to be interpreted in the context of clinical presentation.This short review identified the association of PCT and sepsis.Additional knowledge on the role of PCT might be valuable for more effective and targeted treatment in sepsis.More research is needed to further explore PCT as a possible gold standard biomarker in sepsis. | 2018-12-11T12:15:07.914Z | 2016-04-30T00:00:00.000 | {
"year": 2016,
"sha1": "5c05c0738d409b7319e7caf2bbf77ce95fb5b4a9",
"oa_license": "CCBY",
"oa_url": "https://clinmedjournals.org/articles/jide/journal-of-infectious-diseases-and-epidemiology-jide-2-006.pdf",
"oa_status": "HYBRID",
"pdf_src": "Anansi",
"pdf_hash": "5c05c0738d409b7319e7caf2bbf77ce95fb5b4a9",
"s2fieldsofstudy": [
"Biology",
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
246109555 | pes2o/s2orc | v3-fos-license | CRBPDL: Identification of circRNA-RBP interaction sites using an ensemble neural network approach
Circular RNAs (circRNAs) are non-coding RNAs with a special circular structure produced formed by the reverse splicing mechanism. Increasing evidence shows that circular RNAs can directly bind to RNA-binding proteins (RBP) and play an important role in a variety of biological activities. The interactions between circRNAs and RBPs are key to comprehending the mechanism of posttranscriptional regulation. Accurately identifying binding sites is very useful for analyzing interactions. In past research, some predictors on the basis of machine learning (ML) have been presented, but prediction accuracy still needs to be ameliorated. Therefore, we present a novel calculation model, CRBPDL, which uses an Adaboost integrated deep hierarchical network to identify the binding sites of circular RNA-RBP. CRBPDL combines five different feature encoding schemes to encode the original RNA sequence, uses deep multiscale residual networks (MSRN) and bidirectional gating recurrent units (BiGRUs) to effectively learn high-level feature representations, it is sufficient to extract local and global context information at the same time. Additionally, a self-attention mechanism is employed to train the robustness of the CRBPDL. Ultimately, the Adaboost algorithm is applied to integrate deep learning (DL) model to improve prediction performance and reliability of the model. To verify the usefulness of CRBPDL, we compared the efficiency with state-of-the-art methods on 37 circular RNA data sets and 31 linear RNA data sets. Moreover, results display that CRBPDL is capable of performing universal, reliable, and robust. The code and data sets are obtainable at https://github.com/nmt315320/CRBPDL.git.
Introduction Circular RNA (circRNA) is a special circular endogenous noncoding RNA produced by selective shearing [1,2]. It has been proven to be widely present in Drosophila, mice, the hippocampus and human cells and tissues [3,4]. Although the RNA-binding proteins (RBP) binding sites on circular RNAs are less numerous than those on linear mRNAs, there is still strong evidence to support the interaction of RBPs with circular RNAs [5,6]. On the one hand, circRNAs can regulate RBPs in a variety of ways. CircRNAs can competitively bind to RBPs, regulate the function of RBPs, and act as sponges of RBPs, platforms for RBP assembly, and supertransporters that concentrate certain specific components [7,8]. RBP-adsorbed circRNA can be used as a regulatory factor for target gene transcription and splicing [9]. circRNA can also be used as a bait to retain RBPs in a specific intercellular space and as a scaffold to promote contact between two or more RBPs [10]. On the other hand, the influence of RBPs on circRNAs is becoming increasingly prominent. As a protein that binds to double-stranded or single-stranded RNA, RBPs are present throughout the life of RNA and mediate the maturation [11], transport [12], positioning and translation of RNA [13]. RBPs affect the entire process of the circRNA life cycle, and some RBPs are also involved in the generation of circRNAs, such as Quking (QKI), FUS, and HNRNPL. Moreover, they are involved in almost every aspect of the cyclic RNA life cycle, including generation [14], posttranscriptional regulation [15], and functional execution [16]. Some specific RBPs are tissue-specific or produced under pathological conditions, and their expression defects can cause a variety of diseases and other effects. Multiple studies have shown that the interaction between circular RNA and RBP has an important impact on cancer and other diseases and may be a disease of biomarkers [10,[17][18][19][20][21]. Therefore, predicting the binding site of RNA and RBP can provide insight into the mechanisms underlying diseases involving RBPs and help to further explore the role of circRNA in disease pathophysiology.
As a promising method, machine learning has been used to solve various biological problems, its superiority has been proven many times, and it has gradually been used to identify the binding sites of circular RNA-RBP [22]. Matizka et al. proposed the GraphProt method, which can learn secondary structure characteristics, and used support vector machine (SVM) to predict binding sites and affinity of RBPs in all tissues [23]. Corrado et al. applied recommendation algorithm to recommend RNA targets for RNA-binding proteins based on protein domain composition and RNA predicted secondary structure features [24]. Yu et al. employed the random forest algorithm (RF) to predict specific and general RBP sites based on motif information [25]. The above machine learning models are mainly based on the structural characteristics of RNA sequences to identify the binding sites of circular RNA-RBP [26,27].
Deep learning has fulfilled remarkable accomplishment in the field of bioinformatics recently [28][29][30], which also includes the prediction of RNA-protein interactions The Deep-Bind method utilized convolutional neural network (CNN) to learn binding preference of individual RBPs and obtains better performance [31]. Pan et al. proposed iDeepE method, which uses the global CNN model to predict the binding site by studying RNA sequences [32]. In addition, they further used two separate CNNs and a long-term short-term memory network to learn the sites [ . However, the feature learning network is relatively simple, and there is still potential for improvement in prediction performance.
In our work, we establish a novel computational predictor CRBPDL, which based on an ensemble deep network to identify circRNA-RBP interaction sites. First, we adopted 5 coding schemes to provide comprehensive feature information for model training, including k-nucleotide frequency (KNF), Doc2vec, electron-ion interaction pseudopotential (EIIP), nucleotide chemical properties (CCN) And cumulative nucleotide frequency (ANF). Due to the different distributions of feature descriptors, we first applied convolution filters to the features respectively, and then concatenated them into a feature matrix. Subsequently, to automatically extract high-order local and global context information from feature descriptors, we constructed a deep neural network architecture, which consists of a deep multi-scale residual network (ResNet) and a bidirectional gated recurrent unit with a self-attention mechanism (BiGRUs) network composition. We used deep multi-scale residual networks (MRSN) and BiGRUs to learn local and global contextual information, and can effectively represent highlevel features. Then, used the self-attention mechanism to train the robustness of the model. After model training and selection, we can get the optimized deep learning model (for convenience, the deep learning model before integration is named "sig-CRBPDL"). Finally, the Ada-Boost algorithm was used to integrate the deep learning model. We benchmarked CRBPDL and existing predictors on the unified circRNA dataset. The benchmark test results clearly showed the superiority of our proposed CRBPDL. In addition, CRBPDL has the potential to recognize linear RNA-RBP interaction sites. The benchmark results showed that CRBPDL also has stable performance in predicting linear RNA-RBP binding sites. The structure of the CRBPDL model is shown in Fig 1.
Model performance under different network layers
Network depth has great effects on the performance of deep learning models. Different network depths will lead to diverse results. A relatively shallow network will make the model perform poorly, and an overly complex network will increase the calculation of the model. This section analyzed the model performance changes under different network layers. We compared the increase and decrease: reducing one MSRB block, that is, a 3-layer convolutional network layer, and adding an MSRB block, which means adding a 3-layer convolutional neural network, respectively named CRBP-3 and CRBP+3 for convenience of description. We calculated the prediction performance of CRBPDL, CRBPDL-3 and CRBPDL+3 (AUC as an evaluation index) and running time under 37 data sets (Fig 2A).
First, it can be seen from the scatter plot that the AUC value distribution of CRBPDL is 0.9174, which is higher than that of CRBPDL-3 (0.8995), and the running time is the opposite.
The AUCs of the CRBPDL and CRBPDL+3 (AUC is 0.9011) distributions are not very different, but the running times are quite different. On the other hand, by observing the distribution of the maximum, minimum, and average values in the box chart, it can be found that the midline positions of the three are similar, but the bottom positions of the CRBPDL-3 and CRBPDL+3AUC box charts are lower. The top position of the box chart for CRBPDL+3 time is higher. The prediction performance distributions of CRBPDL-3 and CRBPDL+3 were quite different, and the performance is not stable enough. In contrast, the distribution difference of CRBPDL is smaller, and the stability is better. In terms of time consumption, the performance of CRBPDL-3 is better than that of CRBPDL, but the difference is small. This proves the complexity of the network layer may impact the behavior of the network. It also shows in practical applications, when faced with the needs of different time consumption and prediction effects, both the progressive neural network and the deep neural network have research significance and value.
Model performance under different epoch times
This section statistically analyzes the changes in the loss and ACC of the training set and the validation set during the training phase and accordingly analyzes the impact of epoch on the model performance and the convergence of the model. If the graphs of all 37 data sets are displayed, there are too many pictures. Therefore, we only randomly selected 9 results for display, and can illustrate the effects of different data sets (the loss results of the remaining 28 data sets can be seen in S1 Text). This section mainly analyzes 9 out of 37 data sets which are AGO1, AGO2, U2AF65, DGCR8, FOX2, WTAP, EIF4A3, FMRP, and ZC3H7B. The results of the 9 data sets are shown in Fig 2B. We can see that as the epoch time increases, the train-acc and validation-acc of CRBPDL both show an upward trend, and the overall train-loss and validation-loss show a downward trend and gradually stabilize; the model gradually converges, and training results are gradually optimized. The obvious performance of overfitting is that the performance of the training set is particularly fine, but the performance of the verification set is exceptionally poor. It can be found that in the 5th and 6th data sets of the 9 data sets, the trend of the acc curve is quite different. The performance effect of train-acc is obviously better than that of validation-acc. There is an obvious overfitting phenomenon, but not in the other 7 data sets. The reason for this difference may be that the data volumes of AGO3 and WTAP (that is, the 5th and 6th data sets) are small (1,210 and 892 data points, respectively), and the learning and training process of the CRBPDL model is not sufficient. In contrast, the data volume of the other data sets is on average one hundred times greater, achieving better training results. It can also be seen that the size of the data set is very important for the performance of deep learning neural networks.
Model performance under different learning rate
As a hyperparameter of the neural network, the learning rate can be used to improve the performance of the model. The lower the learning rate is, the slower the gradient rate. When determining the learning rate, it's generally essential to rely on the comparison of old experience and multiple experiments. The section analyzes the effect of the learning rate. To compare the effect, we analyzed three learning rate attenuation schemes (step-based attenuation learning rate scheme, linear learning rate attenuation scheme, polynomial learning rate scheme) and two fixed learning rates (0.002, 0.0005). Fig 2C shows the comparison of the AUC of the network optimization process when uses different learning rates.
On all circRNA data sets, the Adam method achieved an average AUC value of 0.9284, which was significantly better than the 0.8926 of the linear scheme and the effect of two fixed learning rates (average AUC of 0.8167 and 0.8747, respectively). Although it is not much different from the average AUC value of 0.9273 of the step scheme, there are two abnormalities in the step. In contrast, Adam's performance is relatively stable. By performing an experimental comparison of five case, it shows that the Adam linear learning rate plan is always better than other types of plans and has better performance. Therefore, we choose Adam as the learning rate learning plan. First of all, by observing the trend of the line chart, we can find that the AUC value of Fea-A is higher than that of Fea-B and Fea-C on multiple data. In addition, our method Fea-A obtains an average AUC value of 0.9201, which is not only upper than the 0.8928 of the Fea-B, but also superior than the 0.8792 of the Fea-C. For one thing, Fea-B uses 6 hand-designed features, and it is possible that a single hand-designed function is not suitable for advanced network architectures. For another thing, Fea-C is an improvement of one-hot encoding, which only uses feature type references in the pseudo-translation process. This may cause CRBPDL to fail to fully understand enough information in the circRNA-RBP interaction. Furthermore, the experimental results prove the validity of our feature encoding schemes.
Model Performance under different feature encoding schemes
In addition, we analyzed the different performance of the five feature encodes, and the results were shown in Fig 3A. It can be found that on 37 data sets, the Doc2vec coding scheme is relatively better than the other four. It shows that the global text characteristics of RBP binding sites are relatively obvious. Our word vector model seems to have learned the subtle sequence context from semantics, thereby improving the recognition performance. For cir-cRNA data sets, the experimental results show that the self-learning word vector encoding scheme proposed in this paper has a good application prospect. Based on the word vectors obtained, whether the conservative motifs of the binding sites can be analyzed.
Performance of neural network structures with different depths
To prove the effectiveness of our proposed CRBPDL, we input our features into different CNN to compare the prediction performance of different neural networks.We compared its performance with 5 structures: CNN-LSTM, iDeepE [32], ResNet [38], CRIP-RNN [36], and CNN-BiLSTM. CNN-LSTM includes two bidirectional LSTM layers and two fully connected layers; iDeepE combines output features of the global network and local neural network, and two layers of local multichannel neural networks (convolution, ReLU and max pooling) express high-level features and then input the feature map into two fully connected layers; ResNet uses a 21-layer local multichannel network, inserts a shortcut connection between the two networks, and makes the network into a corresponding residual network; CRIP uses two layers, a CNN that extracts high-level features and a RNN that acquires the long-term dependence of sequence; and CNN-BiLSTM uses bi-directional long-short term Memory to integrate data, including two bidirectional LSTM and two fully connected layers. These network structures can be built with reference to the literature or built by themselves, and the parameters of each model have the same parameters as CRBPDL. The experimental result is displayed in Fig 3B. As shown in Fig 3B, we find that the average AUCs of all circular RNA data sets are 0.9174, 0.8778, 0.8854, 0.8877, 0.8760, 0.8733, 0.9148, and 0.9201. When the feature codes of this article are input into different neural networks, the results obtained are different. The difference is based on whether the network structure expresses high-level features accurately. Obviously, the CRBPDL model can learn more valuable sequence information for the identification of cir-cRNA-RBP interaction sites.
To demonstrate how CRBPDL learns efficient feature representation, we take the "WATP" data set as an example, and use t-SNE graphs to visualize feature representation. Both dimensions automatically learn CRBPDL. The original features are shown in Fig 3C. We can find that it is challenging to visually distinguish two categories with primitive characteristics. In addition, the second level of full connectivity after feature representation (Fig 3D) can be used to better identify and separate positive and negative examples. Graphical display shows that CRBPDL can effectively learn excellent feature representations. Moreover, we further analyzed the different performance of MSRN and BiGRU, as shown in Fig 4A. It can be found that although the difference between the two is relatively small, the effect of MSRN is significantly better than that of BiGRU, indicating that in the CRBPDL model, the contribution of MSRN is greater. [42] and five other existing calculation methods are measured by AUC. CSCRSites was based on multiple convolutional thermal coding layers to identify cancer-specific RBP binding sites on circular RNAs. In contrast, CircSLNN used a sequence tagging network to recognize the interaction site. In terms of fairness, the six methods were tested on a unified benchmark data set, using the same sequence similarity threshold. In addition, the same setting environment is used as the model iCircRBP-DHN. The AUC results and average values of other comparative experiments are shown in Table 1, obtained directly from the literature (the maximum value was shown in bold) [37]. As shown in Table 1, the average values of CRBPDL, iCircRBP-DHN, PASSION, CRIP, CSCRites and CircSLNN are 0.9188, 0.908±0.06, 0.884±0.06, 0.876±0.07, 0.842±0.07 and 0.809 ±0.010, respectively. Obviously, our model improves the state-of-the-art performance in 28 of the 37 and accomplishes the supreme average AUCs of 0.9174, specifically in AGO1, AGO2, ALKBH5 and MOV10. And we provide the ROC curve of CRBPDL, and the average ROC curve (Fig 4C). The results fully indicate the enhancement of CRBPDL. At the same time, we also noticed that on the 6 data sets, the performance of CRBPDL has a very small gap with iCircRBP-DHN and PASSION, especially 4 of them are slightly worse than PASSION. The underlying reason may be that PASSION has extracted 6 types. This shows that manual features including richer sequence information can be used, and integrated optimization algorithms can also be used. In addition, CRBPDL is better than CircSLNN, but CircSLNN is a sequence tagging method that can predict the location of the binding site. Therefore, as a new research direction, we can consider whether we can improve the accuracy of predicting the position of the binding site, not just as a binary classification problem.
Prediction performance of CRBPDL on 31 linear data sets
Similar to CRIP and PASSION, our CRBPDL also has the ability to identify linear RNA-protein interactions. To demonstrate the performance of our model CRBPDL, we compare it with ICIRCRBP-DHN, CRIP, iDeepS, DEEPBbind, CSCrites, and CIRCSLNN. To make a fair comparison, we used the same experimental data as the iCircRBP-DHN, and the results of the other comparative experiments were obtained directly from the literature [37]. The experimental results are shown in Table 2. From Table 2, CRBPDL obtained an average AUC of In the remaining 30 data sets, our performance is still better than other methods. In addition, we have given the ACC on 31 data sets (Fig 4D), and we can find that the accuracy on the 31 data sets can meet the identification requirements of linear RNA-RBP binding sites.
Conclusion
In this paper, we design a new deep learning method, called CRBPDL, for circular RNA-RBP interaction site identification. Based on the MSRN framework, CRBPDL first connects the five codes into a single feature vector. Then MSRB is used to automatically explore higher-level local or global context dependencies and obtain high-level sequence features. Subsequently, the output of each MSRB is combined for global hierarchical feature fusion. And add self-Attention to grasp more critical and relevant features and improve prediction performance. Finally, an integrated deep learning network is constructed based on the Adaboost algorithm. Through the visualization of feature representation, this unique architecture has proven to be effective. To verify CRBPDL, we performed predictions of the binding sites of circRNA and linear RNA and evaluated the performance of different methods. The comparison of 37 circular RNA data sets and 31 linear RNAs not only proves the effectiveness of our method but also shows the potential of the model in the identification of circular RNA-RBP interaction sites. Currently, there are few data on known RBP binding sites. The positive and negative samples are unbalanced. Therefore, the most important thing is that future research is to expand the data set, collect RBP binding sites that bind to circRNA, lncRNA or other RNAs, explore their binding characteristics, and develop universal prediction software.
Data sets
To prove the effectiveness of our proposed CRBPDL and make a fair comparison with other tools at the same time, we used the benchmark data set (named as 'circRNA_RBP-37') used in [35,37,42]. The data set consists of 37 RBPs downloaded from the circinteractome database (https://circinteractome.nia.nih.gov/) [19]. The database collects RBP bound to mature circular RNA and RBP bound to the upstream and downstream sequences of mature circular RNA. Since RBP binding may play a role in regulating splicing events near the splicing site, we considered all RBP binding sites in this study. In the end, we obtained a total of 32,216 circular RNAs related to 37 circular RNA data sets. Among them, the positive sample came from the interaction site on the circular RNA verified by the laboratory. In each CLIP-seq peak, the sequence fragment with a length of 101 nucleotides (nt) was centered and extends 50 nucleotides (nt) in both directions. At the same time, negative sequences were randomly selected from the left-over circular RNA fragments. Subsequently, we applied the same postprocessing method to extract the 101 nucleotide length (nt) binding sites/residual intermediate readings in the previous work [17,18]. Since sequence similarity will influence the consistency of the ML, we used CD-HIT to eliminate the sequence with a similarity threshold of 0.8, which is the same as in CRIP and PASSION. After removing sequence redundancy, we got the final data set, namely the positive and negative samples are 335,976 and 335,976 respectively. 80% of the data sets were selected as training set, and 20% were used as test set. Additionally, refer to other studies [17,18], we compared the efficiency of CRBPDL to identify the linear RNA -RBP interaction sites. We downloaded the linear RNA data set from PAS-SION [35] and iDeepS [22], which includes the linear RNA dataset of CLIP-Seq data combined with 31 RBP. Each data set has 5,000 training sets and 1,000 test sets.
Feature encoding k-nucleotide frequencies.
To characterize the local context features of circular RNA sequences, we used KNF coding sequences. KNF describes the frequency of all possible polynucleotides of k nucleotides in the sequence. In this study, we took k = 1, 2, 3, namely singlenucleotide composition frequency, dinucleotide composition frequency and trinucleotide composition frequency. KNF retains a large number of original sequence patterns and integrates a variety of sequence information [43,44]. Compared with traditional single hot spot representation [45], KNF effectively compensates for the lack of information.
Doc2vec. In recent DL model research, to learn more sequence context and semantic information, an increasing number of sequence studies have adopted continuous, high-dimensional word embedding-based coding to substitute one-hot coding, and have achieved good results. Therefore, based on the circRNA corpus of circBase [46], we used the Distributed Memory Model of Paragraph Vectors (PV-DM) model of the Doc2Vec algorithm to vectorize the sequence [47] and train the vectorized model Doc2Vec.model. After that, sequence data were input into the model, 10-mer sequence fragments were taken as circular RNA words, and word embedding training was used to obtain feature vectors. In this way, learning as a continuous distribution representation of global context features expands the vocabulary and can capture the semantics and grammar in these subsequences for long-term dependency modeling.
Electron-ion interaction pseudopotential. The EIIP [48] describes the characteristics of free electron energy on the circRNA sequences. EIIP was widely used to predict the binding sites of RBPs. The EIIP values of the four characters that may appear in the sequence (ie, "A", "T", "C", "G") are 0.1260, 0.1335, 0.1340 and 0.0806. Hence the EIIP coding method can be used to encode DNA sequence as a digital vector. For example, AATCCGA encoding is a numeric vector consisting of (0.1260, 0.1260,0.1335, 0.1340, 0.1340, 0.0806,0.1260).
Chemical characteristic of nucleotide. Each nucleotide has three types of chemical characteristics (CCN): chemical functions (including amino and keto groups), ring structure (including bicyclic purines and monocyclic pyrimidines), and hydrogen bonds (including weak hydrogen bonds and strong hydrogen bonds) [49]. For the ring structure, A and G belong to purines, coded as 1, and C and T belong to pyrimidines, coded as 0. For chemical functions, A and C belong to amino groups, coded as 1; G and T belong to ketone groups, coded as 0. For hydrogen bonds, A and T belong to a weak hydrogen bond, coded as 1, while C and G belong to a strong hydrogen bond, coded as 0. For example, AATCCGA can be encoded as (1,1,1,1,1,1,0,0,1,0,1,0,0,1,0,1,0,0,1,1,1). Accumulated nucleotide frequency. ANF presents the density characteristics of nucleotide sequence [49]. Suppose a circRNA sequence S = s 1 s 2 . . .s i , where i is the length of S. S j = s 1 s 2 . . .s j , j is the length of S j . S j is the j-th prefix sequence of S. Then the ANF calculation formula is:
Multiscale residual network
To obtain rich feature information, a multi-scale CNN layer is constructed to capture highlevel features. Unlike traditional Convolutional Neural Networks, different from traditional CNN, multiscale residual network can improve the information trend flow and gradient of the whole network, reduce the computational complexity and improve the model performance [50].
In the MSRN framework, due to the different distributions of the five feature descriptors, we employed convolution filters on five characteristics with a convolution kernel of 128 and then cascaded, which is a common method to balance the distribution of biological features. Afterwards, the MSRN framework contained a shallow CNN extraction layer, and the size of the convolution kernel was 3. Then, the inception module, including 6 cascaded multiscale residual block (MSRB) modules, was used, and the convolution kernel was 64. Each MSRB includes a 3 convolutional layer. Based on the hierarchical feature fusion structure (HFFS), the output of each MSRB was combined to perform global feature fusion. Subsequently, following input to a layer of convolution kernel, there were 192 filters, and a 1×1 convolution can increase and decrease the number of channels, organize information across channels, and increase feature transformation with a small amount of calculation and nonlinear transformation to improve the network expression ability. After that, there was a merge layer with a dropout value of 0.4.
bidirectional gating recurrent unit
For circRNA sequences, besides local background information, there are also long-chain dependencies [51]. Multiscale residual block network can capture only the dependencies between sequences. Therefore, the study employed a Bidirectional GRU to obtain context information from the front and back at the same time to improve the performance.
Bidirectional GRU has only two gates, namely, the update and the reset gate. The update gate controls the extent to which the state information at the previous moment is brought into the current state. The larger the value of the update gate, the more the state information at the previous moment is brought in. The reset gate is used to control the degree of ignoring the state information at the previous moment. The smaller the reset gate, the more information is ignored. The bidirectional GRU can adaptively change its state according to the input, thereby solving the problem of vanishing gradient in RNN.
Self-Attention
The self-attention mechanism was to adaptively pay attention to and learn an important part according to the needs, and ignore the insignificant part. It was widely used in various deep learning applications, including vision processing, phosphorylation site prediction, drug target prediction, etc. [52]. The intention of the attention mechanism is to neglect insignificant word in the bulk of information, selectively filter out a particle of important information, and and express the importance of the information by calculating the weight of the information.
In this research, in our model CRBPDL, the output matrix of the BiGRU layer and its transposed matrix were input into the attention layer, and different features were given different weights, and important features were selected from the dimensional features. Abandon some secondary features and used sigmoid as the activation function.
Implementation
CRBPDL was implemented using the Keras 1.1.2 library in Python. First, we used 80% of the benchmark data set for the training and 20% for testing. Then, on the training set, 80% for training and 20% for verification. Acc was used to evaluate each parameter setting. The verification data set was applicable to monitor the astringency of each stage in the training process, and the training process can be quitted in advance. The study adopted the update method of the learning rate of the Adam gradient descent algorithm, where the initial learning rate is set to 0.001, the max epochs is 200, the epochs is 30, and the batch size is 50. In addition, we have also adopted a variety of techniques to prevent or reduce overfitting, such as batch normalization [50], dropout [51] and early stopping. We used the selected optimal parameter settings, used all training data to train the model, chose the model with the greatest performance as the base model, employed AdaBoost for ensemble, and applied the integrated model as the computational model. AdaBoost is an iterative algorithm. Its core idea is to train different classifiers (weak classifiers) for the same training set, and then group these weak classifiers to form a stronger final classifier (strong classifier) [53,54]. | 2022-01-22T06:18:41.923Z | 2022-01-01T00:00:00.000 | {
"year": 2022,
"sha1": "b928db708fae7f9b38eb74fce98db47076342d4a",
"oa_license": "CCBY",
"oa_url": "https://journals.plos.org/ploscompbiol/article/file?id=10.1371/journal.pcbi.1009798&type=printable",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "3227f57d86dfa97e3469e63e06bc022011a8a0b5",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science",
"Medicine"
]
} |
234554832 | pes2o/s2orc | v3-fos-license | The Mathematical Model of Transverse Vibrations of the Three-Layer Plate
The article in a flat setting investigated the antisymmetric oscillations of a three-layer plate, which is infinite in plan. It is believed that the plate is not symmetrical in thickness. Based on the exact solutions of the equations of the linear theory of elasticity in transformations, a theory of unsteady transverse vibrations of a three-layer plate is developed. The oscillation equations are derived with respect to two auxiliary functions, which are the main parts of the longitudinal and transverse displacements of the points of some “intermediate” surface of the middle layer. The distance of this surface to the coordinate plane of the plate is arbitrary. All components of the stress tensors and displacement vectors at the points of the layers are expressed, like the vibration equations, through the introduced auxiliary functions. The problem of harmonic antisymmetric vibrations of an elastic three-layer plate is solved.
Introduction
Multilayer plates and shells are widely used in various fields of technology. The sphere of usage of the three-layer plates is extremely wide. It includes such areas as construction, aircraft manufacturing, instrumentation, etc. Therefore, the calculation of such plates for the action of various dynamic loads is widely used in the design and operation of engineering structures, often working in extreme conditions on the effects of explosive, seismic and other loads [1]. Three-layer structures, the appearance of which was dictated by needs, primarily in the military field, fully met the requirements for a combination of mechanical characteristics of materials and functional requirements for engineering structures. The conditions for the intensive use of three-layer and multilayer structures contributed to the need to develop effective methods for calculating such elements [2]. Therefore, simultaneously with the beginning of the use of multilayer structures, theories of their calculation began to develop. Superiority in constructing a theory for calculating multilayer plates belongs to S. G. Lehnitsky, who created the theory, later called the "zig-zag" theory [3]. The use and refinement of the zig-zag theory continues to this day, as evidenced by numerous studies [4]. The work [5] is devoted to the development of a new theory of shear deformation for multilayer and composite plates. The displacement field is approximated using the product of trigonometric and exponential functions. The theory takes into account an adequate distribution of transverse shear deformations over the plate thickness and tangential boundary conditions without stresses on the boundary surface. A similar representation of the shear strain using the exponential function in the displacement field is the subject of [6]. It presents a new model of a multilayer composite structure, which describes the thickness distribution of shear 2 stress taking into account the conditions of the free boundary on the upper and lower surfaces using an exponential function. The number of published studies on the development of new models and theories of unsteady oscillations, homogeneous and multilayer plates and shells subjected to external dynamic loads is small compared with the case of statics [7]. One of the directions for solving this problem is devoted several scientific works [8,9]. These works are based on the method of general solutions in transformations that are used to satisfy the conditions specified on the surfaces of plates and shells. Many scientific works devoted to the creation of models can be cited where deformation is taken into account by attracting certain functions of the hyperbolic [10], power [11] and mixed hyperbolic with irrational [12] types. The new theory of higher-order shear and normal deformation [13] adjoins here for the analysis of bending and free vibrations of multilayer plates with functionally gradient isotropic front plates. A п-order model [14] has also been developed for shear deformation for calculating a functionally graded and composite multilayer plate. The logical continuation of these studies are included several scientific investigations such as [15,16], in which new mathematical models of stationary oscillations of plates and shells are proposed, based on the exact solution of the corresponding three-dimensional viscoelastic problems. The methods developed in these works for developing the theory of oscillations of plates and shells were then used to study layered structures [17,18]. An analyses of a large number of published works on the vibrations of homogeneous and layered plates, taking into account the viscoelastic properties of the material, shows that bending vibrations are the most studied. These studies are still ongoing [19][20][21]. At the same time, the study of studies on unsteady oscillations of three-layer plates, taking into account the new requirements of modern technology, allows us to conclude that there are still many problems that are far from being solved and therefore, new studies in this direction are relevant.
Methods
Consider a three-layer, infinite in plan, elastic plate. We assume that the plate consists of two bearing layers with thicknesses 1 h and 2 h , and the middle layer with thicknesses 0 2h ( Figure 1). In the case when the space between the bearing layers is filled with lighter, i.e. less rigid material, the middle layer is called aggregate. During deriving the equations of vibration, we assume that both the plate as a whole and each of its layers separately strictly obey the mathematical linear theory of elasticity and are described in exact formulations by its three-dimensional equations. Given the unlimited size of the plate, further we will assume that it is under conditions of plane deformation. Therefore, we will consider the plate in a system of rectangular coordinates Oxz and direct the axis Ox along the midline of the cross section, and the axis Oz up, perpendicular to the axisOx . We call the supporting layers of the plate the first and second (in accordance with their thicknesses 1 h and 2 h ) layers, and the middle layer is zero. Therefore, we will consider the plate in a system of rectangular coordinates and direct the axis along the midline of the cross section, and the axis up, perpen-
The equations of oscillation of a three-layer plate
For solving assigned task, the functions of external influences from (3) can be represented as [20] , we also represent potential functions in the form (3.1), substitution of which in (2) gives the ordinary Bessel differential equations with respect to the functions transformed by (5) The solutions of equations (6) in the case of transverse vibrations of the plate, taking into account the antisymmetric effects in the boundary conditions (3), will be Having expressed the transformed displacements m Ũ and m W through solutions (3.3), we expand the hyperbolic functions in them in power series. Further, as the desired functions, we take the main parts of the transformed displacements 0 U and 0 W such a surface of the middle layer of the plate, distance from the surface number which is an satisfying inequality .
which make up a system of two algebraic equations for two unknowns On the other hand, substituting solutions (7) into contact conditions (4) for
as well as (5).Then, boundary conditions (3) can be written as
Substituting the expressions of constants 2 1 (9) and expanding the trigonometric functions in series, on the left-hand sides of (9) by the degrees of the coordinate z and by reversing the system of equations obtained in this way, we will have the general equations of the transverse vibrations of a three-layer plate. These equations have infinitely high orders in derivatives. We assume that the truncation conditions for the infinite series indicated in are satisfied and will be limited to zero or first approximations in the expansions. As a result, we obtain approximate equations of vibration of l -plate length.
The stress-strain state of the plate
.The derivation of formulas for displacements and stresses at the points of the upper and lower bearing layers is also not difficult, but they are more cumbersome. Therefore, for example, we give these formulas in the operator form for displacements 1 U and 1 W , as well as for stress 1 xz , which have the form here L L L u w , , -fourth-order linear differential operators; through 1 k L the inverse operator is indicated , , -also linear differential operators of no higher than fourth order. The above expressions for the stress and displacement components (11) and (12) make it possible to determine the stress-strain state of an arbitrary point of a three-layer plate from the main parts 0 0 W and ) 0 ( 0 U from the results of solving differential equations (10).
Harmonic vibrations of a three-layer plate
As an example, we consider the problem of antisymmetric (transverse) harmonic vibrations of a threelayer plate based on the obtained approximate equations of oscillation. It should be considered that the plate surfaces are free from external loads. Then the right-hand sides of the oscillation equations (10) will be equal to zero. The solution of differential equations (10) with zero right-hand sides will be sought in the form here -circular frequency; kwave number. Substituting (13) into the oscillation equations, we have a system of two homogeneous algebraic equations with respect to 0 W and 0 U 0 7 k the difference is; for h 0 = 0,1 -0,98 ( 120%); for h 0 = 0,05 -0,2 ( 42%); for h 0 = 0,03 -0,15 ( 52%). The smallest values of the elastic modulus and density are from the given series of filler materials of the PCB, which corresponds to large values of frequency ( Figure 5). For example, at 7 k the frequency value for tantalite is 0.86, and for polymer 0.26. In this case, the frequencies of the fiberglass plate are lower (0.44) compared to wood plastic (0.48), despite the fact that the elastic modulus of fiberglass is greater than that of wood plastic. This is because wood plastic is much denser than fiberglass.
Conclusions
-the theory of non-stationary transverse vibrations of an elastic three-layer kin-plate is developed based on general solutions in transformations of equations of the theory of elasticity, in a flat setting; -the developed theory allows us to calculate all the components of the displacement vector and the stress tensor in the sections of the plate as a whole and of all layers through the introduced main parts of the intermediate surface of the middle layer; -the obtained general equations of vibration make it possible to obtain refined equations of the Tymoshenko type and approximate equations of the Kirchhoff type, which can be applied to solve applied problems of engineering practice; -from a comparative analysis of the obtained numerical results it follows that the vibration equations and formulas for determining the SSS developed in the work allow a high degree of reliability to determine the frequencies of antisymmetric vibrations of three-layer plates. Moreover, the frequency analysis performed on the basis of the presented model requires minimal computational resources; -regardless of the thickness of the middle layer, the dependence of the frequency on the wave number is directly proportional. For a fixed value of the wave number, an increase in the thickness of the middle layer of the plate leads to an increase in the vibration frequency, which strongly depends on the filler material. A plate with filler with large values of the elastic modulus and density has a lower oscillation frequency than with filler with lower values of the modulus and density. | 2020-12-24T09:12:30.618Z | 2020-12-18T00:00:00.000 | {
"year": 2020,
"sha1": "f53192b029d5519639153d51c48bbf6bd546c897",
"oa_license": "CCBY",
"oa_url": "https://iopscience.iop.org/article/10.1088/1755-1315/614/1/012062/pdf",
"oa_status": "GOLD",
"pdf_src": "IOP",
"pdf_hash": "910407a02e6d4f30ddcd2e0df2635b51fb29e483",
"s2fieldsofstudy": [
"Engineering"
],
"extfieldsofstudy": [
"Materials Science",
"Physics"
]
} |
119219419 | pes2o/s2orc | v3-fos-license | Effect of valence fluctuations on the ground state properties of SmB6
We argue that because of valence-fluctuation caused dynamical changes (fluctuations) of impurity energies in the impurity band of SmB6, energies of electrons occupying impurity sites can be due to the uncertainty principle only estimated with corresponding uncertainty, so they can be represented by energy intervals of non-zero width. As a conse- quence, both occupied as well as unoccupied states can be found in the impurity band above as well as below Fermi level even in the ground state. Therefore the subsystem of localized charge carriers in the ground state of SmB6 cannot be described by an energy distribution function expected for T = 0 K. This fundamental conclusion adds another reason for absence of resistivity divergence in SmB6 at lowest temperatures, and sheds new light on interpretation of experimental data obtained for SmB6 and similar systems at lowest temperatures.
SmB 6 is a prototypical mixed valence material revealing properties of a narrow-gap semiconductor down to temperatures of a few Kelvin [1][2][3][4]. Paradoxically, at lowest temperatures it exhibits metallic-like conductivity, which is moreover smaller than the Mott's minimum metallic conductivity [1][2][3]. Such temperature non-activated electrical transport of SmB 6 at lowest temperatures cannot be attributed to any scattering scenario known for metals, because extremely high value of the residual resistivity would unconditionally require superunitary scattering with unphysical concentration of scattering centers [1][2][3], at least 80 per unit cell [3]. This fundamental discrepancy between theory and experiment is a subject of long-standing controversy precluding to conclude whether ground state of SmB 6 is metallic, or insulating. In fact, according to Mott-Ioffe-Regel viewpoint conventional Boltzmann transport theory becomes meaningless if the characteristic mean free path of the itinerant conduction electrons becomes comparable to, or less than the interatomic spacing [5][6][7][8]. Therefore the requirement for superunitary scattering implicates that either (i) electrical conductivity is not homogeneous in the volume, i.e. material contains metallic regions forming a conductive path along the sample that is responsible for electrical conductivity at lowest temperatures [9][10][11][12][13] or (ii) electrical transport at lowest temperatures is realized via a hopping process, which however, has to be temperature non-activated [14].
The first mentioned approach is represented by scenarios supposing metallic surface in SmB 6 that could be either of topological nature [9][10][11][12] or due to "trivial" polaritydriven surface states [13]. Several experimental observa-Email address: batkova@saske.sk (Marianna Batkova) tions indicate metallic surface transport in SmB 6 [15][16][17] while there exists huge research effort to prove the existence of topologically protected surface states [18][19][20][21][22][23][24][25]. On the other hand, latest developments provide also evidence of trivial surface states in SmB 6 [26]. It has been also suggested that only the most stoichiometric SmB 6 samples possess a bulk gap necessary for the topological Kondo insulator state [27]. However, up to now, convincing conclusion about the nature of metallic surface states in SmB 6 is still missing.
The second mentioned approach is represented by a recently proposed model of valence-fluctuation induced hopping transport [14]. The model is based on the fact that valence fluctuations (VFs) of Sm ions are intrinsically accompanied by fluctuations of charge, ionic radii and magnetic moments of Sm ions, what unconditionally causes fluctuations of the energies of impurities in the impurity band of SmB 6 . As a consequence there are created favorable conditions for temperature non-activated hops (as in more details summarized below), resulting in temperature non-activated hopping transport [14]. The model intrinsically infers an enhanced conductivity of the surface layer because of higher concentration of lattice imperfections [14], thus resembling a characteristic property of topological insulators [14]. In association with this it should be emphasized here that although the scenario as presented in our previous work [14] has a capability to explain properties of SmB 6 without necessity to consider Kondo topological insulator (KTI) scenario, it does not exclude a possibility of existence of topologically protected surface states in SmB 6 . In fact, both of them can coexist, thus their features can be "mixed" in SmB 6 .
The purpose of this work is to point out that besides valence-fluctuation induced hopping transport as proposed before [14], another consequence of dynamical changes of the energies of impurities in SmB 6 due to VFs is that the energy distribution function of the charge carrier subsystem in the ground state corresponds to presence of occupied states above Fermi level and empty ones below Fermi level, thus resembling effect of thermal broadening. This implicates that temperature of charge carrier subsystem in the ground state of SmB 6 cannot be T = 0 K. For convenience of further discussion let us first recapitulate essential ideas of the original model of valencefluctuation induced hopping transport [14]. In a semiconductor with an impurity band (IB) that contains metallic ions in two different valency states, say M e 2+ and M e 3+ , energy of an impurity is affected by (re)distribution of M e 2+ and M e 3+ ions in its vicinity (because of different interaction energy between the impurity and M e 2+ /M e 3+ ion due to different charge, ionic radii, and magnetic moment) [14]. Considering a hypothetical rearrangement process (RP) causing repetitious changes in distribution of M e 2+ and M e 3+ ions on metal ion positions with a characteristic time constant, t r , this RP unconditionally causes also repetitive changes of the energies of impurities with the same characteristic time constant [14]. Therefore, the energy E i of the impurity i in the impurity band is not constant in time, but varies within the interval E i,min ≤ E i ≤ E i,max [14]. Assuming that Fermi level, E F , lies in the IB formed by one type of impurities (e.g. donors) and is characterized by a constant density of states (DOS), g(E) = g(E F ), and supposing that dynamical changes of the impurity energies are adequately characterized by a typical width of the energy interval E i,max − E i,min = ∆E i ≈ E 0 , then there must exist a subnetwork of impurities i * with energy E i * satisfying the condition [14] The concentration of impurities in this subnetwork is N * = E 0 g(E F ), thus a typical distance between two nearest impurities is R * = [E 0 g(E F )] −1/3 [14]. The impurities defined by the inequality (1) have a unique property: rearrangements of M e 2+ /M e 3+ ions in the vicinity of these impurities cause that some occupied impurity energy levels can shift from the region below E F to the region above E F , and analogously, some empty ones from the region above E F can shift under E F [14]. Such processes, driven by the RP create favorable conditions for electron hops (tunneling) to empty sites of lower energy, i.e. hops with zero activation energy, as depicted in Fig. 1. It was shown [14] that probability of RP-induced hops, P * rp , at sufficiently low temperatures can be expressed as P * rp ∝ ν * rp e −2αR * , where α −1 is a localization length, and ν * rp is defined as the time-averaged probability per unit volume that for a temporarily occupied site belonging to the subnetwork defined by the inequality (1) there will appear an empty state of lower energy in the subnetwork as a consequence of the RP. Accepting the fact that valence fluctuations (VFs) represent a special case of the RP, probability of valence-fluctuation induced hop, P * vf can be expressed as where ν * vf is defined analogously as ν * rp [14]. Direct consequence of Eq. (2) is that hopping probability exponentially increases with decrease of hop distance R * , what implicates an enhanced conductivity in the vicinity of the surface because of higher concentration of lattice imperfections which can play role of hopping centers [14]. Now let us focus on another fundamental impact of the RP. Due to the RP energy levels of impurities defined by the inequality (1) repetitively pass through the Fermi level, resulting in possible temperature non-activated electron hops, as depicted in Fig. 1. In the simplest case, when |E j −E i | is of the order of the Debye energy or smaller, and kT is small compared to |E j − E i |, the intrinsic transition rate γ ij for an electron hopping from a site i with energy E i to an empty site j with energy E j is well approximated by the "quantum-limit" hopping formula [28] where k is Boltzmann constant, R ij is the distance between sites i and j, and γ 0 is some constant, which depends on the electron-phonon coupling strength, the phonon density of states, and other properties of the material, but which depends only weakly on the energies E i and E j or on R ij [28]. According to Eq. (4), the intrinsic transition rate of the electron hopping to a site of less energy decreases exponentially with increasing R ij . Because γ 0 is finite, γ ij must be also finite, so there is always non-zero time interval until an electron can hop (tunnel) to some empty site of less energy, while this time interval increases with increasing distance between sites. Therefore, if the RP is a ground state property of a material, finite γ ij implicates a non-zero probability of finding some occupied energy levels above E F in the ground state. Neglecting electron-electron interactions except that not more than one electron can occupy a single site, time averaged occupation number of site i in thermal equilibrium can be expressed in form [28], what for T = 0 K unconditionally requires that all sites having energy below E F are occupied and all sites having energy above E F are empty. Therefore, if there exist some occupied sites with energy levels above E F in the ground state of a material with the RP, then the charge carrier subsystem in the ground state of this material can not be at absolute zero. Because VFs can be regarded as a special case of the RP, analogous scenario as indicated above can be expected in valence fluctuating semiconductors, too. In principle, VFs are RPs with a short characteristic time constant t r . However, according to Heisenberg relation if t r is short enough, uncertainty of energy of the electron occupying site i (∆E i ≈h/t r ) may cause situation that energy of the electron occupying site i becomes indistinguishable from the energy of empty site j, to which the electron "is going" to hop. This limit case has no essential impact on the valence-fluctuation induced hopping transport as proposed before [14], such as hops between sites i and j with indistinguishable energies can also be treated as hops/tunneling not requiring thermal activation, however, it becomes crucial in discussion on possible effect of VFs on the energy distribution function of localized charge carriers in the ground state of valence fluctuating materials, such as SmB 6 .
The charge fluctuation rate in SmB 6 estimated from phonon spectroscopy studies [29,30] is 200 cm −1 − 650 cm −1 , what corresponds to the characteristic time τ cf (which defines time changes of the impurity energy due to changes of the Coulomb interaction between the impurity and the surrounding Sm 2+ /Sm 3+ ions) of 5.1×10 −14 s − 1.7 ×10 −13 s. According to Heisenberg relation the uncertainty of the energy of electron occupying the impurity site can be estimated as ∆E ≈h/τ cf , what yields for SmB 6 ∆E between 4 meV and 13 meV. Because energies of the impurities fluctuating within the interval of E i,max − E i,min ≈ ∆E 0 lie in the impurity band of the width W IB that is located in the forbidden gap of the width E g , the inequality ∆E 0 ≤ W IB < E g is satisfied. The value of E g detected by many experimental techniques is between 2 and 20 meV in SmB 6 [1][2][3][4][31][32][33][34][35][36][37][38][39], what can be an indication that the above estimated uncertainty of the energy of electron occupying an impurity site (∆E) is greater or comparable with the width of the impurity band (W IB ). Based on this we suppose that the whole interval in which the energy of impurity fluctuates (because of fluctuations of local physical parameters which affect its energy) defines at the same time the energy uncertainty of electron occupying the impurity site (as a consequence of a characteristic time of fluctuations and uncertainty principle). Taking into account this supposition we present energy diagram model of impurity sites in SmB 6 as follows. Let us associate hopping sites in SmB 6 with donortype impurity states having fluctuating energy levels lying in the impurity band of SmB 6 . Each hopping site is characterized by energy interval of the width ∆E 0 defining energy broadening/uncertainty of this site. Thus we treat the hopping center as a narrow band of the energy width ∆E 0 , and we denote it as uncertainty band of the impurity energy (UBIE). The UBIE of a site i is defined by the partial DOS, g * i (E), which is non-zero and constant within the energy interval E i,min , E i,max and zero outside it, and by a time averaged probability of the occupation of site i by an electron, p i ∈ 0, 1 , as depicted in Fig. 2a. Expressing p i using a color scale the UBIE can be depicted as done in Fig. 2b, or in the most simplified form in Fig. 2c. How the situation in the impurity band of SmB 6 in very close vicinity of the Fermi level can look like show illustrations in Fig. 3A. Here the hopping sites are represented by UBIEs crossing the Fermi level (0 < p i < 1), and the partial DOS of the individual impurity, g * i (E), is shown as a bar of corresponding dimensions and position in the energy scale (see Fig. 3Aa). We consider that all UBIEs have identical properties except their position in energy scale, and that their distribution in the impurity band of SmB 6 is uniform. Thus, the total DOS, g * (E), being a sum of all partial DOS of individual impurity states, g * i (E), is constant (see Fig. 3Ab). The time averaged probability of the occupation of site i by an electron, p i , we define as the average value of the function f 0 (E i , T ) = 1/ {1 + exp[(E i − E F )/kT ]} over the energy interval E i,min , E i,max . At such conditions, p i = 1/2 for the UBIE centered at the Fermi level and decreases/increases at increasing/decreasing E i,min (and E i,max ). The time averaged probability of occupation of the state at energy E in the impurity band, f * (E), can be determined as average value of p i of all UBIEs with non-zero g * i (E). The behavior of f * (E), qualitatively depicted in Fig. 3A-c (and also visualized by color representation in Fig. 3A-b) is a continuous function of energy in a close vicinity of the Fermi level, having a finite slope ∂f * (E)/∂E at the Fermi level. This "energy broaden-ing" of f * (E) having its origin in VFs cannot be less than broadening/uncertainty introduced by VFs (expressed by the width of UBIEs, ∆E 0 ). An important point is that energy broadening induced by VFs will take place in SmB 6 also in the case if no thermal broadening can be expected This is a fundamental difference in comparison to a classic semiconductor at T = 0 K (see Fig. 3B), where all impurity energy levels below the Fermi level are occupied and all ones above the Fermi level are empty (as depicted in Fig. 3B-a and Fig. 3B-b), so the energy distribution function, f (E), is a step function (as depicted in Fig. 3B-c). Now let us consider an imaginary experiment. Let SmB 6 sample is in ideal thermal contact with a thermal reservoir at absolute zero, so the charge carrier subsystem of SmB 6 has to be in the ground state. Supposing presence of VFs in the ground state of SmB 6 , the charge carrier subsystem in the ground state of SmB where T CGS represents the temperature of the charge carrier subsystem in the ground state. Essential point is that T CGS has to be non-zero, such as zero value of T CGS would require that f * (E) is a step function, what as follows from the discussion above cannot be the case in SmB 6 . It can be therefore concluded that temperature of charge carrier subsystem in the ground state of SmB 6 cannot be zero, thus neither the electrical conductivity of SmB 6 in the ground state can be zero, such as zero electrical conductivity is the property of a semiconductor with hopping conduction only at T = 0 K. Consequently, although the SmB 6 sample is in ideal thermal contact with a thermal reservoir, which is at temperature T << T CGS , physical properties of SmB 6 that are governed by energy distribution function of charge carrier subsystem will reveal only negligible dependence on the temperature of the thermal reservoir, while a "crossover" to this regime will be observed at temperatures of thermal reservoir comparable with T CGS .
With regard to the discussion in the paragraph above it is worthy to mention that detailed resistivity studies of SmB 6 down to 15 mK reported by Kasuya and coworkers [2] have shown that the best fit of their experimental data down to 100 mK is represented by the formula σ(T ) = 7, 45 × exp [−19, 3/(T + 4, 87)] [Ω −1 cm −1 ], while the resistivity is nearly constant below 100 mK. Surprisingly, such fit corresponds to Arrhenius-type activation process with "shift" in temperature scale by 4,87 K. If we associate this temperature shift with the non-zero temperature of charge carriers in the ground state of SmB 6 , T CGS = 4, 87 K, Kasuya's result can be interpreted by means that at cooling SmB 6 sample, when temperature of the cooling system (thermal reservoir) approaches absolute zero the sample conductivity approaches the value expected for Arrhenius-type activation process at temperature of charge carriers of 4, 87 K. In addition, such estimated T CGS yields surprisingly good correspondence with the fact that plateau in the dc resistivity temperature dependence of SmB 6 samples of various qualities is observed below 5 K [1-4, 15-17, 23, 25, 31, 36, 38].
In association with impact of VFs on the ground state properties of charge carrier subsystem in SmB 6 it should be also emphasized that valence fluctuations of Sm ions are associated with fluctuations of their ionic radii, consequently causing corresponding lattice vibrations. In a "classic" crystal lattice at absolute zero there are no phonons [40], so here arises a question about energy distribution function of phonons, which assist to VFs in the ground state of SmB 6 . A non-zero thermal coupling between the lattice and charge carrier subsystem having a non-zero temperature in the ground state implicates a non-zero ground state temperature of the phonon subsystem, too. However, discussion of this issue is beyond the scope of this paper.
It can be concluded that discussed effect of valence fluctuations on the energy distribution function of localized charge carriers yields alternative explanation for the resistivity saturation in SmB 6 at lowest temperatures, while one does not exclude possibility of existence of metallic surface states in this material (e.g. topologically protected [9][10][11][12] or polarity-driven "trivial" surface states [13]). Moreover, this effect allows us to explain also the ground state properties of nonstoichiometric SmB 6 samples. It would be worthy to emphasize here that it was recently suggested that the presence of even 1 % of Sm vacancies leads to a smearing of the bulk hybridization gap in SmB 6 samples and would result in a breakdown of the KTI state [27]. Thus, only the most stoichiometric SmB 6 samples possess a bulk gap necessary for the topological Kondo insulator state [27]. However, recent electrical resistivity studies of vacant samples Sm 1−x B 6 , for x ranging from 0.03 to 0.2, reveal resistivity saturation at lowest temperatures similar to the stoichiometric SmB 6 [41,42]. Because concentration of Sm vacancies in these samples substantially exceeds level of 1 %, topologically protected surface states should not be present there, what indicates a need for an alternative scenario to KTI allowing to explain the resistivity saturation in such nonstoichiometric SmB 6 samples.
In fact, here discussed effect of valence fluctuations on the ground state properties of SmB 6 brings new light into the problem, and can help to resolve the puzzle, indicating following possibilities of realization of the ground state in SmB 6 : (i) Considering "ideal quality" SmB 6 single crystals without an impurity band or highest quality single crystals with very small impurity band and Fermi level not lying in the impurity band, the bulk should be insulating in the ground state (as there are no impurity states fluctuating around Fermi level, hich could be responsible for the phenomena discussed in this paper). However, because there are still present lattice imperfections in the near-surface region (or at least unterminated Smand B-bonds on the surface) causing increase of density of states in the impurity band, Fermi level in the nearsurface region is supposed to lie in the impurity band, thus the ground state energy distribution function of localized charge carriers in the near-surface region reveals energy "broadening", resembling effect of thermal broadening at non-zero temperature. This infers that near-surface region of highest-quality SmB 6 samples is electrically conductive in the ground state. If, in addition, topologically protected surface states or/and polarity-driven "trivial" surface states are formed there, the (electrically conductive) near-surface region in the ground state of SmB 6 contains also a metallic region/surface. On the other hand, (ii) in "real world" high-quality SmB 6 single crystals with low concentration of lattice imperfections, the impurity band containing Fermi level is present in accordance with experimental observations. Thus, the energy distribution function of localized charge carriers in the near-surface region, as well as in the bulk, reveals energy "broadening" in the ground state. This infers that the near-surface region, as well as bulk of the material, is electrically conductive in the ground state. If, moreover, topologically protected surface states or/and polarity-driven "trivial" surface states are formed there, the near-surface region of such "real world" high-quality SmB 6 crystals contain also a metallic (sub)region. Finally, (iii) in SmB 6 with reasonably high concentration of lattice imperfections, the impurity band containing Fermi level with correspondingly higher DOS can be supposed to be present. Thus, the ground state energy distribution function of localized charge carriers in the near-surface region, as well as in the bulk, reveals energy "broadening", analogously as in the previous case. Therefore, the near-surface region, as well as the bulk of the material is expected to be electrically conductive in the ground state, while conductivity in the near-surface region is enhanced. Deviation from the stoichiometry can cause that topologically protected surface states are not formed, what however still does not exclude possible presence of metallic surface states, e.g. polarity-driven "trivial" ones. If metallic surface states do not exist there, the near-surface region will just reveal enhanced conductivity, as it follows from the scenario of valence-fluctuation induced hopping transport [14].
Finally, we would like to emphasize that similar effects as described here for SmB 6 we expect also in other related materials with valence-fluctuating ground state, e.g. YbB 12 . According to our opinion, the absence of resistivity divergence at lowest temperatures is a fingerprint of this class of materials, being at the same time a proof of valence fluctuating ground state. We believe that here proposed scenario not only represents a base for understanding the underlying physics in valence fluctuating semiconducting compounds at temperatures close to absolute zero, but it also indicates a necessity to consider similar phenomena in many other materials with "dynamical ground state", especially those obeying physical properties which can not be adequately understood presumably supposing the ground state being at absolute zero. This work was supported by the Slovak Scientific Agency VEGA (Grant No. 2/0015/17). | 2017-06-06T07:12:18.000Z | 2016-08-08T00:00:00.000 | {
"year": 2016,
"sha1": "92b04d8f62ccbcf6864b16a1321ec0f341d1ba4b",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "92b04d8f62ccbcf6864b16a1321ec0f341d1ba4b",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
4375291 | pes2o/s2orc | v3-fos-license | Long-Term Temporal Trends of Nosema spp. Infection Prevalence in Northeast Germany: Continuous Spread of Nosema ceranae, an Emerging Pathogen of Honey Bees (Apis mellifera), but No General Replacement of Nosema apis
The Western honey bee (Apis mellifera) is widely used as commercial pollinator in worldwide agriculture and, therefore, plays an important role in global food security. Among the parasites and pathogens threatening health and survival of honey bees are two species of microsporidia, Nosema apis and Nosema ceranae. Nosema ceranae is considered an emerging pathogen of the Western honey bee. Reports on the spread of N. ceranae suggested that this presumably highly virulent species is replacing its more benign congener N. apis in the global A. mellifera population. We here present a 12 year longitudinal cohort study on the prevalence of N. apis and N. ceranae in Northeast Germany. Between 2005 and 2016, a cohort of about 230 honey bee colonies originating from 23 apiaries was sampled twice a year (spring and autumn) resulting in a total of 5,600 bee samples which were subjected to microscopic and molecular analysis for determining the presence of infections with N. apis or/and N. ceranae. Throughout the entire study period, both N. apis- and N. ceranae-infections could be diagnosed within the cohort. Logistic regression analysis of the prevalence data demonstrated a significant increase of N. ceranae-infections over the last 12 years, both in autumn (reflecting the development during the summer) and in spring (reflecting the development over winter) samples. Cell culture experiments confirmed that N. ceranae has a higher proliferative potential than N. apis at 27° and 33°C potentially explaining the increase in N. ceranae prevalence during summer. In autumn, characterized by generally low infection prevalence, this increase was accompanied by a significant decrease in N. apis-infection prevalence. In contrast, in spring, the season with a higher prevalence of infection, no significant decrease of N. apis infections despite a significant increase in N. ceranae infections could be observed. Therefore, our data do not support a general advantage of N. ceranae over N. apis and an overall replacement of N. apis by N. ceranae in the studied honey bee population.
The Western honey bee (Apis mellifera) is widely used as commercial pollinator in worldwide agriculture and, therefore, plays an important role in global food security. Among the parasites and pathogens threatening health and survival of honey bees are two species of microsporidia, Nosema apis and Nosema ceranae. Nosema ceranae is considered an emerging pathogen of the Western honey bee. Reports on the spread of N. ceranae suggested that this presumably highly virulent species is replacing its more benign congener N. apis in the global A. mellifera population. We here present a 12 year longitudinal cohort study on the prevalence of N. apis and N. ceranae in Northeast Germany. Between 2005 and 2016, a cohort of about 230 honey bee colonies originating from 23 apiaries was sampled twice a year (spring and autumn) resulting in a total of 5,600 bee samples which were subjected to microscopic and molecular analysis for determining the presence of infections with N. apis or/and N. ceranae. Throughout the entire study period, both N. apis-and N. ceranae-infections could be diagnosed within the cohort. Logistic regression analysis of the prevalence data demonstrated a significant increase of N. ceranae-infections over the last 12 years, both in autumn (reflecting the development during the summer) and in spring (reflecting the development over winter) samples. Cell culture experiments confirmed that N. ceranae has a higher proliferative potential than N. apis at 27 • and 33 • C potentially explaining the increase in N. ceranae prevalence during summer. In autumn, characterized by generally low infection prevalence, this increase was accompanied by a significant decrease in N. apis-infection prevalence. In contrast, in spring, the season with a higher prevalence of infection, no
INTRODUCTION
The Western honey bee Apis mellifera is a valuable generalist pollinator for many flowering plants in both natural and agricultural ecosystems. In agriculture, commercial pollination of crop plants, that depend on insect pollination for fruit set and seed production, is provided mostly by managed A. mellifera colonies which can, therefore, be regarded as productive livestock. The cultivation of pollinator-dependent crops is expanding all over the world; hence, there is an increasing demand for insect pollination in worldwide agriculture (Aizen et al., 2008Aizen and Harder, 2009). Although, this demand is partially met by a globally increasing number of managed honey bee colonies (Aizen et al., 2008Moritz and Erler, 2016), increasing problems with honey bee health resulting in severe honey bee colony losses pose a serious threat to human food security. Research of the last decade has identified a multitude of factors like pathogens, pesticides, and abiotic stressors being associated with unusually high and inexplicable losses of honey bee colonies (Genersch, 2010;Ratnieks and Carreck, 2010;Cornman et al., 2012;Pettis et al., 2013;Goulson et al., 2015). Among the pathogens studied and discussed in this context are two microsporidian parasites, Nosema apis (N. apis) and N. ceranae, (Cox-Foster et al., 2007;Higes et al., 2008;Genersch, 2010) which infect adult honey bees (Bailey, 1955).
Microsporidia are highly specialized, spore-forming fungi which are optimally adapted to an obligate intracellular parasitic life style (Keeling and Fast, 2002). Outside of host cells, microsporidia exist as metabolically inactive, infective spores. For N. apis and N. ceranae, the infection process starts with the ingestion of infective spores by an adult honey bee. The spores germinate in the midgut thereby extruding the polar tube. If the polar tube pierces a host cell, the sporoplasm is injected into the cell through the polar tube (Bigliardi and Sacchi, 2001;Franzen, 2005). Following the injection of the sporoplasm, it takes about 96 h until the first environmental spores are produced by an infected cell (Gisder et al., 2011). The spores are released into the gut lumen through cell lysis and leave the body of the infected host by defecation (Bailey, 1955;Bailey and Ball, 1991). Heavy Nosema spp.-infections of adult honey bees may result in dysentery (Bailey, 1967). Adult bees suffering from diarrhea will show abnormal defecation behavior, i.e., will defecate inside the hive, resulting in fecal spots on combs and frames. Nest mates cleaning these spots will ingest Nosema spp. spores and become infected (Bailey and Ball, 1991). Infections with Nosema spp. are widespread in honey bee populations. Most infected honey bees do not develop nosemosis and do not show any obvious symptoms like dysentery but may have an increased foraging or flight activity (Woyciechowski and Kozlowski, 1998;Dussaubat et al., 2013) despite impaired orientation and homing skills (Kralj and Fuchs, 2010;Wolf et al., 2014) and may have a suppressed immune system (Antunez et al., 2009;Chaimanee et al., 2012), as well as a reduced life span (Wang and Moeller, 1970;Malone and Giacon, 1996;Fries, 2010).
Initially it was thought that N. apis is specific for the Western honey bee A. mellifera (Zander, 1909), while its congener N. ceranae was described as a microsporidian parasite of the Eastern honey bee A. cerana (Fries et al., 1996), a native of South-and Southeast Asia. Although, experimental infection showed from the very beginning that N. ceranae can also successfully infect A. mellifera (Fries, 1997), it took nearly a decade until the first natural infections of A. mellifera colonies with N. ceranae were reported (Higes et al., 2006;Huang et al., 2007). It soon became evident that N. ceranae was not only much more widespread than expected in the global A. mellifera populations but that is was even the predominant species in many regions (Klee et al., 2007;Chen et al., 2008;Williams et al., 2008;Invernizzi et al., 2009;Chen and Huang, 2010;Yoshiyama and Kimura, 2011;Copley et al., 2012). Based on this epidemiological evidence it was suggested that N. ceranae is replacing N. apis in the honey bee populations worldwide. This process is thought to be driven by an asymmetric within-host competition between N. apis and N. ceranae favoring the spread of N. ceranae (Williams et al., 2014;Natsopoulou et al., 2015) although not all studies observed interspecific competition between N. apis and its congener N. ceranae (Forsgren and Fries, 2010;Milbrath et al., 2015).
However, a pan-European study on the prevalence of N. apis and N. ceranae reported that in South-European countries, such as Italy and Greece, N. ceranae had indeed practically replaced N. apis while this was not observed in Northern Europe (Ireland, Sweden, Norway, and Germany) (Klee et al., 2007). These data pointed to climatic factors differentially influencing assertiveness, establishment, spread, and, hence, prevalence of N. apis and N. ceranae. Experimental evidence exists showing that N. ceranae spores, but not N. apis spores, nearly lose their ability to germinate and, hence, their infectivity when exposed to temperatures close to or below freezing (Fenoy et al., 2009;Fries, 2010;Gisder et al., 2010). In addition, experimental infection of adult bees showed proliferation of N. ceranae-but not of N. apis-to be unaffected by temperatures above 33 • C . These data strongly argue for an advantage of N. ceranae over N. apis in warmer climates. In contrast, the cold-sensitivity of N. ceranae spores might slow down the replacement process in colder climates (Gisder et al., 2010), a hypothesis that could recently be substantiated by mathematical modeling of the replacement process when taking into account the parameters warmer and colder climate (Natsopoulou et al., 2015). However, long term epidemiological data on Nosema spp. prevalence allowing the observation of the spread of the emerging pathogen N. ceranae and evaluating the proposed process of replacement of N. apis by N. ceranae in a given honey bee population have been lacking so far. To fill this gap, we here present our results of a 12 year cohort study on the prevalence of N. apis and N. ceranae in Northeast Germany conducted on a cohort of about 230 honey bee colonies. The duration of the study, and the size of the cohort enabled us to statistically analyse the long term temporal trends in prevalence of N. apis-and N. ceranae-infections in the study area. We also show data from laboratory experiments substantiating our epidemiological data. We provide evidence that the continuous spread of N. ceranae and continuously increasing levels of N. ceranae-infection prevalence at population level not necessarily result in the replacement of N. apis.
MATERIAL AND METHODS
Bee Samples, Field Survey and Molecular Differentiation of N. apis and N. ceranae The data set on Nosema spp. prevalence comprises data from spring 2005 to autumn 2016, which were collected in the course of a 5 year longitudinal cohort-study on Nosema spp. epidemiology (Gisder et al., 2010) and of the still ongoing "German Bee Monitoring Project" . About 23 apiaries located in Northeast-Germany (Figure 1) participated in the projects with 10 colonies ("monitoring colonies") each. Monitoring colonies that collapsed during the study were replaced by colonies from the same apiary, if available by a nucleus colony made from the collapsed colony in the previous year. This procedure ensured that each apiary always contributed 10 monitoring colonies throughout the study period. Due to the long duration of the study, some fluctuation of participating apiaries could not be avoided. However, nearly half of the apiaries (11 of ∼23) participated for more than 9 years and six of them even for the entire duration of the study, i.e., 12 years; at least 20 bee keepers provided samples over a time period of consecutive 5-11 years (Figure 1). When an apiary dropped out, a similar apiary in terms of size, bee race, landscape, region, and history of losses and diseases was chosen as replacement and included in the study as soon as possible. This resulted in an annual mean of 22.67 ± 1.72 (mean ± SD) apiaries participating in spring and 24.0 ± 2.83 (mean ± SD) apiaries participating in autumn. All monitoring colonies were sampled twice a year, in spring and in autumn, resulting in a total of 5,600 honeybee samples collected and analyzed from the participating apiaries over the 12 year study period ( Table 1).
Sampling of bees as well as diagnosis of N. apis and N. ceranae were performed essentially as already described (Gisder et al., 2010). Briefly, from each apiary, a group of 10 bee colonies [annual mean: 10 ± 0.31 (mean ± SD) colonies in spring and 10.01 ± 0.14 (mean ± SD) colonies in autumn] was randomly selected at the beginning of the study or when the beekeeper entered the study and designated "monitoring colonies." From these colonies, bee samples were collected in spring and autumn each year and were stored at −20 • C until analysis. Spring samples collected end of March/beginning of April consisted of dead bees fallen onto the bottom board during the winter season (representing the bees that died over winter) to enable sampling of colonies that collapsed during the winter season (October to March) as well as of surviving colonies. Autumn samples collected in late September/beginning of October consisted of live in-hive bees taken from a super above the queen excluder thus ensuring that only the oldest bees (representing the most frequently infected bees) were sampled (Fries et al., 2013). Diagnosis of Nosema spp. infections was performed by microscopic examination of 20 homogenized bee abdomens according to the "Manual of Standards for Diagnostics and Vaccines" published by the Office International des Epizooties (OIE), the World Organization for Animal Health (Anonymous, 2008). The moderate sample Given are the total number of analyzed colonies per year and season as well as the numbers (n) and proportions (%) of colonies within each infection category.
size is adequate because the experimental unit is the colony (Doull and Cellier, 1961;Doull, 1965). Infection status of the colonies represents detectable levels of infection above 15% with 96% probability of detection (Fries et al., 1984(Fries et al., , 2013Pirk et al., 2013) which can be considered biologically relevant (Higes et al., 2008). For molecular species differentiation, Nosema spp.-positive homogenates were processed and analyzed via PCR-RFLP (restriction fragment length polymorphism) as previously described (Gisder et al., 2010). Results were further verified by re-analyzing randomly selected samples via a recently developed differentiation protocol which is based on the detection of speciesspecific sequence differences in the highly conserved gene coding for the DNA-dependent RNA polymerase II largest subunit. Based on the diagnostic results, four infection categories were defined: Microscopic analysis resulted in the category "Nosema spp." while molecular differentiation allowed for the categories "N. apis" (single infection), "N. ceranae" (single infection), and "co-infection" (infection with both N. ceranae and N. apis) ( Table 1).
Purification of Nosema spp. Spores for In Vitro-Infection
Honey bee colonies of the apiary of the Institute for Bee Research were screened for Nosema spp.-infections by microscopic analysis of 20 randomly collected adult bees (see above and Anonymous, 2008). Nosema spp.-positive samples were molecularly differentiated as previously described to identify samples either containing only N. ceranae or only N. apis spores. Purification of N. apis or N. cerane spores was exclusively performed with freshly sampled bees, because freezing or long-term storage affect spore viability and infection rate (Fenoy et al., 2009;Fries, 2010;Gisder et al., 2010). Midguts were carefully isolated from individual bees by using fresh forceps for each bee. Twenty midguts were pooled in 1.5 ml reaction tubes and spore purification was performed as already described (Gisder et al., 2010). Viability of the purified spores was checked via in vitrogermination. To this end, an aliquot of freshly isolated spores was air-dried onto glass slides for 30 min at room temperature. Germination was triggered by adding 20 µl of 0.1 M sucrose solution buffer directly to the dried spores. Germination process was analyzed under an inverse microscope (VWR, Darmstadt, Germany) at 400x magnification with phase contrast. Nosema spp. spores were counted in a hemocytometer (Neubauer-improved, VWR, Darmstadt, Germany) under an inverse microscope (VWR, Darmstadt, Germany) at 100x magnification. Only those spore preparations that were able to germinate under in vitro conditions were used for cell culture experiments.
In vitro-Infection of Cultured IPL-LD-65Y Cells
The insect cell line IPL-LD-65Y derived from the gypsy moth Lymantria dispar was obtained from the Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ, Braunschweig, Germany) and maintained for routine culture as given in the accompanying data sheet. For in vitro-infection of cultured IPL-LD-65Y-cells, aliquots of about 5 × 10 7 spores, purified as described above, were dried in 1.5 ml reaction tubes (Eppendorf, Hamburg, Germany) in a vacuum concentrator (Eppendorf, Hamburg, Germany) for 30 min at 30 • C. Subsequently, infection of IPL-LD-65Y cells with germinating spores was performed as previously described (Gisder et al., 2011). Briefly, IPL-LD-65Y cells were infected with freshly isolated N. apis or N. ceranae spores with a multiplicity of infection (MOI) of 20. Infected cells (100 µl with 2.5 × 10 5 cells/ml) were seeded in the cavities of six 96-well microtiter plates. N. apis-and N. ceranae-infected cells were incubated at 21 • , 27 • , or 33 • C. Infected cells were centrifuged on glass slides at the time points 24, 32, 48, 72, and 96 h post initial infection and were subsequently Giemsa-stained as described (Gisder et al., 2011). The number of meronts, sporonts, and mature spores of N. apis or N. ceranae was counted under an inverse microscope Eclipse Ti-E (Nikon Instruments, Düsseldorf, Germany) at 600x magnification in 10 individual cells for each time point as well as for each temperature and expressed as mean ± SD.
Statistical Analysis
For statistically analyzing the seasonality of Nosema infections, spring vs. autumn, the Wilcoxon signed rank test was used because the proportions of infected colonies (Table 1) were not normally distributed. In addition, the Spearman rank correlation was determined with R (version 3.2.5, R Development Core Team, 2016) to analyse the relationship between infection categories. The Spearman correlation coefficient determined the strength of the monotonic relationship between season and infection prevalence with effect sizes between 0.10 and 0.29 representing weak correlations, coefficients between 0.30 and 0.49 representing medium correlations, and coefficients of 0.50 or above representing strong correlations.
For each time point, the expected rate of co-infections (E co−inf ) was calculated as the product of the observed rates of single infections with either N. apis (R apis ) or N. ceranae (R ceranae ): E co−inf = R apis * R ceranae . Subsequently, the differences between the observed and expected rates of co-infections were calculated for each time point. Because those differences were normally distributed, a one sample t-test was used to check if these differences were significantly different to zero.
The statistical analysis of temporal trends was performed using RStudio (version 0.99.489) based on R using version 3.2.5. For visualizing infection prevalence data, dotplots were plotted with R, separately for spring and autumn. Generalized linear models (GLM) were fitted with lme4 (Linear Mixed-Effects Models, version 1.1-12) (Bates et al., 2015) for exploring the data set and visualizing the relationship between the dependent variables (Nosema spp.) and the independent variables (year). For statistical analysis of N. apis and N. ceranae prevalence over the 12 year study period we used mixed-effect binary logistic regressions analysis defining year as fixed factor and apiary as random factor to take into account the lack of independence of data within each apiary. Even after 12 years of sampling, the amount of data is still not sufficient to define colony as random factor to fully acknowledge relative data dependence. The sampling consisted of about 230 individual colonies per season, stratified within apiaries, and the prevalence of N. apis-, N. ceranae-, or coinfections at the individual level were analyzed with defining "0" if absent or "1" if present in each colony. Odds ratios (ORs) and 95% confidence intervals [CIs] were used to assess the strength of the associations.
For statistical analysis of the counted number of different developmental stages of Nosema spp. in infected IPL-LD-65Y cells, individual student's t-tests for each time point were performed followed by Benjamini-Hochberg correction (Benjamini and Hochberg, 1995). A p < 0.05 was considered significant for the statistical tests.
Prevalence and Seasonality of Nosema spp.-Infections
The huge data set on Nosema spp.-infection prevalence in Northeast Germany, which was generated during the 12 year longitudinal cohort study, provided a unique opportunity for a comprehensive analysis of the spread and success of Nosema spp., and especially of N. ceranae, in a restricted honey bee population. We first analyzed the seasonality of Nosema spp.-infections based on classical microscopic diagnosis without molecular species differentiation. The data revealed a clear and expected (Bailey and Ball, 1991) seasonality of Nosema spp.-infections for the whole duration of the study period with spring values being always higher than the autumn values of the same year and autumn values being always lower than the spring values of the following year (Figure 2A).
Molecular species differentiation of all Nosema spp.-positive samples enabled analysing the seasonality of N. apis-, N. ceranae-, and co-infections ( Table 1). The same seasonality as already observed for Nosema spp.-infections was also evident for N. apisinfections over the entire study duration despite for the time point "spring 2007" when less colonies where found infected with N. apis than in the preceding autumn 2006 and the following autumn 2007 (Figure 2B). With this exception for "spring 2007, " when only 3.1% of the colonies carried detectable N. apisinfections, the proportion of N. apis-infected colonies varied between 6.4% (spring 2013), and 18.7% (spring 2016). In autumn, the prevalence of N. apis-infected colonies ranged between 1.5% (autumn 2014) and 8.0% (autumn 2005).
For N. ceranae-infections, the described seasonality with higher prevalence in spring than in the following autumn and lower prevalence in autumn than in spring next year could be observed from autumn 2006 onward until spring 2016, whereas between spring and autumn 2016 the prevalence of N. ceranaeinfections did not decrease as expected but instead further increased (from 9.1 to 10.5%; Figure 2B). Spring prevalence from 2007 to 2016 varied for N. ceranae-infections between 7.6% (spring 2015) and 14.9% (spring 2007), while autumn prevalence ranged between 1.2% (autumn 2015) and 8.2% (autumn 2013).
Statistical analysis of the seasonality of Nosema spp.-, N. apis-, N. ceranae-, and co-infections using a Mann-Whitney test confirmed the above given, rather descriptive evaluation (for all infection categories, p < 0.01). Spearman correlation analysis further substantiated this finding (Figure 3). A strong negative correlation (coefficient values between −0.69 and −0.87) was found between season and all infection categories indicating that in each year and for all four infection categories (Nosema spp.-, N. apis-, N. ceranae-, co-infection) the infection prevalence decreased significantly from spring to autumn (for all infection categories: p < 0.01). Medium to strong positive correlations (coefficient values between 0.44 and 0.85) were found between the infection categories implying that all infection categories followed the same prevalence trend. For example, high infection prevalence for N. apis correlated with high infection prevalence for N. ceranae-or co-infections. This correlation was significant for all infection categories (p < 0.05). An interesting question in regard to co-infections was, whether or not the observed prevalence of co-infections in spring was congruent with the expected prevalence. To answer this question, we first calculated the rate of expected co-infections for each year from the rate of observed N. apis-and N. ceranaeinfections in this season. Comparing these values with the observed frequency of co-infections revealed that over the entire study period, the observed prevalence of co-infections was always significantly {one sample t-test; M = 0.037, [0.0195, 0.0546], t (23) = 4.6389, p < 0.01} higher than expected when assuming that the occurrence of co-infections was only influenced by the prevalence of single infections (Figure 4).
Temporal pattern of Nosema spp. Prevalence
For evaluating spread and assertiveness of the emerging honey bee pathogen N. ceranae, we analyzed the temporal patterns of N. ceranae-, N. apis-, and co-infections by plotting and statistically analysing the respective values separately for the spring (Figure 5) and autumn (Figure 6) seasons between 2005 and 2016. While the patterns for N. apis-and coinfections in spring did not show a consistent trend, the pattern for N. ceranae-infection prevalence suggested a continuously increasing trend over the years (Figure 5A). Generalized linear models (GLM) of the prevalence data confirmed this interpretation (Figures 5B-D). Logistic regression analysis (Table 2) demonstrated that the continuous increase in spring prevalence of N. ceranae-infections observed over the entire 12 year study period, i.e., between 2005 and 2016, was on average about 5% per year (Odd Ratio: 1.05 [1.01, 1.1]) and was significant (GLM, Likelihood Ratio test of the model, p = 0.02) (Figure 5B). This increase, however, was not accompanied by any significant (GLM, Likelihood Ratio test of the model, p = 0.95) change in the spring prevalence of N. apis-infections (Odd Ratio: 1.0 [0.96, 1.04]) ( Figure 5C). Likewise, no significant trends (GLM, Likelihood Ratio test of the model, p = 0.17) were observed for co-infections in spring (Odd Ratio: 0.96 [0.9, 1.02]) ( Figure 5D).
The dotplot of autumn prevalence of N. ceranae-, N. apis-, and co-infections ( Figure 6A) showed a different pattern with an increasing trend for N. ceranae-being accompanied by a decreasing trend for N. apis-infections. This finding could be substantiated by GLM-analysis and Likelihood Ratio tests of the models (Table 2, Figures 6B-D). In autumn, the prevalence of Figure 6D).
In vitro-Infection of IPL-LD-65Y Cells
To explain the obvious success of N. ceranae over N. apis in the studied honey bee population in summer, we experimentally analyzed the proliferative capacity of both microsporidian species in infected cells at temperatures between 21 • and 33 • C. To this end, we used an established cell culture model for N. apis and N. ceranae based on experimentally infecting cultured IPL-LD-65Y-cells. This insect cell line derived from Lymantria dispar had been shown to support replication of both microsporidian species (Gisder et al., 2011). Intracellular proliferation of N. apis and N. ceranae at three different temperatures (21 • , 27 • , and 33 • C) was evaluated by determining the number of the developmental stages per cell produced during merogony (meronts) and sporogony (sporonts/spores) of Nosema spp. (Figure 7A). The number of both meronts and sporonts/spores increased for N. apis as well as for N. ceranae over the observation time period of 96 h at all three tested temperatures. However, the number of the different developmental stages varied between N. apis and N. ceranae infected cells depending on incubation time and incubation temperature. At 21 • C, there was no significant difference in the proliferative capacity of N. ceranae and N. apis in infected cells for both meronts and sporonts/spores at all tested time points (24,32,48,72, and 96 h post-infection) (all p > 0.05) ( Figure 7B). However, at 27 • C and even more so at 33 • C, a higher proliferation rate and a faster proliferation of N. ceranae compared to N. apis could be observed. In infected cells which were incubated at 27 • C (Figure 7C), the number of meronts was not significantly different between N. apis and N. ceranae after 32 and 72 h post-infection (p > 0.05) but was significantly different at time points 24, 48, and 96 h post-infection (p > 0.05). More . Dark gray bars represent the number of meronts per cell (mean of 10 cells ± SD), light gray bars represent the number of sporonts/spores per cell (mean of 10 cells ± SD). Statistical analysis of the number of developmental stages was performed with student's t-tests for each time point and temperature. Statistical results given above the bars refer to the comparison of sporonts/spores produced by N. apis and N. ceranae (not significantly different: n.s., p ≥ 0.05; significantly different: *, 0.05 < p < 0.01; **, 0.01 < p < 0.001; ***, 0.001 > p ≥ 0.0001; **** p < 0.0001).
importantly, the number of counted sporonts/spores at 32, 48, 72, and 96 h post-infection was significantly higher (p < 0.05) in N. ceranae-than in N. apis-infected cells (Figure 7C). When the host cells were incubated at 33 • C, the number of N. apis meronts was significantly (all p < 0.01) higher than the number of N. ceranae meronts at 32, 48, 72, and 96 h post-infection ( Figure 7D) and the numbers of sporonts/spores were significantly higher at 32, 48, 72, and 96 h post-infection (all p < 0.01) in N. ceranae-infected host cells than in cells infected with N. apis.
Prevalence of N. ceranae Infections Follows the Same Seasonality as N. apis Infections
Nosema ceranae is an emergent pathogen of the Western honey bee A. mellifera. Its first detection in colonies of A. mellifera dates back a decade (Higes et al., 2006;Huang et al., 2007), although, it obviously switched host from A. cerana to A. mellifera about 40 years ago (Teixeira et al., 2013) and is now endemic in the global A. mellifera population. Several differences between N. ceranae and its congener N. apis have been reported, including differences in virulence, in seasonality of infections, and in temperature dependence of spore germination and biotic potential. Most of these differences seem to work in favor for N. ceranae resulting in its continous spread in the bee population and a supersession of N. apis in many regions (Klee et al., 2007;Chen et al., 2008;Williams et al., 2008;Invernizzi et al., 2009;Chen and Huang, 2010;Stevanovic et al., 2011;Yoshiyama and Kimura, 2011;Copley et al., 2012). However, studies claiming lack of seasonality of N. ceranae or replacement of N. apis in a given honey bee population are rather short-termed studies rarely performed over more than 2 years and most often involving only a limited set of samples. Our epidemiology data based on observing a cohort 230 bee colonies sampled twice a year over 12 years now revealed a different picture at least for the study area.
Nosema apis-infections are known to follow a seasonal pattern with spring prevalence being higher than autumn prevalence. This seasonality can be explained by the pathobiology of N. apis: (i) Only the spores of N. apis are infective; (ii) older bees are more likely to be infected and carry more spores; and (iii) spores are most efficiently transmitted through the fecal-oral route (Bailey, 1967;Bailey and Ball, 1991). Therefore, N. apis transmission within the colony is favored by conditions with low or no brood rearing and forcing adult bees to stay inside the hive for longer periods and to have close in-hive contacts (Bailey, 1967;Bailey and Ball, 1991). These conditions are regularly fulfilled during the winter months in climatic zones with winter temperatures falling below 10 • C not allowing bees to fly out (Winston, 1987). Instead, in these regions honey bee colonies hibernate by longlived adult winter bees forming a winter cluster around the queen bee and not leaving the hive for weeks or months until weather conditions allow cleansing and foraging flights and restarting brood rearing to replace the old winter bees (Winston, 1987). This explains why N. apis-infection levels increase over winter but normally decrease over summer when the rather shortlived summer bees are engaged in foraging, are able to defecate outside the hive, and when newly raised bees regularly replace older more heavily infected bees (Bailey, 1967;Bailey and Ball, 1991;Retschnig et al., 2017).
In contrast, N. ceranae-infections were described to lack this characteristic seasonality (Higes et al., 2006(Higes et al., , 2010 suggesting fundamental differences in pathobiology and preferred routes of transmission which would be interesting to investigate. To analyse this suggested lack of seasonality, we collected bee samples in spring and autumn without gap over 12 years from a cohort of around 230 honey bee colonies and analyzed all samples for the presence of Nosema spp. spores and performed molecular species differentiation in all Nosema spp.-positive samples. Surprisingly, the data clearly disproved that N. ceranaeinfections differ from N. apis-infections in regard to seasonality. Quite the contrary was true: All four infection categories, Nosema spp.-, N. apis-, N. ceranae-, and co-infections, followed the same seasonal pattern with spring prevalence of infection regularly being higher than autumn prevalence suggesting that N. ceranae and N. apis circulating in Northeast Germany are similar in regard to pathobiology and preferred transmission routes. Since reports on the lack of seasonality predominantly stem from South Europe (Higes et al., 2006(Higes et al., , 2010, further experimental studies are necessary to analyse whether the differences in seasonality between the Northern and Southern parts of Europe are due to climatic factors or intraspecies differences in N. ceranae. No Evidence for a General Advantage of N. ceranae Over N. apis and for an Overall Replacement of N. apis by N. Ceranae In many regions of the world, prevalence data collected for N. apis and N. ceranae indicated that N. ceranae has become the dominant species in the worldwide honey bee populations and it was suggested that N. ceranae has replaced or is about to replace its congener globally Martin-Hernandez et al., 2012). However, in Europe, a South to North gradient was observed with N. ceranae being dominant in Southern European countries already 10 years ago while at that time N. apis was still dominant in the Northern part of Europe (Klee et al., 2007) which might reflect an already discussed climatic aspect in N. ceranae spread and assertiveness (Fenoy et al., 2009;Martin-Hernandez et al., 2009;Gisder et al., 2010;Chen et al., 2012;Natsopoulou et al., 2015).
Congruent with this South to North gradient (Klee et al., 2007), at the beginning of our epidemiology study we observed very low levels of prevalence for N. ceranae-infections in Northeast Germany compared to N. apis-infections. This starting condition, the size of the cohort, and the design and duration of the study provided a unique opportunity to follow the spread of the emerging pathogen N. ceranae and analyse the impact of this spread on its congener N. apis, well established in the observed honey bee population. Our epidemiology data show that starting from a very low level, the prevalence of N. ceranaeinfections significantly increased continuously in the observed cohort of honey bee colonies during the last 12 years. This increase was true for both time points of sampling, in spring (showing the development over winter) and autumn (showing the development over summer) clearly indicating that N. ceranae became successfully established and expanded its presence in the honey bee population of Northeast Germany.
With regard to replacement of N. apis by N. ceranae, the obtained epidemiology data showed a complex picture. For assuming a replacement process at the population level, N. apis infection prevalence should have concomitantly decreased during the study period. However, a significant decrease in N. apis-infection prevalence was only observed for autumn indicating that during the bee season in summer N. ceranae successfully competed with N. apis at the population level over the course of the study. Surprisingly and in contrast to autumn, no significant change in N. apis infection prevalence was evident in spring despite a significant increase in N. ceranae prevalence. Therefore, no replacement of N. apis by N. ceranae in the honey bee population of Northeast Germany took place over winter during the last 12 years. Instead, the increase in N. ceranae prevalence in spring came on top of the unaltered N. apis infection prevalence suggesting that the two microsporidian parasites did not compete with each other over winter at the colony and population level. In addition, the long term stability of N. apis-infection frequency in spring indicate, that whatever mechanisms are acting on N. apis during summer and causing its decrease in the population, they are compensated for and reversed during winter preventing a supersession of N. apis through N. ceranae in the observed honey bee population.
Replacement of N. apis by N. ceranae at the population level during summer but not during winter points to different mechanisms acting on or influencing the two microsporidian parasites in summer and over winter. Although the exact mechanism responsible for presence (summer) and absence (winter) of replacement at the population level still remain elusive, experimental data providing explanations at the individual bee level for the increase in N. ceranae infection prevalence over summer exists. In a recent study by Martin-Hernandez et al. (2009), infection experiments with caged bees were performed at different temperatures and the "biotic index" was calculated for both microsporidia as the total N. apis or N. ceranae spore count per day after infection. This "biotic index" was higher for N. ceranae than for N. apis at 25 • C but no significant difference could be observed at 33 • C . Although these results did not provide convincing proof for an advantage of N. ceranae over N. apis during summer, they pointed into an interesting direction. Therefore, we extended the approach and performed infection experiments in cell culture (Gisder et al., 2011), which allowed a detailed analysis of the time course of proliferation and of the proliferative potential of N. ceranae and N. apis at different temperatures. Our in vitro results revealed a significant advantage of N. ceranae over N. apis at 27 • and 33 • C, the normal range of daily maximum temperatures in summers in Northeast Germany. At both temperatures, N. ceranae completed its replicative cycle faster and replicated more efficiently than N. apis. These results were in accordance with a recent study, suggesting a generally higher proliferation rate for N. ceranae compared to N. apis in experimentally infected, caged bees incubated at 30 • C for 20 days (Huang and Solter, 2013). Earlier and higher production of spores, which are transmitting the disease within and between colonies, may translate into higher infection prevalence at population level. These data explain an increase of N. ceranae infection levels, however, they still do not explain the observed replacement of N. apis by N. ceranae over summer.
For replacement of N. apis by N. ceranae, a simple increase in N. ceranae infection prevalence is not sufficient but a successful interspecies competition, with N. ceranae at least more often than N. apis winning the game, is necessary. Again, only experimental data at the individual bee level are available. Co-infection experiments with caged bees and simultaneous feeding of N. apis and N. ceranae spores did not provide evidence for intrahost competition between the two species (Forsgren and Fries, 2010;Milbrath et al., 2015). In contrast, sequential feeding of spores of the two species resulted in within-host competition: The first parasite significantly inhibited the growth of the second, regardless of species (Natsopoulou et al., 2015). This would have prevented the spread of N. ceranae because N. apis had been present in the bee population before N. ceranae arrived and would always have been first. However, this so-called "priority effect" proved to be asymmetric and N. ceranae exhibited a stronger inhibitory effect on N. apis than N. apis on N. ceranae (Natsopoulou et al., 2015). Mathematical modeling proposed that this priority effect will result in a successful replacement process at population level even when taking into account that the cold sensitivity of N. ceranae but not of N. apis spores (Fenoy et al., 2009;Gisder et al., 2010) provides a disadvantage for N. ceranae during cold winters (Natsopoulou et al., 2015).
However, for spring samples our epidemiology data clearly showed that although N. ceranae-infection prevalence increased over time, this increase did not result in a replacement of N. apis. Remarkably, N. apis-infection prevalence in spring remained rather stable over the 12 years study period although the autumn infection prevalence and, hence, the infection prevalence at the beginning of winter, has been declining during this period. Therefore, the two Nosema species rather not competed during winter and the mechanisms promoting the increase of N. ceranae in the studied honey bee population over winter did not influence the prevalence of N. apis. Furthermore, we observed higher than expected co-infection rates in spring suggesting that there is no interspecies within-host competition at colony or population level during overwintering. The co-infection levels rather suggested that an infection with any one of the two microsporidia pre-existing in a colony favored an additional infection of the colony with the other microsporidium. This is in contrast to the above mentioned report (Natsopoulou et al., 2015) showing interspecies withinhost competition with a priority effect favoring the spread N. ceranae over N. apis. However, this inter-species competition was shown at the individual bee level whereas our epidemiology data concern the colony and population levels. And indeed, at the colony and population level it is hardly conceivable how an N. ceranae infection of one bee or colony might inhibit a nestmate or a neighboring colony, respectively, to become infected by N. apis -and the other way round. Actually, the concept of interspecies within-host competition of an obligate intracellular parasite due to competition for the same limited cellular energy resources cannot easily be translated to the colony or population level where limitation or shortage of resources (in this case: new hosts) is not yet the problem. However, if the prevalence of N. ceranae-infections keeps increasing like it did over the last 12 years, within-colony and between-colony competition might become an issue once all colonies are infected with either one of the microsporidia. Therefore, a continuation of this study will further our understanding of the long term epidemiology and interspecies competition at population level of these two important honey bee pathogens.
AUTHOR CONTRIBUTIONS
EG and SG conceived and designed the study and the experiments. SG, VS, and LH carried out the experiments and the microscopic and molecular diagnosis of Nosema spp. SG, VS, and DG performed the statistical analysis. EG supervised all work, SG supervised the laboratory experiments, and DG supervised the statistical analysis. SG and EG wrote the paper. All authors revised the manuscript and approved the final version. | 2017-07-14T17:58:25.221Z | 2017-07-06T00:00:00.000 | {
"year": 2017,
"sha1": "a30b73873d6da74c47c82cacfef55748c2c384cc",
"oa_license": "CCBY",
"oa_url": "https://www.frontiersin.org/articles/10.3389/fcimb.2017.00301/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "a30b73873d6da74c47c82cacfef55748c2c384cc",
"s2fieldsofstudy": [
"Agricultural and Food Sciences",
"Biology",
"Environmental Science"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
12193871 | pes2o/s2orc | v3-fos-license | A bilinear elastic constitutive model applied for midpalatal suture behavior during rapid maxillary expansion
Introduction: This study aims to evaluate the influence of the biomechanical behavior of the midpalatal suture (MPS) during the rapid maxillary expansion (RME) when modeled by the Finite Element Method. Methods: Four simulation alternatives are discussed and, for each analysis, the suture is considered as a functional unit with a different mechanical behavior: (i) without MPS elements, (ii) MPS with Young’s modulus (E) equal to 1 MPa, (ii) MPS with E equal to 0.01 MPa and (iv) MPS with bilinear elastic behavior. Results: The stress analysis showed that, when MPS is not considered in the model, stress peaks are reduced in magnitude and their distribution is restricted to a smaller area when compared to the model with the inclusion of MPS (E=1 MPa). The increased suture stiffness also has a direct influence on MPS displacements after 30 expander activations. Conclusion: The consideration of the MPS in RME computer models influences greatly the calculated displacements between the suture bone ends, even as the stress levels in maxillary structures. Furthermore, as proposed for the described model, the elastic bilinear behavior assigned to MPS allows coherent prediction of stresses and displacements results, being a good representation for this suture overall behavior.
Introduction
Biomechanical studies using the Finite Element Method (FEM) are common in Orthodontics (Holberg and Rudzki-Janson, 2006;Provatidis et al., 2008;Serpe et al., 2014;Wang et al., 2012).Recent computational studies have evaluated stresses and displacements generated in the craniofacial complex due to rapid maxillary expansion (RME).This therapeutic technique basically consists of separating the bone ends of the suture (MPS) using a palatal expander -a device attached to the posterior teeth and supported or not on palatal mucosa.
According to clinical data, a total displacement of 7 mm applied by an expansion screw in 8 to 14 years old patient results in a 2.54 mm increase in the distance between the central incisor crowns, 36% of the expansion screw opening (Ballanti et al., 2010).In another study with patients of the same age span (Weissheimer et al., 2011), the authors noted a "V" shaped suture expansion after RME, with the anterior maxillary wide basis, representing around 50% (32.7-54.7%) of the applied displacement by the expansion device (8 mm) in the anterior region and 36% (27.2-39.2%) in the posterior region.The percentage of the total applied displacement that effectively is transferred to the MPS is one parameter used to access the effectiveness of the proposed procedure.
Biomechanical studies (Gautam et al., 2007;Iseri et al., 1998;Jafari et al., 2003;Lee et al., 2009;Provatidis et al., 2008) attempted to improve the understanding of this phenomenon and to evaluate how this orthodontic therapy affects other regions of the skull.But most of them considered only linear elastic constitutive model for the analysis, which potentially can result in a model that provides an improved representation of the expansion process.Iseri et al. (1998) used a tridimensional (3D) model of the skull with thick shell elements and considered all materials as linear elastic.Simulating RME, an expansion of 5 mm was applied on each half of the expander (overall 10 mm) and, as a result, a greater displacement in dentoalveolar area was obtained, which gradually diminished in upper structures as nasal cavity.High von Mises stress levels were observed in the jaw bone regions near the canines and molars (184.62 and 56.11 MPa,respectively), in the lateral wall of the inferior nasal cavity, in the nasal and the zygomatic bone.The highest stress peak was in the pterygoid plates of the sphenoid bone, near the skull basis (723 MPa).Using the same material properties and 10 mm of expansion, Jafari et al. (2003) observed the following results: 5.31 mm of maximum lateral displacement in the region of central incisors; 1.08 mm of maximum displacement for anterior bone region, located in the anterior edge of the nasal septum.The nasal bone, nasomaxillary suture, nasofrontal suture and anterior border of the frontal process of the zygomatic bone experienced von Mises stresses up to 321 MPa.In the frontal, parietal, temporal and occipital bones the stress levels were between 0 to 40 MPa.Gautam et al. (2007) also evaluated the stress distribution along the facial sutures with finite element analysis using tetrahedral solid elements with the aim to improve accuracy when compared with shell elements.Maximum von Mises stress were found along the frontomaxillary, nasomaxillary, and frontonasal sutures, followed by zygomaticotemporal and zygomaticosphenoid sutures.Tensile and compressive strains were observed along the same suture, as is the case for zygomaticotemporal, zygomaticofrontal and zygomaticomaxillary sutures.Isaacson and Ingram (1964) observed in a clinical study a phenomenon of stress relaxation that occurs in the craniofacial complex between successive activations of the expander screw, as seen in Figure 1.Also considering this phenomenon, and based on data obtained from a Finite Element (FE) study that compared displacement and stress distribution for different models, Provatidis et al. (2007), showed that the most accurate model for the study of RME treatment must contain all skull, complete maxillary dentition, including MPS and consider the phenomenon of stress relaxation that occurs between two successive screw activations.
In a subsequent study, Provatidis et al. (2008) compared the effects of different ossification degrees of the MPS on the craniofacial complex.The model closer to the clinical reality considered all the circummaxillary sutures completely ossified (with Young's modulus equal to the bone) and the palatine and transverse sutures unossified (E = 1 MPa, as proposed by Verrue et al. (2001)).It was found that the lacrimomaxillary, frontomaxillary nasomaxillary, transverse palatal sutures, as well as the suture between the maxilla and the pterygoid process of the sphenoid bone do not influence the result at the RME, while zygomaticomaxillary suture has an influence on its response.Furthermore, the maximum displacement was observed in the region below the hard palate from the central incisors to second premolar, which was dissipated in the frontal and parietal bone.In the occipital bone the displacement was zero.
According to a literature review study by Romanyk et al. (2013), the most common ways to represent the MPS in computer models were either to assume it as an empty space or as a material with the same elastic properties as the cortical bone (E = 13700 MPa) or to use elastic properties similar to those of soft tissues (E = 1 MPa).All studies in this review accepted linear elastic properties for MPS. Lee et al. (2009), comparing three FE models (maxilla without a midpalatal suture, maxilla with suture elements and maxilla without suture elements), suggested that different states of MPS may partly explain the differences in clinical studies of RME.
Considering the findings of previous works and aiming to reach more reliable results than linear elastic models, the present study intends to describe the stress distribution and displacements generated in maxillary complex by tooth-supported palatal expander considering the bilinear elastic behavior of the MPS using FE analysis.It also evaluates the influence of the constitutive model used to model the suture for computational models of the RME technique.
Methods
For computational simulations, a three-dimensional model (Figure 2) consisting of the maxillary bones and the bones closely related to them was developed.The model was based on Cone Beam Computed Tomography (CT) images of a 12 years old patient and this study has been approved by the Research Ethics Committee of UFMG, Brazil (nº 171.147).The images generated by the CT in DICOM format were imported into the computer aided design software (CAD), Simpleware .For three-dimensional reconstruction, 218 cross sections, 285 sagittal and 231 coronal sections were used, with 0.45 mm spacing between them.Segmentation was done according to the tissues contrast, initially divided into bone, tooth and dental pulp elements.The bone was not separated between trabecular and cortical bone.After processing these sections, the generated 3D model was exported in STL format to Rapidform XOR3, software used for solid faces parameterization.The model was then exported in IGES format for SolidWorks 12, another CAD software, to refine the geometry.Considering the model symmetric, it was half-sectioned to reduce the complexity and the number of nodes of the FE model.
The structures that compose the model were considered as isotropic linear elastic, given the limitations in representing the bone biomechanical behavior, and as an initial study of the skull stresses.The real mechanical behavior of the tissues is time dependent, and this characteristic should be further studied to develop a more realistic material model where stress relaxation effects can be considered.The considered mechanical properties, Young's modulus (E) and Poisson's ratio, were as follows: Bone, 10000 MPa and 0.3 (Serpe et al., 2014);Teeth, 20000 MPa and 0.3 (Tanne et al., 1989); Periodontal Ligament, 0.69 MPa and 0.49 (Yoshida et al., 2001); Expander/Steel 200000 MPa and 0.33 (Hibbeler, 2002), respectively.The Young's modulus assigned to the bone (10000 MPa) was estimated in terms of the proportion of cortical bone (13700 MPa) and trabecular bone (1370 MPa) present in the cranium (Serpe et al., 2014), since the model used was composed of a single type of bone tissue aiming to reduce the number of mesh nodes.The consideration of the bone as a homogeneous material is based on an homogenization with the purpose to reduce the computational cost of the analysis, considering that the main focus of the model was to examine the effect of different assumptions for the MPS.For each analysis, the functional unit representing the suture was considered with different mechanical behaviors to evaluate how it influences the results of the simulations.Model (i) represented the maxilla without MPS (MPS unit was removed) while for the remaining models the MPS had different constitutive properties.In models (ii) and (iii), MPS was considered linear elastic, with E equal to 1 MPa and 0.01 MPa, respectively, and Poisson's ratio of 0.49.In addition to these hypotheses, the suture unit was considered to have a bilinear elastic behavior for Model (iv), (E initial = 1 MPa, threshold stress (transition) at 0.1 MPa and E final = 0.01 MPa).
The bilinear model was used to provide a more realistic representation.This model represents the difference in suture stiffness before and after its breakup/separation.The initial elastic modulus (1 MPa) represented the organized connective tissue behavior, which is its main suture component (Provatidis et al., 2007).After its partial failure, the MPS has lower expansion restriction.Preliminary FE analyses were done switching the threshold stress value between 0.5, 0.1 and 0.05 MPa.Based on preliminary FE displacement results and analyzing the parameters to adjust the bilinear model simulations to clinical results (Ballanti et al., 2010;Weissheimer et al., 2011), a stress value of 0.1 MPa was assigned as the threshold for changing the elastic modulus.This represents the stress level at which the disorganization of the MPS functional unit occurs.After reaching this limiting value, the elastic modulus is reduced (E final = 0.01 MPa) to represent an unorganized connective tissue.All these assumptions were made to represent the MPS opening as observed in clinical studies found in the literature (Ballanti et al., 2010;Weissheimer et al., 2011).
The mesh was generated using ANSYS 14.5 and resulted in a model with 462,916 tetrahedral elements and 749,935 nodes.The boundary conditions were applied in the same region where the symmetry was used to restrict lateral movement (Figure 3).In the posterior region of temporal bone, total or partial movement restrictions were applied, to represent the constraints imposed by the skull base.This assumption was done considering that skull bones have displacements limited by the anterior cranial base structures.As in the present work the model is not represented by the full skull, some areas were restrained based in the displacement results obtained by clinical studies (Ballanti et al., 2010;Weissheimer et al., 2011).A partial displacement restriction (yellow areas in Figure 3, marked as "B", "C" and "E") was applied to only Y-axis.All the constraints were imposed only in the posterior region of the model ("D" area in Figure 3), as far as possible from the face, in order not to influence the stress distribution/results in the anterior region of the skull (face).
The medial face of the expander unit was chosen to apply the displacement on the horizontal axis (X axis), as performed by the expander screw ("A" area in Figure 3), and also to restrict the displacement in the vertical and anteroposterior axes, as required by the whole appliance structure.The offsets applied in the simulations were 0.125 mm to one side (overall 0.25 mm -an activation of the unit) and 3.75 mm, simulating the average total displacement in RME (overall 7.5 mm -30 activations).The simulations were performed on the left half of the model considering its symmetry and the results mirrored for visualization.
Results
For all results of tensile and compressive stresses occurring at the bone, the detected maximum (peak) values, shown in the color charts on the following figures, are located in regions where the boundary conditions were applied (posterior region of the model, Figure 3).As it is not a region of interest for this study, these points were not indicated on the following images.
To evaluate the influence of MPS on stress distribution in the initial stages of RME, models (i) and (ii) were simulated with only one palatal expander activation (0.125 mm).As these first models considered the assumption of linear elastic behavior for the MPS, common in the literature (Romanyk et al., 2013), which is a great simplification considering the relaxation present in connective tissues, this analysis aimed to reduce the simulation error, especially for large applied displacements.The results showed that, when MPS was removed from the model (i), the peak of maximum principal stresses was in the range of 1 MPa, close to the orbit and the lingual surface of the alveolar bone, near the first molar (Figure 4).When MPS was considered (Model (ii)), a region of maximum stresses in alveolar bone was observed at buccal and lingual sides close to the teeth used as support, at the beginning of the zygomatic process of the maxilla, nasal floor and internal walls of the nasal cavity, resulting in peaks from 1.5 to 2 MPa.
Regarding minimum principal stresses, there was also difference between models with and without MPS.With the presence of the suture, stress peaks were located in the medial wall of the orbit increased, ranging from 1.5 to 2 MPa (Figure 4).
Dental elements used as support were also evaluated mechanically.For a single screw activation of the tooth-supported expander, results showed tensile stress peak of 10 MPa in the middle third of the first molar palatal root and -11 MPa compressive stress peak near the fulcrum region between the first molar roots (Figure 5).The reaction force found when applied one activation on expansor' medial surface was 25 N.
To assess the stresses generated by the entire displacement applied by the expander (3.75 mm on each side, overall 30 activations), when the MPS has a lower expansion restriction after its partial failure, a bilinear elastic model was adopted (Model (iv)).
The results showed that the maximum principal stress is distributed throughout the orbit (in the anterior and inferior border), canine fossa of malar surface, buccal face of alveolar bone (close to first molar and first premolar), palatal side of the alveolar bone, lateral wall of the nasal cavity, nasal bone and extended along the superior medial orbital wall with peaks of 50.5 MPa.The nasal floor experienced tensile stresses along its length with peaks up to 22.5 MPa.For minimum principal stress, also shown in Figure 6, peaks up to -56.5 MPa were observed in the lower region of the medial orbital wall.In the canine fossa region of the maxillary bone and extending to the zygomatic arch, peak stresses reached -26 MPa.At the nasal floor and the palatal area, results showed compressive stresses in median regions with values between -8.5 and -11 MPa.
When the bilinear behavior for MPS was considered, there was a significant increase in processing time.An Intel (R) X5670 computer, with 2 GHz processor and RAM of 24 GB required approximately 144 hours to complete the simulation, a high operational cost when compared to the linear model, which required an average of 5 minutes of processing time.
The expander showed a large dental effect due to RME.This can be observed in Table 1, which compares the displacement predicted in MPS (desirable skeletal effect of RME) with the one predicted for the tooth occlusal surface.The ratio between the MPS opening and displacement of dental crown is close to 1:3, at premolar and molar regions.
Models (i) and (iii), without and with (E = 0.01MPa) MPS elements respectively, were also simulated when subjected to the total displacement.The simulation resulted in displacement between bone ends of MPS and it was compared to the bilinear model in Table 2.
The higher rigidity of the suture, in the initial region of the bilinear model, reduces the total amount of the suture opening, when compared to cases where only the elastic modulus suggesting disorganized tissue was considered (E = 0.01 MPa).This was also the case when comparing the results with the ones obtained from the model without MPS elements.It was observed that the presence and stiffness of the sutures quantitatively influence the removal of the MPS bone ends, as well as the considered mechanical behavior.
Discussion
The influence of MPS in the FE model was assessed by comparing the stress distribution results, in the presence or absence of MPS elements and also in relation to the mechanical behavior assigned to it.The results obtained from this computational study show that an adequate MPS constitutive model is necessary to analyze RME by FEM, which is in accordance with other studies that suggested a significant influence of MPS on the stresses and strains of adjacent structures (Provatidis et al., 2008;Romanyk et al., 2013).
When the MPS elements were suppressedsimulating surgically sectioned suture -, the stress values were considerably lower (Figure 4).But only when the goal is to simulate the RME assisted surgically, a model without the MPS elements can realistically represent the clinical situation.For the remaining situations it would be required, ideally, carrying out mechanical tests for the determination of a reliable constitutive model.However, in the absence of this ideal constitutive model that should consider the viscoelastoplastic behavior for MPS and damage, a bilinear behavior was proposed in this study.
The bilinear behavior is represented in a constitutive model that has a threshold stress value after which the elastic modulus changes.In this case, when this value is achieved, the elastic modulus for the MPS decreases radically, representing the stress level on MPS before partial rupture.The threshold stress value was estimated by a fitting obtained after repeated FE simulations.For this preliminary tests, it was assigned between 0.5, 0.1 and 0.05 MPa to threshold and the value of 0.1 MPa was chosen as it led to the predicted displacements results for the MPS bone ends (Table 2) found in clinical studies (Ballanti et al., 2010;Weissheimer et al., 2011).
According to some clinical studies, the suture opening depends on the skeletal maturation and usually occurs in a "V" shape, with the fulcrum in the posterior region (Bishara and Staley, 1987;Weissheimer et al., 2011).The displacements variations found in clinical studies were in the range of 32% to 45% in the palatal anterior region and 12 to 32% in the palatal posterior region (Ballanti et al., 2010;Weissheimer et al., 2011) and this displacement was used as a parameter to validate the computational models.The removal of the suture, related to the total displacement applied, varied according to the behavior assumed for MPS (bilinear, linear and absent).The large variation in the displacements observed when changing tissue behavior from linear to bilinear agrees with the statement that different mechanical behaviors of the suture are related to the differences in clinical RME outcomes, as suggested by Lee et al. (2009).This assumption could be extrapolated to clinical results and propose that different ossification stages and imbrications levels of the MPS, that affects the stiffness of the suture, may partly explain the variability of the measures for palatal opening in clinical RME studies.
All models in the present study predicted the "V" shaped opening of MPS (Table 2) but the bilinear model showed smaller displacement than the other cases, probably because this model considers the change in stiffness that occurs in MPS after its rupture.The difference observed on results enlightens the effect of the simplification considered when linear elastic properties were applied for MPS.This fact was sustained also by the stress predicted for the FE models, as biomechanical studies (Gautam et al., 2007;Jafari et al., 2003) that point out to high stresses in various skull bones.
The peak of von Mises stresses up to 321 MPa were observed for nasal bone, nasomaxillary suture, nasofrontal suture, anterior border of frontal process of the zygomatic bone (Jafari et al., 2003).Gautam et al. (2007) obtained von Mises stresses up to 459 MPa in frontomaxillary, nasomaxillary, and frontonasal sutures.However, these results do not represent quantitative reliability due to model simplifications, such as the use of shell element discrete model requiring a simplified geometry (Jafari et al., 2003) and absence of periodontal ligament and linear elastic properties for sutures and bone.As previously mentioned, the bone partially relaxes between expander activations, which makes it difficult to compare the values generated by the sum of displacements directly applied in a model with linear elastic material behavior.
Despite of displacements, the bilinear model showed considerable lower stress peaks on skull bone.To simulate the complete expansion promoted by RME, the bilinear model could generates more acceptable results for predicted MPS removal bone ends and the maximum principal stresses predicted was close to 50 MPa distributed throughout the orbit, canine fossa, alveolar bone (close to, nasal cavity and nasal bone (Figure 5).This model also kept a distribution pattern similar to other studies in the literature (Jafari et al., 2003;Provatidis et al., 2008), reasserting that forces applied by the expander devices are not only restricted to the intermaxillary suture.
The maxillary complex behavior during RME was describe by Isaacson and Ingram (1964) when a gauge was adapt in the palatal expander and showed that forces produced by the expander are not cumulative, it reduced significantly between activations and remain throughout treatment within the range of 9 N and 45 N (Figure 1).This result suggests that the maxillofacial complex absorbs the displacement imposed by palatal expander between activations and the linear model could not be a good option for FEM.Aiming to obtain stress values consistent with the limitations of the linear model, only the efforts of first expander activation was predicted when this constitutive behavior have been applied (Model (ii)).However the linear model showed a distribution pattern of maximum stresses similar to other studies in the literature (Jafari et al., 2003;Provatidis et al., 2008), such as nasal bone, zygomatic arch and nasomaxillary suture, but did not exceed 6 MPa for these regions (Figure 4) when only one activation was applied.This value could be more reliable considering the non-accumulative stress by MPS.
Assuming the force range assessed on palatal expander (Isaacson and Ingram, 1964) and as one more FEM validation, the reaction force of bilinear model (Model (iv)) was requested.The value predicted on the expander appliance for one activation was 25 N, it belongs to clinical range of force evaluated in the palatal expander.Finally, a good representation of the cases studied was obtained, even though the used model did not contain a complete representation of the skull.
Considering the limitation from FE analysis, the obtained results proved that the simplification of complex behavior of some materials has an important influence on stress analysis.In addition, some final remarks can be made: 1.The presence and stiffness of the sutures influence the displacement between the MPS bone ends and the stress provided computationally in maxillary structures; 2. The high level of stresses promoted over deeper craniofacial skeleton structures demonstrates that expander devices effects are not only restricted to the intermaxillary suture, but also distributed in other sutures of the nasomaxillary complex, and showed the role of the entire circummaxillary complex in the RME; 3. For this FEM, the elastic bilinear behavior assigned to MPS allows coherent predicted stress and displacement results similar to available clinical data.
Considering the predicted results for MPS with bilinear properties were more consisted than linear model and it has less computational time than viscoelastic model, the bilinear model should be a good option to represent the MPS on FEM for rapid maxillary expansion.
Figure 1 .
Figure1.Results of force measurement for each expander activation according toIsaacson and Ingram (1964).
Figure 3 .
Figure 3. Finite element model with mesh (left) and the boundary conditions applied in the model (right).The areas were divided in A: region where the displacement related to the rapid maxillary expansion was applied; B, C and E: regions with partial restriction of movement; D: total motion restriction.
Figure 5 .
Figure 5. Maximum and minimum principal stresses distribution in the dental elements when applied one activation on tooth-supported expander (Model (ii)).
Figure 6 .
Figure 6.Maximum and minimum principal stresses distribution when total displacement is applied by tooth-supported expander appliance (Bilinear model for midpalatal suture -Model (iv)).
Table 1 .
Displacement results predicted for midpalatal suture (MPS) and dental crown areas with bilinear model for MPS.
Table 2 .
Predicted displacement results to midpalatal suture (MPS) for different mechanical behavior (% of the total displacement applied by the expander). | 2017-09-18T13:55:09.005Z | 2015-11-27T00:00:00.000 | {
"year": 2015,
"sha1": "f153b19e7d1ccdcec4ee36100c2e9c89e69d85f7",
"oa_license": "CCBY",
"oa_url": "http://www.scielo.br/pdf/reng/v31n4/2446-4740-reng-2446-47400637.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "f153b19e7d1ccdcec4ee36100c2e9c89e69d85f7",
"s2fieldsofstudy": [
"Engineering",
"Medicine"
],
"extfieldsofstudy": [
"Materials Science"
]
} |
18151352 | pes2o/s2orc | v3-fos-license | A Fault-tolerant Structure for Reliable Multi-core Systems Based on Hardware-Software Co-design
To cope with the soft errors and make full use of the multi-core system, this paper gives an efficient fault-tolerant hardware and software co-designed architecture for multi-core systems. And with a not large number of test patterns, it will use less than 33% hardware resources compared with the traditional hardware redundancy (TMR) and it will take less than 50% time compared with the traditional software redundancy (time redundant).Therefore, it will be a good choice for the fault-tolerant architecture for the future high-reliable multi-core systems.
Introduction
With the development of the semiconductor technology, integrating billions of transistors on a single die has been possible and the complexity of the SoC is increasing. Further, more and more processors and IP cores can be implemented on a single die to build a multiprocessor SoC. Along with this trend, the electronic devices are becoming more and more sensitive to external disturbs such as soft errors [1], and fault-tolerant architecture is always used to obtain the reliability.
For fault-tolerant architectures, hardware redundancy and software redundancy are the two popular methods and each kind can be classified further according to whether it's static or dynamic. FTMR (Coming from the N-Modular Redundancy method) is the popular static hardware redundancy method while N-version programming is the typical static software redundancy [2][3] [4]. During these methods, no matter static or dynamic, software redundancy has more flexibility while the hardware redundancy has less hardware resources needed. The proposed architecture in this paper takes a combination of these two methods to make full use of the multi-core system to realize the fault-tolerancy with both the timing and resources constraints taken into consideration.
Laterly, George A.Reis gives a software implemented fault tolerance mechanism named SWIFT with a enhanced control-flow checking mechanism based on the compiler technology [5].This mechanism is useful for software fault -tolerant, but do nothing with the other related hardware modules.
S.Tosun gives a reliability-centric hardware/software co-design framework to partition the hardware/software which takes reliability into consideration which is useful for the initial task scheduling phase [6]. This framework is useful for the initial hardware/software partitioning, but it can't tolerate the transient faults such as soft errors dynamically. These two methods belong to static methods for the reliable system design. In addition, a combined software and hardware technique for the design of reliable IP processor is given by M.Rebaudengo, but it puts software redundancy at the first place and uses partial hardware redundancy, the aim is designing a reliable single-core processor [7].It doesn't consider making full use of the hardware and software resources for the multi-core systems.
The presented architecture is different from the methods listed above, which is trade-off between software redundancy and the hardware redundancy, and can dynamically tolerate the fault based on the BIST (Built-In-Self-Test) structure. Moreover, the new method needs less hardware resources than the hardware redundancy, while which has the flexibility like the software one. The basic idea is using software to replace the hardware module once the fault is detected by the BIST structure, and it's especially suitable for the multi-core systems today. In multi-core systems, as limited by the algorithm's parallelism, not all the cores are busy all the time. So these spare processor cores can be made full use of to realize the fault-tolerant with the architecture proposed here.
The contribution of this paper is as follows: Firstly, a new architecture is come up which makes use of hardware/software co-design [8] and BIST [9][10] technology to realize the fault-tolerant system. Secondly, a methodology based on this architecture is given for time-limited fault-tolerant system designs. Thirdly, experiments are implemented to show the advantages of the architecture and an application of QoS scalable MPEG2 video decoder is given to overcome the disadvantage of this architecture for its usage in real-time multimedia designs.
The next section gives an overview of the architecture for this fault-tolerant system model. The design methodology based on such a fault-tolerant architecture is illustrated in detail in section 3. Section 4 shows the experimental results and gives an application of QoS scalable MPEG2 video decoder to show how to overcome the disadvantage in real-time multimedia designs. Finally, section 5 draws the conclusion and gives the future work related.
Descriptions of the architecture
The proposal is targeted at multi-core SoC which is made up of many processors cores and hardware modules. In this fault-tolerant architecture, hardware-software co-design technology is used to exploit the fault-tolerant metric, which is a trade-off between software flexibility and hardware high-efficiency.
The system architecture for multi-core SoC is shown in figure 1. In this architecture, the multi-core SoC is made up of PEs (Processing Element, in this paper , it refers to processor core ), hardware IP modules and the interconnection network that connects them together. In this system, some IP modules are selected to be fault-tolerant and they are used in the proposed fault-tolerant architecture, such as the parts located in the dashed border. The detailed illustration of these parts is in figure 2. This architecture is made up of three main parts except the on-chip memory, one is the processor (processors), the other is the BIST test structure, and the third is the interconnection network that connects them together. In this architecture, the IP core is put under test and acts as the hardware accelerator. The hardware control unit is used to control all the parts of the BIST structure and supplies the interfaces to the interconnection networks to acquire the test patterns and the related correct results for comparison. The BIST Core is made of three components: the TPG (Test Pattern Generator) which transfers the test patterns from the hardware control unit, the TRA (Test Response Analyzer) which is used to analyze the result and the BIST control unit for the BIST control procedure. The BIST test structure receives the test patterns and the correct results from the memory through the interconnection networks (In DMA (Direct Memory Access) Mode), and once the test is over , the result will be returned to the processor to tell it whether there is fault in the IP core. The test patterns are stored based on a priority-based method which store and select the test patterns in such a way: the test pattern that relates to the most sensitive fault or the one that we care most will be put in the first place and gives the highest priority, and then choose the second one, so on and so forth. During the test procedure, run as more test patterns as possible to get a better fault coverage. After the BIST test, the processor will take response according to this result.
The total flow of designing such a fault-tolerant architecture is as follows: Firstly, partition the hardware and software based on a particular multi-core architecture and parallelize the redundant software code which will act as the function of the hardware module once hardware fault is detected.
Secondly, through the system simulation, an initial estimation of the number of the clock cycles that will be taken to test the module is obtained. And based on this number, modify the machine code on the processor to insert a request signal to the BIST test structure to start the BIST function.
Thirdly, BIST begins and DMA transfer is used to transfer the test patterns and the correct results to the BIST test structure from the memory. While at the same time, the software is still running until when the function taken by the hardware module begins. Then the processor arrives at a break and waits for the result of BIST result.
Fourthly, BIST test ends and the BIST test structure gives an ack signal to the processor and gives a 6ne bit signal to show whether there is fault or not.
Finally, the processor will decide whether use the hardware module or software to realize the function. If software function is used, spare cores will be used to accelerate the software to make full use of the multi-core system. While hardware module is fault-free, the function will be implemented on the hardware module to accelerate the function.
The advantages 6f this architecture are as follows: From the view of the multi-core system design, it makes full use of the idle cores in the multiprocessor system to exploit the fault-tolerant mechanism, in such systems, some processor (like ASIP (Application Specific Instruction Processor) is selected for a specific application, such as the multimedia usage. And in such circumstance, the software running time may be not much more than the IP core and may also meet the need for the real-time applications. In this way, this architecture can be used to efficiently reduce the hardware cost introduced by the hardware-redundancy such as the FTMR.
The most important advantage of this architecture is, compared with the hardware-redundant method, it saves the hardware resources, while with comparison to the time-redundant software-redundant method, it is more time-efficient supposing that the hardware fault rate will be very low. What's more, as the test patterns are transferred using the DMA method from the outside memory, this method will use less hardware than the ordinary defect-oriented BIST structure which stores the specified test patterns with the on-chip memory [11].
Taking the reliability of the whole digital system into consideration. According to the formula, the fault-free probability 6f a single transistor changing with time is as follows [12]: Here, P is the fault-free probability, t is the time and 12 is about 51 10 h , suppose that each transistor contributes equally to the total system ,then the whole reliability (fault-free probability) of the system with N transistors is: Based on this formula, the fault-free probability of our architecture is as follows: In the formula above, the number of the MOS transistors is evaluated by the number of the NAND gates which uses 4 MOS transistors each. Based on this , the proposed architecture needs less hardware resources than hardware-redundancy and needs less memory space than the ordinary software redundancy (say, N-version programming or the time-redundant), so it will be more reliable this advantage will be more significant with the increase of the hardware module scale or the software one.
The Proposed Design Methodology Based on Such A Fault-Tolerant Architecture
For this architecture, scheduling of tasks (both hardware and software tasks) is very important, especially for the real-time applications [13]. Since this architecture is especially useful for heterogeneous multi-core systems, it should be taken into consideration at the system-level design step and give an estimation of the cost to determine whether to use this architecture and how to use it.
Suppose that the memory space of the processor is enough for all the software codes. And since the test structure (apart from the test pattern and test result pattern memory) is very simple compared with the complex IP core under test, the hardware resources taken by it can be ignored. What's more, since the communication time can be overlapped through the DMA transfer and data-prefetch [14], its impact on the whole system is much less than the other parts.
Total runtime for the hardware-redundancy (here, using FTMR as the method) is: In equation (4) M is the memory space used for the storage of the software code whose function equalizes the hardware module.
Total runtime for the software-redundancy (here, using 3 version programming as the method and three versions are running in parallel) is: In equation (6), sr T is the total runtime for the software-redundancy, 1 s T is the same as the one in formula 3, sf T is the software running time for the function that can be accelerated by the hardware.
Total hardware resource for the software-redundancy is: In equation (7), sr H is the total hardware resource for the software-redundancy. Total runtime for the proposed architecture is: In equation (8) are always on the RAM that can be put outside the chip core , so they can be ignored. And for the selection of the number N, it should be a good trade-off between the test-time and the fault coverage, for a particular application, only a small part of the whole test patterns should be mostly cared about, so as less test patterns as possible should be used to get the initial test result.
The total design flow for such fault-tolerant multi-core systems is shown in figure 3. At the beginning, an initial hardware/software partitioning is acquired based on the simulation for the processors and the hardware modules. And the minimum number of test-patterns for each fault-tolerant hardware module can be obtained according to the simulation. Then, the constraints of each fault-tolerant hardware module can be obtained, here, the most 2 important constraints are the hardware resource limit and the running deadline constraint for each one. Supposing the on-chip hardware-resources that can be supplied to this hardware module is HT, the running time deadline that can be taken by this hardware module is TT. (10), if other software-redundant methods (such as recovery block and 2-version programming, etc) can't meet the constraints, the partitioning should be considered twice and refresh it to get a appropriate portioning for such a fault-tolerant system.
After selecting the fault-tolerant architecture for each hardware module is finished, then the total methodology ends and the system level fault-tolerant system is developed.
Experimental results
In this part, three kinds of experiments are implemented for the evaluation of the proposed architecture and an application of QoS scalable MPEG2 video decoder is shown to overcome the drawbacks of such fault-tolerant system for the real-time multi-media systems.
Resource usage evaluation and performance comparison experiments:
Two experiments are implemented, one for combinational circuit module and the other for sequential circuit module. A sorting module is made as the combinational circuit module and an IDCT module is taken as the sequential circuit module, the results are as follows: Supposing the fault rate is 1%, then through the RTL design and use the Design Compiler, the results are as follows: (X is the hardware logic resources needed by the MIPS processor, about 40000 gates). Here, the processor is the MIPS processor which we have designed by ourselves. The test results include the simulation outcomes of both the processor and the hardware module. In these experiments, the software on the processor is for the initialization of the data prepared for the hardware module under test.
The methods chosen for the comparison are as follows: Hardware redundancy uses the popular NMR method (here, using FTMR method which uses three copies of the module and the majority voter to get the right result). Software redundancy uses the popular N-version programming method (Here N = 3 and change little between different versions). Table 1 shows the test result of the running time and hardware logic for the comparison of the three architectures for the case of a combinational circuit module (sorting module). From this table, it can be seen that the running time of the proposed architecture is little more than the hardware redundancy, but it needs much less hardware logic than the hardware redundancy .What's more the running time of the proposed architecture is much less than the software redundancy , while the hardware logic needed is more than the software redundancy. In a word, compare these three methods with a ratio (which is performance/logic number, here, using the 1/ (Running Time) as the performance value), it can be concluded that:
Hardware logic and running time analysis result:
Hardware Redundancy : Software Redundancy : Proposed Architecture = 4.9026 : 2.6344 : 8.8299 This means the proposed architecture will get the best performance per each logic gate. This also shows that the proposed architecture will get the best performance per each logic gate. Table 3 and Table 4 give the energy consumed by each of the fault-tolerant methods, table 3 is for the combinational circuit module while table 4 is for the sequential circuit module. From these results it can be concluded that the energy consumed by the proposed architecture is more than the hardware redundancy method (about 77%), but is much less than software redundancy method (the software redundancy energy is more than 36 times as large as the proposed architecture). The main title (on the first page) should be at the top of the printable area, centered, and in Times 14-point, boldface type. Capitalize the first letters of nouns, pronouns, verbs, adjectives, and adverbs; do not capitalize articles, coordinate conjunctions, or prepositions (unless the title begins with such a word). Leave one blank line after the title.
Communication Cost And Optimization Method
Using the proposed architecture to implement the total system, both power and time cost are considered in the system architecture level, and to reduce the communication cost, both data-prefetch technology and DMA transfer method are used to get the transfer-time overlapped and power will be saved because the processor will no longer need to read the data and write them to the bus. Table 5 shows the comparison for the time and energy per each data fetch before and after the optimization (using 100MHz AMBA [15] bus as the transfer media, and using burst transfer mode). From this table, it can be seen that the DMA transfer overlaps the data fetch with the software running so the time can be ignored, while the energy per fetch can be reduced by nearly 39.7%. So these optimizations can be used to improve the system's performance which uses the proposed fault-tolerant architecture.
Fault-Injection And Coverage Test
Since this architecture mainly targets at the particular faults and the testing time is limited, so the coverage will be lower than typical BIST methods. Table 6 shows the fault-coverage rate of the IDCT module. The fault-injection is on the gate level and 30 random faults are added to the model to simulate the test structure. From this table, it can be seen that the coverage rate is lower than the ordinary test structures. Therefore, this architecture can't ensure the full test coverage but can be used for the specified fault that is most sensitive to the environment.
Fault-free Probability Comparisons
According to equation (3) and the experiment results listed above, the fault-free probability comparisons can be easily acquired and the curves are as follows: Fig 4 and fig 5 give the curves for the IDCT module and the sorting module, from these curves, conclusion can be drawn that the proposed architecture has the highest fault-free probability than the other 2 methods.
QoS Scalable MPEG2 video decoder with this proposed architecture
Since the software running time for the same function is always much more than the hardware accelerators, the proposed architecture can't ensure real-time when a fault occurs. To overcome this drawback, this part gives a QoS Scalable MPEG2 video decoder design for such real-time multimedia applications. MPEG2 [16][17] is a very popular video standard and is selected as the digital TV standard for most digital TVs in China. Once the hardware module of the decoder system has fault in it, it leads the system to a failure, there will be no picture on TV, and the whole system will be shipped to the factory for the repayment. Using the proposed architecture, many faults will be tolerated and the video can be continued in a low QoS mode instead of displaying no picture at all. Using SystemC [18] [19] [20] as the description language, such a MPEG2 decoder with the proposed architecture is implemented and simulated. In this architecture, IDCT module acts as the hardware module and the other functions of the decoder is put on the processor running in software(here ,using the powerful AMD Turion 64 processor as the main processor on which the software is running) . To make the results more reliable, standard test sequences are used to act as the input into the system.
The results are obtained as in table 7: Based on table 7, suppose the processor and the hardware modules work at 550MHz, then the frame-rate for the HD video can be 25.4frames/second with no fault in the system but when a fault occurs, the frame-rate is lowered to 23.6frames/second.Thus, if a fault occurs in the hardware module (IDCT), the decoder can't realize the real-time decoder for the HD video. Using this fault-tolerant architecture, QoS can be lowered that decreases the frame-size to 1280 × 720 once a fault occurs and the frame-rate can be 48.8frames/second. This will be much better than no picture displayed at all in normal systems when a hardware failure occurs, although the QoS is somewhat lowered. This case analysis shows the how to overcome the disadvantage of this fault-tolerant architecture and its usage in multimedia applications.
Conclusions and future work
This paper gives an hardware-software co-design fault-tolerant architecture which is targeted at high-reliability multi-core systems. And through the analysis, conclusion can be drawn that the proposed architecture makes full use of the hardware and the software which are abundant in the heterogeneous multi-core systems. And if N is not large for the specific application it has less hardware resource needed than hardware redundancy and is faster than the normal software redundancy. The example of the MPEG2 QoS Hardware redundancy Software redundancy Proposed Architecture scalable video decoder shows the usage of this architecture for multimedia processing to overcome the drawback of the software running. Therefore, this architecture is very appropriate for the future heterogeneous multi-core systems which need high reliability. In our future work, this architecture will be used in heterogeneous multi-core system designs to acquire the high reliability, especially for the embedded systems. And it will benefit a lot for such designs. The disadvantage of the architecture is that the fault coverage is not very high compared with the normal BIST method, so in the future more attention will be paid to this limitation to improve the architecture. | 2009-10-19T21:01:51.000Z | 2009-10-19T00:00:00.000 | {
"year": 2010,
"sha1": "5d312ddb9c02055334dd4a960daea44121c99ae4",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "5d312ddb9c02055334dd4a960daea44121c99ae4",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
59618581 | pes2o/s2orc | v3-fos-license | Apoptotic Cell-Mimetic Polymers for Anti-Inflammatory Therapy
The field of biomaterials has seen a strong rejuvenation due to the new potential to modulate immune system in our body. This special class of materials is called “immunomodulatory biomaterials”. Generally, three fundamental strategies are followed in the design of immunomodulatory biomaterials: (1) immuno-inert biomaterials, (2) immuno-activating biomaterials, and (3) immuno-tolerant biomaterials. While many applications of immuno-inert biomaterials such as biocompatible medical implants have been already proposed in the past decades, the ability to engineer biological activity into synthetic materials greatly increases the number of their potential uses and improves their performance in more traditional applications. The major focus of researchers is now set on developing immuno-tolerant biomaterials for anti-inflammatory therapies. In this review, we therefore introduce recent developments of immuno-tolerant biomaterials. Especially we introduce an apoptotic cell membrane-inspired polymer and its post-inflammatory effects on immune cells in this article.
INTRODUCTION 1. History of immuno-inert biomaterials
Biomaterials have been historically developed with the aim to transplant then into the human body as substitutes for damaged tissue or a biological function. 1,2 Biomaterials may be metals, ceramics, polymers, or even living cells and tissue. They can be used as surface coatings, fibers, films, or particles for use in biomedical products such as heart valves, hip joint replacements, dental implants, or drug delivery carriers. The first generation of biomaterials was developed under the concept of immune evasion because the biggest challenge of "foreign objects" has been the action of the immune system. Therefore, it was preferable that they be 'inert' and not interact with the biology of the host organism such as proteins, lipids, nucleic acids, sugars, and amino acids. One of the most successful approaches to produce bio-inert materials has been the used of polyethylene glycol (PEG). 3 For example, covalent attachment of PEG to proteins or drugs called "PEGylation" has been shown to improve the safety and efficiency, and many PEGylated pharmaceuticals are currently on the market. This technology can be also used as a means to design anti-fouling surfaces. Grafting PEG to solid surfaces significantly reduces protein adsorption and cell adhesion. [4][5][6] In addition to PEG, zwitterionic polymers have recently been extensively developed for anti-fouling purpose. [7][8][9] Among them, one of the cell membrane lipids, phosphatidylcoline (PtdCho)-inspired polymers such as 2-methacryloyoxyethyl phosphorylcholine (MPS) have been used in various medical devices. [10][11][12] These polymers have extended the applications of biomaterials from traditional implants to biosensing, prodrug carriers, subcellular bioimaging, and cell manipulation. ti-tumor activity is expected when immune systems are activated, while excess activation can cause unexpected symptoms such as allergies (Fig. 1). On the other hand, suppressing the immune system lowers inflammation, but also leads to diminished resistance against infections. Therefore, to obtain immunotherapy with a biomaterial, it is necessary to perfectly understand the biological reactions induced by implanted materials and to adequately design the shape, physical properties, and chemical prop-erties of the material. Fig. 2 shows three types of biomaterials according to their association with the immune system. The first category is called "immuno-inert biomaterials" as described above. The second type is called "immuno-activating biomaterials" designed to exhibit anti-tumor and drug-responsive properties. The third type is known as "immuno-tolerant biomaterials" which suppress and modulate unnecessary uncontrollable inflammation and inhibit rejection responses.
Adjuvant materials
As mentioned above, much effort has been previously made to design biomaterials to minimize the host's immune responses against implanted materials. However, biomaterials can also be designed to activate the host's immune responses and/or provide therapeutic effects. The first application of this concept was the use of non-biological adjuvant materials such as -polyglutamic acid ( -PGA), 13 poly (lactic-co-glycolic acid) (PLGA), 14 or poly (-caprolactone) (PCL). 15 These biodegradable materials can increase the host's immune response to vaccines. New types of pH-responsive nanoparticles have also been developed because the degradation rate for such biodegradable polymers is not fast enough for effective antigen deliveries. 16 The nanoparticles with pH-cleavable crosslinkers are rapidly hydrolyzed under lysosomal acidic conditions (pH 5) and release antigens into dendritic cells (DCs). By contrast, Stayton and co-workers proposed antigen delivery strategies that mimic virus infection mechanisms. They focused on polypropylic acid (PPAA), which has a pKa around pH 6.0-6.5, making it hydrophilic at a biological pH and dramatically hydrophobic at endosomal pH due to the protonation of carboxyl groups along its backbone. 17 Due to the polymer phase transition, PPAA disrupted lipid bilayer membranes and antigens tethered to PPAA by disulfide bonds were released under endosomal reductive conditions.
Hemagglutinating virus of Japan-envelope (HVJ-E)
Anti-tumor immunity can be also induced by viral envelopes such as Hemagglutinating Virus of Japan-envelope (HVJ-E). 18 HVJ-E is a purified product prepared through the complete inactivation of the genome in HVJ by UV irradiation. HVJ-E was originally developed as a novel vector for plasmid DNA, peptides, and drugs because of the fusion proteins of the HVJ such as HN and F are retained after inactivation. Recently, Kaneda and colleagues have reported the tumor-suppressing ability of the inactivated, replication-defective HVJ-E itself. 19 HVJ-E alone induced tumor-specific antitumor immunity by eliciting IL-6 production in dendritic cells (DCs) and eradicated 60% to 80% of tumors growing in mice without exogenous gene expression. Thus, HVJ-E has recently attracted a lot of attention as a new type of therapeutic material for cancer immunotherapy. However, one of the greatest challenges for viral-based vectors is that they tend to fuse with virtually all cells and induce hemagglutination in the bloodstream. Therefore, it has been required to increase the tumor-targeting ability as well as reduce nonspecific biding. To achieve this, the chemical modification of the viral vector is one of the most direct approaches. For example, surface-modified adenoviral vectors (Ads) with a multivalent reactive poly [N-(2-hydroxypropyl) methacrylamide] (PHPMA)-based copolymer has been shown successfully to shield from recognition by antibodies. 20 The authors have fashioned HVJ-E with hyaluronic acid (HA) to act as a ligand for CD44 to improve its anti-tumor activity. 21,22 The functionalized HVJ-E is expected to improve the retention time in the blood stream by inhibiting interactions between HVJ-E and red blood cells. Also, the HA layer can be selectively degraded in cancer tissues by hyaluronan-targeting enzymes. A layer-by-layer (LbL) assembly has been employed for the decoration of the HVJ-E. We have succeeded in forming HA/glycol chitosan (GC) multilayers on HVJ-E for up to six assembly cycles. The outermost HA layer was selectively degraded by hyaluronidase. The HA/GC-coated HVJ-E can be thought of as a potential bomb for cancer immunotherapy because of its ability to target CD44 while preventing nonspecific binding in the bloodstream.
Induction of immune tolerance
In addition to immune-activating biomaterials, researchers have been recently developing immuno-tolerant or suppressing materials for the treatment of transplant rejections and autoimmune diseases. In general, the immune system effectively eliminates foreign substances such as pathogenic microorganisms, but does not respond to self-biomolecules, symbiotic microorganisms, or environmental substances. However, in autoimmune diseases, the immune system misidentifies some molecules as foreign and elicits a response against them. For example, chronic inflammatory disorders such as inflammatory bowel disease, allergic reactions, cancer, obesity, diabetes, collagen disease, and Alzheimer's disease are all caused by the impaired maintenance of immune homeostasis. To induce immune tolerance, several strategies have been developed by using various biomaterials for the delivery of adjuvants, antigens, and drugs. For example, PLGA particles that contained costimulatory anti-sense oligonucleotides for DCs and T cells successfully inhibited immune responses. 23 In another example, PLGA particles modified with anti-DEC205 antibodies, anti-CD1c antibodies, and P-2 peptides were also developed. 24 These examples show the potential application of targeting PLGA particles to maintain immunological tolerance.
However, previous immune-suppressive biomaterials as mentioned above simply operate as delivery carriers. In other words, the materials themselves do not have any anti-inflammatory properties. However, in recent years, material-based immuno-suppressive systems have been developed, and a few specific examples will be introduced here. Stable, organic free radicals such as 2, 2, 6, 6-tetramethylpiperidine 1-oxyl (TEMPO) are often used as catalysts for oxidation reactions in organic synthesis and can act as radical scavengers that detect radical generation reacting systems. Yoshitomi et al. 25 focused on the radical scavenging ability of TEMPO for removing reactive oxygen species (ROS) at inflammatory regions. Normally, large amounts of ROS are produced during inflammation, and exposure to this oxidative stress can cause various diseases such as
Apoptotic cell-inspired biomaterials
Ideally, biomaterials should mimic the living tissue or biomolecules from a mechanical, chemical, biological, and functional point of view. As mentioned in the introduction, one of the most successful biocompatible polymers, MPC polymer, was developed by learning from mammalian cell membrane lipid PtdCho. MPC polymers have been extensively applied to various medical devices, including an artificial joint (Aquala ® , Japan Medical Materials), cardiovascular stent (Endeavor ® : Medtronic), left ventricular assist device (EVAHEART ® , Sun Medical), oxygenator (PrimO2x ® : Sorin), and soft contact lenses (Proclear ® : Cooper Vision). Therefore, we discuss the potential to create a novel immune-suppressing biomaterial by focusing on the lipid membrane in the last section.
In general, cell membranes have a lipid bilayer constructed by phospholipids such as electrically neutral sphingomyelin (Sph) and PtdCho that occupy the outside membrane and anionically charged phosphatidylethanolamine (PtdEA), PtdSer, and phosphatidylinositol (PtdIno) that are located on the cytosol side. 27 Cell membranes keep this heterogeneous structure by active transportation at inner sites by scramblase, and the cytoskeleton is therefore stabilized by anionic phospholipids. However, apoptosis induces the destruction of these highly maintained asymmetrical structures, forcing PtdSer to be exposed on the surface of apoptotic cell membranes, which are then recognized by immune cells (Fig. 3). 28 Apoptosis is a genetically programmed cell death mechanism and is necessary to regulate embolic development and maintain homeostasis. Many studies also revealed that PtdSer acts as a trigger molecule for these anti-inflammatory effects. 29 Voll et al. 30 reported that apoptosis induced leukocytes to downregulate inflammatory cytokine production and upregulate anti-inflammatory cytokines for the production of immune cells. Several therapeutic effects of PtdSer liposomes have been also reported for chronic heart failure, 31 Alzheimer's disease 32 and osteoporosis. 33 Table 1 summarizes the recent progress of anti-inflammatory therapies using apoptotic cells and PtdSer liposomes. [30][31][32][33][34][35][36][37] We have recently designed apoptotic cell membrane-inspired polymers which have a phosphoryl serine (PS) group because PtdSer mimetic polymers are very promising materials for anti-inflammatory therapies (Fig. 4). 38,39 However, the synthesis of PtdSer-mimetic polymers has been diffi- cult as serine has two reactive groups (the amino-and carboxy-groups) that limit the effective and selective modification routes. Therefore, we used a phosphoramidite method which is commonly used in solid phase syntheses of DNA. The advantage of using this reaction is its high selectivity under mild conditions when forming a phosphate ester between two hydroxyl groups. 39 First, we synthesized the methacryloyloxyethyl phosphorylserine (MPS) monomer using 2-hydroxyethyl methacrylate, N--(t-butoxycarbonyl)-L-serine t-butyl ester, and t-butyl tetraisopropylphosphorodiamidite. Next, poly (MPS) was obtained by free radical polymerization and then deprotection and oxidation of the PtdSer using t-butyl peroxide and trifluoroacetic acid. We applied poly (MPS) to lipopolysaccharide (LPS)-activated RAW 264.7 macrophages to investigate the immune-suppressive effects. Amoeboid cell extensions of macrophages in LPS stimulated conditions were inhibited in the poly (MPS)-treated groups. In addition, the effective suppression of NF-B in macrophages was observed over 10 mM of PS groups (Fig. 5). 38 To investigate how macrophages recognize poly (MPS), the localization of rhodamine-modified poly (MPS) on macrophages was studied. A homogeneous distribution of the poly (MPS) in the cytoplasm was observed. By calculating the up-taken amount of poly (MPS) in the cell, we found that 50 pg or more of poly (MPS) within each macrophage is necessary to inhibit inflammatory behavior (Fig. 6). Furthermore, we investigated the time-dependent immunosuppression induction of poly (MPS
CONCLUSION
This review article summarizes recent research examples of immuno-modulating biomaterials and systems. These materials are classified into three categories: (1) immuno-inert, (2) immuno-activating, and (3) immune-tolerant biomaterials. Particularly, we mainly focused on the immuno-tolerant biomaterials in this article. In previous immuno-tolerant biomaterial systems, the materials themselves did not have any anti-inflammatory properties. However, the ability to engineer biological activity into synthetic materials has enabled the design of biomaterials which harness the host's immune responses and provide therapeutic effects. One of the examples is apoptotic cellmembrane mimetic polymers. Advantages of using such polymeric materials for anti-inflammatory therapy is cost-effectiveness as it is inexpensive compared to molecular targeted drugs. However, controlling the immune system is still a challenge due to its methodological complexity. For example, excess activation of the immune system may cause unexpected symptoms such as allergies, while suppressing of the immune system also leads to diminished re-sistance against infections. Therefore, to obtain immunotherapy with a biomaterial, it is necessary to design the shape, physical properties, and chemical properties of the material adequately. In the future, biomaterials will assume an even greater role in anti-inflammatory therapies through biologically inspired design. | 2019-02-12T00:18:43.332Z | 2019-01-01T00:00:00.000 | {
"year": 2019,
"sha1": "548f456b65d0a6399f67c2a6810f67e79aafa141",
"oa_license": "CCBYNC",
"oa_url": "http://cmj.ac.kr/Synapse/Data/PDFData/1057CMJ/cmj-55-1.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "548f456b65d0a6399f67c2a6810f67e79aafa141",
"s2fieldsofstudy": [
"Medicine",
"Materials Science"
],
"extfieldsofstudy": [
"Medicine"
]
} |
245148327 | pes2o/s2orc | v3-fos-license | Case Report: Successful Treatment of Acute Generalized Exanthematous Pustulosis With Secukinumab
Acute generalized exanthematous pustulosis is a severe, usually drug-related reaction, characterized by an acute onset of mainly small non-follicular pustules on an erythematous base. Most cases of acute generalized exanthematous pustulosis (AGEP) clear quickly with a systemic corticosteroid, but severe or recalcitrant cases may need other systemic therapies. In this case, a man in his 40 s with a history of psoriasis consulted a physician about widespread erythema, pustules, target lesions, and fever after the administration of a quadruple antituberculosis drug. Routine laboratory testing revealed elevated white blood cell count and C-reactive protein. The histopathology showed subcorneal pustules, spongiosis as well as lymphocyte and eosinophils infiltration in the dermis. The patient was diagnosed with definitive AGEP according to the diagnostic score from the EuroSCAR study. Cutaneous lesions especially pustules and erythema multiforme-like lesions on the upper arms and palms are crucial for distinguishing AGEP from Generalized pustular psoriasis. The patient was treated with secukinumab as a result of his failure to respond to topical corticosteroids and constrain of systemic steroids. Remission with secukinumab therapy was safe without increased risks of infections. This case indicates that secukinumab is a potential therapy that can rapidly improve the clinical symptoms of AGEP.
Acute generalized exanthematous pustulosis is a severe, usually drug-related reaction, characterized by an acute onset of mainly small non-follicular pustules on an erythematous base. Most cases of acute generalized exanthematous pustulosis (AGEP) clear quickly with a systemic corticosteroid, but severe or recalcitrant cases may need other systemic therapies. In this case, a man in his 40 s with a history of psoriasis consulted a physician about widespread erythema, pustules, target lesions, and fever after the administration of a quadruple antituberculosis drug. Routine laboratory testing revealed elevated white blood cell count and C-reactive protein. The histopathology showed subcorneal pustules, spongiosis as well as lymphocyte and eosinophils infiltration in the dermis. The patient was diagnosed with definitive AGEP according to the diagnostic score from the EuroSCAR study. Cutaneous lesions especially pustules and erythema multiforme-like lesions on the upper arms and palms are crucial for distinguishing AGEP from Generalized pustular psoriasis. The patient was treated with secukinumab as a result of his failure to respond to topical corticosteroids and constrain of systemic steroids. Remission with secukinumab therapy was safe without increased risks of infections. This case indicates that secukinumab is a potential therapy that can rapidly improve the clinical symptoms of AGEP.
INTRODUCTION
Acute generalized exanthematous pustulosis is a rare disease, therefore there are no high-quality evidence-based therapeutic recommendations. As it is, treatment is mainly potent topical or systemic steroids. However, a minority of cases fail to respond to them, highlighting the need for additional treatment options for these refractory cases (1). Secukinumab is an interleukin (IL)-17 inhibitor that is approved to treat psoriasis. However, the efficacy of secukinumab on acute generalized exanthematous pustulosis (AGEP) is unclear. In this review, we reported the successful use of secukinumab in a patient with AGEP.
REPORT OF A CASE
A 41-year-old man presented with a 1-day history of rapidly growing erythema and pustules on the face, trunk, and extremities. One day before, he took isoniazid, rifampicin, ethambutol, and pyrazinamide to treat pulmonary tuberculosis in lazaretto. At admission, the patient was febrile to a temperature of 40 • C. He also had a 2-year history of plaque psoriasis with no history of smoking. Routine laboratory testing revealed white blood cell count of 13,890/µl (normal range, 1,214-5,110/µl) and serum levels of C-reactive protein (CRP) of 7.75 mg/dl (normal range, 0-0.2 mg/dl). Physical examination revealed widespread erythema and dried-up lakes of pus, taking up body surface area of 80%. Pustules and erythema multiforme-like lesions were diffusely spread over the upper arms and palms (Figures 1a-d).
Biopsies of pustule and erythema multiforme-like lesions on the arms were performed. The result demonstrated subcorneal pustules, spongiosis as well as lymphocyte and eosinophils infiltration in the dermis (Figures 1h,i).
The patient was assessed using a diagnostic score according to the EuroSCAR study. The final score was nine, which was defined as the definitive AGEP. The patient was subsequently treated with expectant treatment and topic potent steroids for 4 days, followed by no improvement. However, given the patient's femoral head necrosis, systemic glucocorticoid was not recommended. Repeated computed tomography on the chest showed multiple calcifications and thus excluded the possibility of active tuberculosis. Other infections including hepatitis B and hepatitis C were also excluded. The patient received secukinumab 300 mg after being informed of impersonal medical risks (Supplementary Figure). After the first therapy of secukinumab, all pustules faded at day 3, and erythema substantially darkened with defervescence (Figures 1e-g). The Dermatology life quality index (DLQI) decreased rapidly from 25 to 1 at week 4. There were no clinical symptoms and radiological data related to the progression of latent tuberculosis infection.
DISCUSSION
In this case, the patient had a medication history of quadruple anti-tuberculosis drugs, specific exanthem of erythema multiforme-like lesion, abnormal laboratory results of white blood cell count and C-reactive protein, dermatologic evidence of spongiosis, and subcorneal pustule and eosinophils infiltration. He was then diagnosed with definitive AGEP according to the diagnostic score from the EuroSCAR study. Generalized pustular psoriasis, either the "Annular variant" or "Exanthematic variant, " can also present a similar acute eruption following initiation of medications. The common disease-causing medications include TNF-α inhibitors, lithium, interferons, antimalarials, beta-blockers, rapid withdrawal of systemic corticosteroids. This case provides some clinical reasonings of differential diagnosis between AGEP and GPP. First, cutaneous lesions especially pustules and erythema multiforme-like lesions on the upper arms and palms are crucial for differential diagnosis in this case. Besides, dermatologic evidence of spongiosis, as well as eosinophils infiltration in pathology, favors a diagnosis of AGEP. Finally, the rapid resolution of pustular eruption with secukinumab is suggestive of the diagnosis being AGEP.
Owing to his failure to respond to topical corticosteroids and constrain of systemic steroids, the patient was treated with secukinumab for 1 month. Disease progression was then controlled promptly, with a resolution of pustules and erythema. The great improvement of dermatology-specific health-related quality of life was also observed. Prior to clinical admission to our department, the patient was misdiagnosed as active TB and was sequentially prescribed isoniazid, rifampicin, ethambutol, and pyrazinamide in lazaretto. At admission, active tuberculosis was excluded based on clinical symptoms and results of repeated chest computed tomography. We confirmed that the patient only had latent tuberculosis infection. Thus, this patient did not receive any treatment for his latent TB. The chest radiography was done 1 month later and then every 6 months over a 2-year follow-up. The results showed no change in the lung pathology (calcification). As reported in the literature, patients with latent TB who received no chemoprophylaxis could be safely treated with secukinumab (1, 2).
The underlying mechanism of action of secukinumab on AGEP is related to IL-17. AGEP has been classified as a T cell-related sterile neutrophilic inflammatory response with a predominant T-helper (Th) 1 cell and Th17 (3-5). Th17 cells release IL-17, which affects neutrophil recruitment (6,7). Therefore, the IL-17 pathway is a potential therapeutic target shown to be implicated in AGEP. Further clinical trials are needed to corroborate the safety and efficacy of secukinumab in AGEP.
DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.
ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Ethics Committee of the First Affiliated Hospital of Fujian Medical University. The patients/participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s), and minor(s)' legal guardian/next of kin, for the publication of any potentially identifiable images or data included in this article.
AUTHOR CONTRIBUTIONS
All authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication. | 2021-12-16T14:22:57.364Z | 2021-12-16T00:00:00.000 | {
"year": 2021,
"sha1": "34991c8648a94fd20a3311fbf9c471524d9db9cb",
"oa_license": "CCBY",
"oa_url": "https://www.frontiersin.org/articles/10.3389/fmed.2021.758354/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "9d121f6b0f4c39082dd80fdd7b993d17f7d3fa40",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": []
} |
255844319 | pes2o/s2orc | v3-fos-license | Helical intensity-modulated radiotherapy of the pelvic lymph nodes with a simultaneous integrated boost to the prostate - first results of the PLATIN 1 trial
Definitive, percutaneous irradiation of the prostate and the pelvic lymph nodes in high-risk prostate cancer is the alternative to prostatectomy plus lymphadenectomy. To date, the role of whole pelvis radiotherapy (WPRT) has not been clarified especially taking into consideration the benefits of high conformal IMRT (intensity modulated radiotherapy) of complex-shaped target volumes. From 2009 to 2012, 40 patients of high-risk prostate cancer with an increased risk of microscopic lymph node involvement were enrolled into this prospective phase II trial. Patients received at least two months of antihormonal treatment (AT) before radiotherapy continuing for at least 2 years. Helical IMRT (tomotherapy) of the pelvic lymph nodes (51.0 Gy) with a simultaneous integrated, moderate hypofractionated boost (single dose of 2.25 Gy) to the prostate (76.5 Gy) was performed in 34 fractions. PSA levels, prostate-related symptoms and quality of life were assessed at regular intervals for 24 months. Of the 40 patients enrolled, 38 finished the treatment as planned. Overall acute toxicity rates were low and no acute grade 3 or 4 gastrointestinal (GI) and genitourinary (GU) toxicity occurred. 21.6 % of patients experienced acute grade 2 but no late grade ≥2 GI toxicity. Regarding GU side effects, results showed 48.6 % acute grade 2 and 6.4 % late grade 2 toxicity. After a median observation time of 23.4 months the PLATIN 1 trial can be considered as sufficiently safe meeting the prospectively defined aims of the trial. With 34/37 patients free of a PSA recurrence it shows promising efficacy. Tomotherapy of the pelvic lymph nodes with a simultaneous integrated boost to the prostate can be performed safely and without excessive toxicity. The combined irradiation of both prostate and pelvic lymph nodes seems to be as well tolerated as the irradiation of the prostate alone. Trial Numbers: ARO 2009–05, ClinicalTrials.gov: NCT01903408.
Background
Percutaneous irradiation of locally advanced prostate cancer is the alternative therapy to radical prostatectomy (RP). A direct and valid comparison between radiation therapy (RT) and RP is not possible on the basis of existing studies due to the lack of prospective clinical trials [1][2][3]. When comparing retrospectively the clinical outcome of external beam radiotherapy (EBRT) with case series for RP, similar rates of the five-year (5y) biochemical progression free survival (bPFS) and the 5y-and 10y-disease-specific survival (DSS) were observed for EBRT, while the 10y-bPFS and the 10y-overall survival (OS) were slightly higher in favor of RP. However, this weak evidence does not allow a prioritization of one of the two treatments for locally advanced prostate cancer.
The addition of pelvic lymph nodes to the radiation field raised concerns regarding possible increased side effects. Acute gastrointestinal (GI) and genitourinary (GU) toxicities were more frequently reported in patients treated with WPRT [4,5]. Late complications were also more frequently seen in the whole pelvis (WP)RT group compared to postoperative radiotherapy (PORT) [6][7][8]. IMRT can reduce acute and late toxicities using smaller irradiated volumes of bladder, small bowels and rectum [9,10]. Moderately hypofractionated RT has become routine over the years due to area-wide implementation of image guided (IG)RT and better knowledge of tumor radiation biology [11][12][13]. Hence, in the present study we combined both, advanced IMRT/IGRT techniques for a better tolerability, and a moderate hypofractionated simultaneous integrated boost (SIB) to the prostate for a probably higher biological effectiveness.
The PLATIN (Prostate and Lymph Node Irradiation with Integrated-Boost-IMRT after NHT) phase II trial evaluates an optimized WPRT in patients with locally advanced prostate cancer. Due to image guided IMRT even inhomogeneous dose distributions can be applied accurately. Initiated in 2009, the study was designed to prospectively investigate safety and feasibility of five prostate and lymph node irradiation concepts, each enrolling n = 40 patients. Pelvic lymph nodes are simultaneously irradiated with an integrated boost to either the prostate (PLATIN 1), the prostate and macroscopic lymph nodes (PLATIN 2), the prostate bed (PLATIN 3), the prostate bed and macroscopic lymph nodes (PLA-TIN 4), or to macroscopic lymph nodes in patients with prior PBRT (PLATIN 5). Secondary objectives were a detailed characterization of the toxicity profiles, and the evaluation of quality of life during treatment. Results of the PLATIN 3 arm were published by Katayama et al. [14].
In this article we report on safety and efficacy data applying IMRT treatment of the pelvic lymph nodes with a SIB to the prostate (PLATIN 1).
Methods
From May 2009 to December 2012, 40 patients were enrolled prospectively in the PLATIN 1 trial. Eligibility criteria were, among others, a histological proven prostate carcinoma without lymph node metastases but with an estimated risk of lymph node involvement >20 % according to the Roach formula [15]. In case of a diagnostic lymphadenectomy, a minimum number of ten lymph nodes had to be surgically removed.
Before trial initiation, ethical consent was obtained from the ethics committee of the University of Heidelberg (permit S-034/2009). All patients gave written informed consent before trial enrollment. All reported data were conducted in accordance with the Helsinki Declaration and with national guidelines. Patients received at least 2 months of neoadjuvant AT (bicalutamide or LHRH analogue). With good tolerability AT was continued for at least 2 years after irradiation.
For treatment planning, CT scans with 3 mm slice thickness at full bladder and empty rectum were performed. PTV-P (planning target volume -prostate) covered the prostate (CTV-P) + 6 mm including the seminal vesicles. PTV-L (planning target volume -lymph nodes) included the obturatory, internal and external iliac, common iliac and presacral (down to S3) lymph nodes with a 5 mm margin [16]. Pararectal lymph nodes were not included in the PTV-L. Inverse treatment planning was performed using the Tomotherapy® treatment planning software (Accuray, USA). A total dose of 51.0 Gy was prescribed to 95 % of PTV-L with a SIB of 76.5 Gy to 95 % of PTV-P in 34 fractions. The dose prescription to the lymph nodes of 51 Gy in 1.5 Gy fractions is biologically equivalent to 43.7 Gy, assuming a α/β of 1.5 Gy for prostate cancer; and 48.2 Gy, assuming a α/β of 7 Gy for small bowel. Treatment was performed with full bladder and empty rectum under daily IGRT.
Prostate-specific symptoms and treatment toxicity, using the criteria of the NCI CTC AE version 3.0, were recorded before treatment, weekly during treatment, at the end of treatment, and at 2.5, 6, 12, 18 and 24 months follow-up. For calculation of toxicity rates, only patients with available data at the respective time points were considered. Cumulative GI toxicity was defined as the cumulative incidence of diarrhea, enteritis and proctitis. To facilitate comparison with other publications, only cystitis was included in the calculation of cumulative GU toxicity, as most scoring systems do not include incontinence and erectile dysfunction. Nevertheless, incontinence and erectile dysfunction were recorded.
Quality of life was assessed using the EORTC QLQ-C30 questionnaire before treatment and during followup after 6, 12 and 24 months.
PSA levels were measured before radiotherapy and then every 3 months afterwards, starting from week 10.
Biochemical failure was established according to the Phoenix criteria [17].
A primary endpoint, the safe treatment application rate (STR) was chosen. STR was defined as the proportion of patients receiving treatment as planned and without grade 3-4 toxicity and calculated as the ratio of the number of patients fulfilling this criteria divided by the size of the Intention-to-treat (ITT) population. The ITT population consisted of all patients giving informed consent, fulfilling the inclusion and exclusion criteria and receiving planned treatment for a minimum of 4 weeks after initiation. Based on a one-stage phase II type design, STR of 80 % (null-hypothesis SDR ≤80 %) was tested against the alternative of being at least as large as 95 % in a one-stage phase-II type design using the exact binomial test at the significance level of 0.1 % with a power of 90 %. The null hypothesis would be rejected when SDR would be at least 87.7 %.
Patient characteristics
Among all 40 patients (identical with the ITT population), median follow-up was 23.4 months (range: 2.8 -31.7 months). Median age at inclusion was 70 years (range: 51 -75 years); all patients were high-risk according to the D' Amico risk categories [18]. One patient underwent laparoscopic resection of lymph nodes but rejected prostatectomy. Indication for WPRT was seen by reason of lymph node affection (pN+) without remaining macroscopic lymph node metastasis in situ. All other patients had an estimated risk of lymph node involvement >20 % according to the Roach formula. Both LHRH and antiandrogen therapy were permitted as antihormonal therapies.
Twenty-seven patients received LHRH analogue therapy, seven patients received bicalutamide and six patients both (complete androgen deprivation). Most of the patients still had antihormonal therapy prescribed by their urologist as they were seen in the Department of Radiooncology. Radiotherapy was performed as definitive treatment in 38 patients. Two patients had PSA elevation during NHT and were thus excluded from the study. One patient died 7 months after radiotherapy diagnosed with a metastasized esophageal cancer. For further patient characteristics see Table 1.
Plan quality in terms of organ at risk sparing is shown in Table 2. The anterior rectal wall received a maximum dose of 72.8 ± 1.3 Gy. The rectum received doses ≥60 Gy (9.5 %) and ≥70 Gy (1.6 %). Dose to the small bowel could be kept low in the segmented parts with 9.4 % of the small bowel exposed to ≥40 Gy and a maximum dose of 52.7 Gy. Most of the bladder could be spared from high dose exposure with 6.3 % of the bladder receiving ≥70 Gy.
Treatment safety
After a median observation time of 23.4 months one patient out of 38 died seven months after irradiation suffering from metastasized esophageal cancer. No patient showed acute toxicity ≥grade 3. We defined acute toxicity as side effects within 6 months after therapy. Therefore, at the time of evaluation the PLATIN 1 trial met the prospectively defined statistical criteria of a successful treatment with an SDR of at least 87.7 %.
Gastrointestinal toxicity
Cumulative incidence of acute GI toxicity was 56.8 % (grade 1) and 21.6 % (grade 2). No acute grade 3 or 4 GI toxicity occurred. During treatment, patients suffered Cumulative late GI toxicity was 6.1 % (grade 1). No patient suffered from late enteritis of any grade. Only one patient experienced late proctitis grade 1 at 12 months of follow-up, and one late diarrhea grade 1 at 18 months was observed (see Table 3).
Genitourinary toxicity
The incidence of acute GU toxicity is comparable to other published data with 78.4 % (grade 1) and 48.6 % (grade 2). Cumulative incidence of late GU toxicity was 12.3 % (grade 1) and 6.4 % (grade 2). No patient developed acute or late GU toxicity grade 3/4.
Acute cystitis was reported in 35.1 % (grade 1) and 18.9 % (grade 2) of all patients (see Table 4). Two patients reported stress incontinence grade 1 (occasional, no pads necessary) at the beginning of radiation therapy, which regressed about 12 months thereafter. Four other patients reported on stress incontinence grade 1 in one follow-up. In all cases, the symptoms resolved until the following visit.
One patient required urinary catheterization (for 8 days) during follow-up (6 days after the end of radiotherapy) due to urinary retention. As no other intervention was required, it was rated as grade 2 toxicity. For all other patients the urinary flow has hardly changed during and after therapy. At 24 months of follow-up, all patients were catheter-free.
Within the study, we also evaluated the incidence of adverse effects on libido and erectile dysfunction. However, since hormonal therapy is running within the first 2 years, toxicity analysis is reasonable only after discontinuation of AHT.
Quality of life
Overall health as assessed by the "Global Health Score" of the EORTC QLQ-C30 questionnaire remained almost unchanged at 6, 12 and 24 months follow-up compared to baseline (see Table 5). Scores were on a similar level
Biochemical control and survival
During follow-up, three patients experienced PSA recurrences. All three had discontinued AT after radiotherapy resulting in an actuarial biochemical progression free survival of 91.9 % after 2 years (see Fig. 1a). In one patient, PSA recurrence coincided with the diagnosis of bone metastases; for the other two patients, the localization of recurrence could not be determined. Ten patients quit AT after completion of radiotherapy due to reported intolerance and side effects. At the time of analysis, 15 patients still received AT. Average duration of AT was 13.6 months for the 38 patients remaining in the trial after the end of radiotherapy. At a median of 24 months followup, 37 patients were alive resulting in an actuarial overall survival (OS) of 97.3 % (see Fig. 1b).
Discussion
The procedure of EBRT for prostate cancer has undergone many changes in the past decade. Recently, dose escalation, hypofractionation and the use of IMRT and IGRT became a standard method. Hence, outcome data are slowly emerging. The necessity of prophylactic radiotherapy of the pelvic lymph nodes remains controversial. Many radiation oncology centers electively treat the pelvic lymph nodes because of publications on surgical lymph node sampling and nanoparticle-enhanced MRI studies that revealed a high proportion of occult lymph node metastases [19,20]. In contrast, other centers avoid WPRT because of concerns about excessive toxicity. For more clarity, the effect of WPRT was tested in large prospective trials. Unfortunately, the long-term PFS showed no significant differences between WPRT and PORT [21][22][23]. However, these trials were designed before the dose-escalation era [6,7,24,25]. With a dose-escalation up to 75.6 Gy to the prostate, Aizer et al. could prove the superiority of WPRT [8]. Especially in high-risk situations a benefit with reduced rates of failure were seen in patients treated with doses higher than 75 Gy [26,27].
The present phase II study shows good tolerability of IMRT-based treatment of the prostate and pelvic lymph nodes. We revealed a reduced incidence of toxicities compared to conventionally fractionated schemes. The RTOG 94-13 trial showed late grade ≥3 GI and GU toxicities at 5 years of 3.0 and 4.3 %, respectively, using conventionally fractionation of 1.8 Gy (50.4 Gy WPRT + 19.8 Gy boost to the prostate bed). The GETUG-01 trial applied doses to the prostate and whole pelvis of 66 -72 Gy à 1.8 -2.25 Gy and 45-46.8 Gy à 1.8 Gy. Here, 37.7 % of patients developed late grade ≥2 GU toxicity and 31.7 % developed late grade ≥2 GI toxicity. In our trial, the observed rates of late toxicity compare favorably to the mentioned studies with no grade ≥2 GI toxicity, and 6.4 % grade 2 and no grade 3 or 4 GU toxicity. On the one hand, the lower rates of side effects could be attributed to the modern IMRT/IGRT technique; on the other hand, the moderate hypofractionation could have an important impact: the α/β ratio of prostate cancer is supposed to be lower than the surrounding healthy tissue of rectum and bladder resulting in a probably lower rate of late side effects.
In our trial, patients received at least 2 months of neoadjuvant AT. With good tolerability AT was continued for at least 2 years after irradiation. The addition of antihormonal treatment (AT) in patients with clinically localized intermediate or high-risk prostate cancer showed superior results (bPFS, DSS, OS) in a number of prospective randomized trials [28][29][30][31]. The question of the right timing of adjuvant hormonal treatment (AHT) is still under debate and lasts from 4 months in the RTOG 86-10 trial to 3 years in the EORTC trial [32][33][34][35]. Bolla et al. demonstrated a significantly lower overall mortality in patients under neoadjuvant hormonal treatment (NHT) plus 3 years AHT compared to 6 months AHT [36]. Most randomized phase III trials used short-term NHT in combination with EBRT and showed an improvement either in absence of PSA failure or OS [33,37,38]. One of the possible effects of NHT is the immune modulation of AT resulting in T-cell infiltration of the prostate, which can increase apoptosis [39]. Most studies using long-term AHT for high-risk patients added whole pelvis radiotherapy (WPRT) [30,32,40]. The RTOG 92-02 trial proved that NHT combined with long-term AHT is superior to NHT for locally advanced/ high-risk patients [40]. While it is still unclear whether an additional WPRT is beneficial compared to prostate irradiation alone [21,22,41,42], the mentioned studies also showed no significant improvement of relevant clinical endpoints (OS, bPFS, DSS). The RTOG 94-13 trial planned to evaluate the timing of AT. The role of WPRT became rather complex as the authors found an interaction between field size and timing of AT. No benefit could be found in the trial testing WPRT vs. prostate only RT (PORT) in combination with AHT vs. NHT [21]. However, the results suggest that if a patient chooses NHT, WPRT appears beneficial compared to PORT. The risk of high-grade (> grade 3) GI toxicity is potentially higher under additional WPRT and AT compared to PORT. The evaluation of erectile function and libido based on patient-reported data is prone to reporting bias. Nevertheless, we documented the incidence of adverse effects on libido and erectile dysfunction as well. At this stage evaluation is too early, as for most patients AT is still running within the first 2 years of follow-up. Toxicity analysis is only reasonable if AT is completed.
Our study has shown both safety and efficacy using helical tomotherapy. At the point of analysis, 34 of 37 patients were free from PSA recurrence. However, it is important to note the limitations of our first analysis. The patient number is relatively small. For a reliable evaluation of efficacy, a median follow-up of 23.4 months is rather short. A longer follow-up period is needed to detect differences especially in late toxicity. Also, it will provide more significant reports on biochemical control.
Conclusion
While the role of WPRT of prostate cancer remains to be fully explored, we could demonstrate in the prospective PLATIN 1 trial that prophylactic radiotherapy of the pelvic lymph nodes with a SIB to the prostate can be performed without excessive toxicity. The combined irradiation of both prostate and pelvic lymph nodes seems to be as well tolerated as the irradiation of the prostate alone. | 2023-01-16T14:11:47.403Z | 2015-11-07T00:00:00.000 | {
"year": 2015,
"sha1": "63fe961b765c39006a9d4ed8011b7d4fb1ab4a1e",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.1186/s12885-015-1886-5",
"oa_status": "GOLD",
"pdf_src": "SpringerNature",
"pdf_hash": "63fe961b765c39006a9d4ed8011b7d4fb1ab4a1e",
"s2fieldsofstudy": [
"Medicine",
"Engineering"
],
"extfieldsofstudy": []
} |
184487642 | pes2o/s2orc | v3-fos-license | Optimizing Pipelined Computation and Communication for Latency-Constrained Edge Learning
Consider a device that is connected to an edge processor via a communication channel. The device holds local data that is to be offloaded to the edge processor so as to train a machine learning model, e.g., for regression or classification. Transmission of the data to the learning processor, as well as training based on Stochastic Gradient Descent (SGD), must be both completed within a time limit. Assuming that communication and computation can be pipelined, this letter investigates the optimal choice for the packet payload size, given the overhead of each data packet transmission and the ratio between the computation and the communication rates. This amounts to a tradeoff between bias and variance, since communicating the entire data set first reduces the bias of the training process but it may not leave sufficient time for learning. Analytical bounds on the expected optimality gap are derived so as to enable an effective optimization, which is validated in numerical results.
I. INTRODUCTION
E DGE learning refers to the training of machine learning models on devices that are close to the end users [1]. The proximity to the user is instrumental in facilitating a low-latency response, in enhancing privacy, and in reducing backhaul congestion. Edge learning processors include smart phones and other user-owned devices, as well as edge nodes of a wireless network that provide wireless access and computational resources [1]. As illustrated in Fig. 1, the latter case hinges on the offloading of data from the data-bearing device to the edge processor, and can be seen as an instance of mobile edge computing [2].
Research on edge learning has so far instead focused mostly on scenarios in which training occurs locally at the data-bearing devices. In these setups, devices can communicate either through a parameter server [3] or in a device-to-device manner [4]. The goal is to either learn a global model without exchanging directly the local data [5] or to train separate models while leveraging the correlation among the local data sets [6]. Devices can exchange either information about the local model parameters, as in federated learning [7], or gradient information, as in distributed Stochastic Gradient Descent (SGD) methods [8], [9].
In this work, we consider an edge learning scenario in which training takes place at an edge node of a wireless system as illustrated in Fig. 1. The data is held by a device and has to be offloaded through a communication channel to the edge node. The learning task has to be executed within a time limit, which might be insufficient to transmit the complete dataset. Transmission of data blocks from device to edge node, and training at the edge node can be carried out simultaneously (see Fig. 2). Each transmitted packet contains a fixed overhead, accounting e.g. for meta-data and pilots. Given the overhead of each data packet transmission, what is the optimal size of a communication block? Communicating the entire data set first reduces the bias of the training process but it may not leave sufficient time for learning. We investigate a more general strategy that communicates in blocks and pipelines communication and computation with an optimized block size, which is shown to be generally preferable. Analysis and simulation results provide insights into the optimal duration of the communication block and on the performance gains attainable with an optimized communication and computation policy.
The rest of this letter is organized as follows. In Sec. II, we provide an overview of the model and the associated notations. In Sec. III, we examine the technical assumptions necessary for our work. In Sec. IV, we provide our main result and discuss its implications. Finally, in Sec. V, we consider numerical experiments in the light of our result.
II. SYSTEM MODEL
As seen in Fig. 1, we study an edge learning system in which a device communicates with an edge node, and associated server, over an error-free communication channel. The device has access to a local training dataset X = {x 1 , x 2 , . . . , x N } of N data points {x n } N n=1 , and training of a machine learning model is carried out at the edge node based on data received from the device. As illustrated in Fig. 2, communication and learning must be completed within a time limit T . To this end, the transmissions are organized into blocks, and transmission and computing at the edge node can be performed in parallel.
Data Device
Channel Edge node OH b nc n o w b Server Figure 1. An edge computing system, in which training of a model parametrized by vector w takes place at an edge processor based on data received from a device using a protocol with timeline illustrated in Fig. 2 (OH = overhead).
Training at the edge node aims at identifying a model parametrized by a vector w ∈ R d within a given hypothesis class. Training is carried out by (approximately) solving the Empirical Risk Minimization (ERM) problem (see, e.g, [10]). This amounts to the minimization with respect to vector w of the empirical average L(w) of a loss function (w, x) over all the data points x in the training dataset, i.e., As detailed below, the minimization of the function L(w) is carried out at the edge node using SGD, based on the data points received from the device. In order to elaborate on the communication and computation protocol illustrated in Fig. 2, we normalize all time measures to the time required to transmit one data sample from the device to the edge node. With this convention, we denote as τ p the time required to make one SGD update at the edge node.
As seen in Fig. 2, transmission from the device to the edge node is organised into blocks. In this study, we ignore the effect of channel errors, which is briefly discussed in Sec. VI. In the b-th block, the device transmits a subset X b ⊆ X of n c new samples from its local dataset. At the end of the block, the edge node adds these samples to the subsetX b+1 of samples it has available for training in the b + 1-th block, i.e.,X b+1 =X b ∪ X b with X0 = ∅. The samples in X b are randomly and uniformly selected from the set ∆X b = X \X b of samples not yet transmitted to the edge node. A packet sent in any block contains an overhead, e.g., for pilots and meta-data, of duration n o , irrespective of the number n c of transmitted samples. It follows that the duration of a transmission block is n c + n o .
There are at most B d = N/n c transmission blocks, since B d blocks are sufficient to deliver the entire dataset to the edge node. Therefore, we need to distinguish two cases. As seen in Fig. 2(a), when T ≤ B d (n c + n o ), the device is only able to deliver a fraction of the samples. In particular, denoting as B = T /(n c + n o ) the number of blocks, the fraction of data points delivered at the edge node at time T equals (B − 1)/B d . In contrast, if T > B d (n c + n o ), as illustrated in Fig. 2(b), the edge node has the entire dataset available after B d blocks, that is, for a duration equal to τ l = T − B d (n c + n o ). Henceforth, we refer to this last period as block B l = B d + 1.
During each block b ≤ B d , the edge node computes n p = (n c + n o )/τ p local SGD updates (2). During block B l , the edge node computes n l = τ l /τ p SGD updates. The j-th local update at block b, with j = 1, . . . , n p , is given as where α is the learning rate, and ξ j b is a data point sampled i.i.d. uniformly from the subsetX b = b−1 l=1 X l of samples currently available at the edge node. Note that we haveX B l = X .
The goal of this work is to optimize the number of samples n c sent in each block with the aim of minimizing the empirical loss (1) at the edge node at the end of time T . In the next sections, we present an analysis of the empirical loss obtained at time T that allows us to gain insights into the optimal choice of n c .
III. TECHNICAL ASSUMPTIONS
In order to study the training loss achieved at the edge node at the end of the training process, we make the following standard assumptions, which apply, for instance, to linear models with quadratic or cross-entropy losses under suitable constraints (see the comprehensive review paper [9]): (A1) the sequence of iterates w j b in (2) is contained in a bounded open set W ⊆ R d with radius D = max u,w∈W×W ||w − u|| 2 over which the function (w, x) is bounded below by a scalar inf for all x; (A2) the function (w, x) is continuously differentiable in w for any fixed value of x and is L-smooth in w, i.e., for all (w,w) ∈ W × W, and for all x. This implies for all (w,w) ∈ W × W, and for all x; (A3) the loss function (w, x) is convex and satisties the Polyak-Lojasiewicz condition in w, i.e., there exists a constant c > 0 such that is a minimizer of (w, x). The P-L condition is implied by, but does not imply, strong convexity [9]. We further need to make assumptions on the statistics of the gradient ∇ (w, ξ j b ) used in the update (2). To this end, for each block b > 1, we define the empirical loss limited to the samples available at the edge node at block b as the empirical loss over the samples transmitted at iteration b ≥ 1 as and the empirical loss over the samples not available at the edge at iteration b > 1 Note that we have the identity L(w) First, we observe that given the previously transmitted data samples, the gradient ∇ (w j−1 b , ξ j b ) is an unbiased estimate of the gradient ∇L b (w) of the empirical loss limited to the samples available at the edge node at block b. In formulas, [ · ] is the conditional expectation given the previously transmitted samples. We finally make the following assumption (see, e.g., [9]): (A4) For any setX b of samples available at the edge node, there exist scalars M ≥ 0 and M V ≥ 0 such that
IV. CONVERGENCE ANALYSIS
In this section, we present our main result and its implications on the optimal choice of the number n c of transmitted samples per block. Henceforth, we use the notation E b [ · ] to indicate the conditional expectation E [ · ] as the conditional expectation on the samples selected for the SGD updates in block B l (see Fig. 2(b)). and define Then, for any sequenceX 1 , . . . ,X B the expected optimality gap at time T is upper bounded as if T ≤ B d (n c + n o ); and by The bound (12)-(13) extends the classical analysis of the convergence of SGD for the case in which the entire dataset is available at the learner [9, Theorem 4.6] to the set up under study. The bound distinguishes the case in which the edge node has the entire data set by the last block, and the complementary case, as seen in Fig. 2.
The first term in the bound (13) represents an asymptotic bias that does not vanish with the number of SGD updates, even when all the data points are available at the edge node. It is due to the variance (9) of the stochastic gradient. The bound (12) for smaller values of T also comprises an additional bias term, that is the second term in (13), due to the lack of knowledge about samples not received at the edge node by the end of the training process. In contrast, the last term in bound (12)-(13) accounts for the standard geometric decrease of the initial error in gradient-based learning algorithms. Here, the initial error for each block b is given by E b L(w np b−1 ) − L(w * ) . Note that the additional factor with exponent n l in (13) accounts for the number of updates made after all the samples have been received at the edge node.
The bound (12)-(13) can be in principle optimized numerically in order to find an optimal value to the block size n c . However, in practice, doing so would require fixing the choice of the sequenceX 1 , . . . ,X B , and running Monte Carlo experiments for every randomly selected sample of the sequence of SGD updates (2), which is computationally intractable. Therefore, in the following, we derive a generally looser bound that can be directly evaluated numerically without running any Monte Carlo simulations. This bound will then be used in order to obtain an optimized value for n c . if T ≤ B d (n c + n o ); and by We plot bound (14)-(15) in Fig. 3. These results are obtained for N = 18, 576, T = 1.5N , L = 1.908, c = 0.061, M = 1, M G = 1, τ p = 1, α = 0.0001. We note that L and c represent respectively the smallest and largest eigenvalues of the data Gramian matrix for the example studied in Sec. V. For each value of n o , we mark in the figure both the value of n c that minimizes the upper bound in Corollary 1 and the value of n c at which we have the condition T = B d (n c + n o ). As seen in Fig. 2, this is the minimum value of n c that allows the full transmission of the training set by the last training block.
A first observation is that the optimized value of n c , henceforth referred to asñ c , is generally smaller than the number N of training points in X , suggesting the advantages of pipelining communication and computation. Furthermore, as the overhead n o increases, it becomes preferable, in terms of the bound (14)-(15), to choose larger valuesñ c for the block size n c . This is because a larger value of n o needs to be amortized by transmitting more data in each block, lest the transmission time is dominated by overhead transmission. Finally, for smaller values of n o , the minimumñ c of the bound is obtained when the entire data set is eventually transferred to the edge node, i.e., T > B d (n c + n o ), while the opposite is true for larger value of n o . Interestingly, this suggests that it may be advantageous in terms of final training loss, to forego the transmission of some training points in exchange for more time to carry out training on a fraction of the data set.
V. NUMERICAL EXPERIMENTS
In this section, we validate the theoretical findings of the previous sections by means of a numerical example based on ridge regression on the California Housing dataset [11]. The dataset contains 20640 covariate vectors x n ∈ R 8 , each with a real label y n . We randomly select 90% of the samples to define the set X for training, i.e., we have N = 18576. As for Fig. 4, we choose τ p = 1 and α = 0.0001. The parameter vector is initialized using i.i.d. zero-mean Gaussian entries with unitary power. The loss function is defined as (w, x) = (w T x − y) 2 + λ N ||w|| 2 where w ∈ R 8 and the regularization coefficient is chosen as λ = 0.05.
By computing the average final training loss for each value of n c , we can experimentally determine the optimal value n * c of the block size. We compare the performance using this experimental optimum with the performance obtained using the minimumñ c of the bound (14)-(15). To this end, in Fig. 4, given a fixed overhead size n o , we plot the average training loss L(w j b ) against the normalized training time j for n * c and for the valueñ c obtained from the bound (14)-(15). As references, we also plot as dotted lines the losses obtained for selected values of n c . The choice of the block size n c minimizing the average final loss is seen to be a trade-off between the rate of decrease of the loss and the final attained accuracy. In particular, decreasing n c allows the edge node to reduce the loss more quickly, albeit with noisier updates and at the cost of a potentially larger final training loss due to the transmitted packet being dominated by the overhead. Importantly, determining the optimum block size experimentally instead of using bound (14)-(15) only provides a gain of 3.8% in terms of the final training loss, at the cost of a computationally burdensome parameter optimization.
VI. CONCLUSIONS
In this work, we considered an edge computing system in which an edge learner carries out training over a limited time period while receiving the training data from a device through a communication link. Considering a strategy that allows communication and computation to be pipelined, we have analysed the optimal communication block size as a function of the packet overhead. Among interesting directions for future work, we mention the inclusion of the effect of delays due to errors in the communication channel. In this case, the optimization problem could be generalized to account for the selection of the data rate. Other interesting extensions would be to consider online learning, where data sent in previous packets can be only partially stored at the server, and to investigate a scenario with multiple devices. APPENDIX A PROOF OF THEOREM 1 Using the same arguments as in the proof of [9, Theorem 4.6], we can directly obtain the following inequality for each block b: Note that we have w 0 b = w np b−1 , since the initial parameter at block b is the final parameter obtained at block b − 1. By definition of the local empirical losses (6)-(7), we have the equalitỹ Plugging (17) into (16), we have Iterating this substitution for all blocks b − 1, b − 2, . . . , 2, we obtain While inequality (19) applies for any choice of T , we now specialize the result to the case where the allocated amount of time T is not sufficient to transmit the whole dataset, i.e., T ≤ B d (n c + n o ). (see Fig. 2(a)). According to (6)- (8), for this case, we have the equality Plugging (20) into (19) for block b = B, we then obtain which is (12) in Theorem 1. Finally, we consider the case where there is sufficient time to transmit the whole dataset, i.e., T > B d (n c + n o ) (see Fig. 2(b)). According to (16), we have where (a) arises from plugging (21) in (22) with B = B d . This is (13) in Theorem 1, concluding the proof.
APPENDIX B PROOF OF COROLLARY 1 Defining for all t = 1, . . . , B d , the optimum solution ∆w * b = arg min w ∆L b (w), we can write ∆L b (∆w * b ) ≤ ∆L b (w * ) , and hence also the inequality Writing the Lipschitz continuity property of the gradients (A2) with ∇(∆L b (∆w * b )) = 0 and (A1), we have ∆L b (w which is (14) in Corollary 1. Following the same approach with (22), we obtain which is (15) in Corollary 1, completing the proof. | 2019-06-12T09:52:46.000Z | 2019-06-11T00:00:00.000 | {
"year": 2019,
"sha1": "70fbc0862aa66a36952d08a7a0ae1e41515dfdd1",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/1906.04488",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "688c0f36219c01b9a0b6310b729142193a021358",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science",
"Mathematics"
]
} |
249078466 | pes2o/s2orc | v3-fos-license | Practical Guidance for Diagnosing and Treating Iron Deficiency in Patients with Heart Failure: Why, Who and How?
Iron deficiency (ID) is a comorbid condition frequently seen in patients with heart failure (HF). Iron has an important role in the transport of oxygen, and is also essential for skeletal and cardiac muscle, which depend on iron for oxygen storage and cellular energy production. Thus, ID per se, even without anaemia, can be harmful. In patients with HF, ID is associated with a poorer quality of life (QoL) and exercise capacity, and a higher risk of hospitalisations and mortality, even in the absence of anaemia. Despite its negative clinical consequences, ID remains under-recognised. However, it is easily diagnosed and managed, and the recently revised 2021 European Society of Cardiology (ESC) guidelines on HF provide specific recommendations for its diagnosis and treatment. Prospective randomised controlled trials in patients with symptomatic HF with reduced ejection fraction (HFrEF) show that correction of ID using intravenous iron (principally ferric carboxymaltose [FCM]) provides improvements in symptoms of HF, exercise capacity and QoL, and a recent trial demonstrated that FCM therapy following hospitalisation due to acute decompensated HF reduced the risk of subsequent HF hospitalisations. This review provides a summary of the epidemiology and pathophysiology of ID in HFrEF, and practical guidance on screening, diagnosing, and treating ID.
Introduction
Heart failure (HF) impacts in the region of 26 million people across the world and due to the ageing population its prevalence is still increasing [1]. Although there have been advances to prevent and treat HF, it is still associated with substantial rates of mortality and morbidity as well as diminished patient quality of life (QoL) [1,2].
HF is defined as a syndrome characterised by cardinal symptoms, for example fatigue, breathlessness and ankle swelling, which may occur alongside signs including peripheral oedema, increased jugular venous pressure and crackles in the lung [3]. HF is caused be an abnormality of the heart, which may be functional and/or structural, resulting in increased pressure in the heart and/or a deficient cardiac output while resting and/or exercising [3].
Iron deficiency is an important and frequent comorbid condition in patients with HF [4][5][6][7][8][9]. In these patients, it independently predicts mortality and morbidity, and is also associated with impaired exercise capacity and reduced QoL [4][5][6][7][8][9]. The recently updated 2021 European Society of Cardiology (ESC) guidelines on HF acknowledge the importance of iron deficiency among patients with HF and also provide specific recommendations for diagnosing and appropriately treating the condition [3]. However, iron deficiency remains under-recognised and under-treated in clinical practice [10][11][12][13][14], likely due in part to a lack of practical guidance for clinicians that can be easily followed.
There are three main goals when treating patients with HF with reduced ejection fraction (HFrEF): (1) lessening mortality; (2) preventing recurrent hospitalisations due to HF worsening; and (3) improving functional capacity, clinical status and QoL [3]. Clinical trial evidence has shown that correcting iron deficiency with supplementary IV iron addresses two of the aforementioned treatment goals (reducing recurrent hospitalisations due to HF, and improving HF symptoms, functional status, and QoL) [15][16][17][18]. Hence, correction of iron deficiency in patients with HFrEF is recommended to improve these clinical outcomes [3].
The majority of patients with HF are managed primarily by general internal medicine physicians who play a crucial role in screening, diagnosing and subsequently treating iron deficiency. This article aims to provide a summary of iron deficiency in HF, along with practical guidance for its diagnosis and appropriate treatment. It aims to address the frequently asked questions of 'Why', 'Who', and 'How' to diagnose and appropriately treat iron deficiency in patients with HF.
Role of Iron and the Impact of Iron Deficiency
Iron deficiency is a clinical condition where the available iron is inadequate to fulfil the needs of the body [19]. Iron has a critical role in the function of every cell in the human body [7]. As an essential component of respiratory chain proteins in mitochondria, iron is key for cellular energy generation [20]. While iron is most widely recognised for its role in the transport of oxygen as a vital constituent of haemoglobin (Hb), it also has a major role in non-haematopoietic tissues, such as cardiac and skeletal muscle, which are dependent on iron for oxygen storage, mitochondrial energy production and many other cellular processes [20,21] (Figure 1). Thus, iron deficiency per se, even in the absence of anaemia (i.e., at a normal Hb level), can be harmful. Experimental studies show that iron deficiency directly weakens the ability of human cardiomyocytes to contract in vitro, and that this can be corrected by iron repletion [22]. In patients who have chronic HF (CHF), iron deficiency can be associated with breathlessness on exertion, increased fatigue, reduced exercise capacity [7,23,24], poorer health-related QoL [25,26], worse HF symptoms, increased HF hospitalisation and higher mortality [5,[27][28][29]. These adverse effects are independent of anaemia in patients who have HF and iron deficiency. Furthermore, anaemia does not affect these adverse outcomes in HF when corrected for iron deficiency and other prognostic markers, although patients with both iron deficiency and anaemia have worse outcomes [27][28][29]. Importantly, treatment of iron deficiency with intravenous (IV) iron is associated with improved functional status among patients with HF, even when Hb is normal [15,17,30]. prognostic markers, although patients with both iron deficiency and anaemia have worse outcomes [27][28][29]. Importantly, treatment of iron deficiency with intravenous (IV) iron is associated with improved functional status among patients with HF, even when Hb is normal [15,17,30]. [20,21,31]. ATP, adenosine triphosphate; Fe-S, iron-sulphur; Hb, haemoglobin; TCA, tricarboxylic acid.
Iron Deficiency Prevalence in Patients with Heart Failure
Iron deficiency is one of the most commonly seen comorbid conditions in patients who have HF, with studies reporting that approximately 40−70% of patients with CHF have iron deficiency [5,7,[32][33][34][35][36], regardless of their ejection fraction [9]. Iron deficiency also has a prevalence of up to 80% in patients with acute HF (AHF) [10,37]. Additionally, the prevalence of iron deficiency increases in severe HF (i.e., with higher New York Heart Association [NYHA] class [5]) and when anaemia is present [38].
Iron Deficiency Causes in Patients with Heart Failure
The aetiology of iron deficiency in HF is complex and multifactorial, with contradictory evidence on the precise cause(s) [29]. Factors that may contribute to iron deficiency include reduced appetite, co-administration of proton pump inhibitors, occult gastrointestinal blood loss and comorbidities such as chronic kidney disease and inflammatory activity [27,29,39,40]. The possible driving factors for iron deficiency in HF are summarised in Figure 2. Since hepcidin is tightly regulated by inflammatory activation as part of the antibacterial response mechanism and HF is a condition of increased inflammatory activation, patients with HF may have high levels of circulating hepcidin [29,[41][42][43]. Hepcidin inhibits iron absorption by binding to ferroportin, causing sequestration of iron in the reticuloendothelial system and reducing the available useable iron [29]. There is some evidence that, as HF progresses and iron deficiency develops, the circulating hepcidin levels may become low in patients with CHF [43,44].
Iron Deficiency Causes in Patients with Heart Failure
The aetiology of iron deficiency in HF is complex and multifactorial, with contradictory evidence on the precise cause(s) [29]. Factors that may contribute to iron deficiency include reduced appetite, co-administration of proton pump inhibitors, occult gastrointestinal blood loss and comorbidities such as chronic kidney disease and inflammatory activity [27,29,39,40]. The possible driving factors for iron deficiency in HF are summarised in Figure 2. Since hepcidin is tightly regulated by inflammatory activation as part of the antibacterial response mechanism and HF is a condition of increased inflammatory activation, patients with HF may have high levels of circulating hepcidin [29,[41][42][43]. Hepcidin inhibits iron absorption by binding to ferroportin, causing sequestration of iron in the reticuloendothelial system and reducing the available useable iron [29]. There is some evidence that, as HF progresses and iron deficiency develops, the circulating hepcidin levels may become low in patients with CHF [43,44]. [19,27,29,31,39,40,[43][44][45][46]. DOAC, direct oral anticoagulant; EPO, erythropoietin; GI, gastrointestinal; IL, interleukin; PPI, proton-pump inhibitor; RES, reticuloendothelial system; TNF-α, tumour necrosis factor alpha.
Who and When to Screen for Iron Deficiency?
The 2021 ESC HF guidelines recommend that every patient with HF should be periodically assessed for iron deficiency (and anaemia) including carrying out a full blood count, and measuring both serum ferritin concentration and transferrin saturation (TSAT) (recommendation class I, evidence level C) [3]; plasma iron level is not an adequate mirror of iron deficiency. This recommendation is a noteworthy update to the 2016 ESC HF Figure 2. Causes of iron deficiency in heart failure [19,27,29,31,39,40,[43][44][45][46]. DOAC, direct oral anticoagulant; EPO, erythropoietin; GI, gastrointestinal; IL, interleukin; PPI, proton-pump inhibitor; RES, reticuloendothelial system; TNF-α, tumour necrosis factor alpha.
Who and When to Screen for Iron Deficiency?
The 2021 ESC HF guidelines recommend that every patient with HF should be periodically assessed for iron deficiency (and anaemia) including carrying out a full blood count, and measuring both serum ferritin concentration and transferrin saturation (TSAT) (recommendation class I, evidence level C) [3]; plasma iron level is not an adequate mirror of iron deficiency. This recommendation is a noteworthy update to the 2016 ESC HF guidelines since screening was previously only recommended for new cases of HF [47]. Among the routine blood tests for comorbidities recommended for patients with suspected CHF, iron status (TSAT and ferritin) should also be tested (recommendation class I, evidence level C) [3]. Furthermore, determination of iron status (TSAT and ferritin) is recommended at pre-discharge in patients with AHF [3]. We previously published comprehensive practical recommendations related to diagnosing, treating and monitoring patients with HF and iron deficiency in line with the 2016 ESC HF guidelines [48]. In this article, we have updated our recommendations in line with the 2021 ESC guidelines and recent trial findings, and recommend that clinicians should periodically evaluate iron deficiency and anaemia in all patients with HF regularly as part of the clinical evaluation (i.e., one to two times per year), depending on the iron deficiency severity and HF. Iron status should also be checked in patients with suspected CHF, ambulatory patients or outpatients with worsening HF, and after hospitalisation for AHF. A step-by-step algorithm for screening, diagnosing, treating and monitoring patients with HF is provided in Figure 3.
I, evidence level C) [3]. Furthermore, determination of iron status (TSAT and ferritin) is recommended at pre-discharge in patients with AHF [3]. We previously published comprehensive practical recommendations related to diagnosing, treating and monitoring patients with HF and iron deficiency in line with the 2016 ESC HF guidelines [48]. In this article, we have updated our recommendations in line with the 2021 ESC guidelines and recent trial findings, and recommend that clinicians should periodically evaluate iron deficiency and anaemia in all patients with HF regularly as part of the clinical evaluation (i.e., one to two times per year), depending on the iron deficiency severity and HF. Iron status should also be checked in patients with suspected CHF, ambulatory patients or outpatients with worsening HF, and after hospitalisation for AHF. A step-by-step algorithm for screening, diagnosing, treating and monitoring patients with HF is provided in Figure 3. [48] in line with the 2021 ESC HF guidelines [3]). * TSAT = (concentration of serum iron/total capacity to bind iron) × 100. † Note: The use of ferric carboxymaltose has not been assessed in paediatric patients, and therefore treatment with ferric carboxymaltose is not advised in children less than 14 years of age. Full prescribing information can be found in the latest Summary of Product Characteristics [49]. Hb, haemoglobin; HF, heart failure; HFrEF, heart failure with reduced ejection fraction; ID, iron deficiency; IV, intravenous; LVEF, left ventricular ejection fraction; TSAT, transferrin saturation.
How to Diagnose Iron Deficiency in Patients with Heart Failure
Iron status can easily be determined by measuring two readily available blood biomarkers: ferritin and TSAT [19]. Ferritin is a protein for storing iron within cells that is found in every cell type. Serum ferritin concentration is a surrogate marker for the total iron stored in healthy individuals [45]. TSAT is an indicator of the amount of iron circulating in the body that is available to supply metabolising cells and is defined as the percentage (%) of transferrin which is bound to iron [45].
In patients with HF, iron deficiency should be diagnosed when serum ferritin is <100 µg/L or TSAT is <20% when serum ferritin is 100-299 µg/L [3]. Two different thresholds are used since serum ferritin may be increased in response to inflammation, such as that seen in CHF, since it is an acute-phase reactant and can therefore appear to fall inside the normal range of 100-300 µg/L [19]. In this situation, a TSAT value of <20% is used to confirm the iron deficiency diagnosis [19]. In line with the 2021 ESC HF guidelines [3], ferritin and TSAT should be assessed at the same time to ensure the correct diagnosis of iron deficiency is made.
Although lower ferritin thresholds (e.g., <30 µg/L) are used for diagnosis of iron deficiency in other disease areas, it is important to use the thresholds specified above for the diagnosis of iron deficiency in patients who have HF. It is also critical to note that other laboratory parameters, such as mean values of corpuscular volume, corpuscular Hb and corpuscular Hb concentration are not reliable markers of iron deficiency status [50], so should not be used for determining iron deficiency status in patients who have HF. Furthermore, the measurement of only serum iron should not be utilised as an iron deficiency marker, since serum iron concentrations may differ considerably between individual patients with HF and can also display large diurnal fluctuations [51]. When evaluating iron status, it is also important to check for the presence of anaemia, which should be diagnosed using the Hb thresholds of <12 g/dL in females and <13 g/dL in males [52].
How Should Iron Deficiency in Patients with Heart Failure Be Treated?
Given the serious clinical impact of iron deficiency on patients with HF, it is vital that if diagnosed, this condition is treated.
Recommendations for Correcting Iron Deficiency
The 2021 ESC HF guidelines recommend that IV FCM should be considered for the treatment of iron deficiency in: • Symptomatic patients who have a left ventricular ejection fraction (LVEF) < 45% to alleviate symptoms, improve exercise capacity and QoL (recommendation class IIa, evidence level A) • Pre-and post-discharge follow-up of patients hospitalised for AHF to improve symptoms and reduce rehospitalisation (recommendation class IIa, evidence level B) • Symptomatic patients recently hospitalised for HF with LVEF < 50% to lessen the risk of HF hospitalisation (recommendation class IIa, evidence level B) [3].
Evidence on the Therapeutic Management of Iron Deficiency
Ferric carboxymaltose (FCM), a precision-engineered nanomedicine with a characteristic clinical profile [54], is the most extensively studied IV iron in randomised controlled clinical trials of patients with CHF [15][16][17][18]. Therefore, the majority of the evidencebase for IV iron in HF applies to IV FCM and, as such, FCM is the only iron formulation specifically recommended for the treatment of iron deficiency in the 2021 ESC HF guidelines [3].
Evidence on the Therapeutic Management of Iron Deficiency
Ferric carboxymaltose (FCM), a precision-engineered nanomedicine with a characteristic clinical profile [54], is the most extensively studied IV iron in randomised controlled clinical trials of patients with CHF [15][16][17][18]. Therefore, the majority of the evidence-base for IV iron in HF applies to IV FCM and, as such, FCM is the only iron formulation specifically recommended for the treatment of iron deficiency in the 2021 ESC HF guidelines [3].
The largest randomised controlled trials to evaluate FCM in patients who were irondeficient and had stable CHF (LVEF ≤ 45%) were the FAIR-HF [15], CONFIRM-HF [17], EFFECT-HF [18] and AFFIRM-AHF [16] studies. A summary of the designs and key efficacy and safety findings of these trials is shown in Table 1.
The FAIR-HF study [15] assessed whether treatment with FCM provided a significant improvement of the two primary endpoints: functional capacity as assessed by NYHA functional score and patients' self-reported perception of wellbeing (Patient Global Assessment [PGA]). This treatment benefit was evident after only 4 weeks of treatment with FCM and was sustained for the duration of the 24-week study. FCM treatment was beneficial for the reduction of HF symptoms, and in improving functional capacity and QoL. The treatment benefits of FCM were comparable among patients either with or without anaemia. FCM was well tolerated, and rates of adverse events, serious adverse events, and death were similar in both the FCM and placebo groups. The CONFIRM-HF study [17] evaluated the longer-term efficacy and safety of FCM. In this study, FCM significantly prolonged the Week 24 6-min walk test (6 MWT) distance (a difference of 33 ± 11 metres between the FCM and placebo groups [p = 0.002]), and this treatment effect was maintained until Week 52. Patients treated with FCM also achieved benefits to their PGA, NYHA class, QoL and fatigue score, compared with those receiving placebo. These improvements were statistically significant from Week 24 onwards, and the treatment benefits lasted up to 1 year. Patients treated with FCM were also found to have a significantly reduced risk of hospitalisation due to HF worsening compared with those in the placebo group (hazard ratio [HR]: 0.39 [95% confidence interval (CI) 0.19-0.82], p = 0.009). The mean dose received by patients was 1500 mg of iron over the 12-month study period, and >75% of the patients needed a total of two injections of FCM for correction and maintenance of iron parameters. Analysis of safety outcomes found that the frequency of adverse events and deaths were comparable between the two treatment groups.
The EFFECT-HF study [18] evaluated whether FCM could improve exercise intolerance, based on the assessment of alteration in peak VO 2 from baseline to Week 24. FCM had a favourable effect on peak VO 2 , compared with the control (treatment with standard of care), regardless of baseline anaemia status, and also significantly improved PGA score and NYHA functional class of patients in the study. In this study FCM was mostly well tolerated; there were no hypersensitivity reactions to FCM nor cases of hypophosphataemia reported.
Although the initial randomised, placebo-controlled clinical trials established that IV FCM treatment improved symptoms, functional capacity and health-related QoL of HFrEF patients with iron deficiency, they were not planned or sufficiently powered to assess the treatment effects on hard outcomes, such as hospitalisations and mortality. However, metaanalyses of FCM vs. placebo randomised controlled trials of patients with CHF who have iron deficiency, including the CONFIRM-HF and FAIR-HF studies, indicated that IV FCM treatment reduced the risk of all-cause death or cardiovascular (CV) hospitalisation, CV death or HF hospitalisation, and all-cause/CV death or recurrent CV/HF hospitalisations as combined endpoints [30,56].
Subsequently, the AFFIRM-AHF study evaluated the FCM treatment effect when initiated as early as hospital discharge on mortality and morbidity of patients who were hospitalised due to acute decompensated HF with LVEF < 50% and iron deficiency [16]. Overall, 1108 patients with HF randomised to treatment with FCM (n = 558) or placebo (n = 550) for up to 24 weeks were considered in the analysis [16]. The study reported 293 primary events in the FCM vs. 372 in the placebo groups (rate ratio [RR]: 0.79, 95% CI 0.62−1.01, p = 0.059) for the primary composite endpoint of total hospitalisations for HF and CV deaths for up to 52 weeks, failing to reach the standard statistical significance level of 5% ( Figure 5). The secondary endpoint analyses showed that treatment with FCM significantly reduced the risk of HF hospitalisations by 26% compared with placebo (RR: 0.74, 95% CI 0.58-0.94; p = 0.013), and this treatment benefit was observed for anaemic and non-anaemic patients. Statistically significant treatment benefits with FCM therapy vs. placebo were also observed for the composite endpoint of time to first HF hospitalisation or CV death (HR: 0.80, 95% CI 0.66-0.98, p = 0.03) and for days lost due to HF hospitalisations and CV death (RR: 0.67, 95 CI 0.47-0.97; p = 0.035). Additionally, patients in the AFFIRM-AHF study receiving FCM also had significantly greater improvements in the Kansas City Cardiomyopathy Questionnaire (KCCQ) compared with patients receiving placebo: adjusted mean differences (95% CI) at Week 4 were 2.9 (0.5-5.3, p = 0.018) for overall summary score (OSS) and 2.8 (0.3-5.3, p = 0.029) for clinical summary score (CSS), and at Week 24 were 3.0 (0.3-5.6, p = 0.028) for OSS and 2.9 (0.2-5.6, p = 0.035) for CSS [57]. Treatment with FCM was well tolerated by patients in AFFIRM-AHF and there were no unexpected safety findings [16].
in the AFFIRM-AHF study receiving FCM also had significantly greater improvements in the Kansas City Cardiomyopathy Questionnaire (KCCQ) compared with patients receiving placebo: adjusted mean differences (95% CI) at Week 4 were 2.9 (0.5-5.3, p = 0.018) for overall summary score (OSS) and 2.8 (0.3-5.3, p = 0.029) for clinical summary score (CSS), and at Week 24 were 3.0 (0.3-5.6, p = 0.028) for OSS and 2.9 (0.2-5.6, p = 0.035) for CSS [57]. Treatment with FCM was well tolerated by patients in AFFIRM-AHF and there were no unexpected safety findings [16]. Figure 5. Key primary and secondary outcome results from AFFIRM-AHF [16]. * AFFIRM-AHF primary endpoint narrowly missed statistical significance. AFFIRM-AHF, Study to Compare Ferric Carboxymaltose With Placebo in Patients With Acute Heart Failure and Iron Deficiency; CI, confidence interval; CV, cardiovascular; HF, heart failure; HR, hazard ratio; RR, rate ratio.
Safety and Tolerability of FCM
Evidence from clinical trials has shown that FCM is well tolerated by patients with HF and has a favourable safety profile [15][16][17][18]. The most commonly reported adverse drug reactions in patients who received FCM in clinical trials and real-world practice (occurring in ≥1% to 10% patients) were dizziness, flushing, headache, hypertension, hypophosphataemia, injection-/infusion-site reactions and nausea [49]. Anaphylactoid/anaphylactic reactions are rare (≥1/10,000 to <1/1000) and fatalities have been reported [49]. Moderate or severe hypophosphataemia has more commonly been reported in patients treated with FCM within the cardiology therapy area (9.9%) than the neurology and gastroenterology therapy areas (39% and 47.1%, respectively), but hypophosphataemia does not result in serious clinical outcomes for most patients across the populations studied [58]. Although a higher incidence of hypophosphataemia has been reported with FCM in certain patient subgroups [49,[58][59][60], such as those who have had a kidney transplantation [60], hypophosphataemia was reported at the same frequency in patients with HF who received FCM or placebo (0.2% in each arm) in the AFFIRM-AHF trial [16]. However, it should be noted that the product label specifies that serum phosphate levels should be monitored in those patients who receive multiple higher-dose injections of FCM or receive FCM long term, and in those patients with pre-existing factors that put them at risk for hypophosphataemia [49].
Oral Iron Substitution
Utilisation of oral iron for repletion of deficient iron in patients with HF was specifically evaluated in the 16-week, single, randomised, double-blind, placebo-controlled IRO-NOUT HF clinical trial [61]. This study assessed the effect of oral iron polysaccharide Figure 5. Key primary and secondary outcome results from AFFIRM-AHF [16]. * AFFIRM-AHF primary endpoint narrowly missed statistical significance. AFFIRM-AHF, Study to Compare Ferric Carboxymaltose With Placebo in Patients With Acute Heart Failure and Iron Deficiency; CI, confidence interval; CV, cardiovascular; HF, heart failure; HR, hazard ratio; RR, rate ratio.
Safety and Tolerability of FCM
Evidence from clinical trials has shown that FCM is well tolerated by patients with HF and has a favourable safety profile [15][16][17][18]. The most commonly reported adverse drug reactions in patients who received FCM in clinical trials and real-world practice (occurring in ≥1% to 10% patients) were dizziness, flushing, headache, hypertension, hypophosphataemia, injection-/infusion-site reactions and nausea [49]. Anaphylactoid/anaphylactic reactions are rare (≥1/10,000 to <1/1000) and fatalities have been reported [49]. Moderate or severe hypophosphataemia has more commonly been reported in patients treated with FCM within the cardiology therapy area (9.9%) than the neurology and gastroenterology therapy areas (39% and 47.1%, respectively), but hypophosphataemia does not result in serious clinical outcomes for most patients across the populations studied [58]. Although a higher incidence of hypophosphataemia has been reported with FCM in certain patient subgroups [49,[58][59][60], such as those who have had a kidney transplantation [60], hypophosphataemia was reported at the same frequency in patients with HF who received FCM or placebo (0.2% in each arm) in the AFFIRM-AHF trial [16]. However, it should be noted that the product label specifies that serum phosphate levels should be monitored in those patients who receive multiple higher-dose injections of FCM or receive FCM long term, and in those patients with pre-existing factors that put them at risk for hypophosphataemia [49].
Oral Iron Substitution
Utilisation of oral iron for repletion of deficient iron in patients with HF was specifically evaluated in the 16-week, single, randomised, double-blind, placebo-controlled IRONOUT HF clinical trial [61]. This study assessed the effect of oral iron polysaccharide supplementation at a high dose on exercise capacity among patients with HFrEF (LVEF < 40%) and iron deficiency. Compared with placebo, high-dose oral iron polysaccharide failed to increase exercise capacity, with no significant improvement in the primary endpoint of peak oxygen consumption (peak VO 2 ) or in 6 MWT distance over 16 weeks. The study also showed that oral iron polysaccharide therapy provided negligible recovery of stored iron among patients treated with oral iron therapy [61]. Overall, the IRONOUT HF study findings demonstrated that supplementation with oral iron polysaccharide is not an effective strategy for iron deficiency treatment in patients with HFrEF [61], and consequently the 2021 ESC HF guidelines do not recommend oral iron use in patients with HF [3].
Which Patients with Heart Failure Should Receive IV Iron?
FCM treatment benefit has been confirmed by multiple clinical trials in HFrEF [15][16][17][18]. The FAIR-HF, CONFIRM-HF and EFFECT-HF studies involved patients with stable CHF and NYHA class II/III who had a LVEF ≤ 45% [15,17,18]. The AFFIRM-AHF study involved patients with iron deficiency who had an LVEF < 50% and had stabilised following an episode of AHF [16]. A series of prespecified subgroup analyses of the AFFIRM-AHF study showed a consistent effect of FCM on the composite primary outcome across multiple subgroups [16]. While there were interesting observations in terms of the rate ratios when patients were stratified by chronic kidney disease stage, HF aetiology, and HF history, subgroup analyses are of limited power, and therefore, no definitive conclusions can be made on the basis of the subgroup analyses of the AFFIRM-AHF study.
Little is known about iron deficiency in HF with preserved ejection fraction (HFpEF), and a treatment benefit with IV iron has not been determined in patients with HFpEF since these patients were excluded from previous trials. The aim of the FAIr-HFpEF (NCT03074591) clinical trial, which is currently underway, is to assess the safety and efficacy of IV iron in patients with HFpEF who are iron deficient with or without anaemia ( Table 2) [62].
It should also be noted that parenteral iron must be used cautiously in cases with acute or chronic infection, asthma, atopic allergies or eczema [49]. Additionally, in patients with ongoing bacteraemia, it is recommended that IV FCM treatment should be stopped. Furthermore, a benefit-risk assessment should be carried out in patients who have a chronic infection which considers erythropoiesis suppression [49].
How to Administer IV Ferric Carboxymaltose and Monitor Patients after Treatment
As previously described [48], administration of IV FCM treatment is based on the patient's iron need calculated using their weight and Hb (Figure 3 includes a dosing table) [49]. FCM can be administered by IV injection as a slow undiluted bolus (at a rate of 100 mg per minute, or 1000 mg over 15 min), or an infusion that requires dilution [49].
As an infusion, FCM should not be over-diluted to ensure its stability is maintained [49]. The maximum recommended cumulative FCM dose is 1000 mg of iron equivalent to 20 mL FCM per week. IV iron should only be administered in the immediate vicinity of staff trained to assess and treat anaphylactic reactions, and in a location where full resuscitation facilities are available [49,63]. Following every IV iron injection, observation of the patient for any adverse effects is required for a minimum of 30 min [49].
Iron status should then be re-assessed after three months following iron replacement and further repletion provided as required. As indicated, patients should also be evaluated for loss of blood. It is important to avoid early re-assessment of iron status (i.e., occurring within four weeks of the administration of IV iron) since ferritin markedly increases following such administration, and therefore ferritin levels should not be used early on to indicate iron status. In agreement with the 2021 ESC HF guidelines [3], this working group recommends periodically and regularly evaluating iron deficiency and anaemia in all patients with HF as part of clinical evaluation (i.e., one to two times per year, depending on the severity of iron deficiency and HF). Anaemia and iron deficiency should also be evaluated when HF is decompensated, or when symptoms continue even though a patient has received optimised background treatment for HF. IV iron should then be administered as needed.
Evidence for Erythropoiesis-Stimulating Agent Therapy
The 2021 ESC HF guidelines state that in HF, erythropoiesis-stimulating agent (ESA) treatment of anaemia is not recommended in cases where there are no other indications for this therapy (recommendation class III) [3]. This was determined on the basis of findings from a sizeable randomised clinical trial showing that darbepoetin-alpha did not reduce the risk of HF hospitalisations or all-cause mortality, and the risk of thromboembolic events was found to be increased in patients with HFrEF and mild-to-moderate anaemia [64]. [67] Open-label, randomised, standard of care-controlled trial 1300 patients Key inclusion criteria: -Age ≥18 years -LVEF < 45% within the previous 2 years using any conventional imaging modality -NYHA class II-IV -Iron deficiency: ferritin <100 ug/L and/or TSAT < 20% -Evidence of high risk HF with expectation of survival to discharge including hospitalisation for HF currently or within the past 6 months, OR outpatients in atrial fibrillation with NT-proBNP >1000 ng/L or in sinus rhythm with NT-proBNP >250 ng/L (or BNP 300 pg/mL or >75 pg/mL, respectively) Iron (III) isomaltoside 1000 CV mortality or hospitalisation for worsening HF
Ongoing Research on IV Irons in HF
Prospective, randomised, controlled clinical trials are currently ongoing to investigate the benefit of IV iron on mortality and morbidity outcomes among patients with HF who have iron deficiency (Table 2), and are expected to read out within the next two years. These include the FAIR-HF2 (NCT03036462) [65], FAIR-HFpEF (NCT03074591) [62], HEART-FID (NCT03037931) [66], IRONMAN (NCT02642562) [67] trials, which are evaluating the effects of IV iron vs. placebo among iron-deficient patients with stable CHF.
Conclusions
Iron deficiency is one of the most frequent comorbid conditions in HF and can exist with or without anaemia. Iron deficiency has been recognised as a risk factor for worse outcomes associated with reduced exercise capacity and QoL, worse HF symptoms, and a higher risk for hospitalisations and mortality in patients with HF. Therefore, both the prompt diagnosis and appropriate correction of iron deficiency are crucial. Evidence from prospective randomised clinical trials show that supplementation with IV iron is a highly efficacious treatment in iron-deficient symptomatic patients with CHF. IV iron can provide significant improvements in patients' functional status, exercise capacity and health-related QoL, as well as lessen the risk of recurrent HF hospitalisations among patients hospitalised due to AHF. The 2021 ESC HF guidelines recommend that IV FCM treatment should be considered in: patients with HF who are symptomatic, have a LVEF < 45% and iron deficiency to improve symptoms of HF, increase exercise capacity and QoL; patients hospitalised for AHF and iron deficiency at pre-and post-discharge follow-up to alleviate symptoms and reduce rehospitalisations; and symptomatic patients who have a LVEF ≤ 50% and iron deficiency recently hospitalised for HF, to reduce the risk of HF hospitalisation [3].
Author Contributions: All authors participated in drafting and critically reviewing the manuscript. Before submission, all authors approved the finalised manuscript. All authors have read and agreed to the published version of the manuscript.
Funding: Development of this review article and funds to pay the publication charges for Open Access were supported by Vifor Pharma, Glattbrugg, Switzerland.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable. | 2022-05-27T15:24:06.221Z | 2022-05-25T00:00:00.000 | {
"year": 2022,
"sha1": "5943436b280abe6a5283a9083a7d12822ad11fff",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2077-0383/11/11/2976/pdf?version=1653470764",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "511eb0257b362ae048174f25b88c22890bf8be6a",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
248530206 | pes2o/s2orc | v3-fos-license | The Superoxide Dismutase Gene Family in Nicotiana tabacum: Genome-Wide Identification, Characterization, Expression Profiling and Functional Analysis in Response to Heavy Metal Stress
Superoxide dismutases (SODs) play an important role in protecting plants against ROS toxicity induced by biotic and abiotic stress. Recent studies have shown that the SOD gene family is involved in plant growth and development; however, knowledge of the SOD gene family in tobacco is still limited. In the present study, the SOD gene family was systematically characterized in the tobacco genome. Based on the conserved motif and phylogenetic tree, 15 NtSOD genes were identified and classified into three subgroups, including 5 NtCSDs, 7 NtFSDs and 3 NtMSDs. The predicted results of the transport peptide or signal peptide were consistent with their subcellular localization. Most NtSOD genes showed relatively well-maintained exon-intron and motif structures in the same subgroup. An analysis of cis-acting elements in SOD gene promoters showed that NtSOD expression was regulated by plant hormones, defense and stress responses, and light. In addition, multiple transcription factors and miRNAs are predicted to be involved in the regulation of NtSOD gene expression. The qPCR results indicated specific spatial and temporal expression patterns of the NtSOD gene family in different tissues and developmental stages, and this gene family played an important role in protecting against heavy metal stress. The results of functional complementation tests in the yeast mutant suggested that NtCSD1a, NtFSD1e and NtMSD1b scavenge ROS produced by heavy metal stress. This study represents the first genome-wide analysis of the NtSOD gene family, which lays a foundation for a better understanding of the function of the NtSOD gene family and improving the tolerance of plants to heavy metal toxicity.
Superoxide dismutases (SODs) play an important role in protecting plants against ROS toxicity induced by biotic and abiotic stress. Recent studies have shown that the SOD gene family is involved in plant growth and development; however, knowledge of the SOD gene family in tobacco is still limited. In the present study, the SOD gene family was systematically characterized in the tobacco genome. Based on the conserved motif and phylogenetic tree, 15 NtSOD genes were identified and classified into three subgroups, including 5 NtCSDs, 7 NtFSDs and 3 NtMSDs. The predicted results of the transport peptide or signal peptide were consistent with their subcellular localization. Most NtSOD genes showed relatively well-maintained exon-intron and motif structures in the same subgroup. An analysis of cis-acting elements in SOD gene promoters showed that NtSOD expression was regulated by plant hormones, defense and stress responses, and light. In addition, multiple transcription factors and miRNAs are predicted to be involved in the regulation of NtSOD gene expression. The qPCR results indicated specific spatial and temporal expression patterns of the NtSOD gene family in different tissues and developmental stages, and this gene family played an important role in protecting against heavy metal stress. The results of functional complementation tests in the yeast mutant suggested that NtCSD1a, NtFSD1e and NtMSD1b scavenge ROS produced by heavy metal stress. This study represents the first genome-wide analysis of the NtSOD gene family, which lays a foundation for a better understanding of the function of the NtSOD gene family and improving the tolerance of plants to heavy metal toxicity.
INTRODUCTION
Increasingly severe heavy metal pollution has exerted serious effects on crop growth, yield and quality. However, heavy metals such as Cu, Zn, Fe and Mn are essential for plant growth and development as micronutrient elements but become toxic when present in excess concentrations. Moreover, trace amounts of non-essential elements such as Cd and Hg are highly toxic to plants. For example, exposure of plants to Cd triggers severe symptoms, including chlorosis, root tip browning, stunted growth, and even plant death (Nagajyoti et al., 2010). Generally, heavy metals cause harmful physiological processes in plant cells, including the induction of reactive oxygen species (ROS) generation by changing the intracellular antioxidant defense system, binding the sulfhydryl, histidine and carboxyl groups of proteins and inactivating proteins. The substitution of essential ions at specific sites of proteins causes a loss of function (Hossain et al., 2012). Oxidative stress induced by heavy metals causes oxidative damage to cell membranes, proteins and nucleic acids and even increases cell death (Mittler, 2002). Therefore, scavenging excessive ROS in plant cells is an important strategy for plants to resist the toxicity induced by heavy metals.
Maintaining an optimal intracellular ROS level is essential for plant growth and development (Mittler, 2017). Generally, the ROS generated in chloroplasts, mitochondria and other organelles are maintained in a stable balance by the antioxidant defense system (ADS) in plant cells. The ADS in plants usually contains two categories of molecules: non-enzymatic antioxidant active substances, including ascorbic acid, glutathione, phenolic acid, and flavonoids; and antioxidants, such as superoxide dismutase and catalase. The excessive accumulation of intracellular ROS is scavenged by the complex ADS in plant cells. Among antioxidant systems, superoxide dismutase (SOD, EC 1.15.1.1) plays a pioneer role in scavenging ROS by activating a series of biochemical processes. Its main biofunction is to convert superoxide radicals into oxygen and hydrogen peroxide and protect plant cells from oxidative damage (Su et al., 2021). SODs are metalloproteinases, which are usually encoded by a gene family. According to the different metal cofactors, the SOD family genes in higher plants are classified into three subfamilies: copper zinc SOD (CSD), ferrum SOD (FSD) and manganese SOD (MSD) (Fink and Scandalios, 2002;Abreu and Cabelli, 2010). CSD is most widely distributed in the cytoplasm, chloroplast, peroxisomes, glyoxalic acid cycle and extracellular space. FSD is mainly located in chloroplasts, while MSD is usually located in the mitochondrial matrix and peroxisome (Pilon et al., 2011). In addition to scavenging ROS, SODs also play important roles in electron transport, photosynthesis and signal transmission. For example, CSD loss-of-function in Arabidopsis results in significant inhibition of plant growth and development and decreased chloroplast size, chlorophyll content and photosynthetic activity compared with the wild-type plant (Rizhsky et al., 2003). The Arabidopsis thaliana fsd1 mutant extends fewer lateral roots than the WT strain, but root growth was resumed by expressing FSD (Dvořák et al., 2020). Based on these results, SOD plays an important role in plant growth and responses to environmental changes by maintaining ROS homeostasis.
Superoxide dismutase plays a key role in plant resistance to drought, cold, salinity, and heavy metal toxicity, and regulates plant senescence (Sahu et al., 2017;Pan et al., 2019;Zang et al., 2020;Liu et al., 2021). In recent years, a positive correlation was observed between plant metal toxicity tolerance and antioxidant enzyme activity in heavy metal stress studies (Hasanuzzaman et al., 2020). For example, Ni stress significantly activated the antioxidant enzymes such as SOD and glutathione peroxidase (GPX) in rice leaves (Hasanuzzaman et al., 2019). Cd stimulates the activity of the antioxidant enzymes SOD, CAT and POD in Pseudochlorella pringsheimii to scavenge heavy metal-induced ROS and exhibits a dose-dependent effect (Ismaiel and Said, 2018). Over-expression of the SaCSD gene from Sedum alfredii in Arabidopsis increases Cd tolerance in transgenic plants by scavenging ROS (Li et al., 2017). However, the response of different SOD isozymes to heavy metal stress differed. Significantly higher activity of CSD was observed in the leaves and roots of Tagetes patula under Cd stress, while MSD activity was lower (Liu et al., 2011). In Pisum sativum plants, the activity of CSD localized in the cytoplasm was significantly inhibited by Cd, and FSD was more resistant than CSD, while MSD, the most resistant isoform to Cd, maintained 50% activity under 40 µM Cd stress (Sandalio et al., 2001). Therefore, an in-depth understanding of the fine-tuned mechanisms of ROS homeostasis depends on systematic studies of the functions of plant SOD gene family members.
Genome-wide identification of the SOD gene family has been performed in several plants, including monocots such as rice (Yadav et al., 2019), wheat (Jiang et al., 2019), and sorghum (Filiz and Tombuloglu, 2015), and dicots such as cotton , alfalfa (Song et al., 2018), and rapeseed (Su et al., 2021). Tobacco is the most widely cultivated non-food cash crop worldwide. To date, studies on the genome-wide identification of the SOD gene family in tobacco have not been reported. In the present study, the SOD gene family of tobacco cultivar TN90 was identified at the whole genome level, and its sequence characteristics and gene structure were systematically analyzed. The promoter sequences and cis-acting elements of SOD gene family members were predicted and analyzed. Transcription factors and miRNAs that may be involved in the regulation of SOD gene expression were analyzed. The tissue expression pattern and induced expression pattern of the NtSOD genes in plants under different metal stresses were analyzed using qPCR to further explore the function of the NtSOD genes. This study laid a foundation for further study of SOD gene function in tobacco exposed to heavy metal stress.
Identification and Sequence Analysis of the NtSOD Gene Family
The annotation information of genomic coding and protein sequences of N. tabacum cultivar TN90 were downloaded from the NCBI Genome database 1 . The HMM profiles of CSD (PF00080), FSD (PF02777) and MSD (PF00081) were downloaded from the Pfam database 2 , which was used as the query for searching the deduced protein sequences using the HMMER search program 3 . The candidate protein sequences were submitted to the SMART database 4 and NCBI-CCD website 5 to verify the conserved domains of SOD proteins. The molecular weight and theoretical iso-electric point of NtSOD proteins were calculated using the ProtParam tool in ExPASy web 6 . The subcellular localization of NtSOD proteins was predicted using the BUSCA web server 7 (Savojardo et al., 2018).
Phylogenetic Analysis
Twenty-three SOD proteins were selected to construct an unrooted phylogenetic tree using MEGA 7 with the neighborjoining (NJ) algorithm and 1,000 bootstrap replicates to investigate the phylogenetic relationships of the SOD genes between tobacco and A. thaliana; other parameters all used the default setting. Finally, the iTOL online tool 8 was used to visualize the phylogenetic trees.
Multiple Sequence Alignment, Motif Composition and Gene Structure
Multiple sequence alignment was performed with the MAFFT software (Katoh and Standley, 2013) using the FFT-NS-2 algorithm and subsequently visualized using Jalview software packages to construct consensus sequences of NtSOD family members (Waterhouse et al., 2009). An internal consistency analysis was conducted on each of the three groups of NtSOD genes using BioEdit software. Multiple EM for Motif Elicitation (MEME) suite (http://meme-suite) was used with the default parameters to determine the distribution of conserved motifs within the NtSOD proteins (Bailey et al., 2015). The diagrams of the exon-intron structure of NtSOD genes were generated using the online tool GSDS 9 according to the available coding sequence and their respective genomic sequence (Hu et al., 2015). The composite picture of the phylogenetic tree, motif distribution and gene structure of NtSOD genes was generated using the Gene Structure View program of TBtools software .
Prediction of Cis-Acting Elements
Two thousand-bp sequences upstream of the translation start site were extracted as promoter regions using the Gtf/Gff3 Sequences Extract program of TBtools software to further understand the potential functions of the cis-regulatory elements in NtSOD genes. The cis-regulatory elements in the promoter sequences were analyzed with the PlantCare online tool 10 , and then the results were visualized with the Simple BioSequence Viewer of TBtools software .
Prediction of Transcription Factors and miRNAs Involved in Regulating NtSOD Expression
As a method to better understand the transcription factors and miRNAs involved in regulating NtSOD expression, the promoter and mRNA sequences were used to predict possible transcription factor binding sites and target miRNAs, respectively. In detail, 2000-bp promoter sequences of NtSOD genes were submitted to the PlantRegMap website 11 to predict the transcription factors involved in regulating NtSOD expression with the Regulation Prediction tool at a p value ≤ 1e −6 (Tian et al., 2020). For the miRNA target gene analysis, NtSOD mRNA sequences were submitted to the psRNATarget online server 12 , and then the network map was generated using Cytoscape software (Dai et al., 2018).
Plant Materials and Heavy Metal Treatments
Seeds of tobacco cultivar TN90 were sown in uncontaminated nutrient soil and generated plantlets in a greenhouse at 25 ± 2 • C under a 16/8 h (light/dark) photoperiod. Two-week-old tobacco seedlings were transplanted to hydroponic tanks and cultured with half-strength Hoagland's solution for 7 days. Heavy metal stress was induced by replacing the solution in hydroponic tanks with fresh Hoagland's solution containing 200 µM CuSO 4 , 200 µM ZnSO 4 , 200 µM MnSO 4 , 200 µM Fe-EDTA, or 50 µM CdCl 2 . For ion-deficiency treatments, the solutions were replaced with fresh Hoagland's solution lacking Cu, Zn, Mn and Fe . The tobacco seedlings were cultured with the abovementioned treatments for 7 days. Subsequently, the leaves and roots of the seedlings were separately harvested. For the analysis of the tissue expression profile, five tissue samples, including roots, stems, old leaves, young leaves and flowers, were collected from four-month-old tobacco plants grown in the natural environment. All samples were frozen immediately in liquid nitrogen and stored at −80 • C until total RNA was isolated. Each experimental group consisted of three biological replicates and technical duplicates.
Determination of the Heavy Metal Content in Tobacco Plants
The metal ion concentrations in the roots and shoots of tobacco plants treated with various heavy metals were measured in this study. The roots of tobacco plants were soaked in a 20 mM EDTA solution for half an hour and then washed three times with distilled water to remove heavy metal ions and precipitates adsorbed on the root surface with sterile absorbent paper for laboratory. The plant materials, including roots and shoots, were dried in an oven at 80 • C for 3 days until a constant weight was obtained. Subsequently, the dried roots and shoots were separately ground into powders, and 20 mg of powdered sample was digested with 7 mL of concentrated HNO 3 for 2 h at 115 • C. The contents of Cd, Cu, Zn, Fe and Mn were determined using flame atomic absorption spectrometry (TAS-986, China) at wavelengths of 228.8 nm, 324.8 nm, 213.9 nm, 248.3 nm and 279.5 nm, respectively.
Analysis of the Pattern of NtSOD Gene Expression
Total RNA was extracted from the abovementioned plant materials using an RNAsimple Total RNA kit (DP419, Tiangen Biotech, Beijing, China) according to the manual, and then the quality and concentration of RNA were determined using a BioPhotometer Plus instrument (Eppendorf, Germany). Subsequently, 5 µg of total RNA were used to synthesize first-strand cDNAs via reverse transcription using the GoScriptTM Reverse Transcriptase Kit (Promega, Madison, WI, United States) according to the manufacturer's protocol. All cDNA samples were diluted 25 times with RNase-free water and stored at −20 • C until qPCR was performed.
In the present study, NtSOD gene expression was analyzed based on transcriptome data and qPCR results. The raw transcriptome data (PRJNA208209) of tobacco cultivar TN90 were downloaded from NCBI-BioProject, including SRA data from roots, stems, young leaves, mature leaves, senescent leaves, young flowers, mature flowers, and senescent flowers. The relative abundance of each gene transcript was calculated as transcripts per kilobase million (TPM) values using the Salmon program and visualized in a heatmap using the Heatmap tool of TBtools. qPCR was performed with a CFX96TM real-time fluorescence quantification platform (Bio-Rad, United States) using SYBR Green enzyme (Novoprotein, China) with the following procedure to further verify the results of transcriptome data analysis: 95 • C for 1 min, followed by 45 cycles of 95 • C for 15 s and 60 • C for 30 s. The qPCR primers were designed using Primer Premier 6.0 and are listed in Supplementary Table 1. The relative mRNA expression levels of NtSOD genes were normalized to NtEF1α (accession number: AF120093) (Liu et al., 2022), and relative fold changes were calculated using the 2 − CT method (Livak and Schmittgen, 2001).
Heterologous Expression and Functional Verification of NtSOD Family Genes in Yeast
The CDSs of three NtSOD family members, NtCSD1a, NtFSD1e and NtMSD1b, were cloned by PCR with the primers listed in Supplementary Table 1 to verify the potential function of NtSOD proteins in scavenging ROS produced in response to heavy metal stress. The PCR fragment and pYES2 vector were digested using the restriction enzymes BamHI and EcoRI (TaKaRa, Dalian, China) for 3 h at 37 • C. Subsequently, the CDS of the three genes were inserted into the expression vector pYES2 using the DNA Ligation Kit Ver. 2.1 (TaKaRa, Dalian, China) and transformed into E. coli strain DH5α. The inserted sequences in the resulting recombinant plasmids, named pYES2:NtCSD1a, pYES2:NtFSD1e and pYES2:NtMSD1b, were verified by DNA sequencing (BGI, Shenzhen, China). The recombinant plasmids were amplified in DH5α cells and extracted according to the operation manual of a high purity plasmid extraction kit (Biomed, Beijing, China). The oxidation-sensitive yeast mutant strain yap1 (MATα ura3lys2 ade2 trp1 leu2 yap1:leu2) was transformed with the three recombinant plasmids using the lithium acetate transformation protocol, and the empty plasmid pYES2 was used as the control (Kawai et al., 2010;Rodrigues-Pousada et al., 2019). Yeast cells were cultured in SD/-URA liquid medium until reaching the logarithmic phase (OD 600 = 0.6) to assess the sensitivity of the cells to heavy metal-induced oxidative stress. The cultures were successively diluted 10 times from 1 to 10 −3 and spotted onto SD plates containing 50 µM CdCl 2 , 2.5 mM CuSO 4 , 5 mM ZnSO 4 , 2.5 mM FeSO 4 , or 2.5 mM MnSO 4 and grown for 3 days before being photographed.
Statistical Analysis
Data are presented as the means ± SD derived from at least three biological replicates, unless indicated otherwise. All data were analyzed using one-way ANOVA, and significant differences were analyzed using Dunnett's multiple range test (P < 0.05) with GraphPad Prism 8.0 (GraphPad Software, San Diego, CA, United States).
Identification and Phylogenetic Analysis of NtSOD Genes
According to the results of the HMM search and BlastP, 17 candidate genes were originally obtained from the Pfam SOD family in tobacco. Based on the domain analysis, 5 proteins were shown to have a Cu/Zn-SOD domain (PF00080), 7 proteins contained an Fe/Mn-SOD alpha-hairpin domain (PF00081), and 3 proteins contained an Fe/Mn-SOD C-terminal domain (PF02777). Additionally, 2 proteins contained both a Cu/Zn-SOD domain and an N-terminal heavy metal-associated domain (HMA; PF000403) ( Figure 1B), which were considered copper chaperones for Cu/Zn superoxide dismutase in tobacco (NtCCS) due to their clustering with AtCCS in the phylogenetic tree ( Figure 1A). CCS is essential for transporting Cu to SOD but has no dismutase activity (Cohu et al., 2009). After excluding 2 NtCCS members from 17 candidate proteins, we identified 15 NtSOD proteins in N. tabacum cultivar TN90.
A neighbor-joining phylogenetic rootless tree was constructed using SOD protein sequences from A. thaliana and tobacco to further clarify the evolutionary relationship of SOD genes in tobacco. All NtSOD genes were assigned specific names according to their phylogenetic relationships with AtSOD genes. The results of phylogenetic analysis showed that SOD family members in tobacco were classified into three subfamilies with high bootstrap values, including the Cu/Zn-SOD subfamily (NtCSD1a/1b, NtCSD2a/2b, and NtCSD3), Fe-SOD subfamily (NtFSD1a/1b/1c/1d/1e and NtFSD3a/3b) and Mn-SOD subfamily (NtMSD1a/1b/1c). The NtFSD subfamily exhibited a closer phylogenetic relationship with NtMSD than with NtFSD, potentially because NtFSD and NtMSD have the same conserved domains SOD_Fe_N and SOD_Fe_C ( Figure 1B).
The protein sequences and physicochemical properties of NtSOD family members were characterized in the present study. The NtSOD gene CDSs ranged from 459 bp to 912 bp and encoded NtSOD proteins with lengths ranging from 153-304 amino acids. The molecular weight of the NtSOD proteins ranged from 15.2 kDa to 34.7 kDa, and the theoretical isoelectric points (pIs) ranged from 5.08 (NtCSD1b) to 8.82 (NtMSD1c). Additionally, the results for the predicted subcellular localization showed that SOD proteins in the same subfamily may be located in different organelles. Except for NtFSD1a that is located in the cytoplasm, other members in the NtFSD subfamily are located in chloroplasts. NtCSD2a/2b is located in chloroplasts, while NtCSD1a/1b/3 is located in the extracellular space. NtMSD1c is the only tobacco SOD protein located in mitochondria ( Table 1).
Multiple Sequence Alignment of NtSOD Proteins
Multiple sequence alignment was conducted to analyze the differences among the fifteen NtSOD proteins, and the results are shown in Figure 2. Among the five members of the NtCSD subfamily, NtCSD2a and NtCSD2b had the highest sequence consistency of 95%. NtCSD1b showed low consistency with the other four subfamily members, ranging from 29.2% to 64.8%. The N-terminal and C-terminal regions of NtCSD2a/2b contained chloroplast transporter peptide sequences and transmembrane alpha-helical structures, respectively, suggesting that NtCSD2a/2b was localized in the chloroplast membrane. These results were consistent with the predicted subcellular localization of NtCSD2a/2b. Eight metal binding sites were identified in the protein sequence of the NtCSD gene subfamily, of which three sites bind Cu, four sites bind Zn, and H63 binds both Cu and Zn (Figure 2A). In the NtFSD subfamily, the sequence consistency between NtFSD1b and NtFSD1c, NtFSD1d and NtFSD1e, and NtFSD3a and NtFSD3b was greater than 97%, suggesting that they may be the products of gene doubling events. All NtFSD proteins (except NtFSD1a) have a chloroplast transit peptide and transmembrane alpha-helix, suggesting that these enzymes function on the chloroplast membrane. Although NtFSD1a has a transmembrane structure, the absence of its chloroplast transporter peptide leads to its localization in the cytoplasm, consistent with the predicted subcellular localization. In addition, the four Fe ion binding sites were highly conserved in all members of the NtFSD gene subfamily ( Figure 2B). The three members of the NtMSD subfamily exhibit high sequence consistency, among which NtMSD1b and NtMSD1c have only 5 amino acid differences, and the similarity is 97.8%. The mitochondrial transit peptide and transmembrane alpha-helix are closely linked at the N-terminus of NtMSD, consistent with the prediction that NtMSD is located in mitochondria by the subcellular localization analysis. The four Mn binding sites are extremely conserved in NtMSD subfamily members ( Figure 2C).
Conserved Motifs and Gene Structure Analysis of NtSOD Genes
We further analyzed the conserved motifs of the NtSOD family genes in tobacco, and 7 motifs were identified using MEME software with the default parameters ( Figure 3B). As expected, most of the closely related members in the same subfamily had common motif compositions; however, no common motifs were shared in all fifteen NtSOD genes. Among them, motifs 4 and 6 were associated with the Cu/Zn-SOD domain (PF00080) and were only identified in Cu/Zn-SOD subfamily members. Motif 1, motif 7 and motif 5 were specific to NtFSD and NtMSD proteins, respectively. Motif 2 and motif 3 correspond to the Fe/Mn-SOD alpha-hairpin domain (PF00081) and Fe/Mn-SOD C-terminal domain (PF02777), respectively. Except for NtFSD1a, motif 3 was the common conserved motif in the NtFSD subfamily. Motif2 was shared by the NtFSD subfamily and NtMSD subfamily. The number of introns and gene structure are usually related to the evolution of gene family members . In the present study, cDNA and corresponding DNA sequences of NtSOD family genes were compared to analyze the gene structure ( Figure 3C). NtFSD1b was the smallest gene at 1,124 bp, while NtFSD3b was the longest gene at 9,355 bp. All NtSOD genes contain introns, and the number of introns ranges from 2 to 8. Among them, NtFSD1a has the least number of introns at 2, while NtFSD1c/1d/1e has the most at 8 introns. Notably, the three NtFSD1c/1d/1e genes showed highly similar genetic structures, and their cDNA and DNA sequences were highly consistent, suggesting that they may be the result of gene replication. The same replication event occurs in NtCSD1a/1b and NtCSD2a/2b.
Cis-Acting Elements in the Promoter Regions of NtSOD Genes
Based on accumulating evidence, SOD genes play important roles in the plant response to abiotic stress. The cis-acting elements in the promoter regions of the NtSOD family genes, except NtFSD1e due to incomplete assembly of the tobacco genome sequence, were scanned using plantCARE to better understand the potential regulatory mechanism of NtSOD genes in abiotic stress or hormonal responses of tobacco. Nineteen elements involved in defense and stress responsiveness, phytohormone responsiveness and light responsiveness were detected in the NtSOD promoters and divided into three categories, as shown in Figure 4. These elements were irregularly dispersed in the promoter regions of NtSOD family genes (Supplementary Figure 3). No similar distribution pattern was observed between two genes, even those that are evolutionarily close. Many of the hormone-related cis-acting elements, including ABRE (abscisic acid), AuxRR core/TGA element (auxin), GARE motif/P Box/TATC Box (gibberellin), CGTCA motif/TGACG motif (MeJA), and TCA element (salicylic acid), were identified in the promoter region of NtSOD genes. ABRE and CGTCA/TGACG motifs were widely distributed in NtFSD and NtCSD promoter sequences, suggesting that ABA and MeJA are involved in the regulation of NtFSD and NtCSD. NtMSD was predicted to be regulated by auxin and salicylic acid based on the presence of TGA elements and TCA elements in their promoter regions, respectively. Elements associated with auxin and gibberellin responses are scattered in the promoters of NtSOD family genes (Figure 4). These results suggested that plant hormones might exert a modulatory effect on the regulation of NtSOD gene expression. MYC, one of the key motifs responding to thrilling, was identified in all NtSOD family gene promoter regions. The anaerobic induction regulationrelated element ARE was mainly present in the promoters of NtFSD subfamily and NtMSD subfamily genes, especially in the promoter region of NtMSD1b, which contained 16 ARE elements. In addition, a large number of light-responsive elements were present in almost all promoter regions of NtSOD genes, among which Box4 and G-box were significantly enriched.
Analysis of Potential Regulatory Interactions Between Transcription Factors and NtSODs
The online software PlantRegMap was employed to predict the potential regulatory interactions between transcription factors (TFs) and NtSOD family genes. Sixteen TFs that may be involved in regulating NtSOD family gene expression were identified. The number of potential binding motifs for TFs The similarity between amino acid sequences was recorded in the table, and the chloroplast transit peptide, mitochondrial transit peptide, and membrane transit peptide are marked using green, purple and yellow boxes, respectively, and red boxes mark the metal-binding sites of these SOD proteins.
in NtSOD family gene promoters was counted to conduct a clustering analysis and construct a heatmap ( Figure 5A). The regulatory patterns of NtMSD1a, NtCSD1a, NtCSD1b and NtFSD3a were similar, all of which were regulated by Dof, AP2 and MIKC_MADS. Additionally, three NtSOD genes, NtFSD1c, NtFSD1e and NtMSD1c, constitute another group with similar regulatory patterns, which were all regulated by MYB, ERF and LBD. Notably, the NtCSD3 promoter only contains one C2H2 transcription factor binding site, suggesting that NtCSD3 may not be regulated by transcription factors other than C2H2.
Analysis of miRNAs Targeting NtSOD Genes
MicroRNAs (miRNAs) are a class of single-stranded small (20-24 nucleotides) non-coding RNAs that regulate gene expression by binding to target gene transcripts to inhibit their translation or promote mRNA degradation. psRNATarget online software was used to analyze miRNA binding sites in 15 NtSOD genes and to provide insights into the miRNA-mediated regulation of the NtSOD genes. Nine NtSOD genes were predicted to be targeted by 13 known miRNA families, of which 5 target genes were cleaved by 5 conserved miRNA families, including miR156, miR398, miR482, miR172 and miR164, and 7 target genes were cleaved by unique miRNAs in Solanaceae plants, including miR6025, miR6020, miR6144, miR6150, miR6024, miR6155 and miR6021 ( Figure 5B). As expected, miR398 was predicted to target NtCSD1a/1b. The miR164 family and miR6021 may be involved in regulating the expression of NtMSD1b, which was the only gene regulated by miRNAs in the NtMSD subfamily of genes. Based on accumulating evidence, miR156-SPL and miR164-NAC modules regulate plant abiotic stress tolerance by maintaining ROS homeostasis (Fang et al., 2014;Yin et al., 2019). Interestingly, miR156 and miR164 were associated with the expression of NtCSD1b and NtMSD1b in tobacco, respectively. Therefore, the regulation of NtSOD family genes in tobacco involves a complex regulatory network of transcription factors and miRNAs.
Tissue-Specific Expression Profiles of NtSOD Family Genes
qPCR was used to analyze the relative tissue-specific expression levels of NtSOD family genes in different organs (stem, taproot, lateral root, terminal bud, 1st leaf, 2nd leaf, 3rd leaf and 4th leaf) at the vegetative stage of tobacco grown under normal growth conditions. Taken together, the expression levels of NtSOD family genes in taproots and lateral roots were lower than those in shoots of tobacco plants. Similar expression patterns were observed among NtSOD genes in the same subfamily, which were hierarchically clustered in the heatmap (Figure 6). Additionally, the qPCR results were generally consistent with the results from the transcriptome data (Supplementary Figure 2). NtCSD subfamily genes, except NtCSD3, were expressed at relatively high levels in terminal buds and the 1st leaves but expressed at low levels in mature leaves. NtFSD subfamily genes shared a similar expression pattern and exhibited higher expression levels in leaves than in roots, while NtFSD1d/1e was expressed at high levels in stems. A lower expression level of NtMSD subfamily genes was detected in lateral roots, and no significant difference was observed in the other tissues.
Heavy Metal Stress-Induced Expression Profiles of NtSOD Family Genes
Two-week-old tobacco seedlings were divided into two treatment groups, including a heavy metal toxicity group (hydroponic FIGURE 6 | The relative expression levels of NtSOD genes in different tobacco tissues. The expression data were obtained from the real-time RT-PCR (RT-qPCR) analysis data and are shown as log2 values calculated as averages. High expression levels are shown in red, and lower expression levels are shown in white. Statistically significant differences are indicated using asterisks (Dunnett's test, *p < 0.05, **p < 0.01, and ***p < 0.001). Data are presented as the means ± SD of three replicates. solution containing Cd or excess Cu, Zn, Fe or Mn) and an ion-deficiency treatment group (Hoagland's solution without Cu, Zn, Fe or Mn), to understand the effect of heavy metal stress on the expression pattern of NtSOD family genes. After 7 days of treatment, chlorosis and growth inhibition were observed as significant toxicity symptoms for tobacco seedlings treated with Cd and Cu. Additionally, the growth of roots of tobacco seedlings under Cu stress was substantially inhibited (Figure 7A). In the present study, the ion concentrations in all the treated tobacco seedlings were determined using flame atomic absorption spectrometry. The results were consistent with the expectation. In the ion-deficient stress groups, the ion contents in the roots and shoots of tobacco seedlings were significantly lower than those in the control group. In contrast, the ion content was higher in the heavy metal toxicity group than that in the control group ( Figure 7C). Notably, the copper ions mainly accumulated in the roots of tobacco after treatment with 200 µM Cu, while the concentration of copper ions in the shoots of Cu-treated tobacco was lower than that in the control group. This phenomenon may be due to Cu stress inhibiting the growth and development of roots in tobacco, which blocked ion transport from roots to shoots. Thus, heavy metal stress disrupted ion homeostasis in tobacco plants in the present study.
Antioxidant enzymes, especially SOD, play important roles in scavenging ROS generated in response to heavy metal stress in plants. qPCR was used to analyze the relative expression levels The numbers in the box represent the average ratio of gene expression levels in the treatment and control groups. The asterisks on the number indicate significant differences (Dunnett's test, *p < 0.05, **p < 0.01, and ***p < 0.001). The population size was n = 3. (C) Content of metals contained in tobacco tissues subjected to different heavy metal treatments. The red box indicates that the corresponding metal content is higher than that of the control, while the blue box indicates that the corresponding metal content is lower than that of the control.
of NtSOD family genes in tobacco seedlings with heavy metal toxicity and ion deficiency to clarify the expression patterns of NtSOD family genes in tobacco in response to oxidative stress induced by different heavy metals. Cd exposure significantly inhibited the expression of NtCSD subfamily genes and induced the upregulation of NtMSD subfamily genes ( Figure 7B). The expression levels of NtFSD1a/1b/1c were significantly altered in the tobacco seedlings under Cu-related stress compared with the control group. However, the expression levels of NtFSD1d/1e were not affected by Cu toxicity or Cu deficiency. NtCSD3 expression was downregulated in shoots but not in roots exposed to zinc toxicity. Except for NtCSD2a and NtMSD1a, iron toxicity had no effect on the expression of other NtSOD family genes. On the other hand, iron-deficient stress resulted in the downregulation of NtCSD3 and NtFSD1a/1b/1c expression in shoots and significantly upregulated NtMSD1a expression in roots. Excessive accumulation of Mn in roots upregulated the expression of NtSOD family genes, including NtCSD2a, NtMSD1b/1c, NtFSD1a/1b/1c, NtCSD2b and NtMSD1a. However, Mn-deficient stress only induced the upregulation of NtFSD3a/3b expression in shoots. In summary, NtSOD family genes displayed a wide variety of expression patterns in tobacco seedlings in response to heavy metal toxicity and iondeficient stresses.
Heterologous Expression of NtSOD Genes in Yeast Mutant yap1
The oxidative stress-hypersensitive yeast mutant yap1 was transformed with NtCSD1a, NtFSD1e and NtMSD1b and compared with the control to investigate the ability of NtSOD family genes to resist oxidative stress induced by heavy metals. Four serial dilutions (10 ×) of yeast cells were dropped on solid SD/-URA medium containing different heavy metals, including Cd, Zn and Mn, and cultured for 3 days at 30 • C. The dilution dot assay showed no significant differences in cell growth under normal culture conditions following the heterologous expression of three NtSOD genes in yeast and the control group. The Cd and Zn tolerance in NtMSD1b-overexpressing yeast was stronger than that in the control group. All yeast cells expressing NtSOD genes exhibited better growth than the control under Cu stress and Fe stress. Additionally, the heterologous expression of NtMSD1b or NtFSD1e, but not NtCSD1a, conferred tolerance to high concentrations of Mn (Figure 8). Based on these results, FIGURE 8 | Effect of NtSOD gene expression in oxidation-sensitive yeast yap1 on yeast cell growth in response to different heavy metal stresses (50 µM Cd, 2.5 mM Cu, 5 mM Zn, 2.5 mM Fe, and 2.5 mM Mn). The numbers above the data represent the dilution times of the bacterial solution.
the overexpression of NtSOD genes in the mutant yap1 yeast effectively enhanced the resistance of yeast to heavy metal stress.
DISCUSSION
Tobacco is the most widely cultivated non-food cash crop in the world. ROS accumulation caused by stresses such as drought, high temperature and heavy metals exerts adverse effects on tobacco growth and development (Rizhsky et al., 2002). SOD is the first line of defense against oxidative stress and removes ROS that accumulate in plant cells under stress (Alscher et al., 2002). Previous studies have shown that SOD overexpression confers abiotic stress resistance to tobacco (Pitcher et al., 1991;Zhang et al., 2017;Pan et al., 2019). To date, studies on SOD family genes in tobacco have been limited. A complete characterization of the characteristics and functions of the SOD gene family in tobacco is helpful to enrich our understanding of the SOD gene family in plants. The systematic identification of the NtSOD genes as the best candidate genes for screening and enhancing tobacco heavy metal toxicity tolerance is very important to cope with the increasingly serious soil heavy metal pollution.
In this study, 15 NtSOD genes were identified from the tobacco genome and classified into three subfamilies according to their domains and motifs, including 5 NtCSDs, 7 NtFSDs and 3 NtMSDs (Table 1). Although NtCCS has a conserved SOD_Cu domain (Figure 1), its HMA domain and evolutionary relationship closely with A. thaliana suggested that NtCCS may be a copper chaperone for NtCSD. The N-terminus of the NtCCS protein contains a conserved metal-binding motif MxCxxC, consistent with AtCCS and MaCCS sequences (Chu et al., 2005;Leitch et al., 2009). In previous studies, the CCS gene has been classified as a member of the SOD family (Feng et al., 2016). Since NtCCS lacks the common motif of the NtCSD subfamily, NtCCS does not belong to the NtSOD gene family, and the NtCCS subfamily was not included in the subsequent gene family analysis. Additionally, the NtFSD2 subfamily is lacking in tobacco (Figure 1). Just like the NtMT family genes , The NtFSD2 subfamily gene loss events possibly occurred during the formation of tobacco tetraploid genome.
The phylogenetic analysis showed that NtFSDs and NtMSDs were clustered into one group with a high bootstrap value ( Figure 3A). In addition, an analysis of conserved domains showed that NtFSD and NtMSD shared Motif 3 and Motif 7, while the conserved domains in the NtCSD subfamily were Motif 4 and Motif 6 ( Figure 3B). These results suggest different origins of NtCSD and NtFSD/NtMSD. The emergence of SOD family enzymes is an important event of biological evolution on earth. Before the Great Oxidation Event, Fe and Mn were relatively abundant and highly available in the early reductive ocean, while Zn and Cu were bound to unavailable sulfur-bearing minerals in the crust (Saito et al., 2003). Thus, FSD and MSD were generally considered more ancient than CSD and evolved from a common ancestor . CSD, however, evolved separately in bryophytes. These two groups evolved independently. In addition, according to the distribution characteristics of motifs, we speculated that NtFSD1a might be an intermediate transition type between NtFSDs and NtMSDs ( Figure 3B).
The NtCSDs were mainly localized in the cytoplasm (NtCSD1a/1b) and chloroplasts (NtCSD2a/2b) ( Table 1), and exhibit low amino acid sequence identity (Figure 2). The gene structure analysis revealed significant differences in the number and location of introns between NtCSD1a/1b and NtCSD2a/2b (Figure 3), which is a feature that distinguishes cytoplasmic and chloroplast CSDs. The cytoplasmic CSD in Marchantia paleacea shares high homology with the chloroplast CSD, and MaCSD is speculated to be the closest relative to the original ancestral plant CSD (Fink and Scandalios, 2002). Unlike other NtFSD subfamily members, NtFSD1a is predicted to be expressed in the cytoplasm. Actually, FSDs were also expressed in the cytoplasm of cowpea and L. japonicus (Moran et al., 2003;Rubio et al., 2007). Plant MSDs are generally located in the mitochondria of plants, where NtMSD1c is predicted to be localized in the mitochondria. In addition, MSDs are expressed in peroxisomes (del Río et al., 2003), which explains the prediction that NtMSD1a/1b is expressed in organelle membranes ( Table 1).
The analysis of cis-acting elements in promoters provides important information for studying the regulation of SOD expression. Due to the lack of NtFSD1b promoter information in the tobacco genome database, we extracted 14 NtSOD gene promoter sequences and performed a cis-acting element analysis. A large number of cis-acting elements related to plant hormone response existed in the promoter region of NtSOD family genes (Figure 4), which was consistent with the fact that SOD was regulated by ABA and MeJA in plants (Agarwal et al., 2005;Cao et al., 2009;Lu et al., 2009;Jiang et al., 2015). Plant SODs play an important role in defense and stress responses. Drought, low temperature, heavy metal toxicity, and biotic stresses usually lead to an imbalance in ROS homeostasis, which induces the expression of plant SOD genes. We identified multiple cisacting elements associated with defense and stress resistance in the NtSOD gene promoter region (Figure 4). MYC and LRT, two cis-acting elements associated with freezing injury and low-temperature stress, were distributed in each NtSOD gene promoter region, a prediction that suggests an important role for the tobacco NtSOD family in responding to lowtemperature stress. Overexpression of the NtMSD gene in alfalfa significantly enhanced the cold tolerance and next-year yield of transgenic plants (McKersie et al., 1993). ARE, a key cis-acting element involved in anaerobic induction, is widely distributed in the promoter region of the NtSOD gene, which might explain why SOD activity is increased in plants under hypoxic conditions (Chen and Qualls, 2003). NtSOD gene transcripts were significantly increased under light conditions, and conversely, NtSOD gene expression was significantly downregulated in the absence of light (Tsang et al., 1991). This phenomenon is consistent with our finding of a large number of cis-acting elements associated with light response in the tobacco SOD gene promoter. The analysis of cis-acting elements in the promoter indicated that NtSOD genes play an important role in the tobacco response to stress resistance.
The expression of SOD genes in response to stress is regulated by transcription factors and miRNAs in plants. SOD gene expression levels were significantly upregulated in transgenic peanut overexpressing MuWRKY3, which subsequently effectively enhanced drought stress tolerance in transgenic plants (Kiranmai et al., 2018). The transcription factor SPL7 is the main transcription factor that regulates CSD genes and negatively regulates the expression of CSD genes to adapt to Cu deficiency stress (Mermod et al., 2019). Activation of rat SOD gene expression by ginsenoside Rb2 from a ginseng extract is regulated by the AP2 transcription factor (Kim et al., 1996). Expression of the transcription factor NAC derived from maize (Glycine max) in tobacco resulted in transgenic plants that were more sensitive to drought, and the molecular mechanism was that NAC functioned as a negative regulator, reducing the expression of plant SOD genes and leading to ROS accumulation (Jin et al., 2013). Multiple transcription factor binding sites were predicted in the promoter region of the NtSOD genes, including Dof, AP2, NAC, MADS, and MYB ( Figure 5). This result implies that NtSOD genes have different regulatory modalities to cope with the complex external environment. In recent years, an increasing number of studies have shown that miRNAs play a key role in the posttranscriptional regulation of plant SOD genes. miRNA398 is the first miRNA identified to negatively regulate CSD expression and plays an important role in plant resistance to stress. In A. thaliana, oxidative stress does not directly regulate the expression of CSD genes but represses miRNA398 expression, thereby protecting CSD mRNAs from cleavage (Zhou et al., 2019).
Plant SOD genes show different expression patterns at different growth stages or in different tissues. Some SOD genes exhibit constitutive expression in all tissues (Lin and Lai, 2013), while other SOD genes show tissue-specific expression patterns (Feng et al., 2015). In the present study, we detected the expression patterns of NtSOD genes in different tissues and at different stages using qPCR and nine-pair primers. The results of the qPCR analysis (Figure 6) were consistent with the results obtained from transcriptomic data (Supplementary Figure 2). Furthermore, NtSOD gene expression profiles showed that genes in the same subfamily have similar expression patterns, consistent with the tissue expression profile of cotton SOD family genes . Differences in the expression levels of NtSOD family members were observed in different tissues and leaves at different developmental stages. For example, the higher expression level of NtFSD1a/b/c in the lateral root, compared to other NtSOD genes, implies that it is associated with the development of the lateral root. The fsd1 knockout mutant in Arabidopsis significantly suppressed the development of the lateral roots (Dvořák et al., 2021). NtSOD was expressed at a relatively high level in the terminal bud of tobacco, consistent with the continuous increase in SOD activity observed in buds of Ficus carica during the dormancy to germination period (Sedaghat et al., 2022). Briefly, the expression of tobacco NtSOD family genes show spatiotemporal specificity and play an important role in different tissues and organs and developmental stages.
The regulation of plant SOD gene expression by heavy metals is mainly reflected in two aspects. First, the toxic effects of heavy metal stress on cells induce the accumulation of ROS and a subsequent imbalance of cellular redox homeostasis; ROS removal usually depends on the expression of SOD. Second, metals such as Fe, Cu, Mn and Zn are essential to ensure SOD activity, and their deficiency also induces SOD expression. SOD expression levels in plants are usually upregulated by Cd, which is a non-essential and poisonous heavy metal (Pan et al., 2019). In contrast to our expectations, we found that Cd stress caused a significant decrease in the expression levels of NtCSD subfamily genes, consistent with the observation that Cd treatment exerted a significant inhibitory effect on the CSD mRNA and protein levels and activity in pea (Romero-Puertas et al., 2007). However, Cd stress significantly increased the expression levels of NtFSD1a/b/c and NtMSD1a/b/c (Figure 7), indicating that NtSOD family genes employ a differential division of labor in resisting Cd stress. Compared to toxicity induced by other essential metals, Mn toxicity significantly altered the expression level of NtSOD genes, especially in tobacco roots (Figure 7). A significant increase in SOD activity was detected in soybean roots treated with excess Mn (Santos et al., 2017). The SPL7-miR398 module is an important regulator of Cu homeostasis in Arabidopsis. Cu deficiency induces miR398 expression and results in degradation of the CSD mRNA (Yamasaki et al., 2009). The expression level of the NtCSD subfamily in tobacco was significantly downregulated under Cu-deficient conditions, suggesting that SPL7-mir398 is involved in a conserved regulatory module in plants to address disrupted Cu homeostasis. In addition, Cu deficiency induced the expression of NtFSD1a/b/c to complement the functional deficiency of NtCSD. Fe deficiency downregulated NtFSD expression but induced the expression of the NtMSD genes, consistent with the upregulation of MSD expression in Chlamydomonas reinhardtii under Fe deficiency (Allen et al., 2007). Consequently, NtSOD expression is regulated by a complex regulatory network to counteract the imbalance of heavy metal homeostasis in tobacco.
A yeast (Saccharomyces cerevisiae) mutant strain was used in this study to analyze the biological functions of NtSOD in resisting oxidative stress induced by heavy metal exposure. Yeast is generally acknowledged as a well-understood eukaryote in the field of stress response. The yeast activator protein (YAP) protein family is the largest bZIP subfamily in S. cerevisiae and consists of 8 members (Yap1 to Yap8) (Rodrigues-Pousada et al., 2019). In yeast cells, functional deficiency of Yap1 caused cellular sensitivity to ROS generated by the disruption of metal homeostasis (Schnell et al., 1992). Therefore, utilization of the yeast mutant strain delta yap represents a convenient method to test the ability of heterologously expressed SOD to remove ROS in response to heavy metal stress. In the present study, three tobacco SOD genes, NtCSD1a, NtFSD1e and NtMSD1b, were introduced into the yeast mutant strain yap1. The results of the dilution dot assay showed that all three tobacco SOD genes alleviated oxidative stress induced by heavy metals (Figure 8). Compared to NtMSD1b and NtFSD1e, NtCSD1a may have a slightly weaker ability to remove ROS induced by heavy metals. Thus, NtSOD genes show differences in their abilities to remove ROS generated in response to heavy metals.
CONCLUSION
In the present study, 15 SOD genes were identified in tobacco variety TN90. The NtSOD gene family is divided into three subfamilies: NtCSD, NtFSD and NtMSD. The specific characteristics of the SOD genes were investigated, including the subcellular localization, conserved domain and gene structure. Cis-regulatory elements of the NtSOD gene promoter region in response to plant hormones, abiotic stress and light were also identified. Meanwhile, the transcription factors and miRNAs that may be involved in regulating NtSOD gene expression were predicted. The analysis of the tissue expression profile showed that the NtSOD gene family played an important role in tobacco during the growth and development stages. The expression pattern of the NtSOD gene family in response to heavy metals indicated that the NtSOD genes were the key proteins mediating tobacco resistance to heavy metal toxicity, but their functions were different. In addition, the biological function of the NtSOD gene family in protecting against oxidative stress induced by heavy metals was verified in the yeast mutant strain. These results provide a good understanding of the biological characteristics and functions of the NtSOD gene family in tobacco and provide important information for the analysis of tobacco resistance to heavy metal-induced oxidative stress.
DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.
AUTHOR CONTRIBUTIONS
CH: data curation, methodology, and writing-original draft. CH and LH: investigation. LH: software. TY and SZ: methodology. RL and XJ: validation. WL, RL, and HX: funding acquisition. WL and FZ: writing-review. HX: resources. WL: conceptualization, editing, supervision, and project administration. All authors contributed to the article and approved the submitted version.
ACKNOWLEDGMENTS
We would like to thank Prof. Xiaoling Liao and Dr. Shuang Yang for technical support. | 2022-05-06T13:19:00.413Z | 2022-05-06T00:00:00.000 | {
"year": 2022,
"sha1": "c8ad67726cd46b7378f06090006a0d2e25828a63",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Frontier",
"pdf_hash": "c8ad67726cd46b7378f06090006a0d2e25828a63",
"s2fieldsofstudy": [
"Environmental Science",
"Biology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
8448887 | pes2o/s2orc | v3-fos-license | How friends and non-determinism affect opinion dynamics
The Hegselmann-Krause system (HK system for short) is one of the most popular models for the dynamics of opinion formation in multiagent systems. Agents are modeled as points in opinion space, and at every time step, each agent moves to the mass center of all the agents within unit distance. The rate of convergence of HK systems has been the subject of several recent works. In this work, we investigate two natural variations of the HK system and their effect on the dynamics. In the first variation, we only allow pairs of agents who are friends in an underlying social network to communicate with each other. In the second variation, agents may not move exactly to the mass center but somewhere close to it. The dynamics of both variants are qualitatively very different from that of the classical HK system. Nevertheless, we prove that both these systems converge in polynomial number of non-trivial steps, regardless of the social network in the first variant and noise patterns in the second variant.
Introduction
The dynamics of opinions in society is a very intricate and intriguing process. Especially in today's world, with the pervasive infiltration of social networks like Facebook and Twitter that allow quick broadcasts of opinions, phenomena which were once wild speculations by philosophers, such as the viral spread of memes [Daw76], are now easily observed and quantified. This acceleration of social processes is turning sociology into a quantitative science, where concrete models for social phenomena can be proposed and rigorously tested.
Sociologists have long identified several different processes that determine opinion dynamics, most notably, normative and informational [DG55]. Normative influence refers to the influence that causes people to conform to a group's social norms. On the other hand, informational influence refers to the way people acquire the opinions of others, driven by the assumption that neighbors possess information about a situation. Informational influence is especially relevant in the context of understanding how opinions change, and it arguably is the dominant process for determining trends in, say, fashion, mobile phones and music.
In this work, we focus on quantitative models for informational influence. Such a model should specify how an individual agent updates its opinion using information learned from its "neighbors". By now, this area has been heavily studied; see [Jac08] for a survey. The most basic such model is the classic DeGroot model [DeG74,Fre56,Har59] where each agent's opinion is a real number between 0 and 1, and at every time step, each agent moves to some weighted average of its neighbors' positions, where the neighbors are determined according to an unchanging undirected graph. Such a system always reaches consensus, contrary to the existence of polarized states in society. Another classic model is the voter model [CS73,HL75]. Here, there is an unchanging directed graph among the agents, and at every time step, a random agent selects a random neighbor and adopts the neighbor's opinion as its own. Again, such a system always reaches a consensus, and coalescing random walks [DKS91] can be used to bound the convergence time. In order to explain why consensus doesn't always arise in the real world, one can posit the presence of stubborn agents, agents which never change their own opinions (though they may certainly influence others). More generally, multiple studies [Asc55, DG55,LRL79] have confirmed that even when agents are not stubborn, they usually have a conformity bias, i.e., they assign more weight to opinions that are already close to their own. This notion gives rise to the definition of influence systems or flocking models.
The most popular such flocking model is the Hegselmann-Krause system [Kra97,HK02], and this is the system that we focus on here. In its simplest incarnation, the system consists of n agents placed on the real line, with the i'th agent at x t (i) at time t, and at every time step t 0, the positions update as follows for all i ∈ [n] synchronously: Here, 1 is the confidence bound, as each agent only has confidence in those agents which are within this bound.
The HK system has become quite popular as a mathematically clean and simple formalization of an endogenous dynamical system that captures interesting qualitative properties, such as polarization and conformity bias, found in opinion dynamics 1 . The expectation is that a systematic understanding of the dynamics of the HK system can lend insight into more detailed models and, hopefully, (some aspects of) reality. Indeed, convergence for the HK system is immediate, and the time complexity needed for the system to freeze is known upto a linear factor in n (the best upper bound is O(n 3 ) [BBCN13,MT13] and the best lower bound Ω(n 2 ) [WH14]). Note that it is not a priori clear that there exists a bound which depends only on n and is independent of the agents' initial positions.
But perhaps surprisingly, it has turned out that changing the model in even very simple ways leads to problems which we cannot handle mathematically. For example, one of the most common variations is to let the confidence bound depend on the agent [Lor10,MB12]. That is, 0. This is called the heterogeneous Hegselmann-Krause system. A rigorous proof of its convergence is still missing, although convergence seems clear from simulations! This situation clarifies that we need to develop new technical tools in order to understand the dynamics of influence systems. Indeed, for general influence systems, Chazelle, in an impressive sequence of papers [Cha11,Cha12], developed a new algorithmic calculus to show that general influence systems are almost surely asymptotically periodic. However, these general results do not imply convergence for the specific case of heterogeneous HK.
Our Contributions
In this work, we rigorously study the convergence behavior of two variations of the HK model. Apart from their intrinsic interest from a sociology perspective, their analysis seem to pose similar mathematical difficulties as heterogeneous HK. Nevertheless, we show that for both these systems, we can develop tools to understand their convergence behavior. Specifically, we study the following two models: -Social Hegselmann-Krause: One criticism of the Hegselmann-Krause model is that it only considers informational influence and ignores normative effects altogether. It ignores the fact that individuals belong to groups, and generally, information exchange only occurs between individuals in the same group. To model this fact, we assume that there exists an underlying social network such that two agents can interact with each other only when there is an edge between the two in this graph.
We formally define the social Hegselmann-Krause system as follows: given an undirected graph G on n nodes and a collection of n agents initially at positions x 0 (1), . . . , x 0 (n) ∈ R respectively, the social HK system updates the agents' positions synchronously for t 1 according to Equation The social HK model differs in some very basic ways from the usual HK model, as shown in Figure 1. First of all, it is no longer true that the agents freeze after some time; agents can keep moving by some tiny amount indefinitely as the example in Figure 1a shows. We might hope though that, for every ε > 0, after some time bound that depends on n and ε, the points stay within intervals of ε. Even this is not true as the situation in Figure 1b of the panel shows; there, by making δ > 0 arbitrarily small, the agent initially at 0 can take arbitrarily long to "see" the agent initially at 2 − δ. Finally, unlike the usual HK system, the agents do not preserve their order, as is clear from Figure 1c.
As far as we know, we are the first to formally study the convergence behavior of the social HK model. Fortunato [For05] also investigated the same system but to address a very different problem: Given a random initial configuration of agents in the interval [0, 1], what is the minimum confidence bound that ensures that the agents come to a consensus when the dynamics is that of social HK on a random graph of degree d? Fortunato's empirical result 3 is that this threshold is ∼ 0.2 when d = ω(1) and is 0.5 when d is constant. Perhaps in the style typical of physicists, their work focused on the equilibrium outcome, whereas we study the transient behavior.
Given ε, let us call a step of the dynamical system ε-non-trivial if at least one pair of interacting agents is separatated by distance at most ε.
Theorem 1.1. Given an arbitrary initial configuration of n agents evolving according to the social HK model defined by an arbitrary graph, the number of ε-non-trivial steps in the dynamical system is O(n 5 /ε 2 ).
Chazelle's result [Cha11] implies an n O(n) bound for this system whereas our bound is polynomial.
We also show that the same bound holds when the social network itself changes with time, provided its evolution follows certain constraints. Informally, we require that if two agents interact at time t and they are within each other's confidence bound at time t + 1, then they should keep on interacting at time t + 1. In particular, if edges are never deleted from the social network, then only polynomial number of non-trivial steps take place Non-deterministic Hegselmann-Krause: A different criticism of the HK system is that it is very rigid in that each agent must move to exactly the mass center of the agents within unit distance. In particular, if two agents have the same set of agents within unit distance, then they move to exactly the same opinion at the next time step (and stay together thereafter). This is clearly not very realistic, as the effects of chance and variation are not taken into account.
To address these issues, for any fixed ε ∈ [0, 1], we formally define the ε-non-deterministic Hegselmann-Krause system. The system again consists of n agents placed on the real line, with the i'th agent at - Figure 2: Order not preserved by non-deterministic dynamics x t (i) at time t, and at every time step t 0, the positions update for all i ∈ [n] synchronously as: . Note that we term the system "non-deterministic" instead of "noisy", because the ε i,t 's are not assumed to be random. In fact, they can be entirely arbitrary values in [−ε, ε], even chosen by an adversary depending on the current state. However, note that consensus remains a fixed point of this dynamics.
The dynamics of non-deterministic HK is quite different from that of HK. As Figure 2 shows, the order of agents can change even when there are two agents and ε is arbitrarily small. Also, the fact that agents do not coalesce together once they see the same set of agents complicates the system's behavior significantly. The work here is the first to handle such a general type of non-determinism in doing convergence analysis. In [BBCN13], a one-sided version of noise was discussed but there, the authors could adapt the argument to bound the convergence time of exact HK in a simple way. In contrast, we are not able do so here. Our main result is: Starting with an arbitrary initial arangement of n agents evolving according to the non-deterministic HK model, the number of steps before which all the agents are confined within intervals of length ρ thereafter is O(n 4 + log(1/ρ)/ log n).
Again, in this case too, Chazelle's general result on bidirectional influence systems [Cha11] implies an n O(n) bound whereas we show a polynomial bound.
Our Techniques
Social HK. In [BBCN13], the classical HK system is shown to converge in O(n 3 ) steps by simply showing that the diameter of the system must shrink by a significant amount at each timestep, unless the leftmost and rightmost agents have already frozen. It is not clear how to extend this proof approach to the social HK system, because as Figure 1b shows, the leftmost agent may not be frozen but may take an arbitrary long time before it moves by a significant amount. Therefore, we must search for some other energy function, which is not the diameter but which is also decreasing.
In fact, such an energy function has already been introduced [RMF08]. For a given configuration x = (x(1), . . . , x(n)), define E(x) = (i,j) min(|x(i) − x(j)| 2 , 1). The fact that this energy is decreasing is an important ingredient in several previous works [BHT07,BHT09,BBCN13,EB14]. We generalize this energy function to the social HK setting as: To give a lower bound on the decrease of this energy function, we use an elegant approach introduced recently by Martinsson [Mar15]. Martinsson relates the decrease to the eigenvalue gap of the communication graph, which is well-known to be 1/poly(n). We show that the same approach applies to the social HK system also. We get that the energy decrement in every ε-non-trivial step is Ω( ε 2 n 3 ), and as the energy of any initial configuration lies between 0 and O(n 2 ), this gives an upper bound of O( n 5 ε 2 ) on the number of ε-non-trivial steps. This approach continues to apply even when the social network graph is changing under some natural restrictions.
Non-deterministic HK. To analyze the non-deterministic HK system, we use a mixture of geometric and algebraic techniques similar to the ones presented in [BBCN13]. However, the technical details are significantly more involved. The basic approach in [BBCN13] is to look at the neighbor immediately to the right of the leftmost agent and analyze its influence. However, in our case, we need to partition the neighbourhood of the leftmost agent into two subsets and treat the subsets collectively. One might also notice that proving a lower bound on the movement of the leftmost agent as in [BBCN13] is not enough, as the order of positions is not preserved and the leftmost agent may change over time. To overcome this difficulty, in our work we lower bound the amount by which the diameter (difference between the rightmost and the leftmost agents, whose identities change with time) shrinks within n time steps, or else, we show that some agents separate out leaving behind sub-systems with smaller number of agents, where they both converge independently thereafter.
Social HK model
We reformulate the social HK model in the multidimensional setting. This is very natural when an opinion consists of positions along multiple axes instead of just one.
Let x t (i) ∈ R d (for d 1) be the position of the ith agent at time t, and let G be a fixed undirected graph on n nodes. The dynamics is given by the following equation: where N t (i) = {j : (i, j) ∈ E(G) and x t (j) − x t (i) 2 1}. We denote the new state at time t + 1 by: where P t is a row-stochastic matrix. As mentioned in the Introduction, our proof follows the same line as the recent analysis by Martinsson [Mar15] but with some twists due to the presence of the social network graph. For any configuration x = (x(1), x(2), . . . , x(n)) of n agents, define the communication graph C x so that two nodes i and j are adjacent exactly when (i, j) ∈ E(G) and x(i) − x(j) 2 1. That is, the communication graph is now the conjunction of the social network G and the standard HK communication graph. Also, for a configuration x of n agents, we define its energy as Note that the energy of any configuration lies between 0 and n 2 . For the standard HK system, a very useful fact is that the energy E(x t ) is non-increasing in t; see Theorem 2 of [RMF08] for a proof. In fact, this fact is the driver for the bound on the freezing time of multidimensional HK found in [BBCN13]. Our proof shows that the same energy decreases over time for the social HK system also. For a given state x and for any ordered pair (i, j) ∈ [n] 2 , we say that (i, j) is active if (i, j) ∈ E(C x ). We consequently define the active part of the energy of x as Now, let {x t } be a sequence of configurations evolving according to (4). For simplicity of notation, let E t denote the edge set of the communication graph C xt .
Proof. There are four cases to look at: In this case. we are adding the same term ( x t+1 (i) − x t+1 (j) 2 2 ) to both LHS and RHS. 2. (i, j) / ∈ E t and (i, j) / ∈ E t + 1 In this case too, we are adding the same term (1) to both LHS and RHS.
1. Hence in this case too we are adding a term (1) to RHS which is at least the term ( x t+1 (i) − x t+1 (j) 2 2 1) added to the LHS.
As the inequality is true term-wise, we have LHS RHS.
Proposition 1 (Proposition 2.2 in [Mar15]). For each t 0, let λ t = max{|λ| : λ = 1 is an eigenvalue of P t }. Then: Proof. We reproduce the proof from [Mar15] for completeness. Let A t denote the adjacency matrix of C xt , and let D t denote its degree matrix, that is, the diagonal matrix whose (i, i)'th element is given by the degree of i(Recall that every vertex in C xt has an edge to itself). Observe that P t = D −1 t A t . Recall E t = E(C xt ). We have: where Tr(·) denotes trace. Here, we interpret x t as an n × d matrix. Now consider: This inequality follows by standard linear algebra.
The following is a standard result in spectral graph theory: Proposition 2. For any t 0, we have where diam(C xt ) denotes the graph diameter of C xt . If C xt is not connected, we interpret diam(C xt ) as the largest diameter of any connected component of C xt .
We are now ready to prove our main theorem. Observe that E act (x t ) > ε 2 whenever the t'th step is ε-non-trivial. Theorem 1.1 is therefore immediately implied by the result below.
Theorem 2.2. For any ε > 0, given a social HK system with n agents in R d , there are O(n 5 /ε) values of t for which E act (x t ) > ε.
Proof. Now given any initial configuration, we have diam(C xt ) n (or else, we can decompose the system into independent subsystems and analyze each separately). Applying Proposition 2, it follows that the energy decrement in each step with E act (x t ) > ε is Ω( ε n 3 ) and such steps can hence occur at most O( n 5 ε ) times.
Changing social network
One may ask what happens to the convergence rate when the social network itself evolves with time and is not fixed. Let G t denote the social network graph at time t. To make sure that the above proof carries through, we need to suitably restrict the evolution of G t . Given a sequence of configurations x t = (x t (1), . . . , x t (n)), we again have the communication graph C xt where two nodes i and j are adjacent if (i, j) ∈ E(G t ) and x t (i) − x t (j) 2 1. As before, let E t = E(C xt ).
Definition 2.3. Call a social HK system defined by a sequence of time-varying social networks G t as friendly if it is the case that whenever (i, j) ∈ E t and x t+1 (i) − x t+1 (j) 2 1, (i, j) ∈ E(G t+1 ) (and hence, (i, j) ∈ E t+1 )).
In other words, in a friendly HK system, if two agents interact at time t, and they stay within distance 1 in the next time step, then they keep interacting with each other at time t + 1. Note that the evolution of G t may be endogenous (i.e., depend on the states x t ). We observe that under this natural condition of friendliness, the above proof goes through without any changes.
Theorem 2.4. For any ε > 0, given a friendly social HK system with n agents in R d , there are O(n 5 /ε) values of t for which E act (x t ) > ε.
Proof. The only part of the proof which needs a second look is case 3 in the proof of Lemma 2.1. Now, the definition of friendliness ensures that if (i, j) / ∈ E t+1 and (i, j) ∈ E t , then x t+1 (i) − x t+1 (j) 2 > 1.
Note that without the friendliness assumption, Chazelle [Cha11] shows a bound of n O(n) for the number of non-trivial steps. We conjecture that the friendliness assumption is necessary for a polynomial bound.
Experimental Results
Our analysis above is very general in the sense that our bound for the number of non-trivial steps does not depend on the structure of the social network. Is it true that some social network structures allow faster convergence than others? As a first cut at this question, we explore how the edge density of the social network affects dynamics. Let G(n, p) be the Erdős-Renyi random graph on n nodes, where each pair of nodes is an edge with probability p independently. How does p change the time needed to converge? We study this question when the initial positions of the agents are uniform in the interval [1, n]. Figure 3 summarizes the results of our computer experiments.
We need to clarify what we mean by "convergence time" in the figure. As Figure 1b shows, the time needed to converge can, in general, be arbitrarily long. However, it seems that if the initial positions of the agents are chosen randomly, such pathological cases occur with probability 0, so that after a finite time, all agents are confined to an interval of length 10 −6 ever after. We do not have a rigorous proof of this claim, but all our simulations support it. In view of this, the notion of convergence time we used to arrive at Figure 3 was, for any given n and p, the least time t at which the sum of the movement of all the n agents is less than 10 −6 , averaged over 1000 random initializations of the agents and random graphs from G(n, p) .
What is most interesting about Figure 3 is that there exists a value of p between 0 and 0.3 for which the convergence time is maximized. When p is close to 0, the communication graph consists of many small disconnected components and convergence occurs fast. p = 1 corresponds to the standard HK model. Somewhere in between, the time needed to converge reaches a maximum. The lesson seems to be that opinions take the longest time to converge when the probability of two agents interacting is neither too small nor too large. We also conducted the experiment when the social network was chosen from the Barabási-Albert generative model for scale-free networks. The results there (not shown) are qualitatively similar.
Note that the maximizing value of p seems to decrease slowly with n in Figure 3. It is not clear what the limiting behavior is as n grows. Also, we currently do not have any analytical way to understand the experimental results.
Non-Deterministic Hegselmann-Krause Model
In this Section, we analyze the non-deterministic HK model. Recall that the update rule for the nondeterministic HK model is: where ε i,t is an arbitrary number generated from the interval (−ε, ε) at time t, for every t 0. We analyze the time needed for convergence when ε is sufficiently small. We first establish some notation. Let (t) be the index of the leftmost agent at time t. As already noted in the Introduction, (t) can change with t. Also, let x t ( ) and N t ( ) be shorthand for x t ( (t)) and N t ( (t)) respectively. Similarly, let r(t) be the index of the rightmost agent at time t, and let x t (r) and N t (r) denote x t (r(t)) and N t (r(t)) respectively.
Lemma 3.1. Suppose ε < 1 n−1 . Then, for any agent i ∈ [n] and for all t 0, x t+1 (i) x t ( ). In particular, Proof. Let δ = x t (i) − x t ( ). At time t + 1, without noise, agent i can move to the left by at most δ(1 − 1 n ). By substituting in (11), we get: Remark 3.2. By exactly the same reasoning, the position of the rightmost agent does not move to the right over time if ε < 1 n−1 . Definition 3.3. For all t 0, define the following sets: We next show that any agent in M (t) actually satisfies Lemma 3.1 with strict inequality.
Lemma 3.5. For any agent i ∈ S(t) and for all t 0, x t+1 (i) x t ( ) + 1 n − ε Proof. Consider i ∈ S(t). Let k = |N t ( )| and δ = x t (i) − x t ( ). Then agent i moves to its left by at most: Combining Lemma 3.4 and Lemma 3.5: Corollary 3.6. For any agent i ∈ M (t) and t 0, x t+1 (i) x t ( ) + 1 n − ε Lemma 3.7. Suppose ε < 1 n−1 . For any t 0, if S(t) = ∅, then N t ( ) = L(t) evolves as an independent system and for any ρ > 0, all the agents in L(t) lie within an interval of length at most ρ in time O((log 1/ρ)/(log 1/ε)) time.
Note that case S1 of the above Lemma is irreversible in the sense that by Corollary 3.6, each time it occurs, a subset of the agents converges independently into an interval of an arbitrarily small length ρ. We next establish that case S3 can also occur only a finite number of times.
We are now ready to prove our main theorem.
Theorem 1.2 (recalled) Suppose ε < 1 4n 2 . Starting with an arbitrary initial arangement of n agents evolving according to the non-deterministic HK model, the number of steps before which all the agents are confined within intervals of length ρ thereafter is O(n 4 + log(1/ρ)/ log n).
Proof. Since for any t, |L(t)| n, case S2 can occur consecutively at most n times. So, within every n time steps, case S1 or case S3 must occur at least once. Case S1 can clearly occur at most n times, whereas by Lemma 3.9, case S3 can occur O(n 3 ) times. Hence, after time O(n 4 ) time steps, all agents lie in independent subsystems, each of diameter at most 1. Each of these subsystems, by Lemma 3.7, cluster into intervals of length at most ρ in O(log(1/ρ)/ log(1/ε)) = O(log(1/ρ)/ log n) time steps.
Remark 3.10. Suppose we change the definition of non-deterministic HK models so that each agent is influenced non-uniformly by its neighbors. Specifically, let the update rule be: where each ε i,j,t is an arbitrary 4 number generated from the interval (−ε, ε). Most of the above proof needs no modification. (12) changes to: if ε < 1 4n 2 . Everywhere else, the claims hold straighforwardly using the upperbound of ε on each ε i,j,t . Hence, for this system also, Theorem 1.2 holds.
An interesting aspect about (15) is that an agent might move in the opposite direction than it would move in the classical HK model, whereas in (11), the direction of each agent's movement is the same as in classical HK.
Future Directions
There are a number of open directions suggested by the problems studied in this work.
-In our formulation of the social HK model, we required the underlying social network to be undirected.
This leads to bidirectional dynamical systems. What happens if the social network is directed? Proving convergence for the HK model with a directed social network seems quite challenging because it includes, as a special case, the ill-understood HK model with stubborn agents (i.e., there are some agents with confidence bound 0 while all others have confidence bound 1). To see this, let every non-stubborn agent have edges to all agents and every stubborn agent have no outgoing edges.
-We introduced the notion of friendly social HK systems in Section 2.1 and showed that these allow only a polynomial number of non-trivial steps. We conjecture that friendliness is necessary for a polynomial bound. Can one demonstrate a non-friendly HK system for which there are an exponential number of non-trivial steps? Is friendliness indeed a tight condition for a polynomial bound?
-Is there a rigorous justification for the empirical results reported in Section 2.2? In general, it would be interesting to understand the effect of the social network structure on the dynamics of the HK model.
-For the non-deterministic HK model, it is important to increase the range of ε for which Theorem 1.2 is valid. Note that if ε is allowed to be in [−1, 0], then we could prove convergence for the HK system with stubborn agents (by simply setting ε i,t = −1 for all t if agent i is stubborn and ε i,t = 0 otherwise). Moreover, in the generalized non-deterministic HK systems described in Remark 3.10, if ε is allowed to be in [−1, n − 1], then we can simulate arbitrary heterogeneous HK systems (by setting ε i,j,t = −1 if j / ∈ N t (i) and ε i,j,t = |{k:|xt(k)−xt(i)| 1}| |Nt(i)| − 1 otherwise).
-Can we prove a polynomial bound for the convergence of the non-deterministic HK model in multiple dimensions? Our current proof does not extend while the approach used in Section 2 seems sensitive to the presence of non-determinism. | 2015-03-05T10:15:56.000Z | 2015-03-05T00:00:00.000 | {
"year": 2015,
"sha1": "d2c2318377c20f8c8eaf9a6d9a293bc681115558",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/1503.01720",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "54dd583cf3b0a3f19c87ebea5d6921436c52567b",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Mathematics",
"Computer Science",
"Physics"
]
} |
34157660 | pes2o/s2orc | v3-fos-license | Tributylamine Facilitated Separations of Fucosylated Chondroitin Sulfate (Fucs) by High Performance Liquid Chromatography (HPLC) into its Component Using 1-Phenyl-3-Methyl-5-Pyrazolone (PMP) Derivatization
Monosaccharide characterization study can give valuable information on newly discovered intact mucopolysaccharide. The compound fucosylated chondroitin sulfate (FuCS) is a unique mucopolysaccharide having various bioactive properties reported only in targeted holothurians species. The monosaccharide composition of FuCS from a sea cucumber, Holothuria arenicola was studied by comparing a modified and conventional 1-phenyl3-methyl-5-pyrazolone (PMP) derivatization method where tributylamine addition was found to improve the chromatographic conditions. The monosaccharide proportion of FuCS from H. arenicola was determined to be 1:1:0.44 for glucuronic acid: N-acetylgalactosamine: fucose, respectively. Thus, the FuCS monosaccharide ratio of H. arenicola was first reported and this method is more useful for structurally similar compound analysis.
Introduction
Fucosylated chondroitin sulfate is an unusual compound having various physicochemical activity found in sea cucumbers [1]. The mucopolysaccharide consists of a common chondroitin sulfate chain with the β-D-glucuronic acid moiety O-substituted with sulfated fucosyl at the carbon-3 position [2]. The hydrolysis sequence of this mucopolysccharide into its component starts by removal of sulfate esters at the vicinity of fucose branch followed by fucose release and sulfates in the chondroitin backbone. After that, the chondroitin backbone itself is hydrolyzed into glucuronic acid and acetylgalactosamine. Anticoagulation activity is one of most studied bioactivity expressed by this compound and the structural configuration is a determining factor for various physicochemical properties. Apart from that, the derived components are molecules having novelty pharmaceutical benefit. For example, glucuronic acid or its reduced form, glucuronolactone are natural antioxidant having hepatoprotective function [3], whereas, sulfated fucose are known anticancer agent [4].
Analyzing glycosaminoglycans is challenging due to their polydispersity, sequence heterogeneity and high negative charge density [5]. Monosaccharide composition is often characterized to understand the function of intact carbohydrate chain but saccharides are naturally low in UV absorbency. Therefore, derivatization is often required to label the compound with chromophore tag that allows ionic property changes to promote the desired separation conditions. 1-phenyl-3-methyl-5-pyrazolone (PMP) derivatization is particularly well documented for carbohydrates separation using HPLC since PMP derived carbohydrates are more hydrophobic and thus better retained in the reverse phase material. Separation techniques are involving the use of PMP for improving with the current knowledge that ionpairing reagent facilitates carbohydrates derivatization. To the best of our knowledge, there is currently no reported on the use of ionpairing techniques in the derivatization steps for determination of FuCS monosaccharide composition from sea cucumbers. Here, in this research, we were described the FuCS monosaccharide composition from a sea cucumber, Holothuria arenicola and the efficacy of using tributylamine facilitated derivatization strategy.
Preparation of fucosylated chondroitin sulfate from sea cucumber
FuCS was extracted from the body wall of sea cucumber Holothuria arenicola harvested in Semporna, Sabah, Malaysia (mean weight 60 g) using a modified method from Mourão et al. [6]. Briefly, the sea cucumber was eviscerated and cleaned under running tap water to remove visible impurities. The body wall was carefully cut to separate from other tissue and homogenized. The homogenate was treated with acetone to dehydrate for 24 h. The dried residue was then digested with papain in a 0.1 M sodium acetate buffer solution (pH 6) containing 5 mM EDTA and 5 mM cysteine at 60˚C for 24 h to remove proteins in the sample. The supernatant was collected, precipitated with 5% cetylpyridinium chloride solution and kept in room temperature. The clear precipitate was collected with centrifugation and the resultant pellet was further precipitate first with 95% ethanol and washed twice with 80% isopropanol. The final precipitate was dialyzed against distilled water at room temperature using Spectra/Por membranes (MWCO 1000) for 3 buffer changes (2h; 6h; overnight) and finally lyophilize. The crude extract was fractionated on a FPLC systems (AkTAPRime) equipped with a DEAE Sephadex A-50 column with elution gradient from 0.2 to 2 M NaCl buffer (pH 7). Aliquots of each collected fractions were analyzed by phenol-sulfuric assay in microplate format adapted from Masuko et al. [7]. Positive fractions were lyophilized and afforded 5.7% FuCS extract.
Monosaccharide analysis
Purified FuCS (~1 mg) was hydrolyzed with 2 M TFA (1 ml) at 110˚C for 4 h. The hydrolyzate was dried under a stream of dry nitrogen gas to remove TFA and was derivatized with PMP according to the following conventional and modified method.
Conventional PMP derivatization method
The dried hydrolyzate was derivatized with PMP based on Honda et al. [8]. Briefly, 1 mg of TFA hydrolyzed samples were added into a 2 ml reaction vial containing 100 µl of 0.3 mol/L methanolic PMP solution and subsequently added 50 µl of 0.3 mol/L NaOH. The mixture was left to react at 60˚C for 60 min. After cooling to ambient temperature, the mixture was neutralized with 50 µl of 0.3 mol/L of HCl and diluted to a final volume of 1 ml. Excess PMP reagent was removed with three times 1 ml chloroform extraction. The final aqueous layer was diluted with deionized HPLC grade water prior to HPLC analysis.
Modified PMP derivatization method
Derivatization procedure was adapted from Zhang et al. [9] and replacement used NaOH with 50 µl tributylamine as the catalyst. Without NaOH, the neutralization step with HCl was omitted. The mixture (sample + tributylamine + methanolic PMP) was vortexed and left to react in a thermomixer for 30 min at 60˚C with 300 rpm agitation. The mixture was then centrifuged at 6000 rpm for 10 min upon completely reaction and formed two-layers. The aqueous, lower portion was made to 1 ml. The upper organic layer consisted of tributylamine with excess PMP reagent removing by extraction with three times of equal volume of chloroform. The final aqueous layer was diluted for final concentration prior to HPLC analysis.
HPLC analysis
The PMP derivatized samples (n=3), individual sugar standards (n=3) and mix sugar standards (n=3) were carried out on Waters HPLC system equipped with a model 1515 quaternary pump, a model 2717W autosampler and a 2487 dual UV wavelength detector. The column (Agilent, Zorbax Eclipse XDB-C18, 250 × 4.5 mm) was optimized for PMP separations on Breeze IIHPLC system software. The flow rate was set constant at 1 ml/min. Mobile phase A and B were consisted of water and acetonitrile; used with a gradient run of 10 to 90% B in 18 min and hold for another 3 min during run. The injection volume was set to 20 µl. Single wavelength at 245 nm was used for data acquisition with a sampling rate of 2, 2.0 AUFS and 0.1 times constant. Each samples analysis was done in triplicates.
Sample preparation for GCMS analysis
A confirmation study was done using Gas Chromatography Mass Spectrometry (GCMS) for identification of monosaccharide composition using the N,O-Bis(trimethylsilyl)trifluoroacetamide (BSTFA) derivatization method. The hydrolyzed syrup was subjected to silica gel (60-200 µm mesh size) column chromatography purification using hexane (20 ml), methanol (20 ml) and acetonitrile:water (20 ml, 1:1, v/v) fractions. Fractions were collected and dried using reduced pressure and redissolve with 50 µl of BSTFA regent. The mixture was incubated at 60˚C for 30 min in a heating block and subsequently purge with a constant nitrogen stream to remove excess reagent. The derivatives were made for the final dilutions using acetonitrile before injecting into GCMS.
GCMS analysis
The prepared polar fractions (methanol and ACN:water, n=3) were injected into GCMS system consisting of an Agilent 7890A gas chromatograph system tandem with an Agilent 5975C mass spectrometry detector. Separation was done using capillary column HP-5MS (30 m × 0.25 mm) of 0.25 µm coated film thickness in splitless mode. Injector temperature was adjusted at 250°C and the oven temperature ramp settings were as follow: initial temperature was held constant for 3 min at 156˚C and ramp to 180˚C in 25 min at a rate of 1˚C/min. The flow rate used was a constant 1 ml/min of high purity helium gas (99.9% pure) as the carrier gas. Compound identification was done with matching scan spectral from National Institute of Standards and Technology (NIST) library and the compositions were computed with reference to the abundance of the compounds in chromatogram.
HPLC analysis of monosaccharide composition
The study objective was to investigate the influence of tributylamine on separation of FuCS monosaccharide composition and thus conventional and modified derivatization methods were compared with mix sugar standards. Figure 1a and b were the chromatograms for conventional derivatization method and modified method respectively on derived FuCS monosaccharide composition. The three main monosaccharide retention time were successfully separated among each other as expected. Each of the derivatization peaks were matched with standard monosaccharide derived using the same derivatization manner ( Figure 2) and the order of separations were fucose (Fuc), glucuronic acid (GlcUA) and N-acetyl-galactosamine (GalNAc).
Conventional derivatization method gave ubiquitous spurious peaks probably due to the organic salt formed by addition of sodium hydroxide and during hydrochloric acid neutralization. Zhang et al. [9] demonstrated the addition of triethylamine can facilitate reaction to completeness in a mixture of sugar standards. In our conventional method, GalNAc-PMP derivative had lower peak intensity than GlcUA-PMP. We noticed that the concentration of NaOH was not sufficiently high to facilitate complete reaction and thus have the most obvious decrease in average peak area of 28.13% relative to GlcUA-PMP. The relative standard deviation for reproducibility was calculated to be 2.83% which was acceptable range while stability of derivatized samples when monitored within a 48 h frame at 8 h interval. Peak area was satisfactory with a decrease of 3.87 % and 11.14% at 24 h and 48 h, respectively.
The FuCS structure is common backbone of chondroitin sulfate reported by several authors [6,[10][11][12]. In fact, previous studies on similar compound found almost same amounts of GlcUA and GalNAc with different molar range of Fuc and sulfate [13]. Hence, we were presented the monosaccharide proportion of H. arenicola FuCS to be 1.0:1.0: 0.44 (Table 1) based on the modified method.
Zhang et al. had successfully determined several types of for triethylamine which eventually affects by charge delocalization. Therefore, tributylamine is more readily to accept the protons from the deprotonated monosaccharide species. The ability of the deprotonated hydronium ions was stabilized by tributylamine result in a rapid formation of ducing region which is essential for the PMP reagent to bind (Scheme 1).
To give a better confidence on the monosaccharide composition of purified FuCS, our research group determined the monosaccharide composition using GCMS. Methanol fraction afforded most of the sugar component while the ACN:water fraction has little to trace amount of other sugar species. The hexane fraction consisting of non-polar compounds in trace amount is presumably impurities and byproducts produced during TFA hydrolysis. The main monosaccharide component was identified from HPLC analysis and also presence in the methanol fraction, however, not clear separation (Figure 3) when the use of a semi-polar stationary phase column, HP-5MS. The column used is not specific to carbohydrate analysis and the separation is solely based on the oven temperature, where the volatility of the compounds is reflected in the retention time. Failing the distinct separation by using non-specific column is due to the compounds nature, as the interested monosaccharides using triethylamine as a catalyst during PMP derivatization [9]. The basis of choosing tributylamine in this experiment is that longer alkyl chain length showed better compatibility with on-column retention of saccharides for a typical C18 type stationary phase [5]. Furthermore, longer alkyl chain was presumably better catalyst during PMP derivatization since longer alkyl chain length can give a higher pKa at 10.89 compared to 10.78 compounds (GlcUA, GalNAc, and Fuc) are homological similar. Thus, the obtained total ion chromatograms from GCMS were further tested by using deconvolution strategy for determination its separation integrity. Derivatized fucose with trimethylsilyl was seen clustering in unresolved peaks within retention time 7.846 to 9.314 while derivatized galactosamine signal appear at later retention time within 12.578 to 13.452. Derivatized glucuronic was not observed in methanol fraction but appear within a peak range of retention time 15.770 to 16.312.
Conclusions
The present study using a modified PMP-derivatization with addition of tributylamine is a robust method to determine hydrolyzed FuCS component. The improved chromatographic properties are allowed good separation on main monosaccharide composition of FuCS from H. arenicola. Nevertheless, it is tentatively proposed that non-commercial sea cucumber species can be served as a source of FuCS compound for potential pharmaceutical industry. | 2019-04-07T13:07:05.890Z | 2014-12-10T00:00:00.000 | {
"year": 2014,
"sha1": "0fa24b4a2ffedcb72dc21beedf42ccb8c5f60066",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.4172/2157-7064.1000256",
"oa_status": "HYBRID",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "5a22ea6e876fb0df8feff97f4fbdce21bc5352c5",
"s2fieldsofstudy": [
"Biology",
"Chemistry",
"Environmental Science"
],
"extfieldsofstudy": [
"Chemistry"
]
} |
254688068 | pes2o/s2orc | v3-fos-license | Tachykinins, new players in the control of reproduction and food intake: A comparative review in mammals and teleosts
In vertebrates, the tachykinin system includes tachykinin genes, which encode one or two peptides each, and tachykinin receptors. The complexity of this system is reinforced by the massive conservation of gene duplicates after the whole-genome duplication events that occurred in vertebrates and furthermore in teleosts. Added to this, the expression of the tachykinin system is more widespread than first thought, being found beyond the brain and gut. The discovery of the co-expression of neurokinin B, encoded by the tachykinin 3 gene, and kisspeptin/dynorphin in neurons involved in the generation of GnRH pulse, in mammals, put a spotlight on the tachykinin system in vertebrate reproductive physiology. As food intake and reproduction are linked processes, and considering that hypothalamic hormones classically involved in the control of reproduction are reported to regulate also appetite and energy homeostasis, it is of interest to look at the potential involvement of tachykinins in these two major physiological functions. The purpose of this review is thus to provide first a general overview of the tachykinin system in mammals and teleosts, before giving a state of the art on the different levels of action of tachykinins in the control of reproduction and food intake. This work has been conducted with a comparative point of view, highlighting the major similarities and differences of tachykinin systems and actions between mammals and teleosts.
In vertebrates, the tachykinin system includes tachykinin genes, which encode one or two peptides each, and tachykinin receptors. The complexity of this system is reinforced by the massive conservation of gene duplicates after the whole-genome duplication events that occurred in vertebrates and furthermore in teleosts. Added to this, the expression of the tachykinin system is more widespread than first thought, being found beyond the brain and gut. The discovery of the co-expression of neurokinin B, encoded by the tachykinin 3 gene, and kisspeptin/dynorphin in neurons involved in the generation of GnRH pulse, in mammals, put a spotlight on the tachykinin system in vertebrate reproductive physiology. As food intake and reproduction are linked processes, and considering that hypothalamic hormones classically involved in the control of reproduction are reported to regulate also appetite and energy homeostasis, it is of interest to look at the potential involvement of tachykinins in these two major physiological functions. The purpose of this review is thus to provide first a general overview of the tachykinin system in mammals and teleosts, before giving a state of the art on the different levels of action of tachykinins in the control of reproduction and food intake. This work has been conducted with a comparative point of view, highlighting the major similarities and differences of tachykinin systems and actions between mammals and teleosts. KEYWORDS tachykinins, neurokinins, substance P, endokinin/hemokinin, reproduction, food intake, mammals, teleosts 1 Introduction Tachykinins (TAC) are members of a large family of peptides present from cnidaria [for reviews (1,2)]: to bilateria [for reviews (1)(2)(3)(4)(5)(6)(7)(8):. Tachykinins are usually considered as brain and gut peptides, as they are mainly expressed in neurons from the central nervous system and from the gastrointestinal tract. However, they are also present in non-neuronal cells, such as the immune and inflammatory cells of mammals, and various tissues like the skin of amphibians, as well as the salivary gland of mosquito and octopus, where they serve for exocrine secretion [for reviews (1,4,6,8)]. In addition, in sea squirt, they are found in endostyle and gonad (8), where they act as neurotransmitters of endocrine and local autocrine/paracrine regulations [for reviews (4, 8)].
Since the discovery of the co-expression of neurokinin B, encoded by tachykinin 3 gene (tac3), and kisspeptin/dynorphin in neurons involved in the generation of GnRH pulses in mammals, rekindled attention has emerged for studying tachykinins in vertebrate reproductive physiology. Reproduction is classically controlled by the hypothalamuspituitary-gonad (HPG) neuroendocrine axis in vertebrates [for review (9)]. The gonadotropin-releasing hormone (GnRH), produced and released by hypothalamic neurons, acts on the pituitary to stimulate the synthesis and release of gonadotropins, luteinizing hormone (LH), and follicle-stimulating hormone (FSH). These pituitary hormones act themselves on the gonads to control gametogenesis and the production of sex steroids, mainly estrogens in females and androgens in males. These peripheral hormones exert feedbacks at brain and pituitary levels to regulate GnRH and gonadotropin production. The hypothalamus is also the cerebral center involved in the control of food intake, integrating both external and internal factors and producing neuropeptides stimulating (orexigenic) or inhibiting (anorexigenic) appetite [for reviews (10)(11)(12)(13)].
In vertebrates, feeding and reproduction are linked processes, as the presence of sufficient energy reserves is critical to achieve successful reproduction [for reviews (14,15)]. Any state of negative energy balance thus affects not only central appetite-regulating systems but, often, also reproductive pathways and reproductive performance. Hypothalamic hormones classically involved in the control of reproduction, such as kisspeptin, are reported to regulate appetite and energy homeostasis as well, in mammals [for reviews (16)(17)(18)] and teleosts [for review (19)].
As teleosts represent the most diversified group of vertebrates, with nearly 30,000 species, species-specific regulating mechanisms are often encountered inside this lineage. In addition, some physiological differences exist between fish and mammalian regulatory mechanisms, even if major regulatory features are conserved. Some differences may be due to anatomical specificities of teleost neuroendocrine systems such as direct neuronal innervation and cell regionalization of the pituitary [for reviews (9,20)]. Major breakthroughs in the studies of neuropeptide actions in teleosts have been allowed thanks to recently available published genomes and novel genome editing technics. These new tools are of particular interest and necessity in this group of vertebrates, as due to the teleost-specific whole-genome duplication (3R), teleosts possess an expanded number of genes encoding hormones/peptides that will share initial pleiotropic functions (subfunctionalization) or get new functions (neofunctionalization) [for review (9)].
With a comparative perspective, the purpose of this review is to provide a general overview of the tachykinins and their receptors in mammals and teleosts and then focus on the state-of-the-art literature on the different levels of action of tachykinins in the control of reproduction and food intake in these two groups of vertebrates.
Tachykinin system
Some of the first peptides of the tachykinin (TAC) family were discovered in neurons in mammals and therefore named neurokinins (NK). However, many subsequent data showed their production by non-neuronal cells. Especially the discovery in 2000 by Zhang and collaborators of a third tachykinin gene, PPT-C (21), renamed tac4 and encoding several new tachykinin peptides, with widespread peripheral distributions and with a preferred receptor NK1 receptor, led to debates on their nomenclature [for reviews (22, 23)]. A revised nomenclature was proposed, with the preferred term "tachykinin" compared with "neurokinin," which then appeared inappropriate. Similarly, for tachykinin receptors, for example, the NK1 receptor can no longer be defined only as a substance P (SP) receptor [for reviews (23,24)]. More recently, the Human Genome Organization (HUGO) Gene Nomenclature Committee approved the names TACR1, TACR2, and TACR3 for the three TAC receptors (25). This nomenclature will be adopted in our review. In the following text, we will use TAC for tachykinin peptides and tac for tachykinin genes and transcripts, and likewise TACR and tacr for the receptors.
A recent review highlights the widespread distribution and the functional pleiotropy of TACs and their receptors with a special focus on invertebrates (2) and is complementary to our present comparative review in vertebrates.
Tachykinins
The evolutionary scenario of tachykinins in chordates suggests that an ancestral tac gene in proto-chordates generated four paralogs [(26), for reviews (1,27)] after the two whole-genome duplication rounds (1R/2R) which occurred in early vertebrates (28,29). A possible loss of one of the four paralogs occurred before the split of the ray-finned fish, actinopterygians (leading to teleost fish), and the lobe-finned fish, sarcopterygians (leading to mammals) [(26), for review (1)]. Among teleosts, a specific third-genome duplication (3R) produced a tetraploidization, followed by gene loss or conservation of duplicated paralogs (30, 31). Our recent study shows a wide conservation of the duplicated tachykinin genes in the teleost fish lineage (Campo et al. in preparation), thus increasing the scope of previous research (26).
Vertebrate tac genes consist of five to seven exons that encode a pre-pro-tachykinin (PPT) peptide, named PPT-A or PPT-I for tac1, PPT-B or PPT-II for tac3, and PPT-C or PPT-III for tac4 [for reviews (4, 32-34)]. One or two peptides are cleaved from each of the three PPT [for reviews (1)]. The TAC peptide placed close to the N terminal of PPT is called TAC-related peptide or TACRP, while the other tachykinin peptide closer to the C terminal of PPT is named TAC. For the tac1 gene, TACRP is substance P (SP) and TAC are neurokinin A (NKA), neuropeptide K (NPK), and neuropeptide g (NPg), these last two being NH2-terminally extended forms of NKA. For the tac3 gene, TACRP (only found in teleosts) is TAC3RP or neurokinin B-related peptide (NKBRP) and TAC is NKB. For the tac4 gene, TACRP are hemokinin-1 (HK1), endokinin A (EKA), and endokinin B (EKB), while TAC are endokinin C (EKC) and endokinin D (EKD), depending on the splicing variant ( Figure 1). Indeed in mammals, differential alternative mRNA splicing and precursor processing are observed for each of the three tachykinin genes, and as the different transcripts are regulated in a tissue-specific manner [tac1 (35-37); tac3 (38); tac4 (39); for reviews (40, 41)], these mechanisms are likely to A B FIGURE 1 Comparison of the tachykinin system in a mammal, the human, and in a teleost, the grass carp. In human (A), the tachykinin system comprises three tac genes (tac1, tac2, and tac3) encoding up to 10 different TAC peptides (SP, NKA, NPK, and NPg for the tac1 gene; HK1, EKA, EKB, EKC, and EKC for the tac4 gene; NKB for the tac3 gene) due to the existence of various spliced variants (a, b, g, and d for the tac1 and tac4 genes; (a, b, and g for the tac3 gene). These human TAC peptides bind to three TACR (TACR1, TAC2, and TACR3) with different affinities: SP, HK1, EKA, and EKB for TACR1; NKA, NPK, and NPg for TACR2; NKB for TACR3. In teleosts (B), due to the teleost-specific whole-genome duplication (3R), duplicates for tac1 (tac1a and tac1b), tac3 (tac3a and tac3b), and tac4 (tac4a and tac4b) exist and up to 12 different TAC peptides have been identified up to now. Up to six TACR have been yet characterized: two TACR1 (TACR1a and TACR1b), one TACR2, and three TACR3 (TACR3a1, TACR3a2, and TACR3b). One of the 3R-duplicated tacr2 paralogs (tacr2b) was lost in the teleost lineage but conserved in the eels (elopomorphs) (Campo et al. in preparation). Binding studies with the complete available tachykinin system have only been performed in the grass carp, the unique teleost species for now in which the tac4 gene has been identified and published. In this species, tac4b and tacr2b have not yet been identified and appear in transparency in the figure. For more information, please refer to part 1 of this review. EKA, endokinin A; EKB, endokinin B; EKC, endokinin C; EKD, endokinin D; HK1, hemokinin 1; HK2, hemokinin 2; NKA, neurokinin A; NKB, neurokinin B; NKBRP, neurokinin B-related peptide; NPK, neuropeptide K; NPg, neuropeptide gamma; SP, substance P; TAC, tachykinin peptide; tac, tachykinin gene; TACR, tachykinin receptor (protein); tacr, tachykinin receptor gene. Four peptides can be translated from the gene tac1 ( Figure 1): substance P (SP) or TAC1RP from exon 3; neurokinin A (NKA) or TAC1 from exon 6; neuropeptide K (NPK) or TAC1-NPK from exons 4, 5, and 6; and neuropeptide gamma (NPg) or TAC1-NPg from exons 3, 5, and 6 [for reviews (5,34)]. The tac1 gene produces four different splicing variants (a-, b-, g-, and d-tac1): aand d-tac1 generate only SP; b-tac1 encodes SP, NKA, and NPK; g-tac1 generates SP, NKA, and NPg [for reviews (5,34,45)]. Substance P/TAC1RP was the first neuropeptide ever to be extracted in 1931 [(46); for review (47)]. In their pioneer study, von Euler and Gaddum found in extracts from horse brain and intestine an atropine-resistant factor, which induced contraction of the isolated rabbit jejunum and transient hypotension in anesthetized rabbits (46). They named this new factor substance P, with P for powder. It was only in 1971 that SP was purified and sequenced from bovine hypothalamus (48) and synthetized (49). Neurokinin A/TAC1 was discovered later in extracts of porcine spinal cord by different research groups and named differently at that time: neurokinin a (50), substance K (51), or neuromedin L (52). As SP, it was involved in ileum contraction of guinea pig (50, 52). Further analyses of the pre-pro-peptide structure revealed that peptides other than SP and NKA were encoded by the precursor and that tissue-specific alternative splicing occurred (35, 51): neuropeptide K (NPK) with 36 amino acids and neuropeptide gamma (NPg) with 21 amino acids. Both sequences share the last 10 amino acids in the C terminal with a NKA/TAC1 sequence. NPK, isolated from porcine brain, stimulates guinea pig gallbladder contraction, plasma extravasation, hypotension, and bronchial smooth muscle spasm (53). NPg was isolated from rabbit intestine and found to derive from g-pre-protachykinin, hence its name (54).
Teleosts
The first TAC peptide to be characterized in teleosts was substance P. In 1956, a factor purified from cod brain and intestine extracts was found to have the same properties as those of mammalian substance P (55). Later, a tachykinin of 21 amino acid residues, which possesses mammalian NPg characteristics and was named carassin, was isolated from the brain of the goldfish Carassius auratus (56). Then, SP and NKA were measured by radioimmunoassay in the brain of rainbow trout (57). In goldfish, Lin and Peter described two cDNAs encoding g-PPT that may represent different transcripts resulting from the alternative transcriptional start sites and that contains the sequences of SP, carassin, and NKA (58). The Tac1 gene was first characterized in zebrafish and found in the genomes of goldfish, medaka, and stickleback; it encodes SP and NKA (59, 60). One tac1 gene was then found in many other teleost species, including grass carp (61). It was only recently that a second tac1 gene, likely the result of the 3R, was identified in the grass carp; the duplicated genes were named tac1a and tac1b and shown to encode SPa and NKAa, and SPb and NKAb, respectively (62) (Figure 1). Before that study, it was thought that one of the duplicated tac1 paralogs obtained by 3R was lost in teleosts [for review (1)]. Our recent bioinformatic studies revealed a wide conservation of the 3R-duplicated tac1 genes, even those obtained by the further whole-genome duplication of the salmonids (4R) (Campo et al. in preparation).
Tachykinin 3 2.1.2.1 Mammals
The Tac3 gene has been named tac2 in rodents, while it is in fact an ortholog of human tac3 (5, 63-66); so, for easier reading, rodent tac2 will be replaced by tac3 throughout this review. One peptide is encoded in the tac3 gene: neurokinin B (NKB) or TAC3 (5,32,38,41). NKB was purified from the extract of porcine spinal cord simultaneously by two research groups and given different names at that time: neurokinin b (50) or neuromedin K (67). As SP and NKA, it induces contraction of the guinea pig ileum (50, 67). The structure and gene organization of the neuromedin K/NKB precursor (or pre-protachykinin B) was first determined in bovine (68), then in rat (69). In human, a single gene transcript encoding a single precursor and a single TAC was first revealed (70), but then three TAC3 precursors (a, b, and g) were shown to exist ( Figure 1) (38).
Tachykinin 4 2.1.3.1 Mammals
The molecular cloning of a mouse third PPT gene, PPT-C (later renamed tac4), was reported in 2000 (21). PPT-C mRNA was primarily detected in hematopoietic cells, and its derived peptide was shown to be a crucial factor for the survival of B-cell precursors and thus named hemokinin 1 [HK1 (21);]. Rat HK1 is identical to mouse HK1 (mHK1) (80). In human, the tac4 transcript predicts two tachykinin-like peptides: one, at the N terminus, a homolog of mouse and rat HK1, was named endokinin A (HK1/EKA), and the second at the C terminus was named endokinin C (EKC), in line with their proposed peripheral endocrine roles in contrast to the neuroendocrine/ neuronal role of neurokinins (39). Apart from this tac4 transcript that was named a-tac4, three other splicing variants exist in human (b-, g-, and d-tac4) ( Figure 1): b-tac4 codes for EKB and EKD, while g-tac4 and d-tac4 encode only EKB [(39); for review (66)]. TAC4RP-EKA and TAC4RP-EKB are Nterminal extended versions of TAC4RP-HK1, with different lengths; EKB is a truncated form of EKA and EKD an Nterminally modified version of EKC (39). Thus, the tac4 gene encodes up to five peptides in mammals: HK1, EKA, EKB, EKC, and EKD. Three of them can be translated from the TAC4-RP site and two from the TAC4 site: HK1 or TAC4RP-HK1 from exon 2; EKA or TAC4RP-EKA from exons 1 and 2; EKB or TAC4RP-EKB from exons 1 and 2; EKC or TAC4-EKC from exons 3 and 4; and EKD or TAC4-EKD from exon 4. EKC and EKD are designated tachykinin gene-related peptides [for reviews (5,34,40,66)]. Interestingly, TAC4-EKC and TAC4-EKD that correspond to the TAC4 peptide have substituted the C-terminal methionine by a lysine, thus reducing or suppressing the affinity for all TAC receptors, and they differ by the length of the N terminus (39).
Teleosts
It was not clear whether homologs of HK and EK were present in non-mammalian vertebrates until the tac4 gene was identified in the genomes of various teleosts (71) and recently characterized in brain grass carp (62). The grass carp tac4 gene encodes two mature peptides, hemokinin 1 (HK1) and hemokinin 2 (HK2) (Figure 1), as the mammalian tac4. However, the mammalian tac4 can produce up to four different peptides depending on alternative splicing events that have not been observed in teleosts until now. HK-1 displays very weak activation for neurokinin receptors compared with HK2, likely due to a phenylalanine-to-valine substitution in the Cterminal FXGLM signature motif, leading to an inefficiency on pituitary hormone expression in grass carp pituitary cells (62). Shi and colleagues proposed that the fact that only one TAC4 isoform was isolated up to now in teleosts "might be the result of the non-functionalization by forming pseudogenes or deletion/ mutations leading to the loss of redundant genes" (62). Our recent study demonstrates that the tac4 gene has been duplicated during the 3R and two copies of the gene were conserved in most studied species (Campo et al. in preparation) (Figure 1).
Tachykinin receptors
In vertebrates, TAC peptides bind to three receptors (TACR), belonging to the first-class rhodopsin-like G-proteincoupled receptors (GPCR) (also named as family A GPCRs): TACR1, TACR2, and TACR3. These receptors are normally encoded by five exons that include the seven transmembrane domains, an extracellular N terminus enrolled in peptide recognition, and an intracellular C-terminal end in charge of the cellular response after activation of the receptor [for reviews (34, 44)].
The evolutionary scenario of tachykinin receptors in chordates (26) suggests that an ancestral tac receptor (tacr) gene in protochordates generated four paralogs after 1R/2R in early vertebrates (28, 29). One tacr, the tacr4 gene, would have been lost before the split of the actinopterygians (ray-finned fish) and the sarcopterygians (lobe-finned fish). Further duplication of the tacr genes occurred during the teleost 3R. The 3Rduplicated tacr1 and tacr3 genes are conserved in most teleosts (Campo et al. in preparation), while one of the 3R-duplicated tacr2 paralogs (tacr2b) was subsequently lost in the teleost lineage but conserved in the eels (elopomorphs) (Campo et al. in preparation). A local duplication of tacr3a might have occurred to give rise to tacr3a1 and tacr3a2 genes in the teleost lineage (26) Our recent gene search and phylogenetic study confirms this local duplication, but only in Clupeocephala, not in elopomorphs (Anguilla species) or osteoglossomorphs (bony tongue) (Campo et al. in preparation).
Tachykinin receptors in teleosts
Up to six tachykinin receptors have been characterized in teleosts (Figure 1), results of both whole-genome duplication and local gene duplication. In zebrafish, two 3R-duplicated tacr1 (tacr1a and tacr1b), one tacr2, and three tacr3 (tacr3a1, tacr3a2, and tacr3b) are identified, with tacr3a2 arising from a local duplication of tacr3a1 (26). In the grass carp, the same receptors are found: duplicated tacr1 (NK1Ra and NK1Rb in the article), single tacr2 (NK2R in the article), and three tacr3 (NK3Ra1, NK3Ra2, and NK3Rb in the article) (62).
Using COS-7 cells that expressed zebrafish TACR3a1 (Tac3ra in the article) or zebrafish TACR3a2 (Tac3rb in the article), Biran and collaborators reported that both zebrafish NKBa and NKBRP (NKF in the article) were endogenous ligands of TACR3, while zebrafish NKBb was less effective (71). The same year, a local duplication of tacr3a was reported in zebrafish and binding studies of the three zebrafish TACR3 (TACR3a1, TACR3a2, and TACR3b) were investigated. NKBRPa (NKBa-13 in the article) and NKBRPb (NKBb-13 in the article) have higher potencies for inducing promoter activity of TACR3a1 and TACR3a2 in both CRE and SRE transactivation assays than NKBa-10 (26). For TACR3b, the same three NKB peptides have an inducing effect only using the SRE promoter (26). Zebrafish NKBb (NKBb-11 in the article) cannot activate any of the three TACR3s (26). In the same system, tilapia NKBRP (NKF in the article) was more effective than tilapia NKB in inducing the activity of tilapia TACR3a (Tac3ra in the article) and tilapia TACR3b (Tac3rb in the article) (72). In transfected 293-T cells, goldfish NKBa (NKBa-10 in the article), NKBRPa (NKBa-13 in the article), NKBb (NKBb-11 in the article), and NKBRPb (NKBb-13 in the article) can activate TACR3a1 (Tac3ra in the article), while TACR3b (Tac3rb in the article) can be slightly activated only by NKBa-10 (88).
In the grass carp, three studies investigated the receptor selectivity of TAC peptides, using HEK293T cells transfected with each of the six TACRs identified in this species (61, 62, 89) ( Figure 1). The first two articles were reported before the cloning of the tac4 gene and thus did not include TAC4 peptides (61, 89). The authors found that for TACR1 activation, the potencies were grass carp SP>NKA>NKBa>NKBRPa>NKBRPb>NKBb, and for TACR2 activation, grass carp SP, NKA, NKBa, NKBRPa, and NKBRPb had similar potency, except for NKBb which showed a low potency (61). This reveals that carp TACR2 is a multiligand receptor, which could be activated by various TACs with comparable efficacy and potency. Concerning TACR3 activation, for both TACR3a2 (NK3Ra in the article) and TACR3b (NK3Rb in the article), grass carp NKBRPb, NKBa, and NKBRPa were found to be the most effective compared with NKA, SP, and NKBb (89). Interestingly, in the third publication, HK2, product of the tac4 gene, was shown to be able to activate all six TACRs, but with the highest activity for TACR2 (TACR2>TACR3b>TACR1a≈TACR3a1≈TACR1B>TACR3a2), while HK1 displays a very weak activation for each of TACR isoforms (62). These results suggest that while mammalian hemokinin HK1 exhibits the highest affinity for TACR1, teleost HK2 preferentially stimulates the multiligand receptor TACR2 and may thus have a similar function as other tachykinins through its activation (62).
The physiological role of tachykinins in the regulation of reproduction
Activation of the gonadotropic axis at puberty onset and maintenance of the reproductive state (with generation of GnRH pulse in mammals) are under a complex regulatory network. A major breakthrough occurred in 2003 for reproductive neuroendocrinology with the description of hypogonadotropic hypogonadism in human and mice bearing mutation in either kisspeptin gene (kiss) or its receptor (kissr) (90)(91)(92) and the following multiple studies stating that kisspeptins were the most potent secretagogues of GnRH in all mammals [for reviews (93)(94)(95)]. In addition, the involved kisspeptin neurons of the arcuate nucleus (ARC) in the hypothalamus were shown to co-express neurokinin B and dynorphin and were thus referred to as KNDy neurons, first in sheep (96) then in a variety of mammals [for review (97)]. A model was proposed in mammals for the GnRH pulse generator with NKB stimulating kisspeptin release and dynorphin inhibiting it [for reviews (98)(99)(100)(101)]. All these data rekindle attention to other tachykinins, namely, SP and NKA, in reproductive physiology. In contrast, little is known concerning a potential involvement of tachykinin peptides derived from the tac4 gene, hemokinin, and endokinins, in the control of reproduction, but potential actions at peripheral level are observed ( Figure 2). Some redundancies among the three TACR signaling pathways in the control of reproduction may occur (at least in rodents), as blockade of all three receptors (by the use of an antagonist for all three receptors) is required to inhibit LH secretion [ovariectomized (OVX) rat (104)] and the in vitro stimulatory effect of NKB on KNDy neurons is blocked only in the presence of a cocktail of all three receptor antagonists [male mouse (105)]. In both studies, the use of specific receptor antagonists individually has no effect (104, 105).
At the central level of the HPG axis
The main central targets of tachykinins on the HPG axis are the brain gonadotropin-releasing hormone (GnRH) and pituitary gonadotropins (LH and FSH). A recent review deals with the latest advances in our understanding of the biology of tachykinins in the control of GnRH release in mammals (106). However, as the HPG axis is also controlled by dopamine (DA) in teleosts, amphibians, and seasonal mammals, future studies should aim at investigating the potential regulation of dopaminergic neurons by the products of tachykinin genes. Indeed, in mammals, reports have shown interactions between tachykinins and DA at the hypothalamic level [for review (107)]. For example, Billings and colleagues demonstrated the presence of TACR3 in DA neurons in the ewe, abundantly during anestrous when DA mediates the suppression of GnRH and LH release and considerably less during the breeding season (108).
Inactivation of TAC1 system and reproduction
To date, in humans, hypogonadotropic hypogonadism has never been associated with mutation of either the tac1 gene or tacr1/tacr2 genes. However, in mice, inactivation of these different genes leads to different degrees of reproductive impairments, suggesting the need for the whole tachykinin system to get full reproduction. Knockout of the tac1 gene was obtained in female (109) and male (110) mice, which induced a delay in the onset of puberty in both sexes plus a subfertility in females (109). In contrast, mutant mice for tacr1 are fertile (111). More recently, the characterization of a novel mouse line with congenital ablation of tacr2 has allowed to show partially suppressed basal and stimulated LH secretion, with moderate reproductive impact (normal puberty onset and fertility), in these null mice (112). However, in the same species, impairment A B
FIGURE 2
Direct effects of TAC peptides in the control of reproductive function in mammals and teleosts. TAC peptides can act directly at all the levels of the HPG axis (hypothalamus, pituitary, gonads in both mammals (A) and teleosts (B), and other peripheral reproductive organs in mammals). At the brain level (hypothalamus), the effect of TAC on GnRH is likely kisspeptin-dependent in both mammals (KNDy neurons) and teleosts. At the pituitary level, while TAC action is only stimulatory on gonadotropins in mammals, a species specificity is observed in teleosts with either no, stimulatory, or inhibitory effects. At the peripheral level, TAC can act on ovarian steroidogenesis in mammals (either no, positive, or negative effects) and teleosts (positive effects). In mammals, a positive effect is also noted on the contraction of secondary sex organs (uterus in females; vas deferens, seminal vesicles, and prostate gland in male), as well as on sperm motility in males. For more details and for data from in vivo experiments, please refer to part 2 of this review and to Table 1. +, direct stimulatory effect; -, direct inhibitory effect; 0, no direct effect; cyp11a1, gene encoding cholesterol side-chain cleavage enzyme P450scc; cyp19a1, gene encoding aromatase; Dyn, dynorphin; E2, estradiol; ESR, nuclear estrogen receptor; FSH, follicle-stimulating hormone; GnRH, gonadotropin-releasing hormone; GnIH-R, gonadotropin-inhibitory hormone receptor; GPER, G-protein-coupled (membrane) estrogen receptor; hCG, human chorionic gonadotropin; HK1, hemokinin 1; HK2, hemokinin 2; kiss, kisspeptin; LH, luteinizing hormone; NKA, neurokinin A; NKB, neurokinin B; NPK, neuropeptide K; NPg, neuropeptide gamma; P, progesterone; SP, substance P; T, testosterone.
of TAC1 peptide action does not seem to impact reproduction, as when TACR1 or TACR2 antagonists are ip injected to 8week-old or subcutaneously (sc) to 6-month-old female mice, no effect is observed on reproductive success or litter size (113).
Mammals
Before the description of KNDy neuron involvement in GnRH pulse generation and the renewed interest in SP and NKA as regulators of the HPG, few studies were available for a potential direct effect of TAC1 peptides on GnRH [for review (97)]. Using a perifusion system, Ohtsuka and collaborators (114) were the first to show that SP stimulated the in vitro release of GnRH by rat medio-basal hypothalamus ( Figure 2). Consistent with this direct action of SP on GnRH in rodents, its receptor, TACR1, was shown to be expressed in a fourth of GnRH neurons in the mice (115), and binding sites for SP detected in the rat hypothalamus (116). In addition, SP neurons establish inputs to GnRH neurons in the rat septopreoptic area (117) as well as in human diencephalon (118) and eminence median (119). In contrast, in another mammal, the ewe, no coexpression of SP or TACR1 was detected in GnRH neurons, implying no possible direct effect of SP on GnRH in this species (120). None of GnRH neurons express TACR2 in mice (115) and no detection of NKA binding sites was observed in the hypothalamus of rats (116), suggesting a lack of potential direct effect of NKA on GnRH in both species. However, Sahu and Kalra reported in female rat that the TACR2 agonist, but not the TACR1 one, could suppress GnRH release by fragments of median eminence and arcuate nucleus in culture (121).
Apart from a potential direct action of SP and NKA on GnRH neurons via their respective receptors, data in mammals report effects on kisspeptin neurons, as for NKB in the KNDy network. A stimulatory action of TAC1 peptides on kisspeptin has been demonstrated in rodents. In mice, SP and NKA modulate Kiss1 neurons and kisspeptin release [SP and NKA, males (105); SP and NKA, males and ovariectomized and supplemented with E2 (OVX+E2) females (115); SP, females (109); NKA, females (122)]. SP and NKA are also able to depolarize Kiss1 neurons in male mice (105). In addition, the induction of GnRH release by icv infusion of an SP agonist in adult male and OVX+E2 mice is not observed in Kiss1r-/-mice (115). These results on the kisspeptin-dependent effect of SP on GnRH are consistent with the fact that half of kiss1 neurons express tacr1 in this species (115). In rat, icv injection of SP elevates both gnrh and kiss1 mRNA levels (123).
In other mammalian species, anatomical data suggest a possible action of TAC1 peptides on kisspeptin or GnRH, but in vivo studies are often lacking or report only limited action. In postmenopausal women, SP immunoreactivity is detected within Kiss1 neurons in infundibular nucleus (124) and Kiss-SP occasionally contact with GnRH in the postinfundibular eminence (119). However, to our knowledge, no in vivo data for SP or NKA action on GnRH via kisspeptin are yet available in humans. In the goat, SP fibers are also observed in close apposition within ARC KNDy (125), and high doses of TACR1 and TACR2 agonists are needed, and not effective in all individuals, to induce GnRH pulse generator activity in OVX goats (103). In the male rhesus monkey, an absence of expression of SP in kisspeptin neurons has been reported but SP fibers are observed in close apposition on ARC kisspeptin perikarya (126). In the ewe, only a small proportion of ARC kiss neurons contain tacr1 or SP expression (120).
Recently, results in rat showed that icv injection of SP antagonizes the inhibitory effect of the mammalian ortholog of gonadotropin-inhibitory hormone (GnIH), RFRP-3, on the expression of hypothalamic gnrh and kiss1 (123). These first data suggest that SP may act at different levels of the central control of HPG (GnRH, kisspeptin, or GnIH). More studies in mammals and in other vertebrates should be performed to see whether this effect of SP on GnIH action is conserved throughout evolution.
Teleosts
In zebrafish, direct actions of TAC1 peptides are more likely on GnRH than on Kiss neurons, as associations between TAC1immunoreactive processes and neurons for GnRH3 (the hypophysiotropic GnRH form in this species) in the ventral telencephalic area are observed, while there is no apparent proximity of TAC1 processes to kiss2 mRNA-expressing neurons in the hypothalamus (60). Recently, expression of tacr1a mRNA was reported in several brain regions containing GnRH3, as well as Kiss2, cells such as olfactory bulb, preoptic area, and hypothalamus (127), leading the authors to suggest that TAC1 peptides may act on both neurons. To the best of our knowledge, no functional study has been yet performed to investigate the action of TAC1 peptides on GnRH production in teleosts.
Mammals
A direct action of SP and NKA at the pituitary level is possible as expression of their receptors has been detected in pituitary cells, and sometimes specifically on gonadotrophs, in mammals [tacr1 (128)(129)(130); tacr2 (131)]. Moreover, SP and NKA fibers have been reported surrounding hypophyseal blood capillary vessels in the median eminence and these peptides to be present in the pituitary of mammals [NKA, rat (132); SP, rat (133-136); SP, NKA, rat (137); SP, rhesus monkey (126, 138)]. The first reported study on the in vitro effect of tachykinin on gonadotropins dates back to 1974. Fisher and colleagues observed that SP induced release of LH and FSH by pituitaries, from intact rats, cultured in vitro (139). These preliminary data using few pituitaries and a high dose of SP were followed by a contradictory one, which reported no effect of SP on LH and FSH release by hemi-pituitaries of OVX rats (140). These first data already point out the potential importance of sex steroid on SP action in the control of gonadotropins. A stimulatory effect of SP on LH release from anterior pituitary cells in culture is reported during the peripubertal period in male and female rats but not at the prepubertal age and long after maturation (141). In vitro perifusion of anterior pituitaries from female rats allowed to show an inhibition of GnRH-induced LH release by SP, an effect abolished by the use of the TACR1 antagonist (142). All these results obtained in rat highlight the sex steroid dependence of the in vitro effect of SP on gonadotropins in this species. Data are also available in another mammal, the pig. In cultured porcine gonadotrophs, SP was reported to stimulate LH release without affecting intracellular LH content (143). It also potentiated GnRHstimulated LH release and reversed the GnRH-induced decrease of gonadotroph LH stores, these effects not being blocked by the use of the GnRH-receptor antagonist. A few years later, the same group demonstrated that the SP direct effect on pig gonadotrophs and LH release was extracellular Ca 2+ -dependent and did not involve an effect on lhb transcript levels (144). Altogether, these in vitro data in mammals converge toward a predominantly stimulatory effect of SP on LH release ( Figure 2). Concerning NKA, few data are available. Incubation of hemi-pituitaries with NPK and NPg stimulates LH release in intact male rats but is not significant in castrated animals, while no significant effect is seen with NKA in both situations (145).
In vivo data concerning SP and NKA action on gonadotropins in mammals have been reviewed by Fergani and Navarro (97), and the following text will give a summary of them, complemented with one more recent publication (122). These in vivo studies, which cannot discriminate between direct and indirect effects, report diverse effects of TAC1 peptides on LH and FSH release, mostly depending on the presence or absence of sex steroids and also likely due to species differences [for review (97)]. Most of these studies reported a stimulatory effect of SP on LH release [OVX+E2 female rat (146); prepubertal female and male rats (147); normal men (148); intact female rabbit (149); OVX+E2 female mouse and intact male mouse (115); prepubertal female mouse (109); intact ewe (120)]. However, an absence of effect was sometimes observed [intact adult male rat (145); intact adult female rat (147); OVX female rat (145,146); castrated male monkey (126); OVX or OVX+E2 female goat (103)]. Even inhibition was reported in castrated male rats (145,150). Intravenous (iv) injections of SP failed to induce LH release in the rhesus monkey (126). However, in the cynomolgus monkey, there is a reduction in the duration and amplitude of LH surge after intragastric administrations of the TACR1 antagonist (151). In the ewe, much higher doses of SP compared with NKB are needed to stimulate LH release (120). Similarly, for NKA, various effects were obtained depending on the "sex steroid status" of the animals. Stimulation of LH release was observed in intact mature animals [male rat (145); OVX+E2 female mouse and intact male mouse (115, 147)] and in prepubertal intact male and female rats (147), while inhibition was reported in castrated animals [male rat (145); female rat (145); female mouse (115)]. An absence of effect was also sometimes seen by some authors in OVX adult female rats (121). Recently, the stimulatory action of NKA in the presence of sex steroids during adulthood in female mice was reported to be NKB-independent, as NKA was able to induce LH release in NKB-deficient mice (tac3KO mice) or after blockade of TACR3 by a specific antagonist (122). In addition, the stimulatory effect of NKA was kisspeptin-dependent, as it was absent in Kiss1KO mice (122). Interestingly, the inhibitory action of NKA on LH release in the absence of sex steroids during adulthood in female mice was found to be NKB-and dynorphin-dependent (122). In the ewe, much higher doses of NKA are needed to stimulate LH release, compared with NKB (120). In the female goat, the NKA agonist was inefficient to induce LH release either in OVX or OVX+E2 animals (103). NPK was also shown to modulate gonadotropin as icv injection of the peptide produced a suppression of LH release in ovariectomized rats (145).
A few studies addressed the effects of TAC1 peptides on FSH release in rodents. A sex difference was obtained in prepubertal rats: acute administration of the TACR1 agonist stimulated it only in females, while it was the TACR2 agonist that was able to induce it in males (147). In intact adult rats, the TACR1 agonist had no effect on FSH release in both sexes and the TACR2 one could elevate it in females (147). SP was ineffective in stimulating FSH release in OVX+E2 rats (146). In the intact male mice, central injection of TACR1 or TACR2 agonists induced an elevation of FSH secretion (115).
Teleosts
A direct action of tachykinins at the pituitary level is also possible in teleosts as SP fibers directly innervate the pituitary (58, 152, 153). Moreover, TACR1 expression has been detected in the pituitary [zebrafish (71) To date, only one study investigated the direct effects of peptides encoded by the tac1 gene on teleost pituitary hormone expression and release [ Figure 2 (61)]. Using primary culture of prepubertal grass carp pituitary cells, Hu and collaborators showed that grass carp SP and NKA could elevate prolactin (prl) and somatolactin-a (sla) mRNAs and hormone secretion, without any effect on proopiomelanocortin (pomc), fshb, thyrotropin b (tshb), glycoprotein a-subunit (gp-a), growth hormone (gh), and slb expression. For LH, SP but not NKA could induce a dose-dependent inhibition of lhb mRNA levels (after 24 h of treatment), while both peptides induced a dosedependent stimulation of LH release with a lower potency and efficacy for NKA (after 3 h). This induction of LH release by SP and NKA and the inhibition of lhb mRNA by SP were blocked by the use of the TACR1 antagonist but not of TACR2 or TACR3 antagonists, in agreement with the fact that TACR1 was the only form of TACRs detected in grass carp gonadotrophs. Moreover, SP was able to partially suppress GnRH induction of lhb mRNA, while co-treatment with the TACR1 antagonist enhanced this induction. TAC1 peptides in grass carp can thus have differential effects on LH release and lhb mRNA levels via activation of TACR1 in gonadotrophs. More studies in other teleost species are needed to decipher whether these actions are species-specific or common to all teleosts.
Inactivation of the TAC3 system and reproduction
In 2003, Pintado and collaborators injected intraperitoneally an antagonist of TACR3, the preferential receptor for TAC3 peptides, to 8-week-old female rats and showed no effect on reproductive success or litter size, while a 6-month-old subcutaneous injection of the same antagonist resulted in a reduction in the litter size (113). Later, mutations in the tac3 or tacr3 genes were characterized which lead to hypogonadotropic hypogonadism in human [(154-159); for review (160)], which could be reversed in adulthood (157). Similarly, in mice, tac3 (161) or tacr3 (162) null females show delayed sexual maturation and abnormal estrous cyclicity, which recover in adulthood leading to fertility, although they produced fewer pups per liter. In contrast, timing of sexual maturation and fertility are preserved in tac3 (161) or tacr3 (162) null males. In a teleost, the zebrafish, the knockout of either tac3a, tac3b, or both does not disrupt the reproduction (spermatogenesis and folliculogenesis are not impaired) (163). The impact of tac3 gene mutation should be now studied in other teleost species, as knockout studies of reproductive genes such as the different types of gnrh [gnrh3 (164); gnrh2 (165)] and kiss [kiss1, kiss2, kissr1, and kissr2 (166)] system genes all generate zebrafish with normal gametogenesis, suggesting that this species may have a high compensatory mechanism [for reviews (167,168)] and may not reflect the situation observed in all teleosts.
Mammals
In mammals, the first analysis of tacr3 expression was performed in rodents and showed its presence in GnRH neurons, indicating a possible direct effect of TAC3 on GnRH [rat (169); mouse (170)]. However, an absence of direct regulation of GnRH release by NKB was demonstrated using hypothalamic explants from adult male mice (171). In addition, in the same in vitro system, NKB was able to completely abolish the stimulation of GnRH release induced by kisspeptin (171). These results in mice suggested that NKB could regulate GnRH only via an action on kisspeptin (Figure 2), which is in agreement with the more recent demonstration of a minimal expression of tacr3 in GnRH neurons, but its expression on virtually all KNDy neurons, in this species (115, 172). In female sheep, no TACR3 immunoreactivity was revealed in GnRH neurons, but GnRH neurons and fibers were in proximity to NK3R-containing ones (173). Use of the immortalized GT1-7 cell line, which represents mature post-migratory GnRH neurons with expression of TACR3, allowed to show differential effects on GnRH release depending on the length of exposure: acute treatment with NKB increases GnRH secretion, while long-term treatment decreases it by repressing transcription (174). In the arcuate nucleus of the hypothalamus, TAC3 is co-expressed with kisspeptin and dynorphin in the socalled KNDy neurons [for review (98)]. This was first demonstrated in sheep (96). KNDy neurons project to GnRH neurons and positively regulate their activity, being responsible for the generation of GnRH pulsatility in the hypothalamus of mammals [for reviews (97,100,175)]. Ablation of these neurons in female rats induces hypogonadotropic hypogonadism (176). Most of the studies show that neurons of ARC and particularly KNDy neurons project to the axonic terminals of GnRH neurons (98). Therefore, it is possible to assume that KNDy neurons might act on those terminals in a direct manner or using intermediate neurons to regulate the GnRH release (175). The most accepted hypothesis for mammalian KNDy neurons proposes that TAC3 acts in a positive manner and that Dyn acts in a negative way on the pulsatile release of kisspeptin by KNDy neurons (172, 177). In the ewe also, a high percentage of kisspeptin neurons produces dynorphin and NKB (96). A recent review addresses the question whether the KNDy model for the control of GnRH pulses applies to humans and other primates, compiling data showing that colocalization of kisspeptin and NKB is also observed in rhesus monkeys and humans (178). In addition, the ability of kisspeptin to induce LH release in patients with mutations in TAC or TACR3 tends also toward a proximal action of NKB to kisspeptin in stimulating GnRH secretion (179).
In vivo studies in different mammals have shown the stimulatory effect of NKB on GnRH secretion [prepubertal and pubertal rhesus monkeys: female (180) and male (181); ewe (182); goat (103,168)]. Electrophysiological studies showed that icv administration of the TACR3 agonist (senktide) suppressed the GnRH pulse generator in OVX rats (183), while it induced GnRH release in intact male mice (184).
Teleosts
In tilapia, Mizrahi and colleagues investigated the coexpression of the three different forms of GnRH present in this species with TACR3s. They show that GnRH3 neurons expressed tac3ra, but not tac3rb, while the contrary was observed for GnRH2 neurons, and GnRH1 ones expressed both tacr3 (185). In the striped bass, TAC3 peptides, NKB, and NKBRP (NKF in the article) have an inconsistent effect (no or stimulatory only at the highest dose) on gnrh1 expression by brain slices in culture, while persistently downregulating kiss2 expression (74) (Figure 2).
In contrast to human and rodents, the expressions of tachykinins and kisspeptins are not always expressed in the same neurons in teleosts. In the zebrafish, NKB/TAC3 and NKF/ NKBRP/TAC3RP are expressed in the nuclear lateralis tuberis (NLT), which is the teleost homologous structure to the ARC (71), but kiss2 expression has not been found in this area (186). This finding in zebrafish, however, does not exclude that TAC neurons project on kisspeptin ones. In the striped bass, NKB neurons innervate the largest kiss2 neuronal population in the hypothalamus, which also expresses TACR3, while no expression of TACR3 or no NKB neuronal projection is detected for GnRH1 soma (74). In addition, in this species, TAC3 peptides, NKB and NKBRP (NKF in the article), are able to downregulate kiss2 gene expression in vivo, while having no effect on gnrh1 expression (74). In addition, cotreatment with a NK3R antagonist abolishes the negative effect of TAC3 peptides on kiss2 mRNA levels (74). These results in the striped bass suggest that tachykinin peptides may act preferentially on the kisspeptin system, as in mammals. When injected to goldfish females in mid-vitellogenesis and males in late-spermatogenesis, three NKB peptides (NKBa-13, NKBa-11, and NKBb-13), but not the fourth one (NKBb-11), increase hypothalamic gnrh3 mRNA levels (73). NKBa-10 and NKBa-13 ip injected to goldfish females at the early vitellogenic oocyte stage and males at the early spermatogenesis stage decrease mRNA levels of both hypothalamic kiss2 and gnrh3 (except NKBa-13 on gnrh3) (88). In tilapia, NKBRP injected to mature male tilapia inhibits the expression of brain gnrh-I and kiss2, while NKB has no effect (187). A recent study in the Japanese eel, Anguilla japonica, reports that ip injection of each of the four mature peptides found in this species gives different effects depending on the peptides and the doses used: a low dose of the four peptides had no effect on neither gnrh1 (mgnrh) and gnrh2 (cgnrh) expression, while a high dose of NKBa-10 and NKBb-13 (and not NKBa-13 and NKBb-10) stimulates gnrh1 expression (188). All these data demonstrate that different regulations of gnrh and kiss expression by TAC3 peptides (from none to stimulatory or inhibitory effects) may be encountered among teleosts, depending on the species, the maturity stage, the doses, and the peptides tested.
The non-systematic action of TAC3 peptides on GnRH and kisspeptin in teleosts, compared with the situation observed in mammals, is likely due to the surprisingly non-essential character of these two neuropeptides for reproduction in some teleost species. Indeed, recent knockout studies demonstrated that gnrh3 and gnrh2 in zebrafish (164,189), gnrh1 in male medaka (190), and kiss1 and kiss2 in zebrafish (166) and medaka (191) were dispensable for normal reproductive function. In zebrafish, even triple mutants for gnrh3, kiss1, and kiss2 undergo normal puberty and gonad maturation (192). This lack of effect on reproduction of GnRH and kisspeptin gene editing led many scientists to make assumptions on the possibility of physiological compensatory phenomena in teleosts (for reviews : 167, 193-196). In addition to these knockout results, many data are available, stating reproductive actions of GnRH (for review: 194) and kisspeptins at various HPG levels in teleosts (for reviews : 195, 196 (200)]. To the best of our knowledge, only one in vitro study has investigated the potential direct effect of NKB on gonadotropins in mammals, by using a gonadotroph cell line (201). The authors reported no effect of NKB on lhb and fshb mRNA expression, even if TACR3 was detected in this cell line.
Comparing various tachykinins in vivo, Sahu and Kalra (121) were the first to report that NKB-containing implants, in the third ventricle of OVX rat brain, did not induce any change in LH release. Later, this absence of NKB effect on LH was also shown after either ip or icv administration to intact adult male mice (171). However, evidence for stimulatory effects of NKB on LH has since been documented in many mammalian species [for reviews (97,106,178)], as for example in prepubertal female rats (16,202). In some studies, the stimulatory effect of NKB on LH is only observed under physiological sex steroid levels (i.e., intact or OVX+E2 adult animals) [adult male and female mice (115, 203); adult male and female rats (203, 204); lactating female cattle (205)]. In contrast, in monkeys, NKB is able to stimulate LH release in castrated juvenile (200, 206) and adult (207) males. In the sheep, castration does not prevent the stimulatory action of NKB in adult females (207,208), as compared with intact females [adult (108,209); prepubertal (210)]. A similar situation is observed in the female goat with a stimulatory effect of the iv administered NKB agonist (senktide) (103) or no effect of the icv injected NKB (177), regardless of the gonadal status. Recently, iv administration of senktide has even been shown to be efficient in stimulating LH release in fetal male and female sheep (211). In humans, early studies report no gonadotropin-stimulating effect of NKB iv administered in adult men and women (212, 213), but a series of data obtained by Skorupskaite and collaborators using the TACR3 antagonist given orally show a decrease in overall circulating LH levels and LH pulsatility in adult men (214) and women (215)(216)(217). Few studies report an inhibitory action of senktide on LH release, regardless of the steroid milieu in female rat (183,218) or only in the absence of sex steroids in female mice (172).
These various effects on LH and FSH in mammals could be due to species, physiological status, or mode of peptide administration. Table 1 gives details on all these in vivo studies of NKB action on gonadotropins.
Teleosts
TACR3s (tac3ra1, tac3ra2, and/or tac3rb) are expressed at the pituitary level in various teleosts [zebrafish (26, 71); spotted sea bass (78); grass carp (62, 89, 221, 222)] and in both LH and FSH cells in tilapia (72), making a direct effect of NKB and NKBRP possible on gonadotropin synthesis and release. Interestingly, in the grass carp, Xu and collaborators reported that tacr3a is expressed in both LH and somatolactin a (SLa) cells, while tacr3b expression is only found in SLa cells (89). FSH cells were not investigated in this study (89). In tilapia, when NKB and NKBRP are applied to mature male pituitary cells, they both increase FSH and LH release (72). In culture of pituitaries from mixed sexed juveniles of this species, NKBRP downregulates fshb and lhb mRNAs, while NKB has no effect (187). Still in tilapia, Mun and colleagues recently compared responses of pituitary cells and pituitaries to NKB and NKF in males and females. They reported that expressions of fshb and lhb mRNAs did not show any change after treatment of whole pituitaries with NKB or NKF in both sexes (223). In contrast, the use of primary culture of pituitary cells allowed them to find that NKB could stimulate fshb and lhb mRNAs in female and inhibit them in male, while the contrary was observed with NKF (223). These results highlight major differences according to the maturation stage, to the protein or mRNA, or to the type of culture (primary cell culture versus organotypic culture) concerning the effects TAC3 peptides have on gonadotropin in tilapia. The other studies available suggest in addition to species differences, using the same method, primary cultures of pituitary cells (Figure 2). In the striped bass, the effects of NKB and NKBRP were stimulatory on LH and FSH release but absent on their mRNAs (74). In the European eel, the four peptides encoded by the tac3 gene were able to inhibit lhb mRNAs by pituitary cells in culture but had no effect on fshb mRNAs (76). Some other studies, using this cell culture system, reported no effect of NKB and NKBRP on fshb and lhb mRNAs [grass carp (75); orange-spotted grouper (77)].
In teleosts, most of the in vivo data showed an increase of gonadotropin release and expression after treatment with TAC3 peptides. In zebrafish, homologous (zebrafish) NKBa and NKBRP (NKF in the article) induce LH release when injected to mature females (71). Among the four neurokinin B peptides characterized in zebrafish, NKBb presents a modified C-terminal motif from the typical tachykinin FVGLM to FVGLL, thus losing the final methionine. This change leads to a decreased affinity of this peptide for the two TACR3 and a highly reduced in vivo effect, compared with other neurokinin peptides (71,73). However, the finding of a third TACR3 in the zebrafish genome (26) increases the chances that the NKBb with a different C-terminal motif may be active. In tilapia, ip injections of homologous NKB to mature males increase both FSH and LH plasma levels, while homologous NKBRP induces only LH release (72). In mature female tilapia, ip injection of tilapia NKB and NKBRP has no effect on pituitary lhb and fshb mRNA levels (185). More recently, Mizrahi and colleagues developed specific NKB and NKBRP (NKF in the article) analogs based on the structure of the mammalian NKB analog, senktide (185). When ip injected to mature female tilapia, these analogs increase plasma LH levels as native (tilapia) NKB and NKBRP do, and they are even able to increase FSH release while native ones have no effect (185). Concerning mRNA levels, native NKB and NKB analogs are efficient in stimulating lhb and fshb, whereas native NKBRP has no effect and NKBRP (NKF in the article) analog stimulates only lhb (185). When injected to goldfish females at mid-vitellogenesis and males at late spermatogenesis, homologous NKBa-10, NKBa-13, and NKBb-10, but not NKBb-11, increase pituitary lhb mRNA levels (73). An increase in fshb mRNA levels is observed in females only after administration of goldfish NKBa-13 and NKBb-13, while in males all peptides, except NKBb-11, induce these levels (73). In sexually immature goldfish, NKBa-10 and NKBa-13 ip injected to females at the early vitellogenic oocyte stage and males at the early spermatogenesis stage decrease mRNA levels of both pituitary lhb and fshb (88). In the female orange-spotted grouper Epinephelus coioides at early vitellogenic stages, ip injection of NKB increases pituitary lhb, but not fshb, mRNA levels, while administration of NKBRP has no effect on these expressions (77). A recent study in the Japanese eel reports that ip injection of each of the four mature peptides found in this species gives different effects depending on the peptides and the doses used: a high dose of the four peptides inhibits lhb and fshb expression, while a low dose of NKBa-10 and NKBb-13 stimulates them (188). All these data show various effects of TAC3 peptides on gonadotropin release and expression among teleosts, depending on the species, the maturity stage, the doses, and the peptides tested. Table 1 compiles all these in vivo studies of NKB action on gonadotropins.
TAC4 peptides
Little is known on the effects of TAC4 peptides, HK and EK, on the central reproductive brain-pituitary axis. To the best of our knowledge, the only available study was realized in a teleost, the grass carp. Using transcriptomic analysis of TAC4 peptide effects on pituitary cells, Shi and colleagues have recently demonstrated that HK2 downregulates the pituitary expression (Figure 2). In mammals, the known reproductive role of HK1 takes place at the peripheral level (33), while up to now, none is attributed to EKs.
At the peripheral level
Tachykinins can also act at the levels of the gonads and the secondary sex organs, via paracrine and autocrine effects, in both females and males (Figure 2).
Effects on the female reproductive system 3.2.1.1 Mammals
A review has already been dedicated to tachykinin involvement in mammalian ovarian function (224), and the text below is only a summary. Expression of tac1 and tac3 (mRNA), as well as their receptors, is detected in the mammalian ovary, oocytes, and granulosa cells (113, 224), indicating potential autocrine/paracrine effects. Isolated cumulus granulosa cells in mice express tac1, tac3, and tac4 (113). The control of ovarian steroid secretion by tachykinins in mammals has been previously reviewed (224). Briefly, data were obtained in various mammalian species with different results. In rats, exposure of granulosa cell culture or ovarian fragment in culture with SP and SP analog was unable to modify estrogen or progesterone release (225). In hamster, depending on the age of the animals, treatment of ovaries in culture with SP could stimulate (15-day-old hamsters) or inhibit (adult hamsters) or have no effect (neonatal hamsters) on estradiol release (226). Concerning progesterone release, in the same culture system, SP has a stimulatory effect (neonatal and adult hamsters) or no (15day-old hamsters) effects (226). Using luteal cells in culture and exposure to SP, opposite results were obtained on progesterone release in basal conditions or under stimulation with LH in two different artiodactyla/ungulata species: stimulation in bovine (227) and inhibition in pig (228). In pig, SP treatment did not change estradiol release by granulosa cells in culture but stimulated it by luteal cells (228). These various effects of SP on in vitro ovarian sex steroid release in mammals thus likely depend on species, type of cells, and age of animal.
In the rat placenta, downregulation of tac3 and tacr3 expression is associated with pregnancy (41). NKB placental levels are increased at term labor in women (239). In the placenta of preeclampsia women, elevated circulating NKB and increased tac3 expression are reported as compared with placenta of normal pregnant women (41, 70, 240). TAC3 and TACR3 may contribute to preeclampsia during late pregnancy (41, 241). NKB stimulates the expression of gnrh, kiss, and human chorionic gonadotropin (hCG) by primary cultures of rat placental cells (242).
Teleosts
An autocrine/paracrine action is also possible in teleosts as both NKB and NK3R are expressed in the ovary [zebrafish (26, 71); tilapia (72)]. In the zebrafish, a direct effect of neurokinin B on the ovary is reported, as it stimulates estradiol production and increases the expression of cyp11a1 and cyp19a1 in primary cultures of follicular cells (243).
Mammals
The involvement of tachykinins in the regulation of mammalian testicular function has already been reviewed (236, 244, 245), and the text below is a brief summary. SP inhibits testosterone production and release by isolated Leydig cells in hamster (246,247). Tac1, tac3, and tac4 genes are expressed in the human sperm (248). Tachykinins are likely to enhance the sperm motility by TACR1-and TACR2-dependent mechanisms, as TACR1-and TACR2-(but not TACR3-) selective antagonists can reduce the stimulating effect of phosphoramidon in human (248). Human HK1 also promotes progressive sperm motility ( Figure 2) with a potency similar than that of NKA (lower than that of SP and higher than that of NKB) (249). All classical TACRs seem to be involved in these actions, but the role of TACR1 was predominant (249).
Tachykinins also stimulate contractility of the vas deferens and of seminal vesicles [ Figure 2, for review (236)]. SP (TAC1RP) and NKA (TAC1) are present in the prostate of guinea pig and rat at low levels, of dog abundantly and absent in human prostate [for review (236)], while tac1, tac3, and tac4 mRNA expressions have been detected in human prostate (39, 232). Prostate contraction by tachykinins in human involves TACR2 (250).
Teleosts
An autocrine/paracrine action is also possible in teleosts as both NKB and NK3R are expressed in the testis [zebrafish (71); tilapia (72)]. In tilapia, recent use of NKB antagonists by ip injections on adult males reduced the number of spermatozoa, leading to lower fertility (251), an effect which can be direct as male tilapia have significant amounts of TACR3 in the testis (72).
4 The physiological role of tachykinins in the regulation of food intake As described previously in this review, TAC peptides (SP and NKA) were discovered for their contractile role on the gastrointestinal tract (GIT) in mammals. However, far less direct evidence is available concerning the regulation of food intake by tachykinin peptides when compared with their role in reproduction. Nevertheless, recent data, notably in two teleosts, the sea bass Dicentrarchus labrax, and the grass carp, highlight a potential major regulatory role of TAC3 and TAC4 peptides in the regulation of genes involved in feeding and gut motility.
As gut peptides, tachykinins also have a potential direct role on the GIT in vertebrates. A well-described one is the stimulation of its motility, the first necessary step of food digestion after its intake [for review (257)].
Mammals
4.1.1 Energy state and the tachykinin system TAC3 neurons, mainly as part of KNDy neurons, have been involved in the regulation of both negative and positive energy balance. For example, they mediate the anorexigenic effect of estradiol in young female rats, as selective ablation of KNDy neurons suppresses the post-ovariectomy weight gain (258). In addition, many studies report the regulation of the tachykinin system by a change of energy status. Fasting and caloric restriction (CR) induce a decrease in hypothalamic ARC tac3 and/or tacr3 expression in rodents [pubertal female rat (259); OVX female mice (260); adult female rat (261)]. Nevertheless, in adult male mice, fast increases hypothalamic ARC tac3 and tacr3 (171) while CR and fast (261,262) have no effect in adult female rats. In sheep, chronic food restriction downregulates kiss and tac3 mRNA levels [castrated male sheep (263); OVX ewe lambs (264); for review (265)]. Feeding a high-fat diet does not change ARC mRNA levels for tac3 in pubertal female mice (266) while it has a stimulatory effect in pubertal female rats (267). All these data suggest differential regulation of the tachykinin system (and of its involvement in the control of metabolism) according to species, sexual maturation stage, and/or degree of negative energy balance.
Effects of tachykinins on food intake and GIT motility
Abundant distribution of TAC1, TAC2, and TAC3 receptors is found in the hypothalamic nuclei involved in the control of food intake such as ARC, paraventricular nucleus (PVN), and lateral hypothalamus (LHA) [for reviews (4, 97,106)]. The full tachykinin system is also detected in the neurons and nerve fibers of the mammalian gut with remarkable diversity between species [for reviews (268)(269)(270)]. These distributions point out toward a potential involvement of TAC peptides in the control of feeding and GIT motility.
Administration of NPK to food-deprived rats for 24 h delays the onset of (re)feeding and decreases the cumulative food intake [(271); for review (272)]. Achapu and coworkers show that the inhibition of food intake induced by centrally injected NPK may be due to the intense grooming induced by the injection (273). Similarly, icv injection of SP to food-deprived male rats suppresses food intake, but an increase of locomotor activity is also observed (274). The fact that icv injections of NKA induce an increase in pomc mRNA levels in the rat ARC (275) also argues toward such anorexigenic action of tachykinins in mammals or at least in rodents. However, later, Karagiannides and collaborators consider SP as a novel antiobesity target after showing that the blockade of SP signaling by mean of an TACR1 antagonist leads to a decrease of food intake and body weight in two obese mouse models, an HFD-induced one and a leptin-deficient (ob/ob) one (276). They also report that peripheral injection of SP increases food intake and induces upregulation of hypothalamic npy as well as downregulation of pomc and mRNA levels (276). In male rats, ghrelin negatively regulates the tac1 gene in the hypothalamus and acute icv injection of NPK and NPg (but not SP nor NKA) reduces food intake (277). In addition, in male mice, the hyperphagic effect of peripheral injection of ghrelin disappears in tac1KO animals, suggesting the tac1 requirement in the control of food intake by ghrelin in rodents (277). The first study using tac1-null mice does not show any difference in body size compared with controls (276), while a more recent one reports that these animals have a significantly lower body weight during adulthood and also show increased hypothalamic pomc expression and reduced food intake (278). All these data indicate that, in rodents, TAC1 peptides may function as either endogenous anorexigenic or orexigenic peptides.
The capacity of SP, NKA, and NKB to induce intestine contraction was one of the actions that led to their discoveries (refer to part 1.1. of this review). Their action on motility is observed in all parts of the gut through the tachykinin receptors and has been previously reviewed (279,280).
Energy state and the tachykinin system
In goldfish, a short-term postprandial increase in tac1 mRNA levels (g-PPT in the article) has been reported in both the hypothalamus and the olfactory bulbs (281). In zebrafish, fasting increases the brain expression of tac3 in females (282). In grass carp, food intake can significantly induce hypothalamic tac3a and tac3b mRNA expression (222). Thus, in teleosts, change in energy state may have positive and negative effects on the TAC system.
Effects of tachykinins on food intake and GIT motility
Immunohistochemical studies report high concentrations of tachykinins and their receptors in teleost hypothalamic areas involved in the control of food intake [SP in goldfish (283); SP in sea bass (152); carassin in goldfish (284); TACR1 and TACR3 in A B FIGURE 3 Direct effects of TAC peptides in the control of food intake and gut motility in mammals and teleosts. TAC peptides can act directly at different levels (hypothalamus, pituitary, and gastrointestinal tract) to influence food intake and gut motility. In mammals (A), most of the available studies report the stimulatory effects of tachykinins on gut motility. In teleosts (B), recent in vitro studies are emerging, showing direct effects of TAC3 and TAC4 peptides on the expression of neuropeptides highly expressed in the pituitary and that are involved in the central control of food intake. These TAC peptides can also influence the expression of genes from the gut that control its motility. (73,88); orange-spotted grouper (77); spotted sea bass (78); tongue sole (79)]. While tachykinins and their receptors are expressed in the teleost hypothalamus and its nuclei involved in the control of food intake, to our knowledge, no data have yet shown their direct effect on the expression of neuropeptides such as pomc, npy, or agrp at the brain level. Due to the direct innervation of pituitary cells by hypophysiotropic neurons in teleosts, Hu and colleagues demonstrated a high expression, in the brain and pituitary, of neuropeptides involved in the regulation of feeding (221). They subsequently reported that TAC3 and TAC4 peptides could change the expression of some of these genes, using transcriptomic analysis of TAC peptide effects on grass carp pituitary cells (62, 222). NKB can induce in vitro the expression levels of urotensin 1 (UTS1), cocaine-and-amphetamine-regulated transcript 2 precursor (CART2), proopiomelanocortin b (pomcb), and neuromedin B1 (NMB1) mRNA, all four anorexigenic peptides, an effect that is also reported in vivo after ip injection (222). HK2, a peptide encoded by the tac4 gene in grass carp, upregulates CART2, CART3, and CART5, peptide YY2 (PYY2), UTS1, and NMB1 expression, while downregulating type 2 neuropeptide Y receptor (NPY2R) expression (62). Thus, NKB and HK2 inhibit the expression of the orexigenic pathway (such as NPY one) and stimulate anorexigenic peptides, playing roles of satiety factors (Figure 3). In grass carp, TACR3b mediates this role, while TACR3a modulates NKB action on reproduction (222), indicating a typical case of subfunctionalization where paralogs share initial pleiotropic functions. Tachykinins have also been involved in live prey food preference in hybrid Siniperca chuatsi × Siniperca scherzeri mandarin fish as tac 1 expression is higher in feeders compared with non-feeders in this species (286).
Peripheral action with direct contraction of the smooth gut muscle has also been described in teleosts [for reviews (10,287)]. Substance P stimulates the motility of isolated intestine or stomach in a variety of fish [Pleuronectes platessa, Labrus bergylta, Gadus species, Lophius species, Anguilla species (55); rainbow trout (288, 289); cod Gadus morhua (290, 291); common carp Cyprinus carpio (292); bichir Polypterus senegalensis (293)]. In many of these species, it is demonstrated that the effect of SP is in part direct (cotreatment with tetrodotoxin, a sodium channel blocker) and in part via stimulation of cholinergic and serotonergic neurons (cotreatment with cholinergic or serotonergic antagonists, atropine and methysergide) [ Figure 3, common carp (294):; rainbow trout (288, 289,293)]. NKA is also able to stimulate the motility of isolated trout intestinal muscle and the vascularly perfused trout stomach, but with less efficiency than SP (289). This stimulatory control of gut motility by the tachykinin system takes place at an early stage in development as NKA modulates zebrafish larval gut before or around the time for the onset of feeding (295). More recently, an increase in the expression of tac1 has been detected by RNA-seq in the giant grouper Epinephelus lanceolatus at the onset of feeding (296). Using in vitro stomach and intestine incubation assays in the sea bass, Zhang and collaborators showed that NKB peptides may modulate the expression of hormones (78) (Figure 3), known to have stimulatory activity on GIT motility in vertebrates, such as motilin and ghrelin [for review (297)]. In the stomach, NKBa-13 and NKBb-13 stimulate gastrin mRNA levels, while NKB-10 peptides have no effect (78). NKBb-13 can stimulate stomachal motilin and ghrelin expression, while the other three NKB peptides have no effect (78). In the intestine, NKBa-13, NKBa-10, and NKBb-13 stimulate cholecystokinin mRNA levels, while NKBb-10 has no effect (78). Only NKBb-10 can stimulate intestinal gastrin expression and NKBa-10 motilin expression. None of the four NKB peptides can change ghrelin mRNA levels in the intestine (78).
Conclusions and perspectives
Cumulating evidence places the tachykinin system with not only NKB (TAC3) but also other tachykinin peptides, SP (TAC1) and NKA (TAC1), as major stimulatory actors in the control of reproductive function, in mammals. In teleosts, the two TAC3 peptides NKB and NKBRP and their 3R duplicates can have various effects (stimulatory, inhibitory, or none) mainly according to the species, the maturity stage, and the peptide tested. These sometimes opposite effects of TAC3 peptides on reproductive genes among teleost species are also reported for other neuropeptides involved in the control of the HPG axis, such as kisspeptin and gonadotropin-inhibitory factor (GnIH) (for review: 9). One may take into consideration the variety of reproductive strategies and life cycles among these more than 25,000 different species to try to find explanations, as well as the physiological compensations between neuropeptides that are likely to exist in teleosts, perhaps due to the anatomical direct innervation of pituitary cells and the existence of various 3R paralogs. Concerning potential involvement of TAC1 and TAC4 peptides in teleost reproduction, too few data are available to draw any conclusion. Past studies in mammals and recent ones in teleosts suggest that the tachykinin system may also be involved in the regulation of food intake and metabolism. Even if more studies are still needed, especially concerning the role of TAC4 peptides, it looks like the control of food intake may be taken over by TAC1 peptides in mammals but also by TAC3 and 4 peptides in teleosts. Tachykinins and their receptors thus seem to be part of networks linking metabolism and reproduction and involving central and peripheral hormones, such as kisspeptin, leptin, and ghrelin.
Due to the multiple whole-genome duplication events that occurred in vertebrates, phenomena of divergence and subfunctionalization or neofunctionalization of the ancestral functions are expected and observed, especially in teleosts. Therefore, an analysis of the tachykinin system is recommended for each organism of interest in order to obtain a clear view of the function of this family of peptides and receptors according to vertebrate species.
Future studies should aim at stating whether or not KNDy neurons exist in some teleost species. Too few species have been considered so far. Surprisingly, up to now, no study has yet investigated the possible direct effects of TAC3 peptides on pituitary LH and FSH cells in mammals; this should be performed in the future using primary cultures of pituitary cells from different mammalian species. It would also be interesting to study whether TAC peptides could directly modify the expression of central actors involved in the control of food intake, such as pomc and npy, in both mammals and teleosts. Future directions on the study of the tachykinin system should also include investigations on Mas-related GPCRs (Mrgprs), as TACR1 antagonists have off-target activity on them (298) and substance P recruits these receptors in immune cells to release cytokine contributing to inflammatory pain in mice (299, 300). For example, characterizing Mrgprs in teleosts and knowing whether they are present in the HPG tissues will help to decipher their potential involvement in the reproductive role of the tachykinin system.
Author contributions
All authors contributed to the article and approved the submitted version. | 2022-12-16T14:15:10.293Z | 2022-12-16T00:00:00.000 | {
"year": 2022,
"sha1": "ce2c8146b29742d3ba5eacbb3685d051b6deee05",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Frontier",
"pdf_hash": "ce2c8146b29742d3ba5eacbb3685d051b6deee05",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": []
} |
248268408 | pes2o/s2orc | v3-fos-license | Ulcerative Colitis: Novel Epithelial Insights Provided by Single Cell RNA Sequencing
Ulcerative Colitis (UC) is a chronic inflammatory disease of the intestinal tract for which a definitive etiology is yet unknown. Both genetic and environmental factors have been implicated in the development of UC. Recently, single cell RNA sequencing (scRNA-seq) technology revealed cell subpopulations contributing to the pathogenesis of UC and brought new insight into the pathways that connect genome to pathology. This review describes key scRNA-seq findings in two major studies by Broad Institute and University of Oxford, investigating the transcriptomic landscape of epithelial cells in UC. We focus on five major findings: (1) the identification of BEST4 + cells, (2) colonic microfold (M) cells, (3) detailed comparison of the transcriptomes of goblet cells, and (4) colonocytes and (5) stem cells in health and disease. In analyzing the two studies, we identify the commonalities and differences in methodologies, results, and conclusions, offering possible explanations, and validated several cell cluster markers. In systematizing the results, we hope to offer a framework that the broad scientific GI community and GI clinicians can use to replicate or corroborate the extensive new findings that RNA-seq offers.
INTRODUCTION
Ulcerative Colitis (UC) is a chronic and debilitating inflammatory disease of the colon, and is a distinct condition within a broad group of pathologies termed inflammatory bowel disease (IBD), which includes Crohn's disease (CD) and indeterminate colitis. The etiology of UC remains uncertain, but increasing evidence implicates complex genetic and environmental contributions. Early genomic and immunologic studies were crucial to identifying therapeutic targets and avoiding surgical intervention. Development of infliximab and other monoclonal antibodies against TNFα, and the biologics targeting cytokines (IL-12 or IL-23) that significantly reduce inflammation and induce macroscopic/endoscopic healing, were major breakthroughs in the management of UC (1).
With the development of biologic therapies, the goal of UC treatment extended beyond symptomatic improvement to include both histologic and molecular healing. Several observational studies have shown that histological remission, including epithelial tissue restoration, is associated with lower rates of disease-related complications such as hospitalization, corticosteroid use, and colectomy compared to either resolution of symptoms or endoscopic improvement alone (5)(6)(7). Recognizing that histologic improvements are now included in UC clinical trials of multiple novel therapeutic agents (8), markers of active disease, therapeutic response and remission are urgently needed.
There are numerous UC-related pathologies that contribute to inadequate epithelial maintenance and regeneration. Cell turnover is increased due to autophagy and apoptosis, yet colonocyte differentiation is reduced (9)(10)(11). Barrier defects manifest as abnormal glycosylation and sulfation of mucins, loss of the protective mucus layer, and increased tight junction permeability (12)(13)(14)(15)(16). Microbial dysbiosis also contributes to altered epithelial function, in which reduced butyrate oxidation yields energy-deficient epithelium, and increased exposure to microbial signals stimulates inflammatory cytokine secretion (17)(18)(19)(20). It is not clear if the many recognized pathologies initiate the disease or are consequences that fuel further development. Overall, the spectrum of pathologies highlights the challenges in defining molecular markers of health, UC damage, and epithelial restoration.
Single cell RNA sequencing (scRNA-seq) allows for in-depth transcriptional characterization of epithelial pathologies in UC and identification of possible pharmacologically fruitful targets. In this review, we discuss findings from two independent scRNAseq datasets generated by the Broad Institute (21) and the University of Oxford (22) that compared subsets of epithelial cell types in healthy and UC colonic tissue. Our goal is to highlight novel markers for epithelial UC disease that, when further validated, may lead to identification of pharmacologic targets to promote epithelial healing in UC patients.
EPITHELIAL CELL DIVERSITY REQUIRES SINGLE CELL RESOLUTION TO CAPTURE DISEASE MARKERS
The heterogeneity of cells that compose the normal colonic epithelium originates from a common intestinal stem cell through the actions of multiple niche factors. As cells migrate from the crypt to surface epithelium, they mature into specialized cells for molecular transport, sensing, secretion, and barrier formation. Each epithelial cell type is critical to sustain organ function, and perturbations in the proportion or phenotype of a cell lineage can contribute to disease. Thus, approaches to study epithelial cell identity and function must distinguish among unique populations within the epithelium.
Transcript analyses using mRNA microarrays that preceded RNA-seq analysis allowed for the simultaneous detection of thousands of genes at one time (23) and was essential to demonstrate the presence of a persistent inflammatory state in patients with histologic remission (24,25). A significant disadvantage of the microarray is that they require prior knowledge of the gene sequence under investigation. In contrast, RNA-seq has a wider dynamic range, allows for sequencing of transcripts without prior information, and can detect gene fusions, indels, and single nucleotide polymorphisms (26). scRNA-seq captures detailed molecular snapshots of individual cells that can be compared between healthy and diseased states to identify altered cell populations within nonhomogeneous tissues. scRNA-seq quantifies the expression of each individual gene at a single cell level, offering high sensitivity for low abundance targets that may be masked in a bulk sequencing approach. This technology unveils the heterogeneity within major cell types, describes cell clusters that contribute to the pathogenesis of disease, and allows for a detailed description of the pathways that connect genome and transcriptome to pathology (27). scRNA-seq might be ideal to understand epithelial pathologies associated with UC in relation to healthy cells.
The technical aspects of scRNA-seq are described in great detail in several excellent reviews (28)(29)(30). Typical elements of scRNA-seq analysis include single-cell dissociation, isolation, library construction, and single cell sequencing (27,31,32). In general, the datasets generated by this technique are large and highly complex requiring robust bioinformatic analysis to be fully interpretable (29,33).
The complexity of genetics and environmental factors that fuel UC make it a particularly good candidate for investigation through single cell transcriptomics. scRNA-seq potentially allows for detailed molecular snapshots of the different cell types composing the intestinal mucosa in health and UC. In particular, this technology facilitated the identification of previously unrecognized cell types and their potential transcriptional shift in disease that correlated to disease severity. These new findings could have direct implications in the management of UC. By stratifying patients on the basis of underlying transcriptional and genetic variability, scRNA-seq opens new ways for the development of targeted therapies.
A comparison of parameters and outcomes in the Broad and Oxford scRNA-seq studies is highlighted in Table 1. Notably, the number of subjects sampled and the epithelial cell types, or clusters identified by the two studies, differs. We focus on cell clusters that exhibited most dramatic transcriptional changes with potential relevance to inflammation, host stress response, immune regulation, and epithelial regeneration: BEST4 + cells, colonic M-like cells, goblet cells, colonocytes and stem cells.
BEST4 + CELLS
Both data sets discovered a novel differentiated absorptive surface colonocyte cluster characterized by expression of a Ca2 +sensitive chloride channel (BEST4) and proton-conducting ion channel (OTOP2) proposed to regulate luminal pH. These cells were also enriched in the paracrine hormones guanylin (GUCA2A) and uroguanylin (GUCA2B), which can stimulate fluid and electrolyte transport through action on the receptor guanylate cyclase. Among 46 transcripts significantly elevated in healthy BEST4 + cells in the Broad study, 29 were also detected as significantly elevated in the Broad study ( Figure 1). However, several transcripts were not exclusive to BEST4 + cells and thus do not serve as unique markers. We performed STRING interactome analysis (human Protein-Protein Interaction Networks and Functional Enrichment Analysis using 1 ) of the characteristic transcripts of BEST4 cells which confirmed across both studies that in healthy tissue, BEST4 cell function includes metallothionein expression (HSA-5661231), guanylate cyclase/dehydratase activity (GO:0030250), and proton channel activity (GO:0015252). Surprisingly, proteomic analysis of these BEST4 cells, conducted in the Oxford study, identified several highly upregulated proteins as BAG1 and APOO. These transcripts were not identified as "characteristic" for this cell cluster in Oxford scRNA-seq dataset (the exceptions were BEST4 and CTSE proteins which were only slightly upregulated in comparison to the rest of epithelial cells). These data indicate that further validation at the protein levels are necessary to reliably identify these BEST4 + cells in tissue and better define their role in colonic physiology. 1 https://string-db.org/ UC-associated changes in the BEST4 + cell transcriptome did not completely agree between the studies. The Broad study found the number of BEST4 + cells were substantially reduced in UC, and 16 significantly upregulated transcripts (log 2 FC ≥ 4.0; Figure 1) were putative disease-associated BEST4 cell markers. In contrast, the Oxford study observed stable numbers of BEST4 + cells and listed 28 upregulated transcripts (fold increase for each transcript was not provided) of which only two (CCL20 and S100P) overlapped with the Broad results (Figure 1). MT1E was the only transcript elevated in BEST4 + cells in both studies regardless of disease state. Other UC-associated BEST4 + upregulated transcripts, such as MT1H and MUC1 (Oxford) or TNFAIP2 and NCEH1 (Broad), did not correlate between the studies. Results of the Oxford study suggest that BEST4 + cells maintain host protective functions in UC [metallothioneins (HSA-5661231), antibacterial (HSA-6803157) and antiviral (CL:4665) pathways], whereas no specific function was identified in the elevated transcripts from Broad study.
Because most of the upregulated transcripts in UC-associated BEST4 + cells were elevated in other cells types, we crossreferenced the single cell results with an earlier bulk RNA-seq data set generated from three treatment-naïve UC patients (34). Markers of healthy BEST4 + cells, including BEST4, OTOP2, CA7, MT1H, and MT1G, were also downregulated in active UC analyzed by bulk RNA-seq. Elevation of UC-associated transcripts not specific to BEST4 + cells, including CCL20, S100P and PLA2G2A, were also detected in the bulk RNA-seq study. Upregulated PLA2G2A expression was further confirmed by qRT-PCR and immuno-histochemistry in tissue samples from UC patients and healthy controls, further supporting the bulk RNA-seq and scRNA-seq data (34).
Additional scRNA-seq study of colon biopsies from UC patients and healthy controls of Chinese Han ancestry (35) complemented the Broad and Oxford studies that were conducted in individuals of primarily European ancestry (21,22). In this study, 43,218 cells representing UC-affected sigmoid colon, unaffected proximal colon from UC patients, and sigmoid colon from healthy controls with an average of 1053 genes per cell were mapped into 21 clusters, six of which were epithelial cells (enterocytes, enterocyte progenitors, goblet cells, goblet progenitors, LGR5 + stem cells, and TRPM5 + tuft cells). A BEST4 + cell cluster was not uniquely identified but was found among transcripts in the enterocyte cluster.
COLONIC MICROFOLD (M)-LIKE CELLS
Microfold (M) cells are part of an integrated system of immunosurveillance in the intestinal mucosa, with a principal function of transporting luminal antigens to gut-associated lymphoid tissue (36,37). This activity has led to the idea that M-like cells could contribute to a "leaky gut" (38). There are several types of small intestinal M cells that display a common set of morphologic and functional features but differ in their specific gene expression patterns. Our knowledge of M-cell morphologic, molecular and functional features are mainly based on small intestinal M cells. M-like cells in the colon have been described to emerge under inflammatory conditions in mouse models, but relatively little is known about M-like cells in human colon (39).
According to the Broad study, specific markers of colonic M-like cells in healthy tissue were quite different from M cells in ileal Peyer's patches. Glycoprotein 2 (GP2), the M cell-specific marker that functions as a bacterial uptake receptor in the Peyer's patches, was not reported in colonic M-like cells, and neither were genes such as PGLYRP2, CLEC7A (Dectin-1), nor JAG1. M cells from Peyer's patches and colon shared only two transcripts, CCL20 and SPIB, which are known to initiate M cell differentiation (41). CCL20 is a UC GWAS gene that may serve as a marker of colonic M-cells in healthy tissue (40). Transcripts for both SPIB and TNFSF11, the RANKL receptor, in colonic M-like cells indicate similarities between the pathway of M cell differentiation in ileum and colon. However, SPIB was also present in BEST4 + and Tuft cells, precluding this gene as a specific M-like cell marker. SPIB transcription in BEST4 + cells was confirmed by the Oxford study (22).
We analyzed differences between M-like cells in healthy controls and UC ( Table 2). Not only were M-like cell numbers significantly elevated in inflamed mucosa, these cells also had increased transcription of CCL20 and CCL23, implicating them in the recruitment of other immune cells and propagating inflammation. Only two transcripts, CCL20 and SPINK5, were elevated in M-like cells in healthy colon as well as in UC. However, both transcripts were highly upregulated in UC in other absorptive and secretory cell clusters, thus precluding them from being specific M-like cell markers. Notably, SOX8 did not appear upregulated in any other cell cluster in UC ( Table 2). We conclude that SOX8 could be used to identify inducible colonic M-like cells and study their function in health and UC.
GOBLET CELLS
Goblet cells are critical in maintaining a protective mucus layer that provides separation between epithelia and luminal content. Both studies identified undifferentiated and differentiated transcriptional clusters of goblet cells (GCs) in relationship to the colonic crypt-surface axis. GCs were sub-classified into immature or mature populations (Broad), or grouped into five transcriptionally distinct clusters (Oxford). Among the 64 major GC markers identified by the Oxford study, 35 were also found elevated in GC clusters in the Broad study. Twelve transcripts highly elevated (log 2 FC ≥ 4.0) in both, mature and immature GCs in Broad study were also elevated in this cell type in Oxford study (Figure 2). Both studies are concordant in that there is increasing expression of MUC2 and ZG16 along the crypt-surface differentiation axis, while TFF3, ITLN1, SPINK4, CLCA1, and WFDC2 expression is higher in immature GCs (Figure 2). Surprisingly, only BEST2 and ZG16 were GCspecific in Broad study. Importantly, transcripts elevated in goblet cells (Figure 2) were in agreement with the subsequent study by Li et al. (35), including MUC2, ITLN1, REP15, LRRC26, NPDC1, TPSG1, SERPINA1, CLCA1, and SPINK4.
Of particular interest is the enrichment of mucins other than MUC2 in mature GCs. Both studies agree that transmembrane MUC1 and MUC4 are enriched in differentiated GCs, but are also detected in other cell clusters (transit amplifying (TA) cells). MUC12 is listed as elevated in GC cluster 4 in the Oxford study, but is elevated in colonocytes in the Broad study. MUC13 is enriched in mature GCs and in absorptive cells in both studies. MUC17 was not reported in GCs in any of the studies. Surprisingly, gel-forming MUC5B was elevated in mature GCs in the Oxford study but was not listed in GC clusters in the Broad study (instead, MUC5B was elevated in TA lineages).
The changes in goblet cells in UC through the prism of scRNAseq are striking. It is widely accepted that number of GCs and the luminal mucus layer are significantly reduced in UC-damaged epithelia (41). Both studies found a significant decrease in GC numbers and significantly altered transcriptional signatures (Figure 2). However, the transcriptional sub-populations of spatially distinct goblet cells (crypt bottom vs. top) were preserved in active UC-derived epithelial tissues. Additionally, in the Oxford study, a novel cluster of inflammation-associated GCs was found in UC.
Overall, GC clusters in UC partially lose their secretory identity and are characterized by a mix of transcripts present in both absorptive and secretory lineages. Each GC cluster in UC exhibited highly heterogeneous transcriptional changes, and a consensus for GC markers was not evident. Among 139 transcripts listed in the Oxford study as upregulated in GCs in UC, only 15 (11%) were also found in the Broad study. Additionally, many of these transcripts were also significantly elevated in non-GCs in UC. This significant transcriptional transformation of GCs may explain their functional transformation.
STRING interactome analysis of GC-upregulated transcripts in UC from the Oxford study suggest that GCs gain properties for antigen presentation (HSA04612), stress response (GO:0006950), and immune regulation (GO:0002376). Other studies have alluded to this ability of GCs to act as antigen presenting cells, including delivery of antigens to dendritic cells (42) or microbial sensing and recruitment of inflammatory cells (43).
One of the most striking results from both scRNA-seq datasets is that MUC2 transcript levels were similar between healthy and UC samples, yet MUC2 protein is substantially decreased in patients with severe UC (41,44). The discrepancy in detected MUC2 transcripts and protein in UC prompted us to perform a similar comparison for other goblet cell markers. Proteomic analysis of mucus from healthy or UC sigmoid colon biopsies showed that major core mucus components, including MUC2, FCGBP, CLCA1, and ZG16 were significantly reduced in active UC (42). Protein levels of TFF3, MUC4, 5B, 12 and 13 and ITLN1 did not change in mucus from UC patients. These data suggest post-transcriptional regulation of several core mucus proteins in disease accounts for differences between mRNA and protein expression in active UC.
It was surprising that transcript levels of MUC5AC, a gelforming mucin that is typically associated with surface gastric epithelia, but can be induced in the intestine during infection (45,46), was significantly elevated in inflamed colon. A recent study also found elevated MUC5AC transcript levels in active UC, although there was no significant correlation between MUC5AC expression levels and UC disease severity (47). However, MUC5AC was not reported among the core mucus proteins in active UC (45). MUC5AC is frequently present in colorectal adenomas and colon cancers, thus it's elevation might be due to UC-induced carcinogenesis (48).
In conclusion, according to both scRNA-seq datasets, goblet cell clusters in UC lost some of their secretory hallmarks and were characterized by a mix of transcripts present in both absorptive and secretory lineages.
COLONOCYTES
Colonocytes represent the largest cell pool in colonic epithelial tissue. Numerous studies have characterized the UC inflammation-induced pathologies of colonocytes and suggested that these pathologies must be pharmacologically addressed to achieve a full and sustained remission. Colonocytes can be spatially and functionally divided into three groups: Undifferentiated cells at the bottom of crypts, cells undergoing a transitional differentiation, and fully differentiated surface colonocytes. For the first time, scRNA-seq allows for precise characterization of the stages of cell differentiation and identification of the number of transcriptionally stable transition clusters which traverse from undifferentiated to fully differentiated. We analyzed the cluster specific markers of colonocytes in health and UC disease to gain insights into the molecular-cellular pathways driving UC disease.
Differentiated Absorptive Colonocytes in Health and Ulcerative Colitis Disease
Both studies identified a cluster of terminally differentiated colonocytes (additionally to BEST4 + cells) named Crypt Top Colonocytes (CTC) in the Oxford study or Enterocytes in the Broad study. The CTC cluster included ∼ 264 upregulated transcripts (AUC > 0.7) and Enterocyte cluster included ∼150 significantly upregulated transcripts. Seventeen of these highly elevated transcripts are identified in both studies but neither of these transcripts is cluster-specific ( Figure 3A). However, seven less elevated transcripts are cluster-specific and overlap in both studies (Supplementary Table 1A), and thus may serve as markers of mature colonocytes. ABCB1, an ATP-dependent drug efflux pump for xenobiotic compounds, is one of the common transcripts, and mutations in this gene are associated with IBD. Others are interferon-regulated proteins, including dsRNA-activated antiviral enzyme OAS1 which plays a critical role in cellular innate antiviral response, ubiquitin-like modifier ISG15, and sodium-phosphate symporter SLC20A1, which plays a fundamental housekeeping role in phosphate transport.
Many transcripts highly upregulated in differentiated colonocytes in one ( Figure 3B) or both studies were also detected in other cell clusters (Supplementary Tables 1B,D,E). For example, CTC-specific CDHR2, CLCA4, MS4A12, and SLC9A3R1 were also significantly elevated in differentiated colonocytes in Broad study, but were not cluster-specific. SLC26A3/DRA, the bicarbonate/Cl-exchanger is one of these transcripts and is a well-known marker of differentiated colonocytes validated in numerous studies (49). Low levels of DRA are also detected in undifferentiated colonocytes in both studies. DRA is involved in response to stimulus, immune system process, cell junction organization, bicarbonate transport and regulation of sodium transport. The expression differences in such transcripts between the two studies is likely due to large differences in the number of analyzed cells ( Table 1). A larger number of analyzed cells, as in Broad studies, improves the detection sensitivity for low abundance transcripts. Only a few transcripts are single cluster-specific and could not be detected in other clusters. Instead, the majority of the transcripts form gradients between different cell clusters. Further validation experiments are necessary to test whether inter-cluster copy number difference is large enough to allow use of a particular transcript as the cluster marker of differentiated colonocytes in healthy tissue, as in the case of DRA (Supplementary Tables 1B,D,E).
Only four transcripts were discordant (Supplementary Table 1). These transcripts were elevated in healthy Enterocytes in the Broad study but were among the UC inflammation-elevated transcripts in the Oxford study.
In UC inflammation, only four highly elevated and specific transcripts for differentiated colonocytes were common ( Figure 3C and Supplementary Table 1I); these might serve as UC disease markers of the differentiated colonocyte cluster. In the Oxford study, the UC-affected CTC cluster is characterized by ∼300 upregulated transcripts, with only 17 transcripts shared between inflamed and healthy CTC (Supplementary Table 1D), suggesting substantial structural and functional changes in this cell type in disease. STRING analysis of the CTC-specific transcripts in UC confirmed high representation of immune system processes. In contrast, bicarbonate and sodium transport and cell junction organization pathways were absent from the UC transcriptome. Loss of SCL26A3 was previously linked to inflammatory diarrhea in UC patients (50,51).
Additionally, there were significant changes in the composition of claudins in CTC cells; while CLDN7, 23, and 4 were elevated in healthy epithelium, CLDN1 and eight were elevated in disease. Mutations in NOD2, one of the major UC disease susceptibility genes, have been shown to alter the composition of tight junction proteins including upregulation of CLDN1 (52). scRNA-seq results confirm previous findings and correlate with the less studied CLDN8 in UC. UC-affected CTC cells also lose expression of CDHR5, an important intermicrovillar adhesion molecule that forms calcium-dependent heterophilic complexes with CDHR2 on adjacent microvilli (53). These two proteins control the packing of microvilli at the apical membrane of colonocytes and play a central role in brush border differentiation. Taken together, the CTC transcriptional signature reflects loss of differentiation in UC-inflamed epithelium. Interestingly, transcripts related to viral entry, specifically TMPRSS2, CHMP2A, 2B, and 4B, are downregulated in the CTC cluster in active UC, which might be protective against viral infections, including SARS-CoV2, despite an elevated cytokine profile (54,55).
In the Broad study, ∼111 protein transcripts were upregulated in the UC-inflamed Enterocyte cluster with only 16 being clusterspecific (Supplementary Tables 1I,L). Again, large difference between the numbers of cluster specific elevated transcripts in UC-affected differentiated colonocytes between the studies is probably due to the number of analyzed cells; the less cells are analyzed the more cluster-specific transcripts it would produce. Elevation of the proinflammatory cytokine LIF was previously reported in UC patients, confirming the scRNAseq data. LIF stimulates cell proliferation and thus may play a critical role in tumor development in UC patients (56). Another 13 transcripts elevated in multiple clusters in UC from the Broad study were also specifically elevated in UC-affected CTC cells (Supplementary Table 1J). The Broad study identified a single SAA1 transcript elevated in both healthy (specifically) and UC-affected (non-specifically) Enterocytes. SAA1 expression in epithelium has been shown to serve as an important link between mucosal T cells, microbial communities, and their tissue environment in patients with IBD (57), further supporting these scRNA-seq findings.
Undifferentiated Colonocyte Clusters in Health and Ulcerative Colitis Disease
The pool of immature healthy colonocytes is mainly represented by three distinct clusters of Immature Enterocytes 1 and 2, and Enterocyte Progenitors in the Broad study. We searched for specific cluster markers among the substantially elevated protein coding transcripts (log2fc ≥ 4.0; Supplementary Table 2). DUOX2 and TMEM150B were exclusively elevated in the healthy Immature Enterocytes 1 cluster. Similarly, FOXH1, C14orf180, and HS3ST6 may serve as markers of Immature Enterocytes 2. Surprisingly, no protein transcripts elevated above this threshold were among Enterocyte Progenitors. In addition, there were no transcripts upregulated above the threshold that are common and specific for all three clusters.
In UC-inflamed tissue, according to the Broad study, eight elevated (log2fc ≥ 4.0) transcripts were shared among all three clusters (Supplementary Table 2H). However, these transcripts were upregulated in numerous other clusters, including differentiated absorptive and secretory lineages. Importantly, the expression pattern of these eight transcripts substantially differs between inflamed UC and healthy tissue. In healthy tissue, they are upregulated in a single or few clusters. However, in UC they are spread across many clusters, further demonstrating that the major UC disease phenotype dissociates the molecular boundaries between different cell types and their stage of differentiation. For example, CCL20 appears as an M-like cell marker in healthy tissue, but is broadly elevated in both secretory and absorptive cells in UC. Similarly, in healthy epithelium, GLDN and DUOX2 are elevated exclusively in Enterocyte Progenitors and Immature Enterocytes 1, respectively, but in UC are elevated in at least seven other secretory and absorptive cell clusters. DUOX2 is the only Immature Enterocytes 1 transcript elevated in both healthy and diseased epithelium. It has been shown that epithelial DUOX2 and DUOXA2 form the predominant enzyme system for H 2 O 2 production in human colon and in active UC (58), further supporting these scRNA-seq data.
Elevated MUC1 could be considered as a marker of Immature Enterocytes 1 in active UC (Figure 3D), while in healthy tissue MUC1 is expressed throughout all epithelial clusters. SMAD9, KAZN, and DOCK4 ( Figure 3D) could be considered as markers of Immature Enterocytes 2. Several transcripts highly elevated in the clusters of undifferentiated colonocytes further illustrate the loss of cluster-specificity expression in inflamed UC epithelium compared to healthy tissue. For example, in healthy epithelium, SDR16C5 is elevated in M-like cells, Goblet cells and Tuft cell clusters, MIA is elevated in M-like and Tuft cells; and PRSS22 is elevated in Secretory and Tuft cells; however, all three transcripts are elevated in nearly all the epithelial cell clusters in UC. Importantly, SI, a well-known marker of fully differentiated small intestinal enterocytes was not detected in healthy human colon, but was broadly expressed in UC, including undifferentiated colonocytes. SI has been detected in inflammatory, regenerative, and dysplastic mucosa in Ulcerative Colitis (59). Additionally, OLFM4 is mainly upregulated in Stem and TA1 clusters in healthy epithelium, but in UC is elevated in TA and Absorptive Enterocytes. The expression pattern of these transcripts again shows the loss of cell type specificity in UC.
The transcriptional comparison of immature colonocytes between two studies is challenging and not straightforward due to differences in transcript clustering algorithms. We speculated that Colonocytes and Absorptive Progenitors could be clusters that represent immature colonocytes in Oxford study. Eight transcripts found to be most elevated in all three undifferentiated colonocyte clusters in the Broad study were also significantly elevated in UC Colonocytes in Oxford study. Four transcripts (TNIP3, S100P, DUOX2, and CXCL1) were also elevated in Absorptive Progenitors. Again, similar to the results of the Broad study, none of the transcripts were cluster-specific and all were upregulated in most clusters except EECs in the Oxford study. Additionally, both datasets suggest that DOCK4, a membrane-associated cytoplasmic protein, which functions as a guanine nucleotide exchange factor involved in regulation of adherence junctions between the cells, may be a reliable marker of undifferentiated colonocytes in UC disease. Although its role in UC disease is not known, DOCK4 was recently identified as a regulator of goblet cell differentiation and MUC2 production in the gut (60).
Several transcripts identified as markers of healthy differentiated colonocytes in the Broad and Oxford studies (VAMP8, OAS1, ISG15, SLC20A1; Figure 3) were concordant with data by Li et al. However, the remaining transcripts that were listed as markers of healthy differentiated or undifferentiated colonocytes (Supplementary Tables 1, 2) were not reported. CDC42EP2, a potential marker of UC-affected differentiated colonocytes was also reported by Li et al., while MUC1, a potential marker of undifferentiated colonocytes, was elevated in tuft cells in addition to enterocyte progenitors.
COLONIC STEM CELLS
Altered histology and transcriptional profiles in UC epithelia have been observed to persist even after achieving a disease remission (61), suggesting that the colonic stem cell (SC) may be permanently changed during the course of disease (18). Somatic nonsense or gene mutations have been discovered in UC epithelia, including genes involved in IL-17 signaling, such as NFKBIZ, ZC3H12A, and PIGR (62). Inflammation-induced somatic mutations might provide clonal advantage to these stem cells and propagate UC-related pathologies. We next analyzed stem cell profiles in both scRNA-seq data sets to search for potential signatures associated with UC.
The SC cluster from healthy tissue was not described in the Oxford study, possibly because only 38 cells contained the stem cell marker LGR5. The Broad study identified LGR5 as the most SC-enriched and specific transcript ( Table 3). Among the 18 most elevated SC transcripts (log 2 FC ≥ 4.0), 16 were SC specific, which might substantially broaden the ability to identify and characterize human colonic SC via immunoblotting, immunofluorescence, functional assays and other approaches. OLFM4 is often used as a substitute marker for SC (63) and is elevated (log 2 FC 2.4) in both SC and TA1 cells. Interestingly, several other highly expressed SC-specific transcripts are lincRNAs (RP11-219E7.4 and RP11-760H22.2) and RP11-84C10.4 antisense, all of which might be involved in SC-specific regulation of CFTR.
In the Broad study, 65 transcripts were upregulated in SC in active UC relative to healthy controls. Although a majority of these transcripts were also upregulated at least in one other cell cluster, eight transcripts were specific to SC in UC (Table 3). Surprisingly, the upregulated transcripts are commonly associated with healthy differentiated cell types such as colonocytes (EZR, IFI27) or goblet cells (IER2, HBEGF, NFKBIA, and FAM3D). The functional significance of these UC-induced SC transformations remains to be determined. STRING analysis of the 65 upregulated SC transcripts found associations to stress response (GO:0006950) and cell communication (GO:0010646) as the most inclusive gene ontology pathways. These pathways also included a large set of molecules involved in immune response. Multiple HLA and IFI transcripts, either SC-specific or shared between SCs and other epithelial cell clusters in UC, were significantly elevated, indicating a shift in SC function toward antigen presentation in UC. HSPA6 and CCL20 are UC GWAS genes, possibly linking genetic predisposition to UC with changes in the SC transcriptome. In UC, the fate of SC markers LGR5, OLFM4, or AXIN2 is not clear, as these transcripts were not listed among those enriched or downregulated in UC in the Oxford study. According to the Broad study, AXIN2 was significantly downregulated in stem cells in UC. In contrast, OLFM4 was slightly upregulated in stem cells and much more elevated in all TA clusters and in Enterocytes in UC. The lack of overlap between the two studies on SC markers in inflamed colonic tissue indicates that it is too soon to hypothesize on the mechanism(s) contributing to long-lasting changes in SCs in active UC.
The expression of potential stem cell-enriched transcripts ( Table 3) was not reported in healthy epithelium by Li et al., despite the presence of LGR5 + stem cell clusters in both healthy and UC tissues. In UC, markers elevated in stem cell cluster in either the Oxford or Broad study were discordant with the findings reported by Li et al. (35). EZR, a classical marker of epithelial cells, was found elevated in several epithelial cell clusters, and surprisingly, in fibroblasts. IFI27 was nonspecifically elevated in stem cells, while FAM3D and PRAC1 were elevated in stem cells when expression was normalized to the proximal colon transcriptome but not to healthy sigmoid colon. NFKBIA, JUNB, and IER2 were elevated in other epithelial and/or non-epithelial cell clusters, but not in LGR5 + stem cells.
DISCUSSION
RNA-seq of single cells paves the way for discovery of novel epithelial cell clusters in UC. The major differences between the two studies were in the methods and protocols used to isolate cells and the large difference in the number of analyzed cells. In addition, patient selection is also variable and patients are not comparable across studies ( Table 1). These limitations should not deter researchers from using the findings in these studies, but rather emphasize the need to validate markers by alternative techniques. The Broad and Oxford studies make important contributions to defining the UC epithelia transcriptional landscape, but points of incongruity suggest that expanded profiles from additional subjects are necessary to refine disease-associated hypotheses. Combining scRNA-seq with other -omics into a reference disease atlas would provide a powerful tool to iteratively model, test, and revise mechanistic models in UC experimental settings for drug testing and other clinical studies. Because parameters vary widely across studies, it is not straightforward to determine whether a particular transcript is specifically changed within a single cluster and thus be a reasonable target for further evaluation in human tissue or organoid models. Our analysis acknowledges this gap and identifies some areas in which non-omics validation is important.
scRNA-seq in the Broad and Oxford studies revealed that transcriptional signatures and cell type boundaries become more nebulous in UC. Multiple transcripts were differentially regulated across many clusters, challenging the classical concept of unique cell cluster markers associated with active UC. The lack of clearly separated transcripts reflects a general loss of cell specialization, replaced by cell phenotypes that do not map to healthy tissue. Evaluating markedly altered transcripts in relation to known interactors (interactome analysis) adds depth to cluster analysis, but these strategies are not useful if cell type clusters lack discrete boundaries in disease.
Study interpretations are limited by the heterogeneity of subjects (including gender, age, ethnicity, site of colon biopsy, stage of UC disease, and prior use of anti-inflammatory therapies). The total number of cells acquired and transcripts sequenced also contribute to variability in sensitivity for detecting rare cell types or transcripts. Validation of findings at the protein level is also imperative to advance understanding of the molecular features conserved in UC epithelia.
AUTHOR CONTRIBUTIONS
JS, JF-A, and OK researched data, contributed to discussion of content and writing, generated figures and tables, and reviewed/edited the manuscript before submission. WH and JH contributed to discussion of content for the article and reviewed/edited the manuscript before submission. JI researched data for the article and reviewed/edited the manuscript before submission. All authors contributed to the article and approved the submitted version.
FUNDING
This research was supported by NIH grants U01 DK103168, P01 AI125181 to OK, JI, and JF-A, K01 DK113043 to JF-A, and K01 DK106323 to JI, and by University of New Mexico Department of Internal Medicine start-up funds to JI and OK. | 2022-04-21T13:17:28.803Z | 2022-04-20T00:00:00.000 | {
"year": 2022,
"sha1": "8ed14a1c09c44e51b686dee8a30254e6dd4688cc",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Frontier",
"pdf_hash": "8ed14a1c09c44e51b686dee8a30254e6dd4688cc",
"s2fieldsofstudy": [
"Biology",
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
245837503 | pes2o/s2orc | v3-fos-license | Multi-valued variational inequalities for variable exponent double phase problems: comparison and extremality results
We prove existence and comparison results for multi-valued variational inequalities in a bounded domain $\Omega$ of the form \begin{equation*} u\in K\,:\, 0 \in Au+\partial I_K(u)+\mathcal{F}(u)+\mathcal{F}_\Gamma(u)\quad\text{in }W^{1,\mathcal{H}}(\Omega)^*, \end{equation*} where $A\colon W^{1, \mathcal{H}}(\Omega) \to W^{1, \mathcal{H}}(\Omega)^*$ given by \begin{equation*} Au:=-\text{div}\left(|\nabla u|^{p(x)-2} \nabla u+ \mu(x) |\nabla u|^{q(x)-2} \nabla u\right) \end{equation*} for $u \in W^{1, \mathcal{H}}(\Omega)$, is the double phase operator with variable exponents and $W^{1, \mathcal{H}}(\Omega)$ is the associated Musielak-Orlicz Sobolev space. First, an existence result is proved under some weak coercivity condition. Our main focus aims at the treatment of the problem under consideration when coercivity fails. To this end we establish the method of sub-supersolution for the multi-valued variational inequality in the space $W^{1, \mathcal{H}}(\Omega)$ based on appropriately defined sub- and supersolutions, which yields the existence of solutions within an ordered interval of sub-supersolution. Moreover, the existence of extremal solutions will be shown provided the closed, convex subset $K$ of $W^{1, \mathcal{H}}(\Omega)$ satisfies a lattice condition. As an application of the sub-supersolution method we are able to show that a class of generalized variational-hemivariational inequalities with a leading double phase operator are included as a special case of the multi-valued variational inequality considered here. Based on a fixed point argument, we also study the case when the corresponding Nemytskij operators $\mathcal{F}, \mathcal{F}_\Gamma$ need not be continuous. At the end, we give a nontrivial example of the construction of sub- and supersolutions related to the problem above.
When p and q are constants, such setting is originally due to Zhikov [42] who introduced and studied the integral functional ω → |∇ω| p + µ(x)|∇ω| q dx (1.2) in order to describe models for strongly anisotropic materials. The functional (1.2) also demonstrated its importance in the study of duality theory and in the context of the Lavrentiev phenomenon, see Zhikov [43]. Note that (1.2) is related to the differential operator u → − div |∇u| p−2 ∇u + µ(x)|∇u| q−2 ∇u , which is a special case of (1.1). From the physical point of view, (1.2) describes the phenomenon that the energy density changes its ellipticity and growth properties according to the point in the domain. In the elasticity theory, for example, the modulating coefficient µ(·) dictates the geometry of composites made of two different materials with distinct power hardening exponents q and p, see Zhikov [44]. From the mathematical point of view, the behavior of (1.2) depends on the sets on which the weight function µ(·) vanishes or not. Therefore, we have two phases (µ(x) = 0 or = 0) and so we call it double phase. Even though no global regularity theory for double phase problems exists yet, there are some remarkable results about local minimizers, see [2,3,4,14,15,17,28,29,30,34]. We also refer to the recent overview article in [31]. Let us next formulate the problem under consideration. To this end, let Ω ⊂ R N (N ≥ 2) be a bounded domain with Lipschitz boundary ∂Ω and let Γ ⊂ ∂Ω be a relatively open subset and denote Γ 0 = ∂Ω \ Γ such that ∂Ω = Γ ∪ Γ 0 . We consider the multi-valued elliptic variational inequality of the form u ∈ K : 0 ∈ Au + ∂I K (u) + F (u) + F Γ (u) in W 1,H (Ω) * , (1.4) where W 1,H (Ω) * its dual space of W 1,H (Ω), K is a closed convex subset of the closed subspace V Γ0 of W 1,H (Ω) defined by V Γ0 = u ∈ W 1,H (Ω) : u | Γ0 = 0 , I K is the indicator function related to K, and ∂I K denotes its subdifferential. The lower order multi-valued operators F and F Γ are generated by the multi-valued functions f : Ω × R → 2 R \ {∅} and f Γ : Γ × R → 2 R \ {∅}, respectively. Let for all x ∈ Ω (1. 5) be the critical exponents to p for 1 < p(x) < N , and denote by p ′ (·) the Hölder conjugate to p given by p ′ (·) = p(·) p(·)−1 . We assume the following hypotheses: (H0) p, q ∈ C(Ω) such that 1 < p(x) < N , p(x) < q(x) < p * (x) and 0 ≤ µ(·) ∈ L ∞ (Ω). (F1) f : Ω×R → 2 R \{∅} and f Γ : Γ×R → 2 R \{∅} are graph measurable on Ω×R and Γ × R, respectively, and for a. a. x ∈ Ω the function f (x, ·) : R → 2 R is upper semicontinuous and for a. a. x ∈ Γ, the function f Γ (x, ·) : R → 2 R is upper semicontinuous. (F2) There exist r 1 ∈ C(Ω), r 2 ∈ C(Γ) with 1 < r 1 (x) < p * (x) for all x ∈ Ω, 1 < r 2 (x) < p * (x) for all x ∈ Γ, β ≥ 0, β Γ ≥ 0 and functions α ∈ L r ′ 1 (·) (Ω), α Γ ∈ L r ′ 2 (·) (Γ) such that sup {|η| : η ∈ f (x, s)} ≤ α(x) + β|s| r1(x)−1 for a. a. x ∈ Ω, for all s ∈ R, and for a. a. x ∈ Γ, and for all s ∈ R. We point out that the classical obstacle problem fits in our setting, that is, in Ω with a given obstacle ψ : Ω → R. Originally, the study of obstacle problems is due the pioneering contribution by Stefan [36] in which the temperature distribution in a homogeneous medium undergoing a phase change, typically a body of ice at zero degrees centigrade submerged in water, was studied. Furthermore, we mention the famous work of J.-L. Lions [26] who studied the equilibrium position of an elastic membrane which lies above a given obstacle and which turns out as the unique solution of the Dirichlet energy functional minimized on the closed convex set K.
Before we state our main results, we first give the definition of a weak solution to problem (1.4).
The multi-valued variational inequality (1.4) covers a wide range of elliptic problems which can be deduced from (1.4) by specifying Γ, K, and the lower order terms. To give an idea, let us consider a few examples. in Ω, where ∂u ∂ν A = |∇u| p(x)−2 ∇u + µ(x)|∇u| q(x)−2 ∇u · ν with ν denoting the outward unit normal at Γ. (Ω), then (1.4) becomes the following multi-valued Dirichlet boundary value problem Further special cases can be deduced from (1.4) such as a mixed boundary value problems that arise when |Γ| > 0 and |Γ 0 | > 0, and K = V Γ0 . In Section 6 we will see that (1.4) also includes an important class of generalized variationalhemivariational inequalities.
Our first result is the following existence theorem for (1.4) under a coercivity condition.
Then problem (1.4) has at least one solution in the sense of Definition 1.1.
The proof will be given in Section 3, see also Corollary 3.2, which is a direct consequence of Theorem 1.4. If the coercivity condition (1.7) or appropriate generalized versions of coercivity are not satisfied then problem (1.4) may have no solutions. However, in the noncoercive case we still are able to prove the existence of solutions provided appropriately defined sub-supersolutions for (1.4) exist. In this paper we establish an sub-supersolution method based on the following definition of sub-and supersolutions of problem (1.4). For functions u, v : Ω → R we use the notation Remark 1.7. We note that although variational inequalities are generally nonsymmetric due to the presence of constraints, the notions for sub-and supersolution defined by Definition 1.5 and Definition 1.6, respectively, do have a symmetric structure in the following sense: one obtains the definition for the supersolution u from the definition of the subsolution by replacing u, η, ζ in the definition of subsolution by u, η, ζ, and interchanging ∨ by ∧. Symmetric structure is a main feature of the sub-supersolution concepts for smooth equations, which has been extended here to multi-valued variational inequalities with variable exponent double-phase operator.
Just for illustration, let us apply the above definitions to the special case given by Example 1.2 and assume that f and f Γ are single-valued, that is (1.8) Let u be a subsolution according to Definition 1.5. As K = W 1,H (Ω) and W 1,H (Ω) has lattice structure, condition (i) is trivially satisfied. Condition (ii) yields η(x) = f (x, u(x)) for a. a. x ∈ Ω and ζ(x) = f Γ (x, u(x)) for a. a. x ∈ Γ. For any ϕ ∈ K = for all ϕ ∈ W 1,H (Ω). Since the set {(u − ϕ) + : ϕ ∈ W 1,H (Ω)} equals {ψ ∈ W 1,H (Ω) : ψ ≥ 0}, the last inequality is nothing but the usual notion of subsolution for the boundary value problem (1.8), that is, Similarly, Definition 1.6 for the supersolution u of (1.8) reduces to Next, we suppose the following local boundedness conditions on the multi-valued nonlinearities with respect to the order interval [u, u].
(F3) Let u and u be sub-and supersolutions of (1.4) such that u ≤ u and suppose the following growth conditions for all s ∈ [u(x), u(x)] and for some k Ω ∈ L τ ′ 1 (·) (Ω), k Γ ∈ L τ ′ 2 (·) (Γ). The sub-supersolution method for (1.4) is established by the following existence and comparison result. Theorem 1.8. Let u and u be an ordered pair of sub-and supersolutions of (1.4) fulfilling u ≤ u and let hypotheses (H0), (F1) and (F3) be satisfied. Then problem (1.4) has a solution u ∈ K such that u ≤ u ≤ u a.e. in Ω.
We remark that Theorem 1.8 will be seen as straightforward consequence of a general existence and comparison principle (see Theorem 4.1) which will be proved in Section 4, and which at the same time allows us to order-theoretically and topologically characterize the solution set S of all solutions of (1.4) lying within the interval [u, u]. We have the following characterization of S, see Section 5. Theorem 1.9.
(i) Under the assumptions of Theorem 1.8, the solutions set S is a compact subset of W 1,H (Ω).
9) then (a) any u ∈ S is both a (weak) subsolution and supersolution of (1.4), and (b) S is directed both downward and upward, that is, for all u 1 , u 2 ∈ S, there exists w 1 , w 2 ∈ S such that w 1 ≤ min{u 1 , u 2 } and w 2 ≥ max{u 1 , u 2 }.
(iii) If (1.9) hold then S has smallest and greatest elements, that is, there are u * , u * ∈ S such that u * ≤ u ≤ u * for all u ∈ S.
In Section 6, as an application of Theorem 1.8 and Theorem 1.9, we are going to show that a class of generalized variational-hemivariational inequalities with the double phase operator as the leading operator of the form (1.10) turn out to be a special case of (1.4) only, see Theorem 6.2. In Section 7 we also study the case when the functions f and f Γ need not be continuous (so F and F Γ need not be pseudomonotone anymore). The idea in the proof is the usage of an fixed point argument, see Theorems 2.13 and 7.3. Lastly, in Section 8, we construct nontrivial sub-and supersolutions of (1.4) which can be applied to our results, see Theorem 8.2 and Corollary 8.3.
To the best of our knowledge, our results are new even in the case when p and q are constants. For double phase problems with variable exponents there are only few works, we mention the papers of [1] for the variable exponent Baouendi-Grushin operator, of [16] for single-valued convection problems and of [40] for obstacle problems involving multi-valued reaction terms with gradient dependence. Papers dealing with the constant exponent double phase (1.3) along with multi-valued righthand sides can be found in [38] and [39] who studied obstacle problems involving the special case of Clarke's generalized gradients. Note that all these works are dealing with the coercive case.
Finally, we mention some recent results for single-valued double phase problems without constraints, such as, [13] for eigenvalue problems for double phase problems, [20] for sign-changing solutions based on the Nehari manifold, [21] for general convection problems, [27] for superlinear double phase problems, [33] for double phase problems via Morse theory, [35] for multiple solutions for double phase variational problems and [41] for anisotropic double phase problems. As for multi-valued variational inequalities with leading p-Laplacian type operators we refer to [8] for bounded domains, and [9,10] for unbounded domains.
Preliminaries
In this section we recall some results about variable exponent Sobolev space, Musielak-Orlicz Sobolev spaces and properties of the variable exponent double phase operator. The results are mainly taken from the books of [18] and [22] as well as the papers of [16], [19] and [23].
We suppose now condition (H0) and introduce the nonlinear function H : where the related modular to H is given by The corresponding Musielak-Orlicz Sobolev space W 1,H (Ω) is defined by equipped with the norm where ∇u H = |∇u| H . Moreover, we denote by W 1,H 0 (Ω) the completion of We now equip the space W 1,H (Ω) with the equivalent norm where the modularρ H is given bŷ The next proposition can be found in [16, Proposition 2.14].
(Ω) and let A : X → X * be the nonlinear operator defined by for u, v ∈ X with ·, · being the duality pairing between X and its dual space X * .
The following proposition summarizes the main properties of A : X → X * , see [ In what follows, to shorten notation, we write · = · ρH for the norm in W 1,H (Ω) and · 0 = · 1,H,0 for the norm in W 1,H 0 (Ω). The corresponding dual spaces are denoted by W 1,H (Ω) * and W 1,H 0 (Ω) * , respectively. Given a Banach space X and its dual space X * we denote K(X * ) = {P ⊂ X * : P = ∅, P is closed and convex} .
Let X be a real Banach space with its dual space X * . A function J : X → R is said to be locally Lipschitz at u ∈ X if there exist a neighborhood N (u) of u and a constant L u > 0 such that Definition 2.8. Let J : X → R be a locally Lipschitz function and let u, v ∈ X. The generalized directional derivative J • (u; v) of J at the point u in the direction v is defined by The generalized gradient ∂J : The next proposition collects some basic results, see [12] or [32].
Theorem 2.10. Assume F : Ω × R → 2 R satisfies the following conditions: (i) For a. a. x ∈ Ω and for all u ∈ R, F (x, u) is closed and nonempty; for a. a. x ∈ Ω and for all v ∈ F (x, u).
Remark 2.11. We have an analogous result to Theorem 2.10, where Ω is replaced by Γ. In fact, a straightforward generalization of Theorem 2.10 holds true with Ω being a measure space on which Lebesgue and Sobolev spaces with variable exponents are defined.
Theorem 2.12. Let X be a real reflexive Banach space, let F : D(F ) ⊂ X → 2 X * be a maximal monotone operator, let G : D(G) = X → 2 X * be a bounded multivalued pseudomonotone operator and let L ∈ X * . Assume that there exist u 0 ∈ X and R ≥ u 0 X such that D(F ) ∩ B R (0) = ∅ and for all u ∈ D(F ) with u X = R, for all ξ ∈ F (u) and for all η ∈ G(u). Then the inclusion An important tool in extending our results to discontinuous Nemytskij operators is the next fixed point result, see [ Theorem 2.13. Let P be a subset of an ordered normed space X, and let G : P → P be an increasing mapping, that is, x, y ∈ P with x ≤ y implies Gx ≤ Gy. Then the following holds true: (i) If the image G(P ) has a lower bound in P and increasing sequences of G(P ) converge weakly in P , then G has the smallest fixed point x * given by x * = min{x : Gx ≤ x}. (ii) If the image G(P ) has an upper bound in P and decreasing sequences of G(P ) converge weakly in P , then G has the greatest fixed point x * given by 3. Coercive Case: Proof of Theorem 1.4 In this section, we are going to prove Theorem 1.4. First, recall that the embed- (Ω) * be their adjoints. As a consequence of (F1), for any u ∈ M (Ω), the set of measurable selections of f (·, u),f Similarly, we definef We have the following crucial property of F and F Γ .
Proof. First, let us note thatf (u) ∈ K(L r ′ 1 (·) (Ω)) for all u ∈ L r1(·) (Ω). In fact, the convexity off (u) and the boundedness off (as a multi-valued mapping) follow directly from (F1) and (F2). The proof of the closedness off (u) is a direct consequence of the fact that f (x, t) is a closed bounded interval in R for a. a. x ∈ Ω and for all t ∈ R.
Next, we show that the graph of F is (sequentially) weakly closed in W 1,H (Ω) × W 1,H (Ω) * . Assume that {u n } n∈N and {u * n } n∈N are sequences in W 1,H (Ω) and W 1,H (Ω) * , respectively, such that . It follows from (3.1) and the compactness of the embedding i r1(·) that u n → u in L r1(·) (Ω).
We are now ready to prove Theorem 1.4.
Proof of Theorem 1.4. We are going to apply Theorem 2.12. Since A is continuous, strictly monotone and bounded on W 1,H (Ω) with domain D(A) = W 1,H (Ω), it is a (single-valued) bounded and pseudomonotone mapping from W 1,H (Ω) to 2 W 1,H (Ω) * . It follows from Proposition 3.1 that A + F + F Γ is a pseudomonotone and bounded mapping from W 1,H (Ω) into 2 W 1,H (Ω) * .
We note that ∂I K is a maximal monotone mapping from W 1,H (Ω) to 2 W 1,H (Ω) * with domain D(∂I K ) = K. According to Theorem 2.12, under the coercivity condition (3.7), problem (1.4) has at least one solution.
A straightforward consequence of Theorem 1.4 is the following result. Corollary 3.2. Let hypotheses (H0), (F1) and (F2) be satisfied and suppose that for fixed u 0 ∈ K the following coercivity condition holds Then problem (1.4) has at least one solution.
4.
Noncoercive Case: Proof of Theorem 1.8 In order to prove Theorems 1.8 and 1.9, let us first establish the following general existence and enclosure theorem for solutions of (1.4) if a finite number of sub-and supersolutions exist and f has a local growth between those sub-and supersolutions. Suppose there exist for all s ∈ [u(x), u(x)] and for some k Ω ∈ L τ ′ 1 (·) (Ω), k Γ ∈ L τ ′ 2 (·) (Γ). Then, there exists a solution u of (1.4) such that u ≤ u ≤ u a. e. in Ω.
Proof. First, note that by increasing τ 1 and τ 2 in Definitions 1.5 and 1.6 for each u i and u j and in the growth condition (4.1) appropriately, to simplify the notation we can assume without loss of generality that the functions τ 1 and τ 2 in the definitions of u i (1 ≤ i ≤ k) and u j (1 ≤ j ≤ m) in Definitions 1.5 and 1.6 and in the growth condition 4.1, are the same.
Next, let us define a truncation-regularization function b as follows. For x ∈ Ω and u ∈ R, let
Let us consider the following auxiliary variational inequality: Find u ∈ K and η ∈ L τ ′ 1 (·) (Ω), ζ ∈ L τ ′ 2 (·) (Γ), such that and (4.9) The inequality above is equivalent to the following variational inequality: Find This variational inequality is, in its turn, equivalent to finding u ∈ D(∂I K ) = K, l ∈ (∂I K )(u), and in W 1,H (Ω) * . We observe that ∂I K is a maximal monotone mapping and is a (multi-valued) pseudomonotone bounded mapping from W 1,H (Ω) to 2 W 1,H (Ω) * . Hence, to apply the abstract existence result in Corollary 2.3 of [25], we only need to check the following coercivity condition: There exists u 0 ∈ K such that see (4.10).
In fact, let u 0 be any (fixed) element of K. For any u ∈ K, any l ∈ (∂I K )(u), we have 0 = I K (u 0 ) − I K (u) ≥ l, u 0 − u , i.e., l, u − u 0 ≥ 0. Hence, to prove (4.11), one only needs to show that where η ∈f 0 (u) and ζ ∈f 0Γ (u). It follows from (4.4) that (4.14) From (4.6) and (4.7), by applying Hölder's and Young's inequalities with ε for variable exponents (see e.g. [24]), we get On the other hand, we have for any i ∈ {1, . . . , k}, Hence, Lastly, since A has as a potential functional the following convex functional On the other hand, it follows from (H0) that where |Ω| is the Lebesgue measure of Ω. Hence, it follows from (4.15) that there is a 3 > 0 such that for all u ∈ W 1,H (Ω) (4.20) Combining the estimates from (4.13) to (4.20), we see that for any u ∈ K, see Proposition 3.5 in [16], the estimate in (4.21) implies (4.12). It follows from Corollary 2.3 in [25] that there exist u, η, and ζ that satisfy (4.8) and (4.9).
The proof of Theorem 1.8 is now an immediate consequence of Theorem 4.1.
Proof of Theorem 1.8. In the particular case where m = n = 1, condition (4.1) becomes condition (F3) and Theorem 4.1 reduces to Theorem 1.8.
Extremal Solutions: Proof of Theorem 1.9
In this section we give the proof of Theorem 1.9.
From (F3), {η n } n∈N is a bounded sequence in L τ ′ 1 (·) (Ω) and {ζ n } n∈N is a bounded sequence in L τ ′ 2 (·) (Γ). Using (1.6) with u n , η n , ζ n , and v = v 0 , a fixed element of K, we see that {ρ H (|∇u n |)} n∈N is a bounded sequence and thus the set { ∇u n H : n ∈ N} is also bounded. Hence, {u n } n∈N is a bounded sequence in W 1,H (Ω) and there exists a subsequence {u n l } l∈N ⊂ {u n } n∈N such that u n l ⇀ u 0 in W 1,H (Ω) for some u 0 ∈ K (note that K is weakly closed in W 1,H (Ω)). Thus, u n l → u 0 in L H (Ω) and in L τ1(·) (Ω), and u n l | Γ → u 0 | Γ in L τ2(·) (Γ).
(ii) The proofs for (ii) and (iii) follow the same lines as those for the case of regular Sobolev spaces, thus their outlines are presented here for the sake of completeness. Assuming (1.9), we see that if u 0 ∈ S then u 0 ∧ K ⊂ K and thus u 0 is a subsolution of (1.4) in the sense of Definition 1.5. If u 1 , u 2 ∈ S then they are subsolutions of (1.4) and Theorem 4.1 thus implies the existence of a solution u of (1.4) such that max{u 1 , u 2 } ≤ u ≤ min{u j : 1 ≤ j ≤ m} = u. It is clear that u ∈ S. In fact, since u 1 , u 2 ∈ S, this follows directly from the definition of S and the inequalities (iii) Since W 1,H (Ω) is separable (with the norm topology), so is S. Let {w n } n∈N be a dense sequence in S. Using the directedness of S, we can construct inductively a sequence {u n } n∈N in S such that w n ≤ u n ≤ u n+1 for all n ∈ N. Let u * (x) = sup{u n (x) : n ∈ N} = lim n→∞ u n (x) for x ∈ Ω.
As a consequence of the compactness of S, u n → u * in W 1,H (Ω) and u * ∈ S. Since u * ≥ w n a. e. in Ω for all n ∈ N, from the density of {w n } n∈N in S, we see that u * ≥ u a. e. in Ω for all u ∈ S. The existence of the smallest element u * of S is proved analogously.
Application: Generalized Variational-Hemivariational Inequalities
In this section we are dealing with the generalized variational-hemivariational inequality (1.10) which is of the form where A is the variable exponent double-phase operator given by (1.1). The functions j, j Γ given by j : Ω × R × R → R with (x, r, s) → j(x, r, s), are supposed to be locally Lipschitz with respect to s, and j • (x, r, s; ̺) and j • Γ (x, r, s; ̺) denote Clarke's generalized directional derivatives at s in the direction ̺ for fixed (x, r). In case j and j Γ are independent of r, (6.1) represents a variationalhemivariational inequality. However, in the general case of problem (6.1) the functions s → j(x, s, s) and s → j Γ (x, s, s) may be not locally Lipschitz but only partially locally Lipschitz. This enlarges the class of variational-hemivariational inequalities considerably, and therefore we are calling them generalized variationalhemivariational inequalities. Under hypotheses specified next we are going to show that problem (6.1) is equivalent to some subclass of multi-valued variational inequalities of the form (1.4), which in a sense fills a gap in the literature where both problems are considered independently and separately.
We suppose the following hypotheses on j and j Γ : (J1) The functions x → j(x, r, s) and x → j Γ (x, r, s) are measurable in Ω and on Γ, respectively, for all r, s ∈ R. The functions r → j(x, r, s) and r → j Γ (x, r, s) are continuous for a. a. x ∈ Ω and x ∈ Γ, respectively, and for all s ∈ R. The functions s → j(x, r, s) and s → j Γ (x, r, s) are locally Lipschitz for a. a. x ∈ Ω and x ∈ Γ, respectively, and for all r ∈ R.
(J2) Let s → ∂j(x, r, s) and s → ∂j Γ (x, r, s) denote Clarke's generalized gradient of the functions j and j Γ with respect to the variable s, respectively. Assume the following growth conditions for s → ∂j(x, s, s) and s → ∂j Γ (x, s, s): for a. a. x ∈ Ω, for all s ∈ R, and for a. a. x ∈ Γ, and for all s ∈ R. (J3) Let s → j • (x, r, s; ̺) and s → j • Γ (x, r, s; ̺) denote Clarke's generalized directional derivative of the functions s → j(x, r, s) and s → j Γ (x, r, s) at s, respectively, in the direction ̺ for fixed (x, r). Suppose that s → j • (x, s, s; ̺) and s → j • Γ (x, s, s; ̺) are upper semicontinuous for a. a. x ∈ Ω, and x ∈ Γ, respectively, and for all ̺ ∈ R. Let us define the multi-valued functions f : Ω × R → 2 R and f Γ : Γ × R → 2 R as follows: f (x, s) = ∂j(x, s, s), f Γ (x, s) = ∂j Γ (x, s, s). (6.2) For the so defined multi-valued functions the following lemma holds true. Proof. Hypothesis (F2) follows immediately from (J2). The proof of property (F1) is just a slight adaption of the proof of [11,Lemma 3.2], and therefore can be omitted.
To prove the reverse, let u be a solution of (6.1). In order to show that u is a solution of the multi-valued variational inequality (6.3), we are going to show that u is both a subsolution and a supersolution of the multi-valued variational inequality (6.3), which then by Proposition 6.1 and applying Theorem 1.8 yields the existence of a solutionũ of (6.3) satisfyingũ ∈ [u, u] and thus u =ũ completing the proof.
Discontinuous Multi-Valued Problems
In this section we study discontinuous multi-valued problems. For this purpose, let j : Ω × R × R → R and j Γ : Γ × R × R → R be given functions such that both are locally Lipschitz continuous with respect to the third argument. We denote by s → ∂j(x, r, s) and s → ∂j Γ (x, r, s) Clarke's generalized gradient of j and j Γ with respect to their third variable. Note that we do not suppose any continuity assumptions on r → j(x, r, s) and r → j Γ (x, r, s). This leads to multi-valued functions f : Ω × R → 2 R and f Γ : Γ × R → 2 R given by f (x, s) = ∂(x, s, s) and f Γ (x, s) = ∂j Γ (x, s, s). (7.1) Based on Proposition 2.9 we know that f : The precise problem is stated as follows: Find u ∈ K ⊂ W 1,H (Ω) and τ 1 ∈ C(Ω), where F and F Γ are the multi-valued Nemytskij operators generated by the multivalued functions given in (7.1), that is, : Ω → R : η is measurable in Ω and η(x) ∈ ∂j(x, u(x), u(x)) for a. a. x ∈ Ω} , F Γ (u) = {ζ : Γ → R : ζ is measurable on Γ and ζ(x) ∈ ∂j Γ (x, u(x), u(x)) for a. a. x ∈ Γ} . Remark 7.1. Note that (7.2) can be equivalently written in the form: η ∈ F (u), ζ ∈ F Γ (u) and Definition 7.2. Let Ω ⊂ R N , N ≥ 1, be a nonempty measurable set. A function f : Ω × R m → R, m ≥ 1, is called superpositionally measurable (or sup-measurable) if the function x → f (x, u 1 (x), · · · , u m (x)) is measurable in Ω whenever the component functions u i : Ω → R of u = (u 1 , . . . , u m ) are measurable.
(H3) The functions s → j(x, r, s) and s → j Γ (x, r, s) are locally Lipschitz continuous for all r ∈ R, for a. a. x ∈ Ω and for a. a. x ∈ Γ, respectively. The functions r → j • (r, s; 1) and r → j • Γ (r, s, 1) are decreasing for all s ∈ R for a. a. x ∈ Ω and for a. a. x ∈ Γ, respectively and the functions r → j • (r, s; −1) and r → j • Γ (r, s, −1) are increasing for all s ∈ R for a. a. x ∈ Ω and for a. a. x ∈ Γ, respectively. We have the following existence and enclosure result of extremal solutions for (7.2). Proof. First, we point out that the multi-valued functions f and f Γ defined in (7.1) are no longer upper semicontinuous and so the corresponding Nemytskij operators F and F Γ are not pseudomonotone in general. Hence, we cannot apply Theorem 2.12. Instead we will make use of a fixed point argument based on Theorem 2.13 combined with the existence and comparison results provided by Theorems 1.8 and 1.9.
Here s) and s → f w,Γ (x, s) stand for Clarke's generalized gradients and from Proposition 2.9 (iii) we know that these are upper semicontinuous. Furthermore, we denote by F v , F v Γ , F w and F w,Γ the multi-valued Nemytskij operators related to f v , f v Γ , f w and f w,Γ , respectively. We introduce the following auxiliary problems: Applying hypothesis (H3) we easily see that u, v are sub-and supersolutions of (7.3) and w, u are sub-and supersolutions of (7.4). Moreover, due to (H1) and (H2), the assumptions of Theorems 1.8 and 1.9 are satisfied. Therefore, there exist the greatest solution v * and the smallest solution v * of (7.3) within [u, v] and the greatest solution w * and the smallest solution w * of (7.4) within [w, u]. Furthermore, again by using (H3), we can show that v * ∈ [u, v] is a supersolution of (7.2) and w * ∈ [w, u] is a subsolution of (7.2). This can be shown as it was done in [5, Lemma 4.1].
Step 2: Definition of fixed-point operators We define the following sets: and v is a supersolution of problem (7.2) , W := w ∈ W 1,H (Ω) : w ∈ [u, u] and w is a subsolution of problem (7.2) .
Recall that v * ∈ [u, v] is the greatest solution of (7.3) and w * ∈ [w, u] is the smallest solution of (7.4), we know that the operators are well-defined due to Step 1. As done in [5,Lemma 4.2], using again (H3), one can show that G : V → V is an increasing operator, that is, v 1 ≤ v 2 implies Gv 1 ≤ Gv 2 . Similarly, T : W → W turns out to be increasing as well.
Step 3: Fixed-point argument Using again hypothesis (H3) we are able to show that the range G(V) of G has an upper bound in V and decreasing sequences of G(V) converge weakly in V. The proof is similar to the one in [5, Lemma 4.5] by using Propositions 2.5 and 2.7. Similarly, we show that the range T (W) of T has a lower bound in W and increasing sequences of T (W) converge weakly in W. Now we can apply Theorem 2.13 to T : W → W to get a smallest fixed point and to G : V → V to get a greatest fixed point. By Definition of G, u ∈ [u, u] is a fixed point of G if and only if u is a solution of (7.2). Similar can be said about T : W → W. This finishes the proof.
Construction of Sub-supersolution for a Multi-Valued Obstacle Problem
As an application of the results of the preceding sections, in this section we consider the following obstacle problem when F Γ = 0: u ∈ K : 0 ∈ Au + ∂I K (u) + F (u) in W 1,H 0 (Ω) * , (8.1) where A is the double-phase operator given by (1.1) and K is defined by where s → f 1 (x, s) is a (single-valued) lower semicontinuous function, s → f 2 (x, s) is an (single-valued) upper semicontinuous function, and x → f i (x, u(x)) is a measurable function for any measurable function x → u(x).
(Hf) There exist k i ∈ L r ′ 1 (·) (Ω), i = 1, 2, such that f 1 (x, s) ≤ k 1 (x) and f 2 (x, s) ≥ k 2 (x) for all (x, s) ∈ Ω × R. (Ω), see Proposition 2.5(i). Our existence and comparison result for the obstacle problem is as follows. Proof. We are going to make use of Theorem 1.8. To this end we are going to show that u := u 1 and u := u 2 + M with M ≥ 0 sufficiently large, are sub-and supersolutions of (8.1), (8.2), respectively.
One readily verifies that K given by (8.2) satisfies the lattice condition K ∧ K ⊂ K and K ∨ K ⊂ K.
Hence, as a conclusion of Theorem 1.9 we obtain the following characterization of the set S of all solutions of (8.1),(8.2) lying within the order interval [u, u]. (Ω) and possesses a smallest and a greatest solution. | 2022-01-11T02:15:51.157Z | 2022-01-08T00:00:00.000 | {
"year": 2022,
"sha1": "f6aedee2057383891718d4a8c38172440aeb6d73",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "f6aedee2057383891718d4a8c38172440aeb6d73",
"s2fieldsofstudy": [
"Mathematics"
],
"extfieldsofstudy": [
"Mathematics"
]
} |
259010646 | pes2o/s2orc | v3-fos-license | Design and Implementation of Inductively Coupled Power and Data Transmission for Buoy Systems
: Moored buoys are important components of stereo platforms for ocean observation, which are crucial in underwater exploration. In complex marine environments, power supply and data transmission between moored buoys and underwater sensors are difficult. To solve these problems, an inductively coupled power and data transfer (ICPDT) scheme based on LCCL-S-LC hybrid compensation is proposed. The power transmission was analyzed by establishing an LCCL-S-LC compensation buoy ICPDT system model. The system efficiency and output power were analyzed when the load changed, and the optimal load resistor for maximum system efficiency was determined. A modulation and demodulation circuit used for data transmission was introduced, the compensation topology parameters of each loop of the buoy ICPDT system were deduced, and the crosstalk between power and data was analyzed and reduced. An ICPDT system prototype was built to verify the system’s feasibility and effectiveness when it was powered by 24 V. The LCCL-S-LC topology reduced the interference between data and power transmission. When the measured output power of the system was 61.5 W, the power transmission efficiency was 78.1%, and the data receiving end could achieve correct demodulation when the transmission rate was 100 kb/s.
Introduction
Moored buoys are pieces of marine environment monitoring equipment that use mooring cable to ensure that surface buoys and their underwater anchors remain relatively fixed. With the advantages of long service lives and the ability to host multiple sensors, moored buoys are widely used in marine monitoring [1][2][3]. Various sensors mounted on mooring cables are used to obtain physical information, such as conductivity, temperature, salinity, etc., at different profiles. For power supply, underwater sensors rely on battery packs. However, battery packs must be regularly replaced as they run out, which increases the deployment costs [4,5]. The traditional wired charging method experiences problems with sliding wear and bare connections, which makes it unsuitable for charging underwater sensors. Inductively coupled power and data transmission (ICPDT) is an attractive technology that can simultaneously achieve battery charging and data transmission. Therefore, in this study, we focused on ICPDT technology.
ICPDT systems can realize inductively coupled power transmission from a power source to a load; they can also send instructions to sensors to obtain data. Compared with traditional wired transmission, they do not experience problems with wire-type transmission cable wear or exposed joints. Due to these advantages, ICPDT systems are widely used in electric vehicles, underwater AUVs, etc. [6][7][8][9][10][11]. Compared with traditional ICPDT systems, moored buoy ICPDT systems have the following characteristics: First, moored buoy ICPDT systems are applied for the power supply and data transmission of water buoys and underwater sensors. Underwater sensors are arranged in various sections below sea level to collect temperature and salinity data and are usually several meters away from the buoys for energy and data transmission. The coupling coefficient of traditional primary and secondary coupling coils is low when the distance is too far, and the power transmission power is low or may even be zero [12][13][14]. A contactless power supply system (CLPS) using a cone-type coil was proposed for underwater vehicles [15], which could transmit 500 W of electric power with approximately 96% efficiency when the diameter of the coil was 48 mm. Second, when the closed loops formed by mooring cables are used for power and data transmission, the influence of the resonance formed by the mooring cables on the resonance frequency of the power and data transmission needs to be evaluated, and the loss of data signals when they flow through mooring cable circuits needs to be considered to ensure the efficient transmission of power and the high-speed transmission of data. Third, mooring buoy ICPDT systems use mooring cables as intermediate circuits, allowing for power and data transmission without changing the sensor positions, which reduces the burden of the buoys' mechanical structures.
With the development of ICPDT technology, power transmission technology has become relatively mature, and we focus on how to improve the data transmission performance. At present, the four main methods in ICPDT technology are as follows: The first technique uses two pairs of coupled coils to separately transmit power and data [16][17][18]. The power and the data transmission coils are separate from each other, which reduces the crosstalk between the data and power transmissions. However, adding a pair of coupled coils increases the system's size and weight, and the separation of the two pairs of coils further increases the system volume The second technique transfers power and data by changing the inverter input voltage or load [19][20][21]. Researchers [19] proposed a modulation and demodulation method based on load transfer keying (LSK), which demodulates data by changing the current of the load monitoring/receiving coil, and the voltage information of the load resistance is transmitted to the primary coil to adjust the primary input. The voltage is used to control the load power, and the load output power is approximately 1.1 W. The efficiency of this system is only 60% due to fluctuations in the power received by the load caused by the changes in the input voltage on the primary side. Adjustments to the input voltage and changes in the load lead to changes in the output power, which directly affect the output power and efficiency of the system. The frequency of data transmission is usually lower than the resonant frequency of power transmission, and the data transmission rate is not high. The third technique uses carrier-based ICPDT technology [22][23][24][25], which superimposes data signals onto high-frequency carriers and then superimposes data waveforms onto power waveforms using on-off keying (OOK). Power waveforms are transmitted to the secondary side through electromagnetic coupling; then, the data signals are extracted from the power signals for demodulation. In this case, power transmission has a stronger impact on data transmission, so analyzing and reducing interference are difficult, which increases the difficulty of data demodulation. The fourth technique is the 2FSK -modulation-based ICPDT technique [26,27]. The data transmission frequency based on 2FSK modulation is usually an order of magnitude higher than the power transmission frequency. The interference between data transmission and power transmission is minimal and can be ignored. Most of the interference between power transmission and data transmission can be removed by adding filters. Frequency division multiplexing can also be implemented.
The data transmission mode based on FSK modulation and demodulation has the following advantages: Compared with independent electric energy and data transmission circuits, the circuit structure of the 2FSK method is simple, and data can be transmitted by simply injecting signals into electric energy transmission circuits. The FSK modulation and demodulation frequencies are generally higher than the power transmission frequency, so the data transmission rate can be very high. In addition, the FSK method has a high SNR and a low bit error rate.
In the air, as a medium, ICPDT technology has been applied in various situations, such as the wireless charging of electric vehicles, implantable medical devices [27,28], constantcurrent or constant-voltage inductive charging of UAVs [29], as well as wireless charging of mobile phones, computers, and other devices [30][31][32]. However, inductively coupled power and data transmission systems with intermediate loops in the marine environment have rarely been reported in the literature.
To ensure adequate energy reception and simultaneous data transmission for underwater sensors, in this study, we employed frequency division multiplexing technology and developed an inductively coupled power and data transmission system with shared coupling channels. Through an analysis of the characteristics of four data transmission modes, 2FSK modulation and demodulation technology, known for its high signal-to-noise ratio (SNR) and high data transmission rate, was selected as the preferred data transmission mode.
The main contributions of this study are summarized as follows: (1) The moored buoy ICPDT system model based on the LCCL-S-LC hybrid compensation was established and analyzed in detail, including the power and data transmission modules. (2) The 2FSK modulation and demodulation technology was applied to the buoy data transmission. The demodulation circuit adopts an NE564 module with adjustable loop gain. A notch filter composed of L1/L3 was designed to reduce the interference between power and data transmission. (3) The reliability of the proposed system was verified through experiments, where we observed that, at an output power of 61.5 W, the power efficiency was 78.1%. Additionally, the data receiving side successfully demodulated data at a rate of 100 kb/s. The rest of this paper is organized as follows: Section 2 presents the system structure of ICPDT, and the equivalent circuit mode is established and analyzed. In Section 3, we analyze the crosstalk between power and data and propose methods to reduce the interference. In Section 4, we verify the power and data transmission characteristics of the proposed moored buoy ICPDT system through experiments. The conclusions are drawn and discussions are outlined in Section 5.
System Structure
Although many researchers have studied traditional ICPDT system modeling, the modeling and analysis of the three-loop moored buoy ICPDT system is relatively complex. Moored buoy ICPDT systems are mainly composed of three circuits: the above-water, the mooring cable, and the underwater circuits. The overall structure of the system is shown in Figure 1. A solar panel is installed on the upper part of the buoy body, which provides electricity for the internal battery. An above-water coupler forms a closed loop with the underwater control center's coupler through a mooring cable. Inductive coupling principles are employed for both power and data transmissions in this configuration. The underwater sensor is usually a temperature and salinity depth sensor (CTD) or a Doppler current profiler (ADCP). The mooring cable is not only the load-bearing cable of the buoy but can also be used as a closed intermediate loop for power and data transmission, providing good rigidity and stability. The data received by the above-water system can be transmitted to iridium through the satellite antenna, and the shore-based personnel use an upper computer to receive the data and perform the next step. Figure 2. The DC voltage output by the abovewater power supply system is converted into AC voltage U d through an inverter circuit, and the operating frequency of the inverter is f p . R 1 is the AC equivalent resistance of the above-water circuit; L f 1 , C f 1 , and C 1 are the primary series compensation inductance, parallel compensation capacitor, and series compensation capacitor, respectively; L 1 is the primary wave trap; L p1 andL s1 are the primary and secondary inductance of the abovewater coupling magnetic ring, respectively; L 2 is the self-induction of the mooring cable; C 2 is the series compensation capacitor of the mooring cable circuit.; R 2 is the AC equivalent resistance of L s1 , L 2 , and L p2 . L p2 and L s2 are the primary and secondary inductance of the underwater coupling magnetic ring, respectively; L 3 is the secondary wave trap; C f 3 is the series compensation capacitor of the underwater loop; D 1 , D 2 , D 3 , and D 4 are the four diodes forming a full-bridge rectifier circuit; C F is a filter capacitor; and RL is load. The blue short dotted line represents the data transmission circuit; U md is the modulation data; R in1 and R in2 are the series and parallel resistance, respectively; L dp1 and L ds1 are the primary and secondary inductance of the data injection magnetic ring, respectively; C dS1 forms resonance with L ds1 when the data carrier frequency is f d . The blue dotted line is surrounded by the data receiving circuit. R out is the output resistance of the data receiving circuit; C dp2 is the parallel compensation capacitor; L dp2 and L ds2 are the primary and secondary inductance of the data extraction magnetic ring, respectively; C ds2 forms resonance with L ds2 at f d .
The data demodulation circuit is shown in Figure 3 and consists of 4 modules. The first stage is amplifier Amp1, which can amplify the waveform of the data extraction magnetic ring to an appropriate multiple that can be demodulated. The band-pass filter is composed of RLC in series and parallel. The low-and high-frequency noise is filtered out. Amplifier Amp2 in the third stage is used to amplify the filtered carrier signal; the fourth stage is the FSK demodulation module, which is composed of NE564. U dd represents the final demodulated data.
Power Transmission Analysis
The power transmission of an ICPDT system was analyzed with an equivalent circuit. The system compensation structure adopts LCC-S-LC compensation. To simplify the calculation, the inverter circuit is equivalent to an AC voltage source, and the full-bridge rectifier circuit and electronic load are equivalent to pure resistance R L , which is represented by (1). The simplified circuit is shown in Figure 4, where U d is the output voltage of the inverter , and the frequency is f d , which is represented by (2). The circuit was analyzed by Kirchhoff's law. In the above-water circuit, the series impedance of C 1 , L 1 , and L p1 can be expressed as: Then, the parallel impedance of C f 1 and Z 12 can be expressed as: In the underwater circuit, the series impedance of L 3 , R 3 , C f 3 , and R E can be expressed as: The equivalent impedances Z 1 , Z 2 , and Z 3 of the above-water, mooring cable, and underwater circuits can be expressed as: According to the principle of mutual inductance impedance transformation, the reflection impedances Z 23 and Z 12 can be expressed as: In (7), Z 23 is the impedance converted from the underwater loop to the mooring cable loop, and Z 12 is the equivalent impedance converted from the impedance of the mooring cable loop to the above-water loop.
M 1 and M 2 are the mutual inductances of the above-water coupled magnetic ring and the underwater coupled magnetic ring, respectively. In the experiment, an LCR meter (TH2840B Precision LCR Meter) was used to measure the self-induction and mutual inductance values of the electromagnetic coupler, and we calculated k1 and k2 using the following equation: ω p is the resonant angular frequency of the power transmission system, which is tuned as: Equation (9) describes the conditions that the angular frequency needs to meet when the power transmission system works in the resonance state, where f d is the frequency of power transmission.
By simultaneously solving Equations (1)-(9), the currents of the above-water, the mooring cable, and the underwater circuits can be obtained as: In (10), I 1 , I 2 , and I 3 are the power transmission channel currents of the above-water , the mooring cable, and the underwater circuits.
From (10), the load voltage can be obtained as: The output power and system efficiency can be obtained as: P I N represents the input power of the system. After determining the value of the other components, take the derivative of (12) and (13). When RL is approximately 10.3 Ω and 15.2 Ω, the output power and efficiency reach the maximum value.
After the output voltage is obtained, the transfer function of power transmission can be obtained: The transfer function of power transmission is related to the mutual inductance of the magnetic ring, which is essentially related to the coupling coefficient. The amplitudefrequency response curve of the power transmission shown in Figure 5 can be obtained after substituting the parameters of the system into (14). Figure 5 shows that the power transmission gain has a peak value at 24 kHz, which is the resonant frequency of the power transmission. With the increase in the coupling coefficient, the voltage gain of the load shows an upward trend; that is, the larger the value of the mutual inductance of the power transmission magnetic ring, the greater the power received by the load end.Therefore, a larger coupling coefficient was selected for the experiment. Figure 5 also shows that the load voltage gain has a double-peak characteristic; that is, two maximum points can be seen, which is called the frequency splitting phenomenon of the ICPDT system. This occurs due to the peak separation phenomenon, which is caused by the system being in an over-coupling state. This aspect was not the focus of this study. After determining the working frequency of power transmission, the influence of frequency splitting on the ICPDT system designed in this study could be ignored.
The gap of the coupling ring and coupling coefficient k 1 and k 2 can be determined after determining the influence of the coupling coefficient on the power gain. When determining the values of the other components, the output power and efficiency of the system are related to the load R L , and the output power curve and efficiency curve under the change in R L were drawn using Mathematica, as shown in Figure 6. 6. Curve of (a) system output power and (b) system efficiency with load changes. Figure 6 shows the change curve of the system output power with the load. The load R L continues increasing, and the output power shows a trend of first increasing and then decreasing. When the load is 10 Ω, the maximum output power of the system is 76 W, and, when the load is 15 Ω, the system efficiency reaches a maximum of 79%. For the proposed ICPDT system, high output power and efficiency can be guaranteed by selecting the appropriate load value.
Data Transmission Circuit Design and Analysis
The data transmission circuit was analyzed after paralleling the data transmission channel in the power transmission channel. First, the function of each device in the ICPDT system is described. As shown in Figure 7, L f 1 , C f 1 , and L 1 form a third-order low-pass filter, which can filter out the higher harmonics of the inverter output voltage U d . The interference of U d to data transmission is considerably reduced, thereby reducing the bit error rate of data transmission. L 3 and C f 3 form a second-order low-pass filter, which can reduce the influence of the high-order harmonics of the input voltage of the rectifier circuit on data demodulation. Notch inductors L 1 and L 3 are used to increase the impedance of paths B1 and B2 and increase the data flow of paths A1 and A2. The more the data pass through the data extraction circuit, the stronger the signal received by R out , and the easier the data demodulation becomes. If L 1 and L 3 are not used, most of the data signal flows in B1 and B2, the signal received by R out becomes weaker, and demodulation is more difficult. Due to the existence of L 1 and L 3 , the part of the data signal flowing to the power channel can be ignored in the subsequent analysis of the data transmission circuit.
C ds1 and L ds1 resonate at f d ; C ds2 and L ds2 also resonate at f d . f d is the resonant frequency of data transmission, and f p is the resonant frequency of power transmission. Under the condition that f d satisfies (15), the influence of power transmission on data transmission can be substantially reduced. L ds1 , L dp1 and L ds2 , L dp2 are used as the primary and secondary stages of the data transmission magnetic ring, respectively, and their volume should be reduced as much as possible in the design, so the value should not be too large.
The power of data transmission should be controlled within a certain range to reduce the interference of data transmission to power transmission. Therefore, two conditions should be comprehensively considered to determine the values of L ds1 , L dp1 and L ds2 , L dp2 . R in1 is used to adjust the amplitude of the data signal and limit the power consumption of the data transmission circuit. R in2 is used to remove the voltage spikes generated during the frequency conversion of the two carriers during 2FSK modulation. Among them, the relationship between f d and f p is shown in (15).
The impedance of each loop of the data transmission channel is simplified to obtain the equivalent impedance of each part of the data transmission circuit: where Z 4 and Z 5 are the primary equivalent impedance of the data injection circuit and the secondary equivalent impedance of the data extraction circuit, respectively. Z dt and Z dr are the equivalent total impedances of the data injection and extraction circuits, respectively. According to the principle of mutual inductance impedance transformation, the expressions of Z 35 , Z 23 , Z 12 , and Z 14 are obtained as follows: In (17), Z 35 is the reflected impedance of the data-receiving circuit to the underwater circuit, Z 23 is the reflected impedance of the underwater circuit to the mooring cable circuit, Z 12 is the reflection impedance of the mooring cable circuit to the water circuit, and Z 14 is the reflection impedance of the water circuit to the data transmission circuit. M 3 and M 4 are the mutual inductances between the data-sending and data-receiving magnetic ring, which are represented by the following equation: Based on Kirchhoff's law, the currents of each loop of the data transmission circuit can be obtained as: From (19), we can deduce that the voltage and power received on R out in the data extraction circuit are, respectively: Then, the transmission gain G data of the data carrier can be expressed as: Figure 8 shows the amplitude-frequency response curve of the data forward transmission gain. On the basis of determining the power transmission frequency of 24 kHz, the data transmission gain achieves the maximum value between 1 and 2 MHz. The difference between the power transmission frequency of the ICPDT system and the data transmission frequency is more than 10 times. This power transmission circuit can be equivalent to a short circuit at the data transmission frequency. At this time, the data transmission circuit does not occupy too much load receiving power. The overall efficiency of the system is not substantially reduced. Therefore, 1.3 MHz and 2 MHz were selected as the 2FSK carrier frequencies. The values of L 1 and L 3 are determined based on the data carrier transmission gain G data defined by the above equation. The fourth part produces the variation curve of G data with respect to L 1 and L 3 . By selecting suitable L 1 and L 3 , the interference of power transmission with data transmission can be reduced while obtaining a larger data transmission gain.
Data Transmission Interference with Power Transmission
The influence of data transmission on power transmission is shown as the change in R L received power in the following two states: In the first case, the data transmission loop does not transmit data. In the second case, the data and power are simultaneously transmitted. The power received by R L in the two cases is analyzed below. Figure 9 shows a simplified circuit for crosstalk analysis when only power is transferred, corresponding to the first case . Z dt and Z dr are the total impedances of the datatransmitting and data-receiving circuits, respectively. The ratio of the data transmission frequency to the power transmission frequency is defined as γ , and the expression is: In the case of L ds2 = L dp2 and γ ≥ 10, Z dr can be approximated as follows: To reduce the impact of data transmission on power transmission, the modulus of Z dr is at least one order of magnitude larger than that ofZ 31 , namely: Equation (25) is one of the conditions that L ds2 needs to satisfy. At this time, the parallel impedance of Z dr and Z 31 can be approximated as Z 31 , and the expression of Z pp can be derived from: In the case of L ds1 = L dp1 = L ds2 = L dp2 and γ ≥ 10, Z dt can be approximated as follows: Similar to underwater loop analysis, the modulus of Z dt is at least one order of magnitude larger than that of Z pp , namely: Equation (28) is another constraint that L ds2 should satisfy. Under the two constraints, when only power is transmitted, the influence of the data transmission channel on power transmission can be ignored.
The following is an analysis of the simultaneous transmission of data and power. To simplify the analysis, the simultaneous transmission of data and power is divided into the superposition of two cases: Case 1: only U d is working; Case 2: only U md is working. Case 1 is consistent with the above analysis. Case 2 is analyzed below. Figure 10 depicts a crosstalk analysis circuit when only U md works, where Z dr is the equivalent impedance of the data receiving circuit, Z 32 is the equivalent impedance of the power receiving circuit, and Z dr and Z 32 are both calculated at f d , which are expressed as: When the impedances of L ds2 , L 3 , C f 3 , R L , and R 3 are in an order of magnitude, Z 32 is simplified to: In (30), Z 32 is approximately a pure inductance. Therefore, only when U md works is the power received by R L very small; the input power at U md is also relatively small, most of which is absorbed by R in1 and R in2 . The power received on R out and R E is very small. Therefore, when power and data are transmitted at the same time, the interference of data transmission with power transmission can be ignored.
Power Transmission Interference with Data Transmission
The interference of power transmission with data transmission is divided into two parts: the low-frequency (at f p ) noise received by R out and the high-frequency (at f d ) noise received by R out . The noise is called low-and high-frequency noise in this paper.
The low-frequency noise is filtered out at the data-receiving loop by a band-pass filter. f d is at least one order of magnitude higher than f p . Therefore, the LC resonance can attenuate the low-frequency noise, and only a small part of the low-frequency noise can flow into R out . As shown in Figure 3, a high-performance LC-BPF is added to the demodulation circuit to filter out the low-frequency noise.
Compared with low-frequency noise, high-frequency noise has a more serious impact on data transmission. Most high-frequency noise comes from the harmonics where the power transmission frequency is close to f d . Because the harmonic frequency is close to the data transmission frequency, filtering out the interference signal and restoring the original signal through the filter are difficult. Therefore, a notch filter composed of L 1 and L 3 was designed in this study. By selecting the appropriate values of L 1 and L 3 , the interference of power transmission in data transmission can be minimized.
We mainly analyzed the harmonics at f d . Because the harmonics of the square wave are all odd harmonics, the γ±1 harmonics were analyzed. The voltage gain G AB (γ±1) is defined as: U Rout (γ±1) is the voltage received by R out when only U d works, that is, the power transmission interference voltage received by the data-receiving end. The variation curve of G AB (γ±1) with respect to L 1 (L 3 ) was obtained with the numerical analysis software of Mathematica. The curve of data transmission gain with L 1 (L 3 ) was also obtained. Table 1 shows the value of the circuit components of the ICPDT system. Figure 11 shows the variations in data transmission gain and interference voltage gain with L 1 (L 3 ). The values of G AB (γ−1) and G AB (γ+1) are amplified 100-fold. The red and green curves represent G AB (γ−1) and G AB (γ+1), respectively. The blue curve represents the data transfer gain. The figure shows that, with the increase in L 1 (L 3 ), the data transmission gain continues to increase. Considering the coil volume limitation, L 1 (L 3 ) must meet the power resonance frequency on the basis of ensuring higher data transmission gain. Therefore, 13 µH was selected as the value of L 1 (L 3 ), and the data transmission gain at this time is approximately 0.4. Additionally, a signal amplification circuit is added to the data demodulation side, which can ensure the normal demodulation of data.
Experimental System Structure
According to the aforementioned schematic diagram of the ICPDT system, an experimental platform was built, as shown in Figure 12. The design parameters are shown in Table 1. The system was mainly composed of a DC voltage source, oscilloscope, electronic load, DSP controller, STM32 microcontroller, inverter circuit, data transmission circuit, above-water circuit, mooring cable circuit, underwater circuit, data-receiving circuit, data demodulation circuit, and rectifier circuit. These parts are marked in Figure 12. The DSP controller was used to generate four complementary output PWM signals. The dual frequency modulation signal was implemented using an STM32 microcontroller. Seawater with salinity of 3.2% in the glass tank was taken from the Yellow Sea, China. Table 2 shows the parameters of the coupling magnetic ring in air and seawater. This table shows that the parameters of the coupling magnetic ring in air and seawater only slightly varied and could be ignored. To simulate the real marine environment, we also selected the sea water of the Yellow Sea, China for experiment.
Experiment Results of Power and Data Transmission
When power and data are transmitted at the same time, the voltage and current of inverter output and load receiving are as shown in Figure 13. To verify the output power and efficiency of the ICPDT system when the load changes, experimental tests were conducted by changing the load value. The curve of power and efficiency with load shown in Figure 14 was drawn. The experimental results showed that, when the load value was 15 Ω, the input power of the system was 78.7 W, the output power was directly read from the electronic load, and the result was 61.5 W. The system efficiency of the calculation system from DC input to DC output is 78.1%. The measured output efficiency was similar to the designed output efficiency, which verified the reliability of power transmission. As shown in Figure 14a, with increasing load, the output power increased first and then decreased. When R L = 10 Ω, the system output power reached the maximum value of 61.5 W. When the load changed from 5 Ω to 30 Ω, the output power of the system was always above 40 W, which meets the charging demand of the battery carried by the underwater sensor. Figure 14b shows that, with the constant change in load, the efficiency of the ICPDT system remained above 60%. When R L = 15 Ω, the system efficiency reached the maximum of 78.1%. Because of the existence of the cable loop, the system power transmission from the above-water loop to the underwater loop requires two inductive coupling transmissions. Because the coupling coefficient cannot be one, the efficiency loss of this system is higher than that of the traditional ICPDT system. Figure 15 shows the data modulation and demodulation signals when only data are transmitted. The binary symbol signal was sent with a period of 11001110. The demodulation conditions were tested when the transmission rate was 50 kb/s, 100 kb/s, 150 kb/s, and 200 kb/s, separately.
The experimental results showed that the demodulated data were the same as the original data. The data demodulation signal generated a certain delay, and no error arose in the demodulated signal due to the delay. The designed ICPDT system could demodulate the data signal with a maximum rate of 200 kb/s when the data channel worked alone. Figure 16 shows the data modulation and demodulation signal waveforms when power and data were transmitted at the same time. Due to the addition of power transmission, the amplitude of the carrier signal was seriously affected by the power signal. The frequency corresponding to transmission digital signal "1" is 2 MHz, and the carrier frequency of the digital signal "0" is 1.3 MHz. Figure 15a,b shows that the data were well recovered on the demodulator when the data transfer rate was 50 kb/s and 100 kb/s. As shown in Figure 15c,d, when the data transfer rate was further increased to 150 kb/s and 200 kb/s, the demodulated signal produced a serious error code. Therefore, when the power and data are simultaneously transmitted, the data signal can be transmitted at a rate of up to 100 kb/s and normally demodulated. Figure 16 shows that, compared with data transmission alone, the amplitude of the data carrier considerably changed after the power transmission was added, with a notable fluctuation with the power transmission frequency. Although we implemented measures to reduce the interference of the power with the data transmission in the experiment, completely filtering out one of the mixed waveforms of the power and data was difficult. Some electric energy waveforms are inevitably mixed in the data transmission channel. To enable the normal demodulation of the signal, an operational amplifier with adjustable magnification was added to the front of the demodulation circuit, and the normal demodulation of the signal was realized by adjusting the magnification. The selected 2FSK demodulation chip was an NE564, which has good bandwidth and signal-to-noise ratio, which further reduces the difficulty of signal demodulation.
In the ICPDT system with the same structure, the data transmission rate was compared with that reported in other papers. Table 3 shows that the ICPDT system proposed in this paper has a higher transmission rate and transmission power than those in [28,29,33]. Finally, the current change when the load changes was verified, as shown in Figure 17. By changing the value of load resistance, the current value of the R L receiving terminal was measured. When R L fluctuated in a small range, the current received by the load was basically maintained at approximately 1.5 A; that is, the current was almost stable within a certain range without additional control of the load. This feature can ensure the stability of battery charging and prolong the service life of the battery. Increasing the running time of the sensor carried by a moored buoy is important.
Conclusions
This paper proposes an ICPDT system to enable the power and data transmission between a mooring buoy and underwater sensors, and the ICPDT system was analyzed and designed in detail. The crosstalk between power and data was analyzed, and power and data were successfully transmitted. The experimental results showed that the DCto-DC efficiency was 78.1%. Under load fluctuations, the load receiving current did not considerably change, and the load power could be maintained at a high level. Additionally, the data signal could be normally demodulated when the transmission rate was 100 kb/s. By changing the values of the data gain resistance, the amplification factor of the data transmission loop could be easily adjusted, and the power of the data transmission loop could be limited, thereby improving the efficiency of power transmission. This study provides a reference for the subsequent research on the parallel transmission systems of buoy power and data. On the basis of our findings, we can further explore how to realize the closed-loop control of the whole system. Author Contributions: Methodology, J.X. and X.L.; Software, S.P.; Writing-original draft, X.C.; Writing-review & editing, X.C., J.X. and S.P.; Project administration, X.L. and H.L.; Funding acquisition, X.L. and H.L. All authors have read and agreed to the published version of the manuscript. Institutional Review Board Statement: The study was not involving humans or animals.
Data Availability Statement: Data is unavailable due to privacy. | 2023-06-02T15:18:43.076Z | 2023-05-30T00:00:00.000 | {
"year": 2023,
"sha1": "42eb5f2b0a32a4e33c734b2acdac1d46efaa4129",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.3390/en16114417",
"oa_status": "GOLD",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "5107907eae6a81522d286c80ab1c31d46f8ec187",
"s2fieldsofstudy": [
"Engineering",
"Environmental Science"
],
"extfieldsofstudy": []
} |
249200959 | pes2o/s2orc | v3-fos-license | Homage to Ebola Fighters: Black Labor and Humanitarian Media Campaigns
Abstract During the Ebola outbreak that hit Guinea in 2014, most of the people employed at the Wonkifong Ebola treatment unit were from Africa or Cuba. Despite the recruitment of black personnel, the unit exposes how the humanitarian infrastructure exploited Guinean workers as if their lives were less vulnerable than those of the foreign personnel. The Africanization of aid reveals a post‐colonial segregation at the intersection of race, class, and locality. The article follows Guinean workers in the quarantine unit, as well as their enrolment in media campaigns. Their experience illuminates a triage at the core of Global Health according to which not only were local workers treated as expendable lives, but their stories were silenced. Yet how did Guinean workers inhabit this anti‐black world? The article unfolds the journey of workers during the outbreak and three years later, exploring the strategies they used to produce their own narratives through personal archives. [humanitarian aid, humanitarian media campaigns, race, Ebola, archives]
"I gave him life. It is quite a lot to give. It is the opposite of nothing. And the opposite of nothing is not something. It is everything." - (Kushner 2018) A Visit to an Old Friend In 2018, I was on my way back to Guinea, where I had conducted fieldwork inside an Ebola treatment unit during the massive outbreak that hit the country in 2015 (Gomez-Temesio 2018, 2020b; Gomez-Temesio and Le Marcis 2017). When I had left, several buildings of the capital were covered with portraits of health workers committed to the fight against Ebola. Three years later, though, all these traces of the outbreak had disappeared. A few hours after landing, I was driving to Wonkifong in the district of Coyah, to the former Ebola unit. I had wanted to visit what was left of the unit with Sory, 1 a former Guinean health promoter who used to work there. It was now an abandoned site, he had told me on the phone a few days earlier. Indeed, though the facility had closed only three years before, already it seemed a lifetime ago. There were no roofs or walls, just a concrete slab partly covered by the lush vegetation of Guinean countryside (see Photo 1).
Photo 1. Ruins. Wonkifong Ebola Treatment Unit, 2018. [This figure appears in color in the online issue] Sory looked worried. He had been unemployed since the epidemic reached its end. In 2018, though the virus was no longer an active threat, it outlived that period as a kind of failed promise. Clearly, the afterlife of Ebola did not match the future Sory dreamed of when he enrolled in 2014, having decided to risk his own life to save others. He felt that this place could have represented a future for his past commitment. "I wish something had been done here. It could have been a health center. I could have worked here." He seemed depressed. Suddenly though, he became infused with a residual energy. Jumping on a slab of the former morgue, Sory began to rehearse in mime all the routines that had flowed over that space a few years before. "When a body came, we put him there. Do you remember every time a TV crew came around here? They only got the big shots [the doctors]. But I had so much to tell! Would you film me now as you did when we worked together?" Indeed, during fieldwork, I used my phone to film the unit's daily life. I started modestly, just to keep a trace of the infrastructure. But then Sory turned out to be inclined to star in what he called "my movies." And I began to film and take pictures of him more regularly (see Photo 2).
Wandering among the ruins, Sory remembered one film that took place during a chaotic day in 2015. Not only had a patient infected with Ebola escaped from the isolation zone, but he had been overcome by diarrhea just in front of the staff quarters. The foreign personnel-Cubans, Congolese, and a few Europeans-complained Photo 2. The former morgue. Wonkifong Ebola Treatment Unit, 2018. [This figure appears in color in the online issue] about the mess to the Guinean director. They threatened to lodge a complaint with the World Health Organization (WHO), who officially ruled the place in collaboration with the national government. The patient escaped through a door that had been damaged by an ambulance. Still, the director ordered Sory to clean up the feces. Some workers laughed while he was performing this dangerous task. He shouted at the mocking audience and asked me to start filming: "What is it that you don't understand? I am ready to sacrifice. You do know Henri Dunant? He founded the Red Cross. I live by his ideal. I have volunteered many years at the Guinean Red Cross. I don't care about doing this." Sory wandering among the ruins of Ebola embodies the focus of a long-lasting ethnographic project that brought me to Guinea many times between 2015 and 2018. Following Sory's experience during the outbreak-as well as his own souvenirs of his journey-this article explores anti-blackness in humanitarian aid through the lens of risk and testimony. Not only were Guinean workers treated as expendable lives, but their stories were also neglected-censored or confined to personal archives. While the humanitarian "localization agenda" (Duclos et al. 2019) and more broadly the "Africanization of aid" (Issoufou 2018;Redfield 2012) increasingly push toward hiring local personnel instead of Western expatriates in relief projects, the risks these workers are exposed to dramatically increase without their benefiting from the same incentives as white/Western employees. But as Sharpe warns us, the "non-ending subjection" of blackness also relates to the treatment of black experiences in the public sphere (Sharpe 2016). In 2014, Time magazine elected the "Ebola fighters" personality of the year. 2 It seemed then that people who were ready to sacrifice themselves were enthusiastically celebrated. But is this true for everyone? Here, I address the post-colonial reconfiguration of race in humanitarian aid in two different segments of humanitarian infrastructure: within the mundane details of daily aid relief and within the organization of media campaigns.
Humanitarianism contains within it the most banal and yet unutterable truth. Not all lives have the same value; some can be risked, while others can only be saved. While the literature on humanitarian aid mainly explores the differential value of life between aid workers and aid beneficiaries (Fassin 2007(Fassin , 2009(Fassin , 2018Packard 2016;Redfield 2013), little has been written on local aid workers and even less on the role of race (Benton 2016a(Benton , 2016b. This article, among others (Hirsch 2019(Hirsch , 2021Issoufou 2018;Kingori and McGowan 2016), addresses this gap, shedding light on the enrolment of local workers in Guinea during the Ebola outbreak, the increased risk to which they were exposed, and their limited access to humanitarian media campaigns (see Photo 3).
In the Wonkifong Ebola treatment unit, where I conducted fieldwork in 2015, 3 tasks inside the unit were segregated according to race and locality. The unit had been built by the World Food Program as part of the UN commitment to fighting the outbreak. At its inauguration in December 2014, it was the largest isolation infrastructure in the country and could accommodate up to 100 patients. The Wonkifong unit grew out of a South/South collaboration under the auspices of the WHO. Inside the unit, Guinean health workers collaborated with health teams from the Democratic Republic of the Congo (DRC), deployed by the Economic Community of West African States and the African Union, as well as a Cuban medical brigade hired by the WHO. A French nurse and a Haitian logistics specialist were hired Photo 3. South/South collaboration. Wonkifong Ebola Treatment Unit, 2015.
[This figure appears in color in the online issue] to control functional administration and a European lab processed all the testing. The foreign staff thus consisted mainly of doctors, nurses, and epidemics experts. Guinean workers, meanwhile, were mostly employed as hygiénistes, or aides. They were assigned the most dangerous tasks, such as disinfecting contaminated objects or handling corpses. Most of these workers had no proper medical or humanitarian training. They were mainly male and female youths from the region, unemployed before the outbreak.
That chaotic day in 2015, once he had finished cleaning up after the escapee, Sory told me: "You know it's true, I always wanted to study medicine. As an orphan, I had to quit school. But my dream was to save lives." Sory was bitter. Earlier that week, journalists had come from Cuba to film a news story. As the journalists mainly filmed their own compatriotas, Guinean workers only appeared in the background and were asked questions such as "How do you feel about Cuban citizens who have come to Africa to risk their lives?" Media campaigns took place regularly in the unit. They were mostly conducted by UN agencies or else were initiatives taken by one of the medical contingents.
Advocating for the victims of the Global South is key to the rise of what Fassin identifies as the second age of humanitarianism, introducing an important distinction into the public arena between "those who are subjects (the witnesses who testify to the misfortunes of the world) and those who can exist only as objects (the Photo 4. Sory cleaning after the escapee. Wonkifong Ebola Treatment Unit, 2015. [This figure appears in color in the online issue] unfortunate whose suffering is testified to in front of the world)" (Fassin 2007: 517). Nevertheless, when Fassin identifies two types of humanity, he leaves out how race shapes the ability to testify at the heart of the humanitarian apparatus: among its own employees. Indeed, during the outbreak, humanitarian media campaigns had one thing in common. They hardly ever gave voice to local aid workers. Nevertheless, my argument here does not steer toward a final account of exclusion. The archive, writes Garcia, may "open up the possibility of new historical narratives and modes of subjectivity" (2016: 575). Exploring daily life in the unit, I investigated how Sory and his coworkers were able to tell their own stories through alternative mediums and in intimate circles, and notably through what I would describe as private archives: recorded interviews, pictures and private movies, together with material memories from the unit that represented for them proofs of their involvement (see Photo 4).
An archive, continues Garcia, is "like a place for things you want to save. … Not like the self-storage units people rent … but more like a photo album" (2016: 574). In Wonkifong, I progressively focused an important part of my fieldwork around Sory, and, somehow, we started archiving together. We composed a photo album of his journey during and after the outbreak. Sory liked to share his feelings, and something drew me to him over the other workers. Our interactions were influenced by economic and racial inequalities. I was a white European scholar, and he was a Guinean man engaged in a constant struggle to make a living. Yet as we got to know each other, we shared confidences. I was born from parents who had fled a dictatorship. As the child of a man who experienced extreme violence and until his death was denied political recognition, I always felt drawn to write about others whose stories had been silenced. Sory had lost his father early in life, and this loss could be narrated as the first injustice that opened the door for others: lack of education, financial struggle, the absence of recognition. We both recognized the misfortunes of our fathers as decisive, and in this common understanding we met. Sory always enjoyed the idea that, as well as writing about the work of Wonkifong, I was writing about him. While scientific publications were my own project as a scholar, I came to realize that we also shared a more participative project: archiving. And that through this archival project, Sory was also able to produce his own narrative.
About private archives, Garcia notes that they constituted "texts that embody the key meanings of loss and survival" in her field of study (2016: 575). Accordingly, when doing fieldwork, I focused on how local workers fought silenced stories. These narratives eventually came to be at the core of my ethnographic journey. They shed light on how Guinean workers resisted the de facto relegation of their role by the media, and thus produced their own stories at the heart of an anti-black world. My ethnographic method in the unit, based principally on participant observation and informal discussions with local workers, at the same time produced data and created a space where they could share their thoughts and be recognized by a much wider audience for the journey they had experienced. Keeping a record of our exchanges, this article then established a dialog between the ethnography of a space, its rules and daily life, and the kind of archives that ethnographer and informants might build together, as well as the ones the latter constitute on their own. This project translates here into three different sections. In the first one, I explore the postcolonial reconfiguration of race in humanitarian aid, or how anti-blackness dictates segregated exposure to risk even between persons coming mainly from African and Caribbean regions. In the following section, I explore a media campaign that took place during the outbreak and discuss how local stories were silenced. I then show how I helped curate the private archives of local workers' journeys through my ethnographic involvement in the unit. Finally, a leap in time shows how, in 2018, a secret archive kept by Sory came to embody his own sense of care and sacrifice and helped him live in the present despite the social consequences of the outbreak.
The Enduring Exposure of Black Bodies in a Humanitarian Infrastructure
Humanitarianism, warns Fassin, unfolds a complex ontology of inequality that differentiates in a hierarchical manner the values of human lives (Fassin 2007). In making this argument, he builds on Hannah Arendt's theory of political recognition. Those in the category of bios, a politically recognized life-for Fassin, humanitarian workers and, for Arendt, Roman and Greek masters-are contrasted to zoë, a biological life common to animals and humans-the beneficiaries of aid for Fassin, the slaves of antiquity for Arendt (Arendt 1998(Arendt [1958; Fassin 2009). While Fassin focuses on the humanitarian dichotomy, Arendt draws more broadly on life's daily tasks in antiquity. As she explains, the free man works with his hands, and the slave labors with his body, thereby freeing the masters from the most painful work (1998 [1958], emphasis added). Hence, the segregation of tasks, or the difference between work and labor, is key to understanding who is considered human (therefore having a political life), and whose life is excluded from the realm of humanity. While neither Arendt nor Fassin acknowledged race as crucial to the value given to different groups of people, more recent work on the entanglement of race and medicine sheds light on the enduring exposure of black lives to greater risks, whether in colonial and slave hospitals or under care conditions in contemporary and humanitarian medicine (Gomez-Temesio 2018, 2020a; Gomez-Temesio and Le Marcis 2021). The concept of a black medical superbody, as proposed by Owens, points to an enduring entanglement between race, non-humanity, and labor since the premises of modern medicine were formulated (Owens 2017). While white doctors dismissed black bodies as completely other, they still exploited them as guinea pigs because they thought they could endure more pain than white bodies, pictured as more vulnerable and constituting lives of greater value. More recently, writing on COVID-19, Oyarzun (2020) explored the intersection of race and class among civil servants "volunteered" by the state on the frontline of the COVID-19 pandemic. From colonial to contemporary times, race dictates a segregation of tasks in medical contexts through which black labor is equated with productivity and invulnerability.
In Wonkifong, the segregation of tasks between foreign and local personnel is thus emblematic of an inequality that was shaped not only in two different humanities-one biological, the other political-but also in terms of race and class. In the transnational space that the unit constituted, Guinean caregivers were assigned the more exposed tasks. They cleaned patients and furniture with chlorinated water. They incinerated patients' belongings. They managed corpses. They handwashed all personnel uniforms with chlorinated water. Yet local workers were much more exposed to contaminating fluids (see Photo 5).
Behind the division of labor were important discrepancies in terms of salary. Foreign workers were paid much more than Guinean workers while also enjoying benefits such as daily allowances and professional opportunities at the end of the outbreak. Though Cuban and Congolese earned more than Guineans, their salaries were still ridiculously low compared to those of the Western staff employed at the WHO headquarters in Conakry. Wonkifong was hence the biggest unit of the country while being at the same time the cheapest, employing almost exclusively African and Caribbean low-wage workers as well as Guinean personnel working for even less. Basing aid relief on cheap local labor allowed the deployment of a massive Ebola response in the region (Kingori and McGowan 2016) to protect Western countries from the deadly virus. Consequently, Wonkifong's division of labor enlightens the fact that the "politics of life" (Fassin 2007) that granted less value to the lives of aid beneficiaries also applies to African aid workers (Benton 2016a(Benton , 2016b. Therefore, this inequality does not operate vaguely on different groups of humans but, instead, on different groups of racialized and working poor people. Yet Wonkifong constituted a peculiar anti-black world in that, unlike most of the Ebola infrastructures around the Manor River region (Hirsch 2019(Hirsch , 2021, there were almost no white or Western personnel working there. Its employees were mainly black people coming from Africa and Cuba. 4 Wonkifong thus challenges post-colonial imaginaries around a binary division between white and black people Photo 5. Dressing for the quarantine zone. Wonkifong Ebola Treatment Unit, 2015. [This figure appears in color in the online issue] (Oyarzun 2020), as the segregation of work operated not in terms of skin color so much as in degrees of closeness of black people to white power. Accordingly, the endurance of racial hierarchies also translated through projections of whiteness.
Ezequiel was a Congolese logistician. Some said behind his back that he was close to the DRC secret services and had been dispatched to Guinea to keep an eye on the situation there for his government. True or not, Ezequiel was obeyed by most of the staff. He never went into the quarantine zone but was strategically in charge of the unit's warehouse. His authority was translated into the nickname he was given: "le Blanc," the white man. Once, Ezequiel got sick and vomited in the middle of the unit, greatly disturbing his coworkers. A Guinean worker remembered, "When we saw le Blanc, it was terrible: If he got Ebola, what about the rest of us? If le Blanc can get sick, then we are all really screwed." Ezequiel's whiteness projected authority but also privileges. Ezequiel, as a figure related to whiteness, was not supposed to die of Ebola, unlike his other African colleagues. This anecdote underlines the shared perception of the Guinean staff that privileges were restricted to white people or those close to white power. Ezequiel's supposed proximity to the DRC secret services also sheds light on post-colonial imaginaries that equate corrupted African elites with former colons and contemporary white power. Consequently, Wonkifong provided a perfect setting in which to investigate the post-colonial reconfiguration of race as an integral part of the Africanization of aid.
The psycho-social team of which Sory was part dealt with all the paramedical work of death and life: admissions, family visits, and authorizations to bury. Local workers were thus more exposed to local resentment. One morning, a patient died. Sory called the relatives of the deceased, whose contact details were in the unit's patient register. This book was the most valued item of the unit. Sory's team managed it even though doctors came regularly to check some information, as it was the only place where all the patients' information was gathered. The family came to the unit. In accordance with the biosecurity protocols, the corpse was shown to them in a bag through a fence. 5 When the two young aides lifted the body of the deceased, one of them feared his gloves would tear under the pressure. He panicked and dropped the body. The corpse fell on the ground. The relatives, deeply shocked, started to scream. One told the aides, "Lucky you, we have no guns. Otherwise, we would have shot you." This was hardly uncommon. A few weeks before this incident, a team of workers were assaulted by a mob. Therefore, security in the unit and at the hotels where the foreign workers were staying was increased. Guinean workers, however, enjoyed no such "institutional immunity" (Fassin 2007: 515). Sory said he had to stop going to the market after people threatened him. Indeed, most workers endured social ostracism. Fatie worked alongside Sory in the psycho-social team. She was an agroeconomist who had never managed to find work in her field of study. When she started working in Wonkifong, she was disowned by her fiancé, with whom she had a child. She said, "I don't know if we can go back together. His family opposes it. They said I kill people here." Fatie had to leave her infant with her family while she shared a house with colleagues on the outskirts of Wonkifong.
Labor and work also implied a segregation of space between different groups. Wonkifong had two main spaces: the quarantine zone, where infected patients were isolated; and the rest of the unit, composed of medical and administrative quarters. There was a cafeteria where a local cook prepared meals for a fee. This place was open to everyone but was in fact only used by all the Guineans and some Congolese workers who had no other option. Cubans went back to their hotel, where they enjoyed their own menu, as did some of the Congolese. Europeans at the lab enjoyed meals delivered daily-initially, it seemed, for the patients-by the World Food Program. The cafeteria was hence serving local food for Africans. Hardly anyone was happy with the place. The food was described as too expensive and not sufficient. As Sory put it one day, it was as if "the cook split one corned beef can between 50 people." Plates and spoons were also in short supply, and workers had to share. Collective diarrhea cases were reported. African workers were therefore much more exposed to contamination if Ebola was to spread among the staff.
While local workers took disproportionate risks in comparison to their foreign colleagues, they were not guaranteed the same conditions if they suffered contamination from Ebola. Indeed, while repatriation was a guarantee for foreign workers, Guinean personnel had to be treated in a unit within their country. If the Ebola mortality rate was around 70% in the Mano River countries, it fell to under 20% for repatriated humanitarian personnel (Farmer 2015). The life expectancy of an aid worker could not be determined only on biological grounds (Canguilhem 1966); it also related to work segregation. The journey of local workers thus revealed the dialectics between race and capitalism from colonial times to the era of Global Health, as while black subjects were segregated in every aspect of daily life, their access to low-paying employment still had to be guaranteed.
Walls of Hope, Voices of Oblivion
Since the Biafran war, advocacy has become one of the key elements of humanitarian action in the Global South. Bearing witness, as noble as it may seem to be, still differentiates two sorts of life in the public space: those who can tell their own stories, and those whose stories can only be told by humanitarian institutions (Fassin 2007). During the outbreak, many campaigns publicized the humanitarian commitment and the need to "fight the virus" around the globe. But what lies beneath such mobilizations? In this section, I explore a visual and digital campaign that took place during the outbreak to shed light on how race segregation was embodied in humanitarian media strategies. While black heroes appeared in these campaigns, the voices of Guineans were systematically excluded from the limelight.
"Tribute to Ebola Health Workers" 6 was a campaign launched by the United Nations Development Program (UNDP) in 2015, during which portraits of individuals committed to the Ebola response were posted on the walls of several buildings in Conakry. The campaign was accompanied by the collection of workers' testimonies, archived on a website. I participated in the project by chance when I was asked by an acquaintance at UNDP to gather the testimonies. He thought that as an anthropologist with experience of working in an Ebola unit it would be easier for me to make contact with humanitarian workers (see Photo 6).
We interviewed foreign and local personnel in several of the Ebola units in the country as well as at the WHO headquarters. The atmosphere of the recollection changed dramatically depending on the interviewees. Expatriate workers, used to such media campaigns, mostly had ready-to-go slogans like "Together we can!" In contrast, when we interviewed Guinean workers, they were more eager to share an actual experience. One young man even came back later to give me a letter in which he had recorded his own participation since the beginning of the outbreak. Guinean workers spoke about the patients they had lost, but also about their own losses. Some were driven out of the house they lived in, while others were left by their spouses. They all emphasized the difficult tension between their role during the outbreak and their future return to "real life." The most recurrent themes were messages like "Now that our parents consider us as murderers, who is going to care for us at the end of the outbreak?" or "We want jobs after the epidemic, we don't want to go back to being unemployed." Many mentioned initial studies and prior employment-historian, lawyer, engineer in micro-finance, or salesman for a phone company. Several mentioned having been employed by mining companies. As the outbreak brought mining activities to a standstill, they were suddenly left without income. Hence, Guinean testimonies were not restricted to their experience as Ebola fighters, but also dealt with employment in their country as a disappearing phenomenon. As a Guinean aide confided to me: The worst problem in Guinea, it is not Ebola, it's poverty. We barely survive in this country, and then when Ebola came, people had no money to avoid the virus. Guineans, they might accept that the only solution is to die. The government did not move a finger for us. Did you notice? The best routes in this country are for the miners. And our hospital in Conakry? This is worse than Mozambique during the war.
The recollection process led sometimes to tensions and demonstrated how for expatriates the tribute was immediately recognized as a communication strategy aimed at "showing success" and keeping the money flowing to projects, while local workers tried to gain political recognition. In the Forécariah Ebola treatment unit, managed by the French Red Cross, the chef de mission interrupted the recording, infuriated by the fact that the local aides spent so much time talking. He had presented his own testimony in two minutes and was unable to contain his fury: "You all lazy, back to work or I will fire you!" The recollection was brought to a close. Guineans could not risk losing their jobs. For every story that sees the light of day, Jackson tells us, several others remain untold, censored, or even suppressed (Jackson 2002). The testimonies chosen to be featured on the project website only spoke about heroism, with slogans such as "The fight goes on!" The humanitarian apparatus has the power to constitute truth in the name of others, and here local workers were denied a public appearance as the subject of their own story. The stories displayed in the digital archives were only the fragments fitting the message UNDP wanted to communicate.
In his "Pro Patria Mori," Kantorowicz reminds us that the sacrifice for the homeland always appeals for a counter gift (Kantorowicz 1951)-the political recognition of the value of the sacrificial individual. With the concept of "politics of life," Fassin points out the same link between political existence and sacrifice. Only Western saviors can give their lives for a noble cause-local victims of wars and epidemics are only recognized as biological losses in the overall statistics (Fassin 2007). But what the backstage of a campaign such as UNDP's reveals once again is that racial fault lines segregated the treatment of local workers in daily tasks as in their access to media campaigns. This deprivation of political recognition was linked to the fact that Guinean workers were seen through the lens of locality and economic opportunity. As a foreign doctor told me once, "These [pointing to patients] are their people. If Guinean workers are not the first to care for them, who will? Plus, they are being paid to do so, aren't they?" Different variations along their country, their parents, their job were recurrent in the daily chit chat. Therefore, local workers were described as people benefiting from an economic opportunity more than demonstrating a will to sacrifice. If this interpretation was not necessarily wrong-many Guinean workers found themselves without a job because of the epidemic and thus needed a new means of survival-it still excluded them from the "counter gift of sacrifice": political recognition (Hubert and Mauss 1925;Wool 2015). Their portraits appeared in the campaign, yet voices were silenced when the tale they told was embarrassing for humanitarian organizations.
"We Saved Lives": The Stories I Was Told
Wonkifong was an extraordinary place within the humanitarian landscape of the Manor River region: It offered the unique experience of running an Ebola unit with people who came almost exclusively from the Global South. But the place was excluded from the UNDP tribute for trivial concerns. The recollection was performed and filmed during the rainy season. The road was in poor condition and the producer decided to stay in Conakry so that we could focus on other structures-such as an Ebola unit run by the French army and the WHO headquarters.
When I eventually returned, I found the local staff at Wonkifong discussing a strike. The outbreak had faded out in the region, and the unit was slowly closing down. Most of the local workers had been notified of contract termination. They were offered no prospect of future work. Sory hardly slept at night and feared for his own life. Fatie was scared. She had nowhere to live, as her whole family rejected her. Ebola fighters from elsewhere gained a public tribute as they fitted the archetype of humanitarian heroes. Guinean people and their mundane difficulties, such as unemployment and social ostracism, were left in the dark.
But even though they were denied a space within the big tent of humanitarian aid to share their experience, memorialization was a common concern and inner need for local workers. Most of them regularly took pictures while working. This material was eagerly shared on social media during and after the epidemic. At the end of the tribute, I had spent six months in the unit. I rarely recorded our conversations, as I preferred them to remain informal. But after the closure of UNDP's project, I thought recording might be an interesting approach. I asked the workers what they wanted to tell, to remember. Most of them spent considerable time speaking about the patients they had lost, but they also wished to remember those they had helped to survive. Sory recollected how he had once experienced a conflict with Cuban doctors. It was about a two-year-old boy who had been quarantined alone with no family members present. Sory opposed the placement of the child in the women's tent. He remembered the Cubans gossiping about him: "Who wants to put a kid with unknown males?" But Sory had his reasons: "All the women in the tent were nearing death. ''In the men's tent, on the contrary, there were quite a few men who were already recovering-they could feed the baby". Sory changed the location of the infant at night, when the foreign doctors went home. The next morning, one Cuban almost beat him. Still, the boy remained with the men. He was fed by themand ultimately he survived.
The memory Sory decided to share was highly relevant. According to foreign staff, local aid workers did not abide by all the security protocols or respect collective representations on how to handle sick people. The way the Cuban brigade welcomed me at the unit was emblematic of this. Their chief, probably assuming that as anthropologist I was there to deal with the "natives," told me: "Guineanos, they don't really know how to care for their own. It would be great if you could keep an eye on what's going on here." For most foreign workers, the locality of Guinean employees corresponded to a lack of education. Regarding Cuban workers, it mirrored the fact that Cuba has trained a substantial part of Guinean medical personnel over the years (Kirk and Erisman 2009). And because Ebola was first discovered in The DRC (Piot 2012), Guineans were seen as those who let the virus spread into the region, while the Congolese had always managed to stop an Ebola outbreak before it threatened another country. 7 Therefore, being local was translated into "having to learn something" from foreigners, as well as sharing the blame for the outbreak with the patients quarantined in the unit.
Yet, local workers often found creative ways to care for those in their charge. One worker told me how she used to trick patients to build trust. As she was hidden under her protective equipment, patients could identify her only by her voice, speaking in the different local languages. She usually checked their details in the patients' register and then made up a character identity that she could assume in conversation with the patient. Once a young man refused to eat for an entire day upon his arrival. She explained: He came from a village I knew. I told him, "Hey, do you recognize me? It's me who sells you the best plantain at the stall close to the gas station." It worked! I could see he was relieved. I explained to him, "There is nothing to fear here, just eat and take the drugs, you will get better." She laughed at that memory but while we went on with the recording, she got sad. "There was this little boy. He didn't want to eat. I said to him, ''Don't you recognize me? I am your aunt, I know your mommy very well." 8 His eyes became bigger. "My aunt?" "Yes, I came to see if you are eating to feel better. Please, drink some juice if you want to go back to school." He said, "Ma tante, but I don't go to school." I replied, "If you take your juice, I will send you to school." He seemed so happy. "Auntie, is it true? You are going to send me to school?" He drank but vomited immediately after. I could see he was not going to make it. I started crying. A doctor summoned me. "You can't cry in front of the patients." The day after that, I went to the morgue to check on the corpses. He was there, my little boy. Till this day I remember him.
Fatie also took great pride in the fact that she helped people change their minds about the quarantine unit around her: My own maman, she disowned me. She said I will bring the virus with me and kill her. But then once I asked a colleague to film while I was dressing in the protective equipment. I shared it with my mom, friends, everybody! People understood we were safe, and that we were helping people who got sick.
Guinean workers spoke most about how they managed to make a difference, and about the patients they had lost and were unable to forget. These memories highlighted the fact that if local workers could not share their own stories on humanitarian media campaigns, they could-and knew how to-work. They saved lives-even though their foreign colleagues and the international media campaigns did not acknowledge them.
Of Archives and Love: Three Years Later
We are back in 2018. Wandering among Wonkifong's ruins, Sory and I remembered those who worked there. "Have you seen pictures of Claude on Facebook?" he asked. "Apparently, he is on some humanitarian mission in Asia." Until recently, Sory had kept in touch with his former foreign colleagues. But then his computer broke. He had no money to fix it and then didn't get news from Claude and the others. He kept in touch with the Guineans, though. He told me that Fatie had passed away a few months before. Sory explained. "She had another child. I am not really sure how exactly, but she died afterwards." I remembered her being scared at the closure of the unit. Sory continued: "She got sick right after the baby was born. She got anaemia and, I guess, had no money for the drugs." Fatie was a young mother who had been willing to sacrifice her existence to fight the virus, and then succumbed while birthing another life. She gave the gift of life three times and then died because she was unemployed. She was 32 years old (see Photo 7).
When talking about Fatie, her enthusiastic voice on the recordings came to me and I remembered her careful writing in the patients' handbook. It seemed to me that this book constituted the most important archive of Wonkifong. I wondered suddenly what had become of it. Sory smiled. "I kept it. No one asked for the register when the unit closed. So, I just left with it." Sory did not get to tell his story, but he became the guardian of the stories untold. He revealed, "Nobody knows I kept it. Sometimes I go through it. I remember the people." A self-made curator with no gallery or museum of his own, Sory nevertheless protected his story within the intimacy of his home.
The fact that no one cared to retrieve this register underlined the difference in treatment accorded to what come to be considered as data and material that has the potential to become an archive. Data were produced by the different medical contingents during the outbreak. Data translated patients' experience into numbers and focused on such things as survival rate, immunity, and so on. The patients' register contained information that was considered as data and thus extracted in digital files, but it also remained a material trace of remembrance as it listed every patient with names, home, family contacts, and most of all, anecdotes of their stay inside the unit. The data, extracted from the medical register, were shipped out of the country at the end of the outbreak when foreign teams flew back to their countries. 9 The register was almost forgotten in an abandoned site and eventually rescued by Sory. It then became a personal archive for him. Sory said he feared people from clinical trials might get hold of his register. "Today, everybody does money out of the survivors. With this book, I am the only one who is able to locate all of them." Keeping the book, Sory eventually fulfilled his own dream of caring for others, during and after the outbreak. As recounted by Garcia, being the curator of her informant's prison correspondence, the register became for Sory a material remain that condensed affective and social relations (Garcia 2016: 574). An impromptu archivist of the history of the outbreak in his country, Sory continued to follow his calling in 2018, as in 2015.
While we were talking about the patients' register, Sory suddenly said: "You are in the book." I knew this to be the case, as back in 2015 I had spent hours filling it in, but his smile was enigmatic. I waited for what seemed to be another secret he wanted to share with me. He gave me a photograph of a bride from his pocket. "I could not tell you before I saw you. I married this woman in 2016. Do you recognize her? She is a survivor. A few days after marrying her, I looked at the register, just wanting to see her details, out of curiosity. And I saw your handwriting, you're the one who signed her certificate of departure." Since 2015, my encounters with Sory have always been structured by unequal relationships. I was born white in a Western country, my access to education drove me to work in a profession that brought me great joy and fulfilment. But when he disclosed this hidden archive, it appeared that my hand had helped shape the path of his Ebola afterlife, in which he found some sort of social recovery. Both our journeys colluded here. As children, he dreamed about becoming a doctor and helping others, while I longed to tell forgotten stories about loss and silence. In this we eventually met.
Sory told me the story of his bride. "She was desperate, not only because of Ebola. Her husband had died. I wanted her to find the strength to survive. I told her, if you do it, if you get out of this place, I will marry you when all this is over." And so he did. By marrying a former patient, Sory gained a precious sense of recognition at a time when family and neighbors had turned their backs on him. "People still threaten me. But my wife, she is grateful." "Recognition appears beyond its inscription in language," notes Le Marcis (2012: 488), "it is embodied in acts of care." Sory said to me about his wife, "People treat Ebola survivors as waste. I wanted to show them, they are human beings." Sory and his wife had no children of their own. He confessed that Ebola probably made her infertile, as had happened with many other former female patients. Nonetheless, he had no regrets. Happy was maybe not a term that would describe him, but he had a meaningful life, consistent with his previous commitment: "No matter what happens to me now, when I am at home, I feel good about what I did." Later, when his wife came home, he quickly whispered to me, "Don't tell her I have no job." He didn't want her to be worried. Sory was not recognized for the role he played during the epidemic, but his marriage somehow helped him be the hero in his own history. Looking at Sory's wife that day, nevertheless, I could not help but ask myself: Who was going to tell her story? As a widow who survived Ebola, it seemed the unique path available to her was to marry again with her former savior. Thus, recovery after the epidemic was shaped through dramatic gender inequality.
Epilogue: Not the Only Plot
Greater vigilance is required, Erikson and Wendland warn, "to undo the multiple ways that Global Health depends on steep inequalities for its very existence" (Erikson and Wendland 2014: 3). The Wonkifong unit, the only "South" infrastructure in Guinea during the outbreak, laid bare old and new forms of discrimination that bring to light a robust continuum between colonial medicine and humanitarian aid. About the involvement of black volunteers during the COVID-19 pandemic, Oyarzun writes that in the United States the most precarious people, those without income and health insurance, literally "threw themselves at the "opportunity" to work on the frontlines of the response against a deadly virus"-while if they got infected the State would not protect them (Oyarzun 2020: 583). Ebola-like the COVID-19 pandemic-shed light on the many ways in which black bodies are exploited at the intersection of race, class, and gender (Gomez-Temesio 2020a).
Wonkifong's experience shows us how both segregation of work and the organization of humanitarian media campaigns underline a failure of subjectivity regarding local aid workers. When writing about my informants' experiences during the outbreak, how was I to avoid repeating the "oppressive gaze" (Hartman 1997) in scientific writing? My ethnographic project thus focused not only on the unit's organization but on local workers' imaginaries, regrets, and emotions-and on how they built their own archives. "Slavery was so unprecedented," warns Morrison, that you can let slavery be the story, the plot. … And then you do the worst thing, which is … the center of it becomes the institution and not the people. So, if you focus on the characters and their interior life, it's like putting the authority back into the hands of the slaves, rather than the slave owner. (Morrison 2020: 25) Accordingly, in this piece, I did not stop at "describing" the segregated treatment of local workers but underlined how people inhabit these cramped spaces and how they invest themselves in "innovative resistances" (McKittrick 2013). I investigated how people reflected on their condition and tried to access their "interior life" as far as ethnography could show this: their expectations, dreams, frustrations, and disappointments-how they made sense of the present and prepared a future where life was bearable. It is in this dialog between their thoughts, our discussions, and the material and archival traces of the outbreak that I gained access to an intimate space where the racial segregation that rules humanitarian aid was not the "only plot" (Morrison 2020). As such, the journey local workers made through the epidemic underlines how people may respond to anti-blackness by laying claim to experience on their own terms. The article hence undertook a triple triage at the nexus of humanitarian aid and medical anthropology: the role of race within humanitarian institutions; the voices of local workers in the humanitarian narrative; and the subjectivity of those whose lives we investigate in academic writing.
Back in Wonkifong during the epidemic, I used to take my notebook with me in the unit, though I wasn't always taking notes. Most of the time, I was just busy participating in the unit's life. But when there were conflicts, people started noticing that I rushed to write what was happening. And of course, most of the time, they stopped arguing when they saw me. It became a joke, as Sory summarized it one day, "when it gets hot here, Veronica always gets the pencil out of the pocket!" One day, he said proudly to his colleagues, "Veronica is writing a book. And believe me, I talk a lot in this book." When we were about to leave the ruins in 2018, Sory asked me. "So, your book, is it already out?" I plunged into a long explanation on how writing academic books lasts forever, but, as he looked disappointed, I reassured him I had already published a few pieces in which he was "starring." And I added, "Of course, I changed your name so people could not recognize you." Sory was startled. "Why on earth did you do that? Then nobody will know, it was me, all the time, it was me." Indeed, while I always knew Sory was keen to appear under his real name, until the very last moments of writing and revising this article, I hesitated. I thought I was going against his will. But I also thought he might regret, and that I owed him some droit à l'oubli, a right to be forgotten in the future if he ever wanted to move on from the epidemic. And I mostly feared that, in revealing him, I revealed the identity of his wife, who never manifested a will to have her story of loss and infertility shared with a wider audience. I might be wrong, or not. In this lies an ambivalent inequity that I was never able to overcome, a final limit to my archival project built with Sory where he is forced to remain the anonymous character of a story told by me. | 2022-06-01T06:26:10.740Z | 2022-05-30T00:00:00.000 | {
"year": 2022,
"sha1": "43e41fe99b6943aa5371405ce9ac5a12435066b1",
"oa_license": "CCBY",
"oa_url": null,
"oa_status": null,
"pdf_src": "PubMedCentral",
"pdf_hash": "fdcbcb86731b9fac549d7f4ddec04edd124df6de",
"s2fieldsofstudy": [
"History"
],
"extfieldsofstudy": [
"Medicine"
]
} |
253793628 | pes2o/s2orc | v3-fos-license | Tocotrienol-Rich Fractions Offer Potential to Suppress Pulmonary Fibrosis Progression
Although pulmonary fibrosis (PF) is considered a rare disease, the incidence thereof has increased steadily in recent years, while a safe and effective cure remains beyond reach. In this study, the potential of tocotrienol-rich fractions (TRF) and carotene to alleviate PF was explored. PF was induced in Sprague-Dawley rats via a single intratracheal bleomycin (BLM) (5 mg/kg) instillation. These rats were subsequently treated with TRF, carotene, pirfenidone (Pir) and nintedanib (Nin) for 28 days via gavage administration, whereafter histopathological performance, biochemical functions and molecular alterations were studied in the lung tissues. Our results showed that TRF, carotene, Nin and Pir all ameliorated PF by reducing inflammation and resisting oxidative stress to varying degrees. The related mechanisms involved the TGF-β1/Smad, PI3K/Akt and NF-κB signaling pathways. Ultimately, our findings revealed that, when combined with TRF, the therapeutic effects of Nin and Pir on PF were enhanced, indicating that TRF may, indeed, provide promising potential for use in combination therapy in the treatment of PF.
Introduction
The 2019 outbreak of the coronavirus disease (COVID-19), a communicable respiratory disease caused by the SARS-CoV-2 virus, developed rapidly into a severe and currently ongoing global health challenge [1]. According to the World Health Organization (WHO), as of June 2022 more than 530 million people worldwide had reportedly been infected with the original SARS-CoV-2 or its variants of which over 6.3 million had died. Many of those who recovered from COVID-19, particularly those who suffered severe infection, are reported to be at risk of long-term pulmonary complications [2]. Interstitial pneumonia is a common feature of COVID-19 and can be complicated by acute respiratory distress syndrome (ARDS) of which pulmonary fibrosis (PF) is the recognized sequelae [3]. Autopsy results of COVID-19 patients have shown severe fibrosis in the lungs after an extended course of the disease [4]. Consequently, due to the severity and continuity of the epidemic and the burden of PF after SARS-CoV-2 infection, the management and amelioration of the disease is an increasing topic of medical and scientific discussion. PF, indicated by the scarring of the lungs, is the ultimate and most devastating consequence of various inflammatory lung diseases [5]. While generally a rare disease, in recent years, numerous global studies have reported an alarming increasing trend in the incidence and prevalence of PF [6]. Unfortunately, medical theories and research results have not yet clarified the cause and mechanism of its pathogenesis. Glucocorticoids and immunosuppressive drugs are currently the main response to the disease in China. However, long-term use induces side effects, such as infection and bone necrosis, and large-scale use risks the development of liver necrosis. Moreover, these drugs are only effective in some patients 2. Results and Discussion 2.1. Effects on Weight (g), Feed Intake (g) and Lung Index of TRF As can be seen from Figure 1, there were two significant fluctuations in the body weight of the rats during the 56-day experimental period. The first fluctuation occurred when the model was established on the first day, when the body weight of the rats in each group decreased significantly, thereafter recovering after 7-8 days (Figure 1a). The second fluctuation was observed when the gavaging began on the 28th day, after which the weight stabilized for 2-3 days and then recovered again (Figure 1a). Correspondingly, during these periods of weight loss, feed intake reduced accordingly (Figure 1b). The Nin group had the lowest body weight, corresponding to the lowest feed intake. The lung/body weight ratio (lung index) is an acknowledged index of PF; however, in this experiment, the lung indexes of each group were not statistically different (Figure 1c).
weight stabilized for 2-3 days and then recovered again (Figure 1a). Correspondingly, during these periods of weight loss, feed intake reduced accordingly (Figure 1b). The Nin group had the lowest body weight, corresponding to the lowest feed intake. The lung/body weight ratio (lung index) is an acknowledged index of PF; however, in this experiment, the lung indexes of each group were not statistically different (Figure 1c). Effects on weight (g) and feed intake (g) of TRF: (a) Body weight changes in rat groups recorded every other day during the experiment (n = 15); (b) Feed intake changes in rat groups recorded every other day during the experiment (n = 15); (c) Lung/body weight ratios of rats on the last day of the experiment (n = 15). Data are expressed as the mean ± SD. There were no significant differences between the groups.
Effects on HYP and MMP-7 Level of TRF
As shown in Figure 2, high levels of HYP and MMP-7 contents indicated successful modeling of PF in the model group. The contents of HYP and MMP-7 in the model (BLM) group were significantly higher than those in the control group, rising from 0.11 ± 0.02 and 0.61 ± 0.02 to 6.55 ± 0.28 and 23.40 ± 1.75 mg/g, respectively, in the wet tissue, reflecting significant increases of 455% and 257%, respectively. After treatment with TRF and carotene groups and their combined group, the values of HYP and MMP-7 were reduced, but the therapeutic effect was significantly lower than that of the other drug groups (Nin, Nin + TRF, Pir and the Pir + TRF groups). Among the groups, only the TRF group was statistically different to the corresponding model groups in terms of both HYP and MMP-7 values (p < 0.05). The best therapeutic effects were seen in the Nin + TRF group and PIR group, which decreased by 110% (from 0.61 ± 0.02 to 0.29 ± 0.03 mg/g wet tissue, p < 0.05) and 82% (from 23.40 ± 1.75 to 12.85 ± 1.12 mg/g wet tissue, p < 0.05), respectively. While combined with TRF, the value of HYP and MMP-7 in Nin group and Pir group decreased except the Pir group in MMP-7, which showed better antifibrotic characteristics after added TRF. Effects on weight (g) and feed intake (g) of TRF: (a) Body weight changes in rat groups recorded every other day during the experiment (n = 15); (b) Feed intake changes in rat groups recorded every other day during the experiment (n = 15); (c) Lung/body weight ratios of rats on the last day of the experiment (n = 15). Data are expressed as the mean ± SD. There were no significant differences between the groups.
Effects on HYP and MMP-7 Level of TRF
As shown in Figure 2, high levels of HYP and MMP-7 contents indicated successful modeling of PF in the model group. The contents of HYP and MMP-7 in the model (BLM) group were significantly higher than those in the control group, rising from 0.11 ± 0.02 and 0.61 ± 0.02 to 6.55 ± 0.28 and 23.40 ± 1.75 mg/g, respectively, in the wet tissue, reflecting significant increases of 455% and 257%, respectively. After treatment with TRF and carotene groups and their combined group, the values of HYP and MMP-7 were reduced, but the therapeutic effect was significantly lower than that of the other drug groups (Nin, Nin + TRF, Pir and the Pir + TRF groups). Among the groups, only the TRF group was statistically different to the corresponding model groups in terms of both HYP and MMP-7 values (p < 0.05). The best therapeutic effects were seen in the Nin + TRF group and PIR group, which decreased by 110% (from 0.61 ± 0.02 to 0.29 ± 0.03 mg/g wet tissue, p < 0.05) and 82% (from 23.40 ± 1.75 to 12.85 ± 1.12 mg/g wet tissue, p < 0.05), respectively. While combined with TRF, the value of HYP and MMP-7 in Nin group and Pir group decreased except the Pir group in MMP-7, which showed better antifibrotic characteristics after added TRF.
Effects on Inflammatory Markers of TRF
As shown in Figure 3, the values of IL-1β, IL-6, MPO, TGF-β1 and TNF-α rose from 169.00 ± 8. 12 Effects of TRF and carotene on HYP (a) and MMP-7 (b) levels of TRF in lung tissue, n = 8. Note: * represents p < 0.05 in comparison to the control group; # represents p < 0.05 in comparison to the model group. Figure 3, the values of IL-1β, IL-6, MPO, TGF-β1 and TNF-α rose from 169.00 ± 8.12, 950.60 ± 159.40, 46.89 ± 2.21, 772.20 ± 69.27 and 68.12 ± 3.40 mg/g protein in the control group, respectively, and to 394.50 ± 20.69, 3353.00 ± 586.90, 150.90 ± 9.62, 1744.00 ± 132.70 and 213.20 ± 36.05 mg/g protein (p < 0.05) in the model group, respectively. The TRF, carotene and drug groups all showed excellent anti-inflammatory abilities. The greater effects, particularly those of MPO and TGF-β1, were seen in the four drug groups. For example, the Nin group decreased by 58% (from 150.90 ± 9.62 to 63.37 ± 7.22 mg/g protein), while the carotene group only decreased by 8% (from 150.90 ± 9.62 to 138.80 ± 18.04 mg/g protein); and there was no significant difference in MPO between drug groups and the control group. Effects of TRF and carotene on HYP (a) and MMP-7 (b) levels of TRF in lung tissue, n = 8. Note: * represents p < 0.05 in comparison to the control group; # represents p < 0.05 in comparison to the model group.
Effects on Inflammatory Markers of TRF
As shown in Figure 3, the values of IL-1β, IL-6, MPO, TGF-β1 and TNF-α rose from 169.00 ± 8.12, 950.60 ± 159.40, 46.89 ± 2.21, 772.20 ± 69.27 and 68.12 ± 3.40 mg/g protein in the control group, respectively, and to 394.50 ± 20.69, 3353.00 ± 586.90, 150.90 ± 9.62, 1744.00 ± 132.70 and 213.20 ± 36.05 mg/g protein (p < 0.05) in the model group, respectively. The TRF, carotene and drug groups all showed excellent anti-inflammatory abilities. The greater effects, particularly those of MPO and TGF-β1, were seen in the four drug groups. For example, the Nin group decreased by 58% (from 150.90 ± 9.62 to 63.37 ± 7.22 mg/g protein), while the carotene group only decreased by 8% (from 150.90 ± 9.62 to 138.80 ± 18.04 mg/g protein); and there was no significant difference in MPO between drug groups and the control group.
Effects on Antioxidant Enzymes of TRF
The antioxidant levels of GSH, NO, MDA, CAT and SOD in the lung tissue were also determined, as shown in Figure 4. As with the proinflammatory cytokine influx, the levels of antioxidant enzymes were significantly changed by the administration of BLM. Both the nutritional intervention groups (TRF, carotene and TRF + carotene groups) and drug groups were found to exert certain antioxidant effects and neither was necessarily better than the other. For instance, the CAT and GSH values of the drug groups were lower than those of the nutritional intervention groups, in which even the TRF-group's values were increased to 5.75 ± 0.50 and 15.44 ± 1.86, respectively. That is to say, in these two indicators, the nutritional intervention groups even showed stronger antioxidation than the drug groups.
the nutritional intervention groups (TRF, carotene and TRF + carotene groups) and dru groups were found to exert certain antioxidant effects and neither was necessarily bette than the other. For instance, the CAT and GSH values of the drug groups were lower tha those of the nutritional intervention groups, in which even the TRF-group's values wer increased to 5.75 ± 0.50 and 15.44 ± 1.86, respectively. That is to say, in these two indicator the nutritional intervention groups even showed stronger antioxidation than the dru groups.
Effects on Histological Evaluation of TRF
HE staining was used to identify inflammation, while Masson's trichrome stainin was applied to examine the deposition of collagen in different sections of the tissue sam ples. A quantitative scoring system was used to evaluate the extent of PF and the severit of lung injury (as noted in Figures 5 and 6). However, due to the large number of inte ventions, whether it is necessary to add them to the title remains to be discussed; the H and Masson's trichrome staining, lung specimens from the control group revealed idea intact pulmonary architecture with no evidence of tissue injury or fibrosis, nor any inflam mation, hemorrhage, edema or emphysema. With the exception of the control group, lun architecture was distorted by BLM intervention to varying degrees in the other groups i which the alveoli had begun to obviously expand, the pulmonary interstitium had begu to bleed in varying degrees, and the alveolar septum was narrowed and breaking. Th intervention groups all showed a certain therapeutic effect; however, according to the re sults, the effects on the drug groups were more distinct than those of the nutritional inte vention groups. The results of the Masson's staining were similar to those of the HE stain ing, indicating that the states of inflammation in the groups was consistent with the ex tents of their fibrosis. . Effects on antioxidant enzymes and oxidative stress markers of TRF in the lung tissues of BLM-induced lung fibrosis (a-e); n = 8; Note: compared with CON group, * represents p < 0.05; Compared with MOD group, # represents p < 0.05.
Effects on Histological Evaluation of TRF
HE staining was used to identify inflammation, while Masson's trichrome staining was applied to examine the deposition of collagen in different sections of the tissue samples. A quantitative scoring system was used to evaluate the extent of PF and the severity of lung injury (as noted in Figures 5 and 6). However, due to the large number of interventions, whether it is necessary to add them to the title remains to be discussed; the HE and Masson's trichrome staining, lung specimens from the control group revealed ideal, intact pulmonary architecture with no evidence of tissue injury or fibrosis, nor any inflammation, hemorrhage, edema or emphysema. With the exception of the control group, lung architecture was distorted by BLM intervention to varying degrees in the other groups in which the alveoli had begun to obviously expand, the pulmonary interstitium had begun to bleed in varying degrees, and the alveolar septum was narrowed and breaking. The intervention groups all showed a certain therapeutic effect; however, according to the results, the effects on the drug groups were more distinct than those of the nutritional intervention groups. The results of the Masson's staining were similar to those of the HE staining, indicating that the states of inflammation in the groups was consistent with the extents of their fibrosis.
IHC Determination on Collagen I and Collagen II of TRF
The excessive deposition of extracellular matrix (ECM) is a key feature of fibrosis, and the increased depositions of collagen I and II are one of the makers of PF [25,26]. In our study, the expression levels of collagen I and II were observed via IHC. Increased amounts of both type I collagen (yellow) ( Figure 7) and type II collagen (yellow) ( Figure 8) were detected, and were noted to be deposited in a disordered pattern in areas of fibrosis in model group. While lung tissues from the control group did not show depositions (Figures 7 and 8, control group), TRF, carotene, Nin and Pir treatments all reduced the deposition of both types of collagens in the BLM groups. The reduction of collagen deposition in the nutritional intervention groups was found to be far less than in the drug groups, and the picture magnified 400 times was clearer and more visible. Our results thus suggest that TRF and carotene could treat BLM-induced PF in rats by decreasing the production of collagen I and II to a certain degree, and this finding concurs with the histological evaluation results. Effects on lung histopathological changes of TRF in Masson's Trichrome stained specimens: Lungs from the different groups of rats were collected, and 4-μm thick sections were prepared and used for the assessment of tissue injury; (a) original magnification 100×; (b) original magnification 400×; n = 8. Note: * represents p < 0.05 in comparison to the control group; # represents p < 0.05 in comparison to the model group.
IHC Determination on Collagen I and Collagen II of TRF
The excessive deposition of extracellular matrix (ECM) is a key feature of fibrosis, and the increased depositions of collagen I and II are one of the makers of PF [25,26]. In our study, the expression levels of collagen I and II were observed via IHC. Increased amounts of both type I collagen (yellow) (Figure 7) and type II collagen (yellow) ( Figure 8) were detected, and were noted to be deposited in a disordered pattern in areas of fibrosis in model group. While lung tissues from the control group did not show depositions
IHC Determination on Collagen I and Collagen II of TRF
The excessive deposition of extracellular matrix (ECM) is a key feature of fibrosis, and the increased depositions of collagen I and II are one of the makers of PF [25,26]. In our study, the expression levels of collagen I and II were observed via IHC. Increased amounts of both type I collagen (yellow) (Figure 7) and type II collagen (yellow) ( Figure 8) were detected, and were noted to be deposited in a disordered pattern in areas of fibro- deposition of both types of collagens in the BLM groups. The reduction of collagen deposition in the nutritional intervention groups was found to be far less than in the drug groups, and the picture magnified 400 times was clearer and more visible. Our results thus suggest that TRF and carotene could treat BLM-induced PF in rats by decreasing the production of collagen I and II to a certain degree, and this finding concurs with the histological evaluation results. thus, suggest that TRF and carotene could treat BLM-induced PF in rats by decreasing the production of collagen I and Ⅱ to a certain degree, and this finding concurs with the histological evaluation results. Amelioration of bleomycin-induced PF: Lungs from different rat groups were collected, and 5-μm thick sections were prepared and used for the assessment of tissue injury; (a) representative images of IHC staining for collagen Ⅰ in the lungs; original magnification ×100; (b) representative images of IHC staining for collagen I in the lungs; original magnification ×400; n=8. Note: * represents P < 0.05 in comparison to the control group; # represents P < 0.05 in comparison to the model group. Figure 8. Amelioration of bleomycin-induced PF: Lungs from different groups were collected, and 5-μm thick sections were prepared and used for the assessment of tissue injury; (a) representative images of IHC staining for collagenⅡin the lungs; original magnification ×100; (b) representative images of IHC staining for collagenⅡin the lungs; original magnification ×400; n=8. Note: * represents P < 0.05 in comparison to the control group; # represents P < 0.05 in comparison to the model group.
Effects on TGF-β/Smad signaling pathway
Macrophage chemotaxis was reportedly rendered within a lesion via TGF-β signal activation, the activation and proliferation of fibroblasts were induced, the synthesis of collagen was increased, the expression of a great quantity of fibrotic cytokines and proinflammatory factors was stimulated, and the fibrotic response was further enhanced
Effects on TGF-β/Smad Signaling Pathway of TRF
Macrophage chemotaxis was reportedly rendered within a lesion via TGF-β signal activation, the activation and proliferation of fibroblasts were induced, the synthesis of collagen was increased, the expression of a great quantity of fibrotic cytokines and proinflammatory factors was stimulated, and the fibrotic response was further enhanced and sustained [27]. In this study, the expressions of the TGF-β1 protein, Smad2 protein, Smad3 protein and α-SMA protein were significantly (p < 0.05) downregulated compared to those in the control rats, while Smad7 protein was upregulated. In the WB group diagrams (Figure 9a-f), although there is no evidence of the statistical differences in the histograms (Figure 9c), and the values of the drug group were lower than those of the nutrition intervention groups overall. In the nutritional intervention groups, the treatment effect of the combined treatment group (TRF + carotene) was obviously improved, while the value in the model group was reduced from 1.53 ± 0.17 to 0.43 ± 0.07, and that of the control group was reduced from 0.24 ± 0.10. At the same time, however, gene expression (Figure 9g-j) was found to be more pronounced, which was not completely consistent with the WB results. It is worth mentioning that the TRF values were always statistically different from those of the control group, but similar to those of the model group. Overall, the above findings indicate that carotene and TRF exerted certain antifibrotic abilities, albeit far lower than the drug groups, and the remarkable anti-pulmonary fibrosis mechanisms may be related to the TGF-β/Smad signal pathway.
Effects on TGF-β/Smad Signaling Pathway of TRF
Macrophage chemotaxis was reportedly rendered within a lesion via TGF-β signal activation, the activation and proliferation of fibroblasts were induced, the synthesis of collagen was increased, the expression of a great quantity of fibrotic cytokines and proinflammatory factors was stimulated, and the fibrotic response was further enhanced and sustained [27]. In this study, the expressions of the TGF-β1 protein, Smad2 protein, Smad3 protein and α-SMA protein were significantly (p < 0.05) downregulated compared to those in the control rats, while Smad7 protein was upregulated. In the WB group diagrams (Figure 9a-f), although there is no evidence of the statistical differences in the histograms (Figure 9c), and the values of the drug group were lower than those of the nutrition intervention groups overall. In the nutritional intervention groups, the treatment effect of the combined treatment group (TRF + carotene) was obviously improved, while the value in the model group was reduced from 1.53 ± 0.17 to 0.43 ± 0.07, and that of the control group was reduced from 0.24 ± 0.10. At the same time, however, gene expression (Figure 9g-j) was found to be more pronounced, which was not completely consistent with the WB results. It is worth mentioning that the TRF values were always statistically different from those of the control group, but similar to those of the model group. Overall, the above findings indicate that carotene and TRF exerted certain antifibrotic abilities, albeit far lower than the drug groups, and the remarkable anti-pulmonary fibrosis mechanisms may be related to the TGF-β/Smad signal pathway. (g-j) Representative statistical analysis of TGF-β1 (g), Smad2 (h), Smad3 (i) and Smad7 (j) in mRNA by RT-qPCR; n = 3. Note: * represents p < 0.05 in comparison to the control group; # represents p < 0.05 in comparison to the model group. (g-j) Representative statistical analysis of TGF-β1 (g), Smad2 (h), Smad3 (i) and Smad7 (j) in mRNA by RT-qPCR; n = 3. Note: * represents p < 0.05 in comparison to the control group; # represents p < 0.05 in comparison to the model group.
Effects on PI3K/Akt/mTOR Signaling Pathway of TRF
The PI3K/Akt/mTOR signaling pathway is an important intracellular signal transduction pathway, the inhibition of which can promote autophagy and PF [28][29][30]. Compared with the expressions of PI3K and mTOR proteins in the control group, those in the drug groups were significantly upregulated (p < 0.05), while their downregulation in the drug groups were more pronounced than those in the nutritional intervention groups. The RT-qPCR results (Figure 10e-g) showed that the expressions of PI3K mRNA, Akt mRNA and mTOR mRNA in the BLM-induced rats had increased significantly and were obviously downregulated in the drug groups (p < 0.05), while there were no statistical differences compared to the nutritional intervention groups in PI3K mRNA. Moreover, the protein results were confirmed at the genetic level. These findings suggest that, while TRF and carotene may alleviate PF through the PI3K/Akt/mTOR pathway to a certain extent, the treatment effect was not statistically significant (p > 0.05), and the expressions were more obvious at the gene level.
Effects on PI3K/Akt/mTOR Signaling Pathway of TRF
The PI3K/Akt/mTOR signaling pathway is an important intracellular signal transduction pathway, the inhibition of which can promote autophagy and PF [28][29][30]. Compared with the expressions of PI3K and mTOR proteins in the control group, those in the drug groups were significantly upregulated (p < 0.05), while their downregulation in the drug groups were more pronounced than those in the nutritional intervention groups. The RT-qPCR results (Figure 10e-g) showed that the expressions of PI3K mRNA, Akt mRNA and mTOR mRNA in the BLM-induced rats had increased significantly and were obviously downregulated in the drug groups (p < 0.05), while there were no statistical differences compared to the nutritional intervention groups in PI3K mRNA. Moreover, the protein results were confirmed at the genetic level. These findings suggest that, while TRF and carotene may alleviate PF through the PI3K/Akt/mTOR pathway to a certain extent, the treatment effect was not statistically significant (p > 0.05), and the expressions were more obvious at the gene level. , AKT (f) and mTOR (g) mRNA by RT-qPCR; n = 3. Note: * represents p < 0.05 in comparison to the control group; # represents p < 0.05 in comparison to the model group.
Effects on NF-κB Signaling Pathway of TRF
Nuclear factor kappa B (NF-κB) is widely present in inflammatory cells and, when stimulated to activation, takes part in a variety of pathological and physiological processes. The expression levels of inflammatory cytokines downstream of the NF-κB signaling pathway are especially relevant to the degree of inflammatory responses [31,32]. In this study, compared with the control group, the levels of p-p65 protein, p-IkBα protein, Ikkβ protein, TNF-α mRNA, IFN-γ mRNA, IL-13 mRNA, NF-κB mRNA, IkBα mRNA and Ikkβ mRNA were upregulated significantly (p < 0.05) in model group, especially in their gene expressions. Figure 11 shows that the phosphorylation of the NF-κB signaling pathway was decreased in all groups compared to the control and model groups, but with no significant differences (p > 0.05) between those affected groups. Furthermore, the results of RT-qPCR (also shown in Figure 11), were consistent with, but more obvious than, the WB results. It is worth noting that although the protein expressions of NF-κB in the lung tissue were downregulated in the nutritional intervention groups, and the activation of NF-κB inhibited, the level of downregulation was much lower than that of the drug groups.
Effects on NF-κB Signaling Pathway of TRF
Nuclear factor kappa B (NF-κB) is widely present in inflammatory cells and, when stimulated to activation, takes part in a variety of pathological and physiological processes. The expression levels of inflammatory cytokines downstream of the NF-κB signaling pathway are especially relevant to the degree of inflammatory responses [31,32]. In this study, compared with the control group, the levels of p-p65 protein, p-IkBα protein, Ikkβ protein, TNF-α mRNA, IFN-γ mRNA, IL-13 mRNA, NF-κB mRNA, IkBα mRNA and Ikkβ mRNA were upregulated significantly (p < 0.05) in model group, especially in their gene expressions. Figure 11 shows that the phosphorylation of the NF-κB signaling pathway was decreased in all groups compared to the control and model groups, but with no significant differences (p > 0.05) between those affected groups. Furthermore, the results of RT-qPCR (also shown in Figure 11), were consistent with, but more obvious than, the WB results. It is worth noting that although the protein expressions of NF-κB in the lung tissue were downregulated in the nutritional intervention groups, and the activation of NF-κB inhibited, the level of downregulation was much lower than that of the drug groups.
Discussion
At present, few studies have reported the link between vitamin E and PF. Armutcu et al. illustrated that vitamin E plays a role in BLM-induced lung injury by reducing the inflammation caused by major inflammatory cytokines, such as IL-1β and IL-6, thereby providing a certain degree of protection and regulation [22]. Bese et al. reported that vitamin E supplementation immediately after irradiation could protect rats from radiationinduced PF [23], while Dede showed that vitamin E appeared to offset some of the severe lung damage in BLM-induced rats, as indicated by differences in serum concentrations [24]. In this work, the certain therapeutic effects of TRF and carotene on PF were revealed for the first time and compared with the effects of Nin and Pir. The main findings in our study are: (a) that TRF and carotene attenuate BLM-induced PF to a certain degree; (b) the mechanisms of TRF and carotene in the treatment of PF may be via the inhibition of TGF-β/Smad, PI3K/Akt/mTOR and NF-κB signaling pathways; (c) TRF and carotene have synergistic effects on some therapeutic indexes; and (d) the anti-inflammatory and antioxidant therapeutic effects of Nin and Pir in BLM-induced rats are similar.
The experiment described herein revealed a number of interesting results. In the previous studies on BLM model rats, the body weight generally decreased significantly, which lasted for a long time and recovered to varying degrees after the administration of different interventions [33,34]. However, in this paper, the body weight of the rats after modeling rose continuously after only a short weight loss period of 3-4 days. This weight change is similar to the experimental results of our prevention group [35]. The main difference between the two experimental processes was their modeling methods. The traditional modeling methods of PF are all surgical [36,37]. However, in this study, the tracheal instillation method was employed [38], as it not only reduces mortality but also reduces modeling time. The pathological results showed that our modeling was successful, as the values on the scoring criteria of HE and Masson's staining rose from 0.25 ± 0.16 and 0.37 ± 0.18 to 7.00 ± 0.19 and 7.63 ± 0.26, respectively, through the administration of BLM. The therapeutic effects of TRF and carotene were the same, and the degree of pathological reduction of the two drugs was also similar, while not all therapeutic effects were strengthened by the combination with TRF. For example, the score of the Masson's staining was not significantly reduced by the addition of TRF to Nin or Pir ( Figure 6). The immunohistochemistry results were similar to those of pathology. The collagen deposition was reduced to a certain extent in the nutritional intervention groups, but the total amount of collagen was still much higher than that in the drug treatment groups. Furthermore, while TRF alone did reduce collagen deposition to some extent, its combination with Nin and Pir did not reduce collagen deposition significantly compared to the single drug groups (Figures 7 and 8). In previous studies, the content of HYP in lung tissues has been noted as a quantitative index of fibrosis [39]. According to this indicator (Figure 2a), it was found that TRF, carotene, Nin and Pir alone could have a certain effect on the treatment of PF. However, when the carotene, Nin and PIR were combined with TRF, the values decreased and they played synergistic roles. These findings indicate that the addition of TRF to some drugs may reduce the symptoms of PF and strengthen treatment effects.
Recent studies show that PF can be prevented and treated by relevant natural products via the amelioration of oxidative stress, inhibition of inflammation and regulation of EMT, with mechanisms involving the p38 MAPK, TGF-β1/Smad, Nrf2-Nox4, NF-κB, PI3K/Akt and AMPK signaling pathways [40,41]. Our previously published experiments with a prevention group [35] showed that TRF and carotene can improve the prevention of PF through the TGF-β1/Smad, PI3K/Akt and NF-κB pathways. Nin is known to have the potential to inhibit the protein kinases involved in several molecular pathways and PI3K/AKT, JAK/STAT, TGF-β, WNT/β-catenin, VEGF and MARK are commonly reported in researches [42,43]. Moreover, numerous studies have reported the antifibrosis mechanisms of Pir, most of which involve TGF-β related genes and pathways [44][45][46]. Consequently, the three abovementioned pathways (TGF-β1/Smad, PI3K/Akt and NF-κB) were selected in this to ascertain whether TRF can enhance the therapeutic effect on PF, and if so, what is the specific mechanism involved therein. It was found that all proteins and genes were upregulated or downregulated to varying degrees following different interventions (Figures 9-11). Furthermore, the possible mechanism involved in the inflammatory response was found by downregulating the expression of the NF-κB protein and inhibiting the release of inflammatory downstream cytokines IFN-γ, IL-13 and TNF-α of the NF-κB signal pathway, inhibiting the TGF-β/Smad signaling pathway by downregulating the protein expressions of Smad2/3 and TGF-β1, upregulating Smad7, and suppressing the PI3K/AKT/mTOR pathway by downregulating the levels of the p-AKT, PI3K and mTOR proteins. These results indicate that Nin, Pir, TRF and carotene truly can improve or treat PF via these three pathways, although the up-and downregulation of Nin and Pir were found to be significantly greater than those of TRF and carotene. Following the incorporation of TRF, the improvement of the pathways was enhanced (although there was no significant difference in some proteins), which indicated that the mechanisms by which TRF enhances the effect of drug treatment on PF include the TGF-β1/Smad, PI3K/Akt and NF-κB signaling pathways. These results provide reason to believe that the combination of TRF with appropriate drugs offers good potential in the development of improved effective treatments for PF. However, due to the complexity of TRF, it has not yet been established whether the most effective of its components is indeed γ-tocotrienol and, therefore, further cell experiments are required. Population experiments could be carried out to further confirm whether the addition of TRF in PF treatment can reduce symptoms and improve therapeutic effects, thereby providing more possibilities in the treatment of PF. and Pir (100 mg/kg/D) combined intervention group. Following one week of adaptive feeding, PF modeling via tracheal instillation of bleomycin was performed on all rats except for those in the control group, according to previously reported modeling methods [37]. After 28 days of modeling, the abovementioned interventions were dissolved in 0.5% sodium carboxymethyl cellulose (CMC-Na) and administered to the corresponding groups, while the rats in the control and model (BLM) groups were intragastrically administered with equal volumes of 0.5% CMC-Na for 28 days. Body weight and feed intake were measured daily to monitor changes in growth. All rats were euthanized and their corresponding organs were collected for analysis on the 56th day (28 days of modeling and 28 days of intervention).
Histopathological Analysis
Hematoxylin-eosin (HE) and Masson's trichrome staining were performed according to the manufacturer's instructions. The lung tissues were dehydrated in graded alcohol, then fixed with paraffin blocks and cut into 4 µM flakes. The slices were dewaxed with xylene, rewatered with ethanol and washed with distilled water. Thereafter, they were stained with hematoxylin for 5 min, eosin for 2 min or the Masson compound dyeing solution for 5 min and, finally, with the bright green solution for 5 min. Routine dehydration, tissue removal and sealing were then carried out. Scoring criteria were based on those in a previously published report [47].
Immunohistochemical (IHC) Determination of Collagen I and Collagen II
Fresh lung tissue blocks, sized smaller than 0.5 cm × 0.5 cm × 0.1 cm, were washed with phosphate-buffered saline (PBS), fixed with 4% paraformaldehyde, soaked in ethanol, and then embedded in a stainless-steel mold. The tissue blocks were then cut to a thickness of 5 µm and immersed in xylene overnight. After dewaxing and the addition of water, they were washed with citric acid buffer for 20 min for antigen repair, then extracted with peroxide for 10 min and washed thrice with PBS solution for 5 min each time. Normal goat serum blocking solution was added dropwise on the slices for 20 min at room temperature, whereafter collagen I (AF7001, Affinity Biosciences, Beijing, China) and collagen II antibodies (AF0135, Affinity Biosciences, Beijing, China) were added dropwise, respectively; the slices were left at 4 • C overnight and then washed three times with PBS for 5 min each time. Next, biotinylated rabbit secondary antibody (SP-9001, ZSJQ-BIO, Guangzhou, China) was added to the slices and they were left for 20 min at 37 • C; streptomyces ovalbumin working solution (SA/HRP) labeled with horseradish enzyme was added to the slices, and they were left at 37 • C for 20 min. A diaminobenzidine (DAB) color kit was used for staining, after which the slices were, finally, counterstained and sealed. Image-Pro Plus 6.0 software (Rockville, MD, USA) was used to assess the results.
Statistical Analysis
The data values were presented as the mean ± standard error of the mean (±SEM). One-way analysis of variance was applied for the significance and post-multiple comparison was assessed by Dunnett's t-test (p-values of <0.05 were considered significant). Graph-Pad Prism 9.0 software (San Diego, CA, USA) was used to assess statistical significance.
Conclusions
In summary, our study demonstrated that TRF, carotene, Nin and Pir can all ameliorate PF by reducing inflammation and resisting oxidative stress via the TGF-β1/Smad, PI3K/Akt and NF-κB signaling pathways. Notably, when combined with TRF, the therapeutic effects of Nin and Pir on PF were enhanced. Thus, TRF may be a promising combination therapy for the treatment of PF in the future.
Author Contributions: Y.L. designed this experiment. All authors participated in this experiment. Y.Z., D.X. and Y.W. processed the data. Y.L. wrote the first draft of the original manuscript and W.L. helped with the revision. D.P. and P.W. helped with the tables, while J.X. and S.Y. prepared figures. G.S. and S.W. critically revised the manuscript. Y.L. is the guarantor, had full access to all of the data, and is responsible for the integrity of the data and the accuracy of the data analysis. All authors have read and agreed to the published version of the manuscript.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
Acknowledgments:
We thank the the management of MPOB, in general, for their support of this research and all of those who contributed to this article, especially Shiqing Chen, Fu Juyen and Yoong Jun Hao.
Conflicts of Interest:
The authors declare no conflict of interest. | 2022-11-23T16:32:01.023Z | 2022-11-01T00:00:00.000 | {
"year": 2022,
"sha1": "9c47eaccd93881d02d5e51fd735c031b780af81f",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1422-0067/23/22/14331/pdf?version=1668996687",
"oa_status": "GREEN",
"pdf_src": "PubMedCentral",
"pdf_hash": "97aeeae6f3954e17d6146a73af2f28f2c003d509",
"s2fieldsofstudy": [
"Medicine",
"Environmental Science"
],
"extfieldsofstudy": [
"Medicine"
]
} |
259095252 | pes2o/s2orc | v3-fos-license | Frailty trajectory predicts subsequent cognitive decline: A 26‐year population‐based longitudinal cohort study
Abstract Frailty refers to a decline in the physiological functioning of one or more organ systems. It remained unclear whether variations in the trajectory of frailty over time were associated with subsequent cognitive change. The aim of the current study was to investigate the association between frailty trajectories and subsequent cognitive decline based on the Health and Retirement Study (HRS). A total of 15,454 participants were included. The frailty trajectory was assessed using the Paulson–Lichtenberg Frailty Index, while the cognitive function was evaluated using the Langa–Weir Classification. Results showed that severe frailty was significantly associated with the subsequent decline in cognitive function (β [95% CI] = −0.21 [−0.40, −0.03], p = 0.03). In the five identified frailty trajectories, participants with mild frailty (inverted U‐shaped, β [95% CI] = −0.22 [−0.43, −0.02], p = 0.04), mild frailty (U‐shaped, β [95% CI] = −0.22 [−0.39, −0.06], p = 0.01), and frailty (β [95% CI] = −0.34 [−0.62, −0.07], p = 0.01) were all significantly associated with the subsequent cognition decline in the elderly. The current study suggested that monitoring and addressing frailty trajectories in older adults may be a critical approach in preventing or mitigating cognitive decline, which had significant implications for healthcare.
INTRODUCTION
In recent years, the global population of elderly individuals has steadily risen, leading to heightened concern regarding age-related clinical conditions that carry significant mortality and morbidity risks. 1,2 The cognitive impairment, which has become a major burden on the healthcare system and the economy, is estimated to affect approximately 50 million people, and the number is expected to triple by 2050. 3 The current therapeutic options for cognitive impairment are limited in their efficacy, and often carry the potential for deleterious side effects resulting from pharmacotherapy. Therefore, there is an increasing urgency for preventive measures to reduce the incidence and severity of cognitive impairment. This has become a critical issue that needs to be addressed urgently.
Frailty refers to a medical condition that is characterized by a decline in the physiological reserves of one or more organ systems, resulting in a reduced capacity to cope with physiological stressors. 4,5 The incidence of frailty appears to be highly variable across different studies and populations, and the reported prevalence rates of frailty range from low as 5% to high as 58%. 6 Fried's frailty phenotype has been widely recognized in the identification and characterization of frailty in older adults. 7 The five components of Fried's frailty phenotype include weight loss, limited physical activity, exhaustion, slowness, and weakness, and they reflect a broad range of physiological and functional domains that contribute to the overall concept of frailty. Based on Fried's frailty phenotype, the Paulson-Lichtenberg Frailty Index (PLFI) was found and validated to be an effective tool for assessing frailty in the Health and Retirement Study (HRS) dataset. 8 Recent findings suggested that frailty could pose a potential risk for cognitive impairment. The results showed that elderly individuals who were frail may have a higher likelihood of developing cognitive impairments. 9 Furthermore, there was supportive evidence indicating an association between alterations in the frailty status over time and the onset of cognitive impairment. Nari and colleagues reported that transitions between different frailty statuses (such as transitioning from non-frail to frail) were associated with different levels of cognitive risk. 10 Individuals who demonstrated persistent frailty had the highest risk of cognitive decline, particularly among women. In reality, the decline in frailty was a protracted and gradual process, characterized by a high degree of variability in the underlying processes of change. 11,12 Among older adults undergoing similar transitions in frailty status, there were noticeable differences in the trajectories of their frailty, indicating significant variability in the underlying processes of change. As a result, the utilization of trajectory analysis may offer a more effective means of identifying such patterns.
There may be an association between cognitive function and the frailty trajectory, according to earlier studies, and an association between the trajectory of frailty and cognitive performance. As an illustration, Howrey and colleagues implemented a group-based trajectory model to identify and classify participants into three different frailty groups based on their unique patterns of change in frailty over time: non-frail (n = 331), moderate progressive (n = 855), and progressive high (n = 149). Their subsequent analysis indicated that individuals categorized as being in the progressive high frailty group were more likely to experience a steep decline in cognitive function over time when compared to individuals in the other frailty groups. 13 Furthermore, several other studies have reported findings that support the association between frailty and cognitive function. 14,15 Although the association between frailty and cognitive function has been widely investigated, most studies have primarily focused on examining these variables concurrently rather than exploring the potential role of frailty trajectory on later cognitive decline. As a result, there remains a significant knowledge gap regarding the association between frailty trajectory and subsequent cognitive decline. Addressing this gap will be crucial for developing effective interventions and preventive measures that can improve outcomes and quality of life in older adults who are susceptible to frailty and cognitive impairment. In this study, we sought to enhance comprehension of the potential predictive value of frailty trajectories by utilizing a large cohort of the HRS and presenting new findings by examining diverse trajectories of frailty over time, with longitudinal changes in cognitive function as the outcome variable. To explore the association between frailty trajectories and cognitive decline in older adults, we employed a multilevel model that could account for the mutual correlation among multiple data when analyzing longitudinal data. 16 Moreover, we conducted important subgroup analyses by stratifying the sample by gender, BMI, and loneliness, to reveal the potential heterogeneity of the relationship between frailty and cognition.
Participant characteristics
A total of 15,454 participants were included, with a maximum follow-up period of 8 years. In our study, a total of five distinct frailty trajectories were identified and labeled as follows: no frailty symptoms, emerging frailty, mild frailty (inverted U-shaped), mild frailty (U-shaped), and frailty ( Figure 1), which reflects the different levels of frailty severity, as well as the shape of the trajectories themselves. There were 6701, 630, 1576, 4789, and 1758 participants in these five trajectories, respectively (Table 1). Participants in the frailty trajectory tended to be older, female, with lower educational levels, and were more likely to be widowed or separated/divorced compared to those in the other frailty trajectories.
Association of baseline frailty with the variation of cognition
To investigate the potential association between frailty and cognitive function, univariate analysis was conducted to assess the association between baseline frailty at Wave 9 and subsequent changes in cognitive function. The findings of the univariate analysis indicated that a higher level of frailty was associated with an increased risk of cognitive impairment (β [95% CI] = −2.01 [−2.22, −1.80], p < 0.001; Table 2). After adjusting for time, basic characteristics (age, gender, years of education, marital status, non-housing financial wealth, alcohol consumption status, smoking status, BMI), previous health condition, baseline cognitive function, and loneliness, the significant association between severe baseline frailty and a subsequent decline in cognitive function persisted (β [95% CI] = −0.21 [−0.40, −0.03], p = 0.03). This indicates that severe frailty at baseline was strongly associated with a decrease in cognitive function over time.
Association of frailty trajectory with the variation of cognition
Five trajectories of frailty have been identified, namely: no frailty symptoms, emerging frailty, mild frailty (inverted U-shaped), mild frailty (U-shaped), and frailty. The associations between frailty trajectories and cognitive function have been determined ( Figure 2). In Model 1, a univariate analysis was conducted to investigate the association between frailty trajectory and cognition. Other models were developed to provide a more comprehensive understanding of how the association between frailty trajectories and cognitive function was influenced by gender, BMI, and loneliness (Table S1). In Model A, the association between frailty trajectories and cognitive function was established after adjusting for time, basic characteristics (age, years of education, marital status, non-housing financial wealth, alcohol consumption status, smoking status), previous health condition, and baseline cognitive function. The result showed that mild frailty The results revealed that participants whose frailty trajectories showed frailty, as well as mild frailty (in inverted U-shaped and U-shaped), exhibited an elevated risk of cognitive impairment. Furthermore, this finding emphasizes the significance of accounting for BMI as a potential confounding factor when investigating the association between frailty trajectories and cognitive function among older adults.
Sensitivity analysis of the association of frailty trajectory with the variation of cognition
To further assess the stability and reliability of the model, three sensitivity analyses were conducted ( Figure 3 and Table S2). In the first sensitivity analysis, the analyses were reviewed by employing the cognitive score of participants from waves 10−13 as the outcome. The results showed that participants' frailty trajectories exhibited mild frailty 10], p < 0.01) were significantly associated with later cognition decline. The outcomes of these three sensitivity analyses were consistent with the primary analysis results, which indicated that the primary model in the main analysis could be applied robustly across various population cohorts.
Subgroup analysis of the association of frailty trajectory with the variation of cognition
Gender, BMI, and loneliness were included as stratification factors in the subgroup analyses to address the potential influence of interactions between these three variables and frailty trajectory. The results of the subgroup analysis revealed that the association between frailty trajectories and declining cognitive function was more sig- .00], p = 0.05) were associated with the later cognitive decline. Furthermore, this association remained consistent among participants with a non-obese BMI, as well as those with higher loneliness scores (Figure 4).
DISCUSSION
The present study aimed to provide a thorough examination of the complex relationship between the trajectory of frailty and the following deterioration in cognitive function. The results suggested that there was a substantial association between baseline frailty and subsequent declines in cognitive function over time, with individuals exhibiting higher levels of frailty at baseline being more likely to experience a decline in cognitive function. In addition to the association between baseline frailty and cognitive decline, the present study also examined the association between specific frailty trajectories and subsequent changes in cognitive function, revealing a significant association between mild frailty (inverted U-shaped), mild frailty (U-shaped), and frailty and subsequent decline in cognitive function. These associations were more prominent among females, participants with a non-obese BMI, and participants who report feelings of loneliness. Frailty and cognitive function have been extensively studied due to their potential interconnection in the context of aging-related decline. There appears to be a cyclical relationship between frailty and cognitive impairment, with each potentially exacerbating the other. 9 Prior research has identified an association between cognition F I G U R E 3 Sensitivity analyses of the association between frailty trajectory and the subsequent cognition decline. The model was adjusted for time, basic characteristics, previous health conditions, baseline cognition function, and loneliness. and frailty, as well as changes in frailty over time. 10,17,18 There may be a shared pathogenic underpinning for the association. 19,20 Buchman et al. estimated the rates of change in both frailty and cognition based on the data from 2167 older adults. 14 The majority of participants demonstrated simultaneous worsening of both frailty and cognition. Their findings highlighted a strong association between the baseline frailty level or change in frailty and cognition. This was further validated by Armstrong et al., who conducted a study on older Japanese-American men to investigate the association between cognitive function and transition in frailty status. 17 A total of 2817 men were recruited in the study, and the authors found a significant association between frailty status transition and cognitive decline in older adults, suggesting that cognitive decline may be a consequence of certain frailty transitions. However, it should be noted that the highly selective cohorts that were included in these two researches restricted the generalizability to the general population. Furthermore, many studies have primarily focused on the increasing frailty trajectory, rather than exploring the heterogeneity of frailty trajectories. The diversity of frailty trajectories was demonstrated in the current study. Specifically, the results showed that participants with mild frailty exhibited two different trajectory patterns, including U-shaped and inverted U-shaped trajectories.
Previous research suggested that females were more susceptible to cognitive impairment than males. 21 A prior study investigated how the degree and fluctuations in frailty impact three cognitive domains including memory, speed, and executive function. 15 This study also explored whether the association between frailty and cognition differed between males and females. The study's results showed that females may suffer from a more extensive cognitive decline than males when experiencing frailty. One of the reasons suggested by the study for that was woman has a higher level of frailty than man. 22 In the present study, we found that there was still a gender difference in the association between frailty trajectory and cognitive function, even when individuals had the same degree of frailty. The results suggested that beyond frailty, there were other factors such as pathophysiology that may affect the gender difference in cognitive impairment. 23 Prior studies have identified BMI as a significant factor that influenced both frailty and cognitive function in older adults. A previous meta-analysis has demonstrated that being overweight and being underweight were both related to an elevated risk of frailty among older adults living in the community. 24 On the other hand, the effect of BMI on cognitive function varied based on the age group, and previous studies showed that midlife underweight, obesity, and latelife underweight increase the risk of cognitive impairment while late-life overweight and obesity, decrease the risk F I G U R E 4 The association between frailty trajectory and subsequent cognition decline is stratified by gender, body mass index (BMI), and loneliness. The model was adjusted for time, basic characteristics, previous health conditions, baseline cognition function, and loneliness. of cognitive impairment. [25][26][27] However, there was a dearth of evidence on how BMI affects the association between frailty and cognitive function. In this study, after adjusting for BMI, the results indicated a significant association between mild frailty (inverted U-shaped) and later cognitive impairment. This finding underscored the need to consider BMI as a potential confounding variable when investigating the association between frailty trajectory and cognitive function in older adults. In the subgroup analyses of this study, participants who were not obese showed a strong association between frailty trajectory and cognition, whereas this effect was not prominent in obesity. This finding was consistent with previous research that suggested a higher BMI may have a protective effect in older adults, lowering the risk of cognitive impairment.
Additionally, in accordance with the prior studies, the results of the subgroup analyses involving participants with various levels of loneliness scores suggested that there was a significant and negative association between loneliness and cognitive function. 28 Loneliness was assessed using a three-item scale in HRS that evaluated the participant's sense of companionship, feelings of exclusion, and perceptions of social isolation (i.e., whether participants often feel they lack companionship, whether participants often feel left out, and whether participants often feel isolated from others). A previous study has demonstrated a high correlation between the total score of this three-item loneliness scale used in HRS and the total score of the UCLA loneliness scale. 29 So far, there is no literature that provides a definitive explanation for the cause-and-effect relationship between loneliness and cognitive function. According to prior research findings, there might be a reciprocal relationship between loneliness and cognitive function. 30 Loneliness was related to cognitive decline, potentially due to reduced social interaction and limited intellectual stimulation. Conversely, cognitive impairment may also contribute to reduced social interaction, leading to increased loneliness in turn. 29 The current study exhibits several strengths. First, the study utilized data from the HRS, a nationally representative longitudinal cohort, ensuring that the findings are generalizable to a broader population. This comprehensive investigation of the association between frailty trajectory and subsequent cognition provides important evidence for public health policy and practice. Additionally, the study examined the association between frailty and longitudinal cognition outcomes, allowing for a complete observation of evolution of the entire process and critical evidence of preventable causes of cognitive impairment. 31 This approach is essential for understanding the complex association between frailty trajectory and cognitive decline over time, and developing effective interventions and treatments that can prevent or alleviate cognitive decline in older adults. Moreover, the present study exhibited diverse trajectories of frailty, including U-shaped and inverted U-shaped mild frailty, in contrast to previous studies that focused primarily on the expected worsening tendency of frailty.
The present study has some limitations. First, due to the nature of cohort studies, there might be unobserved confounding variables and residual effects that could confound the association between frailty trajectories and cognitive function. Second, the assessment of frailty in HRS was self-reported, which may not be as precise as data obtained through clinical or laboratory tests. However, previous research has shown that the self-reported data from HRS were comparable to objective measures in identifying frailty, which provided support for the use of self-reported data. 7,8 Nonetheless, further research is needed to explore and validate the generalizability of these findings.
In summary, the present study has revealed the association between frailty trajectories and later cognitive function. The result showed that the associations were more pronounced among females, non-obese individuals, and those who report feelings of loneliness, highlighting the importance of considering these demographic factors in the development of interventions and treatments for cognitive decline. Overall, the results indicated that frailty trajectories could serve as a marker for identifying people at high risk of cognitive deterioration with age. Monitoring and addressing frailty trajectories may be a crucial strategy in preventing or alleviating the cognitive decline in older adults.
Participants
This study was designed based on the HRS, a nationally representative longitudinal survey of more than 37,000 individuals over age 50 in 23,000 households in the USA. 32 The details of the HRS have been described previously. 32 HRS was sponsored by the National Institute on Aging (NIA U01AG009740) and the Social Security Administration. The HRS collected data by interviewing or assessing participants every 2 years. The interviews were conducted by telephone, all respondents read a confidentiality statement when first contacted, and they give verbal consent by agreeing to do the interview. This consent procedure was approved by the Michigan Institutional Review Board. The sample and data for each assessment are identified using the ordinal number of the interview wave. The HRS project was initiated in 1992 and, as of the beginning of this study, the latest available HRS data covered a population of 26 years up until 2018. The longest follow-up time in this study spanned 8 years, from wave 9 to wave 13.
In our study, we included participants who had PLFI data available, which was the sum of the wasting, weakness, slowness, fatigue, and falls for at least one wave in waves 5, 7, or 9 ( Figure 5).
Frailty
In the present study, we used PLFI to evaluate the frailty of participants. 8 PLFI was validated as an effective tool for assessing frailty in the HRS. This index is the sum of F I G U R E 5 Diagrams illustrating the general study design (created with BioRender.com).
the five symptoms, with a total score from 0 to 5: wasting (i.e., the individual reported the loss of at least 10% of bodyweight over 2 years), weakness (i.e., because of health problems, do participants have any difficulty with lifting or carrying weights over 10 pounds, like a heavy bag of groceries), slowness (i.e., because of a health problem, do participants have any difficulty with getting up from a chair after sitting for long periods), fatigue (i.e., have participants had any of the following persistent or troublesome problems: severe fatigue or exhaustion?), and falls (i.e., have participants fallen down in the past 2 years). If the score is in the range 3-5, the individual is deemed as frailty. The assessment for each of the five symptoms was conducted in different waves in HRS, and data for "falls" was only available during odd-numbered waves. As a result, the PLFI data from waves 5, 7, and 9 were utilized to analyze the trajectory of frailty.
Cognition performance
The Langa-Weir Classification is a 27-point scale that included immediate and delayed recall, serial 7s, and back-ward counting items. Respondents who scored from 0 to 6 were classified as having dementia. This classification had been validated against the prevalence of cognitive states assessed by the Aging, Demographics, and Memory Study (ADAMS) and used to evaluate the cognitive function of participants in this study. 33 To investigate the longitudinal changes in cognitive function after PLFI assessment (waves 5,7,9), the Langa-Weir Classification scores from participants in waves 9-13 were determined as the study outcome, to determine the impact of frailty on cognitive function over time.
Covariates
Demographic and economic characteristics, lifestyle factors, and long-term illness conditions were selected as covariates. They included age, gender (male or female), years of education (≤12 or >12), marital status (partnered or without a partner), non-housing financial wealth (in quintile, from Q1 to Q5), alcohol consumption status (yes or no), smoking status (yes or no), body mass index (BMI, underweight, normal weight, overweight, and obesity), loneliness (mean score), and long-term illness such as hypertension, diabetes, cancer, lung disease, heart disease, psych disease, or arthritis (yes or no).
Statistical analyses
Characteristics were compared by using t-tests for continuous variables and chi-square for categorical variables. Latent class trajectory models were used to evaluate the trajectory of frailty over time. This finite mixture model is designed to identify clusters of individuals following similar progressions of some behavior or outcome over age or time. 34 The optimal number of trajectories is usually selected based on the Bayesian information criterion (BIC). In this study, the model selection was based on the available literature, BIC, and clinical plausibility. 35 The multilevel model was established to investigate the association between frailty trajectories and cognitive function variation. β (regression coefficient) and 95% confidence intervals (95% CIs) were calculated to estimate the association. Five models were constructed. Model 1 was the univariate model to analyze the association between frailty trajectory and cognition. Model 2 was additionally adjusted for time based on Model 1. Model 3 was additionally adjusted for basic characteristics including age, gender, years of education, marital status, non-housing financial wealth, alcohol consumption status, smoking status, and BMI based on Model 2. Model 4 was additionally adjusted for the previous health condition based on Model 3. Model 5 was additionally adjusted for baseline cognitive function based on Model 4. Model 6 was additionally adjusted for loneliness based on Model 5.
To demonstrate the effects of gender, BMI, and loneliness on the association between frailty trajectories and cognitive function, we constructed four new models. Model A analyzed the association between frailty trajectory and cognition and adjusted for time, basic characteristics (age, years of education, marital status, non-housing financial wealth, alcohol consumption status, smoking status), previous health condition, and baseline cognitive function. Model B was additionally adjusted for gender based on Model A. Model C was additionally adjusted for BMI based on Model B. Model D was additionally adjusted for loneliness based on Model C.
Three sensitivity analyses were conducted in Model 6 to test the robustness of the results from the primary analyses. In sensitivity analysis 1, we reconducted the analyses by using the cognitive function assessment data of participants from waves 10-13 as the outcome. In sensitivity analysis 2, we excluded the participants with cognition points less than 7. In sensitivity analysis 3, we excluded patients who died within 2 years after the frailty evaluation.
Subgroup analyses were conducted to show the potential influence of interactions between gender, BMI (obesity or other), and loneliness and frailty trajectory. The evaluation of loneliness was the summary score of three questions (i.e., whether participants often feel they lack companionship, whether participants often feel left out, and whether participants often feel isolated from others.), and the higher scores indicated that the participants feel more lonely.
All analyses were conducted using R 4.0.2, and a twosided p-value less than 0.05 was considered significant statistically.
A U T H O R C O N T R I B U T I O N S
Methodology, conceptualization, software, data curation, investigation, software, validation, writing-review and editing: RDL, ZRL, RDH, and YC. Conceptualization, methodology, data curation, writing-original draft, investigation, supervision, writing-review and editing: YLS, XLH, and XCP. Data curation, investigation, writingreview and editing: ZGW, LH, and YYP. All authors have read and approved the final manuscript. Figure 5 was created with BioRender.com (www.biorender. com).
C O N F L I C T O F I N T E R E S T S TAT E M E N T
The authors declare that they have no conflicts of interest.
D ATA AVA I L A B I L I T Y S TAT E M E N T
The HRS database is publicly available and can be accessed by submitting a reasonable request.
E T H I C S S TAT E M E N T
Ethical approval for the HRS Study was obtained from the University of Michigan Institutional Review Board (HUM00061128), and the study has been conducted according to the principles expressed in the Declaration of Helsinki. | 2023-06-08T05:08:26.784Z | 2023-06-01T00:00:00.000 | {
"year": 2023,
"sha1": "470f52d3b8fd903d06834899a63ea26d94a0de7c",
"oa_license": "CCBY",
"oa_url": null,
"oa_status": null,
"pdf_src": "PubMedCentral",
"pdf_hash": "470f52d3b8fd903d06834899a63ea26d94a0de7c",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
248894786 | pes2o/s2orc | v3-fos-license | Outcomes of Transscleral Two-Point Fixation Versus Closed Continuous-Loop Four-Point Fixation of Intraocular Lens in Subluxated Lens Secondary to Marfan Syndrome
Introduction To compare the outcome of two different transscleral fixation approaches for posterior chamber intraocular lens (IOL) implantation, a two-point fixation of the Sensar (Allergan) or CZ70BD (Alcon) IOL and a four-point fixation of the Akreos Adapt (Bausch & Lomb) foldable IOL, for treatment of subluxated lenses in Marfan syndrome (MFS). Methods Fifty-three eyes of 33 consecutive patients with subluxated lenses secondary to MFS were studied. Eighteen patients with MFS (30 eyes) received two-point fixation of the Sensar (16 patients, 26 eyes) or CZ70BD (2 patients, 4 eyes) IOL, and 15 patients with MFS (23 eyes) received four-point fixation of the Akreos Adapt IOL. Preoperative and postoperative ophthalmologic examinations were performed. A primary outcome measure of postoperative complication was studied. Results The mean preoperative best corrected visual acuity (BCVA) in the two-point group was 0.68 ± 0.38 logarithm of the minimum angle of resolution (logMAR), and it improved to 0.30 ± 0.32 logMAR at the final follow-up (p < 0.05). The mean preoperative BCVA in the four-point group was 0.68 ± 0.45 logMAR, and it improved to 0.28 ± 0.28 logMAR at the final follow-up (p < 0.05). The BCVA results did not differ significantly between groups. The intraocular pressure was increased at the final follow-up in the two-point group (p < 0.05). Transscleral two-point fixation of IOL has relatively high incidences of pupillary capture of the IOL. Conclusion The closed continuous-loop transscleral four-point fixation of the Akreos Adapt foldable IOL is more suitable than two-point fixation of a two-haptic IOL in treating subluxated lenses due to MFS.
Key Summary Points
Why carry out this study?
Transscleral two-point fixation of posterior chamber intraocular lens (IOL) may increase the risk for IOL tilt for treatment of subluxated lenses secondary to Marfan syndrome (MFS).
A closed continuous-loop transscleral suture for four-point fixation of the Akreos Adapt foldable IOL has benefits of a reliable and reproducible procedure with improved anatomical and visual outcomes, reduced complications, and decreased surgical times.
What was learned from the study?
This is the first study to compare the outcome of two different transscleral fixation techniques for posterior chamber IOL replacement for treatment of subluxated lenses in MFS.
As a result of more frequent IOL capture in the two-point group, this is the first study to propose that the closed continuousloop transscleral four-point fixation of the Akreos Adapt foldable IOL might be a simple and minimally invasive method for subluxated lens in MFS.
INTRODUCTION
Marfan syndrome (MFS) is a systemic disorder of connective tissue caused by mutation in FBN1, first described in 1896 by a Parisian professor of pediatrics, Antoine-Bernard Marfan [1]. A major feature of the ocular system in MFS is ectopia lentis (subluxation/dislocation of the ocular lens) [2]. A variety of surgical repair methods are available for the management of clinically significant subluxated crystalline lenses in MFS. After initial lens removal, surgical techniques for intraocular lens (IOL) implantation include in-the-bag IOL placement by using capsular tension rings [3] (i.e., \ 6 clock hours of zonulysis) and placement of anterior chamber IOLs [4], iris-fixed IOLs [5], transscleral sutured posterior chamber IOLs [6], even sutureless flanged IOLs [7] in the absence of adequate capsular support (i.e., [ 6 clock hours of zonulysis). However, there is a relatively high rate of postoperative complications such as corneal endothelial damage, peripheral anterior synechia, and glaucoma in anterior chamber IOL placement [8]. The scleral fixated sutured posterior chamber IOL had a more favorable outcome and lower complication rate than the anterior chamber IOL group [9]. Therefore, transscleral ciliary sulcus fixation of posterior chamber IOLs, based on the technique first described by Malbran et al. [10] in 1986, can be considered for patients with MFS.
Numerous techniques of transscleral fixation of posterior chamber IOLs have been developed, including the traditional two-point large-incision fixation of CZ70BD (Alcon) IOL and smallincision fixation of Sensar TM AR40e (Allergan) IOL [11]. However, transscleral two-point fixation of posterior chamber IOLs might increase the risk for IOL tilt and the consequent pupillary capture [6,12]. As published previously [11,13,14], a closed continuous-loop transscleral suture for four-point fixation of the Akreos Adapt foldable IOL has benefits of a reliable and reproducible procedure with improved anatomical and visual outcomes, reduced complications, and decreased surgical times. Thus, in this study, we estimated the outcome and complication of two different surgical methods for treatment of subluxated lens secondary to MFS: placement of a transscleral two-point fixation of the Sensar or CZ70BD IOL or a closed continuous-loop transscleral four-point fixation of the Akreos Adapt (Bausch & Lomb) foldable IOL during pars plana vitrectomy (PPV) and lensectomy.
Participants
The study protocol was approved by the Institutional Review Committee of Eye Hospital of Shandong First Medical University (SDSY-KYY200608) and was conducted in accordance with the tenets of the Helsinki Declaration. Possible complications had been explained, and all patients provided their informed consents prior to inclusion in the examination. The inclusion criteria were patients with subluxated lens ([ 6 clock hours of zonulysis) secondary to MFS, referred after genetic or cardiologic evaluation, and the margin of subluxated lens could be observed without mydriasis by slit lamp (Fig. 1). The exclusion criteria were ocular trauma, preoperative intraocular pressure (IOP) of 25 mmHg or more while receiving treatment with eye drops; scleritis; and postoperative follow-up for less than 1 month.
Intervention
All patients underwent standard preoperative and postoperative ophthalmologic examinations including measurements of uncorrected and best corrected visual acuity (VA), slit-lamp examination, measurement of IOP, and dilated indirect slit-lamp biomicroscopy. Clinical data were collected, including age, sex, best corrected visual acuity (BCVA), refractive error, IOP, and complications.
Technique 1
After a conjunctival peritomy created generally at the 12 o'clock and 6 o'clock, blunt dissection of episcleral tissue and electric coagulation of the surgical field were performed. Transscleral passage of a straight fine needle attached to a 10-0 polypropylene suture (Ethicon, ZW 1713) was performed 1.5 mm posterior to the corneal limbus into the posterior chamber and the needle was then pulled out of the globe using the guidance of 29-gauge insulin needle along the 12 o'clock and 6 o'clock axes, leaving the 10-0 suture traversing the eye from one scleral bed to the other. A 2.65 mm clear corneal incision (CCI) according to the steep axis was made in the anterior limbus using a keratome. The anterior chamber filled with an ophthalmic viscosurgical device. The 10-0 suture was pulled out through the CCI pocket with a hook and was cut off. A foldable IOL of the Sensar AR40e (Allergan) was loaded into the Emerald C Cartridge (Lot CH00841 of AMO Inc., USA) and the cartridge was inserted into the injector. The surgeon removed the injector system and the leading haptic was kept outside of the injector. One suture end was tied at the midpoint of the Fig. 1 Preoperative figure demonstrating the margin of subluxated lens due to MFS without mydriasis using slit lamp haptic (3 mm apart from the end of the haptic). Then the foldable IOL was inserted into the anterior chamber through the CCI using an injector. The trailing haptic was kept outside to prevent the IOL from falling into the vitreous cavity and tied by another suture end. The IOL was placed in the posterior chamber and the haptics were oriented into the sulcus by pulling the sutures. After achieving centration of the IOL, ends of the suture were fixed on the surface of the sclera. Five zigzag passes were done, the suture was tied onto the sclera at about 5 mm posterior to the limbus, and suture ends were cut long ([ 10 mm) and buried under the rectus in all cases. The infusion cannula was removed and the conjunctival peritomy was closed.
The current strategy involves insertion of CZ70BD into the ciliary sulcus. The procedure was the same as previously described [15]. After a standard posterior infusion was placed, a toric lens marker was used to mark the corneal limbus at two points 180°apart along the horizontal axis. Each pair of nasal and temporal sclerotomies was made 2-3 mm posterior to the limbus and 2 mm apart. A 7-mm scleral tunnel incision was constructed. The suture was cut into equal halves and threaded through the eyelet of the CZ70BD IOL. Each end of the suture was pulled out of each corresponding sclerotomy using serrated forceps. The sutures were tied carefully to ensure the IOL was centered and suture tension was balanced.
The IOL used in this procedure was the Sensar AR40e (Allergan) or CZ70BD (Alcon). Sensar AR40e is a foldable three-piece hydrophobic acrylic spherical IOL with two open-loop haptics and has an optic diameter of 6.0 mm and an overall length up to 13.0 mm. The haptics of the Sensar AR40e IOL are made of polymethylmethacrylate (PMMA) and the lens is made of hydrophobic acrylic. Both materials favor mechanical stability of the IOL. The manufacturer's recommendation for A constant is 118.4. CZ70BD is a large-diameter optic (6.5 or 7.0 mm) one-piece PMMA IOL with large modified C-shaped haptics and has an overall length up to 12.5 mm. The manufacturer's recommendation for A constant is 118.8. The optic power was calculated using the SRK/T formula to achieve emmetropia.
Technique 2
The surgery was performed as previously described by Zhang et al. [13,14]. Four sclerotomy sites were marked 2 mm from the limbus and located 3.5 mm superior and inferior to the horizontal corneal markings. A four-haptic foldable IOL of the Akreos Adapt (Bausch & Lomb) was inserted into the anterior chamber using an injector. Transscleral passage of a straight fine needle attached to 10-0 polypropylene suture (Alcon) was performed 2 mm posterior to the corneal limbus into the posterior chamber. The needle was then passed through one eyelet on the bottom-right haptic side of the IOL and pulled out of the globe using the guidance of a 29-gauge insulin needle that passed through one eyelet on the bottom-left haptic side of the IOL, leaving the 10-0 suture traversing the eye and two eyelets of IOL from bottom-right scleral bed to the bottom-left one and paralleling to the horizontal meridian. Next, the needle was pulled out subconjunctivally from bottom-left scleral fixation point to the upper-left one by a needle holder. The same procedures were repeated to introduce the fixation thread into the eye and the other eyelets of the IOL through two other scleral fixation points of the upper side. Finally, the needle was pulled out subconjunctivally from upper-right scleral fixation point to the bottom-right one by a needle holder. The suture was then looped through four eyelets on the IOL haptics and scleral fixation points. Two ends of the suture were tied on the surface of the sclera. The ends of the suture were cut and the suture knot was buried by carefully rotating it into the sclera tunnel.
The IOL used in this procedure was the Akreos Adapt (Bausch & Lomb), which is a foldable hydrophilic acrylic spherical IOL with four closed eyelets at the distal end of the haptics and has an optic diameter of 6.0 mm, an overall length up to 10.5-11.0 mm. The optic power was calculated using the SRK/T formula to achieve emmetropia. The manufacturer's recommendation for A constant is 118.0.
Statistical Analyses
The decimal VA was converted to the logarithm of the minimum angle of resolution (logMAR) units for the statistical analyses. Student's t test was used to compare preoperative and postoperative BCVA and IOP. The chi-square test was used to compare groups at each time point (preoperatively and at the final follow-up postoperatively) with regard to the percentage of eyes in the range of C 20/40, B 20/50, and B 20/200. A p value of less than 0.05 was considered to be statistically significant. The statistical analyses were performed by using the Statistical Package for the Social Sciences program (SPSS 18.0 for Windows; Chicago, IL).
RESULTS
A total of 53 eyes of 33 patients (20 men and 13 women) with MFS were included in the study. The mean age was 19 years (range 4-54).
Visual Acuity
The two-point group contained 30 eyes of 18 patients with a mean age of 19 years (range 4-52). The mean follow-up time after implantation of an IOL was 51.7 ± 38.3 months (range 1-114). Comparison of preoperative and final postoperative BCVA of these patients showed that the mean logMAR decreased from 0.68 ± 0.38 preoperatively to 0.30 ± 0.32 at the final postoperative examination (Fig. 2). This improvement in BCVA was statistically significant (p = 0.0001).
The four-point group included 23 eyes of 15 patients with a mean age of 19 years (range 5-54). The mean follow-up time after implantation of an IOL was 7.7 ± 10.8 months (range 1-37). Comparison of preoperative and final postoperative BCVA of these patients showed that the mean logMAR decreased from 0.68 ± 0.45 preoperatively to 0.28 ± 0.28 at the final postoperative examination (Fig. 3). This improvement in BCVA was statistically significant (p = 0.0007).
To compare the two different approaches, the percentages of eyes with BCVA C 20/40, B 20/50, and B 20/200 preoperatively and at the final follow-up postoperatively were analyzed, and we found that there was no difference between groups (Table 3).
Furthermore, the proportion of all patients defined as legally blind (B 20/200) decreased from 22.64% (12 eyes) before surgery to 7.55% (4 eyes) after surgery. Of the four eyes with a BCVA B 20/200 at the final follow-up, one had a choroidal neovascularization at the fovea, one had severe amblyopia, and the other two suffered from a secondary glaucoma.
Intraocular Pressure
For the two-point group, compared with a mean preoperative IOP of 13.8 ± 3.8 mmHg, IOP was 16.0
Complications
Complications in the two-point group included IOL capture (n = 11, Fig. 4), secondary glaucoma (n = 2), rhegmatogenous retinal detachment that was successfully repaired (n = 2), IOL pigment deposition (n = 2), suture exposed (n = 1), and choroidal neovascularization at the fovea (n = 1). In 11 eyes with iris capture by the IOL, the iris capture was recovered by coroplasty in one eye and a miotic agent was used in other 10 eyes to prevent recurrence.
In the four-point group, no IOL capture, suture erosion, suture loosening, hypotony, scleral atrophy, chronic inflammation, or retinal tear and/or detachment was observed in any patients. The IOLs were well centered within the follow-up period (Fig. 5). It is worth noting that both eyes of a patient with MFS in the fourpoint group had corneal endothelial decompensation before operation, one of which received corneal endothelium transplantation 3 months after IOL implantation and achieved a BCVA of 20/100 at the final follow-up, with mild edema and scattered pigmented keratic precipitates appeared on the corneal graft.
No patient experienced postoperative corneal decompensation, suprachoroidal hemorrhage, hyphema, chronic uveitis, or endophthalmitis in both groups during the follow-up period.
DISCUSSION
In cases of severe lens subluxation secondary to MFS, capsular tension rings may not always be applicable [16], and placement of anterior chamber IOLs had a higher complication rate and more contraindications than the posterior chamber IOLs group [17], so transscleral fixation of posterior chamber IOL approach has become more advisable. As published previously [13,14], a closed continuous-loop transscleral suture for fourpoint fixation of the Akreos Adapt foldable IOL has achieved a reliable and reproducible procedure with improved anatomical and visual outcomes, reduced complications, and decreased surgical times. So in the current study, this procedure of four-point fixation of the Akreos Adapt IOL (15 patients, 23 eyes) was compared with traditional transscleral twopoint fixation of the Sensar AR40e or CZ70BD IOL (18 patients, 30 eyes) in MFS between August 2006 and July 2021. The BCVA at the final follow-up improved in both groups, consistent with previous studies. Asadi and Kheirkhah [6] evaluated the long-term results of transsclerally fixated posterior chamber IOL in children without adequate capsular support, including six eyes with MFS, and demonstrated that BCVA improved postoperatively in 48% eyes by more than one Snellen line. McClellan et al. [5] described 24 eyes from 17 consecutive adult patients with surgically treated subluxated crystalline lenses, including 14 eyes (58%) with MFS, and revealed that placement of iris-sutured posterior chamber IOL at the time of subluxated lens extraction with a pars plana surgical approach yields favorable results in terms of postoperative visual outcomes and surgical complications. The main cause of legal blindness in four eyes postoperatively was coexistent retinal pathologies, amblyopia, and secondary glaucoma.
Transscleral two-point fixation of IOL has relatively high incidences of pupillary capture of the IOL. In my series of 30 eyes in the twopoint group, 11 eyes displayed pupillary capture and were treated medically. No pupillary capture was seen in all cases of four-point scleral suture fixation of the Akreos Adapt IOL. In some series, pupillary capture of the IOL optic has occurred in 7.9-14.3% of cases after scleral fixated sutured posterior chamber IOL implantation [18], but patients with MFS tended to have a higher rate of pupillary capture of the IOL [19]. IOL pupillary capture is usually transient, but it might cause complications such as pupillary block with secondary glaucoma, pigmentary dispersion glaucoma and iritis as well as refractive change [20,21]. In the current study of the two-point group, compared with a mean preoperative IOP of 13.8 ± 3.8 mmHg, IOP increased to 16.0 ± 4.3 mmHg at the final follow-up (p \ 0.05), which might be caused by high rates of pupillary capture of the IOL. Floppy iris, pliable iris, reverse pupillary block, a larger anterior chamber depth, abnormal fluidics of aqueous humor, the IOL material, and inaccurate two-point sclerotomies are suggested as possible causes of pupillary IOL capture [22][23][24]. Scharioth [25] found that the most important prognostic factor for postoperative iris capture in eyes with good IOL positioning is an intraoperative reverse pupillary block, and suggested performing a surgical iridectomy (preferably with a small-gauge vitrectome) with adding postoperative pilocarpine therapy to treat iris capture in patients with transscleralfixated posterior chamber IOLs. In conclusion, four-point scleral suture fixation of the Akreos Adapt is more suitable than two-point fixation of a two-haptic IOL in treating subluxated lens due to MFS.
One limitation of the current study was that only VA outcomes, IOP, and complications were observed. To further compare the two surgical approaches in MFS, more comprehensive clinical data such as endothelial cell density, IOL tilt, and astigmatism need to be collected. Moreover, fewer complications in the four-point group of this study at the final follow-up were probably due to shorter follow-up time. So the small sample size and limited follow-up might have biased our results. In view of the suggestive results obtained, we are recruiting more patients with MFS and collecting more detailed clinical characteristics of patients to confirm our findings.
CONCLUSIONS
This profile demonstrated that both the transscleral two-point fixation of the Sensar or CZ70BD IOL and the closed continuous-loop transscleral four-point fixation of the Akreos Adapt foldable IOL improved BCVA at the final follow-up postoperatively. As a result of more complications in the two-point group, the closed continuous-loop transscleral four-point fixation of the Akreos Adapt foldable IOL is a good option for subluxated lens secondary to MFS.
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/bync/4.0/. | 2022-05-20T13:37:18.543Z | 2022-05-20T00:00:00.000 | {
"year": 2022,
"sha1": "fa1ac8b67b993b7c46cb2e6c501a0852bf9812b1",
"oa_license": "CCBYNC",
"oa_url": "https://link.springer.com/content/pdf/10.1007/s40123-022-00519-6.pdf",
"oa_status": "GOLD",
"pdf_src": "Springer",
"pdf_hash": "fa1ac8b67b993b7c46cb2e6c501a0852bf9812b1",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
201654823 | pes2o/s2orc | v3-fos-license | Clinico-pathological considerations in a 48-years-old female with acute kidney injury: is it lupus nephritis, ANCA-associated vasculitis or something else?
Background The value of ANCA positivity in the setting of systemic lupus erythematous and their pathogenicity remains uncertain. Case presentation We report the case of a 48-year-old female with rapidly progressive kidney failure, arthro-myalgia and weight loss. Auto-immune screening showed anti-dsDNA antibodies, complement consumption and triple ANCA positivity. A first kidney biopsy done at presentation highlighted class IV-G glomerulonephritis with elective extra-capillary involvement and mainly C1q glomerular deposition at immunofluorescence study. After three months of a regimen combining steroids and cyclophosphamide, a second biopsy was performed and showed class IV-G glomerulonephritis with mainly endocapillary proliferation. Conclusion This case is atypical in view of immunological profile and kidney histopathological presentation and evolution and gives rise to discussion in view of recent data on ANCA value in lupus nephritis, and suggests that different auto-immune pathways may be involved in lupus nephritis.
Background
Several lines of evidence indicate that anti-neutrophil cytoplasmic antibodies (ANCAs), especially myeloperoxidase (MPO)-ANCAs, are pathogenic auto-antibodies in ANCAs-associated vasculitis (AAV) [1]. However, ANCAs can also be detected in healthy subjects and in a number of inflammatory/auto-immune diseases where their pathophysiological significance remains debatable [2].
Detection of ANCAs at diagnosis or in the course of systemic lupus erythematosus (SLE) is quite common, reported in up to 20% of patients [3]. According to literature, perinuclear ANCAs (p-ANCAs) rather than cytoplasmic ANCAs (c-ANCAs) are detected using indirect immunofluorescence (IIF) in SLE [3,4].
However, the significance of these ANCAs is quite variable among SLE patients, some of them having ANCAs directed towards minor ANCA antigens (ie, lactoferrin, BPI, elastase.), and others towards major ANCA antigens (MPO or PR3), the latters' being those associated with AAV [3]. Even if ANCA positivity has been assoiated with higher disease activity in SLE patients, their potential role in organ injury still have to be demonstrated [3].
Recently, Tuner-Stokes et al. studied retrospectively more than 200 kidney biopsies from patients with lupus nephritis (LN) and compared histopathological features of those associated with ANCAs at biopsy to those without [4]. As already reported in past studies, a predominance of p-ANCA of MPO specificity was observed in SLE patients [3]. More interestingly, they found that patients with ANCA positivity had more frequently diffuse glomerulonephritis with segmental involvement (class IV-S LN ISN/RPS classification) and lesions of glomerular necrosis, as compared to patients without ANCAs. Moreover, ANCA positive patients had a worse kidney function at the time of biopsy, higher anti-dsDNA antibody levels and lower complement levels. Thus, ANCAs positivity was associated with a specific kidney histological and biological phenotype in SLE patients. These observations allow to reopen an old debate on the potential pathogenic action of ANCAs in LN [5].
Here, we report the case of a 48-years old female referred to our department with systemic symptoms and acute kidney injury. The clinical and biological presentation was very suggestive of SLE, but strong ANCA positivity was also detected. The kidney histology at admission and its evolution 4 months after treatment initiation was uncommon and gives rise to discussion.
Case presentation
A 48-years old Asian female was referred to our hospital with a recent history of weakness, myalgia and arthralgia. She also complained of anorexia with 3 kg weight loss. She had no past medical history and was free of any medication on admission. At presentation, blood pressure was 180/80 mmHg, heart rate was 91/min, temperature was normal. Heart and lung auscultation, as well as abdominal examination, were normal. She had no skin involvement nor lymphadenopathy or synovitis.
Biology showed acute kidney injury with serum creatinine at 209 μmol/L. White blood cell count (4.8 G/L) and platelet count (269 G/L) were normal, but anemia with hemoglobin at 7.8 g/dL was present. C-reactive protein was slightly increased (14 mg/dL). Urinalysis revealed glomerular proteinuria (proteinuria to creatinine ratio (P/C) 3.6 g/g) and microscopic hematuria. Kidney ultrasound examination showed normal sized kidneys and excluded obstruction. Thus, we concluded to acute glomerular syndrome and performed immunological laboratory tests. Antinuclear antibodies (ANA) (1/2560), as well as anti-dsDNA (292 UI/mL), anti-SSA and anti-SSB antibodies were detected. Type 3 cryoglobulinemia and complement consumption (decreased C3, C4 and CH50) were also present. Search for lupus anticoagulant, IgG anti-cardiolipin and anti-beta-2GP1 antibodies was negative. Finally, pANCA were detected at 1/2000 titer using IIF. ELISA showed the concomitant presence of MPO, PR3 and lactoferrine ANCAs with a strong positivity for MPO and lactoferrine ANCAs and a low positivity for PR3 ANCAs. ANCA detection using multiplex technology also detected MPO ANCA at high level, and low PR3 ANCA level.
LN was first considered leading us to perform a kidney biopsy. On optical examination, extra-capillary circumferential cellular or fibro-cellular crescents were observed in 10 of 16 glomeruli. Segmental endocapillary proliferation was absent to very modest, without any lesions of capillary necrosis (Fig. 1). Immunofluorescence analysis showed strong and diffuse mesangial and parietal C1q staining with a granular deposition pattern, while IgG, IgA, IgM, and C3 deposits were quite limited, of mesangial topography and only segmental (Fig. 1).
Following biopsy, an immunosuppressive treatment was initiated with an association of steroids, hydoxychloroquine and pulse intravenous cyclophosphamide (CYC, 500 mg every two weeks) according to the Euro-Lupus protocol trial [6]. Under this regimen, a progressive improvement of both her general condition and kidney function was observed. At month 3 from treatment initiation (after the 6th CYC injection), serum creatinine was 116 μmol/L and P/C ratio decreased to 0.50 g/g. Antinuclear antibodies decreased to 1/200, anti-dsDNA antibodies became undetectable, and complement returned within normal range. ANCAs were still detectable although at lower titer using IIF (1/200), with only MPO ANCAs remaining slightly positives at ELISA and multiplex assays.
At that time, we decided to perform a systematic kidney biopsy to analyze histological response to treatment. On optical examination, only fibrotic crescents were observed in 9/14 glomeruli, none of them being cellular of fibro-cellular. Global lesions of endocapillary proliferation were observed in most glomeruli, without lesions of capillary necrosis (Fig. 2). Immunofluorescence analysis showed diffuse mesangial and parietal C1q staining at a lower intensity as compared to the diagnostic biopsy. IgG, IgA, IgM, and C3 deposits remained limited, in their pattern and intensity (Fig. 2).
After the biopsy results, we decided to continue CYC. However, after the 8th CYC injection, the patient developed toxidermia which we attributed to CYC. At that time, we decided to start mycophenolate mofetil (MMF) at 1.5 g/day.
One year after initial admission and under MMF for 12 months, she has no clinical manifestation of SLE and did not relapse nephritis. Renal function returned to near normal values (serum creatinine of 90 μmol/L, MDRD eGFR 60 mL/min/1.73m 2 ). ANCA detection was negative on IIF evaluation and ELISA, with anti-MPO being still detectable at very low level using multiplex assay. ANA were stable at 1/200, complement was in normal range and search for cryoglobulin was negative.
Discussion and conclusion
The present case represents a very uncommon overlap presentation of SLE and AAV according to both its immunological and kidney pathological aspects. First, our patient presented with a typical SLE serology, but had also circulating auto-antibodies towards three different ANCA antigens, which is an extremely rare condition. Second, the kidney pathological lesions observed at diagnosis, as well as their evolution on control biopsy, were uncommon in comparison to those usually observed in proliferative forms of LN [7,8]. The present case illustrates the complexity of immune mechanisms acting in SLE, and raises the question of the importance of histology findings in the choice of treatment in such situations.
The first atypical aspect in this observation is the coexistence of a typical SLE serology (ANA, anti-dsDNA ab, complement consumption) and ANCA positivity. As mentionned above, ANCAs, mainly p-ANCAs of MPO specificity, are detected in up to 20% of SLE patients, which allows to conclude that it is finally a quite common condition. However, in our patient, ANCAs directed towards three different ANCA antigens, namely MPO, PR3 and lactoferrin, were detected. According to the persistence of p-ANCA pattern while anti-dsDNA disappeared, a cross reactivity anti-dsDNA was unlikely [9]. This condition represents a quite exceptional situation that may be related to particular auto-immune pathways acting in our patient. In contrast with double positive ANCAs, directed towards a major ANCA antigen (MPO or PR3) and towards a minor ANCA antigen, that is finally frequently observed (notably in AAV), ANCA positivity towards both MPO and PR3 appears to be extremely rare [10,11]. Indeed, in a past study, only 23 patients with this phenotype among 3095 consecutive ANCA positive patients, representing less than 1% of patients [12]. Interestingly, most patients with this condition had inflammatory or auto-immune diseases including SLE, but none had AAV. Confirming the rarity of this association, in a recent study analyzing ANCA positivity value in LN patients, double MPO/PR3 positivity was observed in 3/32 patients with ANCA positivity among 254 consecutives LN patients, thus randomly 1% of LN patients [4].
The second atypical aspect of our observation is kidney histopathology at both first and second biopsy. According to the International Society of Nephrology (ISN) / Renal Pathology Society (RPS), the first renal biopsy of our patient responded to the definition of class IV (diffuse, lesions involving > 50% of glomeruli), subtype class IV-G (global, lesions affecting > 50% of the glomerular tuft) [7]. Among class IV-G LN, crescentic forms involving more than 50% of glomeruli, where crescents take up over half space in Bowman's capsule, have been poorly analyzed. Yu et al. observed that these crescentic forms represented 20% of 152 class IV-G LN consecutive patients [13]. Interestingly, as compared to noncrescentic one's, all patients with crescentic class IV-G LN forms presented with rapidly progressive glomerulonephritis (RPGN) and had a poorer renal survival, with higher rates of relapse at 4 years from initial diagnosis. Moreover, while endocapillary proliferation was comparable between groups, crescentic forms were associated with significantly lesser immune complex deposition at immunofluorescence study. ANCA positivity (mainly MPO ANCAs) at biopsy was present in 30% of patients with crescentic form, which was significantly higher as compared to the 2.5% observed in patients with noncrescentic class IV-G LN. These observations show that SLE patients with ANCAs are more likely to present with crescentic LN and thus suggest a role for ANCAs in crescent development. As compared to the study of Yu et al., two particular aspects deserve to be emphasized in our case. First, endocapillary proliferation was only very limited in our patient. Second, immunofluorescence study showed an unbalanced deposition of immune complexes. Indeed, strong and diffuse C1q deposition was present, in contrast to other immune complexes that were mildly represented with only segmental deposition. It is tempting to link this atypical LN presentation to ANCA positivity by suggesting a direct pathogenic role of ANCAs or the involvement of specific auto-immune pathways related to the uncommon tolerance breakdown toward 3 different ANCA antigens observed in our patient.
This overlap between AAV and LN suggests common pathophysiological aspects. One of them could be related to Neutrophil Extracelular Traps (NETs) formation, with a different manner between AAV and LN in NETs formation (lytic versus non lytic formation) [14].
The analysis of the second biopsy is also of great interest. Indeed, all crescents became fibrotic, showing the efficiency of immunosuppressive treatment on the extracapillary lesions. On immunofluorescence evaluation, immune complex deposits were comparable to what was observed on the first biopsy. However, and surprisingly, the endocapillary proliferation became diffuse and global, thus much more pronounced as compared to the first biopsy. It allowed to classify it into IV-G subtype, but no longer because of crescentic involvement. Finally, the present case report showed conversion from crescentic class IV-G to endocapillary proliferative class IV-G. Thus, it shows that lesions can vary between crescentic and endocapillary proliferation forms in the same patient and we suggest that this may be related to modulation of involved pathogenic pathways under immunosuppressive treatment. Interestingly, such observation of transitions between LN classes have been previously reported, especially between class IV-G and IV-S [15]. Biological determinants of these changes remain to be elucidated.
The understanding of biological pathways modulating the histological phenotype of LN may translate into more targeted and efficient therapies. In our patient, given the predominant crescentic presentation at first biopsy, we decided to initiate CYC rather than MMF as induction regimen. We speculated that, even if no fibrinoid necrosis was observed, maybe pathways common to those involved in AAV were acting. We observed a good response of the crescentic lesions to steroid/CYC regimen. Next, at the second biopsy showing mainly endocapillary proliferation, given CYC-related toxidermia, we opted rather to MMF while we were considering to start rituximab before biopsy as used in maintenance treatment of AAV. However, we must acknowledge that all these considerations remain speculatives because some reports have also suggested a better improvement of crescentic LN forms under MMF as compared to CYC [16].
This atypical case report highlights the complexity of LN pathophysiology. Transition from crescentic to endocapillary proliferative forms, which has to our knowledge not been reported previously, suggests that different autoimmune pathways may be involved at different extend and at different timings in the same patient. It also highlights the importance of more systematic kidney biopsy to better understand the pathophysiology of LN, especially in patients with atypical histological and immunological presentations as it was the case in our patient. | 2019-08-28T03:05:27.969Z | 2019-08-27T00:00:00.000 | {
"year": 2019,
"sha1": "739c9a7a2b1e9a67258c67316243b75b498a298c",
"oa_license": "CCBY",
"oa_url": "https://bmcnephrol.biomedcentral.com/track/pdf/10.1186/s12882-019-1531-7",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "739c9a7a2b1e9a67258c67316243b75b498a298c",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
104848052 | pes2o/s2orc | v3-fos-license | The impact of improved aerosol priors on near-infrared measurements of carbon dioxide
The Orbiting Carbon Observatory-2 (OCO-2) was launched in 2014 with the goal of measuring the columnaveraged dry-air mole fraction of carbon dioxide (XCO2 ) with sufficient precision and accuracy to infer regional carbon sources and sinks. One of the primary sources of error in near-infrared measurements of XCO2 is the scattering effects of cloud and aerosol layers. In this work, we study the impact of ingesting better informed aerosol priors from the Goddard Earth Observing System Model, Version 5 (GEOS-5) into the OCO-2 ACOS V8 retrieval algorithm with the objective of reducing the error in XCO2 from real measurements. Multiple levels of both aerosol setup complexity and uncertainty on the aerosol priors were tested, ranging from a mostly unconstrained aerosol optical depth (AOD) setup to ingesting full aerosol profiles with high confidence. We find that using co-located GEOS-5 aerosol types and AODs with low uncertainty results in a small improvement in the retrieved XCO2 against the Total Carbon Column Observing Network relative to V8. In contrast, attempting to use modeled vertical information in the aerosol prior to improve the XCO2 retrieval generally gives poor results, as aerosol models struggle with the vertical placement of aerosol layers. To assess regional differences in XCO2 , we compare our results to a global CO2 model validation suite. We find that the GEOS-5 setup performs better than V8 over northern Africa and central Asia, with the standard deviation of the XCO2 error reduced from 2.12 to 1.83 ppm, due to a combination of smaller prior AODs and lower prior uncertainty. In general, the use of better informed aerosol priors shows promise but may be restricted by the current accuracy of aerosol models.
to clouds and aerosols, which describe an ice cloud, water cloud, and three aerosol types. However, retrieved aerosol optical depths (AODs) from ACOS generally compare poorly to the highly accurate AErosol RObotic NETwork (AERONET, Holben et al., 1998;Nelson et al., 2016). This indicates that the way ACOS handles the scattering effects of clouds and aerosols can potentially be improved.
One choice that impacts how well the retrieved cloud and aerosol parameters perform is the use of prior information to 5 constrain the problem. Often, a constant or climatological value with high uncertainty applied to it is used for the aerosol setup. In this work, we test the hypothesis that using more realistic aerosol priors will allow the retrieved aerosol parameters to better represent the scattering of light in the column and thus reduce the error in retrieved X CO2 . Specifically, we examine the impact of using co-located modeled aerosols from the Goddard Earth Observing System Model, Version 5 (GEOS-5) as prior information on the retrieved X CO2 from real OCO-2 measurements. Global atmospheric models, such as GEOS-5, are highly sophisticated and contain many layers of complex physics to represent aerosol processes in the atmosphere including aerosol dynamic schemes and size-resolved aerosol microphysics (Mann et al., 2014). However, atmospheric models do not perfectly represent reality. There are still large differences between individual models, which are restricted by uncertainties in aerosol emission source characteristics, knowledge of atmospheric processes, and the meteorological field data used (Dubovik et al., 2008). Despite this, it is hypothesized that these models will still be of use in the X CO2 retrievals. We also examine the 15 uncertainties applied to the aerosol priors in the current OCO-2 X CO2 retrieval algorithm to see if using a lower uncertainty, in conjunction with a more realistic aerosol prior, results in an improvement in X CO2 against multiple validation sources. Finally, we test whether vertical aerosol information from GEOS-5 can be successfully ingested. These results impact not only OCO-2, but X CO2 from GOSAT, TanSat, and several other future space-based GHG missions that will also be significantly influenced by the scattering effects of clouds and aerosols. 20 Section 2 gives an overview of the OCO-2 ACOS X CO2 retrieval algorithm and the prior information it uses. Section 3 discusses the two validation datasets used in this study along with filtering and bias correction. Section 4 describes our use of instantaneous modeled aerosols as prior information in the retrieval while Sec. 5 presents the results. Section 6 summarizes the study's results and determines if the use of more realistic modeled aerosols is beneficial in near-infrared retrievals of X CO2 .
2 X CO2 Retrieval Algorithm 25 In this section, we will discuss the current OCO-2 retrieval algorithm, ACOS V8, and its cloud and aerosol parameterization.
The remote sensing of CO 2 using reflected near-infrared sunlight is typically accomplished by measuring absorption in an oxygen line and one or more CO 2 lines. For OCO-2, GOSAT, and TanSat, the O 2 A-band at 0.76 µm is used in conjunction with a weak CO 2 absorption band at 1.6 µm and a strong CO 2 absorption band at approximately 2.0 µm. The main purposes of the O 2 band are to estimate the path length of the photons detected by the satellite and to filter out clouds and aerosols (by 30 screening for large variations from the expected path length). As there are only 1.5-2.0 pieces of information about the vertical distribution of CO 2 from OCO-2 (Connor et al., 2008), the column-averaged dry-air mole fraction of carbon dioxide (X CO2 ) is reported: where N CO2 (z) is the molecular number density of carbon dioxide and N d (z) is the molecular dry air number density at altitude z. An optimal estimation technique (Rodgers, 2000) is used to retrieve carbon dioxide and several other parameters 5 that the measured radiances are sensitive to. Generally, optimal estimation takes a state vector containing priors with associated uncertanties, solves a cost function iteratively to find its minimum value, and outputs an optimized state vector with posterior uncertainties for each variable. A full description of the state vector and retrieval technique of the latest OCO-2 algorithm (V8) can be found in O'Dell et al., 2018. The state vector contains approximately 60 elements, including 20 vertical levels of CO 2 , surface pressure, band-dependent surface albedo, total column water vapor, a temperature profile offset, several cloud and 10 aerosol parameters, and others. The aerosol setup contains five particle types: a water cloud, ice cloud, two aerosol types from a Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2; Rienecker et al., 2011) climatology, and a stratospheric aerosol type. This MERRA-2 climatology is simply the monthly means for one year for each of the five MERRA-2 aerosol types (dust, organic carbon, black carbon, sea salt, sulfate). The two types chosen to be included in the state vector are the two with the highest climatological mean AOD for a given month and location. For example, if dust and organic 15 carbon are the two largest AODs of the possible five in the month of July for a given location, then they are selected as the two types to be retrieved for any July OCO-2 sounding for that location. The vertical profile of these aerosol types are described by Gaussians. The height and magnitude of the Gaussian are retrieved for each of the two aerosol types, water cloud, and ice cloud. The magnitude of the stratospheric aerosol type is solved for, but the height is fixed. The widths of the five Gaussian profiles are fixed, as it has been shown that the radiances are not especially sensitive to the width of a cloud or aerosol layer 20 (Butz et al., 2009;Frankenberg et al., 2012). The prior Gaussian profiles are shown in Fig. 1 , from O'Dell et al., 2018. Thus, nine parameters are included in the state vector that directly describe the clouds and aerosols in the scene. Additionally, the natural log of the AOD is the parameter solved for that describes the magnitude of each Gaussian. This is to prevent the algorithm from attempting to retrieve a negative AOD. The retrieved height of the Gaussian represents the fraction of the surface pressure. For example, if the surface pressure is 1000 hPa and the retrieved height parameter is 0.5, the Gaussian 25 profile will be centered at 500 hPa.
Data
In this work, we use two datasets to evaluate the quality of the OCO-2 X CO2 retrievals in the context of testing the aerosol parameterization. While we expect the retrieved aerosol parameters to improve with the use of a more accurate prior, the retrieved aerosols are still only designed to be effective scattering parameters and thus we will not evaluate their quality in this 30 work. The first validation dataset is 32,175 retrievals co-located with 13 Total Carbon Column Observing Network (TCCON; 4 Atmos. Meas. Tech. Discuss., https://doi.org /10.5194/amt-2018-366 Manuscript under review for journal Atmos. Meas. Tech. Discussion started: 21 November 2018 c Author(s) 2018. CC BY 4.0 License. Wunch et al., 2011) and AERONET sites across the globe. The second is a 30,827 retrievals matched with an ensemble of global CO 2 models where we consider the truth the median of the CO 2 models in places where they agree to within 1 ppm.
These two validation sets complement each other in that TCCON is known to be highly accurate, but with limited spatial coverage. The model validation dataset likely has larger uncertainty than TCCON, but provides excellent spatial coverage.
5
The TCCON validation dataset contained 32,175 OCO-2 measurements taken from 17 September 2014 to 2 May 2016. We co-located the OCO-2 measurements in time and space with the AERONET and TCCON, which were required to both be present and operational at a given site. The co-location criteria was within 1 degree latitude/longitude and +/-30 minutes and the sites selected for use are shown in Fig. 2. As TCCON stations are all located on land, only a small fraction of co-located soundings are over water surfaces.
10 Table 1 lists the TCCON sites used in this study. The soundings were selected from a set of OCO-2 "lite" files (Osterman et al., 2017) that had been pre-filtered (see Sec. 3.3). We then post-processed the retrievals with multiple custom filters in an attempt to remove all scenes contaminated by clouds or aerosols.
Model Validation Dataset
Besides validation against the highly accurate but sparsely located TCCON, a set of global CO 2 models was assembled in order to examine spatial errors. We co-located 30,827 OCO-2 soundings in time and space with a suit of nine global carbon models (Peters et al., 2007;Feng et al., 2011;Baker et al., 2010;Liu et al., 2017;Crowell et al., 2018;Rödenbeck, 2005;Inness et al., 2013;Basu et al., 2013). Only points where all the models agreed to within 1 ppm of X CO2 were used. Work by O'Dell et al., 2018 has shown that using this methodology produces similar error statistics to that of the TCCON validation. The median X CO2 of the nine models for each of the 30,827 soundings was used as the truth metric. The OCO-2 soundings were selected 20 by sorting all the measurements into a 4x4 degree spatial grid and filling all grid boxes with up to 10 soundings. This allowed for excellent global coverage while limiting the demand on the available computational resources needed to run the retrievals.
Filtering and Bias Correction
As OCO-2 struggles with scenes containing clouds and aerosols, multiple strategies are used to try and filter out any scene that is contaminated by scattering particles. For both validation datasets, the O 2 A-band Preprocessor (ABP; Taylor et al., Frankenberg et al., 2005) were applied to every measurement before being selected to run through the retrieval. For each validation set, the approximately 30,000 soundings used in this study were those that had successfully passed through the preprocessors. These soundings were determined to be clear enough to be run through the retrieval. After removing soundings that failed to converge, post-processing filtering techniques were applied to remove additional low-quality retrievals that were 30 not screened out by the preprocessors. These filters included the reduced χ 2 , a delta pressure parameter (from ABP), and the Despite heavy pre-and post-filtering of the dataset to remove cloud and aerosol layers, no atmospheric column is truly 5 free from scattering particles. Thus, a bias correction is typically applied to the final X CO2 in an attempt to mitigate retrieval errors (O'Dell et al., 2018). In the operational V8 product, considerable effort is put into developing a multi-parameter bias correction that reduces the X CO2 bias against several independent truth metrics. In this work, a single parameter bias correction was selected for each validation dataset for simplicity and to ensure a fair comparison across different setups. Additionally, it is hypothesized that an improved aerosol setup might reduce the need for a complex bias correction. The parameter chosen was 10 that which had the largest correlation with X CO2 error. When comparing to TCCON, the retrieved X CO2 was bias corrected by removing a linear fit between the X CO2 error (retrieved X CO2 -TCCON X CO2 ) and the difference between the retrieved surface pressure and the prior surface pressure ("dp"). This was the most correlated parameter in the majority of our TCCON tests and thus was selected as the bias correction parameter. This parameter is correlated with X CO2 biases because any unparameterized clouds and aerosols in the column can make the retrieval think there is a lower surface pressure than in reality. 15 Thus, bias correcting this mistake out is designed to bring the retrieved surface pressures back to realistic values and can approximately account for the improperly parameterized clouds and aerosols. In the case of the model validation dataset, the bias correction parameter was the solar zenith angle. Physically, this represents the removal of artificial biases induced by longer air masses. The reason why this parameter was selected over dp is that the model dataset has excellent latitudinal coverage and thus the air mass is weighted more than dp. TCCON, however, is spatially limited and thus the air mass dependence is not as 20 prevalent when searching for optimal bias correction parameters.
Modeled Aerosol Priors
As discussed in section 2, the OCO-2 retrieval algorithm has several aerosol parameters in its state vector. The prior values for most of these parameters in V8 are fixed or taken from a monthly climatology. Here, we discuss several methods in which we test the use of instantaneous, 3D modeled aerosol data as prior information to improve upon the current priors with the hope of 25 increasing the precision and accuracy of the final OCO-2 X CO2 product.
The GEOS-5 Forward Processing for Instrument Teams (GEOS-5 FP-IT; Rienecker et al., 2008) weather forecasting model, created and maintained by the NASA Global Modeling and Assimilation Office, is designed specifically for instrument teams in that the entire period (2000-current) is run using the same GEOS-5 version to maintain consistency and avoid any unwanted biases from updates to the model. For this work, GEOS-5 FP-IT version 5.12.4, hereafter referred to as GEOS-5, was co-located 30 in time and space with the OCO-2 soundings. GEOS-5 is on a 0.625-degree longitude by 0.5-degree latitude horizontal grid with 72 vertical layers extending to 0.01 hPa with a time-step of 3 hours. The model was linearly interpolated in space and the nearest 3-hour model update was chosen in time. For example, if the OCO-2 sounding was taken at 1900 UTC, the 1800 UTC model run was used. The GEOS-5 aerosol scheme contains 15 different types with up to five different size bins for each type, which we aggregate into five unique types: dust, organic carbon, black carbon, sea salt, sulfate. The aggregation process weights by the typical relative amount of optical depth contributed by each type at 760 nm and uses a typical relative humidity value for the hygroscopic types. Each type has a unique optical properties, including the single scatter albedo, extinction coefficient, and refractive index. Figure 3 shows that GEOS-5 AODs correlate better with AERONET compared to both the climatological MERRA-2 AODs and the corresponding retrieved AOD values from OCO-2 V8. Thus, using the model and assigning it some confidence should result in an improved correlation in retrieved OCO-2 AODs compared to AERONET. 10 Our primary hypothesis in this work is that using instantaneous modeled aerosol data as prior information will result in smaller X CO2 errors when compared to the current operational setup that uses a monthly climatology. Figure 4 shows the first of the two aerosol types selected when using the MERRA-2 monthly climatology and when using the interpolated GEOS-5 model field. Certain features, such as Saharan dust and biomass burning, are generally realistically placed in the climatology but the day-to-day variations of the atmosphere are not present and thus the climatology is not representative of the true state 15 of the atmosphere for a given OCO-2 sounding location. For example, dust is selected over large portions of the high northern latitudes in the GEOS-5 model field, but rarely in the MERRA-2 monthly climatology.
Three methods of varying complexity were chosen to ingest the instantaneous model data: -Using the top two aerosol types and their corresponding AODs -Using the top two aerosol types and fitting a Gaussian to the vertical profile of both types 20 -Using the top two aerosol types and solving for a scale factor on an interpolated 20-layer aerosol profile The methodology for selecting which two (of the five) aerosol types to be included in the state vector is simply sorting them by AOD at 760 nm and selecting the two largest values. The ice cloud, water cloud, and stratospheric aerosol type were always retrieved. The ice cloud and water cloud characteristics were kept the same as V8, while the stratospheric aerosol's optical depth prior and corresponding uncertainty were determined by our setups described below.
Types and Optical Depths
The first method simply takes the top two aerosol types based on sorting by each type's AOD and uses their corresponding AODs as prior information for each type. This method is the simplest of our tests and does not rely on any modeled vertical aerosol information.
Types and Gaussian Fits
The second method takes the largest two aerosol types, as before. The 72-layer GEOS-5 aerosol profiles for both types are then interpolated onto the 20-layer OCO-2 vertical grid. A Gaussian is then fit to that 20-layer profile and the magnitude (optical depth, τ ), height, and width of that Gaussian are fed in to the retrieval as prior information. An example is shown in Fig. 5.
Occasionally, the fit is a poor representation of the vertical profile. This is often the case with the sulfate type, which can have 5 both a lower tropospheric peak and a stratospheric peak, resulting in a profile that cannot be represented with a single Gaussian.
To avoid this issue, the sulfate type, if selected, was fit to below 400 hPa and the stratospheric aerosol type (discussed in Sec. 2) was a separate Gaussian fit for the profile above 400 hPa. This method was chosen to test the hypothesis that ingesting vertical information from the model will lead to an improved parameterization of the scattering and, subsequently, a more accurate X CO2 .
Types and Scale Factors
The third and most complex method takes the largest two aerosol types sorted by AOD, as before. The 72-layer GEOS-5 aerosol profiles for both types are then interpolated onto the 20-layer OCO-2 vertical grid. A scale factor applied to the interpolated profile is then solved for by the retrieval. Because the ACOS retrieval solves for the natural log of the AOD, we solve for an additive scalar which acts as a multiplicative scaling factor when converted to AOD. This means that if the retrieval determines 15 that the scale factor should be 1.5, the GEOS-5 AOD profile magnitude will be multiplied by 1.5. This method is similar to that of Parker et al., 2011, except here we are using a vertical aerosol profile co-located in time and space as the prior, while they use a constant aerosol profile. This method was chosen to test the hypothesis that Gaussian fits are insufficient to realistically parameterize the shapes of true vertical aerosol profiles in the X CO2 retrieval.
20
In addition to these three techniques used to modify the aerosol priors, the prior uncertainty must also be considered when using optimal estimation. In the operational ACOS X CO2 retrieval algorithm, the uncertainty on the aerosol parameters is typically very loose. For example, a prior AOD of 0.1 is approximately -2.3 in ln-space. Assigning the V8 uncertainty of 2.0 to ln(τ ) gives 1-sigma values of -4.3 and -0.3 in ln-space, which equates to about 0.014 and 0.74 in optical depth. This large prior uncertainty is due to the lack of confidence in the monthly climatological priors and because it is believed that the radiances themselves 25 should provide enough information to properly constrain the results. However, using this new instantaneous setup allows us to decrease the prior uncertainty because we have more confidence in the GEOS-5 model compared to the monthly climatology, as demonstrated by Fig. 3. The three setups chosen to test are using the operational uncertainty ("high uncertainty"), 25% of the operational uncertainty ("low uncertainty"), and fixing the prior aerosol values ("no uncertainty"). For example, using our first method above (Sec. 4.1) with low uncertainty means solving for the AODs of the two selected aerosol types with 25% of 30 the operational uncertainty assigned to the prior values. This means that instead of a 1-sigma uncertainty of 2.0 on ln(τ ), we assign a value of 0.5. This equates to a 1-sigma uncertainty range of between 0.06 and 0.16 on a prior AOD of 0.1. lists all the test setups and the corresponding uncertainties on the retrieved ln(τ ) and, except for the scalar profile tests, the retrieved height. The third method, solving for a 20-layer profile (Sec. 4.3), does not lend itself to assigning single values of uncertainty equivalent to the other two setups. Thus, we assigned layer uncertainties of ln (10), 0.5, and 0.001 for each layer to represent the operational, low, and fixed setups. We also ran the operational retrieval with low and no uncertainty on the AODs in an attempt to isolate the impact of only modifying the prior uncertainties. When ingesting vertical information in the second 5 and third methods, we also reduced the uncertainty of the retrieved height. The width was effectively never retrieved, as it was always assigned an uncertainty of 0.001.
After processing all 12 retrieval variants, applying similar post-filtering, and matching soundings to ensure a fair comparison, 19,471 soundings remained in the TCCON validation dataset while 17,355 soundings remained in the model validation dataset.
For both datasets and all tests, there were no significant outliers when it came to the percentage of soundings that converged 10 or remained after post-filtering. Typically, the retrievals with loose priors had slightly more soundings fail to converge and the more complex retrieval schemes lost more soundings in post-filtering, but only by a few hundred soundings.
Results
In this section we discuss the impact of using more intelligent aerosol priors in the OCO-2 X CO2 retrieval by comparing our test setups to both TCCON and an ensemble of global X CO2 models. 15 Figure 6 shows the results of our test setups alongside the operational retrieval variants (top row). Here, we can see the impact that different prior information and different prior uncertainties have on the retrieved X CO2 when compared to TCCON.
TCCON Validation Results
For our nine GEOS-5 test setups, when examining the standard deviation of the error (σ), correlation coefficient (R), and mean absolute error (MAE) versus TCCON, the values are typically best for the setups where only the types and the AODs are 20 ingested (second row). When we apply a Gaussian fit to the modeled profiles and use those heights and widths as priors (third row), we see an increase in the scatter against our validation source along with a worse MAE and R. Finally, when we solve for a scalar on the modeled aerosol profile (bottom row), we see the largest scatter in X CO2 against TCCON, worst correlations, and highest MAEs. It therefore appears that trying to incorporate vertical information from the aerosol model leads to a worse X CO2 . 25 The only GEOS-5 test setup with errors and a correlation coefficient better than the operational retrieval is the middle panel of the second row (green triangles), where the types and AODs were ingested with low uncertainty. This may indicate that it is reasonable to assign some confidence in the modeled AODs from GEOS-5, rather than leaving it mostly unconstrained. The operational retrieval with low uncertainty (blue triangles) does relatively well, but slightly worse than the original V8 (blue squares). 30 For all retrieval setups, fixing the aerosol input (right column) results in worse error statistics. This is likely because models are not perfect and by entirely removing the retrieval's ability to fit for aerosol effects on the radiances it results in large errors in X CO2 . Keeping the prior aerosol uncertainty the same as the operational retrieval (left column), however, appears to allow too much freedom and the aerosol parameters revert to the operational values, regardless of what the prior values are. This is because we are assigning almost no confidence to the prior, so all the information comes from the retrieval. This shows the utility of having semi-constrained aerosol priors to guide the X CO2 retrieval algorithm. The only row where the low uncertainty test does better than the high uncertainty test is when we ingest types and AODs. This again indicates that it may be beneficial 5 to use GEOS-5 modeled types and AODs and assign them some level of confidence that is greater than the V8 constraint.
Model Validation Results
While TCCON gives a robust estimate of the statistical errors for our test setups, it does not allow for regional analysis because of the sparsity of the network. The CO 2 model median validation technique, described in Sec. 3.2, allows for an assessment of regional errors in the test setups compared to a truth metric. For this work, the most promising aerosol setup in the TCCON 10 validation study was selected for further analysis. That is, ingesting the GEOS-5 types and AODs with low uncertainty. Figure 7 shows the operational (V8) X CO2 error and the GEOS-5 aerosol prior setup X CO2 error while the top panel of Fig. 8 shows the difference of the absolute value of the X CO2 errors for the two setups.
Here, regional differences can be seen. Over Northern Africa and Central Asia the operational retrieval (top panel of Fig. 7) is often biased high, which results in a large X CO2 scatter in those regions. The GEOS-5 aerosol setup (bottom panel of Fig. 7), 15 however, shows that those high biased bins have mostly been removed. The difference plot (top panel of Fig. 8) demonstrates an improvement in the error of around 1 ppm for many grid cells in Northern Africa and Central Asia. An additional regional difference is in the Southern Ocean, where the GEOS-5 aerosol setup develops a new high bias of 0.4 ppm in many of the bins, compared to V8 which has a bias of nearly zero (+0.1 ppm). Over the remaining land and ocean regions there is minimal change in X CO2 between the two datasets. While of interest, comparing retrieval setups over the Amazon and high latitudes is 20 difficult due to the lack of soundings. This is because these regions are typically cloudy around solar noon or lack the necessary amount of reflected sunlight to make an accurate retrieval, respectively, so they have been filtered out.
Regarding the plots just discussed, three factors could be contributing to the regional changes in X CO2 between V8 and the GEOS-5 aerosol prior setup. First, the decrease in uncertainty to 25% of V8. Second, the two selected aerosol types. Third, the modified AOD priors of the two aerosol types. In order to isolate the first effect, the middle panel of Fig. 8 shows a comparison 25 between V8 and V8 with low uncertainty. Here, we can see that the impact of simply reducing the prior uncertainty on the retrieved AODs is substantial and accounts for a considerable portion of the improvement over Northern Africa and Central Asia, with the standard deviation of the error being reduced from 2.12 ppm to 1.92 ppm. The ocean remains nearly unchanged, along with much of the remaining land surface. Now that we have isolated the impact of reducing uncertainty, we can compare V8 with low uncertainty to GEOS-5 types 30 & AODs with low uncertainty to determine what impact, if any, using the constrained GEOS-5 aerosol types and priors has on the X CO2 error. The bottom panel of Fig. 8 shows that this change is beneficial over Central Asia (σ XCO 2 error reduced from 2.07 ppm to 1.94 ppm for measurements over land between latitudes 16 N and 56 N and longitudes 52 E and 152 E), has a minimal effect over the Northern Africa (σ XCO 2 error changed from 1.63 ppm to 1.62 ppm for measurements over land between latitudes 6 S and 38 N and longitudes 20 W and 52 E), and a detrimental effect over the Southern Ocean (σ XCO 2 error increased from 1.86 ppm to 1.95 ppm and a positive bias of 0.4 ppm for measurements south of 45 S). This suggests that the improvement in the scatter of the X CO2 error against model validation over the Northern Africa was primarily due to the reduction in uncertainty in the prior AODs, and not the co-located GEOS-5 aerosol prior types and values themselves. Over Central Asia, however, the improvement seems to be a combination of reducing the uncertainty and using instantaneous types 5 and/or AODs. Over the Southern Ocean, using GEOS-5 types and prior AODs results in an increase in scatter and a high bias of 0.4 ppm.
Next, we attempt to distinguish between the impact of using GEOS-5 aerosol types and using their corresponding AODs as prior information. The areas of interest are Central Asia, where using instantaneous priors improved the X CO2 retrieval, and the Southern Ocean, where it worsened the X CO2 retrieval. The top panel of Fig. 9 shows the binned prior AOD for V8, 10 which is derived from a monthly MERRA-2 climatology (as discussed in Sec. 2), while the bottom panel shows the binned prior AOD for our test setup which uses the co-located GEOS-5 modeled data. In general, slightly more variation is seen in the GEOS-5 priors, which is to be expected, as the monthly climatology is a mean of an entire month and thus removes most synoptic variability. The most prominent change, however, is the significant reduction in prior AOD over Northern Africa and Central Asia. The GEOS-5 aerosol setup has a 30% lower mean prior AOD than V8. This is likely because the MERRA-2 15 climatology was created by simply averaging an entire month of data together, including all instances where the optical depth was large due to dust storms, pollution events, biomass burning, etc. Those scenes, however, are typically removed by OCO-2's pre-screeners (ABP and IDP, described in Sec. 3.3) before being processed through the retrieval code and thus the GEOS-5 prior map only includes scenes that were determined to be sufficiently clear to perform retrievals on. This suggests that the MERRA-2 monthly climatology aerosol prior may be artificially high and not entirely appropriate for use in the operational 20 retrieval algorithm. This is also seen in Fig. 3, where there are several MERRA-2 climatology AODs that are significantly high biased against AERONET. Fig. 3, however, is only data from 13 AERONET sites and is thus not representative of the global differences.
In optimal estimation the final, optimized state vector represents a weighted combination of the prior information and the measurement, not just the state of the prior. The top panel of Fig. 10 shows the retrieved AOD for V8 while the bottom panel 25 shows the retrieved AOD for the GEOS-5 aerosol setup. Here, we see an even more dramatic difference in retrieved AODs over the Northern Africa and Central Asia. Over land in these two regions, the retrieved AOD is 52% lower for the GEOS-5 aerosol setup. This is partially due to the reduced uncertainty on the prior AODs for the GEOS-5 aerosol setup, which prevents it from deviating substantially from the prior.
While the differences in X CO2 over Central Asia correlate with a large difference in the magnitude of the retrieved AOD, 30 the same cannot be said of the X CO2 differences over the Southern Ocean, as those differences in retrieved AOD are small Figure 11 demonstrates that changing the types has almost no effect on the Southern Ocean and overall has a small and sporadic effect (note the reduced scale). The mean difference between ingesting GEOS-5 AODs and types and only ingesting GEOS-5 AODs for soundings where the type actually changed is about 0.01 ppm with a standard deviation 5 of about 0.2 ppm. Only a small fraction of soundings had differences larger than 0.5 ppm (2.3% of all soundings). Thus, the selection of an appropriate AOD prior and uncertainty is much more important than the exact aerosol properties in our retrieval parameterization.
After further investigation, the modification to the stratospheric aerosol prior and its corresponding uncertainty is likely the cause of the Southern Ocean 0.4 ppm high bias in the GEOS-5 aerosol setup. As detailed in O'Dell et al., 2018, ACOS V7 10 contained a high bias over water at high southern latitudes due to both the presence of stratospheric aerosol from the Calbuco eruption in 2015 as well as ice build up on the OCO-2 focal plane arrays. Here, we have unintentionally replicated this issue, as the stratospheric aerosol priors from GEOS-5 are near zero (as opposed to a global prior value of 0.006 used in V8). Figure 12 demonstrates that when V8 solves for a large stratospheric AOD in the Southern Ocean, GEOS-5 cannot because it starts with a lower prior and is unable to sufficiently increase it. This result in a significant positive bias in X CO2 (bottom of Fig. 7). When 15 the retrieved V8 stratospheric AOD is greater than 0.012 in the Southern Ocean, the GEOS-5 aerosol setup incurs a positive bias of approximately 1.4 ppm (compared to the positive 0.4 ppm bias for all Southern Ocean retrievals). This indicates that in soundings where large stratospheric AODs are needed to fit the radiances, the lack of AOD usually appears to lead to a high bias in X CO2 .
To further test this hypothesis, we ran a set of soundings where we reverted the stratospheric aerosol prior AOD and prior 20 uncertainty back to V8 values (0.006 and a 1-sigma uncertainty on the ln(AOD) of 1.8). This revealed that the 0.4 ppm bias in the Southern Ocean mostly disappears, as the high bias is reduced back to approximately +0.1 ppm, in agreement with V8.
Discussion and Conclusions
In this study we investigated the impact of using more intelligent aerosol priors in the OCO-2 ACOS X CO2 retrieval algorithm applied to real measurements. We ingested aerosol information from co-located GEOS-5 model data with varying levels of 25 uncertainty and compared to two validation sources, TCCON and a global CO 2 model suite.
We found that ingesting instantaneous AOD information with low uncertainty slightly reduced the standard deviation of the X CO2 error against TCCON from 1.17 ppm to 1.13 ppm. More interestingly, we found that attempting to ingest vertical information from GEOS-5 produced poor results against TCCON, with mean absolute errors more than 50% larger than V8. We hypothesize that this is because accurate vertical placement of aerosol layers still represents a significant challenge in global 30 aerosol models (Koffi et al., 2016). This is due to the large uncertainty in processes related to how aerosols are distributed vertically, partially because of the lack of constraint by global observations (Kipling et al., 2016). Additionally, fixing the Atmos. Meas. Tech. Discuss., https://doi.org/10.5194/amt-2018-366 Manuscript under review for journal Atmos. Meas. Tech. Discussion started: 21 November 2018 c Author(s) 2018. CC BY 4.0 License. ingested aerosol information also makes the retrieval compare poorly to TCCON. This is, again, because the priors will always be imperfect, so the OCO-2 ACOS X CO2 retrieval algorithm needs some amount of freedom to adjust the radiances.
A comparison to our model validation dataset reveals an improvement over Northern Africa and Central Asia when using the GEOS-5 types and AODs with low uncertainty compared to V8, with the scatter in X CO2 error being reduced from 2.12 ppm to 1.83 ppm. Additionally, we see a new high bias in X CO2 of 0.4 ppm over the Southern Ocean. These regional differences 5 in OCO-2 V8 compared to the GEOS-5 type and AOD low uncertainty prior test setup are likely caused by one of the three modifications. First, reducing the uncertainty on the prior AODs in V8 results in an improvement over Northern Africa and Central Asia. Figure 13 shows that by constraining the prior AODs in V8, we prevent the algorithm from retrieving large AODs. Those soundings that retrieved large AODs in V8 but smaller AODs in the low uncertainty setup generally have smaller X CO2 errors (blue pixels in Fig. 13). For soundings where the retrieved V8 AOD is greater than 0.4 over Northern Africa and 10 Central Asia, constraining the prior AODs with low uncertainty reduces the X CO2 bias against the validation dataset by 0.55 ppm on average. Some soundings are outliers, in that they do worse when constrained (red pixels in Fig. 13). Around 8% of the retrievals have an X CO2 error increase of greater than 0.5 ppm. This may be because those scenes actually contain large AODs and thus constraining the priors is hindering the retrieval's ability to properly fit the radiances, but further investigation is needed on this topic. 15 Second, some of the high bias seen in V8 (top panel of Fig. 7) is likely due to the MERRA-2 climatological priors being unrealistically high for the selected OCO-2 soundings. Physically, when there is too much aerosol being retrieved, the CO 2 absorption lines are filled in too much because the retrieval thinks those photons are experiencing a shorter path length than they are in reality. The retrieval then must add more CO 2 to deepen the absorption lines again to match the measured line depths.
This results in the observed high bias in V8 over Central Asia, which is mostly removed when smaller, more appropriate AOD 20 priors are taken from the GEOS-5 dataset (bottom panel of Fig. 8). However, the bottom panel of Fig. 8 also showed that using lower AOD priors alone had no significant net impact on Northern Africa. This difference may be due to Northern Africa having a higher surface albedo than Central Asia. This results in a larger fraction of the signal coming from the surface and less weight placed on any aerosols present. So long as the retrieved aerosols are not very large, as discussed above, the X CO2 results over Northern Africa may be less sensitive to the aerosol prior than other regions. 25 Third, the GEOS-5 types & AODs with low uncertainty setup resulted in a new high bias of 0.4 ppm in X CO2 relative to V8 over the Southern Ocean. It was initially hypothesized that the change in aerosol types may have caused the high bias, but we showed that changing one or both of the aerosol types has a minimal impact on the X CO2 in this setup. After additional tests, we found that the GEOS-5 stratospheric AODs in the Southern Ocean were too small to be able to account for actual stratospheric aerosol in the region along with the effects of an ice accumulation artifact in the OCO-2 radiances. Reverting the 30 stratospheric prior and uncertainty back to that of V8 mostly eliminated the high bias in the Southern Ocean. This conclusion highlights the severe impact that even a small amount of aerosols can have on the retrieved X CO2 .
This work is relevant for several missions set to launch in the near future, including OCO-3, GOSAT-2, MicroCarb, Tansat-2, and GeoCarb. Algorithm development teams will have to decide how they will account for clouds and aerosols in order to minimize errors in their CO 2 measurements. While we have shown the benefits and limitations of using instantaneous modeled aerosol data to inform the OCO-2 X CO2 retrieval, more work is needed to determine the optimal aerosol parameterization for near-infrared measurements of CO 2 . These results may be of use in guiding development for the next version of the OCO-2 ACOS X CO2 retrieval algorithm.
Competing interests. The authors declare that they have no conflict of interest. | 2019-04-10T13:12:29.285Z | 2018-11-21T00:00:00.000 | {
"year": 2019,
"sha1": "b0decf53b95a79f82163886f5ec613c0726e2b17",
"oa_license": "CCBY",
"oa_url": "https://amt.copernicus.org/articles/12/1495/2019/amt-12-1495-2019.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "f6245b8f194607ab88dae44a66aec5b17057fcf5",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": [
"Environmental Science"
]
} |
8531660 | pes2o/s2orc | v3-fos-license | Innate IFN-γ-Producing Cells in the Spleen of Mice Early after Listeria monocytogenes Infection: Importance of Microenvironment of the Cells Involved in the Production of Innate IFN-γ
Production of innate interferon-γ (IFN-γ) is a crucial step in immunological defense against bacteria. However, there is little information regarding cellular mechanisms underlying IFN-γ production in vivo early after bacterial infection. Here we analyze innate IFN-γ production in the spleen of mice early after Listeria monocytogenes (LM) infection ex vivo by flow-cytometry and in situ by immunohistochemistry, and compare them with the IFN-γ-producing cells reported previously in our in vitro coculture system in which cell-cell interaction between lymphocytes and dying bacterial-infected macrophages is required for the production of IFN-γ. In the spleen at 20 h after LM infection, natural killer (NK) cells, a subset of αβ T cells, and subsets of NKT and γδ T cells produced IFN-γ with features similar to the IFN-γ-producing cells in our in vitro coculture system. Immunohistochemistry revealed that LM bacteria were first phagocytosed mainly by ER-TR9+ marginal zone macrophages (MZMs), then forming infectious foci in close vicinity of the marginal zone (MZ) at 20-h postinfection. At this time point, the IFN-γ-producing cells were accumulating at the same site of infectious foci, around which ER-TR9+ MZMs were clustered but most of bacteria were no longer associated with ER-TR9+ MZMs. These results indicate that innate IFN-γ production by innate lymphocytes takes place at infectious foci formed in close vicinity of the MZ, and they also suggest an important role for the microenvironment of the cells accumulated at infectious foci in inducing the production of innate IFN-γ.
INTRODUCTION
Listeria monocytogenes (LM) is a Gram-positive facultative intracellular bacterium that causes severe disseminated infection or local infection like meningitis in immunocompromised individuals and in pregnant women (Vazquez-Boland et al., 2001). Studies of LM infection in mice as a model for antimicrobial defense have brought us marked progress in our understanding of the innate and adaptive immunity (Parham, 1997). During an early stage of LM infection innate immune responses control initial infection, and subsequently T helper type 1 (Th1) adaptive immune responses develop to sterilize LM-infected mice (Unanue, 1997). nterferon-γ (IFN-γ) secreted by the innate immune cells (called "innate IFN-γ") is one of the most important cytokines that contribute to these innate and adaptive immune responses; i.e., IFN-γ plays a crucial role for the activation of macrophage effector functions that are required to limit bacterial growth and control infection, and in the adaptive immunity IFN-γ is responsible for a bias toward Th1 responses (Schroder et al., 2004). Natural killer (NK) cells are thought to be the main source of innate IFN-γ, but other multiple cell types, such as memory type CD8 + T cells, NKT cells, γδT cells, macrophages, and dendritic cells (DCs) are also claimed to produce innate IFN-γ (Hiromatsu et al., 1992;Bancroft, 1993;Frucht et al., 2001;Berg et al., 2005;Berntman et al., 2005;Thale and Kiderlen, 2005). Furthermore, NK1.1 + CD11c + cells have recently been reported to be the primary IFN-γ-producing cells in the spleen early after LM infection in mice (Chang et al., 2007;Plitas et al., 2007). Thus, the cellular sources of innate IFN-γ remain still unsettled.
The spleen has multiple functions such as the phagocytosis of aged erythrocytes, the capture and destruction of pathogens, and the induction of the innate and adaptive immunity. Accordingly, the spleen has a highly organized structure, consisting of the red pulp, distinguishable by the abundance of erythrocytes, and the white pulp where the majority of T cells and B cells reside (Mebius and Kraal, 2005). The white pulp and the red pulp are separated by the marginal zone (MZ), which consists of sinus-lining reticular cells, MZ B cells, dendritic cells, marginal metallophilic macrophages, and marginal zone macrophages (MZMs). Most of the blood flow passes through the MZ and thus macrophages of the MZ are essential for trapping blood-bone pathogens and early control of infection (Aichele et al., 2003). It is widely accepted that upon LM infection of mice, LM bacteria are first trapped by macrophages in the MZ, forming MZ infectious foci and then LM-infected phagocytes translocate from the MZ into the white pulp where they establish secondary infectious foci (Conlan, 1996). The field of infectious foci is considered to be the site for control of bacterial infection through orchestrating innate immune responses (Serbina et al., 2003;Berg et al., 2005;Kang et al., 2008). Also, it may be the site in which antigen presentation to T cells occurs to induce adaptive immune responses especially when they are formed in the white pulp (Muraille et al., 2005). Since IFN-γ is a key cytokine to control bacterial infection, dissecting cellular mechanisms responsible for innate IFN-γ production in the spleen is indispensable for understanding the host defense against bacterial infection. However, there is little experimental data as to where in the spleen and how innate IFN-γ is produced early after LM infection.
In this study, we characterize IFN-γ-producing cells in the spleen of mice early after LM infection by flow-cytometric analysis to determine the cellular source of innate IFN-γ and also we examine the localization of the IFN-γ-producing cells and bacteria within the LM-infected spleens by immunohistochemistry to determine their dynamic relationships. Furthermore, as we have recently shown by using an in vitro coculture system that subsets of NK, NKT, and γδ T cells, which are generically referred to as innate lymphocytes (Bendelic et al., 2001), produce IFN-γ in a cell-cell contact dependent manner in response to dying bacterialinfected macrophages (Kubota, 2006) and have proposed that this IFN-γ production pathway contributes to the host defense at the initial phase of bacterial infection (Kubota, 2010), we compare the innate IFN-γ-producing cells in the LM-infected spleen with the IFN-γ-producing cells in our in vitro coculture system to explore the relevance of our in vitro finding to the in vivo situation in the spleen early after LM infection.
We revealed in this study that early after LM infection, NK cells and, to a lesser extent, a subset of αβ T cells, subsets of NKT and γδ T cells are the cellular sources of innate IFN-γ, and that LM bacteria are first phagocytosed by MZMs, then forming infectious foci in close vicinity of the MZ without being associated with MZMs at 20-h postinfection, and that the innate IFN-γ-producing cells form cell clusters at the same site of infectious foci. This study also demonstrated that there are many similarities between the features of the IFN-γ-producing cells in the LM-infected spleen and those in our in vitro coculture system. These results suggest that the cellular microenvironment similar to our in vitro setting may play an important role in the production of innate IFN-γ by innate lymphocytes in the LM-infected spleen.
BACTERIA
Listeria monocytogenes (EGD strain) were grown in tryptic soy broth overnight, washed with PBS and stored at −80˚C until use.
MICE AND CELLS
C57BL/6 mice were purchased from Oriental Yeast (Tokyo, Japan). Mice were used in accordance with the institutional guideline. Bone marrow-derived macrophage (BMM) cells were grown from marrow cells according to the method described previously (Kubota, 2006). Nylon wool non-adherent (NWNA) spleen cells were prepared according to the method described elsewhere (Hothcock, 1998).
IN VITRO STIMULATION OF NYLON WOOL NON-ADHERENT SPLEEN CELLS BY LM-INFECTED BMM CELLS
This method was described previously (Kubota, 2006(Kubota, , 2010. Briefly, C57BL/6 mice-derived BMM cells (4 × 10 5 cells/ml) were suspended in culture medium without antibiotics, seeded into a 96-well culture plate and were cultured at 37˚C overnight in a CO 2 incubator. After each well was washed with DMEM without antibiotics, bacteria were inoculated into each well at a multiplicity of infection (MOI) of 50, and the plate was centrifuged at 800 g for 5 min and then incubated at 37˚C in a CO 2 incubator. After a 1-h infection period, NWNA spleen cells (5 × 10 6 cells/ml) suspended in culture medium containing gentamicin (100 μg/ml) were added to each well. The plate was incubated at 37˚C for 20 h in a CO 2 incubator. During the last 6 h of culture, GolgiStop (BD Pharmingen) was added, and intracellular IFN-γ staining was conducted and analyzed on FACScan (BD Bioscience).
IN VIVO INFECTION
C57BL/6 mice were infected intravenously (i.v.) with LM with or without fluorescent Fluoresbrite Carboxy BB microspheres (1 × 10 8 ). Mice were inoculated i.v. with 250 μg of BFA 4 h before sacrifice (Lin and Whitton, 2005). At 4-and 20-h postinfection mice were killed and the spleens were excised from the mice for FACS analysis and immunohistochemistry. For FACS analysis, splenocytes were gently dispersed in Ten-broeck Tissue grinder Frontiers in Immunology | Microbial Immunology (Wheaton Scientific, Millville, NJ, USA) and erythrocytes were lysed in NH 4 Cl buffer (155 mM NH 4 Cl, 10 mM KHCO 3 , 0.1 mM EDTA).
INTRACELLULAR IFN-γ ASSAY
This method was described previously (Kubota, 2006). Briefly, splenocytes and NWNA spleen cells pretreated with the blocking anti-FcγR mAb (2.4G2) were stained with mAbs against surface markers, and then the cells were permeabilized using the Cytofix/Cytoperm Plus kit (BD Bioscience) and stained using FITC-conjugated anti-IFN-γ. The data were acquired using FACScan and analyzed using CellQuest software (BD Bioscience).
IMMUNOHISTOCHEMISTRY
Spleen tissues were fixed with 1% formaldehyde in 0.1 M phosphate buffer (pH 7.2) for 1 h, washed with phosphate buffered saline (PBS), and incubated overnight at 4˚C in Holt's gum sucrose solution. They were then embedded in OCT compound (Sakura Tissue-Tek, Tokyo, Japan), frozen in liquid nitrogen, and then 7to 10-μm-sections were cut and air-dried. For staining of MZMs and perlecan, tissues were immediately embedded in OCT compound and fresh-frozen. Fresh-frozen sections (7-10 μm) were fixed with acetone (−20˚C) for 5 min. After blocking with 1% bovine serum albumin (BSA) dissolved in PBS (BSA-PBS) for 1 h, sections were incubated overnight with diluted primary antibodies, washed with PBS (5 min × 3 times). Bound primary antibodies were visualized by incubation for 1 h with fluorochrome labeled species-specific secondary antibodies. Bound ER-TR9 mAb was detected with Alexa Fluor 488-conjugated streptavidin. For double immunofluorescence staining, antibodies were mixed and applied onto sections. After washing with PBS (3 min × 3 times), specimens were mounted with Fluoromount (Diagnostics Biosystems, Pleasanton, CA, USA) and examined under an Axioplan-2 epifluorescence microscope (Zeiss, Tokyo, Japan). Images were captured with a cooled CCD camera (Quantix, Photometrics, Munchen, Germany) and processed using IPlab software (Scanalytics, Fairfax, VA, USA). Deconvolution of some images was performed using Hazebuster software (Vay Tek, Fairfield, IA, USA). All antibodies and Alexa Fluor 488-conjugated streptavidin were diluted 1:100 and 1:500 in 1% BSA-PBS, respectively. Negative control experiments were performed by replacing the primary antibodies with 1% BSA-PBS. Some spleen tissues exhibited background autofluorescence.
A MINOR SUBSET OF αβ T CELLS AND NK, NKT, AND γδ T CELLS ARE THE CELLULAR SOURCES OF INNATE IFN-γ IN THE SPLEEN EARLY AFTER LM INFECTION
To examine innate IFN-γ production in the spleen of mice following infection with LM, we inoculated C57BL/6 mice with LM and analyzed splenocytes ex vivo by three-color flow-cytometry. Detection of cytokine production in vivo has usually been conducted by culturing splenocytes in vitro with brefeldin A (BFA), an inhibitor of Golgi transporter, for a short period of time, followed by intracellular cytokine staining (Emoto et al., 2001;Thale and Kiderlen, 2005;Chang et al., 2007). In the present study, to avoid in vitro manipulation, we employed an approach of in vivo injection of BFA (Lin and Whitton, 2005); i.e., BFA was administered to LM-infected or uninfected C57BL/6 mice 4 h before sacrifice. The spleens were excised from the mice and splenocytes were immediately assessed for IFN-γ production by intracellular cytokine staining in combination with surface staining of NK1.1 and either TCRαβ or TCRγδ. Preliminary experiments showed that the highest number of IFN-γ-producing cells was induced in the spleen of mice injected with LM at a range of doses from 2 to 5 × 10 5 bacteria. When mice were infected with 5 × 10 5 LM, the IFN-γ response in the spleen began around 14 h after LM infection and the maximum response was induced at 20-h postinfection (Figure 1). Thus, in all the following experiments mice were injected with 5 × 10 5 bacteria and the IFN-γ production was assessed at 20-h postinfection. Figure 2 shows that while no IFN-γ producing cells were detected in the splenocytes from control uninfected mice (Figure 2A), 1.1 ± 0.2% of total splenocytes from LM-infected mice produced IFN-γ, of which IFN-γ + percentages of αβ T cells and γδ T cells were 1.0 ± 0.2 and 8.4 ± 1.9%, respectively ( Figure 2B). When dot plots were gated on IFN-γ-producing αβ T cells, 28 ± 1.0% were NK1.1 + , indicating that IFN-γ-producing αβ T cells include NKT cells ( Figure 2B). Of TCRαβ + NK1.1 + (NKT cells), 32 ± 7.5% were IFN-γ + and of TCRαβ − NK1.1 + cells (hereafter denoted "NK" cells because these cells are mostly composed of NK cells), 32 ± 3.3% were IFN-γ + (Figure 2B right panel).
These results indicate that a minor subset of αβ T cells and subsets of NK, NKT, and γδ T cells are the cellular sources of innate IFN-γ production in the spleen at 20 h after LM infection.
We next investigated the localization of IFN-γ-producing cells in the spleen of mice 20 h after LM infection. To this end, part of the spleens used for FACS analysis was fixed with formaldehyde and the sections were stained for IFN-γ ( Figure 2D). Control staining of the spleen was also conducted ( Figure 2C). The results clearly show that the IFN-γ-producing cells (green) identified as being composed of subsets of innate lymphocytes by FACS analysis ( Figure 2B) accumulate at several discrete sites, thus forming multiple cell clusters in the spleen ( Figure 2D). It was of interest that IFN-γ expressed in the cytoplasm of the cells appeared to concentrate beneath the cell membrane (Figure 2D inset). Additionally, a large number of CD11b + cells were stained within the clusters of IFN-γ-producing cells (Data not shown), suggesting that macrophages and granulocytes were attracted to the clusters. We next stained the same section with hematoxylin and eosin to distinguish the white pulp and the red pulp ( Figure 2E). Comparison between Figures 2D,E unveiled that the clusters of IFN-γ-producing cells appear to be localized to the boundary between the white pulp and the red pulp.
We previously reported that dying LM-infected BMM cells induce subsets of innate lymphocytes to produce IFN-γ in a cellcell contact dependent manner in vitro (Kubota, 2006(Kubota, , 2010. Thus, we next asked whether IFN-γ-producing NK1.1 + CD11c + cells are also main IFN-γ-producing cells in our in vitro coculture system. To this end, NWNA spleen cells before culture and NWNA spleen cells cocultured for 20 h with uninfected or LMinfected C57BL/6 BMM cells were subjected to the same analysis as above (Figures 3C-E). Compared to the whole splenocytes ( Figure 3A), CD11c hi cells were hardly detected in NWNA spleen cells (Figure 3C), indicating that DCs are removed from splenocytes after passing a nylon column. As with the whole splenocytes ( Figure 3A), ∼50% of NK1.1 + NWNA spleen cells expressed CD11c, and almost all of the NK1.1 + CD11c + cells were CD3 − and CD244 + (Figure 3C). In contrast to uninfected BMM cells (Figure 3D), LM-infected BMM cells, which underwent cell death during culture as had been reported previously (Kubota, 2006(Kubota, , 2010 induced 44 ± 5% of NK1.1 + CD11c + cells and 42 ± 10% of NK1.1 + CD11c − cells to produce IFN-γ ( Figure 3E). As shown in the right panel of Figure 2E, ∼42% of total IFN-γ-producing cells were NK1.1 + CD11c + cells and thus NK1.1 + CD11c + cells were main IFN-γ-producing cells in our in vitro experimental setting as well as in the spleen in vivo at 20 h after LM infection (Figure 3B right panel). B220 + NK1.1 + CD11c + cell population has been reported as a novel DC subset that has the ability to produce IFN-γ and exhibits properties of both NK cells and DCs (Chan et al., 2006;Taieb et al., 2006). We next examined whether B220 antigen is expressed on the IFN-γ-producing cells responding to LM-infected BMM cells. Figure 3C shows that 4.6 ± 1.4% of NK1.1 + CD11c + NWNA spleen cells expressed B220 antigen. When NWNA spleen cells were cocultured for 20 h with LM-infected BMM cells, this proportion decreased to 2.8 ± 0.1% despite ∼44% of the NK1.1 + CD11c + cell population produced IFN-γ ( Figure 3E). The similar results were obtained with NK1.1 + CD11c + IFN-γ-producing cells in
cells are main IFN-γ-producing cells in both the spleen of mice infected with LM for 20 h and the in vitro 20-h coculture of NWNA spleen cells with LM-infected BMM cells. (A)
Splenocytes from uninfected mice were stained for NK1.1, CD11c, and either CD3, CD244, or B220, and three-color flow-cytometric analysis was performed on FACScan. (B) Splenocytes from LM-infected mice were stained for NK1.1, CD11c, and either intracellular IFN-γ or B220. (C-E) NWNA spleen cells before culture (C) and after coculture with uninfected (D) or LM-infected (E) BMM cells were stained for NK1.1, CD11c, and either intracellular IFN-γ or one of the following surface markers: CD3, CD244, and B220. Representative data from one of three independent experiments are shown. The numbers shown in or outside quadrants represent mean percentages of cells in total splenocytes and the numbers within histograms represent mean percentages of cells expressing surface antigen indicated or intracellular IFN-γ in gated cell populations from three independent experiments.
www.frontiersin.org
LM-infected mice (Figures 3A,B). Thus, B220 + NK1.1 + CD11c + cells did exist, but they were not the specific producers of IFN-γ in both in vitro and in vivo.
SIMILARITY BETWEEN IFN-γ-PRODUCING CELLS IN THE LM-INFECTED SPLEEN AND THOSE IN THE COCULTURE OF NWNA SPLEEN CELLS WITH DYING LM-INFECTED MACROPHAGES
As it was found that NK1.1 + CD11c + cells are primary IFN-γproducers in both the LM-infected spleen and in our in vitro coculture system, further comparison between IFN-γ-producing cells in the LM-infected spleens and those in our in vitro coculture system (Kubota, 2006(Kubota, , 2010 was performed and the results are summarized in Figure 4. Since IL-12 and IL-18 can synergistically stimulate lymphocytes to produce IFN-γ in vitro (Dinarello and Fantuzzi, 2003), analysis of IFN-γ production by NWNA spleen cells cultured for 20 h with IL-12 and IL-18 has also been included in this comparison. As shown in Figure 4A, our in vitro coculture and the in vivo infection induced similar proportions of each individual cell type to produce IFN-γ (0.8 ± 0.2 and 1.0 ± 0.2% of αβ T cells, 46 ± 6 and 32 ± 3.3% of "NK" cells, 38 ± 11 and 32 ± 7.5% of NKT cells, and 12 ± 3.4 and 8.4 ± 1.9% of γδ T cells, respectively), whereas, compared to the above two stimulations, the IL-12/IL-18 stimulation resulted in the production of IFN-γ by relatively large proportions of each individual cell type (4.5 ± 1.6% for T cells, 95 ± 3.5% for "NK" cells, 87 ± 3.5% for NKT cells and 37 ± 7.0% for γδ T cells). Thus, the results indicate that the IFN-γ-positive proportions of each individual cell type in the spleen at 20-h postinfection resemble to those observed in our in vitro cocultures of NWNA spleen cells with dying LM-infected BMMs.
Frontiers in Immunology | Microbial Immunology
Because IFN-γ-producing γδ T cells in the spleen of mice infected with LM were found to be biased to NK1.1 + γδ T cells (Figure 2B), we next compared the proportions of NK1.1 + cells in IFN-γ-producing γδ T cells among the three different stimulations by three-color flow-cytometry ( Figure 4C). In our in vitro coculture and in the in vivo infection, 76 ± 2.1 and 69 ± 7.3% of IFN-γproducing γδ T cells were NK1.1 + , respectively, while 25 ± 4.3 and 22 ± 3.5% of normal counterparts were NK1.1 + (Figure 4Cii,iii) In contrast, in the IL-12/18 stimulation only 24 ± 7.0% of IFN-γproducing γδT cells were NK1.1 + (Figure 4Ci) although when the whole NK1.1 + γδT cells were gated, most of them were IFN-γ + (Data not shown). The results indicate that in our in vitro coculture system as well as in the LM-infected spleens, IFN-γ-producing γδ T cells are biased to the cells with a NK1.1 + phenotype.
Overall, the comparison among the three different stimulations reveals that there are many similarities between the features of IFN-γ-producing cells in the spleen at 20-h postinfection and those in our in vitro cocultures of NWNA spleen cells with dying LM-infected BMMs.
CLUSTERS OF THE IFN-γ-PRODUCING LYMPHOCYTES ARE LOCALIZED TO THE SAME SITE OF INFECTIOUS FOCI FORMED IN CLOSE VICINITY OF THE MZ
We next determined the precise localization of IFN-γ-producing cells and bacteria within the LM-infected spleen. To this end, we excised the spleens from mice treated in the same way as the case of FACS analysis except that 3 × 10 8 fluorescent microspheres were simultaneously injected into mice, and then the spleen sections were examined by immunohistochemistry using antibodies specific to LM, MZM, IFN-γ, and perlecan. Perlecan is a major heparan sulfate proteoglycan constituting various basement membranes (Lokmic et al., 2008). Microspheres have been reported to be taken up by macrophages in the MZ and by CD11c + DCs in the red pulp when injected into mice (Iyoda et al., 2002). In preliminary experiments we confirmed that the injection of microspheres did not affect the total number of IFNγ-producing cells in the LM-infected spleens (Data not shown). When cryosections were stained for LM (red) and ER-TR9 + MZMs (green; Figure 5), ER-TR9 + MZMs and microspheres (blue) were observed in the same area of the sections at 4-h postinfection, delineating the MZ, and most of LM bacteria were associated with ER-TR9 + MZMs (Figure 5A). At 20-h postinfection, LM formed infectious foci and ER-TR9 + MZMs were found around the infectious foci, where most of bacteria were no longer associated with ER-TR9 + MZMs (Figure 5B).
To determine the location of infectious foci, we next stained spleen tissues excised from LM-infected mice at 16-h postinfection ( Figure 6A) and 20-h postinfection ( Figure 6B) by using anti-LM antibody (red) and anti-perlecan antibody (green). Staining of perlecan, which detects vascular basement membranes surrounding marginal sinuses and central arterioles, delineated the MZ and the central arteriole. Most of the bacteria were found in close vicinity of the MZ, forming infectious foci in both the 16 and 20-h sections. It should be noted, however, that in the 20-h sections, some bacteria were also observed in the white pulp, occasionally in periarteriolar lymphoid sheath surrounding the central arteriole.
We next examined the positional relationship between infectious foci and clusters of IFN-γ-producing cells. To this end, because both of the Abs against LM and IFN-γ were rabbit immunoglobulin and therefore unable to be used for double immunofluorescence staining, consecutive sections were separately stained with anti-LM (Figures 7A,C) and anti-IFN-γ (Figures 7B,D) antibodies, respectively, and a series of comparisons between those consecutive sections were conducted. As shown in Figure 7, infectious foci (Figures 7A,C; red) and clusters of IFN-γ-producing cells (Figures 7B,C; red) were always observed in the same area of consecutive sections in close vicinity of the MZ, which were delineated by microsphers (blue). This result indicates that clusters of IFN-γ-producing cells are localized to the same site of infectious foci.
Taken together, these results indicate that bacteria are mainly trapped by ER-TR9 + MZMs at 4-h postinfection and that at 20-h postinfection infectious foci and clusters of IFN-γ-producing cells are formed at the same sites in close vicinity of the MZ, where most of bacteria are no longer associated with ER-TR9 + MZMs.
DISCUSSION
The present study demonstrated that NK cells and, to a lesser extent, a minor subset of αβ T cells, subsets of NKT and γδ T cells are the cellular sources of innate IFN-γ in the spleen of mice early after LM infection. Contrary to the previous report of Chang et al. (2007) that γδ T cells were not involved in the cellular sources of IFN-γ early after LM infection, our study clearly showed that a subset of γδ T cells participates in the innate IFN-γ production. This discrepancy may be due to the fact that Chang et al. (2007) cultured splenocytes ex vivo with BFA for a short period of time before flow-cytometric analysis while we employed a technique of in vivo administration of BFA which can avoid in vitro artificial effects (Lin and Whitton, 2005). In addition, our results indicate that NK1.1 + γδT cells are main innate IFN-γ-producers among γδ T cell populations in vivo early after LM infection. This result is compatible with the recent reports that γδ T cells with a NK1.1 + phenotype are efficient IFN-γ-producers among γδ T cell populations (Stewart et al., 2007;Haas et al., 2009). The important role of NK1.1 + γδ T cells in innate IFN-γ production has also been suggested by the results of other investigators (Emoto et al., 2001).
It has been claimed that macrophages and DCs can produce IFN-γ under certain conditions, playing an important role in the initial stages of infection (Frucht et al., 2001). However, this notion has been challenged by the recent reports showing that contaminant NK cells with a CD11c + phenotype may be the source of IFN-γ in macrophage and/or DC populations, and that there were no IFN-γ + MHC class II + macrophages/DCs among ex vivo splenocytes from mice early after LM infection (Laouar et al., 2005;Schleicher et al., 2005;Thale and Kiderlen, 2005). Our study using a technique of in vivo administration of BFA clearly indicates that CD11c hi DCs did not produce IFN-γ in the spleen. Thus, albeit not excluding a possibility that DCs might produce IFN-γ at later stages in infection, it is unlikely that DCs become a source of IFN-γ in early innate immune responses against bacterial infection.
Recently, it has been reported that NK1.1 + CD11c + cells are the primary IFN-γ-producing cells early after LM infection in mice (Chang et al., 2007;Plitas et al., 2007). NK1.1 + CD11c + cells or B220 + NK1.1 + CD11c + cells have been reported to be multifunctional cells with attributes of both NK cells and DCs, and they are referred to as NK dendritic cells (NKDCs) and IFN-γ-producing killer DCs (IKDCs), respectively (Pillarisetty et al., 2005;Chan et al., 2006;Taieb et al., 2006). The present study confirmed that NK1.1 + CD11c + cells are indeed the primary cellular source of IFN-γ in both the spleen early after LM infection and our in vitro coculture system. However, the results presented here also demonstrate that among normal splenocytes, ∼50% of NK1.1 + NK cells express CD11c antigen along with the CD244, an NK receptor. Hence, the IFN-γ-producing NK1.1 + CD11c + CD3 − cell is most likely to belong to the NK cell type based on the definition of its surface markers. This conclusion is consistent with that of Laouar et al. (2005). In addition, IKDCs have recently been reported to be as an activated form of NK cells (Blasius et al., 2007;Vosshenrich et al., 2007). Thus, although it cannot be excluded that a particular subset of NK cells may have a DC-like function (Terme et al., 2009), our results show that NK cells are the major source of innate IFN-γ in the spleen, and emphasize the important role of NK cells for the innate response to bacterial infection, as has been proposed by other investigators (Bancroft, 1993;Unanue, 1997;Thale and Kiderlen, 2005).
As a whole, our comparative study of IFN-γ-producing cells in vivo and in vitro reveals that the features of the IFN-γ-producing cells in the spleens early after LM infection resemble to those observed in our in vitro coculture system in which NWNA spleen cells are cultured with dying macrophages resulting from bacterial infection (Kubota, 2006(Kubota, , 2010. In our coculture system, cell-cell interaction between NWNA spleen cells and dying bacterialinfected macrophages plays an important role in inducing IFN-γ. Thus, the similarity between the IFN-γ-producing cells in vitro and in vivo suggests that the cellular microenvironment of our in vitro setting that leads to the production of IFN-γ may reflect the in vivo pathophysiological situation in the spleen early after LM infection. Our immunohistochemical study showed that in the spleen early after LM infection innate IFN-γ-producing cells form clusters at the same site of the infectious foci located in close vicinity of the MZ. The location of infectious foci reported here is in accordance with that suggested by Jablonska et al. (2007). However, this is in contrast to the results of other investigators that infectious foci were predominantly localized to the T cell zone (Serbina et al., 2003;Muraille et al., 2005;Chang et al., 2007;Aoshi et al., 2008;Kang et al., 2008). The reasons for this discrepancy are not clear, although conflicting data may be due to the different doses of LM injected into mice and the different time points of excising the spleens from mice after LM infection. It is generally considered that LM-infected macrophages first accumulate at the MZ and the red pulp to form MZ infectious foci, followed by the translocation of infected macrophages into the white pulp where they establish secondary infectious foci (Conlan, 1996). Thus, it is most likely that we were mainly observing primary infectious foci formed in close vicinity of the MZ. Other investigators, usually using lower doses of LM bacteria for infection, appear to be observing secondary infectious foci, i.e., infected phagocytes translocated from the MZ into the white pulp (Muraille et al., 2005;Aoshi et al., 2008). In fact, in addition to the preferential existence of infectious foci in close vicinity of the MZ, we also observed some of the bacteria reaching the white pulp. Since both innate and adaptive immune responses are mounted in the spleen upon bacterial infection, our histological observations seem to reflect the complexity of host reactions that are proceeding during LM infection, and the different location of infectious foci may represent distinct stages of innate and adaptive immune responses.
At an early stage of LM infection blood-borne LM is reported to be mainly trapped by macrophages in the MZ (Aichele et al., 2003;Jablonska et al., 2007). Consistent with those reports, our immunohistochemical study showed that at 4-h postinfection most of the bacteria are trapped by ER-TR9 + MZMs. However, at 20-h postoinfection when bacteria multiplied and formed infectious foci, MZMs were found around the infectious foci but most of the bacteria were no longer associated with MZMs. The reason most of the bacteria were not found in MZMs at this time point is not clear at present. Our study comparing IFN-γproducing cells between in vitro and in vivo suggests a possibility that IFN-γ production in the LM-infected spleen occurs under the microenvironment similar to our in vitro coculture system in which cell death in bacterial-infected macrophages is required for inducing IFN-γ (Kubota, 2006(Kubota, , 2010. Therefore, a plausible explanation may well be that in infectious foci LM-infected MZMs undergo cell death that is associated with the production of IFN-γ by innate lymphocytes (Kubota, 2010). Similarly, Jablonska et al. (2007) have argued that LM-infected MZMs in infectious foci might be killed by bacterium. The rationale for their argument is based on the facts that LM-infected MZMs secrete chemokines, thereby becoming the condensation nucleus of inflammatory cell clusters and that when CCL2 chemokine is depleted by injecting anti-CCL2 Abs into mice, uninfected MZMs are not attracted to infectious foci but also the infected MZMs that should have become the condensation nucleus are undetectable in the infectious foci. Thus, one can envisage that early after LM infection in mice LM bacteria are initially trapped by MZMs in the spleen and then the LM-infected MZMs accumulate at discrete sites of the MZ where they attract inflammatory cells by secreting chemokines and that during this period LM-infected MZMs undergo cell death, thereby triggering attracted innate lymphocytes to produce IFN-γ as proposed previously (Kubota, 2010). One must, however, await further evidence in support of this assumption.
In summary, our study revealed the cellular sources of innate IFN-γ in the spleen early after LM infection and shed light on the dynamic relationships among the IFN-γ-producing cells, bacteria, and MZMs in the LM-infected spleen. Furthermore, this study highlights the importance of the cellular microenvironment in which innate lymphocytes and bacterial-infected macrophages www.frontiersin.org interact and cooperate to produce innate IFN-γ and ultimately to control bacterial infection. Further study using our in vitro coculture system will be instrumental in understanding the cellular microenvironment that leads to the production of innate IFN-γ in vivo early after LM infection. | 2014-10-01T00:00:00.000Z | 2011-03-28T00:00:00.000 | {
"year": 2011,
"sha1": "dceaf9501a8d7fbf93458bddfe1bc5eb69623bb9",
"oa_license": "CCBY",
"oa_url": "https://www.frontiersin.org/articles/10.3389/fimmu.2011.00026/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "dceaf9501a8d7fbf93458bddfe1bc5eb69623bb9",
"s2fieldsofstudy": [
"Biology",
"Medicine"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
36020737 | pes2o/s2orc | v3-fos-license | Electronically Stabilized Copoly(Styrene-Acrylic Acid) Submicrocapsules Prepared by Miniemulsion Copolymerization
This work reports the preparation and characterization of poly(styrene-acrylic acid) (St/AA) submicrocapsules by using the miniemulsion copolymerization method. AA was introduced to miniemulsion polymerization of St to increase the zeta potential and the resulting electrostatic stability of St/AA submicrocapsules. Phytoncide oil was adopted as the core model material. Miniemulsion copolymerization of St and AA was conducted at a fixed monomer concentration (0.172 mol) with a varying monomer feed ratio [AA]/[St] (0.2, 0.25, 0.33, 0.5, and 1.0). Concentrations of initiator (azobisisobutyronitrile; 1.0 × 10−3, 2.0 × 10−3, 3.0 × 10−3, and 4.0 × 10−3 mol/mol of monomer) and surfactant (sodium dodecyl sulfate; 0.6 × 10−3, 1.0 × 10−3, and 1.4 × 10−3 mol) were also controlled to optimize the miniemulsion copolymerization of St and AA. Dynamic light scattering and microscopic analyses confirmed the optimum condition of miniemulsion copolymerization of St and AA. Long-term colloidal stability of aqueous St/AA submicrocapsule suspension was evaluated by using TurbiscanTM Lab. In this work, the optimum condition for miniemulsion copolymerization of St and AA was determined ([AA]/[St] = 0.33; [SDS] = 1.0 × 10−3 mol; [AIBN] = 2.0 × 10−3 mol/mol of monomer). St/AA submicrocapsules prepared at the optimum condition (392.6 nm and −55.2 mV of mean particle size and zeta potential, respectively) showed almost no variations in backscattering intensity (stable colloids without aggregation).
Zeta potential is the potential difference between the dispersion medium (water, in this case) and the stationary layer of fluid attached to the dispersed particle [19]. Therefore, the zeta potential can be used as a key indicator of the stability of colloidal dispersions because the magnitude of the zeta potential indicates the degree of electrostatic repulsion between particles in a dispersion. For small particles, high zeta potential will confirm the electrostatic stability, preventing aggregation in the dispersion. If the zeta potential is small, this electrostatic repulsion may not be implemented by exceeding attractive forces, resulting in the flocculation of particles. Mandzy et al. [20] reported that reagglomeration of the TiO 2 nanoparticle dispersion could be prevented by electrostatic stabilization when its zeta potential value was less than −30 mV or greater than +30 mV. Many other researchers have investigated the effects of zeta potential on particle stability for metallic powders including zirconia and titania [21][22][23][24].
Polystyrene nanocapsules have been adopted as vehicles for photochromophores [25], phase change materials [26], and so on, by virtue of their ease of preparation and thermal stability. However, chronic particle aggregation of polystyrene nanoparticles has constrained their practical usage. Even though several researchers have investigated the miniemulsion polymerization of styrene (St) and acrylic acid (AA) to enhance the colloidal stability of polystyrene nanocapsules, copolymerization of AA and St with over 20 wt % of AA has rarely been considered for water compatibility of AA [27][28][29].
In this work, we adopted AA to miniemulsion polymerization of St to increase the zeta potential and resulting electrostatic stability of St/AA submicrocapsules containing phytoncide oil. To this aim, St/AA submicrocapsules were prepared at over 20 wt % of AA content. Mean particle size and zeta potential of St/AA submicrocapsules were then analyzed by dynamic light scattering (DSL). Dispersability of submicrocapsules was also evaluated by Turbiscan TM Lab to confirm the sink stability over time.
Materials
St and AA were purchased from Aldrich Chemical (St. Louis, MI, USA) and used as monomers after distillation. Azobisisobutyronitrile (AIBN) and divinyl benzene (DVB) were also provided by Aldrich Chemical (St. Louis, MI, USA) and used as initiator and crosslinking agent, respectively. Dae Jung Chemical (Siheung, Korea) supplied sodium dodecyl sulfate (SDS), which is used as surfactant. n-Hexadecane was purchased from Alfa Aesar (Ward Hill, MA, USA) and used as cosurfactant. Distilled water was of Milli-Q quality (Millipore, Billerica, MA, USA). Phytoncide oil was provided by CNG Co. (Daegu, Korea). All the reagents were of either HPLC grade or American Chemical Society analytical grade.
Miniemulsion Copolymerization
SDS (0.6~1.4 × 10 −3 mol) was dissolved in distilled water to prepare the water phase. St and AA were homogeneously mixed with phytoncide oil, and n-hexadecane to form the oil phase. These solutions were emulsified with homogenizer for 10 min at 19,000 rpm to prepare micelles in submicrometer range. AIBN (initiator, 1.0~4.0 × 10 −3 mol/mol of monomer) was added at 60 • C and copolymerization was maintained for 6 h. The pH of the reaction system was slightly lowered with the increase of the amount of AA, ranging from 2.3 to 2.6. Detailed experimental conditions are presented in Table 1.
Characterization
The size and morphology of St/AA submicrocapsules were analyzed by scanning electron microscopy (SEM) (SU8220, Hitachi, Tokyo, Japan). The mean particle diameters and zeta potentials of St/AA submicrocapsules were also determined using DLS (ELS8000, Photal Otsuka Electronics, Osaka, Japan) equipped with vertically polarized light supplied by a He-Ne laser, operated at 10 mW. DLS measurements were performed at room temperature. The dispersion stability of St/AA submicroparticles in the aqueous phase was evaluated from the time variation of backscattered light by using the Turbiscan TM Lab (Formulaction LAB, Toulouse, France). All the measurements were repeated three times.
Effects of Monomer Feed Ratio
The aim of miniemulsion copolymerization of St and AA is to increase the stability and prevent the coagulation of submicrocapsules containing functional materials in the core (phytoncide oil, in this case) by virtue of the enhanced zeta potential of submicrocapsules due to -COO − functional groups of AA. The salient feature of our current study is that miniemulsion copolymerization of St with AA yields electronically stabilized submicrocapsules which can be used as carriers for the lypophilic core. over 0.33, however, there was a steep increase in the mean particle size of the submicroparticles. As is well known, copolymerization behavior is mainly affected by the different reactivities of monomers. Compared with St, AA polymerizes vary rapidly, especially at low pH. Therefore, the increase of AA led inevitably to longer polyacrylic blocks in the copolymer formation at the particle-water interface, and increased the particle diameter. The increase in particle size with the increase in the amount of AA results in the formation of a hydrophilic swollen layer around the particle [27]. Absolute value of zeta potential ( Figure 1b 33 was attributed to the particle agglomeration in the aqueous phase due to the increased hydrophilic swollen layer of the particles and resulting reduction in surface area [29,30]. This can also be confirmed from SEM photographs presented in Figure Figure 3 shows the effects of surfactant concentration on the mean particle size and zeta potential of St/AA submicrocapsules prepared at 0.33 and 2.0 × 10 −3 mol/mol of monomer for [AA]/[St] and AIBN concentrations, respectively. The mean particle size of the submicrocapsules decreased with the increasing surfactant concentration (Figure 3a). This result is in good agreement with those of other researchers, suggesting that more SDS molecules are available for stabilizing the oil-water interfacial area generated during homogenization at a higher level of SDS [29,[31][32][33]. The zeta potential of submicrocapsules, however, showed the optimum value at 1.0 × 10 −3 mole of surfactant concentration ( Figure 3b). Figure 3 shows the effects of surfactant concentration on the mean particle size and zeta potential of St/AA submicrocapsules prepared at 0.33 and 2.0 × 10 −3 mol/mol of monomer for [AA]/[St] and AIBN concentrations, respectively. The mean particle size of the submicrocapsules decreased with the increasing surfactant concentration (Figure 3a). This result is in good agreement with those of other researchers, suggesting that more SDS molecules are available for stabilizing the oil-water interfacial area generated during homogenization at a higher level of SDS [29,[31][32][33]. The zeta potential of submicrocapsules, however, showed the optimum value at 1.0 × 10 −3 mole of surfactant concentration (Figure 3b). Figure 3 shows the effects of surfactant concentration on the mean particle size and zeta potential of St/AA submicrocapsules prepared at 0.33 and 2.0 × 10 −3 mol/mol of monomer for [AA]/[St] and AIBN concentrations, respectively. The mean particle size of the submicrocapsules decreased with the increasing surfactant concentration (Figure 3a). This result is in good agreement with those of other researchers, suggesting that more SDS molecules are available for stabilizing the oil-water interfacial area generated during homogenization at a higher level of SDS [29,[31][32][33]. The zeta potential of submicrocapsules, however, showed the optimum value at 1.0 × 10 −3 mole of surfactant concentration (Figure 3b). Elimelech and O'Melia [34] reported that the collision efficiency of colloidal particles is independent of particle size. They also suggested that coupling of electrodynamics and hydrodynamics, and the possible effects of surface roughness have a significant effect on the kinetics of particle-particle interaction. With the increase in the concentration of SDS, a large number of smaller micelles can be obtained. Therefore, St/AA submicrocapsules prepared from these micelles are supposed to have a relatively thin and low molecular weight electrostatic polyacrylic shell. This meager swollen layer will become vulnerable to aggregation and, as a result, show lower zeta potential. This reduction in the absolute value of zeta potential at higher SDS content can be confirmed by Figure 4 which presents SEM photographs of St/AA submicroparticles with varying surfactant concentrations. Figure 5 shows the effects of the initiator concentration on the mean particle size and zeta potential of St/AA submicrocapsules prepared at 0.33 and 1.0 × 10 −3 mol for [AA]/[St] and SDS concentrations, respectively. As is well known, the initiator concentration greatly affects the reaction rate and extent of radical polymerization. The increase in initiator concentration at a fixed amount of surfactant will increase the possibility and frequency of droplet nucleation in a micelle. At this point, preference of AA for copolymerization with St is no longer maintained. Therefore, the hydrophilic swollen layer composed of longer polyacrylic blocks and the resulting zeta potential of St/AA submicrocapsules were assumed to be diminished for higher initiator concentrations. On the other hand, a limited amount of initiator could not properly activate the droplet nucleation. At an initiator concentration of 1.0 × 10 −3 mol/mol of monomer, it was proposed that submicrocapsules were formed insufficiently, which was confirmed by large particle size and low zeta potential. As shown in Figure Elimelech and O'Melia [34] reported that the collision efficiency of colloidal particles is independent of particle size. They also suggested that coupling of electrodynamics and hydrodynamics, and the possible effects of surface roughness have a significant effect on the kinetics of particle-particle interaction. With the increase in the concentration of SDS, a large number of smaller micelles can be obtained. Therefore, St/AA submicrocapsules prepared from these micelles are supposed to have a relatively thin and low molecular weight electrostatic polyacrylic shell. This meager swollen layer will become vulnerable to aggregation and, as a result, show lower zeta potential. This reduction in the absolute value of zeta potential at higher SDS content can be confirmed by Figure 4 which presents SEM photographs of St/AA submicroparticles with varying surfactant concentrations. Elimelech and O'Melia [34] reported that the collision efficiency of colloidal particles is independent of particle size. They also suggested that coupling of electrodynamics and hydrodynamics, and the possible effects of surface roughness have a significant effect on the kinetics of particle-particle interaction. With the increase in the concentration of SDS, a large number of smaller micelles can be obtained. Therefore, St/AA submicrocapsules prepared from these micelles are supposed to have a relatively thin and low molecular weight electrostatic polyacrylic shell. This meager swollen layer will become vulnerable to aggregation and, as a result, show lower zeta potential. This reduction in the absolute value of zeta potential at higher SDS content can be confirmed by Figure 4 which presents SEM photographs of St/AA submicroparticles with varying surfactant concentrations. Figure 5 shows the effects of the initiator concentration on the mean particle size and zeta potential of St/AA submicrocapsules prepared at 0.33 and 1.0 × 10 −3 mol for [AA]/[St] and SDS concentrations, respectively. As is well known, the initiator concentration greatly affects the reaction rate and extent of radical polymerization. The increase in initiator concentration at a fixed amount of surfactant will increase the possibility and frequency of droplet nucleation in a micelle. At this point, preference of AA for copolymerization with St is no longer maintained. Therefore, the hydrophilic swollen layer composed of longer polyacrylic blocks and the resulting zeta potential of St/AA submicrocapsules were assumed to be diminished for higher initiator concentrations. On the other hand, a limited amount of initiator could not properly activate the droplet nucleation. At an initiator concentration of 1.0 × 10 −3 mol/mol of monomer, it was proposed that submicrocapsules were formed insufficiently, which was confirmed by large particle size and low zeta potential. As shown in Figure Figure 5 shows the effects of the initiator concentration on the mean particle size and zeta potential of St/AA submicrocapsules prepared at 0.33 and 1.0 × 10 −3 mol for [AA]/[St] and SDS concentrations, respectively. As is well known, the initiator concentration greatly affects the reaction rate and extent of radical polymerization. The increase in initiator concentration at a fixed amount of surfactant will increase the possibility and frequency of droplet nucleation in a micelle. At this point, preference of AA for copolymerization with St is no longer maintained. Therefore, the hydrophilic swollen layer composed of longer polyacrylic blocks and the resulting zeta potential of St/AA submicrocapsules were assumed to be diminished for higher initiator concentrations. On the other hand, a limited amount of initiator could not properly activate the droplet nucleation. At an initiator concentration of 1.0 × 10 −3 mol/mol of monomer, it was proposed that submicrocapsules were formed insufficiently, which was confirmed by large particle size and low zeta potential. As shown in Figure 5b, the St/AA submicrocapsule prepared at 1.0 × 10 −3 mol/mol of monomer of the initiator concentration shows no surface potential (almost zero potential) which means that immense coagulation of particles will occur. Polymers 2017, 9, 291 6 of 10 5b, the St/AA submicrocapsule prepared at 1.0 × 10 −3 mol/mol of monomer of the initiator concentration shows no surface potential (almost zero potential) which means that immense coagulation of particles will occur.
Colloidal Stability
In this work, the long-term stability of St/AA submicrocapsules containing phytoncide oil was investigated by evaluating both the optical transmission and the photon backscattering profiles of the aqueous dispersion of submicrocapsules by using the Turbiscan TM Lab (Formulaction, Toulouse, France) [35]. Measurements were carried out using a pulsed near infrared LED at a wavelength of 880 nm for 24 h. Two different synchronous optical sensors received the light transmitted through and backscattered by samples at an angle of 180° and 45° with respect to the incident radiation, respectively. Transmitted and backscattered light flux were correlated as a percentage to those of reference standards. If there was no variation greater than 2% in the backscattering profile, it can be considered as a stable formulation [36].
The Figure 6. Variations of transmission and backscattering profiles were not correlated to the destabilization processes below the sample height of 5 mm and over that of 40 mm, the values having been determined by enclosed air in the bottom and/or on the top of the cylindrical glass tube, respectively. Analysis of the submicrocapsule dispersion showed that backscattering was the prevalent signal and there were almost no variations in the backscattering intensity for the entire test height. From this result, it can be confirmed that St/AA submicrocapsules prepared at the optimum condition could be formulated as stable colloids without aggregation.
Colloidal Stability
In this work, the long-term stability of St/AA submicrocapsules containing phytoncide oil was investigated by evaluating both the optical transmission and the photon backscattering profiles of the aqueous dispersion of submicrocapsules by using the Turbiscan TM Lab (Formulaction, Toulouse, France) [35]. Measurements were carried out using a pulsed near infrared LED at a wavelength of 880 nm for 24 h. Two different synchronous optical sensors received the light transmitted through and backscattered by samples at an angle of 180 • and 45 • with respect to the incident radiation, respectively. Transmitted and backscattered light flux were correlated as a percentage to those of reference standards. If there was no variation greater than 2% in the backscattering profile, it can be considered as a stable formulation [36].
The Figure 6. Variations of transmission and backscattering profiles were not correlated to the destabilization processes below the sample height of 5 mm and over that of 40 mm, the values having been determined by enclosed air in the bottom and/or on the top of the cylindrical glass tube, respectively. Analysis of the submicrocapsule dispersion showed that backscattering was the prevalent signal and there were almost no variations in the backscattering intensity for the entire test height. From this result, it can be confirmed that St/AA submicrocapsules prepared at the optimum condition could be formulated as stable colloids without aggregation.
Conclusions
It was shown that St/AA submicrocapsules can be obtained by miniemulsion copolymerization of AA and St with lypophilic core materials. There was an optimum condition for obtaining electronically stable St/AA submicrocapsules in the aqueous phase.
With the increase of [AA], the mean particle size and zeta potential of St/AA submicrocapsules were greatly decreased due to anionic AA and polyacrylic blocks in micelles and submicrocapsules, respectively. However, at over [AA]/[St] = 0.33, the mean particle size of St/AA submicrocapsules was increased, which was ascribed to the particle agglomeration in the aqueous phase due to the increased hydrophilic swollen layer of the particles and the resulting reduction in surface area and zeta potential. The increase in the applied amount of surfactant assures the reduction in size of micelles and resulting St/AA submicrocapsules. At higher [SDS] of over 1.0 × 10 −3 mol, however, the formation of the weakened swollen layer, causing particle aggregation and a decrease in zeta potential, was suspected. Mean particle size and zeta potential of St/AA submicrocapsules were at the optimum condition at an initiator concentration of 2.0 × 10 −3 mol/mol of monomer. Droplet nucleation was not effectively activated at a lower initiator concentration. At a higher initiator concentration, particle aggregation was manifested by virtue of the reduction in the portion of polyacrylic blocks in the swollen layer around the submicrocapsules.
Conclusions
It was shown that St/AA submicrocapsules can be obtained by miniemulsion copolymerization of AA and St with lypophilic core materials. There was an optimum condition for obtaining electronically stable St/AA submicrocapsules in the aqueous phase.
With the increase of [AA], the mean particle size and zeta potential of St/AA submicrocapsules were greatly decreased due to anionic AA and polyacrylic blocks in micelles and submicrocapsules, respectively. However, at over [AA]/[St] = 0.33, the mean particle size of St/AA submicrocapsules was increased, which was ascribed to the particle agglomeration in the aqueous phase due to the increased hydrophilic swollen layer of the particles and the resulting reduction in surface area and zeta potential. The increase in the applied amount of surfactant assures the reduction in size of micelles and resulting St/AA submicrocapsules. At higher [SDS] of over 1.0 × 10 −3 mol, however, the formation of the weakened swollen layer, causing particle aggregation and a decrease in zeta potential, was suspected. Mean particle size and zeta potential of St/AA submicrocapsules were at the optimum condition at an initiator concentration of 2.0 × 10 −3 mol/mol of monomer. Droplet nucleation was not effectively activated at a lower initiator concentration. At a higher initiator concentration, particle aggregation was manifested by virtue of the reduction in the portion of polyacrylic blocks in the swollen layer around the submicrocapsules.
From the above consideration, the optimum condition for miniemulsion copolymerization of St and AA was determined as follows; [AA]/[St] = 0.33, [SDS] = 1.0 × 10 −3 mol, and [AIBN] = 2.0 × 10 −3 mol/mol of monomer. Long-term colloidal stability was also evaluated for the optimum condition of St/AA submicrocapsules by measuring the optical transmission and the photon backscattering profiles of the aqueous dispersion. It is shown that backscattering was the prevalent signal and there were almost no variations in backscattering intensity, which means that St/AA submicrocapsules prepared at the optimum condition have superior colloidal stability without aggregation and sedimentation. These electrically stable St/AA submicrocapsules can be applied to various fields such as a targeted and/or sustained release drug delivery system, emulsion templating of 3D objects, additives for polymeric resins, highly dispersive fiber finishing agent, and so on. | 2017-08-21T03:25:37.558Z | 2017-07-01T00:00:00.000 | {
"year": 2017,
"sha1": "a222cbe6e8ee0ec650f6b70ceea053354f85b2b8",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2073-4360/9/7/291/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "a222cbe6e8ee0ec650f6b70ceea053354f85b2b8",
"s2fieldsofstudy": [
"Chemistry",
"Materials Science"
],
"extfieldsofstudy": [
"Materials Science",
"Medicine"
]
} |
229283850 | pes2o/s2orc | v3-fos-license | Parent Mobile Phone Use in Playgrounds: A Paradox of Convenience
Creating social and physical environments that promote good health is a key component of a social determinants approach. For the parents of young children, a smartphone offers opportunities for social networking, photography and multi-tasking. Understanding the relationship between supervision, mobile phone use and injury in the playground setting is essential. This research explored parent mobile device use (MDU), parent–child interaction in the playground, parent attitudes and perceptions towards MDU and strategies used to limit MDU in the playground. A mixed-methods approach collected naturalistic observations of parents of children aged 0–5 (n = 85) and intercept interviews (n = 20) at four metropolitan playgrounds in Perth, Western Australia. Most frequently observed MDU was scrolling (75.5%) and telephone calls (13.9%). Increased duration of MDU resulted in a reduction in supervision, parent–child play and increased child injury potential. The camera function offered the most benefits. Strategies to prevent MDU included turning to silent mode, wearing a watch and environmental cues. MDU was found to contribute to reduced supervision of children, which is a risk factor for injury. This is an emerging area of injury prevention indicating a need for broader strategies addressing the complex interplay between the social determinants and the developmental younger years.
Introduction
For the parents of young children, a mobile device, such as a smartphone, offers opportunities for communication, social networking, photography, personal organisation, work flexibility and multi-tasking [1][2][3][4]. However, it can also be a source of distraction that brings about feelings of guilt and concern for poor role modelling of mobile device use (MDU) for young children during their early development [1,[4][5][6][7][8]. Parent MDU can have detrimental effects on family relationships including interaction and attachment [3,5,9,10], reducing parents' responsiveness when children try to attract their attention [3,5] and social support [11].
The early years of life represent a vital time during which lifelong trajectories for health are determined by the complex interplay between the social determinants [12]. Social determinants of health are conditions in the environments in which children are born, live, learn, interact and play [13]. Creating social and physical environments that promote good health is a key component of a social determinants approach to health [14]. The social environment includes interactions with family and parent-child play encompasses social relationships in physical settings such as playgrounds [12,14,15]. Of interest, outdoor play is deemed fundamental for the physical, cognitive, emotional and social well-being of children [15,16], and has been declared a right of every child by the United Nations Convention on the Rights of the Child [17]. Outdoor play in early childhood has been linked to the social determinants that shape health and development, emphasising children's participation in play is influenced by multiple interacting social and contextual factors [15]. Very few studies have focused on understanding the social (e.g., parental supervision) and contextual (e.g., playground setting) factors that may influence behaviours and the relationship with the injury.
Whilst parent supervision is recognised as being a protective factor for injury in young children [18][19][20], there is evidence that there is a reduction in supervision quality as a result of parent MDU [21,22], and an association between parent MDU and injury in young children [23]. A recent US study found that children's safety needs were more often at risk when parents used their phones than when they were distracted by other factors, such as talking with other adults or reading printed material [24]. A review of the impact of parent MDU on parent-child interaction found increased phone ownership and related increases in parents' lack of attention to children could be associated with increases in childhood injuries [25]. US research has also reported a positive association between young children presenting at hospitals with injuries and the rapid adoption of smartphone use [23,26]. One study concluded that distraction from increased mobile phone use was directly associated with an increase in emergency hospitalisation for injuries in children aged 0-5 years; however, there was no association with children aged 6-10 years [23]. Hiniker et al. (2015), on the other hand, reported parents did not use their phone in the playground setting because they believed it would compromise their child's safety and lower their ability to respond. There is an increasing interest in understanding the impact of mobile device use by parents/carers of children aged 0-5 years [1,25], which commands an investigation into which characteristics of MDU are most absorbing of their attention [9,25,27] when supervising children.
The aim of the naturalistic study was twofold: first, to observe mode and duration of parent and caregiver MDU (telephone call, scroll/type or camera), parent/caregiver interaction and the coinciding behaviour (specifically: supervision, interaction and child injury) in the playground setting. Second, to explore parent/caregiver attitudes and perceptions of MDU and strategies used to limit MDU in the playground setting.
Study Design
A mixed-methods approach was used to collect quantitative and qualitative data from parent or caregiver/child dyads in playgrounds in Perth, Western Australia (WA). Trained researchers worked in teams of two to conduct covert, naturalistic observations (n = 85). In addition, intercept interviews were conducted with parents/caregivers (n = 20). The sample size for the observations was informed by naturalistic observational methodologies for data collection conducted in New South Wales, Australia (n = 50) [4] and New Jersey, US (n = 60) [27], whilst one other study in the US collected 171 observations over almost double (seven parks) [24] compared with four in this study. The sample size for the interview numbers (n = 20) were consistent with similar studies [4,27].
Study Setting
The study was undertaken in four playgrounds across a range of Socio-Economic Indexes for Areas (SEIFA) [28] in metropolitan Perth, WA. The playgrounds included: (i) a fenced playground with several different play areas, built specifically for 6-year-olds and under; (ii) a fenced playground with low-level equipment for toddlers; (iii) an ungated single playground adjacent to a number of BBQ pits, and grass/benches for picnics; and (iv) an ungated single playground. Observations were undertaken on school days between the hours of 9:30 a.m. and 12:30 p.m., a time when older siblings have been dropped off at school and a popular time for 0-5 years to attend playgrounds [4,27].
Study Participants
Participants were parent or caregiver/child dyads where the parent/caregiver appeared to be under 45 years, attending the playground by themselves (not with another adult) and caring for at least one independently mobile child who appeared to be under 5 years. The estimation of the child's age was based on general appearance and height. Parents caring for more than one child were included in the study. Where parent/carers had multiple children under their care, the researchers chose the child that best fitted the criteria-under 5 years and independently mobile. Participants were not notified they were being observed so as to obtain naturalistic data. Participants who left the playground or were lost from the site (into a café/toilet) within the first 10 min of the observation were excluded from the study. Researchers recruited participants for intercept interviews (n = 20). Parent/Carers appearing to meet the participant criteria were approached and invited to participate in the study. No parent/carers declined to be interviewed. Participants were provided with an information sheet and consent form prior to the commencement of interviews. Ethics approval was obtained for this study by Curtin University Research Ethics Committee no: HRE2016-0027.
Quantitative
The two researchers positioned themselves within the boundary of the playground to observe participants, one observation at a time. At the beginning of the observation period, the researchers synchronised stopwatches to concurrently record data for a period of up to twenty minutes. Participants were observed for MDU (mode and duration of use), caregiver supervision, caregiver/child interaction and child injury potential with data recorded during each minute.
The first researcher collected data for caregiver MDU using an adapted, previously trialled mobile device timing audit [4,27]. During each minute of the observation, the researcher observed the dominant mode of caregiver MDU (telephone call, scroll/type or camera) and the duration of caregiver MDU (recorded as a minimum of 0 s and a maximum of 60 s). The summed duration for each minute of observation was later categorised as; "No MDU"; "1-10 s"; "11-20 s"; "21-30 s"; "31-40 s"; "41-50 s" and "51-60 s". The second researcher collected data on caregiver supervision, caregiver-child interaction and child injury potential using a combined observation audit [4,27]. During each minute of the observation, the researcher observed and recorded the dominant behaviour.
Qualitative
Caregivers appearing to meet the participant criteria were approached and invited to participate in the study. Participants were provided with an information sheet and consent form prior to the commencement of interviews. The researcher recorded the intercept interviews using a structured interview schedule. Each interview took approximately 10 min to complete. The interviews were conducted by two researchers, field notes were kept by both researchers and compared for consistency at the completion of each interview period.
Measures
Data from the mobile timing device audit were collapsed and used to create three categorical variables: (1) MDU: this binary variable categorised minutes as either "No MDU" (0 s of MDU) or "MDU" (1-60 s of MDU). (2) MDU Mode: this variable categorised minutes as "Telephone call" (using the device telephone call function) "Scroll/type" (using the device touchscreen function), or "Camera" (using the device camera function).
Data from the caregiver-child observation tool for supervision, interaction and injury potential were collapsed: (1) Supervision: "Constant supervision" (caregiver watching, following, mediating, redirecting the child or remaining in close proximity); "Intermittent supervision" (caregiver sought visual contact with the child intermittently); "No supervision" (caregiver had no contact with the child). (2) Interaction: "Caregiver-child play" (caregiver and child play together), "Independent play" (child plays without caregiver), "Verbal interaction" (caregiver and/or child talking or calling to each other), "Hold/touch" (physical contact between caregiver and child), "Sitting/eating/drinking" (caregiver and child in close proximity having a drink or food). (3) Injury Potential: Child injury risk was measured using an adapted injury risk behaviour checklist developed by Dotson (2013). Injury potential was categorised as "Increased injury" potential (unsafe play behaviours, i.e., child passes within moving radius of equipment, child uses equipment in an unintended manner), "Decreased injury" potential (safe play behaviours, i.e., child takes precaution, child stops swing and dismounts) or "Inadequate carer supervision" (i.e., child moves out of view of caregiver, the caregiver does not give child direction on how to behave safely).
Quantitative Data
The data for the 85 observed participants were broken down into one-minute blocks which resulted in a total of 1532 min of observation time which was entered into SPSS version 23 [29]. Descriptive statistics were used to describe the parent/caregiver participant characteristics and the children under the parent/caregiver's supervision. The quantitative analysis explored the coinciding MDU, supervision, interaction and injury risk behaviours within each minute of the observation. A series of cross-tabulations with Pearson chi-square tests with a significance level of p < 0.01 were conducted to test the association between MDU, MDU Mode, MDU Duration and the outcome variables Supervision, Interaction and Injury Potential.
Qualitative Data Collection and Analysis
The interviews were transcribed verbatim by one researcher (KB) and were divided among the research team for open coding (KB, JL, JJ). The first researcher then collated the coded interviews using NVIVO 12 [30] and identified relationships between the codes to form the emerging themes which were agreed upon by the research team. Data saturation was reached during the data collection process (n = 20) and no new concepts emerged. Identified themes were consistent with work previously undertaken by Australian researchers [4]. The general inductive approach is a straightforward easily used, systematic set of procedures for analysing qualitative data and provides reliable findings [31]. Participant quotes to support the themes have been de-identified and presented in the results.
Participant Characteristics
Participants were mostly female (n = 72, 85%) and caring for one child under the age of 5 years (n = 68, 80%). Of the 85 parent/caregiver-child dyad observations (n = 85), 47 were in high socio-economic status (SES) playground locations and 38 in low SES playground locations.
Mobile Device Use-Mode and Duration
During the observation period, caregiver MDU was observed among 70% (n = 59) of the caregiver-child dyads, 30% (n = 16) of parents/carers did not use their mobile device, with total caregiver MDU comprising 23.5% of the observed time. The most frequently observed mode of MDU was "Scrolling/typing" (n = 272, 75.5%), followed by telephone call (n = 50, 13.9%) and using the camera (n = 38, 10.5%). The duration of MDU was either short of "1-10 s" (n = 91) or almost a full minute ("51-60 s") (n = 92). Table 1 presents a comparison of supervision behaviours between minutes of "No MDU" and minutes of "MDU". For all observed minutes, the proportion of "No supervision" was higher when using a mobile device (MDU) compared with "No MDU" (29.4% and 5.2%, respectively). The proportion of "Constant supervision" was similar for "MDU" (45.8%) and "No MDU" (43.6%). A significant association was found between MDU and caregiver supervision χ 2 (2) = 191.67, p ≤ 0.001.
Caregiver Mobile Device Use and Caregiver-Child Interaction
"Independent play" (32.6%, n = 500) and "Caregiver-child play" (35.4%, n = 542) each made up one-third of the interaction behaviour of all observed minutes (Table 2). However, the proportion of "Caregiver-child play" with MDU was half that of no MDU (20.3% and 40.0%, respectively), while the proportion of "Independent play" for MDU was double that of "No MDU" (50.3% compared with 27.2%, respectively). The results show similar proportions for "Verbal interaction" within the "MDU" and "No MDU" groups (22.2% and 20.8%, respectively). A significant association was found between MDU and caregiver-child interaction χ 2 (4) = 81.95, p ≤ 0.001.
Caregiver Mobile Device Use and Child Injury Potential
Increased child injury potential was observed among 9.5% (n = 146) of all observed minutes. The proportion of observed "Increased injury potential" for "No MDU" was 5.7% (n = 67) compared with 21.9% (n = 79) with "MDU" ( Table 3). The highest proportion of "Increased injury potential" was observed in the "51-60 s" (n = 30, 32.6%) MDU duration category. When comparing child injury potential across the mode of caregiver MDU, the highest proportion of "Increased injury potential" was observed for "Telephone call" (n = 13, 26%).
Reported MDU Behaviours
Almost all participants reported using their mobile device at the playground that day (n = 19, 95%). Participants made short statements around their reason for their MDU in the playground on that day such as "Had to answer a couple of business calls" (P13) and "Just to check if I've received any messages" (P21). The most frequently reported reason for MDU included methods of communications (calls, text, email) (n = 12, 60%). Other reasons included photography (n = 9, 45%) and using the device to undertake specific tasks for personal administration (n = 3, 15%).
Perspectives on Caregiver MDU
Four dominant themes emerged from the participant interviews; these were: convenience, distraction, security and making memories.
Convenience
Participants cited essential daily communication including telephone calls and text-messaging, connecting through social media networks, easy access to information, and photography, for example, "If you need to know where a street is you can just google it. Taking photos of my child, sending them to different people would be an advantage and most importantly, if there's an emergency I can call someone." (P19).
In addition, other activities were web browsing for news, sports scores, attending to work-related matters, or activities for personal "admin", such as banking, shopping and coordinating services such as car repairs. The ability to undertake these activities anytime, anywhere "I guess you can multitask, you can get those little things done when you have a chance" (P20) contributed to the convenience and the need to carry a mobile device.
Distraction
In contrast, most participants reported that notifications, social media and the obligation to check in with work were a distraction, for example, "If you've got your phone it's very easy to be distracted and to keep it constantly in use and to be distracted from doing this I'll just check Facebook." (P2) Whilst the obligation to check in with work was also a distraction, "You can feel obligated to check work emails, like I'm just stepping out of work for today just to watch him for a minute but you do feel obligated to be checking, too available." (P5).
Participants highlighted a paradox between the convenience and distraction of MDU, giving reasons why they would not use their device in the playground. For example: "I guess that you have to be conscious that it doesn't distract you from spending this time with the kids . . . , and giving them your full attention because you want to be interacting, you want to be there with them in the moment". (P20) However, there was a tension between distraction and convenience, "It's like a psychological kind of thing where I'm drawn to answer that beep and that takes me away from my kid. It's on silent now. It's just a distraction. I use it for my convenience rather than other people's convenience." (P16).
Participants believed that brief moments of MDU, such as making a telephone call would have the least impact on supervision, whereas lengthy periods viewing the device for text messaging or social media were more likely to diminish the quality of supervision, "I think if you're scrolling through Facebook, then that obviously requires a lot more attention than if you're communicating on a phone. I think you're much more able to supervise if you're only using it in its basic sense whereas if you're typing a text or you're writing an email, scrolling through Facebook then I think your ability to supervise is impaired. Coz obviously it takes more concentration." (P5).
Using the camera was viewed differently "Other than photos, anytime you are with your kids you should be supervising them so in public unless you've got somebody else with you who can take over that role. I always have eyes on (child) even through the camera on my phone. I think that's the main thing." (P16).
Participants commented on the social acceptability of MDU within the playground setting with work tasks or essential telephone calls deemed more acceptable distractions than social media, along with some criticism of this behaviour. For example, "There's not too many things that are more important than watching a kid, you know if he's close to the road you don't need to use your phone but if he's in the playground and I need to take a phone call for any particular reason. But I don't think there is any reason that you need to use Facebook or something like that if the kid needs watching." (P5).
Security
Participants frequently reported the benefit (n = 14, 70%) of having their mobile device in case of an emergency within the playground, "The phone is just there as a secondary device as an emergency or if you need to contact someone." (P21); or outside the playground setting, "I really just try not to answer your phone unless you absolutely have to. You don't want to miss a call either, if it's something important like family or an emergency." (P18).
When considering supervising children, participants made comparisons to other settings, for example, being in and around water, "She actually just turned around to talk to me for a few seconds and that's all it took for a one year old to nose dive into the pool . . . I think so much can happen in the blink of an eyelid, you can't take your eye off them." (P18). Or other settings like driving a car, "You can't have your eyes on two things at once, it's impossible. Other people maybe can, I don't think I can. It's like the driving, they wouldn't have it banned if it were safe. Surely." (P22).
Making Memories
Photography was highlighted as a valuable aspect of MDU which enhanced their experience at the playground, "To capture precious moments, to be in contact with other people that aren't with you. Family overseas, you can post a photo of Facebook they can kind of be a part of it as well." (P10) It enabled them to capture memories with their children and share these with loved ones at a later date.
Caregivers were concerned about the ability to interact with and supervise children while using a mobile device, particularly if it involved using social media. However, photography was commonly indicated as an acceptable mode of MDU in the playground indicating the perception that photography can both mitigate safety concerns and enable interaction with children. For example: "I'm supposed to be looking after her, engaging with her, supervising her and making sure that she's safe unless it's for the camera." (P16).
Strategies for Limiting MDU around Children
Caregivers were forthcoming about their desire to limit MDU, which stemmed from wanting to spend quality time with their child, avoid distraction in risky settings, and role model appropriate technology use, so as to not raise device focussed children.
All participants nominated strategies to limit their MDU use in the playground setting, with the most common strategy defined as abstinence, "Don't bring it out, don't bring it, wear a watch!" (P6). Other participants, however, referred to restricting the functionality of their device by putting it on silence or only using it for a camera, "Put it on silent. I don't have my phone on when I'm meant to be watching her because I know how quickly things can change as well. Also having it on silent I don't kind of have that niggling feeling to look at all the beeps that go off. Just think of it as a camera when you're at the park, nothing else." (P16).
Participants reported 'downtime', which were times when they perceived their child was occupied and required less supervision or interaction. One participant explained "Most things require you to look at the screen. When she's on the swing. I'm not as concerned, coz she's strapped in, she's holding on. I look at it frequently, don't get me wrong, I'm not some saint." (P15).
Some participants reported the need to role-model positive MDU behaviours for their children, "I don't really think they need to come out of your bag when you're at the playground or at schools or in social situations with your own friends, you should be engaging with the people there in front of you. Family mealtimes, it's really important to sit down together, have dinner, have a discussion. I think kids today will miss out on a lot of social skills if they see everyone on their smartphone all the time." (P19). This resulted in curbing their MDU.
Discussion
This naturalistic study explored mobile device use by parent/caregivers of children aged 0-5 years in four playgrounds in metropolitan Perth, Western Australia. The study found that parent texting/scrolling, telephone calls and camera use were a common occurrence in the playground setting (70% of observation participants) but occupied a small part of their time at the playground. This finding is consistent with other research in Australia and the US [4,27], however, the paradox between convenience and distraction was a salient theme of the study which has also been found in other research [7,24,32,33]. Of interest, parents and carers suggested that the practicality of having a mobile phone close-by was reassuring in the event of an emergency, thus creating a tension between having the mobile phone close by or not at all. Child-injury prevention agencies and playground designers should explore broad strategies to promote the importance of supervision to prevent child injury, reduction in MDU in social settings and the promotion of parent-child play. Going forward, policies and programmes must embrace the social determinants of health that play a vital role in the development years.
It has been suggested that the longer the time spent on a mobile device, increases the negative impact on supervision, and decreases interaction with the child [24]. This finding was supported by our observations, for as the time on the mobile device increased, the supervision and caregiver-child play decreased, being replaced by no parent/carer supervision, and independent play by the child. Furthermore, when parents were on the mobile device for a full minute, the potential for injury increased, compared with those parent/carers not using their mobile device. The injury potential increased when the supervision was interrupted by scrolling and increased again when on a telephone call. There is recent evidence suggesting children's safety needs are compromised when parents use their mobile phones [34], and smartphone ownership by parents' may help explain the increase in young children presenting at ED with injuries [23,26]. Interviewees acknowledged the potential for interrupted supervision, especially if there was an opportunity to multi-task (e.g., telephone calls or check work emails). However, worthy of exploration is the notion that a parent/carer may take the child to the playground more frequently because they can use their smartphone to multi-task, e.g., access e-mails, and in turn the child may get injured more, not because the caregiver is distracted but simply because the child goes to the playground more [23]. Our findings support a recent review where parents reported the need for uninterrupted supervision of children in and around playgrounds, roads and waterways where there is the potential for childhood injury [25]. Further research to fully understand the level of device distraction, multi-tasking and childhood safety to provide guidelines for parents' use is a research priority.
This study found that the tasks (telephone call, scroll/type or camera) undertaken on the mobile device influenced supervision and the opportunity for the social interplay between parents and their children. Participants reported that scrolling/texting resulted in a break from visual supervision, however, using the camera function not only maintained supervision but was often associated with play interaction between parents and their child. Of interest, parent/caregivers did not connect camera use with being distracted, and this finding is consistent with another recent Australian study that found parents value mobile devices as a way to capture memorable moments [4]. Noteworthy, the longer the time spent on taking or making a telephone call was found to be most closely associated with no supervision and minimal play between the parent and the child. Parental behaviours have a critical role as a social determinant of child development. For example, positive reinforcement while on play equipment, displays of warmth and affection result in fewer child behaviour problems and positive peer relations that, in turn, enhance a child's health [12]. An investment in early child development as a determinant of health will translate to learning skills and increased well-being across the life course.
Using a social determinants lens, early childhood development opportunities are affected by various social and environmental factors, including relationships with parents and caregivers [12]. The playground, specifically controlled, gated playground environments purpose-built for those under 6 years of age, provide co-benefits for the child and parent/carer. There is an opportunity for children's involvement in independent, active and risky play [34][35][36]; and a time for parents to 'take a break' from constant supervision acknowledging that some minor injuries are an inevitable aspect of early childhood [37]. As such, factors including a child's age, physical ability and the playground design have the potential to influence the level and type of supervision, and the benefit of parents using their mobile device whilst at the playground. Furthermore, recent literature supporting risky play, i.e., a play that is thrilling and exciting but includes the possibility of physical injury, has been positively associated with physical and social health in children [36]. However, despite the differing designs of the four playgrounds in this study, ranging from a large gated setting with multiple play areas to a small single ungated play area, similar patterns of MDU and supervision were observed.
Further exploration of the mitigation of potential risks of MDU and the interplay of the benefits might be an area worthy of future research.
Social acceptability and parent role-modelling were identified as influences of parent MDU. Quick calls, text messages and work were deemed more acceptable than using the device for social media or entertainment. Observations of other adults MDU in the playground were reported to influence interviewee's attitudes to their own MDU and how this relates to 'good parenting' but also the importance of role-modelling non-intentional use [27] to their children and to other adults in the playground setting. These perspectives contrast to prior work where caregivers were less critical of others and reserved judgement about the appropriateness of others' MDU [27]. These findings may highlight a changing social climate that is becoming more critical of parent/caregiver MDU and may indicate a need for realistic population-wide strategies to raise awareness of the risks associated with the distraction caused by mobile phones.
Most interviewees indicated that they had experienced success in limiting their own MDU and offered a range of strategies to help minimise use when caring for children. These strategies included abstinence, achieved by leaving the device in a bag, at home or in the car, choosing a phone function mode such as a silent or camera, and synchronising MDU during times when they perceived children required less interaction or supervision. In addition, a less often mentioned but observed strategy was caregiver proximal supervision i.e., following the child around the playground and keeping close whilst using a mobile device. This behaviour enabled the caregiver to maintain 'Constant supervision' whilst engaging in MDU [20]; however, it did not support caregiver-child interaction and exploit the time for parents and children to interact in a play setting.
Strengths and Limitations
To our knowledge, this is the first research to examine an association between the characteristics of parent/caregiver MDU (duration and mode) and caregiver-child interaction, caregiver supervision and child injury potential, providing new evidence. A mixed-method approach provided a range of perspectives and strategies that parents use to limit their MDU around children. There were a number of limitations including: an over-representation of females; small sample size, however, this is consistent with other similar research [4,24,27]. The approximate age of the child in the parent/child dyad was not estimated (other than being under 5 years), so we did not have an approximated age break down to complete any additional analysis by age, and half of the playgrounds being located in higher SES, metropolitan areas.
Conclusions
The majority of caregivers reported that whilst MDU was convenient for communication, personal organisation and security, it was an unwanted distraction in the playground, where supervising and interacting with children should be the priority. However, the exception was the camera, which was highly valued by caregivers for making memories and also offered the most support for maintaining supervision and interaction through play. Mobile device use was found to contribute to reduced supervision of children, which is a risk factor for injury. This is an emerging area of injury prevention that indicates the need for broader strategies addressing the complex interplay that exists between the social determinants and the developmental younger years. Finally, the caregiver perspectives from this research are valuable for the development of realistic and effective strategies that support parents and caregivers to achieve their desired MDU. | 2020-12-17T06:16:33.755Z | 2020-12-01T00:00:00.000 | {
"year": 2020,
"sha1": "4f99a51aefe24996c3de61a204ac7851d28d1c75",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2227-9067/7/12/284/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "670cc166b1168bdf49f0503229545a8cf5038293",
"s2fieldsofstudy": [
"Sociology",
"Education"
],
"extfieldsofstudy": [
"Computer Science",
"Medicine"
]
} |
9305747 | pes2o/s2orc | v3-fos-license | Radiation therapy induces the DNA damage response in peripheral blood.
Stereotactic body radiation therapy (SBRT) is a radiotherapy modality that delivers highly conformal, ablative doses to a well-defined target. Here, using a semiquantitative multiplexed assay to analyze ATM and H2AX phosphorylation, we show that ATM kinase activity in peripheral blood mononuclear cells is induced following SBRT. This observation of a systemic ATM kinase-dependent DNA damage response in the peripheral blood is unprecedented and promotes the use of ATM serine-1981 phosphorylation as a predictive biomarker for DNA damaging modalities and ATM inhibitors.
INTRODUCTION
Over half of the 1,638,910 Americans diagnosed with cancer in 2012 received radiotherapy. Stereotactic body radiation therapy (SBRT) is a treatment modality that relies upon target localization and motion compensation to deliver a conformal, ablative dose of radiation. While external beam radiotherapy is typically delivered in 25-40, 1-2 Gy fractions, 5 days per week, SBRT is delivered in [1][2][3][4][5] Gy high dose fractions on nonconsecutive days. Although there is rapid dose fall-off from tumor into normal tissues, blood flow through the targeted volume is not restricted. We examined the induction of the DNA damage response in peripheral blood mononuclear cells (PBMCs) in blood drawn from patients before and after the first fraction of SBRT.
Since the majority of cancer therapies are DNA damaging, there is a pressing need for pharmacodynamic biomarkers of DNA damage. Histone H2AX serine-139 phosphorylation generates gamma-H2AX which was originally associated with the induction of DNA double strand breaks (DSBs) [1]. While it is now clear that gamma-H2AX is induced by agents that do not induce DSBs [2,3], gamma-H2AX remains a widely used biomarker of DNA damage [4]. ATM is a kinase that initiates DNA damage signaling following exposure to ionizing radiation (IR) and other agents that induce DSBs. ATM kinase activity is associated with ATM serine-1981 phosphorylation in cells [5]. ATM serine-1981 phosphorylation is induced following 5 cGy [6] and facile extraction makes this phosphoprotein an excellent candidate biomarker.
RESULTS AND DISCUSSION
To demonstrate the utility of ATM as a biomarker, we developed a semiquantitative multiplexed immunoblot to analyze ATM serine-1981 phosphorylation, total ATM protein, gamma-H2AX and H2AX. Human cancer cells www.impactjournals.com/oncotarget were irradiated and whole cell extracts were resolved by SDS-polyacrylamide gel electrophoresis. Separate PVDF membranes were co-immunoprobed with antiphosphoserine-1981 ATM and anti-pan-ATM antibodies as well as anti-gamma-H2AX and anti-H2AX antibodies. A dose-dependent induction of ATM serine-1981 phosphorylation was detected in the cancer cell lines HT29 (colorectal), H460 (lung), and MDA-MB-231 (breast), and this induction was greater than that of gamma-H2AX at each dose ( Figure 1A-C).
To determine whether ATM serine-1981 phosphorylation is induced in PBMCs ex vivo, human blood was drawn from three healthy volunteers and exposed to IR. A dose-dependent induction of ATM serine-1981 phosphorylation was detected in PBMCs isolated from irradiated blood and this induction was greater than that of gamma-H2AX at each dose ( Figure 1D-F). The dynamic range of ATM serine-1981 phosphorylation induction was 40-fold in PBMCs and 8-fold in cell lines suggesting that this response may be greater in cells that have not been cultured in vitro.
To test the hypothesis that ATM serine-1981 phosphorylation may be induced in the peripheral blood of SBRT patients, we analyzed ATM serine-1981 phosphorylation and gamma-H2AX in PBMCs isolated from blood drawn before and after the first fraction of SBRT.
SBRT induced ATM serine-1981 phosphorylation and gamma-H2AX 5.3-fold and 3.7-fold, respectively, in a patient with inoperable non-small cell lung cancer (NSCLC) (Figure 2A). The prescribed planning target volume (PTV) was 15.1 cm 3 with 18 Gy per fraction for a maximum of 20 Gy within the target volume and a beamon time of 164 seconds. The maximum and mean dose to the heart was 2.0 and 1.0 Gy, respectively. Assuming the heart contains 280 cm 3 of a total blood volume of 4,500 cm 3 , PBMCs within the heart of this patient received 6 cGy (280/4,500 x 1 Gy). This approximation is consistent with the 5 cGy that induces ATM serine-1981 phosphorylation in fibroblasts [6] and PBMCs ( Figure 1E). SBRT induced ATM serine-1981 phosphorylation and gamma-H2AX 4.1-fold and 1.4-fold, respectively, in a patient treated following pancreatectomy for pancreatic adenocarcinoma ( Figure 2B). The prescribed PTV was 17.5 cm 3 with 12.0 Gy per fraction for a maximum of 13.9 Gy within the target volume and a beam-on time of 904 seconds.
SBRT induced ATM serine-1981 phosphorylation 3.2-fold in a patient with metastatic renal cell carcinoma with bone metastases ( Figure 2C). No induction of gamma-H2AX was observed. The prescribed PTV was 126.3 cm 3 with 8.0 Gy per fraction for a maximum of 9.4 Gy within the target volume and a beam-on time of 171 seconds.
SBRT induced ATM serine-1981 phosphorylation and gamma-H2AX 1.2-fold and 2.8-fold, respectively, in a patient treated for adenocarcinoma of the gastroesophageal junction post esophagectomy ( Figure 2D). The prescribed PTV was 10.2 cm 3 with 8.0 Gy per fraction for a maximum of 9.0 Gy within the target volume and a beam-on time of 100 seconds.
To our knowledge this is the first documentation of ATM serine-1981 phosphorylation in patients following radiotherapy. ATM serine-1981 phosphorylation has unequaled sensitivity for the detection of DNA damage in cells cultured in vitro [5,6] and in human tissues [7]. The data presented here show that ATM serine-1981 phosphorylation has an even greater dynamic range in PBMCs irradiated ex vivo. ATM serine-1981 phosphorylation may be an excellent biomarker for exposure to IR and other DNA damaging agents in human patients and as such may have utility in predicting therapeutic response and as measure of radiation exposure in counter-terrorism efforts.
Finally, it has long been postulated that systemic ATM kinase inhibition may be an excellent means to increase the efficacy of targeted radiotherapy [8,9]. Our data show, for the first time, that even the most guided radiation therapy is associated with a systemic DNA damage response. Therefore, systemic ATM kinase inhibition may be associated with DNA damage injury in circulating PBMCs. In this regard the absence of detectable ATR expression in PBMCs [10] is interesting because it suggests that the biological responses to DNA damage in PBMCs may be strictly dependent on ATM and as such ATR kinase inhibition may be a better means to selectively increase the efficacy of targeted radiotherapy. Our observation of a systemic ATM kinase-dependent DNA damage response through the peripheral blood following SBRT is unparalleled and promotes the use of ATM serine-1981 phosphorylation in evaluating the biological effects of DNA damaging modalities and ATM inhibitors.
Blood Collection and PBMC Isolation
Human blood (~8.0 mL/tube) was drawn into Cell Preparation Tubes (CPT) (Becton Dickinson, Franklin Lakes, NJ) during a single venipuncture event in accordance with an Institutional Review Board (IRB) protocols (#PRO13020420, #PRO10020482, #PRO10050157, #PRO11010576). PBMCs were isolated by centrifugation at 1500 x g for 30 min at room temperature (RT). The PBMC layer was transferred to a new 15 mL polypropylene conical tube and washed with 3-4 volumes of Plasma-Lyte A USP. The cells were pelleted by centrifugation (430 x g for 10 min at RT). The PBMCs were resuspended in 3.0 mL of Plasma-Lyte A. Immediately, 20 microliters of the cell suspension was removed and used to estimate PBMC density with a Beckman Coulter Z1 particle counter before an additional wash of 10 mL in Plasma-Lyte A. After re-pelleting, aspiration and disposal of the supernatant, the PBMC density was adjusted to 3 x 10 6 cells/mL in Plasma-Lyte A. Single-use aliquots of 0.5 mL were then transferred into new 2.0 mL micro-tubes. Lastly, the PBMCs were recovered as cell pellets by centrifugation (10,000 x g for one min at RT), and flash-frozen in a dry ice/ethanol bath before storage at −70 °C until use. To preserve specimen integrity, all CPT tubes were processed within three hours of collection.
Whole Cell Extract Preparation
Frozen cell pellets were resuspended in DTTmodified Laemmli buffer (50 mM Tris-HCl (pH 6.8) containing 2% (w/v) SDS, 10% (v/v) glycerol, 200 mM DTT and 0.002% (w/v) bromophenol blue) freshly supplemented with protease and phosphatase inhibitor cocktails, vortexed and pulsed sonicated in an ice-cold cup-horn sonicator (Cole-Parmer, Vernon Hills, IL). Lysates were heated to 95 °C for 7 min, vortexed and chilled briefly on ice. After centrifugation at 16,000 x g for 5 min at RT, supernatants were transferred into new 1.5 mL micro-tubes.
Detailed Immunoblot Procedure
To avoid bias in protein levels due to possible plasma protein contamination, PBMC lysates were normalized by cell number [11]. Aliquots of 25 microliters (5 x 10 5 cells) were loaded per well onto 4-15% TGX Criterion™ 18-well precast gels in a Criterion™ vertical midi-format electrophoresis cell (Bio-Rad Laboratories, Hercules, CA). Proteins were fractionated under a constant 200 V applied for ~40 min. Resolved proteins were electrophoretically transferred onto 0.45 micrometer Immobilon-FL polyvinylidene difluoride (PVDF) transfer membranes (Millipore, Billerica, MA) with 100 V applied for 1.5 h in a Criterion blotter transfer tank (Bio-Rad Laboratories, Hercules, CA). Following electroelution transfer, membranes were rinsed with water and incubated in 0.5x Odyssey® blocking buffer (20 mL/membrane) for 1 h at RT with gentle agitation. The membranes were cut horizontally into three strips. The top membrane strip (above 75 kDa) was co-immunoprobed with an antiphosphoserine-1981 ATM and anti-pan-ATM antisera mixture diluted to 1:500 and 1:2500, respectively, in 0.5x Odyssey® blocking buffer containing 0.1% Tween 20 (OBB-T). The middle membrane (25-75 kDa) was immunoprobed with anti-GAPDH diluted 1:20,000 in OBB-T. Likewise, the bottom membrane strip (below 25 kDa) was simultaneously probed with a biotinylated monoclonal rabbit anti-gamma-H2AX antibody and a mouse monoclonal anti-H2AX antibody, each diluted to 1:1000 in OBB-T. Membranes were incubated overnight at 4 °C with gentle agitation. After removing the antisera, the membranes were rinsed with TBS, washed thrice with 50 mL of TBS-T, for periods of 10 min each. The ATM membrane was immunoprobed with highly cross-absorbed anti-IgG detection antibodies that were individually matched to the specific host species in which the primary ATM antibodies were generated. The individual detection antisera mixture antibodies, a goat anti-rabbit IgG and a goat anti-mouse IgG antibody were labeled with NIR fluorochrome dyes (800 and 700 nm) and diluted 1:4000 and 1:5000, respectively. The H2AX membrane was similarly probed except both antisera in the mix were diluted 1:5000 and in addition, the mixture included streptavidin conjugated to an 800 nm fluorophore diluted 1:10000. All detection reagents were prepared in OBB-T containing 0.02% (w/v) SDS. The strips were incubated in the dark for 1 h at RT with gentle agitation. After incubating, and washing thrice for 5 min each in 50 mL of TBS-T, to remove Tween-20, membranes were washed twice more for 10 min in 50 mL of TBS. All five washes were carried out undulating in the dark at RT. Before proceeding, membranes were dried at RT for at least 1 h in the dark.
Immunoblot Imaging and Signal Quantification
Dry membrane strips were visualized using an Odyssey® CLx infrared imaging system (model 9120). Signal quantification was performed utilizing the version 2.1 Image Studio software (LI-COR Biosciences-US, Lincoln, NE). Shapes were drawn around bands of interest, and the signal was calculated as the sum of the individual pixel intensity values (Total) for the selected shape minus the product of the average intensity values of the pixels in the background (Bkg) and the total number of pixels enclosed by the shape (Area). Thus, Signal = Total -(Bkg x Area). The selected background value applied for each individual shape was the median of three pixels above and below the shape. The ratio of the phosphorylated to panprotein signal was used to correct for loading variations between lanes. | 2017-04-29T02:36:57.659Z | 2013-06-26T00:00:00.000 | {
"year": 2013,
"sha1": "30c255f3448f3517475f469b9204934c0d8f8a92",
"oa_license": "CCBY",
"oa_url": "http://www.oncotarget.com/index.php?journal=oncotarget&op=download&page=article&path[]=1084&path[]=1373",
"oa_status": "HYBRID",
"pdf_src": "PubMedCentral",
"pdf_hash": "30c255f3448f3517475f469b9204934c0d8f8a92",
"s2fieldsofstudy": [
"Biology",
"Medicine"
],
"extfieldsofstudy": [
"Medicine",
"Biology"
]
} |
253038144 | pes2o/s2orc | v3-fos-license | Nanostructured Hybrids Based on Tantalum Bromide Octahedral Clusters and Graphene Oxide for Photocatalytic Hydrogen Evolution
The generation of hydrogen (H2) using sunlight has become an essential energy alternative for decarbonization. The need for functional nanohybrid materials based on photo- and electroactive materials and accessible raw materials is high in the field of solar fuels. To reach this goal, single-step synthesis of {Ta6Bri12}@GO (GO = graphene oxide) nanohybrids was developed by immobilization of [{Ta6Bri12}Bra2(H2O)a4]·4H2O (i = inner and a = apical positions of the Ta6 octahedron) on GO nanosheets by taking the advantage of the easy ligand exchange of the apical cluster ligands with the oxygen functionalities of GO. The nanohybrids were characterized by spectroscopic, analytical, and morphological techniques. The hybrid formation enhances the yield of photocatalytic H2 from water with respect to their precursors and this is without the presence of precious metals. This enhancement is attributed to the optimal cluster loading onto the GO support and the crucial role of GO in the electron transfer from Ta6 clusters into GO sheets, thus suppressing the charge recombination. In view of the simplicity and versatility of the designed photocatalytic system, octahedral tantalum clusters are promising candidates to develop new and environmentally friendly photocatalysts for H2 evolution.
Introduction
The generation of molecular hydrogen using sunlight is one of the most attractive energy production pathways for the future. Within the framework of sustainable chemistry, there is a special interest in the use of photo-and electroactive materials formed by abundant elements in the Earth's crust to obtain H 2 from sunlight and water, as inexhaustible sources of energy. In particular, the relative abundance of tantalum (1-2 ppm) has promoted the preparation of Ta-derived photocatalysts such as Ta 2 O 5 and tantalum (oxy)nitrides for water reduction and splitting under photocatalytic conditions [1]. Among these and other tantalum-derived materials, nanostructures (such as nanowires [2,3], nanorods [4], nanosheets [5,6], nanotubes [7,8], nanoplates [9], nanoflowers [10], and nanoparticles [11,12]) have emerged as materials with improved photo-and electrocatalytic performance in water splitting without the presence of co-catalysts. Low-nuclearity tantalum clusters appear as another kind of nanosized molecules, of which photocatalytic properties remain unexplored. Concretely, the [{Ta 6 X i 12 }L a 6 ] n+/− (X i = halogen; L a = halogen, O-donor ligand) clusters, which belong to the large family of metal atom clusters defined by Cotton [13], show interesting photo and redox properties. In their cores, 6 metal atoms of the cluster are arranged in a regular octahedral geometry, interconnected by direct metal-metal bonds, and further linked to 12 inner ligands (X i ) and 6 apical ligands (L a ) located at the edge-bridging and terminal positions of the tantalum cluster, respectively ( Figure 1). Various applications of the octahedral halide-bridged clusters fields of crystallography, medicine, optics, and catalysis have been pro {Ta6Br i 12} clusters are used as a commercial tool for phase determinati by X-ray crystallography and as a radiographic contrast agent [14,15]. halide clusters have been employed to build nanostructured hybrid glazing applications in energy-saving buildings [16][17][18][19]. Catalytic app (X i = Cl, Br) cluster compounds and derived composites, involving hyd ring closure reactions under high-temperature conditions, were rep chemistry of octahedral tantalum halide clusters not only encompasses polymeric matrices or other molecules for obtaining the respective hy pramolecular materials [22] but also involves ligand exchange reaction ical positions [23]. This combination of ligand reactivity and the robu core, which keeps the cluster integrity intact during these transformat search for a catalytic application of this cluster family.
The preparation of tantalum bromide and chloride [{Ta6X i 12}X a 2(H Br) aqua clusters was developed by Koknat et al. in the 1970s [24] and analog was reported [25]. Because of its stability and availability, the b Br) is the most studied of the three, and its preparation, properties, an solvents and oxidizing agents were recently thoroughly revisited [ 1980s, Vogler and Kunkely showed that the bromide cluster was redo tochemical reduction of water to molecular hydrogen and in the presen property makes this cluster compound a suitable photocatalyst in the water and sunlight. The easy ligand exchange in the [{Ta6Br i 12}(H2O) a 6 which is the dominant species in acidic solutions of [{Ta6Br i 12}Br a 2(H only involved in the mechanism of the H2O-to-H2 transformation, but coordinative cluster anchoring onto functionalized supports for the hybrid materials with enhanced activity in photocatalysis. Graphene-based photocatalysts have gained increasing interest tive to increase the yield of photocatalytic H2 production in solar en Various applications of the octahedral halide-bridged clusters of tantalum in the fields of crystallography, medicine, optics, and catalysis have been proposed. For instance, {Ta 6 Br i 12 } clusters are used as a commercial tool for phase determination of biomolecules by X-ray crystallography and as a radiographic contrast agent [14,15]. Recently, tantalum halide clusters have been employed to build nanostructured hybrid materials for solar glazing applications in energy-saving buildings [16][17][18][19]. Catalytic applications of {Ta 6 X i 12 } (X i = Cl, Br) cluster compounds and derived composites, involving hydrogen transfer and ring closure reactions under high-temperature conditions, were reported [20,21]. The chemistry of octahedral tantalum halide clusters not only encompasses their inclusion into polymeric matrices or other molecules for obtaining the respective hybrid [16][17][18][19] or supramolecular materials [22] but also involves ligand exchange reactions, mostly in the apical positions [23]. This combination of ligand reactivity and the robustness of the cluster core, which keeps the cluster integrity intact during these transformations, favors further search for a catalytic application of this cluster family.
The preparation of tantalum bromide and chloride [{Ta 6 [24] and recently the iodide analog was reported [25]. Because of its stability and availability, the bromide cluster (X = Br) is the most studied of the three, and its preparation, properties, and reactivity toward solvents and oxidizing agents were recently thoroughly revisited [26]. Already in the 1980s, Vogler and Kunkely showed that the bromide cluster was redox active in the photochemical reduction of water to molecular hydrogen and in the presence of HCl [27]. This property makes this cluster compound a suitable photocatalyst in the H 2 evolution from water and sunlight. The easy ligand exchange in the [{Ta 6 O, is not only involved in the mechanism of the H 2 O-to-H 2 transformation, but it is also useful for coordinative cluster anchoring onto functionalized supports for the preparation of new hybrid materials with enhanced activity in photocatalysis.
Graphene-based photocatalysts have gained increasing interest as a viable alternative to increase the yield of photocatalytic H 2 production in solar energy conversion in chemical energy [28][29][30][31]. The most common photocatalysts, such as TiO 2 , CdS, and BiVO 4 , show low photocatalytic activities due to the low surface area and high recombination rate of photo-induced charge carriers owing to their band energies. One strategy to enhance photocatalyst reactivity for visible-light water splitting is to blend it with graphene materials in order to obtain hybrid nanostructured materials. In this sense, GO is an excellent co-catalyst in photocatalytic applications due to its high surface area, excellent electronic mobility, and high adsorption ability associated with the oxygen functionalities [29]. Among photocatalytic applications of the GO-derived composites, photocatalytic degradation of pollutants and H 2 evolution from water has become the most relevant, but applications in the photoreduction of CO 2 , atmospheric ammonia synthesis, and photocorrossion suppression applications still need further exploration [31][32][33][34]. In mimicking the mechanism of Gratzel cells, which includes dyes as light harvesters capable of absorbing visible light and injecting electrons into the conduction band of a semiconductor material (such as TiO 2 ), it is highly desirable to develop hybrid cluster-GO photocatalysts without the need of precious metals. GO serves as a support of alternative metal clusters and nanoparticles with an absorption window from the UV to the visible region. Their immobilization avoids metal aggregation and suppresses the charge recombination by channeling electrons into the GO substrates. Recently, we immobilized molybdenum octahedral clusters, which are optimal photosensitizers and electron transfer catalysts, onto GO surfaces in order to obtain nanohybrids which resulted in efficient water reduction photocatalysts [35,36]. In this work, the [{Ta 6 Br i 12 }Br a 2 (H 2 O) a 4 ]·4H 2 O cluster was systematically immobilized on GO nanosheets by taking advantage of the labile nature of the apical ligands of the cluster and the coordinating ability of the oxygen functionalities of the GO support. The {Ta 6 Br i 12 }@GO hybrids developed in this work were fully characterized and used as photocatalysts for the reduction of water vapor to H 2 . Their catalytic ability in the H 2 evolution reaction (HER) was explored under light irradiation and the synergy between the cluster and the support and its recyclability was discussed.
Instrumentation
Combustion chemical analysis of the samples was carried out using a Fisons EA 1108-CHNS-O analyzer (ThermoFisher Scientific, Waltham, MA, USA). Inductively coupled plasma atomic emission spectrometry (ICP-AES) analyses for the determination of atomic tantalum of the solid materials were performed after aqua regia digestion at 180 • C for 24 h in reflux, and the resulting solutions were measured in a Varian 715 spectrometer (Palo Alto, CA, USA). Fourier transform infrared spectroscopy (FTIR) spectra were measured on KBr pellets with a Nicolet 8700 Thermo spectrometer (ThermoFisher Scientific, Waltham, MA, USA). Samples were grounded with dry KBr in agate mortar and pressed by vacuum. Raman spectra were acquired from solid materials on an aluminum sample holder and under atmosphere, using a Renishaw "Reflex" spectrometer (Wotton-under-Edge, U.K.) equipped with an Olympus optical microscope. The excitation wavelength was 514 and 785 nm generated by an Ar + ion laser. The laser power on the sample was 30 Mw, and a total of 10 to 40 acquisitions were taken. Spectra registered under N 2 atmosphere were performed in a sealed reactor equipped with a quartz window. The UV-vis-NIR diffuse reflectance (DRS) were collected in the range of 200 to 2000 nm with a Varian Cary 5000 spectrophotometer. UV-vis spectrophotometric analysis for samples in solution was performed using a Varian Cary 50 UV-vis Agilent analyzer equipped with a Xe lamp as the light source and a Czerny-Turner model dual beam monochromator with 10 × 10 mm quartz cuvettes. The powder X-ray diffraction (PXRD) data of the K 4 [{Ta 6 lytical Cubix-Pro diffractometer equipped with a PANalytical X'Celerator detector. This equipment employed monochromatic CuKα X-ray radiation (L1 = 1.5406 Å, L2 = 1.5444 Å, I2/I = 0.5) and a tube voltage and intensity of 45 kV and 40 mA, respectively. It uses a variable slit with an irradiated sample area of 5 mm and the goniometer arm length is 200 mm. The diffractogram of the powder samples was obtained at room temperature in 2θ range of 2-90 • . The morphology and composition of the materials were characterized by scanning electron microscope (SEM) using a ZEISS model ULTRA55 FESEM coupled to an Oxford Instruments energy dispersive X-ray (EDS) detector. Irradiation experiments were performed with a spotlight Lightnincure LC8 model, 800-200 nm, 150 W, equipped with a fiber optic light guide with a spot size of 1.0 cm diameter. Molecular hydrogen production was monitored by gas chromatography (GC) on the Agilent 490 Micro GC System, equipped with a column coated with a zeolite molecular sieve (CP-Molsieve 5Å, Agilent J&W) and a conductivity detector (TCD). Ar was taken as the carrier gas and the flow rate was set to 5 mL min −1 . The inlet and detector temperatures in the GC run were 110 • C and 220 • C, respectively, and the isothermal oven temperature profile was set at 62 • C with an initial column pressure of 15 psi.
Synthesis and Characterization of Materials
The preparation of K 4 [{Ta 6 Br i 12 }Br a 6 ] was performed as described by Messerle et al. with minor modifications [39]. A mixture of TaBr 5 (6.00 g, 10.32 mmol), KBr (1.912 g, 15.92 mmol), and Ga granulate (0.64 g, 9.12 mmol) was sealed in an evacuated quartz ampule and placed into a 270 • C preheated furnace for 45 min. During this step, the ampule was taken out every 15 min. and shaken vigorously (hand protection is necessary to prevent thermal burns)-three times in total-in order to evenly spread the molten gallium. Then, the ampule was heated to 300 • C and held at this temperature for 12 h. After this procedure, the ampule was shaken to homogenize the reactants and returned into the furnace chamber. The temperature was raised to 400 • C and held for 24 h. The ampoule was removed, allowed to cool to ambient temperature, and the product was extracted to obtain [{Ta 6 6 Br i 12 }Br a 6 ] was added to 3 mL of Milli-Q water and the solid material was ground in a mortar for 15 min until a homogeneous dark green mixture was obtained. The solid was extracted by filtration through a medium porosity fritted glass funnel by adding 40 mL of water in order to remove the residual GaO(OH) from the K 4 [{Ta 6 Br i 12 }Br a 6 ] precursor and other impurities in the material. In order to prevent cluster oxidation under standard environmental conditions, the dark green filtrate was treated with a small portion of SnBr 2 (ca. 5 mg) followed by HBr (50 mL). The solution was kept at 4 • C, and after 20 h a dark green solid precipitated. The translucent supernatant was carefully removed by decantation, the solid was washed with diethyl ether and dried by rotatory evaporation at 60 • C for 3 h to give 355 mg of a green microcrystalline material identified as [{Ta 6 (40 mg in 100 mL of THF), which was previously sonicated during 1 h in a round bottom flask. The mixture was magnetically stirred at 40 • C for 2 h or 16 h. Next, for each reaction, the solid was separated by filtration under vacuum and it was washed with THF under an inert atmosphere. The resulting products were dried under vacuum to provide black solids labeled as {Ta 6 Br i 12 }@GO-20S (42 mg) and {Ta 6 Br i 12 }@GO-20L (44 mg), "20" referring to the percentage in weight (w/w) of cluster with respect to GO, and "S" and "L"-to the shortest (2 h) and longest reaction times (16 h), respectively. This synthetic procedure was extended to {Ta 6 Br i 12 }@GO-5L (5% w/w) and {Ta 6 Br i 12 }@GO-40L (40% w/w), and the amount of the solid obtained was 41 mg and 38 mg, respectively. These materials were stored under N 2 atmosphere in a MBraun dry box, and characterized by UV-vis-NIR DRS, FTIR, Raman, DRX, and SEM-EDS techniques. The amount of Ta was determined by ICP analyses. Elemental analyses provided the C and H content (wt %) of each sample, namely: {Ta 6 3 µmol) and phenol (0.3 g, 3.2 mmol) in THF (4 mL) in a round bottom flask; (ii) the cluster/benzoic acid solution, which was prepared following the same methodology, but using 3 mg (1.32 µmol) of cluster precursor, 3 mg (24.5 µmol) of benzoic acid, and THF (4 mL); and (iii) the control solution of the cluster dissolved in THF. All solutions were kept at 40 • C with constant stirring for 24 h. All the manipulations were performed under argon and using Schlenk techniques. An aliquot of the resulting mixtures was diluted and placed in a spectroscopic cuvette at the end of the reaction.
Photocatalytic H 2 Evolution Procedure
The photocatalytic reactions were carried out in the presence of aqueous mixtures in the vapor phase and under deaerated conditions. The photoreactor was a double cylindrical quartz reactor (110 mL of total volume) in which the two vessels (1 and 2) were connected with a quartz bridge (2 cm length). Figure S1 illustrates the experimental layout for vapor water photoreduction. Water (30 mL MilliQ water), HBr 0.7 M, and the sacrificial electron donor (methanol, 19.85% v/v) were loaded into the reactor vessel (1) and purged with Ar (30 min). The photocatalysts (5 mg of {Ta 6 Br i 12 }@GO or [{Ta 6 Br i 12 }Br a 2 (H 2 O) a 4 ]·4H 2 O materials) were dispersed in methanol (0.2 mL), and carefully deposited in the reactor vessel (2) by drop casting with simultaneous slow evaporation under a current of Ar, until a thin film was obtained. The reactor was sealed and pressurized with argon up to 0.3-0.5 bar and connected to an electrical heating ribbon that allowed heating of the reactor vessel (1) to 80 • C in order to achieve the evaporation of the water/sacrificial mixture. The vessel (1) was irradiated during 24 h (standard irradiation time) with a Hamamatzu Xe lamp with a spotlight placed at a distance of 5 cm above the reactor surface. The gas phase samples (500 µL) were collected with a Hamilton syringe and injected into the GC-TCD spectrometer. The quantification of the amount of H 2 was determined after 24 h of irradiation. The molecular hydrogen peak area was calculated to the corresponding concentration using the standard calibration curve as reference. The micromoles of H 2 produced were calculated taking into account the ideal gas law (n = PV/RT). Control experiments were done for the {Ta 6 Br i 12 }@GO-20L material (Table S1) and confirmed the lowest amounts of gas production obtained. All the experiments showed the detection of H 2 and the atmospheric gases, exclusively.
Reuse experiments were carried out for three cycles under the same conditions of the above experiments. The aqueous mixture at the end of the reaction was first removed, whereas the catalytic reactor with the cluster-based materials was purged under argon stream for 1 h. After, a fresh solution of H 2 O/MeOH/HBr was put on the reactor, and a second purge step was carried out in order to deoxygenate the system. This procedure was repeated for each reuse cycle. The percentage of the amount of H 2 produced was calculated with respect to the value of hydrogen obtained in the first use.
Results and Discussion
In this research, the {Ta 6 analytical, and spectroscopic techniques, and were applied as catalysts in the HER from water and light. The following two subsections encompass the results and discussion derived from this investigation.
Preparation and Characterization of Tantalum Materials
The surface of GO has different oxygen donor functionalities, mainly represented by hydroxyl, carboxyl, and epoxy groups, that are suitable for coordinative metal cluster anchoring [35,36] O with GO in hot THF, but cluster decomposition was observed. This was confirmed by SEM-EDS analyses of the solid, and through UV-vis analysis of the colored filtrate obtained after separation of the solid material. The EDS analysis showed a disproportion in the bromine (>0.4 wt %) and tantalum (ca. 0.8 wt %) content, which agrees with the decomposition of the cluster after reaction with GO in air. UV-vis analyses in milli-Q water solutions showed the disappearance of the band at 645 nm, a shift of the band at 750 nm to 715 nm, and a new band around 870 nm, which indicate full oxidation of the {Ta 6 Br i 12 } 2+ to {Ta 6 Br i 12 } 3+ species [26,27]. These results agree with the reported air-promoted cluster oxidation and decomposition towards Ta 2 O 5 described for {M 6 X i 12 } 2+ (M = Ta, Nb; X i = halogen) cluster species [40,41].
The synthesis of the {Ta 6 Br i 12 }@GO nanohybrids was then performed under an inert atmosphere. Different tantalum cluster loadings were used in order to obtain the nanohybrids, namely: {Ta 6 [{Ta6Br i 12}Br a 2(H2O) a 4]·4H2O, for which the diffraction patterns are indistinguishable in terms of the intensity peaks and 2θ values [26]. This indicates that the [{Ta6Br i 12}Br a 2(H2O) a 4]·4H2O cluster material is supported onto GO sheets in its crystalline form. The morphology of GO and chemical composition of the resulting {Ta6Br i 12}@GO nanohybrids were examined by SEM and EDS techniques, respectively. The flaky texture of the GO (Figure 3a) reflects its characteristic layered microstructure and the larger interspaces of the layer and the thinner layer edges of GO can be clearly appreciated [43]. Regarding the {Ta6Br i 12}@GO materials (Figure 3b-d), the {Ta6Br i 12} 2+ cluster crystals appear as small particles of lighter color with characteristic hexagonal shapes ( Figure S2) that are homogeneously distributed onto the graphenic surface. It should be noted that, in the case of the materials with a higher content of the cluster, the crystals form aggregates as nanocrystals with a larger size. The morphology of GO and chemical composition of the resulting {Ta 6 Br i 12 }@GO nanohybrids were examined by SEM and EDS techniques, respectively. The flaky texture of the GO (Figure 3a) reflects its characteristic layered microstructure and the larger interspaces of the layer and the thinner layer edges of GO can be clearly appreciated [43]. Regarding the {Ta 6 Br i 12 }@GO materials (Figure 3b-d), the {Ta 6 Br i 12 } 2+ cluster crystals appear as small particles of lighter color with characteristic hexagonal shapes ( Figure S2) that are homogeneously distributed onto the graphenic surface. It should be noted that, in the case of the materials with a higher content of the cluster, the crystals form aggregates as nanocrystals with a larger size. In order to find more evidence for the presence of the {Ta6Br i 12} cluster units in the {Ta6Br i 12}@GO nanohybrids, several EDS spectra were recorded and analyzed in different regions of the sample ( Figure S3). The results revealed that the material was composed uniquely of carbon, oxygen, bromine, and tantalum. Tantalum was exclusively detected In order to find more evidence for the presence of the {Ta 6 Br i 12 } cluster units in the {Ta 6 Br i 12 }@GO nanohybrids, several EDS spectra were recorded and analyzed in different regions of the sample ( Figure S3). The results revealed that the material was composed uniquely of carbon, oxygen, bromine, and tantalum. Tantalum was exclusively detected in the brightest zones, and the averaged Ta/Br ratio found corresponds to 0.48, which matches the expected value (0.43) for the Ta Figure 4. The spectra of the graphenic materials show a wide band centered approximately at 3430 cm −1 that is associated with the stretching modes of the hydroxyl groups [44] that remain unaltered after cluster immobilization. The spectrum of a crystalline sample of [{Ta 6 Br i 12 }Br a 2 (H 2 O) a 4 ]·4H 2 O shows water libration bands at 3040 cm −1 and 3243 cm −1 , and appears in the fingerprint of the nanohybrids after cluster immobilization. Apart from the signal at 1627 cm −1 assigned to the bending vibration of water molecules, the spectrum of the pure tantalum cluster material does not show additional signals in the range of 500-2700 cm −1 . Note that the characteristic {Ta 6 Br i 12 } 2+ cluster core IR vibration bands at lower frequencies are not detectable within the registered spectral window [45].
The FTIR identification of {Ta 6 Br i 12 }@GO shows the two strong characteristic C=O vibration bands at 1726 and 1584 cm −1 , and a shoulder at 1627 cm −1 . The first band is shifted with respect to that of the GO and is associated with the carbonyl group vibration of the GO material, whereas the shoulder at 1627 cm −1 , which corresponds to carboxylic/adsorbed water vibration bands, remains unaltered. A new and intense band appears at 1584 cm −1 after cluster immobilization. This band was also detectable in the spectrum of the (TBA) 2 Mo 6 I i 8 @GO nanohybrid and was attributed to the interaction of the hexametallic Mo 6 I i 8 cluster units with the carboxylate functionalities of the graphene support. This band confirms, for the first time, the interaction of the {Ta 6 Br i 12 } cluster units with carboxylate funtionalities. In the {Ta 6 Br i 12 }@GO spectra, the hydroxyl vibration region characteristic of the GO appear distorted: the band at 1210 cm −1 , which is characteristic of the C-OH graphene vibrations of GO, shifts to longer wavelenghts (1234-1244 cm −1 ) and increases in intensity. This peak is associated to a new Ta-O-C vibration of the coordinated carboxylate and alcoxo/alcohol groups. This is not surprising because octahedral tantalum clusters readily and rather selectively exchange apical ligands with alcohols [46,47] and alcoxo ligands [48,49], being the alcoxo cluster species commonly bonded to the oxidized {Ta 6 [50,51], which is present as an impurity in the [{Ta 6 Br i 12 }Br a 2 (H 2 O) a 4 ]·4H 2 O precursor (1% w/w determined in aqueous solution by UV-vis spectroscopy). This band can be detected after cluster immobilization, but only in the spectra from the samples with higher tantalum loadings. The low intensity of this signal suggests that cluster degradation was efficiently prevented during and after the cluster immobilization reaction. These results unequivocally confirm that cluster immobilization occurred by partial ligand exchange between the aqua ligands and the carboxylic and hydroxyl groups of the GO surface to afford the coordinative anchoring of the {Ta 6 Br i 12 } cluster units.
immobilization reaction. These results unequivocally confirm that cluster immobili occurred by partial ligand exchange between the aqua ligands and the carboxyli hydroxyl groups of the GO surface to afford the coordinative anchoring of the {Ta cluster units. The {Ta6Br i 12}@GO hybrids and the crystalline precursor were also characteriz Raman spectroscopy. The most representative spectrum was measured under 785 n {Ta6Br i 12}@GO-20L (Figure S4a,b). The most intense bands are centered at 1342 cm 1598 cm −1 (G band), which correspond to the characteristic D and G graphenic b respectively [42]. Whereas the [{Ta6Br i 12}Br a 2(H2O) a 4]·4H2O compound is highly under 785 nm excitation at low Raman shifts, the nanohybrid show only two centered at 140 cm −1 and 153 cm −1 (Figure S4b), which are assignable to vibratio {Ta6Br i 12} tantalum cluster units [45]. This behavior suggests that the GO str influences the electronic properties of the cluster thus indirectly indicatin coordinative nature of the cluster immobilization. This is also evident from R characterization of both materials under 514 nm: the nanohybrid shows a wide and The {Ta 6 Br i 12 }@GO hybrids and the crystalline precursor were also characterized by Raman spectroscopy. The most representative spectrum was measured under 785 nm for {Ta 6 Br i 12 }@GO-20L ( Figure S4a,b). The most intense bands are centered at 1342 cm −1 and 1598 cm −1 (G band), which correspond to the characteristic D and G graphenic bands, respectively [42]. Whereas the [{Ta 6 Br i 12 }Br a 2 (H 2 O) a 4 ]·4H 2 O compound is highly active under 785 nm excitation at low Raman shifts, the nanohybrid show only two bands centered at 140 cm −1 and 153 cm −1 (Figure S4b), which are assignable to vibrations of {Ta 6 Br i 12 } tantalum cluster units [45]. This behavior suggests that the GO strongly influences the electronic properties of the cluster thus indirectly indicating the coordinative nature of the cluster immobilization. This is also evident from Raman characterization of both materials under 514 nm: the nanohybrid shows a wide and low-intensity band at low Raman shifts ( Figure S4c), while the spectrum of the crystalline compound shows the characteristic pattern of the {Ta 6 Br i 12 } cluster compounds, with the most intense cluster bands at 122 cm −1 , 172 cm −1 , and 233 cm −1 [26], and with no additional bands at higher Raman frequencies. Similar Raman spectra were registered under N 2 atmosphere, therefore, cluster decomposition under air was discarded.
The UV-vis-NIR DRS identification of the {Ta 6 Br i 12 }@GO materials also confirmed the presence of the metal cluster units. Figure 5 illustrates a comparison between the absorption spectra of the nanohybrids, [{Ta 6 Br i 12 }Br a 2 (H 2 O) a 4 ]·4H 2 O, and GO materials. The spectra of {Ta 6 Br i 12 }@GO materials show a good correspondence with the most energetic absorption bands of the molecular material. In general, these bands are comparable to those assigned in the DRS and UV-vis spectra reported for {Ta 6 Br i 12 } cluster compounds and derived composites [16][17][18][19]. The cluster bands in the Vis-NIR region are most sensitive in the samples with the highest loadings of cluster material. The spectra of {Ta 6 Br i 12 }@GO-20S and 20 L materials show signals at 630 nm, 735 nm, and the most intense band at 890 nm. These bands are not detectable in the graphenic support and differ in shape and wavelength with respect to the molecular material. The bands detected at lower wavelengths (750 nm and 641 nm) for the crystalline [{Ta 6 Br i 12 }Br a 2 (H 2 O) a 4 ]·4H 2 O cluster material correspond to the local excitations within the cluster core of the tantalum halogenated aqua clusters [25,26], and show a hypsochromic shift in the spectra of the nanohybrids due to the interaction of the {Ta 6 Br i 12 } 2+ cluster core with the oxygen functionalities of the GO. A band at 850 nm in the spectrum of [{Ta 6 Br i 12 }Br a 2 (H 2 O) a 4 ]·4H 2 O is associated with the effect of high crystallinity of the material under DRS acquisition conditions. The appearance of the intense band at 890 nm in the spectra of the nanohybrids, and the increase in the intensity of the band at 735 nm (relative to the above-mentioned cluster bands) with the reaction time, suggest the presence and accumulation of the oxidized {Ta 6 Br i 12 } 3+ cluster species [16][17][18][19]. This is not unexpected, since previously reported 3+ and 4+ species coordinated to alcohol and alcoxo ligands have been isolated in ligand exchange reactions starting from the 2+ clusters in air [46][47][48][49]. In our case, the partial Ta 6 cluster oxidation could be due to a redox reaction between the cluster and the oxygen functionalities on the GO surface during the immobilization process. Reaction time supports this hypothesis: the band at 890 nm appears at a lower relative intensity in {Ta 6 Br i 12 }@GO-20S with respect to {Ta 6 Br i 12 }@GO-20L spectra. In addition, it is known that the cluster oxidation is favored in solution in the presence of ketones under ambient conditions [16][17][18][19]26], and keto groups also form part of the graphenic support. Thus, in absence of O 2 , the presence of oxidized {Ta 6 Br i 12 } 3+ cluster units is to be ascribed to the reduction in some of the graphenic functionalities by the cluster. Analyses of the materials from carbon suggest a partial reduction in the graphenic support after the immobilization process. The carbon content for the hybrid materials is ca. 12 % higher than that expected from pure GO and cluster mixtures in their corresponding ratios. This is in agreement with the ability of the {Ta 6 Br i 12 } 2+ cluster species to be oxidized to the 3+ and 4+ species, and of GO to be reduced towards graphene via chemical or electrochemical means [52,53]. Alternatively, the reaction between [{Ta 6 Br i 12 }Br a 2 (H 2 O) a 4 ]·4H 2 O and carboxylic and hydroxyl functionalities was assessed by UV-vis spectroscopy and in homogeneous conditions by using benzoic acid and phenol as respective models. UV-vis spectra show that after 24 h of reaction, no spectral changes of the cluster precursor material were detected (Figures S5 and S6). These results evidenced that the detection of {Ta 6 Br i 12 } 3+ cluster species is exclusively ascribable to the redox coupling between O-functionalities supported onto GO sheets.
Photocatalytic Performance of Tantalum Nanomaterials
Gas phase photocatalytic water transformations constitute a promising alternative to more common liquid phase methodologies because soft conditions used in this case assure stability and recovery of the catalyst [35,36,54,55]. This also sets up a sustainable pathway
Photocatalytic Performance of Tantalum Nanomaterials
Gas phase photocatalytic water transformations constitute a promising alternative to more common liquid phase methodologies because soft conditions used in this case assure stability and recovery of the catalyst [35,36,54,55]. This also sets up a sustainable pathway to hydrogen production via electrolysis of moisture from air [56,57]. In our research, we investigated the catalytic performances of the {Ta 6 Br i 12 }@GO nanohybrids and the [{Ta 6 Br i 12 }Br a 2 (H 2 O) a 4 ] and GO nanomaterials towards water reduction of H 2 in the presence of vapor water mixtures and under photochemical conditions. Water was mixed with hydrobromic acid as the proton source and methanol as the sacrificial electron donor.
After 24 h of irradiation, all the materials selectively produced H 2 but different activities were found (Figure 6a). Crystalline [{Ta 6 Br i 12 }Br a 2 (H 2 O) a 4 ]·4H 2 O showed higher catalytic efficiency than unmodified GO. The performance of the crystalline cluster material, however, is low, due to its low surface area which prevents interaction between gas molecules and cluster sites. In contrast, the cluster immobilization onto GO sheets allows reactive species such as water, methanol, and H + to be adsorbed on the more exposed surface of the material, and to promote the reaction, giving superior H 2 yields. The {Ta 6 Br i 12 }@GO-20L nanomaterial provided the highest hydrogen production (84 µmol/g cluster ) among all the nanohybrids and controls tested. This amount of H 2 is superior to the sum of the yields achieved on the individual counterparts, which proves both the synergetic effect and the true hybrid nature of this nanomaterial. In fact, the activity of the cluster sites of {Ta 6 Br i 12 }@GO-20L and the molecular materials is 706 µmol·g −1 vs. 101 µmol·g −1 atomic tantalum (Figure 6b). This indicates that the performance of {Ta 6 Br i 12 }@GO-20L is seven-fold higher than for the microcrystalline [{Ta 6 Br i 12 }Br a 2 (H 2 O) a 4 ]·4H 2 O, which demonstrates the enhancement of the catalytic efficiency of the cluster site supported onto the graphenic surface. The activities found for the other nanohybrids are lower, with 32 µmol/gcluster, 55 µmol/gcluster, and 67 µmol/gcluster for {Ta6Br i 12}@GO-20S, {Ta6Br i 12}@GO-5L, and {Ta6Br i 12}@GO-40L, respectively. The {Ta6Br i 12}@GO-20L provides the highest H2 production, while the {Ta6Br i 12}@GO-5L and -20S materials, with the lowest tantalum loadings, are catalytically less active. Interestingly, the amount of supported cluster and the catalytic yields are not always proportional and the activity diminishes with the highest (ca. 40% w/w) tantalum loadings. This may be due to the too-large size of the crystallites of the tantalum cluster in {Ta6Br i 12}@GO-40L, or to an agglomeration of the same, which limits the occurrence of successful charge transfer between the cluster and the GO, thus decreasing the effectiveness of the photocatalytic process.
It is worth stressing that these catalytic yields are superior to those reported for similar metal cluster-anchored GO nanocomposites, that is, for (TBA)2Mo6I i 8@GO (30 µmol·g −1 ) [35] and (TBA)2Mo6Br i 8@GO (<60 µmol·g −1 ) [36] in vapor or liquid phases, respectively, and with the same sacrificial reductant, whereas the reaction rate is in the i The activities found for the other nanohybrids are lower, with 32 µmol/g cluster , 55 µmol/g cluster , and 67 µmol/g cluster for {Ta 6 6 Br i 12 }@GO-5L and -20S materials, with the lowest tantalum loadings, are catalytically less active. Interestingly, the amount of supported cluster and the catalytic yields are not always proportional and the activity diminishes with the highest (ca. 40% w/w) tantalum loadings. This may be due to the too-large size of the crystallites of the tantalum cluster in {Ta 6 Br i 12 }@GO-40L, or to an agglomeration of the same, which limits the occurrence of successful charge transfer between the cluster and the GO, thus decreasing the effectiveness of the photocatalytic process.
It is worth stressing that these catalytic yields are superior to those reported for similar metal cluster-anchored GO nanocomposites, that is, for (TBA) 2 Mo 6 I i 8 @GO (30 µmol·g −1 ) [35] and (TBA) 2 Mo 6 Br i 8 @GO (<60 µmol·g −1 ) [36] in vapor or liquid phases, respectively, and with the same sacrificial reductant, whereas the reaction rate is in the same order of magnitude (Table S2). The optimal values obtained with {Ta 6 Br i 12 }@GO nanocomposites also exceed the best value, achieved with [{Ta 6 Br i 12 }Br a 2 (H 2 O) a 4 ]·4H 2 O (45 µmol·g −1 ) in aqueous HCl under non-catalytic conditions [27]. It is worth stressing that the highest catalytic activity found in this work (4 µmol·g −1 ·h −1 ) is of the same order of magnitude as that reported in the literature for other tantalum photocatalysts, such as MTaO 3 (M = Li, Na, Mg) and MTa 2 O 6 (M = Mg, Ba) tantalates [58], used in the heterogeneous system and in the liquid phase in water/methanol mixtures without co-catalysts (Table S2). To the best of our knowledge, there is no previous work associated with the photocatalytic performance of tantalum-GO composites towards water reduction.
A plausible photosensitation mechanism can be proposed considering the low optical band gap (1.44 eV) of the {Ta 6 Br i 12 } 2+ cluster associated to its good visible-NIR cluster absorbance, which is responsible for efficient utilization of solar energy. The energy positioning of the LUMO orbitals (−3.48 eV) of the cluster complex [59] guarantees the catalytic role of the cluster unit for water reduction reaction from the thermodynamic point of view. The role of GO as co-catalyst is based on its semiconductor nature, due to the presence of conductive sp 2 and nonconductive sp 3 carbon domains on the GO surface [60][61][62]. The band gap of GO (2.9-3.7 eV) depends on the coverage, arrangement, and relative ratio of the epoxy and hydroxyl groups [28]. Even if some partial reduction in GO occurs, the positioning of the conduction band (CB) of GO is practically independent of the degree of oxidation (from −0.52 and −0.75 eV vs Ag/AgCl) [61,63]. The ability to separate the photogenerated electron-hole pair, superior electron mobility, high surface area, and easy surface chemical modification make GO and GO-based materials excellent photocatalysts for water reduction under UV and visible irradiation [29,64]. Taking into consideration these characteristics, the coordinative anchoring of the Ta 6 Br i 12 cluster units onto GO sheets is expected to improve the performance of both components in visible light region, as we have proved experimentally. Upon excitation by light, the HOMO-LUMO transition of the cluster compound reaches the excited state as the first step of cluster activation [27]. The {Ta 6 Br i 12 } 2+ /{Ta 6 Br i 12 } 3+ system acts as an electron shuttle and, in the following step, injects two electrons into the CB of GO, which are invested in the reduction of water to provide H 2 (Figure 7). Although we have not yet conducted mechanistic studies, the electron transfer from methanol into the valence band of GO would be feasible and provide necessary electrons to reduce the oxidized Ta 6 cluster thus closing the catalytic cycle. The role of GO, therefore, is to enhance the efficiency of the electron transfer from cluster sites to water molecules and to prevent electron recombination. This behavior is associated with the coordinative binding of the cluster sites and the high exposition of cluster sites anchored onto GO sheets. The improvement of the catalytic performance in the nanohybrids, in comparison with the tantalum cluster and GO separately, is a consequence of the high charge mobility and surface area of GO, which are responsible of the synergetic behavior between the cluster and GO.
The recyclability of {Ta 6 Br i 12 }@GO-20L was tested in three consecutive runs under the same reaction conditions and the results are represented in Figure 8. The material proved to be stable after three reuse cycles, and the H 2 production decreased slightly in the first recycling cycle (ca. 10%), but in the following two reuse cycles it seemed to remain stable with minimum variation, demonstrating the recyclability of the material.
recombination. This behavior is associated with the coordinative binding of the sites and the high exposition of cluster sites anchored onto GO sheets. The impro of the catalytic performance in the nanohybrids, in comparison with the tantalum and GO separately, is a consequence of the high charge mobility and surface area which are responsible of the synergetic behavior between the cluster and GO. The recyclability of {Ta6Br i 12}@GO-20L was tested in three consecutive runs un same reaction conditions and the results are represented in Figure 8. The material to be stable after three reuse cycles, and the H2 production decreased slightly in t recycling cycle (ca. 10%), but in the following two reuse cycles it seemed to remain with minimum variation, demonstrating the recyclability of the material. The recyclability of {Ta6Br i 12}@GO-20L was tested in three consecutive runs un same reaction conditions and the results are represented in Figure 8. The material to be stable after three reuse cycles, and the H2 production decreased slightly in recycling cycle (ca. 10%), but in the following two reuse cycles it seemed to remai with minimum variation, demonstrating the recyclability of the material.
Conclusions
Nanostructured {Ta 6 Br i 12 }@GO hybrids were synthesized in a single step from GO and [{Ta 6 Br i 12 }Br a 2 (H 2 O) a 4 ]·4H 2 O precursors in THF under inert conditions. The immobilization was achieved with different cluster:GO ratios and reaction times, providing four different nanohybrids, with materials obtained with longer reaction times being more effective. The integrity of the cluster units was preserved after immobilization, as proven by ICP-AES and EDS. The PXRD and SEM identifications of the nanohybrids show high crystallinity of [{Ta 6 Br i 12 }Br a 2 (H 2 O) a 4 ]·4H 2 O phases immobilized onto GO layers. Cluster material was anchored through carboxyl and hydroxyl functionalities as evidenced by FTIR and Raman analyses. UV-vis-NIR identification of the samples showed both the presence of {Ta 6 Br i 12 } 2+ and {Ta 6 Br i 12 } 3+ cluster units, with the latter ascribable to a partial oxidation of the material associated with a reduction in the functionalities of GO during cluster immobilization and to the coordinative coupling between inorganic-organic materials.
Selective photoreduction of water into H 2 was achieved with {Ta 6 Br i 12 }@GO hybrids as catalysts. All the nanohybrids exhibited higher reaction yields than their precursors, which agrees with a synergetic effect between the cluster and graphenic counterparts. The enhancement of the photocatalytic behavior of the {Ta 6 Br i 12 } units and GO is due to the coordinative immobilization of the cluster onto the GO surfaces, which promotes the electron transfer from the photoexcited cluster to the GO. The best catalytic performance corresponds to the hybrid with 20% (w/w) cluster tantalum loading, and its efficiency is comparable to similar molybdenum cluster GO nanocomposites. The recyclability and the high output of this material were proven for three consecutive runs. The simplicity of the preparation of the nanohybrids and the photocatalytic design, the stability and recovery of the catalyst, and the efficient recycling make the developed methodology superior and more advantageous for converting water into H 2 .
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/nano12203647/s1, Figure S1: Experimental setup for photocatalytic hydrogen production in vapor phase conditions; Table S1: Amount of H 2 produced in control tests conducted under standard photocatalytic conditions; Figure S2: SEM image of [{Ta 6 Br i 12 }Br a 2 (H 2 O) a 4 ]·4H 2 O; Figure S3: EDS analysis of {Ta 6 Br i 12 }@GO-20L showing peaks of the mass percentage of bromine and tantalum. The Cu signal appears to not be significant because the calculated concentration of this metal was below the quantification limit of the technique (<0.05% w/w); Figure S4: Raman spectra of {Ta 6 (a,b), and at 514 nm (c); Figure S5: UV-vis spectra of the (i) and (ii) solutions; Figure S6: UV-vis spectra of the (iii) solution; Table S2: Optimal catalytic activities of {Ta 6 Br i 12 }@GO-20L compared with octahedral molybdenum cluster-based photocatalysts and selected tantalum photocatalysts for H 2 production from water; Refs. [12,35,36,58,[65][66][67] are cited in supplementary materials. | 2022-10-21T15:04:26.881Z | 2022-10-01T00:00:00.000 | {
"year": 2022,
"sha1": "b11ce447489c36b8f6e2bb7e7c88651e49c8ae41",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2079-4991/12/20/3647/pdf?version=1666078641",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "8ae1d901bf91df8cc9545fa28060cc8472345c58",
"s2fieldsofstudy": [
"Chemistry",
"Materials Science",
"Environmental Science"
],
"extfieldsofstudy": [
"Medicine"
]
} |
51925814 | pes2o/s2orc | v3-fos-license | Plasma potassium ranges associated with mortality across stages of chronic kidney disease: the Stockholm CREAtinine Measurements (SCREAM) project
Abstract Background Small-scale studies suggest that hyperkalaemia is a less threatening condition in chronic kidney disease (CKD), arguing adaptation/tolerance to potassium (K+) retention. This study formally evaluates this hypothesis by estimating the distribution of plasma K+ and its association with mortality across CKD stages. Methods This observational study included all patients undergoing plasma K+ testing in Stockholm during 2006–11. We randomly selected one K+ measurement per patient and constructed a cross-sectional cohort with mortality follow-up. Covariates included demographics, comorbidities, medications and estimated glomerular filtration rate (eGFR). We estimated K+ distribution and defined K+ ranges associated with 90-, 180- and 365-day mortality. Results Included were 831 760 participants, of which 70 403 (8.5%) had CKD G3 (eGFR <60–30 mL/min) and 8594 (1.1%) had CKD G4–G5 (eGFR <30 mL/min). About 66 317 deaths occurred within a year. Adjusted plasma K+ increased across worse CKD stages: from median 3.98 (95% confidence interval 3.49–4.59) for eGFR >90 to 4.43 (3.22–5.65) mmol/L for eGFR ≤15 mL/min/1.73 m2. The association between K+ and mortality was U-shaped, but it flattened at lower eGFR strata and shifted upwards. For instance, the range where the 90-day mortality risk increased by no more than 100% was 3.45–4.94 mmol/L in eGFR >60 mL/min, but was 3.36–5.18 in G3 and 3.26–5.53 mmol/L in G4–G5. In conclusion, CKD stage modifies K+ distribution and the ranges that predict mortality in the community. Conclusion Although this study supports the view that hyperkalaemia is better tolerated with worse CKD, it challenges the current use of a single optimal K+ range for all patients.
A B S T R A C T
Background. Small-scale studies suggest that hyperkalaemia is a less threatening condition in chronic kidney disease (CKD), arguing adaptation/tolerance to potassium (K þ ) retention. This study formally evaluates this hypothesis by estimating the distribution of plasma K þ and its association with mortality across CKD stages. Methods. This observational study included all patients undergoing plasma K þ testing in Stockholm during 2006-11. We randomly selected one K þ measurement per patient and constructed a cross-sectional cohort with mortality follow-up. Covariates included demographics, comorbidities, medications and estimated glomerular filtration rate (eGFR). We estimated K þ distribution and defined K þ ranges associated with 90-, 180-and 365-day mortality. Results. Included were 831 760 participants, of which 70 403 (8.5%) had CKD G3 (eGFR <60-30 mL/min) and 8594 (1.1%) had CKD G4-G5 (eGFR <30 mL/min). About 66 317 deaths occurred within a year. Adjusted plasma K þ increased across worse CKD stages: from median 3.98 (95% confidence interval 3.49-4.59) for eGFR >90 to 4.43 (3.22-5.65) mmol/L for eGFR 15 mL/min/1.73 m 2 . The association between K þ and mortality was U-shaped, but it flattened at lower eGFR strata and shifted upwards. For instance, the range where the 90-day mortality risk increased by no more than 100% was 3.45-4.94 mmol/ L in eGFR >60 mL/min, but was 3. 36-5.18 in G3 and 3. 26-5.53 mmol/L in G4-G5. In conclusion, CKD stage modifies K þ distribution and the ranges that predict mortality in the community. Conclusion. Although this study supports the view that hyperkalaemia is better tolerated with worse CKD, it challenges the current use of a single optimal K þ range for all patients.
I N T R O D U C T I O N
Hyperkalaemia, i.e. excess of circulating potassium (K þ ), is a potentially life-threatening electrolyte disorder [1]. Although it is agreed upon that there appears to be a U-shaped relationship between serum K þ and death [2][3][4][5][6], there remains a controversy as to the exact level of K þ at which the clinician should become concerned and consider altering therapy such as the renin-angiotensin-aldosterone system (RAAS) blockade [7][8][9][10]. The optimal serum K þ may depend on the underlying conditions. Maintenance of total body K þ homoeostasis is a major function of a healthy kidney. Chronic kidney disease (CKD), with subsequent failure to augment distal tubular K þ secretion, arises as one of the most important clinical predictors of hyperkalaemia occurrence [6,11]. In addition to K þ retention, accompanying comorbidities and medications make CKD patients particularly prone to a state of chronic hyperkalaemia [5,6].
Intriguingly, there is evidence suggesting that hyperkalaemia is a less threatening condition in CKD [3,[12][13][14][15][16], arguing adaptive mechanisms that render better tolerance to elevated K þ in circulation [17][18][19][20][21][22]. Most of these studies are of small scale and they include high-risk patients. In clinical practice, chronic mild hyperkalaemia is often interpreted as the normal state in patients with advanced CKD [6,18]. This study evaluates this hypothesis by formally estimating K þ distribution in relation to estimated glomerular filtration rate (eGFR) and by exploring K þ ranges that predict mortality across worse CKD severity stages in a large healthcare system. As a secondary objective, and because plasma is the predominant source of K þ assessment in Sweden, we explored whether the observed optimal ranges approximated to current clinical K þ thresholds.
Data sources and study population
Included individuals were selected from the Stockholm CREAtinine Measurements (SCREAM) project [23], a healthcare utilization cohort from the region of Stockholm, Sweden. SCREAM included all laboratory tests and healthcare utilization data from any individual who had plasma creatinine measured at least once during 2006-11 [23]. All standard laboratory tests performed for these individuals during that period were extracted; the data set was then linked to regional and national administrative databases with complete information on demographic data, healthcare use, diagnoses, validated renal outcomes, vital status and pharmacy-dispensed medications. The Regional Ethical Review Board in Stockholm approved the study.
Eligible patients for this study were all adult individuals (!18 years old) with at least one plasma K þ measurement recorded in outpatient care between 2006 and 2011 with a concurrent plasma creatinine measurement (up to 1 year prior to index date). Implausible plasma creatinine (<25 or >1500 lmol/L) and plasma K þ (>10 mmol/L) values were discarded, and if multiple measurements were recorded on the same day, then their median value was considered. We censored at migration from the county or 31 December 2011; we also discarded individuals undergoing maintenance dialysis. Finally, we sampled at random one measurement per individual and set the index date as the date in which such measurement occurred. Baseline demographics, comorbid history and drug use (up to 6 months before index date) were defined at that point. The detailed flow chart of selection process for this study is depicted in Supplementary data, Figure S1.
Biochemical assessments and study covariates
Biochemical assessments were performed as part of healthcare encounters and by three different laboratories that provide services to the region [23]. Inter-and intra-laboratory variation was considered minimal, as the three laboratories were frequently being audited for quality and harmonization. Plasma creatinine measurements were standardized to isotope dilution mass spectrometry. Creatinine was used to calculate eGFR using the 2009 CKD-EPI creatinine-based equation [24], and categorized as <30 (mL/min/1.73 m 2 , eGFR category G4-G5), between 30 and 60 (eGFR G3) or !60 (eGFR G1-G2). Scandinavian countries adopted in the early 2000s plasma as Optimal potassium ranges across CKD stages the standard source for K þ in healthcare [25]. Plasma K þ tests accounted for 91% of all K þ measurements (the remaining being in serum) [11] and were measured by potentiometric titration.
Other study covariates were age at index date and sex. Clinical history of comorbidities (listed in Supplementary data, Table S1) was defined from relevant International Classification of Diseases, 10th edition codes by established algorithms [26]; the use of selected K þ -sparing medications (listed in Supplementary data, Table S2) at index event was defined as a drug purchase during the 6 months before index date using Anatomical Therapeutic Chemical classification system codes.
Mortality outcomes
Mortality was ascertained via linkage with the Swedish living status registry that is updated monthly and has no loss to follow-up. The main study outcome was 90-day mortality; secondary outcomes were 180-and 365-day mortality.
K þ distribution within CKD strata
We modelled plasma K þ distribution using quantile regression that calculates the percentiles of the outcome distribution as a function of other covariates. We modelled 2.5, 50.0 and 97.5 percentiles, including all covariates (demographics, comorbidities and medications detailed above) present in at least 1% of individuals and all pairwise statistical interactions between age, sex and eGFR. eGFR and age were modelled by using restricted cubic splines, with knots at 15, 30, 45, 60, 90 mL/min/ 1.73 m 2 and 45, 65, 75 years, respectively. Plasma K þ reference range (or normal range) was defined as the set of values where 95% of the population falls within (i.e. the interval between the 2.5th and 97.5th percentiles of the distribution). We predicted distribution ranges using the quantile regression models and predicted the percentiles under different combinations of eGFR categories, age categories, sex, comorbidities and drug use, setting the adjustment level for the remaining covariates to their median (or most frequent) value.
K þ ranges associated with mortality within CKD strata First, we modelled the association between K þ values (as a continuous variable) and mortality using Cox regression models. The concept of 'optimal' K þ range is based on concentrations or levels that are associated with optimal health or minimal risk. The limits are defined as those where the health risks exceed a certain threshold. In this study, K þ ranges were defined as the range of values within which the mortality risk, quantified via the predicted hazard ratio (HR) from the above-mentioned models for a given eGFR category, increased by no more than 50% or 100%. We considered ranges where the mortality risk increased by no more than 50% as clinically relevant, and ranges where the mortality risk increased by no more than 100% as our most conservative estimation of clinical ranges. The central point, and reference value of each safe range was the value that yielded the lowest predicted risk for each eGFR category. We modelled K þ using restricted cubic splines with five knots at default percentiles, and we included an interaction term with eGFR category to test for effect modification. We adjusted each model by eGFR value, demographics, comorbidities and drug use. At every step, we graphically assessed and found satisfying the proportional hazards assumption for the validation models by plotting Schoenfeld residuals against ranks of time.
As a sensitivity analysis, we repeated the random sampling of index K þ per individual generating 99 additional study data sets. We performed all analyses again on these 99 data sets and compared them to evaluate if the random sampling process influenced the robustness and consistency of our results.
Statistical analyses
We present continuous variables as median with interquartile interval (IQI); categorical variables are presented as number with proportion. All data cleaning, tidying and statistical analyses were performed using R [27].
K þ distribution within CKD strata Median plasma K þ values and reference ranges [95% confidence interval (CI)] by eGFR strata are presented in Figure 1A. Median K þ levels were higher with lower kidney function, with a noticeable point of inflexion at the level of 45 mL/min/1.73 m 2 . Reference ranges (95% CI) for plasma K þ by eGFR (A) and density distribution of plasma K þ by eGFR category (B). In (A), the red line represents the 50th percentile, and black lines represent 2.5th and 97.5th percentiles, with their corresponding 95% CIs in shadows. (A) The output of a quantile regression model adjusted by eGFR, age, sex, comorbidities (myocardial infarction, heart failure, peripheral vascular disease, cerebrovascular disease, dementia, rheumatoid disease, chronic obstructive pulmonary disease, peptic ulcer disease, liver disease, cancer, diabetes mellitus, hypertension) and ongoing medication (NSAIDs, angiotensin-converting enzyme inhibitor, angiotensin receptor blockers, mineralocorticoid receptor antagonists). eGFR and age were modelled by using restricted cubic splines, and all pairwise statistical interactions between eGFR, age and sex were included.
Optimal potassium ranges across CKD stages For instance, a median plasma K þ of 3.98 mmol/L was observed in participants with eGFR between 91 and 105 mL/min/1.73 m 2 compared with median plasma K þ of 4.56 mmol/L in participants with eGFR 15 mL/min/1.73 m 2 ( Table 2). The lower reference limit was rather stable across lower eGFR categories, but the upper reference limit increased with worse eGFR categories, from 4.45 for individuals with eGFR >105 to 5.95 mmol/L for individuals with eGFR 15 mL/min/1.73 m 2 ( Table 2). The distribution of plasma K þ values by eGFR strata is presented in Figure 1B, and it shows a gradual shift towards higher plasma K þ values across eGFR categories. Reference ranges by age, adjusting for renal function and all other covariates, were stable with minimal variation as age increased (Supplementary data, Figure S2 and Supplementary data, Table S4). The shift in K þ ranges in relation to kidney function was similarly observed when segregating results by age and sex (Supplementary data, Table S5 and Supplementary data, Figure S3).
As a sensitivity analysis, we compared these ranges with those obtained by replicating the random K þ selection process 99 times more. Supplementary data, Table S6 shows the reference ranges obtained by selecting the mean and median extremities across all the replicates, the narrowest interval obtained and the widest interval obtained. The difference between the median value and normal ranges presented in the main analysis and those obtained from the replicates were considered as minimal.
Follow-up for mortality
During the first 90 days of follow-up, 11 833 participants died (1.42%). Furthermore, 20 615 (2.48%) and 33 869 (4.07%) participants died within 180 and 365 days, respectively. K þ values outside the previously estimated distribution range were associated with increased risk of mortality; multivariable-adjusted HRs ranged between 1.72 (95% CI 1.62-1.82) and 2.97 (95% CI 2.76-3.19), with greater effect sizes for short-term mortality outcomes (Supplementary data, Table S7). Sensitivity analyses obtained by replicating the selection process showed consistent results in terms of magnitude and direction of the association (Supplementary data, Table S8).
eGFR-specific K þ ranges based on risk of mortality
As expected, the absolute risk of mortality is higher in individuals with lower renal function (Supplementary data, Figure S4). The adjusted relative mortality risk associated with K þ levels by eGFR category is depicted in Figure 2. Overall, the Ushaped association was gradually flattened with lower eGFR strata and for each predetermined follow-up time (P for interaction <0.001). Table 3 depicts K þ concentration associated with the minimum mortality risk across eGFR strata. Although the lowest 90-day mortality risk was observed for a plasma concentration of 4.02 mmol/L in individuals with normal renal function (G1-G2), this value went up to 4.34 mmol/L (8% increase) in individuals with eGFR G4 or more. The same pattern was observed throughout the different observation periods. This was accompanied by a widening in the K þ ranges that predicted different risk scenarios: whereas the lower K þ limit remained rather stable (between 3.4 and 3.7 mmol/L through all strata), the upper K þ limit presented larger differences with lower eGFR. For instance, in eGFR G1-G2, while the range where the 90-day mortality risk increased by no more than 100% was 3.45-4.94 mmol/L, it widened to 3.36-5.18 in G3 and 3.26-5.53 mmol/L in G4-G5. Sensitivity analyses showed similar results in the 99 random data set selections (Supplementary data, Tables S9-S11), and a 'dose-dependent' association was also observed while further splitting the eGFR category G3 into G3a (eGFR <60-45 mL/min/1.73 m 2 ) and G3b (eGFR <45-30 mL/min/1.73 m 2 ) (Supplementary data, Figure S5 and Supplementary data, Table S12). Multivariable association between K þ levels and 90-, 180-and 365-day mortality, by eGFR category. Models are adjusted by eGFR category, eGFR value, age, sex, comorbidities (myocardial infarction, heart failure, peripheral vascular disease, cerebrovascular disease, dementia, rheumatoid disease, chronic obstructive pulmonary disease, peptic ulcer disease, liver disease, cancer, diabetes mellitus, hypertension) and ongoing medication (NSAIDs, angiotensinconverting enzyme inhibitor, angiotensin receptor blockers, mineralocorticoid receptor antagonists). K þ was modelled by using restricted cubic splines with five knots, and a statistical interaction between K þ and eGFR category was included.
When applying the K þ range of 3.5-5.0 mmol/L, the HR for 90-day mortality associated with out of range K þ values was significantly lower (non-overlapping 95% CI) in patients with eGFR G4-G5 (1.73, 95% CI 1.51-1.95) when compared with eGFR G1-G2 (2.66, 95% CI 2.48-2.85). When applying our eGFR-specific ranges, the HR was similar across eGFR strata (Supplementary data, Figure S6 and Supplementary data, Table S13). A considerable proportion of individuals would be reclassified by our proposed eGFR-specific ranges (Supplementary data, Table S14).
D I S C U S S I O N
The main findings of this large healthcare-based study are: (i) patients with CKD had progressively higher concentration of plasma K þ ; (ii) the association between plasma K þ and mortality was U-shaped, with the lowest mortality observed in the range of 4.0-4.3 mmol/L depending on the underlying kidney function; and (iii) K þ ranges for mortality prediction varied by underlying kidney function, being shifted upwards in individuals with moderate and severe/end-stage CKD. Thus, although our study supports the view that hyperkalaemia is better tolerated in CKD, it challenges the current use of a single clinical K þ range for all patients.
The kidney plays a fundamental role in K þ homoeostasis [6], and our study convincingly showed an important shift in both crude and multivariable-adjusted plasma K þ distribution with lower eGFR. How this chronic K þ accumulation may influence clinical outcomes is not well known, but small-scale reports suggested hyperkalaemia to be a less threatening condition in CKD. These studies compare the mortality associations of current clinical K þ thresholds in patients with CKD: Einhorn et al. [12] were the first to suggest a stronger association between hyperkalaemia and 1-day mortality among patients with normal renal function than among patients with CKD; patients with end-stage renal disease in the study by Nakhoul et al. [15] also had lower death hazards with hyperkalaemia relative to the entire cohort; An et al. [13] showed, in 923 consecutive hospital admissions, a graded decrease in risk of death among patients with extreme levels of hyperkalaemia as CKD stage worsened; finally, a prospective analysis of sustained hyperkalaemia and outcomes in 820 patients with creatinine clearance <50 mL/ min suggested that serum K þ in the ranges of 5.0-6.0 mmol/L was well tolerated and associated with lower mortality risk when compared with 4.0 mmol/L [16]. Of note, similar effect differences were reported in acute myocardial infarction cohorts when comparing dialysis versus non-CKD patients with hyperkalaemia [3,14]. We could indeed observe the same HR attenuation in our study when using the currently established single clinical thresholds.
Our study expands previous evidence with the novel demonstration that the U-shaped association between K þ and death flattens across worse CKD stages. Based on this flattening, we suggest K þ ranges for different scenarios of risk prediction, and show a clear gradual shift in the upper K þ threshold across stages of CKD. We note that our most conservative range estimation in individuals with normal kidney function (3.45-4.95 mmol/L for 90-day mortality risk) approximates to the currently accepted single threshold of 3.5-5.0 mmol/L. However, this same estimation suggests a difference of 0.5 mmol/L in the upper threshold of individuals with CKD G4-G5 (3.26-5.53 mmol/L). This is consequent with the biological plausibility of a physiological adaptation to chronic hyperkalaemia in these patients [17][18][19][20][21][22]. Yet, our results disagree with the size effect attenuation reported in previous small studies [16], and we note that plasma K þ >5.5 mmol/L strongly and independently predicted mortality in all eGFR strata, suggesting that K þ concentrations in these ranges continue to deserve attention, strict monitoring and treatment to minimize adverse outcomes. Recently [28], a meta-analysis of the CKD prognosis consortium reported no difference in the association between serum K þ and mortality by eGFR strata among 1.2 million participants from 27 international cohorts. The diversity of populations considered in that analysis (10 general population cohorts or healthcare extractions, 7 high cardiovascular risk cohorts and 10 CKD-referred cohorts) may hamper comparison with our results. Furthermore, this research question was not formally addressed, given serum K þ within each eGFR strata was compared against the same reference category (4.2 mmol/L) that is based on the lowest risk prediction for the whole population.
The mechanisms behind this possible adaptation to increased K þ in CKD are not well described and cannot be ascertained from our observational analyses. Of the possible mediators of the association between hyperkalaemia and death, cardiac arrhythmia caused by alteration in the transmembrane potential is probably the most relevant. K þ has a major role in maintaining the resting membrane potential, with implications in regulating transmission of nerve impulses and cardiac rhythm. In myocardial cells, there are about 10 different K þ ion channels that are responsible for the intricate maintenance of a stable resting membrane potential through repolarization of the cell [29]. The K þ channels display different properties and movements of K þ are also influenced by other ions such as protons, calcium and magnesium. CKD patients may to some extent adapt to the higher K þ levels through different mechanisms (intestinal and tubular secretion), thereby controlling hyperkalaemia. Nonetheless, we are not aware of any studies showing adaptation to higher K þ levels in other cells, e.g. myocardiocytes. However, K þ ion channels of the myocardium are sensitive to endogenous factors and may change both the number and the functioning as a result of structural [30,31] and environmental [32,33] alterations. We hypothesize that similar mechanisms may take place in CKD patients by slowly increasing K þ levels that are already in the high normal range. Finally, because of the interdependence between K þ , acidosis and calcium/magnesium, it would be interesting to determine whether, for instance, acidosis modulates the association between circulating K þ and subsequent mortality.
A second novelty in our study is the focus on plasma K þ determinations. Plasma K þ is becoming standard and preferred to serum in many countries, owing to its faster turn around and not being influenced by pre-analytical issues (such as thrombocytosis, tourniquet compression or violent sample handling, thus reducing the risk of false positives-e.g. pseudokalaemias) [34]. Despite this, plasma K þ is not currently considered by Optimal potassium ranges across CKD stages clinical guidelines, which either refer to serum or do not make distinction between the sources of K þ determination [7][8][9][10]. Our plasma-derived K þ ranges may potentially be helpful to estimate risk in clinics or regions that use this method.
These findings may have clinical implications. On one hand, we add robust evidence to ongoing discussions on the need to develop disease-specific 'optimal' K þ ranges [35]. Although current clinical guidelines suggest maintaining K þ within a range of 3.50-5.00 mmol/L, observational studies exploring mortality-associated ranges in elderly [2], and patients with myocardial infarction [3] or hypertension [4], suggest these ranges to be narrower, in the range of 3.5-4.5 mmol/L. Our less conservative (<50% risk increase) observed range in individuals with normal kidney function (3.61-4.61 mmol/L for 90-day mortality) would agree with these studies, but again suggests a difference of 0.5 mmol/L in the upper threshold of individuals with CKD G4-G5 (3.50-5.09 mmol/L).
On the other hand, clinical guidelines recommend some clinical decisions based on these K þ thresholds, such as not to initiate RAAS blockade therapy when K þ is >5.0 mmol/L [7][8][9][10]. Given the widely demonstrated effectiveness of these therapies and our observed shift in K þ ranges, interventional studies and/or post hoc analyses of existing trials may be needed to reevaluate these assumptions. Studies should ideally follow to confirm our findings and, specially, to report on serum K þ ranges across renal function categories. If we consider that serum K þ concentration can be typically 0.1-0.4 mmol/L higher than that measured in an anticoagulated plasma sample [36], then this may have larger consequences in both the use of these therapies and the management of dyskalaemias in persons with CKD.
Our study was observational, and we recognize that despite the use of robust methods, there is residual/unknown confounding, and that the relationship between K þ and death may represent reverse causation. It is also conceivable that the relative importance of hyperkalaemia decrease with lower eGFR as the underlying absolute risk of mortality is larger and patients die from other causes. Hyperkalaemia is also less common in individuals with normal renal function. Hence, the increased risk of mortality in that group may be a proxy of severe acute illnesses that we may not have accounted for in the analysis (detection bias). Nevertheless, even if K þ is just a marker of illness, clinical ranges are useful to aid in clinical decisions and may still vary by CKD stage. As study strengths, SCREAM and the Swedish National Health registries contain uniquely detailed information, and we were able to extract reliable and complete data on K þ tests, comorbid illnesses, concomitant medication and date of death, thus overall strengthening the reliability of our findings.
To conclude, our study shows that individuals with CKD require higher plasma K þ to predict a mortality risk similar to that of persons with normal renal function. Although we hope that our efforts open discussions on the need for disease-specific as opposed to a single clinical K þ range for all patients, we note that physiological adaptation in CKD incompletely explains why mortality risk could be diminished once hyperkalaemia has been established. As recently discussed by Montford and Linas [1], frequent metabolic derangements (such as metabolic acidosis) and structural heart disease (such atrial fibrillation or heart failure) in CKD patients should theoretically predispose them to more toxicity with hyperkalaemia due to lower arrhythmogenic potential. Trials that randomize patients with CKD to different K þ targets would be necessary to definitively establish the optimal range for maintaining K þ levels.
S U P P L E M E N T A R Y D A T A
Supplementary data are available at ndt online.
F U N D I N G
This study was supported by an institutional grant from AstraZeneca to Karolinska Institutet. In addition, we acknowledge additional grant support from the Swedish Heart and Lung Foundation, the Stockholm County Council, Vifor Fresenius Medical Care Renal Pharma and the Martin Rind's and Westman's Swedish Foundations.
C O N F L I C T O F I N T E R E S T S T A T E M E N T
J.-J.C. received research grants from Viforpharma, Astellas, Merck Sharp & Dome; H.X. was employed by AstraZeneca; T.J. was lecturing and consulting for Astra-Zeneca, Merck and Aspen; L.H.L. received research grants and was consulting for AstraZeneca and ViforPharma, and consulting for Bayer. The other authors have no conflict of interest to declare. | 2018-08-14T19:12:27.365Z | 2018-08-06T00:00:00.000 | {
"year": 2018,
"sha1": "99a8bae55100ae3b477678dae31bff2997d621a0",
"oa_license": "CCBYNC",
"oa_url": "https://academic.oup.com/ndt/article-pdf/34/9/1534/33455992/gfy249.pdf",
"oa_status": "HYBRID",
"pdf_src": "PubMedCentral",
"pdf_hash": "95c7d6c782783160ece18813ef34f750aae22684",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
41095582 | pes2o/s2orc | v3-fos-license | Analysis of the albumin level, neutrophil-lymphocyte ratio, and platelet-lymphocyte ratio in Guillain-Barré syndrome
ABSTRACT The purpose of this study was to investigate the prognostic value of the pretreatment and post-treatment albumin level, neutrophil-lymphocyte ratio (NLR), and platelet-lymphocyte ratio (PLR) in subtypes of Guillain-Barré syndrome (GBS). A retrospective analysis of 62 patients with GBS treated between 2011 and 2015 in Dicle University Hospital, Turkey, was carried out. The pretreatment and post-treatment albumin, NLR, and PLR were documented, together with acute inflammatory demyelinating polyradiculoneuropathy (AIDP), acute motor axonal neuropathy, motor sensory axonal neuropathy, and Hughes’ scores. Post-treatment albumin levels in GBS were significantly reduced, and albumin level was negatively correlated with the Hughes scores. Elevated pretreatment NLRs and PLRs were significantly associated with AIDP. There were no correlations between the Hughes scores, NLR, and PLR. The results point to a negative correlation between albumin levels and GBS disability and suggest that the NLR and PLR may be promising blood biomarkers of AIDP.
levels may be related to the course and outcome in GBS 4 . The neutrophil-to-lymphocyte ratio (NLR) is calculated from the white blood cell count and is a novel prognostic and inflammatory marker in patients with neurological diseases 5,6 . The platelet-to-lymphocyte ratio (PLR) is a new biomarker of inflammation 7 . The PLR is thought to be a sensitive marker and to be a prognostic factor in many malignancies 8 . Changes in the levels of acute phase reactants, such as albumin, the NLR, and the PLR have not been well studied in patients with GBS. In this study, we aimed to evaluate the albumin level, NLR, and PLR in patients with GBS. We also evaluated the association between disease prognosis and the albumin level, NLR, and PLR.
METHODS
This study was conducted retrospectively in the Neurology Department of Dicle University, Diyarbakir, Turkey. The data for the study were extracted from the medical records of patients who attended the hospital between January 2011 and January 2016. The study included 62 patients with GBS. Demographics, age, sex, clinical features, electrophysiology, subtype, and treatment-related outcomes were assessed. A diagnosis of GBS was based on the criteria of the Brighton Collaboration GBS Working Group 9 . The strength of the proximal and distal muscles of the upper and lower limbs was classified as 0-5, according to the criteria of the Medical Research Council. Each patient was evaluated according to Hughes et al. 's disability score at the time of hospital admission and discharge 10 .
All the patients underwent physical and neurological examinations, liver and kidney function tests, and lipid profiling. In all cases, a complete blood count was also obtained, and electrolyte levels were tested. Electromyography (Nihon-Kohden) was performed in each patient. The classification of the patients as having an axonal or demyelinating subtype was based on the electrodiagnostic criteria of Hadden et al. 11 , and an AMSAN diagnosis was based on the criteria of Rees et al. 12 .
Exclusion criteria included severe heart failure, autoimmune disease, diabetes mellitus, malignant hypertension, Cushing' s syndrome, central nervous system vasculitis, congenital vascular disease, trauma, dissection, thyroid and kidney dysfunction, liver failure, and local and systemic infection.
Venous blood samples were collected when the patient initially presented to the emergency department or intensive neurology care unit (pretreatment-1) and 96-120 h after the first observation (post-treatment-2). The serum albumin levels were measured using a Beckman Coulter CX9 (Beckman Coulter, Inc., Brea, CA) chemistry analyzer. At our hospital, a serum albumin range of 3.5-5.5 gr/dL is considered normal. Hematologic indices were measured using an automated hematology analyzer system (Abbott Cell-Dyn 3700; Abbott Laboratory, Abbott Park, ILs). All subsequent analyses were based on absolute cell counts. The baseline NLR was measured by dividing the neutrophil count by the lymphocyte count, and the PLR was measured by dividing the platelet count by the lymphocyte count.
Statistical analysis
The statistical analyses were performed using SPSS software, version 20.0 (SPSS Inc., Chicago, IL). Continuous data are presented as mean ± standard deviation (SD). Between-group differences in continuous variables were determined by a Student's t test or the Mann-Whitney U test for variables, with or without a normal distribution, respectively. To test whether the data showed a normal distribution, the Kolmogorov-Smirnov test was used. Categorical variables were summarized as percentages and compared with a one-way ANOVA test. Relationships between the variables were examined by calculating Pearson's and Spearman's correlation coefficients. To find independent associates of the Hughes' score, variables with a p value of ≤ 0.05 in a bivariate correlation analysis and univariate analysis were selected for multiple linear regression analyses. The cut-off values and corresponding sensitivity and specificity values for the prediction of the AIDP based on the serum albumin level, NLR, and PLR were estimated by receiving operator characteristic (ROC) curve analysis. A p value of < 0.05 was accepted as the threshold for determining statistical significance.
RESULTS
Sixty-two patients were enrolled in this study. Of the patients with GBS, 36 were men (58.1%), and 26 were women (41.9%). The mean age of the patient group was 48.0 ± 19.84 . Four of the patients died. Intravenous immunoglobulin was administered to all the patients.
The mean serum albumin levels were 3.58 ± 0.55 (2-4.6) at the first observation (albumin-1), and 30.6% of the patients (n = 19) had hypoalbuminemia. The mean serum albumin levels after 96-120 h (albumin-2) were 3.32 ± 0.59 (1.5-4.5), and 54.8% of the patients (n = 34) had hypoalbuminemia. Three patients were treated with 100 ml of 25% albumin via intravenous infusion. The albumin-1 (baseline/pretreatment) levels were significantly lower than the albumin-2 (post-treatment) levels (p < 0.05). The albumin-1 and -2 levels were negatively correlated with the Hughes' scores (admission/discharge). Thirty-five of the patients had AIDP, 12 had AMAN, and 15 had AMSAN. The Table shows the comparisons of the demographic features and laboratory findings among the subgroups. In the patients with AIDP, the neutrophil-1 and -2 levels (pre-and post-treatment levels, respectively) and NLR-1 (pretreatment) were significantly higher than those of the patients with AMAN and AMSAN. (p < 0.05). The pretreatment PLR (PLR-1) was significantly higher in the patients with AIDP when compared with those with AMAN (p < 0.05) The neutrophil-1 and -2 and lymphocyte-1 (pretreatment) and -2 (post-treatment) levels were significantly higher in the patients with AIDP, as compared to those of the patients with AMSAN (p < 0.05). When the results of the pre-and posttreatment measurements were compared, there were no correlations between the Hughes' scores (admission/discharge) and neutrophil-1 and -2, lymphocyte-1 and -2, platelet-1 and -2, NLR-1-and 2, and PLR-1 (p > 0.05).
DISCUSSION
This study demonstrated that serum albumin levels decreased in GBS patients in the subacute period and that there was a negative correlation between albumin levels and Hughes' scores (admission/discharge). The NLR and PLR increased in AIDP during the acute period. To the best of our knowledge, this is the first clinical study to evaluate the association of GBS subtypes with serum albumin levels and the NLR and PLR.
The serum albumin concentration depends on various factors, such as the synthesis, rate of degradation, distribution, and exogenous loss of albumin, as well as nutritional intake and colloid oncotic pressure changes. The presence of systemic inflammation affects the synthesis of albumin 13,14,15,16 . Albumin is a late-reacting negative acute-phase protein 17 .
The present study demonstrated the following: hypoalbuminemia is common in patients with GBS, it decreases after the subacute period, and there is a negative correlation between albumin levels and GBS disability. The mean pre-and post-treatment serum albumin levels of the AIDP group were lower than those of the other groups. Such decreases in mean albumin levels in AIDP are thought to be mainly due to inflammation, hemodilution, or an acute phase response. Pathophysiological changes in GBS depend upon the subtype. Immune reactions directed against epitopes in Schwann cell surface membrane or myelin can cause AIDP 18 . Cellular and humoral immune responses participate in these pathophysiological processes, with infiltration of epineural and endoneural small vessels by lymphocytes and monocytes causing segmental myelin degeneration throughout the nerve 19 . In demyelination forms of GBS, Berciano et al. showed that spinal root sections had extensive and almost pure macrophage-associated demyelination, with the occasional presence of T lymphocytes and neutrophil leukocytes 20 .
On the other hand, immune reactions against epitopes in the axonal membrane cause AMAN and AMSAN 21 . In these variants of GBS, the axon is affected, without an inflammatory response 21 . The primary immune process is directed at the nodes of Ranvier, leading to functional axonal involvement with conduction block caused by paranodal myelin detachment, node lengthening, sodium channel dysfunction, and altered ion and water homeostasis 22 . These pathophysiological processes may be rapidly reversed in some patients or it may progress to axonal degeneration. Acute motor axonal neuropathy involves the motor nerves of the ventral roots, peripheral nerves, and preterminal intramuscular motor twigs 23 . In AMSAN, sensory nerves are also affected. These pathophysiological processes may increase the importance of lymphocytes and neutrophils as diagnostic features of GBS subtypes. In this study, neutrophilia was detected in AIDP. According to some studies, the NLR and PLR are new biomarkers of the presence of inflammation 24,25 . Alan et al. 24 showed that the NLR and PLR might be associated with the presence and severity of Behçet's syndrome. Kokcu et al. 26 reported that the NLR, platelet count, and PLR were elevated in late stages of ovarian cancer. They also claimed that the PLR was an independent prognostic factor of the stage of epithelial ovarian cancer. In the present study, NLR-1 was a statistically significant biomarker in AIDP, and PLR-1 was a statistically significant biomarker in AIDP but not AMAN. When the results of the pretreatment and post-treatment measurements were compared, there were no correlations between the Hughes' scores (admission/discharge) and neutrophil-1 and -2, lymphocyte-1 and -2, platelet-1 and -2, NLR-1 and -2, and PLR-1. The data demonstrated that a pretreatment NLR value of 3.275 predicted the presence of the acute period of AIDP with 83% sensitivity and 93% specificity. A pretreatment PLR of 121.8 predicted the presence of the acute period of AIDP, with 74% sensitivity and 70% specificity.
In conclusion, decreased albumin levels may exacerbate GBS-related disability. Decreased NLRs and PLRs may indicate the presence of AIDP, but they are not associated with the severity of the disease. The NLR may be a useful diagnostic marker of AIDP. Larger prospective studies are needed to support the findings of the present study. | 2017-08-28T11:52:59.948Z | 2016-03-07T00:00:00.000 | {
"year": 2016,
"sha1": "8badc6b608addf0d7926b4c22180ac1fd59de065",
"oa_license": "CCBYNC",
"oa_url": "http://www.scielo.br/pdf/anp/v74n9/0004-282X-anp-74-09-0718.pdf",
"oa_status": "GOLD",
"pdf_src": "Thieme",
"pdf_hash": "6ecdd638cc443acd39c9d0b489dc1c3b9876c9a1",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
206558140 | pes2o/s2orc | v3-fos-license | Saccharomyces cerevisiae Gle2/Rae1 is involved in septin organization, essential for cell cycle progression
Abstract Gle2/Rae1 is highly conserved from yeast to humans and has been described as an mRNA export factor. Additionally, it is implicated in the anaphase‐promoting complex‐mediated cell cycle regulation in higher eukaryotes. Here we identify an involvement for Saccharomyces cerevisiae Gle2 in septin organization, which is crucial for cell cycle progression and cell division. Gle2 genetically and physically interacts with components of the septin ring. Importantly, deletion of GLE2 leads to elongated buds, severe defects in septin‐assembly and their cellular mislocalization. Septin‐ring formation is triggered by the septin‐regulating GTPase Cdc42, which establishes and maintains cell polarity. Additionally, activity of the master cell cycle regulator Cdc28 (Cdk1) is needed, which is, besides other functions, also required for G2/M‐transition, and in yeast particularly responsible for initiating the apical–isotropic switch. We show genetic and physical interactions of Gle2 with both Cdc42 and Cdc28. Most importantly, we find that gle2∆ severely mislocalizes Cdc42, leading to defects in septin‐complex formation and cell division. Thus, our findings suggest that Gle2 participates in the efficient organization of the septin assembly network, where it might act as a scaffold protein. © 2017 The Authors. Yeast published by John Wiley & Sons, Ltd.
Introduction
Septins are highly conserved eukaryotic proteins that also in human are increasingly recognized as novel components of the cytoskeleton (Mostowy and Cossart, 2012). Their dysfunction is linked to various diseases, including cancer, neurological disorders and infections. All septins are GTPbinding proteins that form hetero-oligomers and higher-order structures resulting in filaments, bundles or rings (Mostowy and Cossart, 2012), which are necessary to control cellular processes that require localization, for instance at the division site (Joo et al., 2005;Kinoshita and Noda, 2001) or the plasma membrane (Hagiwara et al., 2011;Sellin et al., 2011). Septins control cellular processes by being scaffolds for protein recruitment and by establishment of structures that provide diffusion barriers important for cell division (Mostowy and Cossart, 2012). Their ability to form filaments was shown to be crucial for septin function and in case of errors activate the morphogenesis checkpoint to halt cell division (Kim et al., 2011;Lew, 2003;McMurray et al., 2011). Septins associate with cellular membranes, actin filaments and microtubules (Kinoshita et al., 2002;Sellin et al., 2011;Surka et al., 2002;Tanaka-Takiguchi et al., 2009). However, the regulatory mechanisms for the directed and timely septin assembly are only partly understood.
Here we show that Gle2 (RAE1 in humans) is involved in proper septin organization. Gle2 was identified as a Nup116-and Nup100-associated protein, which helps to sustain the structural integrity of the nuclear pore complex (NPC) (Ho et al., 1998). Deletion or mutation of GLE2 leads to NPC-clustering (Bucci and Wente, 1997) and accumulation of poly(A) + containing RNAs in the nucleus (Bailer et al., 1998;Murphy et al., 1996). The protein is highly conserved from Saccharomyces cerevisiae to metazoans and an involvement in mRNA-export has also been documented for Saccharomyces pombe (Yoon et al., 2000) and human Blevins et al., 2003).
Interestingly, in addition to its involvement in mRNA export, a mutation in S. pombe RAE1 (rae1-1) leads to an arrest in cell cycle at the G 2 /M boundary with perturbations of the cytoskeleton (Brown et al., 1995;Whalen et al., 1997). Crystal structure analysis revealed that the kinetochore checkpoint protein hBub3 and Gle2/Rae1 both are seven-bladed WD40 repeat propeller proteins, which are typical scaffold proteins, and studies in human cells revealed that they are both involved in the progression through mitosis (Larsen and Harrison, 2004;Larsen et al., 2007;Reddy et al., 2008;Ren et al., 2010). There, a Rae1-Nup98 complex interacts with the early Cdh1 activated form of the anaphase promoting complex (APC Cdh1 ) (Jeganathan et al., 2005). Ubiquitinylation of securin and mitotic cyclins by the APC with subsequent proteasomal degradation leads to chromosome segregation and entry into mitotic exit (Baker et al., 2007). Defects in this process cause chromosome missegegation and subsequent aneuploidy, leading to cancer and in particular leukaemia Jeganathan et al., 2005). Another cell cycle-related function of Rae1/Gle2 is the localization of an mRNA/Rae1 complex to microtubules (Kraemer et al., 2001;Sitterlin, 2005), where it is required for microtubule dynamics and spindle assembly (Blower et al., 2005).
All of these findings argue for a broad but in detail still undefined role of Gle2/Rae1 in the cell. Our study unravels an involvement for Gle2 in cell cycle regulation and in particular in septin-ring formation, which is essential for cytokinesis.
Yeast strains, plasmids and oligonucleotides
All yeast strains used in this study are listed in the Supporting information Table 1 and plasmids in Table 2. Plasmids and yeast strains were generated by conventional methods. Unless stated differently all yeast strains derived from the BY4741 strain background.
Drop dilution tests
Cells were spotted in serial dilution (10 7 to 10 3 cells/10 μL per drop) onto rich medium (Figures 1b, 3a and 4a) or selective medium ( Figure S1b). Plates were incubated for 3 days at the indicated temperatures.
Synthetic genetic array screen
Synthetic genetic array (SGA) analyses were carried out as described using a Singer RoToR HDA (Tong and Boone, 2006). The query strain was a gle2Δ::natMX4 derivative of Y7092 (HKY1163), which was kindly provided by C. Boone, University of Toronto. The library was a collection of temperature sensitive mutants, also kindly provided by C. Boone. Growth defects were detected by comparing the growth of double mutants with the combined growth of single and double mutants. As a measure for growth, colony areas were taken, which were quantitated from plate scans using 'Balony' (Young and Loewen, 2013).
Cell cycle arrest and flow cytometric analysis
Overnight cultures were diluted in rich medium to a density of 0.5 × 10 7 cells/mL and incubated at 25°C for 2 h. Cells were arrested in their cell cycle by addition of α-factor to a final concentration of 30 μg/mL and incubated for 2 h at 25°C. After addition of another 10 μg/mL α-factor per milliliter culture and 1 h incubation, the 0 min time point was taken and cells were fixed with 70% ethanol. The rest of the culture was washed five times with fresh medium to remove the α-factor and brought into the same volume of fresh medium as before. Samples were taken at time points indicated in the experiments and fixed as described above. For flow cytometry, fixed cells were washed with 50 mM sodium citrate pH 7.0 and treated with 0.25 mg/mL RNase A at 50°C for 1 h. After removal of RNase by washing with sodium citrate, samples were sonicated (15 s, 30% output level, Branson Sonifier 250) with a micro-tip to separate cells from each other. Samples were washed twice 460 G. Zander et al. with sodium citrate and incubated with 0.2 μL Sytox-Green® (Thermo Fisher) per milliliter suspension at room temperature in the dark for 30 min. Analysis of the cells was performed using the BD FACS Canto Cytometer.
Determination of chromosome loss rates
Chromosome loss rates were determined according to Zhu et al. (2015). A gle2Δ::kanMX4 deletion was introduced into RLY8492 (HKY1600) and confirmed via PCR analysis. Two independently isolated clones were analysed. RLY8492 served as wild type and RLY8496 (HKY1602) (mad1Δ) as positive control. Overnight cultures of each strain were grown in SC medium lacking leucine at 25°C. These cultures served to determine the GFP À /GFP + ratio at starting time. The cultures were diluted in YPD to a cell density of 2 × 10 6 cells/mL and grown at 25°C to cell densities of 0.75-1.5 × 10 8 cells/mL. The theoretical number of doublings was calculated for each culture. Cells were fixed with 4% formaldehyde and analysed by flow cytometry. Chromosome loss rates were calculated as described previously (Zhu et al., 2015).
Microscopic studies
For the analysis of live cells as depicted in Figure 1 (c) and Figure S2, cells were grown in rich medium at the indicated temperatures overnight, harvested and examined directly. For green fluorescent protein (GFP) microscopy cells were arrested in cell cycle as described above and fixed with 4% formaldehyde for a maximum of 5 min. Samples were washed twice with P-solution (0.1 M potassium phosphate buffer pH 6.5, 1.2 M sorbitol), permeabilized with 0.5% Triton® X-100 in P-solution on a polylysine coated slide and DNA was stained with Hoechst 33 342 (Sigma). Fluorescent in situ hybridization experiments were used for visualization of poly(A) + RNAs (Figure 2c and Figure S2) as described before (Zander et al., 2016). Cells were grown to log phase and shifted to 37°C for 1 h before they were fixed with 4% formaldehyde for 1 h. Zymolyase (Amsbio) treatment resulted in spheroblasts that were further permeabilized with 0.5% Triton® X-100 in P-solution on a polylysine coated slide. Samples were pre-hybridized with Hybmix (50% deionized formamide, 5× SSC, 1× Denhardts, 500 μg/mL tRNA, 500 μg/mL salmon sperm DNA, 50 μg/mL heparin, 2.5 mM EDTA pH 8.0, 0.1% Tween® 20, 10% dextran sulphate) for 1 h at 37°C and hybridized with a Cy3-labelled oligo d(T) 50 probe (0.5 μM) in Hybmix at 37°C overnight. DNA was stained as described above. For microscopic studies a Leica AF6000 microscope was used and pictures were obtained by using the LEICA DFC360FX camera and processed with the LAS AF 2.7.3.9 software (Leica).
Quantification
All experiments shown in this work were performed at least three times independently with the exception of the SGA screen and the chromosome misseggregation. Error bars represent the standard deviation. p-Values shown in Figure 3(f) were calculated using a two-tailed, two-sample unequal variance t-test. p-Values shown in Figures 3(h) and 4(d, e) and Figure S3(a) were calculated using a two-tailed, two-sample equal variance t-test. p-Values are indicated as follows: *** p < 0.001, ** p < 0.01, * p < 0.05. For quantification of cells with displayed phenotypes (Figures 2c, 3h and 4d and Figure S3) for each experiment a minimum of 100 cells were counted. For Figures 2(d) and 3(c) three times 20 cells were measured.
Gle2 interacts with cell cycle regulators
In order to characterize cellular functions of Gle2 we performed an sSGA analysis with temperature-sensitive (ts) alleles of over 600
465
Gle2 and septin formation essential genes (kindly provided by C. Boone). We prepared a gle2Δ strain, crossed it with the library and analysed haploid segregants, a method described earlier (Tong and Boone, 2006).
More than 100 mutant alleles show genetic interactions with the deletion of GLE2 (Figure 1a). Surprisingly, the amount of interacting genes involved in nuclear transport or RNA processing was quite small (13 alleles). However, we found many genes involved in cell cycle progression and regulation, such as genes encoding proteins of the APC, the kinetochore, the spindle and the cytoskeleton (Figure 1a). To confirm these interactions we generated new double mutants of the APC (cdc20-2), Cks1 (cks1-38), important for G1/S and G2/M transition, and members of the mitotic exit network (cdc14-8, cdc15-2) with gle2Δ via tetrad dissection. Detailed analysis of these mutants showed enhanced growth defects (Figure 1b), increased cell size and defects in morphology, which reflected mostly malfunction at different stages of cell division, when combined with gle2Δ ( Figure S1a). Interestingly, the abnormalities in growth and morphology of cks1-38 were suppressed by high copy (2 μ) GLE2 ( Figure S1b-d), suggesting a direct interaction of these proteins. Indeed, physical interactions of Gle2 specifically with the cell cycle regulators Cks1, Cdc15 and Cdc16, but not Cdc14, are shown in co-immunoprecipitation (co-IP) analyses (Figure 1c). Interactions with the RNA-binding protein Gle2 are insensitive to RNase treatment, suggesting that they are not mediated and dependent on RNA. These findings support an involvement of Gle2 in regulation of the cell cycle.
To address if Gle2 alone affects cell cycle regulation, we performed flow cytometry experiments
466
G. Zander et al. and found significant cell cycle delay in cells deleted for GLE2 (Figure 2a). Wild-type and gle2Δ cells were arrested in the G 1 phase of the cell cycle using the mating pheromone α-factor. Washing away this factor re-starts the cell cycle and SYTOX®-green staining of the DNA allowed monitoring the synchronous population going through replication and cytokinesis. About 40 min after release, the portion of cells with a diploid (2 N) genome is~46% in the wild-type strain. In contrast, less than half (~20%) of the cells in gle2Δ have a 2 N content (Figure 2a, bottom). At 60 min most of the wild-type cells have reached the 2 N stadium, while in gle2Δ this does not happen until 100 min, suggesting that cells lacking GLE2 face trouble entering S-phase and progress from there. These data argue for an already early function of Gle2 in cell cycle progression, although its exact role remains to be studied in more detail.
Our findings that Gle2 genetically and physically interacts with components involved in the regulation of the APC support research in higher eukaryotes that also linked Rae1 with the APC (Jeganathan et al., 2005). As the APC is a major regulator of the correct timing for chromosome segregation and we found a physical interaction of Gle2 with the APC-component Cdc16 (Figure 1 c), we addressed whether Gle2 is required for proper chromosome segregation, using a GFPbased quantitative chromosome transmission fidelity assay that allows sensitive and quantitative detection of chromosome loss (Zhu et al., 2015). We found that the deletion of GLE2 causes massive chromosome missegregation (Figure 2b). Interestingly, the effect is much stronger than that of the spindle assembly checkpoint regulator Mad1, which controls proper attachment of the microtubules to the chromosomes and else delays division of the sister chromatids. Deletion of MAD1 leads to a~10-fold higher loss of the mini-chromosome compared with wild type, an increase that is more than doubled in gle2Δ (~21-and~25-fold higher than wild type). This defect in maintaining chromosomal stability in gle2Δ might result from misorientation of the mitotic spindle or from a general perturbation of cell cycle controlling complexes. Nevertheless, this striking effect underlines the general importance of Gle2 in cell cycle regulation.
Given the involvement of Gle2 in mRNA export, one might speculate that the cell cycle perturbations seen in gle2Δ might be due to a shortage of proteins evoked by insufficient nuclear export of the respective mRNAs. However, analysis of mRNA export shows only very minor defects (Figure 2c and Figure S2) and mutants that have stronger mRNA export defects like rat7-1 show none of the morphological phenotypes that can be observed for a deletion of GLE2 ( Figure S2).
This involvement of Gle2 in cell cycle regulation is a new finding for S. cerevisiae and in accordance with data from higher eukaryotes that identified a role for Gle2/Rae1 in the microtubule organization, cell cycle regulation and prevention of aneuploidy , showing once more that basic principles are conserved in all eukaryotes.
Gle2 is involved in septin organization
Besides the interactions of Gle2 with cell cycle regulators, we found a novel interesting group of genes that are important for cell division that belong to the septin family and its regulatory network (Figure 1a). Drop dilution tests with mutants of all septins revealed that their combination with gle2Δ leads to significantly reduced growth compared with the single mutants ( Figure 3a). Strikingly, cdc10-1 gle2Δ double mutants show a drastic increase in the defects in morphology with about 10-fold elongated cells compared with wild type (Figure 3b and c), clearly indicating defects in entering isotropic bud growth and separation of mother and daughter cells. These data suggest that Gle2 might be important for septin-ring formation. To support our findings, we investigated physical interactions of Gle2 with the septin Cdc10 by co-IP analyses and found strong physical interactions (Figure 3d).
Because the interaction of GLE2 with CDC10 is quite strong on a genetic level and the two proteins show a very stable physical interaction, we analysed this aspect in more detail. Cdc10 together with Cdc3, Cdc11 and Cdc12 is one of the four main septins in yeast. Their ordered interaction leads to formation of hetero-octameric filaments that localize to the incipient bud site (McMurray et al., 2011). Over the course of budding the single filaments interact with each other and build a highly structured meshwork called the septin ring. This ring is necessary for correct bud formation 467 Gle2 and septin formation and cell division and represents a barrier between mother and daughter cell (Bi and Park, 2012).
To investigate if the formation of the septin ring would be affected by missing Gle2, we analysed the interaction between two septins in gle2Δ. Strikingly, co-IPs clearly showed a reduced interaction of Cdc10 and Cdc11 in gle2Δ cells (Figure 3e). In fact, quantification of several of these co-IPs revealed a~60% reduced interaction of the septins when GLE2 was deleted (Figure 3 f). These findings could suggest a direct function of Gle2 in assisting septin assembly.
The reduced septin interaction in gle2Δ could be a result of incorrect hetero-octamer formation itself or hindered multimerization of the filaments, or it might be that already assembled filaments are rather unstable in gle2Δ cells. Another possibility could be that the correct cue that triggers localized formation is missing. Therefore, we first investigated possible disturbance of the septin-ring localization by using GFP-tagged versions of septin proteins that allowed monitoring of the formation and localization of the septin ring during cell division with GFP microscopy. After synchronization with α-factor we took samples of wild-type and gle2Δ cells every 20 min. Septin rings become visible about 40 min after release ( Figure S3a and S3b) and reach their maximum after around 80-100 min. While in the wild type nearly all cells form a visible septin ring, <80% of the gle2Δ cells show this structure ( Figure S3a). In addition to the reduced amount in gle2Δ, the most striking difference from wild-type cells is the change in localization of septin rings. At 80 min after release, cells are in the middle of the budding event with a clearly distinguishable bud and the mother-budneck visible. In wild type the septin ring is located at the mother-daughter border, while this signal can be found prominently at the bud tip and not the bud neck in gle2Δ cells (Figure 3g). Not only for Cdc10, but also for Cdc11, this mislocalization is observed, indicating that the GFP-signal really represents the septin ring and not a defect in a single septin protein alone. This wrong localization of the septin ring to the bud tip in gle2Δ is not a rare event. Quantification of mislocalized septin rings in gle2Δ indeed revealed that around 50% of the cells show this defect (Figure 3h). This is highly significant compared with wild type in which <1% of the cells have a mislocalized septin ring. So when Gle2 is missing, the septin ring can either not assemble properly or cannot be maintained at its natural position at the mother-bud-neck.
Gle2 is involved in the Cdc42-mediated apicalisotropic switch
Gle2 is a WD40--propeller protein, a typical protein structure that recruits regulators such as kinases or phosphatases (Reddy et al., 2008). The organization of the septin filaments into a ring is tightly regulated and coupled with other checkpoints of the cell cycle (Bi and Park, 2012) and the lack of the correct placement of the septin ring in gle2Δ could be a reason for defects in regulation, which in turn might argue for a role of Gle2 as a scaffold for septin-regulating proteins.
The GTPase Cdc42 acts as a major regulator in cell cycle progression and morphology in all eukaryotes. In yeast, Cdc42 controls bud emergence, septin recruitment and the switch between apical and isotropic bud growth (Johnson, 1999). While its localization to the bud tip during bud emergence is required for bud growth, it is distributed along the daughter cell membrane in G 2 /M-phase (Bi and Park, 2012). This re-localization is triggered by the kinase Cdc28 (Johnson, 1999). Strikingly, both proteins show strong genetic interactions with gle2Δ ( Figure 4a). Moreover, Cdc42 and Cdc28 physically interact with Gle2 as shown by co-IPs (Figure 4b). Most importantly, we show that Gle2 is required for correct Cdc42 localization, as shown by synchronized cells. In wild type, Cdc42 is located at the bud tip with a peak at 50 min after release from the cell cycle arrest (Figure 4c and d). In contrast, in gle2Δ cells Cdc42 remains localized to the bud tips after 50 min and even after 120 min upon release (Figure 4c and d). Another apparent difference is the shape and length of the newly formed bud (Figure 4c). Quantification shows that cells lacking GLE2 form buds that are significantly elongated compared with wild type (Figure 4e). This phenotype could be a result of the prolonged stay of Cdc42 at the bud tip, which delays the apical-isotropic switch.
Together, our data have identified an involvement of Gle2 in cell cycle regulation and the APC-mediated chromosome separation, underlining that this function is highly conserved from S. cerevisiae to humans. Additionally, we found a novel function for Gle2 in septin organization that is important for cell cycle progression. As
468
G. Zander et al. this novel function presumably occurs before its APC-mediated function, it will be interesting to see how far they are connected. One could speculate that Gle2 as a WD40 repeat propeller protein might be a scaffold for septin-complex formation.
In this function it could provide a platform for proteins and complexes that regulate the bud emergence, growth and cell cycle in general.
Gle2 interacts not only with the GTPase Cdc42, but also with the kinase Cdc28/Cdk1 (Figure 4b), which as the Clb1-2/Cdc28 complex coordinates re-localization of Cdc42, important for the apical-isotropic switch in the daughter cell and entering of the G 2 /M-phase (Johnson, 1999). A localization of Clb2/Cdc28 to the bud neck has been shown previously (Eluere et al., 2012;Hood et al., 2001) and it is tempting to suggest that Gle2 supports this as a scaffold and allows a coordinated regulation of these processes. Whether Gle2/Rae1 impacts septin-complex formation in humans remains to be shown; however, owing to the fact that the septins are increasingly recognized as important components of the cytoskeleton and as such are involved in the organization of cytokinesis (Mostowy and Cossart, 2012), a function of Gle2/Rae1 in this process is most appealing.
Supporting information
Additional Supporting Information may be found online in the supporting information tab for this article. Figure 1C. (B) Western blot shown in Figure 3D. (C) Western blot shown in Figure 3E. (D) Western blot shown in Figure 4B. | 2018-04-03T02:19:29.647Z | 2017-08-04T00:00:00.000 | {
"year": 2017,
"sha1": "95b91cd7445911808415116af54391be71562710",
"oa_license": "CCBYNC",
"oa_url": "https://onlinelibrary.wiley.com/doi/pdfdirect/10.1002/yea.3249",
"oa_status": "HYBRID",
"pdf_src": "Wiley",
"pdf_hash": "42a5cea886422965ab8c9f6bc85b08f434f0b45d",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
118691015 | pes2o/s2orc | v3-fos-license | Measurement of Galactic Logarithmic Spiral Arm Pitch Angle Using Two-Dimensional Fast Fourier Transform Decomposition
A logarithmic spiral is a prominent feature appearing in a majority of observed galaxies. This feature has long been associated with the traditional Hubble classification scheme, but historical quotes of pitch angle of spiral galaxies have been almost exclusively qualitative. We have developed a methodology, utilizing two-dimensional fast Fourier transformations of images of spiral galaxies, in order to isolate and measure the pitch angles of their spiral arms. Our technique provides a quantitative way to measure this morphological feature. This will allow comparison of spiral galaxy pitch angle to other galactic parameters and test spiral arm genesis theories. In this work, we detail our image processing and analysis of spiral galaxy images and discuss the robustness of our analysis techniques.
INTRODUCTION
Approximately 60% of galaxies in the local Universe are spiral (Buta 1989). A considerable number of these spiral galaxies show Grand Design structure, where the spiral pattern is uniform and spans the entire disc of the galaxy. In these galaxies, the spiral pattern is often logarithmic in nature (Seigar & James 1998), and so their appearance is scale independent. The best geometric measure for logarithmic spirals is the pitch angle, and this can be measured for any galaxy in which spiral structure can be discerned, independently of the distance to the galaxy.
It is worth noting that spiral structure has been observed for over 150 years 1 with no clear standard for quantitative measurement having emerged, even though it correlates well with other important features of galaxies, such as central supermassive black hole (SMBH) mass (Seigar et al. 2008). Furthermore, spiral arm pitch angle could serve as a means to discriminate between rival theories for the formation of spiral structure in galactic discs. The aim of this paper is to present one such method, based on a Two-Dimensional (2-D) Fast Fourier Transform (FFT) algorithm 2 , which decomposes images into spirals of different pitch angles and numbers of arms.
A long-standing and quite successful theory of spiral structure in galaxies is the quasi-stationary density wave model (Lin & Shu 1964). As gas enters this density wave, it is compressed to a density at which stars can form (Roberts 1969;Shu et al. 1972). These star forming regions, along with stars and gas, conglomerate together into spiral arms in the disc regions of spiral galaxies, with star forming regions found on the leading edges of arms, and dust (in the form of dust lanes) seen on their trailing edges.
A recently formulated rival theory proposes that spiral arms are composed of identifiable groups of stars in highly eccentric and chaotic orbits, which originate near the ends of galactic bars. These orbits, though chaotic, keep the stars grouped in relatively narrow tubes known as manifolds, which are responsible for the observed spiral structure (Athanassoula et al. 2009b(Athanassoula et al. ,a, 2010. According to this theory, galaxies with stronger bar poten-tials should have more open spiral structure. In a recent study of 27 galaxies, Martínez-García (2012) found that ≈ 60% of galaxies corroborate this theory and that galaxies in which the spiral arms maintain a logarithmic shape for azimuthal ranges greater than 70 • seem to corroborate the predicted trend. Spiral galaxies are classified into three main types of spiral structure: grand design, flocculent (Elmegreen 1981), and multi-armed. Grand design spirals are welldefined two-armed galaxies and theoretical efforts have naturally focused on explaining these very striking patterns. Flocculent spirals are less regular with sporadic spiral arm segments. It has been proposed that the origins of this kind of spiral are quite different from grand design spirals, the products of stochastic self-propagating star formation being acted upon by the differential rotation of the disc to create segments with the appearance of spiral arms (Seiden & Gerola 1982). Multi-arm spirals have distinct spiral arms, not necessarily symmetrically spaced. It is likely that they formed as a result of galaxy harassment (frequent high speed galaxy encounters within clusters; Moore et al. 1996). Spiral arm generation from external forces has been proposed to explain the genesis of multi-arm spiral structure in our own Milky Way (e.g., Purcell et al. 2011).
In general, logarithmic spirals are good approximations of the shape of galactic spiral arms (Seigar & James 1998). Logarithmic spirals are defined in polar coordinates as r = r 0 e θ tan (φ) (1) where r is the radius, θ is the central angle, r 0 is the initial radius when θ = 0 • , and −90 • ≤ φ ≤ 90 • is the pitch angle. The limits for the absolute value of the pitch angle are 0 • and 90 • , which produce a circle and a line, respectively. Pitch angle is defined as the angle between the line tangent to a circle and the line tangent to a logarithmic spiral at a specified radius. Small pitch angle absolute values are associated with tightly wound spirals and high absolute values with loosely wound spirals. The sign of the pitch angle indicates the chirality of the spiral, with positive pitch angles indicating clockwise outward winding and negative pitch angles indicating counterclockwise outward winding (as seen from a given observer's position, i.e., above or below the galactic plane). In this paper we present a method for determining reliable galactic spiral arm pitch angles. Given sufficient quality images, our software can reliably measure pitch angles by iterative 2-D FFT analyses. The paper is outlined as follows: §2 describes the observations of the images we use and our procedure to prepare those images for measurement through our software. §3 details the need for and the nature of our iterative adaptation to the FFT software, along with interpretation of its results. §4 discusses how we determine errors on our measured pitch angles. §5 describes our image analysis and related tools for further image refinement and evaluation. Finally, in §6 we present a discussion of our results and a few possible applications of the code.
OBSERVATIONS AND DATA ANALYSIS TECHNIQUES
The galaxy images we use in this paper (unless mentioned otherwise) come from the Carnegie-Irvine Galaxy Survey (CGS 3 ; Ho et al. 2011). This is a statistically complete, magnitude-limited sample of 605 bright (B T < 12.9 mag), Southern (δ < 0 • ) galaxies observed using the SITe2k CCD camera (with a pixel scale of 0.259 pixel −1 ) on the 2.5 m du Pont telescope at the Las Campanas Observatory in Chile. The overall quality of the images is high, both in terms of resolution (median seeing ∼ 1 ), field-of-view (8.9 × 8.9 ), and depth (median limiting surface brightness ∼ 27. 5, 26.9, 26.4, and 25.3 mag arcsec −2 in the B, V, R, and I bands, respectively). All CGS images have been oriented to have up as North and left as East (before we subsequently rotate images for deprojection purposes). In this paper, we use a subset of galaxies from the CGS sample in order to test our methods. For images not included in the CGS sample, we use the NASA/IPAC Extragalactic Database (NED) 4 to acquire images.
Two-Dimensional Fast Fourier Transformations of
Galaxy Images A program called 2DFFT (Saraiva Schroeder et al. 1994) accomplishes the 2-D FFT decomposition of images. 2DFFT itself is an adaptation of the FOURN routine from Numerical Recipes in C (Press et al. 1989) that deals with CCD (charge-coupled device) images. The program is intended to analyze face-on or deprojected galaxy orientations. The decomposition is modeled on logarithmic spirals. As pointed out by Considere & Athanassoula (1988), this method does not assume that observed spiral structures are logarithmic. It only decomposes the observed distributions into a superposition of logarithmic spirals of different pitch angles and number of arms, which can be thought of as building blocks. This is analogous to the usual Fourier method of decomposing signals into a superposition of sinusoidal functions of different frequency. As per Puerari et al. (2000), the amplitude of each Fourier component is given by (2) where u ≡ ln r, r (radius) and θ (central angle) are in polar coordinates, r min is the inner radius, r max is the outer radius of the user-defined calculation annulus, and D is a normalization factor written as I(u, θ) is the distribution of light of a given deprojected galaxy, in a (u, θ) plane, m represents the number of arms or harmonic modes, and p is the variable associated with the pitch angle (φ), defined by with p max being the value of p with the highest amplitude for a given harmonic mode (see Figure 1). As currently defined, the code calculates Equation 2 for 0 ≤ m ≤ 6. Fig. 1b (right) -A(m, p) values for a deprojected B-band image of NGC 5054 with a measurement annulus defined by an inner radius of 160 pixels (41.4 ) and an outer radius of 508 pixels (132 ). This indicates a peak in the three-armed spiral harmonic mode at pmax = 3.50. The equivalent single value pitch angle via Equation 4 is −40.60 • (Note -subsequent sections and figures will revise this measurement with improved methods).
Additionally, the code reports a phase angle (Φ) for the orientation of the spiral arm pattern, calculated as where An important step in measuring the pitch angle of a galaxy, regardless of the method, is to deproject the galaxy to a face-on orientation. This process assumes that a galaxy with the plane of its disc parallel to the plane of the sky will be circular. A circular galaxy with random inclination appears on the sky as an ellipse. Thus, a circular galaxy can be described by its position angle (P A; orientation of the semi-major axis in degrees East of North) and its axis ratio. In turn, the axis ratio can be further incorporated to characterize the angle of inclination (α) from the plane of the sky defined by where a is the semi-major axis and b is the semi-minor axis. Thus, an inclination angle of 0 • and 90 • describes a face-on and an edge-on galaxy, respectively. The position angle and axis ratio can be determined easily from images using various programs, e.g., SExtractor (Source Extractor; Bertin & Arnouts 1996) or the ELLIPSE routine in IRAF 5 (Tody 1986;Jedrzejewski 1987). ELLIPSE works 5 IRAF (Image Reduction and Analysis Facility) is distributed by the National Optical Astronomy Observatory (NOAO), which by iteratively fitting isophotes interactively to a galaxy image and reporting various parameters; most importantly, position angle and ellipticity (1 − (b/a)).
With the position angle and axis ratio well-defined, the galaxy can be readily deprojected. This is accomplished by rotating the image of the galaxy (see Figure 2a) by −P A (see Figure 2b) and then stretching the x-axis by the a/b axis ratio (see Figure 2c). We use the IRAF routines ROTATE and MAGNIFY to accomplish the image rotation and stretching, respectively. This procedure effectively aligns the semi-major axis of the galaxy with the y-axis on the image and then stretches the semi-minor axis to an equal length as the semi-major axis, thus creating one unique radius for the galaxy and turning what was an ellipse into a circle. The process of deprojection is conducted in order to minimize errors in the resulting measurement of pitch angle. However, as discussed later in §4.1, precise deprojection is not necessary for the measurement of the pitch angle. Deprojection increases the range of galactic radii over which valid pitch angles may be measured, and thus decreases error in those measurements. As a result, our assumption of galaxies being intrinsically circular is not especially critical to the measurement of spiral arm pitch angle. In the case of highly inclined (i.e., nearly edge-on) galaxies, much of the spiral arms are hidden from sight and recovery of the intrinsic geometry via deprojection becomes increasingly difficult with higher inclination angles. However, we have had success with deprojection on galaxies up to α = 77.47 • for the case of IC 4831. Of course, the ability to exis operated by the Association of Universities for Research in Astronomy (AURA) under cooperative agreement with the National Science Foundation (NSF). tract meaningful information from a deprojected image will strongly depend on the resolution of the image. For images of low to moderate resolution, it is unlikely to be able to meaningfully analyze galaxies with α > 60 • .
Image Cropping
After deprojection, the next step is to determine the center of the galaxy. We assume that the center of the galaxy is the brightest region within the galactic nucleus using the IRAF routine IMCNTR to determine the brightest pixel location within a specified search region. The apparent center of a galaxy is strongly affected by the interstellar extinction. As a result, different wavebands may yield slightly different center coordinates based on this routine. We have conducted a test of 10 randomly selected galaxies from the CGS sample and have determined that, on average, the positions of the central coordinates vary by a distance of 1.986 pixels (0.514 ) between B and I band images. This insignificant discrepancy is made even less important by further findings in §4.1, which show that measurement of pitch angle does not critically depend on location of the precise center of the galaxy. The IMCNTR-determined coordinates are then adopted as the center of the galaxy. The image is then cropped about the galaxy, with the center of the galaxy as the center of the cropped image (see Figure 2c), and the resulting image made to be a perfect square, as required by the 2DFFT code, with oddnumbered pixel-sized edges in order to allow a unique median pixel location as the exact center of the image 6 .
Star Subtraction
FFT image analysis is widely used for its mimicry of the human eye's ability to pick out symmetries and repetitions even in noisy or cluttered images. The 2DFFT code can measure pitch angles, in spite of the presences of many non-spiral features in a galaxy's image. An interesting feature is that though the code models the spiral 6 As required by the 2DFFT code, the input image file must be in the form of a text file. We use the IRAF routine WTEXT to convert .fits images to .txt files.
image as a superposition of spirals of different numbers of arms and different pitch angles, it measures the correct pitch angle even for harmonic modes where the number of spiral arms is incorrect. Thus, for a sufficiently lownoise image, it is not even necessary to correctly infer the number of spiral arms in the galaxy in order to accurately measure its pitch angle.
When noise is introduced, this agreement in pitch angle measurement between the different harmonic modes is the most obvious casualty. One important source of noise is the presence of bright foreground stars, especially when they are superimposed on the disc of the galaxy itself (see Figure 3). Nevertheless, the 2-D FFT of the harmonic mode with the correct number of arms (most commonly, m = 2) will still usually give a stable value, which seems to correspond to the correct pitch angle even when other harmonic modes show no reliable measure. In order to increase confidence in our pitch angle measurement, we reduce the noise by subtracting the foreground stars. Since the IRAF ELLIPSE function does not always work when analyzing star-subtracted images (it fails to start if light from the center of a galaxy has been removed and it cannot locate the center), it is best to measure the ellipticity before performing star subtraction 7 . Often, this results in several of the harmonic modes coming into at least rough agreement with the one harmonic mode previously selected as the best single example. This suggests that foreground star contamination is a leading source of noise in the images and that star subtraction is a useful step. At the same time, it may not always be required for an accurate measurement. This is corroborated by Martínez-García (2012), who finds from a study of 27 galaxies, that the presence of foreground stars does not affect the value of pitch angle in general.
Image Measurement
Following pre-processing, the first step is to specify an inner and outer radius of the galactic disc. The inner radius is the location where the spiral arms begin, i.e., where the galactic bar or bulge terminates; the outer radius is the location where the spiral arms cease, usually the outermost visible radius of the galaxy. Thus, an annulus is defined within which the 2DFFT code conducts its Fourier Decomposition. We then take this previously established procedure a step further by automating the code to measure many annuli so the final quoted pitch angle is not determined solely by one user-determined annulus. Our modifications and motivations behind the modifications to the code are detailed in the following section.
PITCH ANGLE AS A FUNCTION OF INNER RADII
The greatest source of human error lies in choosing an inner radius. Whereas it is seemingly easy for the user to visually identify the edge of the galaxy (i.e., the outer radius), it is significantly more difficult for the user to accurately specify the cessation of the bar/bulge feature of a galaxy (i.e., the inner radius). Furthermore, slight error in specification of the outer radius has little ill effect, whereas slight error in specification of the inner radius may have significant effect. To illustrate this result, consider different values in outer radii: underestimation results in the full length of spiral arms not being measured; overestimation results in the sky being measured at the edge of the spiral arms. Since our FFT computations are luminosity biased, sky inclusion does not significantly affect the computations. On the other hand, consider different values in inner radii: overestimation results in the full length of spiral arms not being measured; underestimation results in a bright bar/bulge feature being measured in addition to spiral arms, this last case being the worst possible scenario.
As a result of this observed sensitivity to inner radius selection, we run 2DFFT iteratively at different inner radii. This allows the user to specify an outer radius and calculate pitch angles for all possible inner radii within the defined outer radius. A scripting utility has been created to calculate pitch angles at all possible inner radii, given an outer radius, for a galaxy (see Figure 4). Additionally, we have modified the memory allocation of the original code to allow for input image sizes up to 2048 × 2048 pixels.
The inner radius is a numerical artifact which should not affect the measurement of pitch angle. Therefore, we seek a harmonic mode in which we find a range of inner radius over which the measured pitch angle appears to be the most stable and consistent with the observed appearance of the galaxy. We aim, typically, for a measurement of pitch angle with an associated error of 2 • to 4 • . The resulting pitch angles can be plotted vs. inner radius in order to visually identify stable pitch angle regions as a function of inner radius beyond the influence of a potential bar or oblate bulge feature. Stable regions are selected by several criteria; the stable region must be of the same sign (chirality) as the observed spiral arm windings in the image, it should be of the same harmonic mode as the visually observed number of spiral arms, there must not be any erratic fluctuation in pitch angle, and the region of stable pitch angles must be contiguous. In certain cases, the resultant pitch angle agrees in multiple harmonic modes, therefore, selection of the harmonic mode is not critically important. This allows us to focus on picking a stable pitch angle, even when the correct m value is ambiguous. To understand the code's behavior, we have conducted tests with very low noise images, artificially created logarithmic spirals. Not surprisingly, the code finds it trivial to measure the Run 2DFFT at inner radius, r min = 1 and outer radius, rmax pitch angle of such an image. For these synthetic spirals (see Figure 5), the apparent stable regions are easily perceptible. However, in real galaxies, careful image inspection and other techniques (see §5 and its subsections) are sometimes required to pick out more visually elusive stable regions amidst the range of harmonic modes available.
To be clear, our method does not avoid having to inspect each image individually. This is a necessary and not totally undesirable requirement. Computer vision methods are currently under development to measure galactic spiral arm pitch angle (e.g., Davis & Hayes 2012). The human eye has proven itself as the most reliable tool for the geometric classification of galaxies (e.g., Galaxy Zoo; Lintott et al. 2008). The human operator is required only to inspect the image qualitatively for signs of gross error, not to re-perform any measurements quantitatively.
An unavoidable side effect of deprojection is an oblate distortion of a galaxy's nuclear region. Initially, a nuclear galactic bulge is considered to be spheroidal. Therefore, a perfectly face-on galaxy (no deprojection required) with a nuclear bulge does not hinder the selection of an inner radius. This can be seen in the case of a synthetic logarithmic spiral with a nuclear bulge component added (see Figures 5c and 5g). Since a galaxy is deprojected according to the outer region of the galaxy and not the inner region, the nucleus can be distorted to an oblate spheroid. This creation of a non-spherically symmetric feature can negatively affect the calculated pitch angle in the innermost regions of a galaxy.
The largest and likeliest source of error due to inner radii determination is when barred galaxies are measured. Galactic bars are linear features and therefore have high pitch angles (φ 90 • ). Inclusion of a high pitch angle feature into the measurement annulus of the 2DFFT code results in a significant overall biasing of the resulting pitch angle towards the high side. This is always the case, because the highest practical limit for spiral arm pitch angle is significantly lower than the pitch angle produced by a galactic bar. The effect of galactic bars are illustrated in the example of a synthetic twoarmed spiral with a bar component added (see Figures 5d and 5h).
ERROR DETERMINATION
The most obvious error is the variance about the mean pitch angle over the selected stable region in inner radii. The error is found by calculating the mean and standard deviation of the sample of pitch angles over the selected stable region. This standard deviation of pitch angle over the selected stable region is then weighted by the length of the stable region compared to the total length from the innermost spiral structure to the edge of the galaxy. Based on our observation from running synthetic logarithmic spirals through our code (see subsequent subsections), reliable pitch angles are not measurable for inner radii selected beyond ≈ 90% of the selected outer radius. At this point, too much information has been ignored for the code to accurately measure a pitch angle.
In addition, it is important to consider the resolution of the 2DFFT code due to a discrete step size (see Figure 6). 2DFFT captures −50 ≤ p ≤ 50 values in discrete steps of 0.25 for six harmonic modes (1 ≤ m ≤ 6). Therefore, only discrete values of pitch angle are produced by the subsequent conversion of p → φ. The step size of the discrete Fourier transform is the analog of the frequency stepsize in One-Dimensional (1-D) discrete Fourier Transforms, the smallest measurable frequency. This leads to a necessarily higher precision in the lower regime of pitch angle absolute values and in the higher order harmonic modes. The quantized error of the mean pitch angle due to the resolution of the code (see Figure 7) is added in quadrature to the previously determined standard deviation of the mean pitch angle to give a total error. The final error is therefore where E φ is the total pitch angle error, m is the quantized error for the dominant harmonic mode, σ is the standard deviation about the mean pitch angle, β is the distance (e.g., in pixels) from the innermost stable spiral structure (i.e., beyond the influence of a bulge or bar) to 90% of the selected outer radius of the galaxy (0.9r max ), and λ is the length (in the same units as used for β) of the stable range of radii over which the pitch angle is averaged. Figure 8 serves as a good example of our error determination and its subsequent reduction by the use of star subtraction. Equation 7 reflects the fact that in our method, a human researcher rather than a computer makes the final selection of pitch angle. That is, a balance of two main principles governs Equation 7: the fluctuation across and the length of a chosen stable region of pitch angle as a function of inner radius. Our process ensures that the error about the mean pitch angle is appropriate, based on the choices made by the user. For example, an erratic "stable" region or a short stable region will both be punished with appropriately high errors. Thus, a careful selection of stable region is required so as not to produce substantial errors.
Inclination Angle and Galactic Center Position
Errors The problem of making a poor choice of inner radius was addressed by altering the code so that it calculates a pitch angle for all possible inner radii. Other user-defined parameters have little impact on our results, and thus do not require such measures. The most important step is deprojection, which requires the user to measure the galaxy's inclination angle, presuming that the galaxy's disc is inherently circular. Tests with a synthetic twoarmed spiral with pitch angle of −20 • (see Figure 5b) demonstrate that measurement of pitch angle is correct for any even number of arms and for inner radii up to ≈ 90% of the outer radius (see Figure 5f). For a one-armed synthetic spiral, all harmonic modes are in agreement (see Figure 5e).
When the synthetic two-armed spiral (see Figure 5b) is shrunk along one axis incrementally to simulate an in- creasingly inaccurate deprojection, the results show that there is still a stable region of inner radii with the correct measure of pitch angle (see Figure 9a). Similarly with a real two-armed galaxy, NGC 5247 (see Figure 10b), it is of interest that an incorrect choice of inclination angle merely causes a gradual reduction in the length of the stable region over which the selected inner radii yield the correct pitch angle (see Figure 9b). Thus, deprojection is still an important step, but is unlikely to be a significant source of error when using the script, which calculates pitch angle for a wide variety of possible inner radii.
Similarly, when choosing the center of the galaxy image, tests with a synthetic two-armed spiral (see Figure 5b) and a real two-armed galaxy (NGC 5247, see Figure 10b) suggest that incremental errors in centering only gradually reduce the stable region without affecting the actual measure of pitch angle (provided the stable region of roughly constant pitch angle remains lengthy enough to be found, see Figure 11). Overall, these tests are a testament to the robustness of the 2DFFT algorithm.
Bulges and Bars
Our synthetic two-armed spiral was also used to study the effects of circular bulges and bars in galactic nuclei on pitch angle measurements. When a circular bulge component is added to the synthetic two-armed spiral (see Figure 5c), the even numbered harmonic modes are unaffected, whereas the odd harmonic modes are systematically different with the lower harmonic modes being the worst (see Figure 5g). In contrast, when a bar component is added to the synthetic two-armed spiral (see Figure 5d), the resulting value of the measured pitch angle is significantly increased at inner radii, with the correct pitch angle value returning after the inner radius is beyond the extent of the bar (see Figure 5h). As an example, NGC 1365 (see Figure 12) displays a similar bar to the two-armed synthetic spiral with a bar added. These results confirm that circular bulges should not affect pitch angle, whereas the barred geometry can significantly bias pitch angle measurements towards higher values. We are therefore confident in the necessity of our efforts to systematically exclude barred nuclei from the pitch angle measurement annulus.
Problems with Underlying Presumptions
So far we have presumed that the pitch angle of a logarithmic spiral is a meaningful quantity to measure in images of disc galaxies. Certainly there are very many disc galaxies for which logarithmic spiral patterns are the most obvious feature of the disc, as the human eye perceives it. Nevertheless, two important objections might be made concerning the measurement of pitch angles as a useful characteristic of galaxies. One is that the pitch angle may be different for the same galaxy when viewed at different wavelengths. The other is that the pitch angle might vary with the radius of the disc, in other words that the spiral is not truly logarithmic.
The Effect of Wavelength on Pitch Angle
It is important to consider the possibility of different pitch angles arising in different wavebands of light and what physical processes that might imply. For instance, optical B-band images tend to trace the bright massive star forming regions of a galaxy and near-infrared (NIR) images tend to trace the old stellar populations in galaxies (Seigar & James 1998;Eskridge et al. 2002). The old stellar population traces the spiral density wave (Seigar & James 1998). Furthermore, a spiral that appears flocculent in the B-band may appear to have a weak grand design spiral in the near-infrared (Thornley 1996). Kendall et al. (2011) used a 1-D FFT analysis on optical and NIR images of grand design spiral galaxies to measure their m = 2 pitch angles and concluded that a good correlation exists between galaxies being grand design in the infrared and in the optical. Seigar et al. (2006) demonstrates that a 1:1 relation exists between the B and NIR band pitch angles for a sample of 66 galaxies from a combination of the CGS (Ho et al. 2011) and the Ohio State University Bright Spiral Galaxy Survey (OSUBSGS; Eskridge et al. 2002). Alternatively, Grosbol & Patsis (1998) propose a contrary view. They find a systematic trend of arms being tighter in bluer colors by investigating five galaxies in B, V, I, and K filters. Admittedly, two of their five galaxies are tight spirals for which little or no change in pitch angle is observed, but it seems that more work with multiple filters is required.
Using our method, we have remeasured a subset of 47 of the galaxies appearing in Seigar et al. (2006, they used an earlier version of this method) and have also identified a seemingly 1:1 relation (see Figure 13 and Table 1). Therefore, despite seemingly small-scale differences between spiral arms in different wavelengths of the optical-NIR spectrum, the overall structure of the spiral arms, Fig. 8a (left) -a stable mean pitch angle of −17.52 • is determined for the m = 3 harmonic mode from a minimum inner radius of 36 pixels (9.32 ) to a maximum inner radius of 208 pixels (53.9 ), with an outer radius of 258 pixels (66.8 ). This stable region of 172 pixels (44.5 ) occupies 67% of the galactic disc. Equation 7 yields E φ = 3.17 • with λ = 172 pixels (44.5 ), β = 196 pixels (50.8 ), σ = 2.75 • , and 3 = 0.47 • . The final determination of pitch angle is therefore −17.52 • ± 3.17 • . Fig. 8b (right) -a stable mean pitch angle of −17.98 • is determined for the m = 4 harmonic mode from a minimum inner radius of 9 pixels (2.33 ) to a maximum inner radius of 235 pixels (60.9 ), with an outer radius of 264 pixels (68.4 ). This stable region of 226 pixels (58.5 ) occupies 86% of the galactic disc. Equation 7 yields E φ = 2.61 • with λ = 226 pixels (58.5 ), β = 229 pixels (59.3 ), σ = 2.56 • , and 4 = 0.35 • . The final determination of pitch angle is therefore −17.98 • ± 2.61 • . This result is barely different from the result without star subtraction; the main difference is the redetermination of the dominant harmonic mode. The percent difference in mean pitch angle is 2.59% with a 17.67% reduction in error from the original. and thus the proposed density wave, is consistent across the optical-NIR spectrum. This is in opposition to the prediction of the density wave theory that different pitch angles are expected for spirals when observed in different bands (Hozumi 2003). Although, the expected difference in pitch angle across wavelength is probably small enough that an extremely high precision fit would be necessary to falsify this prediction of density wave theory.
From our experience, we have become accustomed to preferring B-band images in general due to their characteristic clarity of galactic stellar components. However, our comparison of pitch angles in different wavebands has convinced us that we can typically measure pitch angle across a wide range of electromagnetic wavelengths. In that regard, we have successfully measured pitch angles of galaxies in the extreme cases of far-ultraviolet and 21 cm radio wavelength images when no other imaging data was available.
Variable Pitch Angle with Galactic Radius
Occasionally, spiral arms may appear to change pitch angle in the outer region of the disc, sometimes discontinuously. These are more the exception than the rule and we have generally preferred to measure the inner part of the disc in such cases or use more elaborate processing methods (see §5.1) to mitigate the severity of pitch angle variability. Considering the case where a dichotomy ex-ists between the pitch angles measured in the inner and outer regions of a galactic disc, the code can be made to run iteratively for two separate regions of the galaxy and average the results to yield an average pitch angle for the disc. However, if the pitch angle results are subsequently used for building relationships to processes in the nucleus of a galaxy (e.g., Seigar et al. 2008), pitch angles for the innermost portion of a galaxy perhaps make the most physical sense and are furthermore not as susceptible to extragalactic interaction. It is also likely that the entire extent of a galaxy might not display logarithmic spirals. If so, our stable regions are selected to only highlight clearly logarithmic sections of spiral arms.
To illustrate the case of measuring pitch angles of interacting galaxies, we have selected perhaps the most famous case of interacting galaxies, M51 (see Figure 14a). M51 consists of M51a (NGC 5194) and its companion dwarf galaxy M51b (NGC 5195). Due to M51a's clear interaction with M51b, its well-defined spiral structure is seen to depart from regularity close to the companion. For this case, our typical method of measuring pitch angle across the entirety of the galactic disc knowingly samples the outer 40% of the galaxy, which is clearly seen to be disrupted (see Figure 14b). Just as in the case of iterative determination of pitch angle as a function of inner radius to omit interior regions, by alternatively selecting Fig. 9.-Inclination angle tests on pitch angle output for a synthetic two-armed logarithmic spiral (left) and a real two-armed galaxy (right). Fig. 9a (left) -Pitch angle results for different errors in inclination angle for the synthetic two-armed spiral in Fig. 5b. Even at a high degree of simulated inclination angle error, the mean pitch angles remain approximately the same despite a gradually shrinking stable region across inner radii. Fig. 9b (right) -Real two-armed galaxy inclination test using NGC 5247 (see Fig. 10b). Three angles of inclination are tested: Original -the galaxy before deprojection, SExtractor -incorporates the deprojection according to SExtractor -Center error tests on pitch angle for a synthetic two-armed logarithmic spiral (left) and a real two-armed galaxy (right). Errors of 10, 20, 30, 40, and 50 pixels from the previously determined center are used for both. Fig. 11a (left) -Pitch angle results for different errors in center determination for the synthetic two-armed spiral in Fig. 5b. As the error increases, the stable region gradually decreases, yet the approximate mean pitch angle remains about the same. Fig. 11b (right) -Real two-armed galaxy center test using a B-band image of NGC 5247 (see Fig. 10b) after deprojection (P A = 38.71 • & α = 25.18 • ) was performed. The same case is true for the real galaxy image; the mean pitch angle remains constant despite a decreasing stable region with increasing error. an outer radius interior to disrupted outer regions, we can confine our measurements to only the stable portions of M51a (see Figure 14c) and other similar galaxies.
Flocculence
Concerning the measurement of pitch angle of different types of spiral galaxies, flocculent spirals provide perhaps the biggest challenge. From our experience with flocculent galaxies, we find that our code most often finds them to have high-valued harmonic modes. Their characteristics can range from fragmented arms at best to chaos at worst. Elmegreen & Elmegreen (1987) defined a sys-tem of arm classes and descriptions to categorize spirals into categories with varying degrees of flocculence. They defined 12 arm classes (classes 10 and 11 are no longer in use) with 12 having the most orderly spiral structure and 1 the least. Galaxies with arm classes 1-4 are considered flocculent, and those with arm classes 5-12 are grand design.
We have subsequently ascertained the arm classes (based on blue images from the Palomar Observatory Sky Survey) given by Elmegreen & Elmegreen (1987) for the galaxies we have measured for this paper and listed all available arm classes in Table 2. We have also cre- ated two plots (see Figure 15) of pitch angle absolute value vs. arm class (see Figure 15a) and pitch angle error vs. arm class (see Figure 15a) in order to investigate possible dependencies on arm classes. No clear relationship can be found from either plot, thus measurement of flocculent spirals do not appear to be inherently less precise than grand design spirals. However, our method of measuring pitch angle is very much dependent on the visual inspection conducted by the user. When initially inspecting images of possible candidate spiral galaxies for subsequent pitch angle measurement, it is more natural to be drawn to grand design spirals. This selection bias can be seen in selection of our sample for this paper without prior knowledge of their arm classes. Of the 48 galaxies listed in Table 2, only 8 galaxies are classified as being flocculent, with the remaining galaxies all classified as being grand design spirals. It is our practice to only attempt pitch angle measurement on galaxies that display convincing evidence of definable spiral structure from image inspection. From our study of the CGS sample thus far, we have been able to convincingly measure pitch angles for 62% of the spiral galaxies we have examined; 17% of the galaxies were rejected due to their high angle of inclination and the remaining 21% were omitted due to a lack of discernible spiral structure (of this 21%, among those with arm classifications from Elmegreen & Elmegreen (1987), 60% were classified as flocculent).
IMAGE ANALYSIS
The sign of the pitch angle and the number of harmonic modes are very important for correct image analysis. As the pitch angle is calculated over all possible values of inner radii for a galaxy, it is not uncommon for pitch angle to vary drastically in different harmonic modes. Different harmonic modes will have different signs of pitch angle and even across one harmonic mode, sign changes may occur. The most apparent feature to the human eye, for discernable spiral arms, is the chirality of the spiral arms. As a result, harmonic modes that favor opposing chirality can immediately be ruled out after a quick visual inspection of the image.
For galaxies with visually distinctive spiral arms, it is simple enough to count the number of spiral arms by eye and adopt that number of arms as the correct harmonic mode. However, in flocculent galaxies or galaxies with galactic arm spurs, it maybe necessary to adopt other methods in selecting the correct harmonic mode. Typically, the harmonic mode with the largest region of stable pitch angle across inner radii is the most valuable for our purposes. Nonetheless, other aspects of the code can lend a hand in identifying the dominant harmonic mode. The easiest method is by plotting the amplitude of p max as a function of inner radius. This will help identify the harmonic mode with the strongest amplitude over the largest radial range of the galaxy. For NGC 5054 (see Figure 16b), the m = 3 harmonic mode is dominant (Block et al. 1999) over the outer 84.8% of the galaxy's radius. In many cases, several harmonic modes agree reasonably well as to the pitch angle. In addition, producing an image of the Inverse FFT of a harmonic mode can help visually identify the "correct" harmonic mode (see §5.2).
Symmetrical Component Significance
It is a likely possibility that all of the arms of a spiral galaxy, especially galaxies with multiple arms, might not be perfectly symmetric. This could be the result of tidal disruption, galaxy harassment, etc. Whatever the reason, slight imperfections should be handled by the robustness of the FFT. A common trend among galaxies we have analyzed is that some galaxies exhibit spiral arms which gradually tighten, or decrease in absolute value of pitch angle, toward the outer regions of the galaxy (see Figures 10a and 17a). For galaxies with drastically asymmetric spiral arms or arms which demonstrate variable pitch angle, we use the method of Elmegreen et al. (1992) to isolate the symmetrical component of a galaxy (see Figure 18) and then we perform a pitch angle determination on the symmetrical component (see Figures 17b and 19b). Symmetric parts of galaxies are illustrated by making images from successive rotations and subtractions. The procedure of Elmegreen et al. (1992) is (8) where for m ≥ 2, S m is the image displaying the m-fold symmetric part of a galaxy made from the original image F , and the subscript T stands for truncation, meaning that pixels with negative intensities are set to zero. For the case of a two-armed spiral galaxy, the S 2 image consists of all bright features in the original image that have equally bright features diametrically across the galaxy. This procedure highlights symmetric emission, such as spiral arm spurs, star formation regions, etc., but it introduces spurious absorption features. For example, if there is a dust lane in only one arm, then only the bright part of that arm will appear in both arms of the S 2 im- Fig. 14b (middle) -A stable mean pitch angle of 19.13 • is determined for the m = 2 harmonic mode from a minimum inner radius of 54 pixels (16.5 ) to a maximum inner radius of 229 pixels (69.8 ), with an outer radius of 653 pixels (199 ). This stretch of 175 pixels (53.4 ) occupies 27% of the galactic disc. Equation 7 yields E φ = 4.76 • with λ = 175 pixels (53.4 ), β = 534 pixels (163 ), σ = 1.54 • , and 2 = 0.78 • . The final determination of pitch angle is therefore 19.13 • ± 4.76 • . Due to the interaction with its companion galaxy, M51a shows a significant departure from a constant pitch angle in the outer regions of the galaxy. This is seen in Fig. 14a and at the noticeable sign change in this plot at an inner radius of 389 pixels (119 ). Fig. 14c (right) -A stable mean pitch angle of 16.26 • is determined for the m = 2 harmonic mode from a minimum inner radius of 54 pixels (16.5 ) to a maximum inner radius of 276 pixels (84.2 ), with an outer radius of 389 pixels (119 ). This stretch of 222 pixels (67.7 ) occupies 57% of the measurement annulus. Equation 7 yields E φ = 3.20 • with λ = 222 pixels (67.7 ), β = 296 pixels (90.3 ), σ = 2.36 • , and 2 = 0.57 • . The final determination of pitch angle is therefore 16.26 • ± 3.20 • . This alternate pitch angle measurement isolates the inner portion of the galaxy out to the clear break from constant pitch angle seen in Fig.14b. As a result, the unstable outer portion of the galaxy has been ignored and a more accurate pitch angle has been determined for the purer inner structure of this interacting galaxy. Fig. 15a (left) -Pitch angles (black crosses) arranged into their arm classes with binned averages (red squares connected by red line segments). No clear trend is recognizable between pitch angle and arm class. Fig. 15b (right) -Pitch angle errors (black crosses) arranged into their arm classes with binned averages (red squares connected by red line segments). No clear trend is recognizable between pitch angle errors and arm class. (5) waveband/wavelength; col. (6) telescope/survey imaging source; and col. (7) arm class from Elmegreen & Elmegreen (1987). Source (1) CGS; source (2) Palomar 48 inch Schmidt; source (3) KPNO 2.1 m CFIM; and source (4) UK 48 inch Schmidt. a 103aE emulsion. b In addition to spiral arms in the disc of the galaxy, NGC 1097 displays rare m = 2 nuclear spiral arms in the bulge. These arms display an opposite chirality to the disc arms with φ = −30.60 • ± 2.68 • . c IIIaJ emulsion. Fig. 1b). However, the pitch angle can be seen to tighten (decrease) as the inner radius increases. Fig. 16b (right) -Plot of the amplitude of pmax as a function of inner radius for NGC 5054, indicating the m = 3 component as the dominant harmonic mode for the galaxy. The m = 3 harmonic mode is dominant from an inner radius of 77 to 456 pixels (19.9 to 118 ), constituting about 75% of the galaxy's radius.
age; this gives the false impression that there is a dust lane in the other arm also. This method appears to decrease our error estimates when performed. For galaxies with apparent initial symmetry, the mean pitch angle is not changed significantly; e.g., NGC 5247 (see Figure 19), the percent difference in mean pitch angle is 10.48% with essentially the same error. This tool seems most useful for galaxies that display variable pitch angle (see Figure 17). Error estimates can be reduced drastically for these cases; e.g., NGC 5054, for which the percent difference in mean pitch angle is 4.19% with a 72.03% decrease in error. Moreover, this process can act as an effective substitute for star subtraction.
Two-Dimensional Inverse Fast Fourier Transform
One of the most powerful tools provided by 2DFFT is the ability to run an Inverse FFT. After having deprojected the images and identified the dominant harmonic modes, we can calculate the inverse of the transforms according to Seigar et al. (2005). The inverse transform can be written as where S m (u) = D e 2u 4π 2 p+ p− G m (p)A(p, m)e ipu dp.
G m (p) is a high-frequency filter used by Puerari & Dottori (1992). For the logarithmic spiral governed by Equa-tion 4, it has the form This filter is also used to smooth the A(p, m) spectra at the interval ends (p − = −50 and p + = 50 with dp = 0.25) (Puerari & Dottori 1992). Equation 9 is designed as such, to allow the user to create an inverse transform for a selected number of harmonic components. For example, the inverse transform can be calculated for one component, e.g., m = 2, or any number of components can be combined to yield a composite result, e.g., m = 2, 3, & 4.
Once an Inverse FFT is created, it can be directly compared to the deprojected image of the galaxy, allowing us to effectively observe what the code is seeing. Figure 16a and Figure 20 show images of spiral galaxies overlaid with contours representing the results of Inverse FFTs of the same galaxy. The contours are the real part of the complex spatial function of Equation 9. The use of these images to analyze a galaxy can lead to more confident determination of pitch angle.
DISCUSSION AND FUTURE WORK
Our modified version of 2DFFT is a powerful tool for accurately measuring galactic spiral arm pitch angle. Our software, combined with careful image and data inspection, comparative pitch angle selection, and a selfregulating error determination allows for reliable pitch angle measurements. We hope that quantitative determinations of spiral arm pitch angle will aid in galaxy classification, in the indirect study of central black hole masses and more generally in our understanding of galac- Fig. 17a (left) -B-band pitch angle as a function of inner radius for NGC 5054, after star subtraction and deprojection (P A = 160 • & α = 53.84 • ) were performed (see Fig. 10a). A stable mean pitch angle is difficult to define since the pitch angle is seen to continually decrease from an inner radius of about 200 to one of about 425. A rough mean pitch angle of −24.52 • (a significant departure from the single value, non-iterative measurement of −40.60 • , see Fig. 1b) is determined for the m = 3 harmonic mode from a minimum inner radius of 123 pixels (31.9 ) to a maximum inner radius of 434 pixels (112 ), with an outer radius of 508 pixels (132 ). This stretch of 311 pixels (80.5 ) occupies 61% of the galactic disc. This measurement has a significant error due to the unstable pitch angle. Equation 7 yields E φ = 12.84 • with λ = 311 pixels (80.5 ), β = 334 pixels (86.5 ), σ = 11.92 • , and 3 = 0.92 • . The final determination of pitch angle is therefore −24.52 • ± 12.84 • . Fig. 17b (right) -For the pure symmetrical component of NGC 5054 (see Fig. 18a), the B-band pitch angle (same deprojection parameters as Fig. 17a) as a function of inner radius is far more stable. A stable mean pitch angle of −25.57 • is determined for the m = 3 harmonic mode from a minimum inner radius of 91 pixels (23.6 ) to a maximum inner radius of 253 pixels (65.5 ), with an outer radius of 593 pixels (154 ). This stretch of 162 pixels (42.0 ) occupies 27% of the galactic disc. Equation 7 yields E φ = 3.72 • with λ = 162 pixels (42.0 ), β = 443 pixels (115 ), σ = 1.31 • , and 2 = 1.00 • . The final determination of pitch angle is therefore −25.57 • ± 3.72 • , a percent difference of 4.19% in mean pitch angle with a 72.03% reduction in error from the original. tic morphology and its evolution. One important advantage of this means of describing galaxies is its relative ease of acquisition, since only imaging data is required to measure it. Also, it provides us with great opportunity to test competing theories behind galactic spiral arm genesis (Martínez-García 2012).
Comparison to Other Methods
Our method adds one dimension to recently published FFT methods (e.g., Kendall et al. 2011). In doing so, we are able to effectively use more of the inherent information in the images. Alternatively, 1-D methods identify radial peaks in intensity azimuthally about the galactic center by use of radial search segments that ultimately do not utilize the full resolution of the image. Admittedly, our method comes at a higher price in terms of computing power to analyze the full resolution of an image, but with modern computational power, this computational expense is trivial and is easily handled by modern processors. Ultimately, users of both 1-D and 2-D FFT methods are still obliged to visually inspect images. As for flocculent galaxies, 1-D FFT methods are admittedly only equipped to handle grand design spirals. Our 2-D FFT, though it may occasionally encounter trouble with high degrees of flocculence, will perform adequately with flocculent galaxies and with the support of additional image analysis methods (see §5 and its subsec-tions) and multi-wavelength imaging, it can confidently approach any galaxy with hints of spirality. Additionally, our pitch angle measurements are unique in that fact that we measure and quote pitch angles for multi-armed galaxies (dominant harmonic modes m > 2). Other researchers may have analyzed and discussed the influences from higher order harmonic modes, but in the end always publish pitch angles resulting from harmonic modes m ≤ 2.
In order to compare the results of our method to other published methods, we have chosen a well-studied sample of galaxies whose pitch angles have been determined independently, using different techniques in the literature. For this sample, we have selected the results of Martínez-García (2012), Kendall et al. (2011), Ma (2001, Grosbol &Patsis (1998), andKennicutt (1981) as references with which to compare our measurements (see Table 3). These five references provide a nice spread in measurement methods. Kennicutt (1981) used Hα photographic plates and subsequent by-hand geometric measurements of nearby Sa-Sc galaxies to calculate average pitch angles determined from the two main arms in each galaxy 8 ; Fig. 10a). Fig. 18b (right) -m = 2 symmetrical component (inverted color and with the same deprojection parameters as Fig. 10b) of NGC 5247 (see Fig. 10b). Grosbol & Patsis (1998) utilized accurate surface photometry and derivation of axisymmetric components to calculate m = 2 pitch angles for five galaxies from the residuals of intensive image processing and Fourier components of the azimuthal intensity variations; Ma (2001) visually selected points along spiral arms in CCD images of galaxies and fit logarithmic spirals to the points; Kendall et al. (2011) employed 1-D FFT decomposition to calculate m = 2 pitch angles for a sample of grand design spiral galaxies; and Martínez-García (2012) used both a "slope method" and adopted a 2-D FFT algorithm similar to our own, but without our analysis of pitch angle as a function of inner radius. Table 3 compares 38 of our pitch angle measurements against available measurements from five independent sources. The mean difference between the measurements for the same galaxies are as follows: ∆φ 1 = −5.13 • ± 19.41 • (the difference between our measured pitch angles and those of Martínez-García 2012) or ∆φ 1 = −0.11 • ± 7.38 • if the outlying measurement of NGC 4995 is disregarded, ∆φ 2 = 1.73 • ± 3.58 • (the same for Kendall et al. 2011), ∆φ 3 = 0.66 • ± 9.97 • (the same with the average arm measurement from Ma 2001), ∆φ 4 = 2.93 • ± 1.71 • (the same for Grosbol & Patsis 1998), and ∆φ 5 = 5.15 • ± 8.66 • (the same for Kennicutt 1981). These differences are comparable to the mean error in our measurements for this sample: φ E = 3.34 • ± 1.94 • .
NGC 7083
For individual measurements, several differences can be explained by our selection of a different harmonic mode from that chosen by the other group. For example, our measurement of NGC 7083 differs from the measurement by Grosbol & Patsis (1998); −19.44 • ± 3.21 • and −15.0 • ± 1.0 • , respectively. This can be explained by our selection of the m = 3 harmonic mode and their selection of the m = 2 harmonic mode (see Figure 21). In short, we believe this is a three-armed galaxy, not two-armed, a finding supported by the strength of our code's m = 3 harmonic mode and by visual inspection. It is of note that we find our measurement of the m = 4 harmonic mode's pitch angle to be −15.38 • ± 2.97 • (see Figure 21c), which coincides with the measurement of Grosbol & Patsis (1998). However, our measurement of the m = 2 harmonic mode is not possible due to its chiral instability. Despite the nice agreement of pitch angle between our m = 4 pitch angle and their m = 2 pitch angle, we find strong indications that the dominant harmonic mode is m = 3 (see Figure 21b). degrees compared to our measurement of 13.00 • ± 2.88 • . Their enormously high measured pitch angle is most sensibly unphysical, along with any measurement of φ > ∼ 60 • .
NGC 1365
A big discrepancy can be seen in comparison of our measurement of NGC 1365, −34.81 • ± 2.80 • , to the pitch angle absolute value measurements of Ma (2001), 13.8 • &17.8 • , andKennicutt (1981), 18 • ± 3 • . We find strong indications both from our code (see Figure 12) and visually (see Figure 22) that the pitch angle beyond the large bar is on the high side. In order to visually compare the fit of two logarithmic spirals with different pitch angles, the scaling must be adjusted. According to Equation 1, the radius of a logarithmic spiral with a higher pitch angle will grow much more rapidly than a logarith- Fig. 19.-Fig. 19a (left) -B-band pitch angle as a function of inner radius for NGC 5247 after deprojection (P A = 20 • & α = 28.36 • ) was performed (see Fig. 10b). A stable mean pitch angle of −28.76 • is determined for the m = 2 harmonic mode from a minimum inner radius of 1 pixel (0.259 ) to a maximum inner radius of 331 pixels (85.7 ), with an outer radius of 565 pixels (146 ). This stretch of 330 pixels (85.5 ) occupies 58% of the galactic disc. Equation 7 yields E φ = 5.73 • with λ = 330 pixels (85.5 ), β = 508 pixels (132 ), σ = 3.56 • , and 2 = 1.70 • . The final determination of pitch angle is therefore −28.76 • ± 5.73 • . Fig. 19b (right) -For the pure symmetrical component of NGC 5247 (see Fig. 18b), a stable mean B-band pitch angle (same deprojection parameters as Fig. 19a) of −31.94 • is determined for the m = 2 harmonic mode from a minimum inner radius of 9 pixels (2.33 ) to a maximum inner radius of 235 pixels (60.9 ), with an outer radius of 486 pixels (126 ). This stretch of 226 pixels (58.5 ) occupies 47% of the galactic disc. Equation 7 yields E φ = 5.75 • with λ = 226 pixels (58.5 ), β = 428 pixels (111 ), σ = 2.83 • , and 2 = 2.06 • . The final determination of pitch angle is therefore −31.94 • ± 5.75 • , a percent difference of 10.48% in mean pitch angle with essentially the same error as the original. As a characteristic example, NGC 5247 displays a similar pattern of agreement among even harmonic modes as the two-armed synthetic spiral with an added symmetrical bulge component (see Figure 5g). mic spiral with a lower pitch angle. In order to allow the radius of the φ = 16.5 • logarithmic spiral to grow at the same rate as the three higher pitch angle logarithmic spirals in Figure 22, we multiplied its resulting radius by a factor of 2.8. When optimally scaled, our high pitch angle measurement and their low pitch angle measurements can be brought into rough agreement. The difficulty in accurately measuring pitch angle increases as the amount of angular wrapping around a galaxy decreases, i.e., θ max π 2 for NGC 1365 whereas the spirals do not become significantly different until θ > ∼ 3π 4 at the edge of Figure 22. Furthermore, NGC 1365 is a well-documented case of variable pitch angle (Ringermacher & Mead 2009) with high pitch angle near the arm-bar junction and low pitch angle in the outermost regions of the galaxy. Our code correctly identifies this high pitch angle near the arm-bar junction and provides us with our desired innermost stable pitch angle of the galaxy (see §4.3.2).
Pitch Angle -SMBH Relation
Strong evidence suggests that SMBHs reside in the nuclei of most galaxies (Kormendy & Richstone 1995;Kormendy & Gebhardt 2001). Additionally, it has recently been shown that a correlation exists between the pitch angle of spiral arms and SMBH mass in disc galaxies (Seigar et al. 2008). The relation is such that more massive SMBHs reside in galaxies with low pitch angle spiral arms (i.e., those that are tightly wound) and the least massive SMBHs are found in galaxies with high pitch angle spiral arms. Given that a significant fraction of galaxies in the Universe have spiral or barred spiral morphologies (Buta 1989), we wish to improve upon existing methods for measuring spiral arm pitch angle in order to quantify their structure. This measure can in turn be used to estimate the central black hole mass.
One of the current widely used relationships to SMBH mass is stellar velocity dispersion of the bulge/spheroid (Gebhardt et al. 2000a;Ferrarese & Merritt 2000). This technique requires spectroscopy of the galactic nucleus. Pitch angle determination only requires optical imag- Fig. 21b (middle) -Plot of the amplitude of pmax as a function of inner radius for NGC 7083, indicating the m = 3 component as the dominant harmonic mode for the galaxy. The m = 3 harmonic mode is dominant from an inner radius of 145 to 384 pixels (37.6 to 99.5 ), constituting about 61% of the galaxy's radius. Fig. 21c (right) -A stable mean pitch angle of −19.44 • is determined for the m = 3 harmonic mode from a minimum inner radius of 143 pixels (37.0 ) to a maximum inner radius of 319 pixels (82.6 ), with an outer radius of 390 pixels (101 ). This stretch of 176 pixels (45.6 ) occupies 45% of the galactic disc. Equation 7 yields E φ = 3.21 • with λ = 176 pixels (45.6 ), β = 208 pixels (53.9 ), σ = 2.67 • , and 3 = 0.58 • . The final determination of pitch angle is therefore −19.44 • ± 3.21 • . Figure 12a) overlaid with a m = 2 spiral with φ = −34.81 • (solid red lines) representing our bestfit pitch angle measurement (see Figure 12b), φ = −32.01 • (short dashed red lines) representing our lower limit fit, φ = −37.61 • (long dashed red lines) representing our upper limit fit, and φ = −16.5 • (alternating short-long dashed blue lines) representing the average fit of Ma (2001) and Kennicutt (1981). ing. Compared to simple optical imaging, which is widely and readily available, spectroscopy is time intensive. Other methods such as Reverberation Mapping (Gebhardt et al. 2000b) require long-term campaigns to obtain multi-epoch spectra and require significant telescope time and allocation. Even black hole estimates from single-epoch spectra (Vestergaard 2002) require spectroscopy. Other techniques such as bulge luminosity estimates (Kormendy 1993;Kormendy & Richstone 1995;Magorrian et al. 1998;Marconi & Hunt 2003;Häring & Rix 2004) require bulge decomposition. One specific bulge luminosity estimate incorporates Sérsic Index (Sérsic 1963) measurements of elliptical galaxies and the bulges of disc galaxies and relates them to SMBH mass (Graham & Driver 2007).
Evolution of Pitch Angle with Redshift
No matter how a logarithmic spiral is scaled, pitch angle is unaffected. This allows pitch angle measurements for distant galaxies to be considered equally valid as those for local galaxies. For distant galaxies, as long as spiral arms are detectable, it is possible to measure a pitch angle. Unlike other methods, details such as distance, extinction, etc. do not need to be known in order to measure pitch angle. Additionally, the measurement of pitch angle is independent of cosmological assumptions.
Upon examining the GOODS (Great Observatories Origins Deep Survey; Dickinson et al. 2003) fields, we have identified 224 spiral galaxies with spectroscopic (Barger et al. 2008) and photometric (Wolf et al. 2004) redshift (z) data for GOODS North and South, respectively. Of these 224 spiral galaxies, 179 galaxies lie in the range z ≤ 1, 43 galaxies in the range 1 < z ≤ 2, and two galaxies with redshift greater than z = 2. So far, we have measured their pitch angles using the previous version of 2DFFT and are planning on remeasuring the sample using the new iterative version of the code. This work demonstrates that it is not uncommon to be able to measure pitch angle for galaxies beyond a redshift of one. Current work also includes artificially redshifting (Barden et al. 2008) this GOODS sample of spiral galaxies in order to test the completeness of the GOODS fields (Shields 2012). Artificial redshifting allows us to predict at what distance spiral arms are no longer visible and thus pitch angle is immeasurable. Results thus far show no indications of a relationship between pitch angle and redshift (Shields et al. 2010), but this matter will be further explored. | 2012-05-29T18:30:30.000Z | 2012-02-21T00:00:00.000 | {
"year": 2012,
"sha1": "9f5f303b564d65a2568cb8ea10dadaafc0e5e7ae",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/1202.4780",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "9f5f303b564d65a2568cb8ea10dadaafc0e5e7ae",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
59463028 | pes2o/s2orc | v3-fos-license | Planning Method of B6 in Kaunos Design Methods of a Rock-Cut Tomb in Southeast of Karia
The aim of this study is to identify the planning method of B6 in Kaunos. Assuming proportional relationships of each part were applied in the construction, as in the ancient Greek temples and stoas, the planning of B6 was investigated, leading to a conclusion that rough calculation of the scale has been determined first as a scheme design, followed by dimensional adjustment of each part in the execution stage. Further, the study identified that the planning is based on a grid technique using 3 feet as a grid, with proportional design and dimensional adjustments employed in the façade.
Introduction
It has been said that the ancient Greek architecture, as shown in the architectures of temples, stoas, etc., is highly formalized and consistent with predetermined style and structure according to the types of the buildings. However, Hellenistic tombs in the ancient Mediterranean world are so varied that it is said that there never existed the same form twice [1]. It is not hard to imagine that the varied architectural forms of tomb building during this period and a newly emerged value of accepting varieties have contributed greatly to Roman architecture that was to come later, in terms of, for example, the provision of the varied architectural languages and the universalization of freedom in choosing the languages. With 'the particularity among the ancient Greek architecture' and 'possibility of the contribution to the coming Roman architecture' of these Hellenistic tombs as the starting point, the author has been studying Hellenistic tombs, with the ultimate aim of systematically categorizing the tombs in order for them to be incorporated into the narrative of the history of architecture. However, as far as the author is aware, there is no study that dealt with Hellenistic tombs exhaustively other than those by Fedak [1] and the author himself. Comprehensive studies of Hellenistic tombs are at an early stage, and it is difficult to achieve the aim of system-atically categorizing the tombs in order for them to be incorporated into the narrative of the history of architecture. Therefore, based on the conclusion suggested by Fedak that 'a rich variety of appearance is the characteristic of Hellenistic tombs, however, clear tendencies and biases can be found in the forms, if chronological and geographical limits are imposed [1] the author has been investigating characteristics of the forms of Hellenistic tombs and the planning process of built-tombs [2][3][4][5][6][7].
In the previous paper the author investigated the rock-cut Greek temple style tombs [8] in southeast Karia ( Figure 1). The dates of construction of most of these tombs are unclear, partly because these rock-cut tombs stand high on a cliff face conspicuously with very few burial items remaining intact. Contrary to the general ancient Greek architecture that is ashlar masonry, these rock-cut tombs are cut out of cliff rocks usually at considerably elevated positions, using special construction methods. And knowing the architectural devices for con-structing a building at such a height or the knowledge on how the practical difficulties of constructing such rock-cut tombs affected the design decisions, even if the rock-cut tombs were built outside the Hellenistic era, may be useful in analyzing the rock-cut tombs of the Hellenistic era. With these points in mind, the constraint on estimated construction period as Hellenistic era was excluded from our discussion to focus on the rock-cut tombs that imitated ancient Greek temples and are situated in the southeast of Karia in the previous study, identifying a possibility that dimensional adjustment based on visual effect was carried out in these tombs.
Keeping the perspective, the present study mainly aims to analyze planning processes of the rock-cut tombs, and then, from the plan-ning point of view, discuss the possibility of having visual effect adjustment and the method of executing such planning. However, for want of space, this article focuses on Kaunos B6 in southeast of Karia ( Figure 2), and aims to identify how the original ground planning was carried out. B6 was chosen since it is the best conserved ex-ample among the of the planning method using the measurement data obtained from a 1/100 scale drawing; however, it is viable by the following reasons. Roos recorded several actual measurements in his report. For example, he reported the diameter of a small disc-shaped decoration under the capital of the antae as 21.5 cm. Meanwhile, when calculated using the afore-mentioned method suggested by the author, the diameter of the small disc-shaped decoration under the capital is 21.1 cm, showing a very small difference of 4mm from the 1/100 scale figure documented in Roos's report. Accordingly, this is thought to prove the highest accuracy of Roos's drawing, as well as the feasibility of acquiring good measurement data using the calculation method devised by the author. In addition, considering the characteristic of ancient Greek architecture of 'planning based on simple proportional relations' as mentioned in former studies [20], it is possible to analyze the planning method using data that allows a margin of a certain degree. In other words, 'proportional relations' in ancient Greek architecture usually refer to integer ratios, such as 1:2 and not 0.9:2.1. Above all, what must be presented here is a planning method based on a concept that is consistent throughout the whole plan, such as 'measurement of a part is used as the module to define all measurement' or 'dimensions of upper parts are all obtained as a result of the lower dimensions' . Therefore, even though there might be a slight discrepancy, it should not be a big problem, unless it is big enough to spoil the proportional relations which were supposed to be used in the original planning. Moreover, if the discrepancy is so big as to spoil the proportional relationship, we cannot deduce that a consistent and rational planning method was used. Further to this, the B6 tomb is relatively large in size, with larger parts. Therefore, for example, if a designer deter-mines a length of a part as 50 cm in relation to a 100 cm long part using a proportion of 2:1, and the measured length of the latter part constructed is 105 cm, we can get a proportion of 2:1 by calculating the ratio between 105 cm and 50 cm (105:50/2:1) where a discrepancy is 5 cm. The study concluded that the tomb was planned to use both proportional relations and a grid method, and a certain degree of discrepancy in measurement data is allowed also in the discussion regarding the grid method. When using the grid method, all parts, not just some part, must be determined in relation with the grid of equal intervals. In this case, it is unlikely that a discrepancy or discrepancies can accidentally place every part of the structure onto a grid with equal intervals. On the contrary, if all parts were deter-mined according to the grid method, even if some parts might have a certain degree of discrepancy, the overall structure would be placed on the grid, from which we can deduce the possibility of a planning process based on a grid method. Therefore, it can be said to be viable to use the 1/100 scale drawing to analyse the construction method even if the grid method planning had been conducted. As described above, in larger tombs such as B6, it is possible to explore whether proportional relations or the grid method were used or not, and if so, what their values were, even if there are some discrepancies in dimensional values, since the values themselves are rather large with minimal influence exerted by the discrepancy. Therefore, it is possible to analyse the planning method using the measurements taken from the 1/100 scale drawing.
One point to note is the difference between ashlar masonry buildings and the rock-face masonry buildings. Whether it is a temple, stoa or built-tomb, the ruins of an ashlar masonry building are usually buried underground and only exposed by excavation. In those circumstances, the condition of the ruin and components are generally good without much erosion by the weather. With this in mind, when the planning method of these ruins is analysed, the acceptable tolerance value between the design dimensions calculated using the planning method and the actual measurements is usually set rather small. On the other rock-cut Greek temple style tombs in southeast Karia, which makes it the preferred subject for the analysis, under the current situation where the planning method of these rock-cut tombs is still unknown. Fedak specified B6 as having been built in the Hellenistic era from its architectural form. Though it is risky to determine the date of construction purely by architectural form, if Fedak's judgment is to be believed, this present investigation can be positioned as a study into the planning method of Hellenistic tombs.
Materials and Methods
Multiple studies have been conducted on the planning methods of ancient Greek temples and stoas based on the Ten Books on Architecture by Vitruvius [8]. Though most of these studies are focused on individual buildings, Horiuchi [9] Hayashida [10][11][12][13][14] and Coulton [15][16][17][18] among others have studied the planning methods systematically with additional analysis of their own, and demonstrated that regular proportional relations between each dimension were used in temples and stoas in ancient Greek architecture. As far as the author is aware, no study has specifically examined planning methods of rock-cut tombs, however, the author's previous studies have found simple proportional relations between each dimension on the Lion Tomb at Amphipolis and the Nereid Monument at Xanthos, showing the possibility that planning methods using proportional relations were adopted for tombs in the Hellenistic period as well. So, in this study, the question whether a planning method using proportional relations was used in B6 is examined firstly by identifying any regular proportional relations between each dimension.
And of course, thinking of the time when the monument was actally built, each dimension must be expressed in the 'yardstick' of that time or the 'ancient measures' . After analysing the planning method using the identified proportional relations, design dimensions will be worked out by the planning method and the ancient measures, then the design dimensions and the actual measurements will be compared to verify the planning method suggested in the study. The an-cient measures is expressed in 'foot/feet' . In the past Dinsmoor reported that there were two kinds of 'ancient foot'; the Doric foot (1 foot=approximately 0.326m) and the Ionic foot (1 foot = approximately 0.294 m) [19]. However, in recent studies the ancient foot is not limited to these two measurements. Therefore, one foot in this study is assumed to be somewhere in the range of 0.294m to 0.330m as suggested in other studies. The smallest unit is called a dactyl, four times of a dactyl is called a palm and four times of a palm is a foot. Therefore dimensions of each part of the building need to be expressed as what fit into these units. Hayashida has also suggested the possibility that units of one third and one fifth also existed as well as dactyl, palm and foot. Since one third and one fifth are considered simple and basic divisional numbers, they are used in the analysis in this paper. In other words, the fractions are expressed with the denominators of 2, 3, 4, 5, 8, and 16 when converting into ancient feet. When analyzing, the dimensions of each part of the B6 tomb were taken from the Roos's report [20]. Roos's reported 'the aspect of facade' , 'size and shape of each part' , 'found artifacts' , 'types and examples' , etc. of rock-cut tombs in Kaunus in the southeast of Karia. Also included in the report are 1/100 and 1/50 scale drawings of the whole tombs as well as 1/10 and 1/5 scale detailed drawings of capitals and bases of the columns and the antae, for those tombs with exquisitely detailed designs or interesting features. Since the B6 tomb, a subject of interest, is quite large in scale, a 1/100 scale drawing of the whole tomb as well as 1/10 scale detailed drawings of the parts are avail-able. In this study, the 1/100 scale drawing and the 1/10 scale drawings created by Roos were scanned into CAD to obtain dimensions [21]. Some may wonder about the validity of the analysis
General Description of B6
The ancient city of Kaunos situated in the modern day city of Dalyan in southwest Turkey, across the river that runs at the west side of the city. B6 is carved out on the rock cliff facing the river and can be seen from the city of Dalyan on the other side. As stated ear-lier, a ruin of ashlar masonry buildings is usually buried underground and only exposed upon excavation. Rock-cut tombs, on the other hand, we have been exposed to the weather since ancient times, resulting in more erosion and deterioration. Considering this, B6 is in relatively good shape without many parts missing, except the acroteria ( Figure 3).
The tomb and its surroundings are separated from the rest of the rock face. The ground is rectangular, approximately 6.8×10.2m and the height from the foundation platform to the top of the acroterion is approximately 9.3m. The order is Ionian. The façade consists of 2 bands of foundation platform supporting 2 columns, flanked by a pair of anta, and which, in turn, support the architrave and the pediment on top. In other words, B6 is a rock-cut tomb that imitates an in-antis style Greek temple. However, the tops of the antae go through the lower fascia of the architrave, showing another difference from the formal in-antis prototype.
The sepulcher is located behind the pronaos and contains 3 catafalques positioned in three-pointed-star layout, not side by side. The catafalques are part of the tomb, not separate pieces. If equal-sized catafalques are positioned in a three-pointed-star layout, the floor hand, rock-cut tombs have been exposed to the weather for a long time, resulting in more deterioration compared to a temple, stoa or builttomb built in ashlar masonry. Further, with the technical difficulties of carving directly onto a rock face, accuracy in execution may have been reduced. These factors can cause larger discrepancy between the design dimensions derived from the planning method and the actual measurements. With the above consideration in mind, a yardstick value for such discrepancy was set using comparisons of measurements of a A symbol '*' indicates the dimension for a façade part. Places with this mark have different dimensions on the façade side and the backside. The façade side dimension is used in the study since it is natural to assume that more attention must have been paid on the façade side that was intended to be seen. • A symbol '**' indicates the scheme design dimension. shape of the sepulcher should be square. However, in B6, the catafalque at the left side of the entrance (in this article, left and right are always seen facing the façade.) is smaller than the rest, which makes the floor shape of the sepulcher a rectangle of approximately 3.2 × 2.6 m. The dimensions of each part are shown in the (C) column of the Table 1.
Setting the yardstick value for tolerance
As mentioned earlier, taking the technical difficulties of the construction on a cliff face as well as the severe erosion into consideration, it was decided to set yardstick values for tolerance. Rockcut tombs are carved out of a cliff face at an elevated position and an observer cannot see the back, sides and inside of a tomb. Because of this, there is a possibility that less accurate works might have been carried out at such places. In fact, in the B6 tomb, the back wall and the right-side wall do not make a right angle, which makes the depths widely different on the right and the left, as well as the widths of the tombs at the front and the back. Therefore, this study set two separate yardstick values for tolerance, one for the facade side which was intended to be seen, and another for the rest, including the back, the sides and the interior, which were not intended to be seen. For those parts which were not intended to be seen such as the back, the sides and the interior, the yardstick tolerance was set as follows using the values of the 'back wall thickness' of both on the right and left side; that is, (0.998-0.941)/{(0.998+0.941)/2}×100=5.88 (%). For the fa-cade side, which was intended to be seen, the yardstick tolerance was set as follows using the 'distance between the inward-facing side of anta and the outward-facing side of the column'; that is, (1.076-1.036)/ {(1.076+1.036)/2} = 3.79 (%).
Investigation into the planning method
In this study, each part of B6 is expressed simply with initials. For example, the width of anta is expressed as 'W・A' . The positions of each symbol are shown in Figure 4 and the (A) and (B) columns in Table 1.
Reconstruction of the designing method: As a result of examining proportional relations among dimensions of various parts, many parts were found which contained a ratio of dimensions that cannot be expressed simply by integers. This revealed that Tomb B6 cannot be designed only by a designing method using simple integer ratios. On the other hand, it was confirmed that the thickness of front and back walls, the width of the anta, and the width of side passage have relatively similar dimensions. Equi-interval grid lines were thus drawn on the plan of Tomb B6 using dimensions of those parts, and it was found that the planar shape of Tomb B6 is defined approximately by those grid lines. In light of the above facts, assuming that an ancient foot used for Tomb B6 was between 0.294 m and 0.330 m, a method such as this, which is further described below, can be proposed as the designing method of Tomb B6: per-form basic design by using the grid method with 1 grid ("grid" is hereinafter abbreviated as "G") stipulated to be equivalent to 3 feet and then adjust dimensions of each part in detailed design.
In the assumed basic design of Tomb B6, let the "width of the whole tomb (W・WT)" including passage around the tomb be 9G [21,22], the "depth of the whole tomb (D・WT)" likewise including passage around the tomb be "8G", and the "width of the side passage (W・ SP)" and the "width of the back passage (W・BP)" be respectively 1G. The width of the sepulcher (W・S) and the depth of the sepulcher (D・S) are determined to be 3G and 2G, respectively. The width of the anta (W・A) and the thickness of the back wall (T・BW) and the front wall (T・FW) are each determined to be 1 G. Let the width of the foundation platform covering upper and lower levels be 1 G, with the width at each level determined to be (1/2) G. As a result, the "width of the tomb (W・T), " the "depth of the tomb (D・T), " the "width of the pronaos (W・P), " and the "depth of the pronaos (D・P)" are 7G, 6G, 5G, and 2G, respectively (Column (D) of Table 1 and Figure 4). On the other hand, the "axial intercolumniation (I), " the "lower diameter of the column (LD・C), " the "width of the plinth (W・PI), " and the "width of the opening, " are determined on the basis of proportional relations. That is, the "axial intercolumniation (I)" is deter-mined as (6/16) C・DA by dividing the "center-to-center distance between antae (C・DA)" into the ratio of "5:6:5" (Figure 5). The "lower diameter of the column" is determined as (1/4) I based on "axial intercolumniation (I)" by using the ratio of "1:4. " The "width of the plinth (W・PI)" is determined as "LD・C + (1/3) LD・C" by adding (1/6) LD・C, derived from the "lower diameter of a column (LD・C)" by using the ratio of "1:6, " to both left and right sides of the lower diameter of the column. The "width of the opening (W・ O)" is considered to have been determined as (1/3) W・P on the ba-sis of the "width of a pronaos" by using the ratio of "1:3. " It is considered that dimensions of certain parts were adjusted in consideration of the functionality of the tomb and construction constraints after finishing the above basic design (Column (E) of Table 1). For instance, let us take a look at the tomb in the direction of its depth. The "depth of the sepulcher (D・S), " which was planned to be 2 G, has the value increased by 2 feet. As a reason why the depth of the sepulcher had to be increased, it is conceivable, for instance, that it was required to place a sarcophagus of an appropriate size in the sepulcher. For, without this expansion of the sepulcher, the length and width of a sarcophagus that could have been placed there would have been 60 cm smaller than those of the one actually placed, and such a sarcophagus would have caused problems in the burial of a corpse. The "thickness of the front wall (T・FW)" became "1 G + (1/16) T・FW**, " expanded by the value calculated from the "thickness of the front wall (T・FW)" in the basic design by using the ratio of "1:16. " This adjustment in dimension is considered to have been made in order to decorate the jambs of the door on the lateral surface of the front wall with reliefs without decreasing the thickness of the front wall. As the value of this amount of dimensional adjustment, (1/16) T・FW**, resultantly coincides with a mark on an ancient measure, namely 3/16 foot, it can be said to be possible that it was not calculated by using the said ratio as described above but determined as 3/16 foot directly in accordance with an ancient measure from the beginning. If, however, this amount of dimensional adjustment is assumed to have been derived by using ratios, the designing method proposed by the author is the one based on a consistent concept as a measure was used in designing those parts that were not conspicuous to those who would look at the tomb in consideration of construction constraints and ratios were used in designing those parts that were conspicuous to them. As noted above, this dimensional adjustment is considered to have been made in order to decorate the jambs of the door with reliefs. In this case, it seems natural to think that the amount of dimensional adjustment was derived from the thickness of the front wall using a ratio by considering how thick the decorating reliefs should be relative to the original thickness of the wall to strike a good balance. Therefore, in this paper, the amount of dimensional adjustment for the thickness of the front wall is considered to have been derived using a ratio. Returning to the subject of dimensional adjustment, the "depth of a pronaos (D・P)" has a value slightly reduced, by 1 foot. In light of this, 1 foot out of 2 feet added to the depth of the sepulcher mentioned above and (1/16) T・FW** added to the thickness of the front wall are considered to have been subtracted from the "depth of a pronaos (D・P). " In light of the fact that the "width of the back passage" has a value reduced by 1 foot, the remaining 1 foot out of 2 feet added to the depth of the sepulcher is considered to have been subtracted from the "the width of the back passage (W・BP). " It should be noted that length added by the "depth of the sepulcher (D・S)" and the "thickness of the front wall (T・ FW)" could have been handled by increasing the "depth of the whole tomb (D・WT)" instead of subtracting it from the "depth of a pronaos (D・P)" and the "width of the back wall. " That was not the case with Tomb B6, however. The following reason is conceivable for this. If the "depth of the whole tomb" was increased, the volume of rock that had to be excavated in order to construct Tomb B6 had to be increased for the amount obtained by "the increment of the depth of the whole tomb × the width of the whole tomb × the height of the whole tomb. " Therefore, only a slight increase in the "depth of the whole tomb" would have significantly increased the work for con-structing the tomb. On the other hand, since human eyes are not so sensitive to depth wise changes compared with lateral changes, the impact of some change in the "depth of a pronaos (D・P)" on the appearance of the tomb is considered to be small. In addition, since the back passage of the tomb is out of the sight of those who see the tomb, a change in the "width of the back passage" does not affect the appearance of the tomb at all. Therefore, the amount of dimensional adjustment generated in the "depth of the sepulcher" and the "thick-ness of the front wall" is considered to have been handled by subtracting it from such parts as the "depth of a pronaos" and the "width of the back passage. " Such an approach of adjusting dimensions in parts that are unlikely to affect the appearance of the tomb is one found in the designing method of the Nereid Monument at Xanthos.
With regard to the width direction, the "width of the sepulcher (W・S)" was expanded by 1 foot to 10 feet. While the actual opening serving as an entrance to the sepulcher was established in the lower right corner of the apparent door (Figure 3), the sepulcher was shifted to the right so that the central axis of the actual opening and that of the sepulcher coincide ( Figure 6). If the sepulcher had not been shifted right to align with the actual opening, the sarcophagus would have been pushed out to the position of the actual opening, which would have caused problems in carrying a corpse into the sepulcher. Therefore, it can be said that the sepulcher needed to be shifted to the right to align with the position of the actual opening. While the "width of the sepulcher (W・S)" was expanded by 1 foot, as noted above, even if the width of the sepulcher had remained 9 feet, the original size, it would still have been possible to place a sarcophagus having a sufficient size to put a corpse into. Therefore, it is difficult to think that the reason for expanding the "width of the sepulcher" was to accommodate a sarcophagus large enough. While several reasons for expanding the "width of the sepulcher" are conceivable, we can think of the following reason for instance. That is, the ratio of the length to the width of the sarcophagus was 1:3 except for the small one placed on the left of the entrance (Figure 7). If the "width of the sepulcher" is divided using this ratio, if the "width of the sepulcher" is 10 feet, the length of the sarcophagus is 7.5 feet and its width is 2.5 feet, resulting in proportional dimensions. Therefore, it can be in-ferred as a reason to have increased the "width of the sepulcher" by 1 feet to improve the workability of the sarcophagus. Alternatively, the interval between sarcophagi placed on the right and left of the sepulcher is 5 feet in the clear, which is a wellrounded number (Figure 7). This part of the sepulcher is considered to require a certain amount of space for conducting the burial of a corpse and rituals such as flower offering. Thus, assuming that the designer of Tomb B6 gave priority to keeping the width of this part 5 feet, if the width of the sepulcher had remained 9 feet, the width of each sarcophagus would have been 2 feet. If the width of each sarcophagus had been 2 feet, the width of space in which a corpse is actually placed would have been approximately 40 cm, subtracting the width of the two sidewalls, 12 cm wide each, constituting a sarcophagus. As this approximate value of 40 cm is smaller than the shoulder width of a typical adult, it would have been impossible to place a corpse in the sarcophagus appropriately.
The width of the sepulcher may have been increased in order to avoid this. Returning to the subject of dimensional adjustment, the "width of an anta (W・A)" was also somewhat increased to decorate the side of an anta with reliefs. This increment is (1/8) W・A*, which is derived from the "width of the anta (W ・ A*)" in the basic design by using the ratio of 1:8. In accordance with this increase in the width of an anta, the "width of the tomb (W・T)" is considered to have been changed to "8 G + (1/4) W・A*" by Adding the increment in the width of antae on the right and left. The "width of the whole tomb (W・WT)" is also considered to have been changed to "9 G + (1/4) W・A*" by adding the increment in the width of antae on the right and left. As the value of this increment in the width of antae resultantly coincides with a mark on an ancient measure, namely 3/8 foot, it can be said to be possible that it was not calculated by using a ratio but determined as 3/8 foot directly in accordance with an ancient measure from the beginning. If, however, this increment in the width of antae is assumed to have been determined by using ratios, the designing method proposed by the author is the one based on a consistent concept as a measure was used in adjusting dimensions of those parts that were not conspicuous to those who would look at the tomb in consideration of construction constraints and ratios were used in designing those parts that were conspicuous to them. In addition, antae are components of a façade, and it needs to be considered how much adjustment to the original anta width re-sults in the balanced anta width relative to a façade. Therefore, it seems natural to conceive that this increment in anta width was derived by using a ratio for the purpose of proportional adjustment rather than adjusted by using a measure in one way or the other. Therefore, in this paper, the increment in the width of antae is considered to have been derived from the width of antae itself by using a ratio. This kind of planning process in which a scheme design is adjusted at the execution planning stage was observed in other temples and stoas as well as the Nereid Monument in Xanthos.
In light of the above design process, for parts where dimensional adjustments using ratios were made, numbers coinciding with marks on an ancient measure were used in ratios applied, for instance, 1:2 (=1/2), 1:3 (=1/3), 1:8 (=1/8), 1:16 (=1/16) ( Table 1, Column E). On the other hand, the dimension of each part of Tomb B6 in the basic design was determined on a 3-foot grid. Therefore, the value of a dimensional adjustment calculated by using the ratio of numbers mentioned above is necessarily a well-rounded number coinciding with a mark on an ancient measure (Table 1, Column (F)). For in-stance, the width of an anta was considered to have been planned on a 3-foot grid in the basic design, as noted above, and a value derived from itself using the ratio of 1:8 was considered to have been added in the detailed design. In this case, the increment in anta width is 3/8 foot, which coincides with a mark on an ancient measure. It is difficult to be considered as accidental that ratios used in dimensional adjustment all coincide with numbers found on marks on an ancient measure as found above. Therefore, this fact can be conceived as one piece of evidence that supports that the designing method proposed in this paper is appropriate. In addition, under the designing method proposed in this study ratios are used in executing design and dimensional adjustment for the façade, which is conspicuous, namely with regard to such lengths as axial intercolumniation, the diameter of a lower column, the width of a plinth, the width of an anta, and the width of an opening. On the other hand, inconspicuous parts are de-signed by the grid method, and dimensional adjustment is made in multiples of an ancient foot. This can be interpreted as follows. Ratios were used in designing the façade, which is conspicuous, in giving priority to design, and the designing method using the grid method was basically used, with adjustment made in dimensions easily measured by a measuring stick, in designing other parts, giving priority to construction constraints. Namely, the designing method proposed in this paper is based on a consistent concept, and a possibility is considered small that a designing method with such a consistent concept was accidentally devised.
Some may wonder that, in light of the above result, there is a possibility that dimensional adjustments were made after executing basic design by using ratios instead of using the grid method. Design using proportional relations, however, is considered to have been used to create harmonious architectural appearance, to begin with, as Vitruvius said. Therefore, it is difficult to think that, if design was executed using proportional relations, a dimensional adjustment was made that broke those proportional relations. As a matter of fact, as indicated in preceding studies on ancient Greek designing methods, even if design was executed using proportional relations, when it came to overall dimensions such as the width and depth of a building, for instance, some revisions were made by later detailed adjustment even if proportional relations were used in the early stage of design. Except for such situations, however, it was rarely found that an adjustment is easily made to dimensions derived by using proportional relations. On the other hand, as shown by the preceding study by Horiuchi, under the grid method, instances are found where the overall shape is determined by reference to grid lines with dimensions of each part being adjusted later. Therefore, in light of the results of preceding studies, it seems appropriate to think that basic design by the grid method was executed in the case of Tomb B6.
Examination of design dimensions:
In light of the fact that Tomb B6 was constructed in the ancient time, design dimensions of each part derived by using the above de-signing method must, of course, be represented in "ancient measures, " yardsticks used at the time. Here, the validity of the designing method proposed in this paper is verified by calculating design dimensions using the designing method introduced in the preceding section and ancient measures and examining errors against measured dimensions.
For this purpose, the length of an ancient foot used for Tomb B6 needs to be estimated first. As noted in the preceding section, the "width of the tomb (W・T)" is considered to have been planned to be 21 3/4 feet. Therefore, by dividing the measured dimension of the "width of the tomb" at 7.088 m by 21 3/4 feet, 1 foot is obtained as 0.326 m. The reason why the ancient foot is calculated on the basis of the "width of the tomb" is as follows. The dimension such as the "width of a tomb" that defines the overall shape of a tomb has a large impact on its visual impression. If this dimension deviates from the design dimension, the visual impression of the tomb will be different from what has been assumed at the time of design. If the construction of a tomb starts from detailed parts, errors resulting from work on each part can accumulate, making it more possible that the width of the tomb differs from the design dimension. Therefore, it seems natural to think that, in constructing a tomb, the width of the tomb was precisely established first and then work on detailed parts started. Therefore, in this paper, assuming that the builder of Tomb B6 also paid a minute attention to ensuring the width of the tomb, an ancient foot is calculated on the basis of the width of the tomb. Let me add, however, that taking an average of lengths of an ancient foot calculated on the basis of parts on the façade yields the same value of one ancient foot at 0.326 m. Table 1 list values of the design dimension of each part calculated on the basis of the designing method introduced in the preceding section using this ancient measure with one foot = 0.326 m as well as errors between the design and measured dimensions. As shown in Table 1, except for the depth of the right side of the tomb, which was obviously deformed in work, an error between the design and measured dimensions is small with every part. As indicated above, the designing method proposed in this paper is based on a consistent concept. In addition, under the designing method proposed in this paper, every ratio used in dimensional adjustment is expressible as a fraction found in marks on the ancient measure. As a result, the amounts of dimensional adjustment are necessarily calculated as values coinciding with marks on an ancient measure (Columns (E) and (F) of Table 1). Of course, reasonable grounds leading to a dimensional adjustment can be inferred for each part where the adjustment was made. Since a probability is considered low that these state of affairs occur simultaneously by accident, there can be said to be a sufficient possibility that the designing method presented in this paper was used for Tomb B6.
Columns (F), (G), and (H) in
It should be noted that, while, under the designing method proposed in this paper, dimensional adjustments are made in detailed design after basic design, as indicated by Column (F) of Table 1, a wall thickness was never decreased by dimensional adjustment even if it was increased. A wall thickness of at least 1G or 3 feet was secured. Perhaps, the designer of Tomb B6 may have thought, in executing design, that it was necessary to ensure a wall thickness of at least 1G or 3 feet in constructing a rock-cut tomb of the scale of Tomb B6. An attention is intended to be paid to the relationship between wall thickness and the scale of a tomb in analyzing the design-ing methods of other rock-cut tombs in Kaunos.
Conclusions
This paper thus analyzed the plan designing method of Tomb B6. The designing method proposed in this paper are characterized as follows.
1) It is possible that, in designing the plan of Tomb B6, after the approximate scale of the tomb and the layout of chambers were determined in the basic design, dimensions of parts were adjusted in the detailed design.
2) While the basic design basically used the grid method with 1 grid equivalent to 3 feet, ratios were considered to have been used in designing the façade. In the detailed design, dimensional adjustments based on ratios were made in the façade, and dimensional adjustment based on a measure were made elsewhere. This can be interpreted as follows. The designer of Tomb B6 gave priority on design in designing the façade, which was conspicuous, and on construction constraints in designing other parts.
3) In dimensional adjustments regarding wall thickness, no adjustment was made that resulted in a wall thickness below 3 feet or 1 G while some adjustments resulted in a wall thickness above 3 feet. This suggests a possibility that the designer of Tomb B6 thought, in executing design, that it was necessary to ensure a wall thick-ness of at least 1 G or 3 feet in constructing a rock-cut tomb of the scale of Tomb B6.
As indicated above, this study showed a series of designing methods that have a possibility of having been used for Tomb B6 and clarified several characteristics of those designing methods. In the future, it will be necessary to analyze the designing methods of other rock-cut tombs adjacent to Tomb B6 and further examine whether the designing methods proposed in this paper and their characteristics are products of chance or not. The next paper, however, will continue to focus on Tomb B6 to clarify the designing method for its elevation surface. It is intended to examine the presence or absence and the methodology of visual compensation reported by the author in the preceding study from the perspective of a designing method. | 2019-04-16T13:22:09.009Z | 2017-05-22T00:00:00.000 | {
"year": 2017,
"sha1": "102a7ee4a760a78c3349f1e390642d5c0c39af30",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.4172/2165-784x.1000276",
"oa_status": "HYBRID",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "8d0f8087a07fa2a5d7022134457da3ca75813c06",
"s2fieldsofstudy": [
"Engineering",
"History"
],
"extfieldsofstudy": [
"Engineering"
]
} |
33540251 | pes2o/s2orc | v3-fos-license | A general analysis with trilinear and bilinear R-parity violating couplings in the light of recent SNO data
We analyse an extension of the minimal supersymmetric standard model including the dominant trilinear and bilinear R-parity violating contributions. We take the trilinear terms from the superpotential and the bilinear terms from the superpotential as well as the scalar potential. We compute the neutrino masses induced by those couplings and determine the allowed ranges of the R-parity violating parameters that are consistent with the latest SNO results, atmospheric data and the Chooz constraint. We also estimate the effective mass for neutrinoless double beta decay in such scenarios.
I. INTRODUCTION
It is firmly established that non-zero neutrino masses can provide a solution to the observed solar [1,2] and atmospheric neutrino deficits [3]. This requires two large mixing angles (very recent SNO data [1] favour the large mixing angle solution, LMA, for the solar anomaly) and two hierarchical neutrino mass squared differences (∆m 2 solar ≪ ∆m 2 atm ). We assume that the three mass eigenstates are given by m i (i = 1, 2, 3) and we parametrise the rotation matrix from neutrino flavour (f ) to mass (i) eigenstates as: V f i = R 23 (θ 23 )R 13 (θ 13 )R 12 (θ 12 ). The MNS matrix (analogous to the CKM mixing matrix for the quark sector) neglecting CP phases is given by [4], where c ij ≡ cos θ ij and s ij ≡ sin θ ij . For the solar anomaly, which we take to be a consequence of ν e -ν µ oscillation, the relevant mass squared difference is ∆m 2 12 = ∆m 2 solar , while for the atmospheric case, the oscillation being between ν µ and ν τ , the relevant mass squared difference is ∆m 2 13 ≈ ∆m 2 23 = ∆m 2 atm . After the inclusion of the SNO data, the MSW-LMA oscillation is the most favoured solution with ∆m 2 solar = (2.5 − 19.0) × 10 −5 eV 2 and sin 2 2θ 12 = 0.61 − 0.95 [5]. The SuperK atmospheric neutrino data suggest ∆m 2 atm = (2 − 5) × 10 −3 eV 2 with sin 2 2θ 23 = 0.88 − 1.0 [3]. The Chooz [6] and Palo Verde [7] long baseline reactor experiments bound sin 2 θ 13 ∼ < 0.04. Concerning the absolute masses, the recent claim for the evidence of neutrinoless double beta decay (0νββ) by the Moscow-Heidelberg Collaboration [8], although not yet firmly established, constrains the (ee)-element of the neutrino mass matrix in the flavour basis to lie between 0.1 and 0.5 eV. Tritium β-decay requires m νe ∼ < 2.2 eV [9]. Cosmological analysis from the recent 2dF Galaxy Redshift Survey constrains i m i ∼ < 2.2 eV [10].
Neutrino masses can be generated in the R-parity violating (R p / ) Supersymmetric Standard Model, where R p is defined as (−1) 3B+L+2S [11,12]. Here B, L and S are the baryon number, lepton number and spin of a particle, respectively. Strict phenomenological bounds on B and/or L violation exist in the literature [13]. In this paper we assume that B violating couplings are absent and generate neutrino Majorana masses via two units of L violation. For this purpose, we allow both bilinear (µ i ) and trilinear (λ, λ ′ ) interactions in the superpotential as well as the bilinear soft terms (B i ) 1 , where the vector L J = (H d , L i ) with J : 4..1, and the soft supersymmetry-breaking potential is As has been extensively discussed in refs [14][15][16][17] (and references therein), field redefinitions of the H d , L i fields correspond to basis changes in L J space and consequently the Lagrangian parameters will be altered. We use the basis-independent parameters constructed in [15,16] and write the neutrino mass matrix in terms of δ i µ , δ i B , δ ijk λ , δ ipq λ ′ , which in the basis in which the sneutrino vevs are zero correspond to µ i /|µ|, B i /|B|, λ ijk , λ ′ ipq , respectively. Calculations of neutrino masses in the context of R p / theories have focused on tree-level contributions from the bilinear µ i -parameter or loop contributions from the trilinear λ and λ ′ couplings [18]. Recently a detailed phenomenological analysis has been done for a model including both λ, λ ′ and the superpotential bilinear parameter µ i [19,20]. In the basis-independent approach, a phenomenological analysis has been done including only the purely bilinear contributions, δ i µ from the superpotential and δ i B from the soft Lagrangian [17]. The possibility that loops involving δ µ were important was also discussed in [17].
Our purpose in this paper is to perform for the first time an analysis in a model which includes both the superpotential and soft bilinear parameters δ i µ and δ i B along with the superpotential trilinear couplings δ ijk λ , in the mass insertion approximation. A similar analysis can be performed including the trilinears δ ipq λ ′ . In our analysis we take all R p / parameters to be real. Thus we update the analysis in [17] by including the trilinear loop contributions alongside the bilinear tree and loop terms. We interface the neutrino mass matrix constructed out of the R p / parameters with the constraints on the solar and atmospheric mass squared splittings, the corresponding mixing angles, and the Chooz constraint on θ 13 . We then examine the nature of the mass spectrum and we check the consistency with cosmological data. We also obtain the effective mass for neutrinoless double beta decay in our scenario. Furthermore, we observe the impact of adding the trilinear contributions to the numerical results obtained in [17].
II. PARAMETRIZATION OF THE MASS MATRIX
The R p / model we are considering generates a single neutrino mass at tree level proportional to δ i µ δ j µ . There are additional loop corrections to the mass matrix when non-vanishing δ i B , δ ijk λ (and δ ipq λ ′ ) are included, leading to more than one non-zero mass eigenvalue. This enables us to fit the data on mass squared splittings and mixing angles. The relevant types of loops we consider in the mass insertion approximation are: • the well-known loops involving the trilinear R p / couplings λ or λ ′ at the neutrino vertices I and II in Fig. 1 (with lepton/slepton or quark/squark as propagators), which give contributions proportional to δ λ δ λ (or δ λ ′ δ λ ′ ); • the Grossman-Haber diagrams [14], in which there are gauge couplings at the neutrino vertices while there are two types of R p / interactions contributing to the ∆L = 2 Majorana mass in the diagram of Fig. 1. The first kind have R p / couplings located at positions III + IV (slepton-Higgs mixing) with contributions proportional to δ B δ B . In the second type, the R p / interactions are located at positions V + IV (neutrino-neutralino and slepton-Higgs mixing) with contributions proportional to δ µ δ B ; • the diagram of Fig. 2, where two units of L violation come from positions V (neutrino-neutralino mixing) and II (λ or λ ′ vertex). The contribution to the neutrino mass is proportional to δ µ δ λ (or δ µ δ λ ′ ).
The last two types of loops have frequently been overlooked in analytic estimates of neutrino masses. Exact formulae for these diagrams can be found in [16]. Our formulae are robust (see later) and good enough for an order-of-magnitude estimate.
The conventional loops (Rp / at I + II) and the Grossman-Haber loops (Rp / at III + IV or V + IV) contributing to the neutrino mass.
Setting all unknown sparticle masses equal to M S = 100 GeV, and neglecting the mixing angles among the sparticles, we have a neutrino mass matrix of the form [ We have included for completeness the contributions arising from the δ λ ′ terms which we put to zero in our numerical analysis. The simplest case to consider with a common δ i µ ≡ δ µ , δ i B ≡ δ B and δ ink λ ≡ δ λ does not work as it gives only two non-zero masses and an eigenvector of the form (1/ √ 2, −1/ √ 2, 0) which cannot accomodate two large mixing angles for θ 12 and θ 23 .
In our numerical analysis we take : δ i µ , δ i B , δ ink λ ≡ δ λ , δ ink λ ′ = 0, for i = 1, 2, 3, i.e., seven independent parameters. Thus, the neutrino mass matrix elements will be given by, where we have employed the hierarchy of the charged fermion masses to keep only the dominant terms.
III. RESULTS AND CONCLUSIONS
We have performed a general scan of parameter space made up by the seven parameters that appear in the mass matrix allowing tree-level contributions to either dominate over the loop corrections, to be on the same order as these, or to be much smaller than the loop terms. The fitted values of the couplings, obtained by using the atmospheric and Chooz data together with the preferred solar MSW-LMA solution from SNO, are presented in table I. Here we have taken cos β = 1 primarily to ensure a comparison with results in [17], which too employed the same value of cos β, on the same footing. We stress that although this choice of cos β leads to an unacceptably low tan β, we still use this value for illustration and effective comparison with previous results, noting at the same time that an order-of-magnitude estimate of the couplings is not sensitive to this choice. We also show in table I the allowed ranges of i m i , and m eff ≡ i V 2 ei m i = i |V ei | 2 m i (since we have assumed the V matrix to be real). This last quantity is the effective mass relevant for neutrinoless double beta decay (the neutrino masses induced by R p / interactions are Majorana type). In Fig. 3 we present the allowed region in the |δ B | = i (δ i B ) 2 versus |δ µ | = i (δ i µ ) 2 plane for the combined fit taking cos β = 1. We present our results for both δ λ = 0 and δ λ = 0. It can be clearly seen that the allowed region increases when we admit non-zero values of δ λ . This is mainly due to the presence of the δ µ δ λ terms in the mass matrix (originating from Fig. 2) which can take either sign and thus can accomodate a larger region of parameter space.
The resulting fit strongly prefers a hierarchical mass pattern in our scenario, although our analysis cannot make a distinction between the inverted and the normal hierarchy. The inclusion of a non-zero δ λ to the set of non-vanishing δ i µ and δ i B allows us to have all three non-vanishing neutrino masses while if only the latter two vectors are non-zero there are only two non-zero eigenvalues [17]. The maximum value of m eff we predict can be tested in the next generation of neutrinoless double beta decay experiments. We have also checked that the cosmological bound restricting the sum of eigenvalues to be at most ∼ 2 eV is fulfilled.
To conclude, we outline the novel features of our analysis: (i) the trilinear R p / interactions along with both types (originating from superpotential and soft terms) of bilinear R p / terms have been employed for the first time to analyse the neutrino mass matrix; (ii) the seven parameter neutrino mass matrix has been interfaced with the latest SNO results favouring MSW-LMA solution along with the SuperK atmospheric and Chooz data; (iii) by including a non-vanishing δ λ we have updated the analysis in [17] and the numerical impact of adding this term has been highlighted in Fig. 3; (iv) the mass spectrum in this kind of scenario implies a hierarchical pattern, although a distinction between the normal and inverted pictures cannot be made; (v) our prediction of m eff can be tested in the next generation of neutrinoless double beta decay experiments (GENIUS, Majorana, MOON, EXO) [21]. We finally make a remark that allowing a non-vanishing δ λ ′ will not qualitatively change the pattern of our fit.
FIG. 3. We present the available region in the δB ≡ |δB | versus δµ ≡ |δµ| plane for a combined fit using Chooz, SuperK and the SNO results for the MSW-LMA solar oscillation solution, with cos β = 1. The circles are solutions for δ λ = 0, the crosses are for δ λ = 0. This figure should be compared with Fig. 5 of ref. [17]. | 2017-09-17T18:32:01.102Z | 2002-08-01T00:00:00.000 | {
"year": 2002,
"sha1": "906c07ec4ffa9a4d1804627f05994ab35b3e5aa6",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/hep-ph/0208009",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "906c07ec4ffa9a4d1804627f05994ab35b3e5aa6",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.