Add files using upload-large-folder tool

preprint/preprint__0ea74b5c8de9b0e46fcdecaea8a66b6a4b59669369e79a3310d00aabe8f4f20a/preprint__0ea74b5c8de9b0e46fcdecaea8a66b6a4b59669369e79a3310d00aabe8f4f20a_det.mmd

@@ -0,0 +1,573 @@
+<|ref|>title<|/ref|><|det|>[[44, 106, 890, 210]]<|/det|>
+# Maternal Plasma Cell-Free RNA as a Predictor of Early and Late-Onset Preeclampsia Throughout Pregnancy  
+
+<|ref|>text<|/ref|><|det|>[[44, 229, 422, 277]]<|/det|>
+Tamara Garrido- Gómez tgarrido@fundacioncarlossimon.com  
+
+<|ref|>text<|/ref|><|det|>[[42, 303, 945, 370]]<|/det|>
+Carlos Simon Foundation - INCLIVA Health Research Institute https://orcid.org/0000- 0002- 6584- 4832  Nerea Castillo- Marco  
+
+<|ref|>text<|/ref|><|det|>[[44, 350, 936, 391]]<|/det|>
+Carlos Simon Foundation, iPremom Pregnancy healthcare Diagnostics https://orcid.org/0000- 0002- 4817- 4777  
+
+<|ref|>text<|/ref|><|det|>[[44, 397, 668, 440]]<|/det|>
+Teresa Cordero  Carlos Simon Foundation, iPremom Pregnancy healthcare Diagnostics  
+
+<|ref|>text<|/ref|><|det|>[[44, 444, 441, 485]]<|/det|>
+Marina Igual  iPremom Pregnancy healthcare Diagnostics  
+
+<|ref|>text<|/ref|><|det|>[[44, 490, 950, 531]]<|/det|>
+Irene Muñoz- Blat  Carlos Simon Foundation - INCLIVA Health Research Institute https://orcid.org/0000- 0003- 0543- 1064  
+
+<|ref|>text<|/ref|><|det|>[[44, 536, 440, 577]]<|/det|>
+Carla Gómez- Álvarez  iPremom Pregnancy healthcare Diagnostics  
+
+<|ref|>text<|/ref|><|det|>[[44, 582, 440, 624]]<|/det|>
+Neus Bernat- González  iPremom Pregnancy healthcare Diagnostics  
+
+<|ref|>text<|/ref|><|det|>[[44, 628, 440, 670]]<|/det|>
+Ángela Gaspar- Doménech  iPremom Pregnancy healthcare Diagnostics  
+
+<|ref|>text<|/ref|><|det|>[[44, 675, 440, 716]]<|/det|>
+Érika Ortiz- Domingo  iPremom Pregnancy healthcare Diagnostics  
+
+<|ref|>text<|/ref|><|det|>[[44, 721, 440, 762]]<|/det|>
+Alba Vives  iPremom Pregnancy healthcare Diagnostics  
+
+<|ref|>text<|/ref|><|det|>[[44, 767, 950, 809]]<|/det|>
+Sheila Ortega- Sanchís  Carlos Simon Foundation - INCLIVA Health Research Institute https://orcid.org/0000- 0001- 6956- 4910  
+
+<|ref|>text<|/ref|><|det|>[[44, 814, 275, 855]]<|/det|>
+Rogelio Monfort- Ortiz  University Hospital La Fe  
+
+<|ref|>text<|/ref|><|det|>[[44, 860, 280, 901]]<|/det|>
+Petr Volkov  Carlos Simon Foundation  
+
+<|ref|>text<|/ref|><|det|>[[44, 906, 774, 948]]<|/det|>
+Juan Luis Delgado  Fetal Medicine Unit Murcia- IMIB Arrixaca https://orcid.org/0000- 0003- 1687- 0222
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[42, 42, 773, 103]]<|/det|>
+Laura Hernandez-Hernandez Fetal Medicine Unit Murcia-IMIB Arrixaca Esther Canovas Fetal Medicine Unit Murcia-IMIB Arrixaca https://orcid.org/0000-0003-4595-2219  
+
+<|ref|>text<|/ref|><|det|>[[42, 108, 772, 150]]<|/det|>
+Maria M Gil Hospital Torrejon https://orcid.org/0000- 0002- 9993- 5249  
+
+<|ref|>text<|/ref|><|det|>[[42, 156, 636, 199]]<|/det|>
+Belén Santacruz Hospital Universitario de Torrejon, Universidad Francisco de Vitoria  
+
+<|ref|>text<|/ref|><|det|>[[42, 203, 636, 270]]<|/det|>
+Nieves Luisa Gonzalez- Gonzalez University of La Laguna  
+
+<|ref|>text<|/ref|><|det|>[[42, 275, 353, 317]]<|/det|>
+Walter Plasencia Hospital Universitario de Canarias  
+
+<|ref|>text<|/ref|><|det|>[[42, 322, 470, 365]]<|/det|>
+Alfredo Perales- Marín University Hospital La Fe, University of Valencia  
+
+<|ref|>text<|/ref|><|det|>[[42, 369, 275, 411]]<|/det|>
+Beatriz Marcos- Puig University Hospital La Fe  
+
+<|ref|>text<|/ref|><|det|>[[42, 416, 729, 457]]<|/det|>
+Ana Palacios Hospital Alicante  
+
+<|ref|>text<|/ref|><|det|>[[42, 462, 731, 504]]<|/det|>
+Iñigo Melchor Corcóstegui BioCruces Health Research Institute https://orcid.org/0000- 0002- 4190- 3180  
+
+<|ref|>text<|/ref|><|det|>[[42, 509, 430, 551]]<|/det|>
+Alicia Martin- Martinez Complejo Hospitalario Universitario Insular  
+
+<|ref|>text<|/ref|><|det|>[[42, 555, 430, 597]]<|/det|>
+Taysa Benitez- Delgado Complejo Hospitalario Universitario Insular  
+
+<|ref|>text<|/ref|><|det|>[[42, 602, 944, 644]]<|/det|>
+Carlos Simon Carlos Simon Foundation- INCLIVA Health Research Institute https://orcid.org/0000- 0003- 0902- 9531  
+
+<|ref|>text<|/ref|><|det|>[[42, 682, 102, 700]]<|/det|>
+Article  
+
+<|ref|>text<|/ref|><|det|>[[42, 720, 135, 738]]<|/det|>
+Keywords:  
+
+<|ref|>text<|/ref|><|det|>[[42, 758, 330, 777]]<|/det|>
+Posted Date: January 17th, 2025  
+
+<|ref|>text<|/ref|><|det|>[[42, 796, 475, 815]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 5684050/v1  
+
+<|ref|>text<|/ref|><|det|>[[42, 834, 914, 876]]<|/det|>
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+<|ref|>sub_title<|/ref|><|det|>[[42, 894, 253, 912]]<|/det|>
+## Additional Declarations:  
+
+<|ref|>text<|/ref|><|det|>[[42, 917, 519, 936]]<|/det|>
+Table 1 is available in the Supplementary Files section.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[41, 44, 951, 134]]<|/det|>
+Yes there is potential Competing Interest. N.C- M., M.I., T.G- G., C.S. are inventors on a patent application (EP24383276.3) covering methods for determining the risk of preeclampsia. N.C- M., T.C., M.I., C.G- A., N.B- G., A.G- D., E.O- D., A.V., T.G- G. are employees of iPremom Pregnancy Healthcare Diagnostics. C.S. is a founder of iPremom Pregnancy Healthcare Diagnostics  
+
+<|ref|>text<|/ref|><|det|>[[41, 183, 944, 227]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on October 20th, 2025. See the published version at https://doi.org/10.1038/s41467-025-64215-2.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[78, 84, 797, 105]]<|/det|>
+# 1 Maternal Plasma Cell-Free RNA as a Predictor of Early and Late-Onset 
+
+<|ref|>sub_title<|/ref|><|det|>[[78, 115, 483, 136]]<|/det|>
+## 2 Preeclampsia Throughout Pregnancy 
+
+<|ref|>text<|/ref|><|det|>[[78, 153, 880, 316]]<|/det|>
+3 Nerea Castillo-Marco¹,², Teresa Cordero¹,², Marina Igual², Irene Muñoz-Blat¹, Carla Gómez-Álvarez², 4 Neus Bernat-González², Ángela Gaspar-Doménech², Érika Ortíz-Domingo², Alba Vives², Sheila Ortega- 5 Sanchís¹, Rogelio Monfort-Ortiz³, Petr Volkov¹, Juan Luis Delgado⁴, Laura Hernandez-Hernandez⁴, 6 Esther Canovas⁴, Maria del Mar Gil⁵,⁶,⁷, Belén Santacruz⁵,⁷, Nieves Luisa Gonzalez-Gonzalez⁸, Walter 7 Plasencia⁹, Alfredo Perales-Marín³,¹⁰, Beatriz Marcos-Puig³, Ana María Palacios-Marqués¹¹,¹²,¹³, Íñigo 8 Melchor¹⁴, Alicia Martín-Martínez¹⁵, Taysa Benitez-Delgado¹⁵, Carlos Simón¹,²,¹⁰,¹⁶, Tamara Garrido- 9 Gómez¹,², on behalf of the PREMOM Consortium¹⁷. 
+
+<|ref|>sub_title<|/ref|><|det|>[[78, 333, 300, 349]]<|/det|>
+## 10 PREMOM Consortium: 
+
+<|ref|>text<|/ref|><|det|>[[78, 366, 880, 644]]<|/det|>
+11 Aiartzaguena A, Alonso-Menéndez S, Amezcua A, Andrada-Ripolles C, Arias-Valdés EM, Arnal-Burró
+12 AM, Barbazán MJ, Batalla-Urrea R, Baviera-Royo P, Bondía M, Burgos J, Cabrera-Leon CR, Campillos-
+13 Maza JM, Casanova MC, Cuenca-Gómez D, De Bonrostro-Torralba C, De Leon-Socorro S, Del Campo A,
+14 Delgado-Gonzalez JL, Diaz-Lozano P, Fabre M, Ferrer E, Florez-Herrero S, Francés-Ferré J, García-
+15 Izquierdo O, García-Sousa V, Gibbone E, Goiti H, Gomez A, Gonzalez E, Gregorio-González SE,
+16 Hernández-Suárez M, Herrero-Serrano R, Jimenez-Mendez A, Jodar-Perez MA, Larrea-Ortíz Quintana
+17 MM, Lasierra-Beamonte A, Lobo-Valentín RM, López-Soto A, Lozano-Moreno A, Macías-Alonso MJ,
+18 Martín-Arias A, Martín-Medrano EM, Martínez-Cendán JP, Martínez-Rivero I, Meabe A, Melchor JC,
+19 Mico Y, Montesinos-Albert M, Moreno-Reviriego A, Nieto-Tous M, Orive-Boluda A, Oros D, Padrón-
+20 Pérez E, Paules C, Peña-Lobo SM, Pérez-Pascual E, Recio V, Reula-Blasco C, Rodríguez L, Romero MI,
+21 Rubert L, Ruiz-Martínez S, Ruiz-Peña AC, Salinas A, Sanchez-Martínez E, Satorres-Pérez E, Villar-
+22 Graullera E. 
+
+<|ref|>sub_title<|/ref|><|det|>[[78, 661, 195, 676]]<|/det|>
+## 23 Affiliations: 
+
+<|ref|>text<|/ref|><|det|>[[78, 688, 864, 915]]<|/det|>
+24 ¹ Carlos Simon Foundation, Valencia, Spain
+25 ² R&D Department, iPremom Pregnancy healthcare Diagnostics, Valencia, Spain
+26 ³ Department of Obstetrics, University Hospital La Fe, Valencia, Spain.
+27 ⁴ Fetal Medicine Unit Murcia-IMIB Arrixaca
+28 ⁵ Obstetrics and Gynecology Department, Hospital Universitario de Torrejón, Madrid, Spain.
+29 ⁶ Obstetrics and Gynecology Department, Hospital Universitario La Paz, Madrid, Spain.
+30 ⁷ School of Medicine, Universidad Francisco de Vitoria, Madrid, Spain.
+31 ⁸ Department of Obstetrics and Gynecology, University of La Laguna, Tenerife. Spain.
+32 ⁹ Department of Obstetrics and Gynecology, Hospital Universitario de Canarias, Tenerife. Spain.
+33 ¹⁰ Department of Pediatrics, Obstetrics, and Gynecology, University of Valencia, Valencia, Spain.
+34 ¹¹ Obstetrics and Gynecology Department, Dr. Balmis General University Hospital, Alicante, Spain.
+35 ¹² Alicante Institute for Health and Biomedical Research (ISABIAL), Alicante, Spain.
+36 ¹³ Department of Gynecology, Miguel Hernández University, Alicante, Spain.
+37 ¹⁴ Obstetrics and Gynecology Service, BioCruces Health Research Institute, Hospital Universitario Cruces (Basque Country University), Biscay, Spain.
+38 ¹⁵ Department of Obstetrics and Gynecology, Complejo Hospitalario Universitario Insular, Materno Infantil, Las Palmas, Spain.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 97, 863, 140]]<|/det|>
+16 Department of Obstetrics and Gynecology, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA, USA 17 A complete list of contributors of the PREMOM Consortium is provided in the Supplementary Information.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 200, 186, 215]]<|/det|>
+## Abstract  
+
+<|ref|>text<|/ref|><|det|>[[115, 230, 883, 584]]<|/det|>
+Early- onset (EOPE) and late- onset preeclampsia (LOPE) pose significant challenges to maternal and child health, highlighting the need for early, non- invasive risk identification. In this prospective longitudinal study, we followed 9,586 pregnant women, collecting blood samples each trimester: 9- 14 weeks (T1), 18- 28 weeks (T2), and after 28 weeks or at preeclampsia diagnosis (T3). Plasma cell- free RNA (cfRNA) signatures were analyzed in women who developed EOPE (n=42) or LOPE (n=43) and compared to matched normotensive controls (n=75). Mapping cfRNA origins and performing differential abundance analysis provided insights into multi- organ impacts, revealing distinct transcriptional features of EOPE and LOPE. We developed a first- trimester EOPE predictive model using 36 transcripts, achieving 83% sensitivity, 88% specificity, and an AUC of 0.85, detecting risk 18.0 weeks before onset. A second- trimester model based on 87 cfRNA transcripts, predicted EOPE 8.5 weeks prior to onset with 87% sensitivity, 84% specificity, and an AUC of 0.85. For LOPE model, detecting risk 14.9 weeks before onset, used 92 cfRNAs, with 86% sensitivity, 89% specificity, and an AUC of 0.88. EOPE models were enriched for decidua- associated transcripts, highlighting the maternal involvement in this subtype, while LOPE models showed diverse tissue responses, paving the way for improved subtype differentiation and tailored interventions to mitigate preeclampsia risks.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[68, 85, 216, 100]]<|/det|>
+## 73 Introduction  
+
+<|ref|>text<|/ref|><|det|>[[67, 116, 886, 184]]<|/det|>
+Maternal and infant mortality during pregnancy and labor are critical indicators of community and national health<sup>1,2</sup>. Most pregnancy complications arise from disorders that develop during the periconeptional phase, particularly during embryonic implantation and early placentation<sup>3</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[67, 198, 886, 336]]<|/det|>
+Preeclampsia — a life- threatening obstetric syndrome— is characterized by new- onset of hypertension after 20 weeks of gestation, accompanied by signs of kidney, liver, or brain damage<sup>4</sup>. Each year, preeclampsia contributes to 14% of maternal deaths worldwide, leaving a lasting impact on survivors' health<sup>5</sup>. It also constitutes a significant public health burden, incurring \$1.03 billion in maternal healthcare costs and an additional \$1.15 billion for neonatal care in infants born to mothers affected by preeclampsia within the first year after birth in the United States<sup>6</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[67, 350, 884, 514]]<|/det|>
+The heterogeneity of preeclampsia is notable, differentiated by the timing of onset and severity of symptoms. Early- onset preeclampsia (EOPE) arises before 34 weeks of gestation, necessitating emergency delivery to mitigate risks to maternal and fetal health<sup>7,8</sup>. In contrast, late- onset preeclampsia (LOPE) manifests after 34 weeks and can lead to severe maternal organ damage such as kidney, liver, or brain damage<sup>8- 11</sup>. Therefore, there is an urgent need for straightforward, non- invasive methods for early diagnosis of preeclampsia in the first trimester to implement preventive strategies effectively<sup>12- 14</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[67, 528, 886, 666]]<|/det|>
+Since the maternal decidua regulates the initial steps of maternal- embryo communication, decidualization resistance (DR) — characterized by defective endometrial cell differentiation— results in abnormal placentation, which has been associated with the etiology of major obstetric syndromes, including preeclampsia<sup>15- 19</sup>, even though symptoms may manifest later in gestation<sup>15- 19</sup>. Recently, we provided an in- depth multi- omics characterization of DR in former EOPE patients, further underscoring the uterine contribution to this pathological condition<sup>20</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[66, 680, 884, 891]]<|/det|>
+Analyzing plasma cell- free RNA (cfRNA) through liquid biopsy (i.e., from a blood sample) has emerged as a promising non- invasive tool for molecular monitoring in pregnancy, offering insights into physiological and pathological events<sup>21- 23</sup>. In this study, we prospectively analyzed the cfRNA profiles in 9,586 pregnant women across the three trimesters of pregnancy, comparing EOPE and LOPE with normotensive controls. This approach facilitated the characterization of the circulating transcriptome by mapping the tissue origins and transcriptional changes associated with EOPE and LOPE, revealing that both subtypes display distinct transcriptional differences compared to controls. Our research identified cfRNA profiles that exhibited robust predictive performance for EOPE in both the first (averaging 18.0 weeks before diagnosis) and second trimesters (averaging 8.5 weeks prior to clinical
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 881, 174]]<|/det|>
+onset), as well as for LOPE in the second trimester (14.9 weeks prior to clinical onset). Monitoring cfRNA profiles not only aids in predicting the risk of developing preeclampsia but also allows the differentiation of both subtypes of preeclampsia and the evaluation of different organ damage in affected patients, providing insights into their prognosis.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 223, 176, 237]]<|/det|>
+## Results  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 256, 587, 273]]<|/det|>
+## Clinical study design and participants baseline characteristics  
+
+<|ref|>text<|/ref|><|det|>[[115, 288, 882, 475]]<|/det|>
+A total of 9,586 pregnant women with singleton pregnancies were enrolled in this prospective and longitudinal study in fourteen tertiary hospitals in Spain (ClinicalTrials.gov Identifier: NCT04990141). Uncomplicated pregnancies that progressed to term (>37 weeks) were classified as normotensive controls, while those diagnosed with EOPE or LOPE, were categorized according to current established ACOG<sup>4</sup> and FIGO<sup>24</sup> clinical guidelines. This comprehensive approach enabled us to characterize cfRNA profiles throughout the progression of pregnancy in both controls and preeclampsia patients (EOPE and LOPE), as we collected samples during each trimester and at the time of preeclampsia diagnosis (Extended Data Fig. 1a).  
+
+<|ref|>text<|/ref|><|det|>[[115, 489, 882, 892]]<|/det|>
+In the study cohort, we included 85 patients who developed EOPE or LOPE, enabling a comparable modeling of both preeclampsia subtypes. Overall, this subset included EOPE (n=42), LOPE (n=43) patients and normotensive controls (n=75). We selected participants in the control group matched by gestational age at sample collection, maternal age, and parity to mitigate confounding effects due to biological variation (Extended Data Fig. 1b,c). Maternal characteristics, clinical symptoms, and birth outcomes are summarized in Table 1. There were no significant differences in maternal age, parity or smoking status between patients and controls. However, body mass index (BMI), a known preeclampsia risk factor, was notably higher in EOPE patients than in the control group (p = 0.0005); at the same time, LOPE patients didn't have a statistically higher BMI index (p = 0.1142). Additionally, ethnicity varied significantly between EOPE and controls (p = 0.097), but not between LOPE and controls (p = 0.1132). Natural conception rate was lower in EOPE patients compared to controls (p = 0.0426) but did not differ significantly in LOPE (p = 1). EOPE was diagnosed at 30.0 ± 3.4 weeks, with severe symptoms in 76.2% of patients; LOPE was diagnosed at 36.5 ± 1.8 weeks, with severe symptoms in 41.9% of patients. Birth outcomes for EOPE and LOPE included higher rates of small for gestational age, preterm birth, cesarean delivery, and lower fetal weight (p < 0.0001). Specifically, preterm deliveries occurred in 87.8% of EOPE patients and in 41.9% of LOPE patients, with cesarean sections required in 69.0% and 44.2% of patients, respectively. In contrast, all deliveries in the control group
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 881, 197]]<|/det|>
+occurred at term, and only \(16\%\) involved cesarean sections. Fetal sex did not differ between groups. EOPE patients had significantly higher rates of stillbirth \((11.9\%)\) and post- delivery complications \((p< 0.001)\) , with \(35.2\%\) of mothers and \(50.0\%\) of neonates requiring intensive care. In comparison, among patients with LOPE, \(18.6\%\) of mothers and \(16.3\%\) of newborns required intensive care, whereas no mothers or newborns in the control group needed intensive care.  
+
+<|ref|>text<|/ref|><|det|>[[115, 212, 882, 399]]<|/det|>
+For each participant, we analyzed total cfRNA from three peripheral blood samples collected between 9- 14 weeks (T1), 18- 28 weeks (T2), and at the time of diagnosis of EOPE and LOPE or after 28 weeks (T3) (Fig. 1). Data on the gestational weeks of blood sample collection are summarized in Supplementary Table 1. Due to clinical emergencies necessitating the termination of pregnancy, the third sample could not be collected from fourteen EOPE patients and seven LOPE patients. For the development of predictive models, the cohort was randomly stratified into discovery (70% of patients) and validation (30% of patients) sets, with the discovery set used to create the predictive model and the validation set to assess its accuracy (Extended Data Fig. 1a).  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 446, 781, 464]]<|/det|>
+## Profiling the tissue origin and dynamics of cfRNA in EOPE and LOPE through pregnancy  
+
+<|ref|>text<|/ref|><|det|>[[115, 479, 882, 618]]<|/det|>
+We analyzed a total of 29,871 cfRNA transcripts after applying quality filtering and normalization processes. To determine the tissue origins of the identified transcripts, we compared our cfRNA dataset to the Human Protein Atlas database<sup>24</sup>, focusing on transcripts classified as "enriched" or "enhanced" in specific tissues or organs. Our experimental protocol detected over \(90\%\) of these classified transcripts for each targeted organ or tissue of interest (Fig. 2a), indicating a robust coverage of tissue- specific cfRNA signatures in our dataset.  
+
+<|ref|>text<|/ref|><|det|>[[115, 633, 882, 819]]<|/det|>
+We then calculated the organ/tissue- specific signature score for patients and controls at three time points during pregnancy (T1, T2 and T3) (Fig. 2b). In EOPE patients, a significant increase in cfRNA transcripts from the liver, kidney, and decidua was identified at T2 \((p< 0.05)\) , indicating tissue specific damage approximately eight weeks before diagnosis (Fig. 2b). At T3, when clinical symptoms appear, EOPE patients displayed a significantly higher signature score \((p< 0.0001)\) for additional organs including brain, lungs, placenta, and lymphoid tissues, signaling widespread organ injury (Fig. 2b). In contrast in LOPE patients, tissue- specific transcripts suggesting organ damage was only observed at T3 \((p< 0.001)\) , with lower levels of significance than those in EOPE (Fig. 2b).  
+
+<|ref|>text<|/ref|><|det|>[[115, 834, 881, 900]]<|/det|>
+To decode cfRNA dynamics throughout pregnancy, we performed a differential abundance analysis at each time point, elucidating molecular changes in the circulating transcriptome associated with disease progression and offering insights into underlying mechanisms. At the time of diagnosis (T3),
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 881, 246]]<|/det|>
+we identified 24,336 transcripts with significantly altered expression in EOPE patients compared to controls (FDR < 0.05) (Extended Data Fig. 3a and Supplementary Table 2). In contrast, LOPE patients exhibited 11,859 differentially abundant transcripts (FDR < 0.05) (Extended Data Fig. 3b and Supplementary Table 3). Notably, only 251 cfRNAs showed differential abundance in T2 for EOPE patients (FDR < 0.05), whereas no differentially abundant cfRNAs were detected in T1 for either EOPE or LOPE patients, nor in T2 for LOPE. These findings suggest that transcriptomic alterations emerge as EOPE progresses, while LOPE remains largely unchanged.  
+
+<|ref|>text<|/ref|><|det|>[[115, 260, 883, 543]]<|/det|>
+Gene ontology overrepresentation analysis within the differentially abundant cfRNAs at diagnosis revealed biological processes indicative of fetal and maternal organ- specific damage (FDR < 0.05) (Extended Data Fig. 3c and Supplementary Table 4). Both, EOPE and LOPE patients displayed significant enrichment in key biological processes, including transport across the blood- brain barrier, renal water homeostasis, regulation of blood pressure and cognition, which are hallmark processes of the pathology. Importantly, signatures of fetal tissue damage were identified in both EOPE and LOPE, with a notably greater impact in EOPE patients. Distinct biological processes were associated with either EOPE or LOPE. In EOPE, overrepresentation analysis revealed significantly enriched pathways related to neuronal death, renal filtration, and immune dysfunction – including interleukin- 8 production, response to interleukin- 4, neutrophil- mediated immunity, and antimicrobial humoral immune response. In contrast, LOPE cfRNA profile showed signatures linked to heart and brain function (p < 0.05), suggesting significant damage to these organs.  
+
+<|ref|>text<|/ref|><|det|>[[115, 557, 882, 717]]<|/det|>
+Thus, cfRNA profile analysis at diagnosis (T3) underscores more extensive transcriptomic alterations in EOPE compared to LOPE, highlighting an exacerbated proinflammatory state as a defining feature. These findings underscore the impacts of the disease on multiple organ systems and suggest that cfRNA profiling can serve as a powerful tool for characterization of preeclampsia subtypes. Additionally, the identification of distinct biological processes linked to each preeclampsia subtype emphasizes the need for tailored therapeutic approaches targeting specific dysfunctions observed in EOPE and LOPE.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 768, 644, 785]]<|/det|>
+## Early prediction of EOPE and LOPE in the first trimester of pregnancy  
+
+<|ref|>text<|/ref|><|det|>[[115, 801, 882, 890]]<|/det|>
+Given the evidence that cfRNA profiles reflect molecular changes throughout pregnancy, disruptions in these pathways may help identify pregnancies at risk for EOPE or LOPE. Here, we developed a model for EOPE risk assessment based on plasma cfRNA profiles in the first trimester (T1), approximately 18.0 weeks before clinical onset.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 882, 246]]<|/det|>
+Our optimal predictive model for EOPE utilized 36 cfRNA transcripts (Supplementary Table 5) and was evaluated with a leave- one- out cross- validation approach to estimate disease risk. In a hold- out validation set, the model achieved a sensitivity of \(83\%\) and specificity of \(88\%\) , with an area under the receiver operator characteristic curve (AUC) of 0.85 (Fig. 3a and Supplementary Table 6). Nearly all samples were correctly classified, with minimal misclassifications observed reinforcing the model's robustness and indicating no evidence of overfitting (Fig. 3b). Relative contribution of individual cfRNA transcripts to the model's performance are detailed in Fig. 3c  
+
+<|ref|>text<|/ref|><|det|>[[115, 260, 882, 373]]<|/det|>
+Further analysis of these 36 transcripts revealed that 17 (47.2%) were identified as markers of DR in women with a history of severe preeclampsia, including CBR3, MMP7, MDK, TRIB1, PAEP<sup>20</sup>. The model also incorporates cfRNA transcripts known to be disrupted in preeclamptic placentas, such as RFLBN<sup>25</sup>, and CD74<sup>26</sup>, as well as others associated with intrauterine growth restriction, such as CCL4L2<sup>27</sup> and MYL6<sup>28</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[115, 389, 882, 550]]<|/det|>
+Using the same computational approach, we developed a predictive model for LOPE in the first trimester (T1), with predictions averaging 24.9 weeks before clinical onset. However, the model's performance in the validation set was limited, achieving a sensitivity of \(69\%\) , specificity of \(67\%\) , and an AUC of 0.68 (Fig. 3d and Supplementary Table 6), underscoring the challenges in early LOPE prediction. Misclassified samples are shown in Fig. 3e, and the relative contribution of individual cfRNAs to predictive accuracy detailed in Fig. 3f. While predictive capability was limited, analysis of the selected cfRNAs offers insights into LOPE mechanisms.  
+
+<|ref|>text<|/ref|><|det|>[[115, 566, 881, 656]]<|/det|>
+Further exploration revealed that several of these cfRNAs map to protein- coding genes with known roles in cardiovascular, hepatic, and immune functions, including PRR23D1, SnoRD126, CD52, TRDV3. Unlike EOPE, no cfRNA transcripts in this model were associated with decidua, underscoring distinct pathophysiological pathways for EOPE and LOPE.  
+
+<|ref|>text<|/ref|><|det|>[[115, 672, 881, 737]]<|/det|>
+In conclusion, our findings demonstrate the effectiveness of cfRNA signatures in predicting EOPE during the first trimester, while LOPE prediction remains challenging, likely reflecting fundamental differences in pathophysiology between EOPE and LOPE.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 788, 675, 805]]<|/det|>
+## Early prediction for EOPE and LOPE in the second trimester of pregnancy  
+
+<|ref|>text<|/ref|><|det|>[[115, 820, 881, 907]]<|/det|>
+We next investigated the potential for early detection of EOPE and LOPE in the second trimester (T2). The most effective predictive model for EOPE was based on 87 cfRNA transcripts (Supplementary Table 5), achieving a sensitivity of \(87\%\) and specificity of \(84\%\) , with an AUC of 0.85 in the validation set (Fig. 4a and Supplementary Table 6). Misclassified samples are shown in Fig. 4b, and importance scores for
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 880, 125]]<|/det|>
+each transcript are illustrated in Fig. 4c. This model reliably identifies patients at risk for EOPE between 18 and 28 weeks of gestation, approximately 8.5 weeks before clinical onset.  
+
+<|ref|>text<|/ref|><|det|>[[115, 140, 883, 374]]<|/det|>
+Further investigation into the tissue- specific origin of these transcripts revealed that 32 (36.8%) are associated with DR signature previously described in endometrial tissue from women with a history of severe preeclampsia, including CCL20, CXCR4, IGF1, RBP4, SQSTM1, WNT5A<sup>20</sup>. The persistence of decidual contributions as EOPE approaches underscores the maternal decidua's role in its pathophysiology. The model also includes inflammatory mediators such as SQSTM1, IL1B, CCL20, FASLG and TREM1, as well as transcripts encoding T cell receptors (e.g. TRAV21, TRBV27, TRBV5- 7). Additionally, it incorporates anti- inflammatory mediators like ALOX5AP, an immunosuppressive gene linked to recurrent miscarriage<sup>29</sup>, and IL19. Transcripts such as RBP4, which directly influences blood pressure regulation<sup>30</sup>, NRBF2 involved in autophagy and liver protection<sup>31</sup>, and WNT5A, and WNT5A, a key regulator of placental growth<sup>32</sup>, further support the model's clinical relevance.  
+
+<|ref|>text<|/ref|><|det|>[[115, 389, 881, 479]]<|/det|>
+The top- performing predictive model for LOPE at T2 included 92 cfRNAs (Supplementary Table 5), achieving a sensitivity of 86%, specificity of 89%, and an AUC of 0.88 in the validation cohort (Fig. 4d and Supplementary Table 6). An analysis of misclassified samples is shown in Fig. 4e, with the contributions of individual cfRNAs to predictive accuracy detailed in Fig. 4f.  
+
+<|ref|>text<|/ref|><|det|>[[115, 494, 882, 655]]<|/det|>
+Pathway enrichment analysis revealed that this model included cfRNAs related to immune function, such as CFHR1 and CFHR3, involved in complement activation, and immunoglobulin transcripts (e.g., IGKV3D- 20, IGKV3D- 11, IGHV5- 10- 1, IGHV3- 69- 1), and CXCR5, linked to B- cell migration<sup>33</sup>. Additionally, the model incorporates a cfRNA corresponding to HISLA, highly expressed in the liver<sup>34</sup>, and LINC01419<sup>35</sup>. Notably, most predictive cfRNAs were classified as non- coding RNAs or pseudogenes with no annotated function. In contrast to the EOPE model, this LOPE model includes only two cfRNAs related to DR, HES4 and SPEF1.  
+
+<|ref|>sub_title<|/ref|><|det|>[[116, 707, 201, 721]]<|/det|>
+## Discussion  
+
+<|ref|>text<|/ref|><|det|>[[115, 737, 882, 900]]<|/det|>
+Previous efforts to develop screening tests for preeclampsia have primarily focused on circulating biomarkers related to placental dysfunction, such as sFLT1 and PlGF<sup>36</sup>. These tests have been validated for use starting at 23 weeks of gestation, with their strongest predictive accuracy typically observed within two weeks of symptom onset. Consequently, they are recommended for patients with suspected preeclampsia<sup>37,38</sup>. While these tests are particularly useful for short- term prediction, placental dysfunction- based tests are also utilized as early as the first trimester. They are often combined with maternal epidemiological factors and ultrasound or Doppler parameters. However,
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 881, 174]]<|/det|>
+they face significant limitations in their effectiveness and application<sup>39- 42</sup>. In settings where guidelines from the National Institute for Health and Care Excellence (NICE) and the ACOG are applied, screening primarily relies on pregnancy- related factors and maternal characteristics. While this approach minimizes additional costs, it has low sensitivity (<41%)<sup>43,44</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[115, 188, 882, 374]]<|/det|>
+Predictive models based on cfRNAs from liquid biopsy, grounded in biological plausibility and applicable early in pregnancy, offering potential improvements for the clinical management of preeclampsia<sup>21,23,45</sup>, yet they have not been clinically applied. Building on this foundation, we prospectively collected blood samples from 9,586 pregnant women across three gestational trimesters (9–42 weeks), generating a comprehensive longitudinal dataset of cfRNA profiles related to EOPE or LOPE progression. The performance metrics demonstrate substantial advancements in leveraging cfRNA signatures for early detection of EOPE in both the first and second trimesters, as well as LOPE in the second trimester.  
+
+<|ref|>text<|/ref|><|det|>[[115, 389, 882, 670]]<|/det|>
+Our study represents an advancement over previously published research in this field. By leveraging a study design that provides a dynamic and comprehensive view of cfRNA changes across all three trimesters of pregnancy, we capture critical information spanning pre- manifestation stages to disease onset. This approach enables earlier detection and offers deeper insights into disease progression compared to prior studies. Additionally, our clinically well- defined cohort allows for precise differentiation between EOPE and LOPE. The transcriptional insights we have uncovered highlight distinct pathophysiological trajectories for EOPE and LOPE, paving the way for more targeted and patient- specific interventions. Moreover, the cfRNA- based algorithms developed in our study demonstrate improved predictive performance, offering clinically relevant tools for distinguishing between EOPE and LOPE. These algorithms outperform current state- of- the- art methods, providing a robust framework for earlier and more accurate diagnosis, ultimately enhancing patient care and outcomes.  
+
+<|ref|>text<|/ref|><|det|>[[115, 686, 882, 870]]<|/det|>
+Our analysis highlighted the tissue- specific origins of the detected cfRNAs, offering further insight into the pathophysiology of both subtypes of preeclampsia. For EOPE patients, early signs of tissue distress were observed in the liver, kidney, and decidua at T2, suggesting that these organs may be affected up to eight weeks before clinical diagnosis. By the time of the clinical onset (T3), cfRNA levels associated with critical organs such as the placenta, brain and lungs showed marked elevation in EOPE, indicating widespread organ involvement likely due to apoptotic processes releasing cfRNA into circulation. In LOPE patients, although cfRNA levels also increased by T3, levels were lower compared to EOPE. Furthermore, differential abundance analysis at diagnosis revealed distinct transcriptomic profile in
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 880, 125]]<|/det|>
+EOPE, with more pronounced cfRNA changes than in LOPE reflecting potential differences in severity and inflammatory response between both preeclampsia subtypes.  
+
+<|ref|>text<|/ref|><|det|>[[115, 140, 881, 254]]<|/det|>
+These distinctions between the subtypes extended to the biological roles of cfRNAs included in the predictive models. In models predicting EOPE, a substantial proportion of cfRNA transcripts were associated with genes involved in decidualization and DR, along with some placental- related transcripts. In contrast, cfRNA transcripts associated with LOPE prediction reflecting broader systemic contributions including placental malfunction.  
+
+<|ref|>text<|/ref|><|det|>[[115, 270, 881, 431]]<|/det|>
+A limitation of our study is that the control group consisted of women without obstetric complications. Although this design allows a clear distinction between patients with preeclampsia and those with uncomplicated pregnancies, it may not fully reflect the diverse clinical backgrounds encountered in real- world settings, where women who develop other obstetric conditions later in pregnancy may also be screened. In addition, external validation is required to further validate of the diagnostic performance, which is currently underway (ClinicalTrials.gov Identifier: NCT06716242), to ensure the generalizability of the predictive models before their clinical application.  
+
+<|ref|>text<|/ref|><|det|>[[115, 446, 881, 560]]<|/det|>
+Overall, by analyzing circulating RNAs from a single blood sample at T1 or T2, our approach provides a reliable, standardized diagnostic measure that minimizes subjective interpretation and reduces variability in clinical decision- making. This streamlined strategy simplifies risk stratification, improving both the accuracy and efficiency of preeclampsia screening and facilitating personalized patient monitoring.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 611, 321, 627]]<|/det|>
+## MATERIAL AND METHODS  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 644, 219, 660]]<|/det|>
+## Study design  
+
+<|ref|>text<|/ref|><|det|>[[115, 676, 882, 910]]<|/det|>
+This prospective, multicenter case- control study involved fourteen hospitals across Spain (ClinicalTrials.gov Identifier: NCT04990141). Given the incidence rate of preeclampsia, the cohort size was designed to capture a minimum of 30 patients of EOPE over the course of the study. Approval was obtained from the relevant Clinical Research Ethics Committees at each site, and written informed consent was collected from all participants prior to blood collection and sample anonymization. A total of 9,586 pregnant women were enrolled based on the following criteria: signed informed consent, age over 18, singleton pregnancy, and first blood sample collection within 9–14 gestational weeks. Each participant provided 20 mL of peripheral blood in the three trimesters of pregnancy, coinciding with routine clinical follow- up: (T1) 9–14 weeks, (T2) 18–28 weeks, and (T3) >28 weeks or at the time of preeclampsia diagnosis. Gestational age was confirmed via ultrasound during the first trimester.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 883, 410]]<|/det|>
+Clinical data for each participant were recorded in an electronic data capture system. All blood samples were processed to isolate plasma and stored at \(- 80^{\circ}C\) until pregnancy outcomes were available. Preeclampsia patients were diagnosed following ACOG<sup>4</sup> and FIGO<sup>46</sup> guidelines, as per the clinical protocol of each hospital involved. EOPE patients (n=42) and a subset of LOPE patients (n=43) were selected and matched with normotensive pregnant women with uncomplicated pregnancies as controls (n=75). Participants in the control group were selected based on matching gestational age at the time of blood collection, maternal age, and parity, utilizing Euclidean distance for optimal pairing. Patients and controls were randomly stratified following a 70:30 proportion into two cohorts: discovery and validation sets. Sample sizes for the EOPE, LOPE, and control groups in each cohort are detailed in Supplementary Table 6. The discovery set was used for feature selection, model training, and optimization, with model performance assessed by leave-one-out cross-validation. The optimal model from this process was then applied to the validation set to assess the predictive performance, yielding metrics based on an unexposed sample set. The bioinformatic workflow is detailed in Extended Data Fig. 4.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 428, 405, 445]]<|/det|>
+## Blood sample processing and storage  
+
+<|ref|>text<|/ref|><|det|>[[117, 460, 882, 550]]<|/det|>
+Peripheral blood samples (20 mL) were collected in Streck Cell- Free DNA BCT tubes (Illumina, San Diego, CA), stored, shipped at room temperature, and processed within seven days to obtain the plasma fraction. All blood samples were centrifuged for 15 min at 1,600 x g and \(4^{\circ}C\) . Plasma was transferred to a new collection tube and stored at \(- 80^{\circ}C\) until use.  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 567, 524, 584]]<|/det|>
+## cfrNA isolation, library preparation, and sequencing  
+
+<|ref|>text<|/ref|><|det|>[[115, 599, 883, 857]]<|/det|>
+Plasma supernatant samples (n=457) from the study patients (n=160) were centrifuged for 10 min at 13,000 x g. Following the manufacturer's protocol, cfRNA from 2 mL of plasma was isolated using MiRNeasy Serum/Plasma Advanced Kit (Qiagen, Hilden, Germany). According to the manufacturer's protocol, cDNA libraries from total cfRNA samples were prepared using Illumina RNA Prep with Enrichment (L) Tagmentation (Illumina, San Diego, CA). cDNA libraries were quantified using an Agilent D1000 ScreenTape in a 4200 TapeStation system (Agilent Technologies Inc). Libraries were normalized to 10 nM and pooled in equal volumes. The pool concentration was quantified by qPCR using the KAPA Library Quantification Kit (Kapa Biosystems Inc) and an Agilent D1000 ScreenTape in a 4200 TapeStation system (Agilent Technologies Inc). The mean value was used to establish pool concentration, which was then sequenced in a NextSeq 500/550 High Output kit with 2.5 cartridges of 150 cycles (Illumina, San Diego, CA).  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 873, 333, 889]]<|/det|>
+## Sequencing data processing
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 881, 220]]<|/det|>
+Raw reads were aligned to the human reference genome (GRCh38 Gencode v38 Primary Assembly) using STAR (2.7.10a). The SAM/BAM files were further processed using SAMtools (v.1.6). Only reads with mapping quality more significant than \(90\%\) were maintained (MAPQ score obtained from the alignment). The duplicated reads were removed with Picard MarkDuplicates (v.2.27.4). The mapping and the quantification of the reads were done using featureCounts (v.2.0.1). Read statistics were estimated using FastQC (v.0.11.9) and ResqQC (v.5.0.1) and summarized using MultiQC (v.1.13).  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 238, 300, 254]]<|/det|>
+## Sample quality filtering  
+
+<|ref|>text<|/ref|><|det|>[[115, 270, 882, 430]]<|/det|>
+Three key quality parameters related to the sequencing process were estimated for each analyzed sample: RNA degradation, DNA contamination, and rRNA fraction as previously defined<sup>22,47</sup>. Samples were retained for further analysis if they met the established cut- off values for each parameter: RNA degradation (cut- off: \(40\%\) ), DNA contamination (cut- off ratio: 3), and rRNA fraction (cut- off: \(15\%\) ). Principal Component Analysis (PCA) was used as an additional quality control measure. Samples deviating by more than 3 standard deviations from the mean of the first and second components for each dataset were excluded from the analysis.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 448, 361, 464]]<|/det|>
+## Differential abundance analysis  
+
+<|ref|>text<|/ref|><|det|>[[115, 480, 882, 618]]<|/det|>
+CfRNAs differentially abundant between EOPE or LOPE patients and controls at each time point (T1, T2, T3) were identified using the limma- Voom method from the Bioconductor package limma (v3.60.5). For the T3 samples, comparisons only included patients whose samples were collected at the time of EOPE or LOPE diagnosis and gestationally matched control samples collected during routine medical appointments. Genes with False Discovery Rate (FDR) less or equal to 0.05 were considered statistically significant.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 635, 275, 650]]<|/det|>
+## Enrichment analysis  
+
+<|ref|>text<|/ref|><|det|>[[115, 667, 881, 755]]<|/det|>
+Gene Ontology (GO) analyses were performed to identify biological processes using the enrichGO function from the clusterProfiler R package (v4.2.2). The input consisted of cfRNAs that were differentially abundant between EOPE and controls, as well as LOPE and controls (FDR \(< 0.05\) ). The p- value adjustment method used was FDR, with a significance threshold set at 0.05 (FDR \(< 0.05\) ).  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 773, 445, 789]]<|/det|>
+## Estimating signature scores for each tissue  
+
+<|ref|>text<|/ref|><|det|>[[115, 805, 881, 894]]<|/det|>
+Gene sets for each tissue of interest were derived from the Human Protein Atlas database<sup>24</sup>, which includes gene expression data across tissues, focusing specifically on transcripts classified as either "enriched" or "enhanced" within those tissues. The signature score in our dataset was calculated by summing the log- transformed, normalized counts of each gene in the set. For the T3 samples,
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 881, 150]]<|/det|>
+comparisons included only those patients whose samples were collected at the time of EOPE or LOPE diagnosis and gestationally matched control samples collected during routine medical appointments, to avoid potential bias. Differences between groups were assessed using the Wilcoxon rank- sum test.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 167, 223, 183]]<|/det|>
+## Data splitting  
+
+<|ref|>text<|/ref|><|det|>[[117, 198, 882, 287]]<|/det|>
+Our study cohort was divided into discovery and validation sets to develop and evaluate the predictive models, following best practices to prevent overfitting in artificial intelligence. Using stratified sampling based on obstetric outcomes (patient/control groups) and the scikit- learn library (v1.5.1) in Python, \(70\%\) of participants were allocated to the discovery set.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 304, 329, 320]]<|/det|>
+## CfRNA count normalization  
+
+<|ref|>text<|/ref|><|det|>[[115, 336, 883, 617]]<|/det|>
+CfRNAs were filtered based on their detection value, and only cfRNAs with levels over more than 0.5 counts per million reads (CPMs) in \(\geq 70\%\) of discovery samples after removing outlier samples were kept. Discovery set CPMs were normalized using the "deseq median ratio normalization" with pydeseq2 (v0.4.1). The validation set were then normalized with the same algorithm using size factors, from discovery set as described in MLSeq package<sup>48,49</sup>. Batch effect and other possible confounding factors were assessed using PCA, hierarchical clustering with Spearman correlation as a distance metric, and variance component analysis. A differential abundance analysis was performed on the processed counts of the discovery set, comparing the patients and controls with the shrunk log2 fold changes from pyDeseq2 (v0.4.1) using the default parameters and options. The ranking of genes, based on their adjusted and non- adjusted p- value was obtained to see the differential abundant cfRNA of controls versus patients. Finally, the normalized counts of each sample of discovery and validation sets were re- scaled to 0- 1 range with a min- max scaling process.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 634, 252, 649]]<|/det|>
+## Feature selection  
+
+<|ref|>text<|/ref|><|det|>[[115, 666, 883, 899]]<|/det|>
+The Lasso regression model was used to select the more relevant cfRNAs to discriminate between patients and controls. The discovery dataset was used with a lasso regression algorithm (v1.5.2, sklearn.linear_model.Lasso) with a penalty term (alpha) of 0.5 and the case condition as a dependent part and the cfRNA abundance levels as the independent components, resulting in a regression formula that assigns a coefficient to each cfRNA variable, indicating the correlation between the condition and each variable. The number of cfRNAs selected was determined by a minimum coefficient threshold, which determined whether a cfRNA was relevant or not. Different minimum coefficient thresholds, ranging from 0 to the maximum coefficient in increments of 0.05, were tested to determine the optimal set of cfRNAs. The F1- score was calculated for each set of cfRNAs using the strategy of leave- one- out cross- validation, and the set that yield the highest F1- score metric was selected. Lasso
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 84, 881, 150]]<|/det|>
+regression was chosen over other feature selection methods due to the relatively small sample size, which can lead to model overfitting. The penalty term of the model helps to counteract overfitting by shrinking and selecting features with less importance<sup>50</sup>.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 166, 405, 183]]<|/det|>
+## Algorithm selection and optimization  
+
+<|ref|>text<|/ref|><|det|>[[115, 198, 883, 479]]<|/det|>
+For the development of the optimal predictor, the discovery set was used, with cfRNA selection performed using the lasso regression method as previously described. . Six different algorithms were tested with python (v3.10.6): support vector machine (v1.5.2, sklearn.svm.SVC), Elastic Net Linear Regression (v1.5.2, sklearn.linear_model.EastNet), Lasso Linear Regression (v1.5.2, sklearn.linear_model.Lasso), Random Forest (v1.5.2, sklearn.ensemble.RandomForestClassifier), XGBoost (v1.7.6 xgboost. XGBClassifier) and TabPFN (v0.1.10 tabpfn.TabPFNClassifier). Each algorithm was trained with the best parameters calculated with a grid search applied with a cross- fold strategy. The evaluation of the predictive capacity of each model was done with a leave- one- out cross- validation with the discovery samples. The algorithm providing the best F1- score was selected for each group of samples: EOPE in the first trimester (EOPE T1), in the second trimester (EOPE T2), LOPE in the first trimester (LOPE T1) and LOPE in the second trimester (LOPE T2). The resulting chosen algorithms were TabPFN for T1 EOPE and T2 LOPE and support vector machine for T2 EOPE and T1 LOPE.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 496, 315, 512]]<|/det|>
+## Predictive model training  
+
+<|ref|>text<|/ref|><|det|>[[115, 528, 883, 737]]<|/det|>
+For each model (EOPE T1, LOPE T1, EOPE T2, LOPE T2), the ML algorithm showing the highest F1- score and its best parameters was trained with the discovery dataset. To evaluate the predictive capacity with the discovery data, a strategy of leave- one- out was performed. The selected algorithm was trained N number of times. In each iteration, one sample was isolated, and the rest were used to fit the model. The fitted model was used to predict the label of the isolated sample, and the result of the prediction was added to a pool of predicted labels that were used to calculate the discovery leave- one- out metrics. Finally, the algorithm was fitted with all the discovery samples, and the obtained trained model was used to predict the labels in the validation dataset and evaluate the performance with never seen samples.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 754, 256, 769]]<|/det|>
+## Model evaluation  
+
+<|ref|>text<|/ref|><|det|>[[115, 787, 881, 874]]<|/det|>
+We evaluated the predictive performance of each model (EOPE T1, LOPE T1, EOPE T2, LOPE T2) using two approaches: (1) leave- one- out cross- validation in the discovery dataset, and (2) predictions on the validation dataset using the final model. Model performance was assessed with key metrics, including accuracy, sensitivity, specificity, AUC, and F1- score.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 892, 245, 907]]<|/det|>
+## Data availability
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 84, 881, 245]]<|/det|>
+The cell- free RNA- sequencing data generated for this manuscript has been uploaded to Gene Expression Omnibus database under the accession number (REDACTED 11 AUGUST 2025). The uploaded data include Raw count matrices obtained with FeatureCount. The raw sequences are not publicly available due to privacy concerns. However, they are available from the corresponding authors (C.S, carlos.simon@uv.es; T.G, tgarrido@fundacioncarlossimon.com) upon reasonable request and with permission of the Institutional Review Board of the Spanish hospitals involved.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 296, 217, 308]]<|/det|>
+## REFERENCES  
+
+<|ref|>text<|/ref|><|det|>[[110, 303, 876, 912]]<|/det|>
+1. Organization WH. 2018 Global Reference List of 100 Core Health Indicators (plus health-related SDGs). Geneva; 2018.  
+2. Say L, Chou D, Gemmill A, et al. Global causes of maternal death: a WHO systematic analysis. Lancet Glob Health. Jun 2014;2(6):e323-33. doi:10.1016/S2214- 109X(14)70227-X  
+3. Moreno I, Capalbo A, Mas A, et al. The human periconceptional maternal-embryonic space in health and disease. Physiol Rev. Jul 01 2023;103(3):1965-2038. doi:10.1152/physrev.00050.2021  
+4. ACOG Practice Bulletin No. 202 Summary: Gestational Hypertension and Preeclampsia. Obstet Gynecol. Jan 2019;133(1):1. doi:10.1097/AOG.0000000000003019  
+5. Hinkle SN, Schisterman EF, Liu D, et al. Pregnancy Complications and Long-Term Mortality in a Diverse Cohort. Circulation. Mar 28 2023;147(13):1014-1025. doi:10.1161/CIRCULATIONAHA.122.062177  
+6. Stevens W, Shih T, Incerti D, et al. Short-term costs of preeclampsia to the United States health care system. Am J Obstet Gynecol. Sep 2017;217(3):237-248.e16. doi:10.1016/j.ajog.2017.04.032  
+7. Lisonkova S, Joseph KS. Incidence of preeclampsia: risk factors and outcomes associated with early- versus late-onset disease. Am J Obstet Gynecol. Dec 2013;209(6):544.e1-544.e12. doi:10.1016/j.ajog.2013.08.019  
+8. Lisonkova S, Sabr Y, Mayer C, Young C, Skoll A, Joseph KS. Maternal morbidity associated with early-onset and late-onset preeclampsia. Obstet Gynecol. Oct 2014;124(4):771-781. doi:10.1097/AOG.0000000000000472  
+9. de Havenon A, Delic A, Stulberg E, et al. Association of Preeclampsia With Incident Stroke in Later Life Among Women in the Framingham Heart Study. JAMA Netw Open. Apr 01 2021;4(4):e215077. doi:10.1001/jamanetworkopen.2021.5077  
+10. Conti-Ramsden FI, Nathan HL, De Greeff A, et al. Pregnancy-Related Acute Kidney Injury in Preeclampsia: Risk Factors and Renal Outcomes. Hypertension. Nov 2019;74(5):1144-1151. doi:10.1161/HYPERTENSIONAHA.119.13089  
+11. Lisonkova S, Bone JN, Muraca GM, et al. Incidence and risk factors for severe preeclampsia, hemolysis, elevated liver enzymes, and low platelet count syndrome, and eclampsia at preterm and term gestation: a population-based study. Am J Obstet Gynecol. Nov 2021;225(5):538.e1-538.e19. doi:10.1016/j.ajog.2021.04.261  
+12. Rolnik DL, Wright D, Poon LC, et al. Aspirin versus Placebo in Pregnancies at High Risk for Preterm Preeclampsia. N Engl J Med. Aug 17 2017;377(7):613-622. doi:10.1056/NEJMoa1704559  
+13. Davidson KW, Barry MJ, Mangione CM, et al. Aspirin Use to Prevent Preeclampsia and Related Morbidity and Mortality: US Preventive Services Task Force Recommendation Statement. JAMA. Sep 28 2021;326(12):1186-1191. doi:10.1001/jama.2021.14781  
+14. Bibbins-Domingo K, Grossman DC, Curry SJ, et al. Screening for Preeclampsia: US Preventive Services Task Force Recommendation Statement. JAMA. Apr 25 2017;317(16):1661-1667. doi:10.1001/jama.2017.3439
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[58, 87, 875, 899]]<|/det|>
+15. Garrido-Gomez T, Dominguez F, Quiñonero A, et al. Defective decidualization during and after severe preeclampsia reveals a possible maternal contribution to the etiology. Proc Natl Acad Sci U S A. 10 03 2017;114(40):E8468-E8477. doi:10.1073/pnas.1706546114
+16. Garrido-Gomez T, Castillo-Marco N, Clemente-Ciscar M, et al. Disrupted PGR-B and ESR1 signaling underlies defective decidualization linked to severe preeclampsia. Elife. 10 28 2021;10doi:10.7554/eLife.70753
+17. Garrido-Gomez T, Castillo-Marco N, Cordero T, Simón C. Decidualization resistance in the origin of preeclampsia. Am J Obstet Gynecol. 02 2022;226(2S):S886-S894. doi:10.1016/j.ajog.2020.09.039
+18. Ng SW, Norwitz GA, Pavlicev M, Tilburgs T, Simón C, Norwitz ER. Endometrial Decidualization: The Primary Driver of Pregnancy Health. Int J Mol Sci. Jun 08 2020;21(11)doi:10.3390/ijms21114092
+19. Woods L, Perez-Garcia V, Kieckbusch J, et al. Decidualisation and placentation defects are a major cause of age-related reproductive decline. Nat Commun. Sep 05 2017;8(1):352. doi:10.1038/s41467-017-00308-x
+20. Muñoz-Blatt I, Pérez-Moraga R, Castillo-Marco N, et al. Multi-Omics based mapping of Decidualization Resistance in patients with a history of Severe Preeclampsia. Nat Med. Jan 07 2025. doi:10.1038/s41591-024-03407-7.
+21. Munchel S, Rohrback S, Randise-Hinchliff C, et al. Circulating transcripts in maternal blood reflect a molecular signature of early-onset preeclampsia. Sci Transl Med. Jul 01 2020;12(550)doi:10.1126/scitranslmed.aaz0131
+22. Moufarrej MN, Wong RJ, Shaw GM, Stevenson DK, Quake SR. Investigating Pregnancy and Its Complications Using Circulating Cell-Free RNA in Women's Blood During Gestation. Front Pediatr. 2020;8:605219. doi:10.3389/fped.2020.605219
+23. Rasmussen M, Reddy M, Nolan R, et al. RNA profiles reveal signatures of future health and disease in pregnancy. Nature. 01 2022;601(7893):422-427. doi:10.1038/s41586-021-04249-w
+24. Uhlén M, Fagerberg L, Hallström BM, et al. Proteomics. Tissue-based map of the human proteome. Science. Jan 23 2015;347(6220):1260419. doi:10.1126/science.1260419
+25. Nevalainen J, Skarp S, Savolainen ER, Ryynänen M, Järvenpää J. Intrauterine growth restriction and placental gene expression in severe preeclampsia, comparing early-onset and late-onset forms. J Perinat Med. Oct 26 2017;45(7):869-877. doi:10.1515/jpm-2016-0406
+26. Przybyl L, Haase N, Golic M, et al. CD74-Downregulation of Placental Macrophage-Trophoblastic Interactions in Preeclampsia. Circ Res. Jun 24 2016;119(1):55-68. doi:10.1161/CIRCRESAHA.116.308304
+27. Medina-Bastidas D, Guzmán-Huerta M, Borboa-Olivares H, et al. Placental Microarray Profiling Reveals Common mRNA and lncRNA Expression Patterns in Preeclampsia and Intrauterine Growth Restriction. Int J Mol Sci. May 20 2020;21(10)doi:10.3390/ijms21103597
+28. Timofeeva AV, Fedorov IS, Brzhozovskiy AG, et al. miRNAs and Their Gene Targets-A Clue to Differentiate Pregnancies with Small for Gestational Age Newborns, Intrauterine Growth Restriction, and Preeclampsia. Diagnostics (Basel). Apr 20 2021;11(4)doi:10.3390/diagnostics11040729
+29. Wang F, Jia W, Fan M, et al. Single-cell Immune Landscape of Human Recurrent Miscarriage. Genomics Proteomics Bioinformatics. Apr 2021;19(2):208-222. doi:10.1016/j.gpb.2020.11.002
+30. Kraus BJ, Sartoretto JL, Polak P, et al. Novel role for retinol-binding protein 4 in the regulation of blood pressure. FASEB J. Aug 2015;29(8):3133-40. doi:10.1096/fj.14-266064
+31. Lu J, He L, Behrends C, et al. NRBF2 regulates autophagy and prevents liver injury by modulating Atg14L-linked phosphatidylinositol-3 kinase III activity. Nat Commun. May 22 2014;5:3920. doi:10.1038/ncomms4920
+32. Lu Z, Hardt J, Kim JJ. Global analysis of genes regulated by HOXA10 in decidualization reveals a role in cell proliferation. Mol Hum Reprod. Jun 2008;14(6):357-66. doi:10.1093/molehr/gan023
+33. Stelzer G, Rosen N, Plaschkes I, et al. The GeneCards Suite: From Gene Data Mining to Disease Genome Sequence Analyses. Curr Protoc Bioinformatics. Jun 20 2016;54:1.30.1-1.30.33. doi:10.1002/cpbi.5
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[58, 80, 876, 757]]<|/det|>
+34. Fagerberg L, Hallström BM, Oksvold P, et al. Analysis of the human tissue-specific expression by genome-wide integration of transcriptomics and antibody-based proteomics. Mol Cell Proteomics. Feb 2014;13(2):397-406. doi:10.1074/mcp.M113.035600  
+35. Dang H, Chen L, Tang P, et al. LINC01419 promotes cell proliferation and metastasis in hepatocellular carcinoma by enhancing NDRG1 promoter activity. Cell Oncol (Dordr). Oct 2020;43(5):931-947. doi:10.1007/s13402-020-00540-6  
+36. Zeisler H, Llurba E, Chantraine F, et al. Predictive Value of the sFlt-1:PlGF Ratio in Women with Suspected Preeclampsia. N Engl J Med. Jan 07 2016;374(1):13-22. doi:10.1056/NEJMoa1414838  
+37. Hund M, Allegranza D, Schoedl M, Dilba P, Verhagen-Kamerbeek W, Stepan H. Multicenter prospective clinical study to evaluate the prediction of short-term outcome in pregnant women with suspected preeclampsia (PROGNOSIS): study protocol. BMC Pregnancy Childbirth. Sep 18 2014;14:324. doi:10.1186/1471-2393-14-324  
+38. Thadhani R, Lemoine E, Rana S, et al. Circulating Angiogenic Factor Levels in Hypertensive Disorders of Pregnancy. NEJM Evid. Dec 2022;1(12):EVIDoa2200161. doi:10.1056/EVIDoa2200161  
+39. Chaeemsaithong P, Sahota DS, Poon LC. First trimester preeclampsia screening and prediction. Am J Obstet Gynecol. Feb 2022;226(2S):S1071-S1097.e2. doi:10.1016/j.ajog.2020.07.020  
+40. Rana S, Lemoine E, Granger JP, Karumanchi SA. Preeclampsia: Pathophysiology, Challenges, and Perspectives. Circ Res. Mar 29 2019;124(7):1094-1112. doi:10.1161/CIRCRESAHA.118.313276  
+41. Agrawal S, Cerdeira AS, Redman C, Vatish M. Meta-Analysis and Systematic Review to Assess the Role of Soluble FMS-Like Tyrosine Kinase-1 and Placenta Growth Factor Ratio in Prediction of Preeclampsia: The SaPPPhirE Study. Hypertension. Feb 2018;71(2):306-316. doi:10.1161/HYPERTENSIONAHA.117.10182  
+42. MacDonald TM, Walker SP, Hannan NJ, Tong S, Kaitu'u-Lino TJ. Clinical tools and biomarkers to predict preeclampsia. EBioMedicine. Jan 2022;75:103780. doi:10.1016/j.ebiom.2021.103780  
+43. ACOG Committee Opinion No. 743: Low-Dose Aspirin Use During Pregnancy. Obstet Gynecol. Jul 2018;132(1):e44-e52. doi:10.1097/AOG.0000000000002708  
+44. Visintin C, Mugglestone MA, Almerie MQ, et al. Management of hypertensive disorders during pregnancy: summary of NICE guidance. BMJ. Aug 25 2010;341:c2207. doi:10.1136/bmj.c2207  
+45. Moufarrej MN, Vorperian SK, Wong RJ, et al. Early prediction of preeclampsia in pregnancy with cell-free RNA. Nature. Feb 2022;602(7898):689-694. doi:10.1038/s41586-022-04410-z  
+46. Poon LC, Shennan A, Hyett JA, et al. The International Federation of Gynecology and Obstetrics (FIGO) initiative on pre-eclampsia: A pragmatic guide for first-trimester screening and prevention. Int J Gynaecol Obstet. May 2019;145 Suppl 1(Suppl 1):1-33. doi:10.1002/ijgo.12802  
+47. Tong YK, Lo YM. Diagnostic developments involving cell-free (circulating) nucleic acids. Clin Chim Acta. Jan 2006;363(1-2):187-96. doi:10.1016/j.cccn.2005.05.048  
+48. Goksuluk D, Zarrasiz G, Korkmaz S, et al. MLSeq: Machine learning interface for RNA-sequencing data. Comput Methods Programs Biomed. Jul 2019;175:223-231. doi:10.1016/j.cmpb.2019.04.007  
+49. Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 2014;15(12):550. doi:10.1186/s13059-014-0550-8  
+50. Ji Hyung Lee ZS, Zhan Gao. On LASSO for predictive regression. Journal of Econometrics. 2022;229(2):322-349.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[148, 130, 660, 752]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 770, 881, 907]]<|/det|>
+<center>Fig. 1. Overview of sample collection, preeclampsia diagnosis, and delivery time points across patient and control groups. Bar graph illustrating the number of samples collected at each gestational week for the EOPE (a), LOPE (c) and control (e) groups. Color represents the time point of sample collection: T1 (9-14 gestational weeks); T2 (18-28 gestational weeks); T3 (at the time of preeclampsia diagnosis or >28 gestational weeks). Density plot showing the relative frequency of preeclampsia diagnosis and delivery across gestational weeks for the EOPE (b), LOPE (d) and control (f) groups. </center>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[170, 85, 722, 628]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 641, 880, 897]]<|/det|>
+<center>Fig. 2. CfRNA abundance by organ/tissue origin in EOPE, LOPE patients and controls. (a) Total number of transcripts detected in different organs/tissues of interest. The percentage indicates the proportion of cfRNA transcripts identified relative to the total transcripts catalogued for each organ/tissue in the Human Protein Atlas database. (b) Box plots show cfRNA abundance scores by tissue of origin at each time point, calculated as the sum of log-transformed CPM-TMM normalised counts. The color represents the group. The horizontal line represents the median, with the lower and upper edges marking the 25th and 75th percentiles, respectively. Whiskers extend to 1.5 times the interquartile range. Sample sizes for each time point and group are as follows: T1 (EOPE, n=41; LOPE, n=43; control, n=75); T2 (EOPE, n=40; LOPE, n=41; control, n=73); T3 (EOPE, n=18 vs. control, n=36; LOPE, n=23 vs. control, n=39). P-values were determined by Wilcoxon rank-sum test with two tails. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. </center>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[166, 80, 800, 730]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 750, 880, 911]]<|/det|>
+<center>Fig. 3. Performance evaluation and feature importance analysis of the first trimester (T1) predictive models for EOPE and LOPE. Receiver operating characteristic curve for the predictive models of EOPE (a) and LOPE (d). The X-axis represents the False Negative Rate, while the Y-axis represents the True Positive Rate. Violin plots displaying the distribution of correctly and misclassified patients and controls based on the classifier score obtained from the predictive model for EOPE (b) and LOPE (e). The X-axis shows the real obstetric outcome, while the Y-axis shows the predicted outcome. Participants with a classifier score above the threshold were classified as preeclampsia patients, and those below the </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[57, 83, 883, 199]]<|/det|>
+threshold were predicted as controls. Bar plot illustrating each cfRNAs contribution to the performance of the predictive models of EOPE (c) and LOPE (f). The X-axis shows the feature importance scores, which quantify the relative contribution of each cfRNA to the model's predictions, with higher scores indicating features that play a more significant role in discriminating between outcomes. CfRNAs associated with DR are marked with an asterisk. AUC, area under the curve.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[144, 93, 825, 875]]<|/det|>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 882, 390]]<|/det|>
+Fig. 4. Performance evaluation and feature importance analysis of the second trimester (T2) predictive models for EOPE and LOPE. Receiver operating characteristic curve for the predictive models of EOPE (a) and LOPE (d). The X-axis represents the False Negative Rate, while the Y-axis represents the True Positive Rate. Violin plots displaying the distribution of correctly and misclassified patients and controls based on the classifier score obtained from the predictive model for EOPE (b) and LOPE (e). The X-axis shows the obstetric outcome, while the Y-axis shows the predicted outcome. Participants with a predicted probability above the threshold were classified as preeclampsia patients, and those below the threshold were predicted as controls. Bar plot illustrating each cfRNAs contribution to the performance of the predictive models of EOPE (c) and LOPE (f). The X-axis shows the feature importance scores, which quantify the relative contribution of each cfRNA to the model's predictions, with higher scores indicating features that play a more significant role in discriminating between outcomes. cfRNAs associated with DR are marked with an asterisk. AUC, area under the curve.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 406, 293, 421]]<|/det|>
+## Extended Data Figures  
+
+<|ref|>image<|/ref|><|det|>[[150, 450, 860, 808]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[115, 840, 882, 906]]<|/det|>
+Extended Data Fig. 1. Study Design and clinical/epidemiological variables matched for the selection of patients and controls. (a) Workflow from sample collection to patients and controls stratification into discovery and validation sets. (b) Box and bar plots showing that EOPE patients (n=42) and controls
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[60, 82, 883, 270]]<|/det|>
+(n=75) groups are matched for gestational age at sample collection (T1: 9- 14 gestational weeks; T2: 18- 28 gestational weeks; T3: at the time of preeclampsia diagnosis or >28 gestational weeks), maternal age and primiparity. (c) Box and bar plots showing that LOPE patients (n=43) and controls (n=75) are matched for gestational age at sample collection, maternal age, and primiparity. In box plots, the horizontal line represents the median, with the lower and upper edges marking the 25th and 75th percentiles, respectively. Whiskers extend to 1.5 times the interquartile range. Color represents the EOPE, LOPE or control group. P-values were determined using the two- tailed Wilcoxon rank- sum test, with Bonferroni adjustment for p- values. \*\*\*\*P < 0.0001.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[150, 95, 825, 740]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[60, 760, 884, 897]]<|/det|>
+Extended Data Fig. 2. Sequencing quality parameters applied to identify outlier samples. a, RNA degradation. b, rRNA fraction. c, DNA contamination. The dashed line represents the quality threshold; samples exceeding this threshold were identified as outliers. In box plots, the horizontal line represents the median and the lower and upper edges mark the 25th and 75th percentiles, respectively. Whiskers extend to 1.5 times the interquartile range. Color represents the group. d, e. Principal component analysis at the three different time points (T1: 9-14 gestational weeks; T2: 18-28 gestational weeks; T3:
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[60, 83, 883, 172]]<|/det|>
+at the time of diagnosis in disease or \(>28\) gestational weeks) of sample collection for EOPE compared to controls (d) and LOPE compared to controls (e). Color represents the group. Arrow points out the samples deviating more than 3 standard deviations from the mean of the first and second components, which were considered outliers.  
+
+<|ref|>image<|/ref|><|det|>[[166, 192, 848, 530]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[60, 545, 884, 733]]<|/det|>
+Extended Data Fig. 3. Differentially abundant cfRNAs in early-onset and late-onset preeclampsia at diagnosis and enriched biological processes. a,b, Volcano plots showing differentially abundant cfRNAs in EOPE \((n = 20)\) (a) and LOPE \((n = 34)\) (b) compared to controls \((n = 36)\) . Significant changes in cfRNA abundance is represented by color. c, Dot plot illustrating significantly enriched biological process using as input cfRNAs differentially increased in patients compared to controls. The plot shows common pathways in EOPE and LOPE as well as pathways unique to each preeclampsia subtype. Shape represents the group. The gradient color represents the ratio \((\%)\) calculated by dividing the number of transcripts altered by the total number of transcripts described in the pathway.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[123, 85, 875, 305]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[66, 350, 884, 685]]<|/det|>
+Extended Data Fig. 4. Bioinformatic workflow to develop the predictive models. Participants were divided into two groups: a discovery set (70% of samples) for model training and optimization, and a validation set (30% of samples) for independent performance assessment. In the discovery phase, cfRNA counts were normalized, and features (cfRNAs with predictive potential) were selected using a Lasso regression method. Several cfRNA sets were generated by adjusting Lasso coefficient thresholds, and each set was evaluated for model performance using the F1 score, with the set achieving the highest F1 score (80% in this example) chosen. The selected cfRNA set was then used to train and evaluate multiple models via a leave-one-out cross-validation approach, where each sample was excluded sequentially, the model trained on the remaining samples, and then tested on the excluded sample. This process ensured each sample was used for both training and testing, yielding robust performance metrics. The model with optimal performance was selected and applied to the validation set. In the validation phase, the cfRNA matrix was filtered to include only model-selected features, which served as input for the predictive model. Based on correct and incorrect classifications, final performance metrics for the model were calculated.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 736, 177, 750]]<|/det|>
+## TABLES  
+
+<|ref|>text<|/ref|><|det|>[[115, 767, 864, 785]]<|/det|>
+Table 1. Maternal characteristics and pregnancy outcomes for the selected subset of participants.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 834, 313, 850]]<|/det|>
+## SUPPLEMENTARY TABLES  
+
+<|ref|>text<|/ref|><|det|>[[66, 867, 880, 909]]<|/det|>
+Supplementary Table 1. Gestational age at blood sample for patients and controls in the selected subset of participants.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[60, 83, 881, 120]]<|/det|>
+701 Supplementary Table 2. Differentially abundant cfRNA transcripts at diagnosis in early- onset preeclampsia patients compared to normotensive controls. 
+
+<|ref|>text<|/ref|><|det|>[[60, 140, 881, 180]]<|/det|>
+703 Supplementary Table 3. Differentially abundant cfRNA transcripts at diagnosis in late-onset preeclampsia patients compared to normotensive controls. 
+
+<|ref|>text<|/ref|><|det|>[[60, 199, 881, 238]]<|/det|>
+705 Supplementary Table 4. Gene ontology analysis for increased abundant cfRNA in early-onset preeclampsia and late-onset preeclampsia. 
+
+<|ref|>text<|/ref|><|det|>[[60, 257, 881, 296]]<|/det|>
+707 Supplementary Table 5. cfRNAs composing the predictive models for early-onset preeclampsia and late-onset preeclampsia at the first and second trimester of pregnancy. 
+
+<|ref|>text<|/ref|><|det|>[[60, 315, 881, 354]]<|/det|>
+709 Supplementary Table 6. Summary of predictive model performance metrics for early-onset preeclampsia and late-onset preeclampsia during the first and second trimesters of pregnancy.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[42, 42, 312, 70]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[42, 93, 768, 113]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[59, 131, 235, 310]]<|/det|>
+- Table1.xlsx- Supp.Table1.xlsx- Supp.Table2.xlsx- Supp.Table3.xlsx- Supp.Table4.xlsx- Supp.Table5.xlsx- Supp.Table6.xlsx
+
+<--- Page Split --->

preprint/preprint__0eccaef55dcb59ca1b81c7f04d09c150ba54c7571b656e22e7118c670e4141a7/preprint__0eccaef55dcb59ca1b81c7f04d09c150ba54c7571b656e22e7118c670e4141a7.mmd

@@ -0,0 +1,405 @@
+
+# Interrogating Site Dependent Kinetics over SiO2- Supported Pt Nanoparticles  
+
+Christian Reece  
+
+christianrecee@fas.harvard.edu  
+
+Harvard University https://orcid.org/0000- 0002- 3626- 7546  Taek-Seung Kim  Harvard University https://orcid.org/0000- 0001- 8137- 0326  Christopher O'Connor  Harvard University https://orcid.org/0000- 0002- 9224- 9342  
+
+## Article  
+
+# Keywords:  
+
+Posted Date: September 6th, 2023  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 3235489/v1  
+
+License: © © This work is licensed under a Creative Commons Attribution 4.0 International License.  Read Full License  
+
+Additional Declarations: There is NO Competing Interest.  
+
+Version of Record: A version of this preprint was published at Nature Communications on March 7th, 2024. See the published version at https://doi.org/10.1038/s41467- 024- 46496- 1.
+
+<--- Page Split --->
+
+# Interrogating Site Dependent Kinetics over \(\mathrm{SiO_2}\) -Supported Pt Nanoparticles  
+
+Taek- Seung Kim, Christopher R. O'Connor and Christian Reece\*  
+
+Rowland Institute at Harvard, Harvard University, Cambridge, MA 02142  \*Corresponding author: christianreece@fas.harvard.edu  
+
+## Abstract  
+
+A detailed knowledge of reaction kinetics is key to the development of new more efficient heterogeneous catalytic processes. However, the ability to resolve site dependent kinetics has been largely limited to surface science experiments on model systems. Herein, we can bypass the pressure, materials, and temperature gaps, resolving and quantifying two distinct pathways for CO oxidation over \(\mathrm{SiO_2}\) - supported \(2\mathrm{nmPt}\) nanoparticles under operando conditions. We find that the pathway distribution directly correlates with the distribution of well- coordinated (e.g., terrace) and under- coordinated (e.g., edge, vertex) CO adsorption sites on the \(2\mathrm{nmPt}\) nanoparticles as measured by in situ DRIFTS. We conclude that well- coordinated sites follow classic Langmuir- Hinshelwood kinetics, but under- coordinated sites follow non- standard kinetics with CO oxidation being barrierless but conversely also slow. This fundamental method of kinetic site deconvolution is broadly applicable to other catalytic systems, affording bridging of the complexity gap in heterogeneous catalysis.
+
+<--- Page Split --->
+
+Heterogeneous catalytic processes are the foundation of the chemical industry. However, our ability to rationalise and predict the behaviour of these complex industrial processes has been largely limited due to the significant complexity gap<sup>1</sup> that exists between our fundamental understanding of catalytic systems and their application. Surface science experiments over planar model catalysts have been able to precisely resolve intrinsic catalytic kinetics and dynamic catalytic behaviour<sup>2- 6</sup>; however, their application to “real-world” catalytic systems is often limited due to perceived pressure, material and temperature gaps<sup>7</sup>. These gaps are even more apparent in small (≤ 5 nm) supported nanoparticle systems where the metals no longer retain their bulk-like properties and notable support- metal interactions can exist<sup>8,9</sup>. Therefore, a method of directly measuring intrinsic kinetics over complex multi- faceted supported nanoparticle catalysts is highly desirable.  
+
+As the rate of a reaction is defined by the specific geometric and electronic characteristics of an “active site”, we propose that measurement of an intrinsic rate constant (directly related to the free energy of the reaction) would be specific to a given active site. Therefore, a precisely resolved kinetic coefficient (or set of kinetic coefficients) can act as a parameterised representation of a given active site. CO oxidation has been utilised as a fundamental model reaction for understanding surface catalytic processes for decades<sup>4,10- 13</sup>. In particular, CO oxidation over Pt catalysts has attracted significant attention due to its complex and dynamic surface chemistry, with the reaction generally understood to proceed via a Langmuir- Hinshelwood mechanism<sup>14,15</sup>. More recently, CO oxidation was used as a probe to study atom utilisation from nanoparticles to the single- site limit<sup>13,16</sup>. Due to the variations in physical structure with changing nanoparticle size<sup>17</sup> and reaction environment<sup>18,19</sup> understanding the catalytic role of geometric surface sites (i.e., structural- sensitivity) has been<sup>20,21</sup>, and still is<sup>22</sup> of great interest. For example, the activity of Pt catalysts has been considered size- dependent for C–H bond activation in the methane reforming reaction<sup>23,24</sup>, but conversely their activity for CO oxidation has typically been regarded as size- independent<sup>15,25</sup>. However, the possibility of site- dependence for CO oxidation on non- reducible oxide- supported Pt catalysts was recently reported<sup>26,27</sup> where through a combination of operando DRIFTS and steady- state kinetic measurements it was thought that different reaction kinetics for under- coordinated (UC) and well- coordinated (WC) sites exists on the metallic Pt surface. However, as the study was performed under steady state conditions their analysis was limited to simple reaction orders and apparent activation energies.  
+
+Non- steady state techniques such as Temporal Analysis of Products (TAP, see section: The Temporal Analysis of Products Experiment) are an effective way of measuring intrinsic kinetics as they can provide information about each sequential elementary step for the overall reaction. The TAP experiment serves to bridge the perceived pressure, material, and temperature gaps with the peak pressure during a pulse over the catalyst being on the order of 1 mbar<sup>28</sup> (similar to other operando methods) and by using a packed bed microreactor (allowing powdered samples) which can be heated to the reaction temperature. As the pulse of reactant gas contains significantly fewer molecules than the number of reactive sites on the catalyst, a single pulse is not considered to change the surface significantly. However, by repeatedly pulsing it becomes
+
+<--- Page Split --->
+
+possible to dynamically evolve the catalyst state using a technique known as chemical calculus<sup>29</sup> making titration- like experiments extremely powerful<sup>30- 34</sup>.  
+
+In this work we demonstrate that with the combination of TAP experiments, kinetic modelling, and DRIFTS measurements it is possible to precisely resolve site- dependent kinetics on working catalysts under operando conditions. We provide a detailed insight into the site- dependent intrinsic kinetics for CO oxidation over a well- defined 2 nm- sized Pt/SiO<sub>2</sub> catalyst. Using isotopic labelling, we deconvolute the production of CO<sub>2</sub> that arises from the reaction with preadsorbed <sup>13</sup>CO<sup>*</sup> and from the adsorption/reaction of reactant <sup>12</sup>CO in the gas phase under CO oxidation conditions. We identify two distinct kinetic features in the adsorbed <sup>13</sup>CO<sup>*</sup>- driven CO<sub>2</sub> production, which when combined with isothermal titration experiments are identified as two distinct pathways for CO oxidation. Regression of a kinetic model to the TAP exit flux curves was used to quantify the distribution of each pathway and calculate the intrinsic kinetics for the surface reaction of adsorbed oxygen with preadsorbed CO<sup>*</sup> as a function of temperature and coverage. Finally, by combining this data with comparable DRIFTS measurements, we are able to deduce site- specific kinetics as a direct relationship between the distribution of kinetic pathways and the distribution of well- coordinated (e.g., terrace like) and under- coordinated (e.g., edge, vertex) sites.  
+
+## Results  
+
+Initial characterisation of pristine and CO<sup>*</sup>- covered 2 nm- sized Pt/SiO<sub>2</sub> catalysts. Uniform and well- dispersed Pt nanoparticles were synthesized via a conventional polyol method (Supplementary Fig. 1) and were deposited on a SiO<sub>2</sub> support with a concentration of 0.72 wt%. The average size and distribution of the nanoparticles was measured to be \(1.82 \pm 0.51\) nm (Fig. 1a) on the fresh catalyst, and \(1.94 \pm 0.37\) nm (Fig. 1b) over the spent catalyst. As no sintering of the nanoparticles occurred throughout the duration of the TAP experiments, we can directly correlate the catalytic behaviour to this narrow distribution of particles approximately 2 nm in diameter. The catalytic activity of the 2 nm Pt/SiO<sub>2</sub> catalyst was tested under steady- state CO oxidation conditions (2.5% CO, 5% O<sub>2</sub>) with an apparent activation energy of \(89 \pm 3\) kJ/mol measured between 130 and 160 °C, and reaction orders of \(1.02 \pm 0.06\) in CO and \(1.22 \pm 0.1\) in O<sub>2</sub> (Supplementary Fig. 2). These values match well with previous results over Pt catalysts with similar average particle sizes<sup>15,26,35- 37</sup>, demonstrating that the Pt/SiO<sub>2</sub> catalyst used in this work is comparable with ones reported previously, albeit with a narrower size distribution. Finally, we see excellent reproducibility for all TAP experiments performed in this work (Supplementary Fig. 3).
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Fig. 1. Characterisation of Pt/SiO₂ catalyst and TAP CO oxidation experiments. a,b, Representative TEM images with particle size distribution (inset) of (a) fresh and (b) spent Pt/SiO₂ catalyst after all of the TAP experiments. c, Argon normalised exit flux curves of m/z = 28 (CO), 32 (O₂), and 44 (CO₂) at (i) 25 and (ii) 350 °C for a pulse set of 6.6% CO 13.4% O₂ gas mixture in an inert Ar tracer over CO*-covered Pt/SiO₂ catalyst. d, Temperature-dependent integrated exit flux of m/z = 44 (CO₂) normalised via Ar on (i) pristine and (ii) CO-covered Pt/SiO₂ catalyst for CO oxidation while heating from 25–350–25 °C at a heating rate of 8 °C/min. </center>  
+
+The fresh 2 nm Pt/SiO₂ catalyst was loaded into a home- built TAP reactor³⁸ and the catalytic activity for CO oxidation was measured by pulsing an oxygen rich CO/O₂ mixture (1:2 molar ratio, 6.6% CO 13.4% O₂) over a pristine and CO*- covered catalyst while heating from 25 to 350 °C at a heating rate of 8 °C/min. At 25 °C no conversion of CO or O₂ and no production of CO₂ was measured (Fig. 1c-(i)), whereas at 350 °C near 100% conversion of the CO to CO₂ is recorded (Fig. 1c-(ii)). To simplify the comparison between the experiments, the exit flux curves for every pulse set in the experiment were integrated and normalised to the inert Ar tracer (Fig. 1d). The pristine Pt/SiO₂ catalyst (Fig. 1d-(i)) shows a gradual increase in activity from 0% CO conversion at room temperature to almost 100% CO conversion above 100 °C. Similar reactivity for CO₂ production between the heating and cooling steps was found indicating the catalyst state remains consistent during the experiment. However, the CO*- covered catalyst (Fig. 1d-(ii)) shows increased production of CO₂ compared with the pristine catalyst. The excess O₂ in the reactant gas mixture is able to react with the preadsorbed CO* and act as a titrant, sequentially removing the preadsorbed reactive sites with increasing temperatures. By 350 °C it is assumed
+
+<--- Page Split --->
+
+that almost all of the preadsorbed \(\mathrm{CO}^*\) has either been reacted or desorbed off the catalyst surface as the activity during cooling was the same as that measured over the pristine catalyst. Interestingly, during the heating ramp the \(\mathrm{CO}^*\) - covered \(2\mathrm{nm}\mathrm{Pt / SiO}_2\) catalyst showed two catalytic features, with peaks in \(\mathrm{CO}_2\) production around 100 and \(200^{\circ}\mathrm{C}\) . However, further experiments were required to precisely understand the cause for these two kinetic features.  
+
+Identifying pathways for oxidation of preadsorbed \(\mathrm{CO}^*\) . In an attempt to further rationalise the two kinetic features of observed in Fig. 1d-(ii), isotopically labelled \(^{13}\mathrm{CO}\) was used to prepare a \(^{13}\mathrm{CO}^*\) - covered catalyst. This allows a precise deconvolution of the \(\mathrm{CO}_2\) produced from the reaction of the \(6.6\%\) \(\mathrm{CO}13.4\%\) \(\mathrm{O}_2\) gas mixture over the catalyst \((\mathrm{m} / \mathrm{z} = 44)\) and the \(\mathrm{CO}_2\) produced from the reaction of \(\mathrm{O}_2\) with preadsorbed \(^{13}\mathrm{CO}^*\) \((\mathrm{m} / \mathrm{z} = 45)\) . The heating rate experiment was repeated on the \(^{13}\mathrm{CO}\) - covered \(\mathrm{Pt / SiO}_2\) catalyst under the exact same conditions as used in Fig. 1d-(ii). Interestingly, the reaction of gas phase \(\mathrm{CO}\) and \(\mathrm{O}_2\) with the catalyst (Fig. 2, red triangles) was similar to that measured over the pristine catalyst whereas the preadsorbed \(^{13}\mathrm{CO}^*\) - driven \(\mathrm{CO}_2\) production (Fig. 2, orange circles) shows two well- defined kinetic features around 100 and \(200^{\circ}\mathrm{C}\) . The total \(\mathrm{CO}_2\) production \((^{12}\mathrm{CO}_2 + ^{13}\mathrm{CO}_2\) , Fig. 2, black circles) also matched well with the \(\mathrm{CO}_2\) measured in Fig. 1d-(ii). As in other temperature programmed techniques such as Temporal Programmed Oxidation (TPO), the two peaks in the measured signal would indicate two different kinetic pathways (often prescribed to different sites) for the oxidation of preadsorbed \(\mathrm{CO}\) with gas phase \(\mathrm{O}_2\) .  
+
+![](images/Figure_2.jpg)
+
+<center>Fig. 2. Isotopic labelling TAP experiment. Temperature-dependent integrated exit flux of \(\mathrm{m} / \mathrm{z} = 44\) \((\mathrm{CO}_2)\) and \(\mathrm{m} / \mathrm{z} = 45\) \((^{13}\mathrm{CO}_2)\) normalised via Ar from the TAP experiment where an oxygen rich \(\mathrm{CO} / \mathrm{O}_2\) mixture (1:2 molar ratio, \(6.6\%\) \(\mathrm{CO}13.4\%\) \(\mathrm{O}_2\) ) was pulsed over a \(^{13}\mathrm{CO}^*\) -covered \(\mathrm{Pt / SiO}_2\) catalyst while heating from \(25–350^{\circ}\mathrm{C}\) at a heating rate of \(8^{\circ}\mathrm{C} / \mathrm{min}\) . </center>
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Fig. 3. Isothermal \(\mathbf{O}_2\) titration experiments at 100 and \(200^{\circ}\mathrm{C}\) over the \(\mathbf{CO}^{*}\) -covered Pt/SiO2 catalyst. a,b, Integrated Ar normalised exit flux of \(\mathrm{O_2}\) and \(\mathrm{CO_2}\) during the \(\mathrm{O_2}\) titration experiments over the \(\mathrm{CO}^{*}\) -covered \(\mathrm{Pt / SiO_2}\) catalyst (left), and the apparent rate constant \((k^{\prime}_{app})\) as a function of cumulative \(\mathrm{CO_2}\) produced (right) at (a) 100 and (b) \(200^{\circ}\mathrm{C}\) . Yellow lines are provided to guide the eye to the two kinetic regimes. </center>  
+
+To directly probe the kinetic features at 100 and \(200^{\circ}\mathrm{C}\) , isothermal \(\mathrm{O_2}\) titration experiments were carried out. First, CO was pulsed over the catalyst at the reaction temperature until it was saturated with adsorbed \(\mathrm{CO}^*\) . Then, the preadsorbed \(\mathrm{CO}^*\) was titrated off sequentially with a series of \(\mathrm{O_2}\) pulses. This affords calculation of the apparent rate constant for the oxidation of \(\mathrm{CO}^*\) as a function of CO coverage \(^{31,34}\) . In the first few pulses \(\sim 100\%\) conversion of \(\mathrm{O_2}\) is observed at both temperatures (Fig. 3- left). Then, as \(\mathrm{O_2}\) is repeatedly pulsed, the \(\mathrm{CO_2}\) production decreases, and the exit flux of \(\mathrm{O_2}\) increases up to the saturation state on the Pt surface, confirming that complete removal of the reactive \(\mathrm{CO}^*\) occurs. To estimate the kinetics of the preadsorbed \(\mathrm{CO}^*\) consumption, we plot the temperature corrected apparent rate constant as a function of cumulative \(\mathrm{CO_2}\) produced (indicative of \(\mathrm{CO}^*\) coverage) in Fig. 3a,b- right, which is calculated using \(^{31}\) :  
+
+\[k_{app}^{'} = \frac{X}{1 - X}\sqrt{T}\approx k_{app}\approx k_{a}\theta_{CO^{*}} \quad (1)\]  
+
+Where \(k_{app}^{\prime}\) is the temperature corrected apparent rate constant, \(X\) is the fractional \(\mathrm{O_2}\) conversion, \(T\) is the temperature (K), \(k_{app}\) is the apparent rate constant (s \(^{- 1}\) ), \(k_{a}\) is the intrinsic rate constant for the reaction of \(\mathrm{O_2}\) with preadsorbed \(\mathrm{CO}^*\) , and \(\theta_{CO^*}\) is the coverage of
+
+<--- Page Split --->
+
+preadsorbed \(\mathrm{CO^{*}}\) . On account of the usage of the same catalyst bed throughout all the pulse experiments, variations in \(k_{app}\) show a direct correlation to the intrinsic rate constant for the \(\mathrm{O_2}\) titration experiment \(^{39,40}\) . Specifically, a linearly decrease in \(k_{app}\) with \(\theta_{CO^{*}}\) (or \(\mathrm{CO_2}\) produced) indicates the presence of a unique intrinsic rate constants \(\left(\mathrm{k_a}\right)\) for the reaction of \(\mathrm{O_2}\) with \(\mathrm{CO^*}\) through the following relationship \(^{31,41}\) :  
+
+\[\frac{\Delta k_{app}}{\Delta\theta_{CO^{*}}} = \frac{(1 - \epsilon)}{\epsilon V} k_a \quad (2)\]  
+
+Where \(\epsilon\) is the void fraction of the packed bed, and V is volume of the catalyst zone. We identify two linear regimes for the reaction of \(\mathrm{O_2}\) with preadsorbed \(\mathrm{CO^*}\) in both isothermal titration experiments (Fig. 3, yellow line), indicating that two intrinsic rate constants, and as such two pathways exist simultaneously at both temperatures. We identify a slow and fast intrinsic reaction rate for \(\mathrm{CO_2}\) production, as shown by the gradients in Fig. 3- right. At high relative coverage of \(\mathrm{CO^*}\) the "fast" pathway dominates, at lower \(\mathrm{CO^*}\) coverages the "slow" pathway dominates during the titration experiment. Coupling this insight with the two kinetic regimes seen in the temperature programmed experiments, we feel confident in claiming that at least two different pathways for the oxidation of \(\mathrm{CO^*}\) by \(\mathrm{O_2}\) exist on the \(2\mathrm{nm}\mathrm{Pt / SiO_2}\) catalyst.  
+
+Quantifying active species using MZTRT and DRIFTS. Due to its well- defined nature, it is possible to precisely resolve the intrinsic kinetics of catalytic processes using the TAP experiment \(^{40,42}\) . For linear (first order or pseudo first order) reactions Multi- Zone TAP Reactor Theory \(^{43,44}\) (MZTRT) is a powerful and efficient tool for simulating TAP exit flux responses. The model of the experiment was built using the generalised form of MZTRT (see Supplementary I) with the Symmetric Thin- Zone assumption applied to the catalyst zone \(^{40}\) . For the reaction of oxygen with preadsorbed \(\mathrm{CO^*}\) , a three- pathway model was identified as the most likely candidate (see Supplementary II; Fig. 4a and Supplementary Fig. 4). The experimental results for two catalytic features ( \(\mathrm{CO_2}\) production around 100 and \(200^{\circ}\mathrm{C}\) in Fig. 2) and two kinetic regimes (two pathways for the reaction of \(\mathrm{O_2}\) in Fig. 3) would indicate two separate pathways for oxidation of \(\mathrm{CO^*}\) . As the oxygen balance (i.e., oxygen released as \(\mathrm{CO_2}\) / oxygen consumed) is not always 1 throughout the experiment a third pathway of irreversible oxygen adsorption is necessary in the model. It is very important to note that the model regression is performed on each set of exit flux curves individually (Fig. 4b) with the model fit to both the shape and magnitude of Ar, \(\mathrm{O_2}\) , and \(\mathrm{CO_2}\) exit flux curves. The kinetic model can be broken down into two parts. First, the apparent adsorption rate constants \(k_{a,1}'\) and \(k_{a,2}'\) represent the irreversible adsorption and subsequent reaction of gaseous \(\mathrm{O_2}\) with preabsorbed \(\mathrm{CO^*}\) , and the adsorption rate constant \(k_{a,3}'\) represents the irreversible adsorption of \(\mathrm{O_2}\) to sites where no further reaction with \(\mathrm{CO^*}\) takes place. Second, the intrinsic rate constants \(k_{r,1}\) and \(k_{r,2}\) represent the rate at which the surface reaction between adsorbed oxygen and \(\mathrm{CO^*}\) occurs for pathways 1 and 2, respectively. In short, the adsorption constants \(k_{a}'\) control the magnitude of the \(\mathrm{O_2}\) and \(\mathrm{CO_2}\) exit flux curves, whereas the surface reaction constants \(k_{r}\) control the shape. As each pulse set is regressed individually during the isothermal \(\mathrm{O_2}\) titration experiment (Fig. 4c), a set of rate constants for the reaction of oxygen with preabsorbed \(\mathrm{CO^*}\) can be calculated at each point (Fig. 4d). Models of varying complexity were tested, but only the three- pathway model was able to precisely recreate the experimental data without being overparameterised (see Supplementary II;
+
+<--- Page Split --->
+
+Supplementary Fig. 5). The model fitting exhibits a high degree of confidence and consistency with all TAP experiments in this work, although decreased confidence in the intrinsic surface reaction constants is observed at the limit of very low \(\mathrm{CO_2}\) production.  
+
+![](images/Figure_4.jpg)
+
+<center>Fig. 4. Model fitting of isothermal \(\mathrm{O_2}\) titration experiment using MZTRT. a, Schematic of the three-pathway model. b, Experimentally measured and model exit flux curves for pulse set 40 of \(\mathrm{O_2 / Ar}\) during \(\mathrm{O_2}\) titration experiment at \(100^{\circ}\mathrm{C}\) . c, Experimentally measured and model fitted integrated exit flux curves for the whole \(\mathrm{O_2}\) titration experiment at \(100^{\circ}\mathrm{C}\) (corresponding to Fig. 3a). d, Rate constants calculated from model fitting with \(95\%\) confidence intervals included. Inset shows small but non-zero value for \(k_{r,1}\) . </center>
+
+<--- Page Split --->
+![](images/Figure_5.jpg)
+
+<center>Fig. 5. Characterisation of \(\mathbf{CO}^*\) covered catalyst via DRIFTS, TAP, and kinetic modelling. a, DRIFT spectra for CO adsorption on the \(\mathbf{CO}^*\) -covered \(2\mathrm{nm}\mathrm{Pt} / \mathrm{SiO}_2\) catalyst where \(\mathbf{CO}^*\) was preadsorbed at 35, 100, 200, and \(350^{\circ}\mathrm{C}\) . The spectra were obtained after cooling the catalyst to \(35^{\circ}\mathrm{C}\) . All DRIFTS measurements were acquired at \(35^{\circ}\mathrm{C}\) which avoids the influence of temperature on vibrational features \(^{45}\) . b, Illustration for adsorbed \(\mathbf{CO}^*\) sites (well-coordinated and under-coordinated) on the surface of a model Pt nanoparticle (regular, truncated octahedron; 586 atoms; \(\sim 2.17 \mathrm{nm}\) -size). c, Integrated Ar normalised exit flux of \(\mathrm{m} / \mathrm{z} = 44\) ( \(\mathbf{CO}_2\) ) during TPO experiments on the \(\mathbf{CO}^*\) -covered \(\mathrm{Pt} / \mathrm{SiO}_2\) catalyst where \(\mathbf{CO}^*\) was preadsorbed at 25, 100, 200, and \(350^{\circ}\mathrm{C}\) . For TPO experiments, the \(\mathbf{CO}^*\) was preadsorbed at \(25 - 350^{\circ}\mathrm{C}\) on the catalyst and the catalyst was cooled to \(25^{\circ}\mathrm{C}\) . Then \(\mathrm{O}_2\) was repeatedly pulsed over the catalyst while being linearly heated to \(350^{\circ}\mathrm{C}\) at \(8^{\circ}\mathrm{C} / \mathrm{min}\) . d–g, Deconvoluted \(\mathbf{CO}_2\) production pathways and \(\mathbf{CO}_2\) production pathway ratios calculated using the regressed kinetic model (MZTRT) for the TPO experiments where \(\mathbf{CO}^*\) was preadsorbed at (d) 25, (e) 100, (f) 200, (g) \(350^{\circ}\mathrm{C}\) . </center>  
+
+To rationalise if the two pathways for oxidation of the preadsorbed \(\mathbf{CO}^*\) were correlated to the geometric structure of metallic Pt sites (e.g., terrace, edge, vertex sites) on the \(2\mathrm{nm}\) nanoparticles, a series of DRIFTS and TPO experiments were performed where CO was preadsorbed at \(25 - 350^{\circ}\mathrm{C}\) (Fig. 5). In general, it is known that the binding energy for adsorbed CO on the under- coordinated (UC) sites (e.g., edge, vertex sites) is higher than that on the well- coordinated (WC) sites (e.g., terrace sites). This implies that when CO is adsorbed at higher temperatures, an increase in populated UC sites relative to the populated WC sites would be expected \(^{46}\) .
+
+<--- Page Split --->
+
+From the DRIFTS investigation in Fig. 5a, it is shown that the pre- adsorption of CO at 35, 100, 200, and \(350^{\circ}\mathrm{C}\) (35CO, 100CO, 200CO, 350CO, respectively) populates linear bound CO to WC sites, UC sites, and bridge bound CO as evidenced by three \(\mathrm{v(C - O)}\) bands (Fig. 5a,b features 1- 3). The 35CO yields \(\mathrm{v(C - O)}\) bands centred at \(2075\mathrm{cm}^{- 1}\) with a small high frequency shoulder (feature 1), a small band near \(2042\mathrm{cm}^{- 1}\) (feature 2), and a broad band at \(1805\mathrm{cm}^{- 1}\) (feature 3). Increasing the adsorption temperature to \(100^{\circ}\mathrm{C}\) causes a slight decrease in the intensity of feature 1 and a shift in features 2 and 3 to lower frequency. For the 200CO and 350CO, there is a dramatic decrease in the intensity and slight shift to lower frequency of feature 1. The decrease in the intensity is attributed to the reduction of the total number of adsorbed \(\mathrm{CO}^{*}\) with increasing temperature. In turn the surface is predominantly relatively strongly bound CO on UC sites. A larger shift to lower frequency of features 2 and 3 is observed, which clarifies feature 2 as a distinct peak from feature 1. The position of the \(\mathrm{v(C - O)}\) frequencies as a function of CO adsorption temperature are reported in Supplementary Table 1. Feature 1 with a peak maximum from \(2075\mathrm{cm}^{- 1}\) is assigned to the collective oscillation of linear bound CO to WC sites (linear WC) where the high frequency shoulder is attributed to a dense CO phase for high CO coverages. Feature 2 with a peak maximum from \(2042\mathrm{cm}^{- 1}\) is assigned to the collective oscillation of linear CO to UC sites (linear UC). The frequency of the \(\mathrm{v(C - O)}\) for linear CO on the \(2\mathrm{nm}\mathrm{Pt}\) nanoparticles catalyst is in agreement with previous investigations when considering the frequency is dependent on the CO coverage \(^{26,47 - 57}\) , Pt coordination environment \(^{26,35,51 - 63}\) , and nanoparticle size \(^{26,57 - 61}\) . Feature 3 with a peak maximum from \(1805\mathrm{cm}^{- 1}\) is assigned to the collective oscillation of bridge bound CO to Pt sites and is also in reasonable agreement with previous work \(^{49 - 54,64 - 67}\) . In contrast to the linear CO vibrational features, the identification of unique bridge CO features for different Pt coordination environments was not possible. The \(\mathrm{CO}^{*}\) adsorption site information from the DRIFTS analysis is graphically summarised in Fig. 5b.  
+
+Alongside the DRIFTS investigation, comparable TPO experiments were performed where \(\mathrm{CO}^{*}\) was preadsorbed at temperatures from \(25 - 350^{\circ}\mathrm{C}\) and was titrated using \(\mathrm{O_2}\) pulses while heating from \(25 - 350^{\circ}\mathrm{C}\) at \(8^{\circ}\mathrm{C / min}\) (Fig. 5c). Interestingly, the 25CO TPO shows an onset temperature for \(\mathrm{CO_2}\) production at around \(40^{\circ}\mathrm{C}\) , whereas the TPO results of both 100CO and 200CO show 10- 20 times higher \(\mathrm{CO_2}\) production at \(40^{\circ}\mathrm{C}\) . It is believed that when CO is preadsorbed at \(25^{\circ}\mathrm{C}\) the surface is fully covered in \(\mathrm{CO}^{*}\) meaning there are no vacant sites for adsorption/dissociation of \(\mathrm{O_2}\) (i.e., CO poisoning). We prescribe the increased activity for \(\mathrm{CO_2}\) production in the 100CO and 200CO TPO experiments to the increased number of vacant sites on the partially \(\mathrm{CO}^{*}\) - covered surface. This is supported by an additional TPO experiment where CO is pulsed to saturation over the 100CO catalyst at \(25^{\circ}\mathrm{C}\) . The \(100\mathrm{CO} + 25\mathrm{CO}\) TPO experiment shows similar tendency to that of 25CO (Supplementary Fig. 6), which implies that the low temperature \(\mathrm{CO_2}\) production at \(25^{\circ}\mathrm{C}\) is vacancy driven. Further, isothermal titration experiments performed at \(25^{\circ}\mathrm{C}\) show no \(\mathrm{CO_2}\) production (Supplementary Fig. 7). When combined, this would indicate that \(\mathrm{CO_2}\) production by oxidation of preadsorbed \(\mathrm{CO}^{*}\) is driven by a Langmuir- Hinshelwood type reaction (as is expected for CO oxidation over Pt) \(^{14,15}\) . However, we cannot rule out other methods such as \(\mathrm{CO}^{*}\) assisted \(\mathrm{O_2}\) dissociation \(^{15}\) . Interestingly, the low- temperature oxidation of preadsorbed \(\mathrm{CO}^{*}\) would indicate that the Langmuir- Hinshelwood surface reaction between oxygen and \(\mathrm{CO}^{*}\) over \(2\mathrm{nm}\mathrm{Pt}\) nanoparticles has a very low activation barrier, which is counter to previous work on single crystals and supported Pt catalysts with activation energies ranging from 37 to \(85\mathrm{kJ / mol}^{15,36,68,69}\) .
+
+<--- Page Split --->
+
+Similar to the isotopically labelled experiments, two kinetic features are observed: a spike around \(80^{\circ}\mathrm{C}\) (first feature) and a long shoulder from \(100–350^{\circ}\mathrm{C}\) (second feature) appear in the 25CO TPO. The first kinetic feature appears at low temperatures \((< 50^{\circ}\mathrm{C})\) in both the 100CO and 200CO which we ascribe to the low- temperature \(\mathrm{CO}^{*}\) conversion mentioned above. Due to decreasing \(\mathrm{CO}^{*}\) coverage throughout the titration experiment, the \(\mathrm{CO}_{2}\) production would be expected to decrease. However, in the 100CO experiment a distinct second catalytic feature appears around \(125^{\circ}\mathrm{C}\) , which suggests the presence of temperature- dependent kinetics. Further, the second catalytic feature disappears as the adsorption temperature increases from 100 to \(200^{\circ}\mathrm{C}\) , which corresponds to the large decrease in WC sites from the DRIFTS results in Fig. 5a. This strongly suggests that site dependent kinetics does exist for the oxidation of \(\mathrm{CO}^{*}\) over \(\mathrm{SiO}_{2}\) - supported \(2\mathrm{nm}\mathrm{Pt}\) nanoparticles.  
+
+To calculate the intrinsic kinetics at each point in the TPO experiments, the rate constants for the model in Fig. 4a were calculated by regressing the MZTRT Symmetric Thin- Zone model to every pulse set in the experiments. All rate constants for the 25CO, 100CO, 200CO, and 350CO TPO experiments are shown in Supplementary Fig. 8. Also, the MATLAB script used to process the 25CO TPO experiment is included alongside the Supplementary Information. As during the TAP experiment the pulse size is sufficiently small that coverage of \(\mathrm{CO}^{*}\) species is not changed by any appreciable amount during a pulse, the first order irreversible adsorption rate constants \(k_{a,1}^{\prime}\) and \(k_{a,2}^{\prime}\) are pseudo first order and are proportional to the concentration of the adsorbed \(\mathrm{CO}^{*}\) species involved in each pathway \((k_{a,3}^{\prime}\) would be proportional to the number of empty sites). This means that it becomes possible to deconvolute the amount of \(\mathrm{CO}_{2}\) produced from each pathway as shown in in Fig. 5d- g. From the MZTRT Symmetric Thin- Zone model, the conversion of \(\mathrm{O}_{2}\) in each pulse set can be calculated using \(^{40,70}\) :  
+
+\[X_{O_{2}} = \frac{\left(k_{a,1}^{\prime} + k_{a,2}^{\prime} + k_{a,3}^{\prime}\right)\left(L / 2D_{e}\right)}{1 + \left(k_{a,1}^{\prime} + k_{a,2}^{\prime} + k_{a,3}^{\prime}\right)\left(L / 2D_{e}\right)} \quad (3)\]  
+
+Where \(X_{O_{2}}\) conversion of the reactant, \(D_{e}^{R}\) is the diffusivity of the reactant \((\mathrm{cm}^{2} / \mathrm{cm}^{3})\) , \(L\) is the length of the catalyst bed (cm), and \(k_{a,n}^{\prime}\) represents the apparent pseudo- first order adsorption/reaction constant for each individual site included in the model (cm \(\mathrm{s}^{- 1}\) ) and is the parameter that is calculated during the model fitting. The apparent pseudo- first order adsorption/reaction constant is linearly related to the intrinsic adsorption/reaction rate constant through the following relationship \(^{41,43}\) :  
+
+\[k_{a,n}^{\prime} = k_{a,n}\theta_{n}L_{cat}\frac{S_{v}(1 - \epsilon_{b})}{\epsilon_{b}} \quad (4)\]  
+
+Where \(k_{a,n}\) is the intrinsic adsorption/reaction constant \((\mathrm{cm}^{3} \mathrm{mol} \mathrm{s}^{- 1} \mathrm{s}^{- 1})\) and \(\theta_{n}\) is the concentration of \(\mathrm{CO}^{*}\) in the case of sites 1 and 2, and the coverage of empty irreversible adsorption sites in the case of site 3 \((\mathrm{mol} \mathrm{cm}^{- 3})\) , \(L_{cat}\) is the length of the catalyst zone \((\mathrm{cm}) S_{v}\) is the surface area of the catalyst per volume of catalyst \((\mathrm{cm}^{2} / \mathrm{cm}^{3})\) . This is a slight modification to
+
+<--- Page Split --->
+
+forms of the equation previously published<sup>40</sup>, as the MZTRT Symmetric Thin- Zone model does not explicitly include a value for \(L_{cat}\) and so it is lumped into the apparent pseudo- first order adsorption/reaction constant \(k_{a,n}^{\prime}\) . For each pathway included in the model, the conversion of oxygen specific to each pathway can be deconvoluted using the following:  
+
+\[X_{O_{2},n} = \frac{k_{a,n}^{\prime}(L / 2D_{e})}{1 + \left(k_{a,1}^{\prime} + k_{a,2}^{\prime} + k_{a,3}^{\prime}\right)(L / 2D_{e})} \quad (5)\]  
+
+The amount of \(\mathrm{CO}_{2}\) produced through can be calculated from the conversion of oxygen specific to pathways 1 and 2 by considering the reaction stoichiometry using the following:  
+
+\[M_{0,norm}^{CO_{2},n} = 2X_{O_{2},n} \quad (6)\]  
+
+From the pathway deconvolution in Fig. 5d- g, we find that pathway 1 is dominant for \(\mathrm{CO}_{2}\) production at low- temperatures, whereas at elevated temperatures pathway 2 becomes the dominant pathway. The \(\mathrm{CO}_{2}\) yield from each pathway can be easily calculated by summing the deconvoluted \(\mathrm{CO}_{2}\) production throughout the TPO experiment (Supplementary Fig. 9). We find that the \(\mathrm{CO}_{2}\) yield of pathway 1 stays relatively constant up to a CO adsorption temperature of \(200^{\circ}\mathrm{C}\) , whereas the \(\mathrm{CO}_{2}\) yield of pathway 2 rapidly decreases. Further, when \(\mathrm{CO}\) is preadsorbed at \(350^{\circ}\mathrm{C}\) , the amount of \(\mathrm{CO}\) species related to pathways 1 and 2 becomes approximately equal on the surface.  
+
+To identify if any correlation between the two pathways and geometric surface sites exists, we summarised the ratio of the \(\mathrm{CO}_{2}\) yield of pathway 1 to pathway 2 calculated from the kinetic model with the quantitative DRIFTS analysis for the population of linear UC and WC \(\mathrm{CO}^{*}\) sites in Fig. 6. The fraction of pathway 1 from the kinetic model (Fig. 6, black) increases with increasing CO adsorption temperature. From the quantitative deconvolution of the DRIFTS spectra for the population of UC and WC \(\mathrm{CO}^{*}\) sites (see Supplementary III for details on quantification), the fraction of populated linear CO at UC sites is also increasing with CO adsorption temperature. Deconvolution of the bridge CO to WC and UC sites is not possible, but there is a shift to lower frequency of the bridge peak maximum with temperature that is consistent with increasing the relative population of UC Pt sites. Most notably, we find a straightforward relationship between the amount of each pathway as calculated from the kinetic model, and the total amount of UC and WC sites calculated using DRIFTS. Due to this direct relationship, we claim that pathway 1 mainly occurs on UC sites, and pathway 2 occurs mainly at WC sites.
+
+<--- Page Split --->
+![](images/Figure_6.jpg)
+
+<center>Fig. 6. Comparison between pathway ratio (TPO experiment) and area ratio of \(\mathbf{CO}^*\) sites (DRIFTS investigation). A direct correlation between the total amount of pathway 1 and pathway 2 calculated from the kinetic model, and the total amount of UC and WC sites from the DRIFTS investigation is found. </center>  
+
+Measuring the Intrinsic Kinetics of CO oxidation on UC and WC sites. Along with the reactive site information, the intrinsic surface reaction constants \(k_{r,1}\) and \(k_{r,2}\) (related to the reaction of adsorbed oxygen with adsorbed CO for pathways 1 and 2) from the isothermal titration and TPO experiments were calculated and are shown in Fig. 7. As mentioned previously, the reaction over UC sites is ascribed to pathway 1, whereas the reaction over WC sites is ascribed to pathway 2. It can be seen that the rate constant for pathway 1 \((k_{r,1})\) is significantly lower than pathway 2 \((k_{r,2})\) under all conditions probed, with pathway 1 being a slow reaction between adsorbed oxygen and CO and pathway 2 being fast. Interestingly, pathway 1 shows no dependence on temperature or the coverage of CO in both the isothermal O2 titration (Fig. 7a,b) and the TPO (Fig. 7c,d) experiments but is unintuitively a slow reaction. One such rationalisation for this behaviour based on transition state theory would be that the reaction results in a significant loss in entropy in the transition state \(^{71}\) . As adsorbed oxygen is considered to be immobile under these conditions, it could be hypothesized that a CO molecule strongly bound to a vertex of a nanoparticle has a large number of degrees of freedom, but during the transition state to make \(\mathrm{CO_2}\) the species would become immobile due to the strongly bound oxygen, losing a large number of degrees of freedom. However, as CO adsorption on Pt is not well described by DFT \(^{72,73}\) , this would be difficult to confirm, so this idea remains conceptual. Experiments using molecular beam scattering \(^{14,68}\) or velocity resolved kinetics \(^{74}\) do not observe this barrierless reaction, but it should be noted that these were performed on Pt single crystals, whereas this work is performed on 2 nm Pt supported nanoparticles, and as such the nanoparticle size effect must also be considered. Further, due to the slow nature of the reaction, this pathway would be difficult to isolate as it would certainly be limited by CO desorption under steady-state conditions at any appreciable CO pressure.
+
+<--- Page Split --->
+
+We find that pathway 2 shows more classic kinetic behaviour. Under isothermal conditions we find a linear dependence on the coverage of \(\mathrm{CO}^*\) (Fig. 7b) indicating lateral interactions between adsorbed \(\mathrm{CO}^*\) molecules and/or adsorbed \(\mathrm{O}^*\) , which is expected based on previous work<sup>36</sup>. Under non-isothermal conditions an exponential increase in rate with increasing temperature (Fig. 7d) is observed above \(80^{\circ}\mathrm{C}\) . It should be noted that as the surface coverage is also changing during the TPO experiment, a classic Arrhenius style analysis to calculate activation energies and pre-exponential factors is non-trivial and will be attempted in future publications. Further, we observed complex kinetic behaviour in pathway 2 below \(80^{\circ}\mathrm{C}\) (which was not recreated in the isothermal \(\mathrm{O}_2\) titration experiments) which cannot be described by classical Arrhenius kinetics. It could be hypothesized that some restructuring of the Pt nanoparticles is occurring around \(100^{\circ}\mathrm{C}\) , but this will be evaluated in future work combining these methods with techniques such as X- ray Absorption Spectroscopy.  
+
+![](images/Supplementary_Figure_12.jpg)
+
+<center>Fig. 7. Calculated rate constants \((k_{r,1}\) and \(k_{r,2}\) ) for the surface reaction between adsorbed oxygen and CO from the three-pathway kinetic model with \(95\%\) confidence intervals overlaid. a,b, Isothermal titration experiments where CO was adsorbed at 100 and \(200^{\circ}\mathrm{C}\) . Increasing total \(\mathrm{CO}_2\) produced is correlated with decreasing CO coverage. c,d, TPO experiments where CO was adsorbed at temperature ranging from \(25 - 200^{\circ}\mathrm{C}\) . When the production of \(\mathrm{CO}_2\) is sufficiently low in the TPO experiment \((>200^{\circ}\mathrm{C})\) the signal/noise ratio of the \(\mathrm{CO}_2\) exit flux curves significantly decreases, which in turn decreases the confidence in the model fitting, particularly for pathway 2, as shown in Supplementary Fig. 12. </center>
+
+<--- Page Split --->
+![](images/Figure_8.jpg)
+
+<center>Fig. 8. Summary of the three-pathway model for the reaction of \(\mathbf{O}_2\) with preabsorbed \(\mathbf{CO}^*\) over \(\mathbf{SiO}_2\) -supporrted 2 nm Pt nanoparticles. </center>  
+
+The combination of TAP, kinetic modelling, and DRIFTS measurements provides unparalleled levels of kinetic insight into the dynamic site dependent activity for CO oxidation over 2 nm- sized \(\mathrm{Pt / SiO_2}\) catalysts. The precisely defined nature of the TAP experiment means that fine kinetic features are resolvable by modelling the exit flux curves using efficient analytical functions. By coupling this insight with DRIFTS measurements, we have been able to identify that site- specific kinetics exists for CO oxidation over 2 nm Pt nanoparticles with the three pathways for the interaction of oxygen with \(\mathrm{CO}^*\) summarised in Fig. 8. We find two pathways for the oxidation of \(\mathrm{CO}^*\) and one pathway for the irreversible adsorption of oxygen (no reaction). Pathway 1 mainly occurs at the UC sites and has slow kinetics, is coverage independent, and is temperature independent. On the contrary, pathway 2 mainly occurs at WC sites with fast kinetics, is highly dependent on the \(\mathrm{CO}^*\) or \(\mathrm{O}^*\) coverage and shows an exponential increase with temperature. These results serve as a significant insight into understanding the kinetics of various reactive sites in heterogeneous catalysis. Typically, it has been widely accepted that increasing the number of UC sites is advantageous for CO conversion<sup>22,27</sup>. However, even though the reaction at the UC sites is found to be barrierless, it has slow kinetics which may not be optimal for CO conversion.  
+
+Once the reaction pathways have been assigned to specific sites, this method of kinetic site deconvolution can be applied generally, allowing features such as restructuring, sintering, and even potentially metal- support interactions to be understood. The quantitative nature of the TAP experiment means simultaneous calculation of the number of active sites, their distribution on the surface, and the intrinsic kinetics of the surface reactions is now possible. We believe that our approach is general enough to apply to other catalytic systems and can serve as a new toolkit in the characterisation of heterogeneous catalysts.  
+
+## Methods
+
+<--- Page Split --->
+
+Preparation of Pt/SiO₂ catalyst. In a typical synthesis of 2 nm- sized Pt nanoparticles, \(^{75}\) 100 mg of chloroplatinic acid hydrate \((\mathrm{H}_{2}\mathrm{PtCl}_{6} \cdot \mathrm{xH}_{2}\mathrm{O}, 99.9\%)\) , Sigma- Aldrich), 20 mg of poly(vinylpyrrolidone) (PVP, \(\mathrm{M}_{\mathrm{w}} = 40000\) , Sigma- Aldrich), 2.5 mL of sodium hydroxide (NaOH, 1 N) were dissolved in 10 mL of ethylene glycol (Sigma- Aldrich) in a 50 mL three- neck roundbottom flask. The flask was heated to 80 °C and evacuated for 30 min with vigorous stirring, then heated to 200 °C. The solution was kept at 200 °C for 2 hr with Ar gas purging. After the reaction, the solution cooled down to RT. The colloidal suspension was diluted to 50 mL of acetone and centrifuged at 6500 rpm for 10 min twice, repeatedly. Then, the as- synthesized Pt nanoparticles were re- dispersed in 40 mL of ethanol. Then, 10 mL of the Pt- dispersed solution was dropped onto the 0.3 g of \(\mathrm{SiO}_{2}\) powder (pretreated at 700 °C for 1 hr; 10–20 nm, Sigma- Aldrich) with vigorous stirring. The suspension was sonicated for 20 min, subsequently evaporating the solvent in a vacuum at 50 °C overnight. The microstructure of the Pt/SiO₂ catalyst was investigated via XRD measurement (Supplementary Fig. 1b) and shows diffraction peaks at 39.5 and 46°, corresponding to the (111) and (200) planes of reference Pt (JCPDS #04- 0802). For characterisation and TAP experiments, the Pt/SiO₂ catalyst was calcined at 500 °C for 1 hr in air condition to remove majority of carbonaceous capping agent with a ramp up rate of 1 °C/min. To prepare the metallic Pt nanoparticles, the Pt/SiO₂ catalyst was reduced at 300 °C for 1 hr with a ramp up rate of 1 °C/min under 10% \(\mathrm{H}_{2}\) in Ar flow.  
+
+Characterisations. The morphology and size distribution of Pt/SiO₂ catalyst were investigated by transmission electron microscopy (TEM; ARM 200F, JEOL) at 200 kV. The concentration of catalyst was determined using inductively coupled plasma optical emission spectroscopy (ICP- OES; 5110 ICP- OES; Agilent). The catalyst microstructure was measured by X- ray diffractometer (XRD; D2 Phaser, Bruker). The steady- state flow catalytic activity was measured in a home- built ambient pressure flow reactor that has been previously described \(^{76}\) with all gas mixtures reported balanced in Ar.  
+
+Temporal Analysis of Products experiments. The TAP technique has been described extensively in the literature \(^{41,42,77}\) , but it is summarised here. During the TAP experiment a nanomole pulse of gas \((\sim 10^{15}\) molecules, 108 \(\mu\) s pulse width) is sent into a packed bed microreactor that is held at ultra- high vacuum \((\sim 10^{- 9}\) torr). During the experiment, the pulsed gas diffuses through the packed bed via Knudsen Diffusion where it can interact with the catalyst surface. Eventually the gas diffuses out the exit of the microreactor and the exit flux of is measured via mass spectrometry. Due to the precisely defined nature of Knudsen Diffusion, the shape (and magnitude) of the exit flux curves provides highly resolved kinetic insight, in particular when coupled with kinetic modelling of the exit flux curves (see section: Modelling of Temporal Analysis of Products Pulse Responses).  
+
+For the TAP experiments, the Pt/SiO₂ catalyst is first calcined at 350 °C by injecting 1000 pulse sets of large O₂ pulses (160 \(\mu\) s pulse width) until the CO₂ signal (m/z = 44) becomes near- zero to minimise the decomposition of the remaining carbonaceous capping. Before all pulse/transient response experiments, the Pt/SiO₂ catalyst is reduced at 350 °C by injecting 600 pulse sets of large H₂ pulses (160 \(\mu\) s pulse width) to achieve a metallic Pt surface. We utilise a home- built TAP reactor \(^{38}\) where the microreactor contains a layer of commercial sand (29.7 mm; 50–70 mesh SiO₂; Sigma- Aldrich) followed by a layer of Pt/SiO₂ catalyst (5.4 mg) followed by a final layer of commercial sand (34.4 mm) for a total reactor length of 64.1 mm. The exit flux of
+
+<--- Page Split --->
+
+CO, \(\mathrm{O_2}\) , Ar, and \(\mathrm{CO_2}\) is monitored via mass spectrometry. The integrated exit flux of the Ar tracer is used for normalisation of all pulse experiments (see Supplementary IV). As the mass spectrometer can only investigate one \(\mathrm{m / z}\) value per pulse, multiple pulses are used to scan the whole range of \(\mathrm{m / z}\) values and are combined to one pulse set<sup>38</sup>. All gas mixtures reported are balanced in Ar.  
+
+Diffused Reflectance Infrared Fourier Transform Spectroscopy experiments. DRIFTS experiments were carried out in a low- temperature reaction chamber (Harrick Scientific) equipped with ZnSe windows, mounted inside the sample compartment of a Bruker Invenio FT- IR spectrometer using a Praying Mantis diffuse reflectance accessory (Harrick Scientific). The catalyst sample was prepared by pressing approximately \(2\mathrm{mg}\) of \(2\mathrm{nm}\) Pt/SiO<sub>2</sub> onto a 304 stainless- steel mesh ( \(150 \times 150\) mesh). The DRIFTS reactor was loaded by placing the catalyst- containing mesh on top of approximately \(110\mathrm{mg}\) of \(120\mathrm{gr}\) SiC, an inert support with high thermal conductivity. There can exist large temperature gradients between the thermocouple contact in a DRIFTS reactor cell and the catalyst surface temperature exposed to the infrared beam.<sup>78</sup> Therefore, a thermocouple was mounted in physical contact with the bottom of the stainless- steel mesh and the temperature gradient to the catalyst surface at \(350^{\circ}\mathrm{C}\) was less than \(20^{\circ}\mathrm{C}\) as calibrated by an optical pyrometer. All DRIFTS experiments used a total volumetric flow rate of \(100\mathrm{scm}\) . Each absorbance spectrum was obtained by averaging 200 background and sample scans at a resolution of \(4\mathrm{cm}^{- 1}\) using a liquid- nitrogen- cooled \(\mathrm{HgCdTe}\) (MCT) detector, while the Praying Mantis diffuse reflectance accessory and FT- IR spectrometer was purged with dry \(\mathrm{N_2}\) . The background measurement was acquired after the catalyst sample was annealed at \(350^{\circ}\mathrm{C}\) for \(30\mathrm{min}\) in \(5\%\) \(\mathrm{H_2}\) in Ar and cooled to \(35^{\circ}\mathrm{C}\) in Ar. The sample measurements were acquired after the catalyst sample was annealed at \(30^{\circ}\mathrm{C / min}\) in Ar to the CO adsorption temperature, the temperature was maintained for \(10\mathrm{min}\) in \(0.1\%\) CO in Ar until saturation was achieved, and the sample was cooled to \(35^{\circ}\mathrm{C}\) in Ar. A description of the quantitative analysis of the DRIFTS spectra is provided in the Supplementary (see Supplementary III).  
+
+Modelling of Temporal Analysis of Products Pulse Responses. To simulate the TAP exit flux response curves, Multi- Zone TAP Reactor Theory<sup>43,44</sup> (MZTRT) was utilised with the catalyst zone being approximated as a Thin Zone<sup>40</sup> in the centre of the microreactor between two layers of inert packing. To perform the curve fitting first, the experimentally measured signals were normalised to the inert Ar tracer and corrected using their corresponding calibration factors (see Supplementary IV). Then, the curves were further normalised to the concentration of reactant gas in the pulsed mixture such that the integrated area under the reactant curve is 0 at \(100\%\) conversion and 1 at \(0\%\) conversion. Next, the diffusivity of Argon in the packed bed reactor was calculated by fitting a one- zone TAP model to the Ar exit flux curve. As the Knudsen diffusivity is proportional to \(\sqrt{1 / M}\) where \(M\) is the molecular weight of the gas, it becomes possible to calculate the diffusivity of the reactant ( \(\mathrm{O_2}\) , \(M = 32\) ) and product ( \(\mathrm{CO_2}\) , \(M = 44\) ) gases by scaling the diffusivity relative to the inert gas (Ar, \(M = 40\) ). When performing the fitting of the reactant and product curves, the diffusivities of the gases, the reactor length, and the void fraction of the reactor were all fixed, with the only variables being the rate constants for the corresponding model. It is very important to note the regression is performed on each set of exit flux response curves (i.e., exit flux plotted as a function of time) individually. Therefore, the rate constants are calculated separately at each pulse set (and catalyst state) during the experiment. All curve fitting
+
+<--- Page Split --->
+
+was performed in the MATLAB environment using the lsqcurvefit function, with the \(95\%\) confidence intervals for the fitted variables evaluated using the nlparci function. A full description of the MZTRT model used in this work and how the model fitting is performed is available in Supplementary Information I, II, IV, and an example of the MATLAB script used to simulate the TAP experiments is included alongside this paper.  
+
+## Data availability  
+
+The authors declare that the data supporting the findings of this study are available within the article and its Supplementary Information files. An example MATLAB script for the modelling of the TAP response curves is also included alongside the data for the 25TPO experiment. All other relevant data is available from the authors upon reasonable request.  
+
+## Acknowledgements  
+
+C.R. gratefully acknowledges the Rowland Fellowship through the Rowland Institute at Harvard.  
+
+## Author contributions  
+
+Taek- Seung Kim: Conceptualization, Investigation, Formal analysis, Writing - original draft  Christopher R. O'Connor: Investigation, Writing - original draft  Christian Reece: Conceptualization, Formal analysis, Supervision, Writing - original draft, Writing - review & editing  
+
+T.S.K performed the synthesis, characterisation and TAP experiments. C.R.O. performed the DRIFTS experiments. C.R. performed the TAP modelling and supervised the project. All authors in frequent discussions and contributed significantly to writing the manuscript.  
+
+## Additional information  
+
+Supplementary information accompanies this paper. The MATLAB script accompanying this paper requires the Curve Fitting Toolbox and the Statistics and Machine Learning Toolbox.  
+
+## Competing financial interests  
+
+The authors declare no competing financial interests.  
+
+## References  
+
+1. Mou, T. et al. Bridging the complexity gap in computational heterogeneous catalysis with machine learning. Nat. Catal. 6, 122-136 (2023).  
+
+2. Zhdanov, V. P. Impact of surface science on the understanding of kinetics of heterogeneous catalytic reactions. Surf. Sci. 500, 966-985 (2002).
+
+<--- Page Split --->
+
+3. Libuda, J. & Freund, H. J. Molecular beam experiments on model catalysts. Surf. Sci. Rep. 57, 157-298 (2005).  
+
+4. Van Spronsen, M. A., Frenken, J. W. M. & Groot, I. M. N. Surface science under reaction conditions: CO oxidation on Pt and Pd model catalysts. Chem. Soc. Rev. 46, 4347-4374 (2017).  
+
+5. Nguyen, L., Tao, F. F., Tang, Y., Dou, J. & Bao, X.-J. Understanding catalyst surfaces during catalysis through near ambient pressure X-ray photoelectron spectroscopy. Chem. Rev. 119, 6822-6905 (2019).  
+
+6. Lee, J. D. et al. Dilute alloys based on Au, Ag, or Cu for efficient catalysis: from synthesis to active sites. Chem. Rev. 122, 8758-8808 (2022).  
+
+7. Reece, C. & Madix, R. J. Moving from fundamental knowledge of kinetics and mechanisms on surfaces to prediction of catalyst performance in reactors. ACS Catal. 11, 3048-3066 (2021).  
+
+8. Haruta, M. et al. Low-temperature oxidation of CO over fold supported on TiO₂, α-Fe₂O₃, and Co₃O₄. J. Catal. 144, 175-192 (1993).  
+
+9. Reske, R., Mistry, H., Behafarid, F., Roldan Cuenya, B. & Strasser, P. Particle size effects in the catalytic electroreduction of CO₂ on Cu nanoparticles. J. Am. Chem. Soc. 136, 6978-6986 (2014).  
+
+10. Freund, H.-J., Meijer, G., Scheffler, M., Schlögl, R. & Wolf, M. CO oxidation as a prototypical reaction for heterogeneous processes. Angew. Chem. Int. Ed. 50, 10064-10094 (2011).  
+
+11. Kim, T.-S. et al. Catalytic synergy on PtNi bimetal catalysts driven by interfacial intermediate structures. ACS Catal. 10, 10459-10467 (2020).
+
+<--- Page Split --->
+
+12. Kim, T.-S. et al. Catalytic boosting on AuCu bimetallic nanoparticles by oxygen-induced atomic restructuring. Appl. Catal. B: Environ. 331, 122704 (2023).  
+
+13. Maurer, F. et al. Tracking the formation, fate and consequence for catalytic activity of Pt single sites on \(\mathrm{CeO_2}\) . Nat. Catal. 3, 824–833 (2020).  
+
+14. Campbell, C. T., Ertl, G., Kuipers, H. & Segner, J. A molecular beam study of the catalytic oxidation of CO on a Pt(111) surface. J. Chem. Phys. 73, 5862–5873 (1980).  
+
+15. Allian, A. D. et al. Chemisorption of CO and mechanism of CO oxidation on supported platinum nanoclusters. J. Am. Chem. Soc. 133, 4498��4517 (2011).  
+
+16. Yang, X.-F. et al. Single-atom catalysts: a new frontier in heterogeneous catalysis. Acc. Chem. Res. 46, 1740–1748 (2013).  
+
+17. Ding, R. et al. Size-dependent shape distributions of platinum nanoparticles. Nanoscale Adv. 4, 3978–3986 (2022).  
+
+18. Vogt, C. et al. Dynamic restructuring of supported metal nanoparticles and its implications for structure insensitive catalysis. Nat. Commun. 12, 7096 (2021).  
+
+19. Chee, S. W., Arce-Ramos, J. M., Li, W., Genest, A. & Mirsaidov, U. Structural changes in noble metal nanoparticles during CO oxidation and their impact on catalyst activity. Nat. Commun. 11, 2133 (2020).  
+
+20. Somorjai, G. A. & Carrazza, J. Structure sensitivity of catalytic reactions. Ind. Eng. Chem. Fund. 25, 63–69 (1986).  
+
+21. Beniya, A. et al. CO oxidation activity of non-reducible oxide-supported mass-selected few-atom Pt single-clusters. Nat. Commun. 11, 1888 (2020).  
+
+22. Vogt, C. & Weckhuysen, B. M. The concept of active site in heterogeneous catalysis. Nat. Rev. Chem. 6, 89–111 (2022).
+
+<--- Page Split --->
+
+23. Wei, J. & Iglesia, E. Mechanism and site requirements for activation and chemical conversion of methane on supported Pt clusters and turnover rate comparisons among noble metals. J. Phys. Chem. B 108, 4094–4103 (2004).  
+
+24. Ramallo-López, J. M. et al. Complementary methods for cluster size distribution measurements: supported platinum nanoclusters in methane reforming catalysts. J. Mol. Catal. A-Chem. 228, 299–307 (2005).  
+
+25. Engel, T. & Ertl, G. A molecular beam investigation of the catalytic oxidation of CO on Pd (111). J. Chem. Phys. 69, 1267–1281 (1978).  
+
+26. Kale, M. J. & Christopher, P. Utilizing quantitative in situ FTIR spectroscopy to identify well-coordinated Pt atoms as the active site for CO oxidation on \(\mathrm{Al}_2\mathrm{O}_3\) -supported Pt catalysts. ACS Catal. 6, 5599–5609 (2016).  
+
+27. Neumann, S. et al. Insights into the reaction mechanism and particle size effects of CO oxidation over supported Pt nanoparticle catalysts. J. Catal. 377, 662–672 (2019).  
+
+28. Darshini Srinivasan, P. In situ and operando tools and methods for characterization of heterogeneous catalysts. (University of Kansas, 2019).  
+
+29. Yablonsky, G. S. Decoding complexity of chemical reactions. Theor. Found. Chem. Eng. 48, 608–613 (2014).  
+
+30. Reece, C., Redekop, E. A., Karakalos, S., Friend, C. M. & Madix, R. J. Crossing the great divide between single-crystal reactivity and actual catalyst selectivity with pressure transients. Nat. Catal. 1, 852–859 (2018).  
+
+31. Karatok, M., Madix, R. J., Van Der Hoeven, J. E. S., Aizenberg, J. & Reece, C. Quantifying oxygen induced surface enrichment of a dilute PdAu alloy catalyst. Catal. Sci. Technol. 11, 7530–7534 (2021).
+
+<--- Page Split --->
+
+32. Widmann, D. & Behm, R. J. Activation of molecular oxygen and the nature of the active oxygen species for CO oxidation on oxide supported au catalysts. Accounts Chem. Res. 47, 740-749 (2014).  
+
+33. Shekhtman, S. O., Yablonsky, G. S., Glevaes, J. T. & Fushimi, R. R. Thin-zone TAP reactor as a basis of 'state-by-state transient screening'. Chem. Eng. Sci. 59, 5493-5500 (2004).  
+
+34. Shekhtman, S. O., Goguet, A., Burch, R., Hardacre, C. & Maguire, N. CO multipulse TAP studies of 2% Pt/CeO₂ catalyst: Influence of catalyst pretreatment and temperature on the number of active sites observed. J. Catal. 253, 303-311 (2008).  
+
+35. Ding, K. et al. Identification of active sites in CO oxidation and water-gas shift over supported Pt catalysts. Science 350, 189-192 (2015).  
+
+36. Bourane, A. & Bianchi, D. Oxidation of CO on a Pt/Al₂O₃ catalyst: from the surface elementary steps to lighting-off tests: II. kinetic study of the oxidation of adsorbed CO species using mass spectroscopy. J. Catal. 209, 114-125 (2002).  
+
+37. Engel, T. & Ertl, G. Elementary steps in the catalytic oxidation of carbon monoxide on platinum metals. in Advances in Catalysis vol. 28 1-78 (Elsevier, 1979).  
+
+38. Brandão, L., High, E., Kim, T.-S. & Reece, C. Simplifying the temporal analysis of products reactor. Preprint at https://doi.org/10.26434/chemrxiv-2023-d1wgj (2023).  
+
+39. Shekhtman, S. O., Maguire, N., Goguet, A., Burch, R. & Hardacre, C. Three primary kinetic characteristics observed in a pulse-response TAP experiment. Catal. Today 121, 255-260 (2007).  
+
+40. Shekhtman, S. O., Yablonsky, G. S., Chen, S. & Glevaes, J. T. Thin-zone TAP-reactor - theory and application. Chem. Eng. Sci. 54, 4371-4378 (1999).
+
+<--- Page Split --->
+
+41. Gleaves, J. T., Yablonskii, G. S., Phanawadee, P. & Schuurman, Y. TAP-2: An interrogative kinetics approach. Appl. Catal. A-Gen. 160, 55-88 (1997).  
+
+42. Fushimi, R. Temporal analysis of product (TAP). in Springer Handbook of Advanced Catalyst Characterization 899-934 (Springer, 2023).  
+
+43. Constales, D., Yablonsky, G. S., Marin, G. B. & Gleaves, J. T. Multi-zone tap-reactors theory and application: I. The global transfer matrix equation. Chem. Eng. Sci. 56, 133-149 (2001).  
+
+44. Constales, D., Yablonsky, G. S., Marin, G. B. & Gleaves, J. T. Multi-zone TAP-reactors theory and application. III Multi-response theory and criteria of instantaneousness. Chem. Eng. Sci. 59, 3725-3736 (2004).  
+
+45. Geissler, P. L. Temperature dependence of inhomogeneous broadening: on the meaning of isosbestic points. J. Am. Chem. Soc. 127, 14930-14935 (2005).  
+
+46. Hammer, B., Nielsen, O. H. & Nørskov, J. K. Structure sensitivity in adsorption: CO interaction with stepped and reconstructed Pt surfaces. Catal. Lett. 46, 31-35 (1997).  
+
+47. Meunier, F. C., Elgayyar, T., Dembélé, K. & Kaper, H. Stability of Pt-Adsorbed CO on Catalysts for Room Temperature-Oxidation of CO. Catalysts 12, 532 (2022).  
+
+48. Lundwall, M. J., McClure, S. M. & Goodman, D. W. Probing terrace and step sites on Pt nanoparticles using CO and ethylene. J. Phys. Chem. C 114, 7904-7912 (2010).  
+
+49. Hayden, B. E. & Bradshaw, A. M. The adsorption of CO on Pt(111) studied by infrared reflection-absorption spectroscopy. Surf. Sci. 125, 787-802 (1983).  
+
+50. Carrasco, E. et al. Infrared spectra of high coverage CO adsorption structures on Pt(111). Surf. Sci. 604, 1320-1325 (2010).
+
+<--- Page Split --->
+
+51. Hopster, H. & Ibach, H. Adsorption of CO on Pt(111) and Pt 6(111) × (111) studied by high resolution electron energy loss spectroscopy and thermal desorption spectroscopy. Surf. Sci. 77, 109-117 (1978).  
+
+52. McClellan, M. R., Gland, J. L. & McFeeley, F. R. Carbon monoxide adsorption on the kinked Pt(321) surface. Surf. Sci. 112, 63-77 (1981).  
+
+53. Yoshinobu, J., Tsukahara, N., Yasui, F., Mukai, K. & Yamashita, Y. Lateral displacement by transient mobility in chemisorption of CO on Pt(997). Phys. Rev. Lett. 90, 248301 (2003).  
+
+54. Luo, J. S., Tobin, R. G., Lambert, D. K., Fisher, G. B. & DiMaggio, C. L. CO adsorption site occupation on Pt(335): a quantitative investigation using TPD and EELS. Surf. Sci. 274, 53-62 (1992).  
+
+55. Hayden, B. E., Kretzschmar, K., Bradshaw, A. M. & Greenler, R. G. An infrared study of the adsorption of CO on a stepped platinum surface. Surf. Sci. 149, 394-406 (1985).  
+
+56. Banholzer, W. F. & Masel, R. I. A reflection-absorption infrared study of carbon monoxide and nitric oxide adsorption on platinum(100). Surf. Sci. 137, 339-360 (1984).  
+
+57. Fanson, P. T., Delgass, W. N. & Lauterbach, J. Island formation during kinetic rate oscillations in the oxidation of CO over Pt/SiO₂: a transient Fourier transform infrared spectrometry study. J. Catal. 204, 35-52 (2001).  
+
+58. Brandt, K. & Hughes, M. R. The interpretation of CO adsorbed on Pt/SiO₂ of two different particle-size distributions. Surf. Sci. 286, 15-25 (1993).  
+
+59. Avanesian, T. et al. Quantitative and atomic-scale view of CO-induced Pt nanoparticle surface reconstruction at saturation coverage via DFT calculations coupled with in situ TEM and IR. J. Am. Chem. Soc. 139, 4551-4558 (2017).
+
+<--- Page Split --->
+
+60. Kappers, M. J. & Van Der Maas, J. H. Correlation between CO frequency and Pt coordination number. A DRIFT study on supported Pt catalysts. Catal. Lett. 10, 365–373 (1991).  
+
+61. Hossain, M. J., Rahman, M. M. & Jafar Sharif, Md. Preference for low-coordination sites by adsorbed CO on small platinum nanoparticles. Nanoscale Adv. 2, 1245–1252 (2020).  
+
+62. Liu, A. et al. Getting insights into the temperature-specific active sites on platinum nanoparticles for CO oxidation: A combined in situ spectroscopic and ab Initio density functional theory study. ACS Catal. 9, 7759–7768 (2019).  
+
+63. Kustov, L. M., Ostgard, D. & Sachtler, W. M. H. IR spectroscopic study of Pt/Kl zeolites using adsorption of CO as a molecular probe. Catal. Lett. 9, 121–126 (1991).  
+
+64. Di, M. et al. Chasing \(\mathrm{PtO_x}\) species in ceria supported platinum during CO oxidation extinction with correlative operando spectroscopic techniques. J. Catal. 409, 1–11 (2022).  
+
+65. Klimavicius, V., Neumann, S., Kunz, S., Gutmann, T. & Buntkowsky, G. Room temperature CO oxidation catalysed by supported Pt nanoparticles revealed by solid-state NMR and DNP spectroscopy. Catal. Sci. Technol. 9, 3743–3752 (2019).  
+
+66. Persson, B. N. J., Tüshaus, M. & Bradshaw, A. M. On the nature of dense CO adlayers. J. Chem. Phys. 92, 5034–5046 (1990).  
+
+67. Baró, A. M. & Ibach, H. New study of CO adsorption at low temperature (90 K) on Pt (111) by EELS. J. Chem. Phys. 71, 4812 (1979).  
+
+68. Zaera, F., Liu, J. & Xu, M. Isothermal study of the kinetics of carbon monoxide oxidation on Pt(111): Rate dependence on surface coverages. J. Chem. Phys. 106, 4204–4215 (1997).  
+
+69. Mhadeshwar, A. B. & Vlachos, D. G. A thermodynamically consistent surface reaction mechanism for CO oxidation on Pt. Combust. Flame 142, 289–298 (2005).
+
+<--- Page Split --->
+
+70. Shekhtman, S. O. & Yablonsky, G. S. Thin-zone TAP reactor versus differential PFR:  
+
+analysis of concentration nonuniformity for gas–solid systems. Ind. Eng. Chem. Res. 44,  
+
+6518–6522 (2005).  
+
+71. Dumesic, J. A., Rudd, D. F., Aparicio, L. M., Rekoske, J. E. & Trevi��o, A. A. The  
+
+microkinetics of heterogeneous catalysis. (American Chemical Society, 1993).  
+
+72. Janthon, P., Viñes, F., Sirijaraensre, J., Limtrakul, J. & Illas, F. Adding pieces to the  
+
+CO/Pt(111) puzzle: the role of dispersion. J. Phys. Chem. C 121, 3970–3977 (2017).  
+
+73. Feibelman, P. J. et al. The CO/Pt(111) puzzle. J. Phys. Chem. B 105, 4018–4025 (2001).  
+
+74. Neugebohren, J. et al. Velocity-resolved kinetics of site-specific carbon monoxide oxidation on platinum surfaces. Nature 558, 280–283 (2018).  
+
+75. Song, H. C. et al. The effect of the oxidation states of supported oxides on catalytic activity:  
+
+CO oxidation studies on Pt/cobalt oxide. Chem. Commun. 55, 9503–9506 (2019).  
+
+76. High, E. A., Lee, E. & Reece, C. A transient flow reactor for rapid gas switching at  
+
+atmospheric pressure. Review of Scientific Instruments 94, 054101 (2023).  
+
+77. Morgan, K. et al. Forty years of temporal analysis of products. Catal. Sci. Technol. 7, 2416–2439 (2017).  
+
+78. Meunier, F. C. Pitfalls and benefits of in situ and operando diffuse reflectance FT-IR spectroscopy (DRIFTS) applied to catalytic reactions. React. Chem. Eng. 1, 134–141 (2016).
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+SiteDependentKineticsSI.pdf  23InterrogatingSites3PathwayModel.zip
+
+<--- Page Split --->

preprint/preprint__0eedd8efcf3fab683389f91e77a829502b5c698130c328f60b9e1bf314f4cc29/images_list.json

@@ -0,0 +1,107 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Figure 1. The framework of Retro-MTGR. Molecules in the form of graphs first are represented by an MPNN-based atom encoder to learn initial atom embeddings. The atom embedding enhancer (AEE) further boosts the atom embeddings by contrastive learning on molecules and their synthons w.r.t. molecule embeddings. The Reaction-Center Perceptron (RCP) leverages a bond-level readout on enhanced atom embeddings to learn bond embeddings, which are sequentially augmented by extra bond energies and then recognize reaction centers among bonds. After that, the leaving group predictor (LGP) learns LG embeddings based on a leaving group co-occurrence graph and measures the proximity between them and synthon embeddings (involving atoms and bonds in reaction centers) to predict appropriate LGs for given synthons.",
+    "footnote": [],
+    "bbox": [
+      [
+        150,
+        90,
+        825,
+        560
+      ]
+    ],
+    "page_idx": 7
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2. Ablation comparison. Compared with the three variants (red, green and purple bars), Retro-MTGR (blue bars) achieves the best retrosynthesis prediction in terms of top-1, top-3, and top-5 in the case of both unknown and known reaction types.",
+    "footnote": [],
+    "bbox": [
+      [
+        152,
+        205,
+        825,
+        456
+      ]
+    ],
+    "page_idx": 16
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Figure 3. Breaking energy distribution of chemical bonds. The X-axis shows the bond energy (0-900) divided into 20 intervals. The Y-axis shows the frequency of chemical bonds falling in different bins. As illustrated, bonds having bond energy \\(> 360 \\mathrm{kJ / mol}\\) are usually ordinary bonds. However, \\(45.33\\%\\) of ordinary bonds overlap with reaction centers in the case of breaking energies \\(< 360 \\mathrm{kJ / mol}\\) .",
+    "footnote": [],
+    "bbox": [
+      [
+        152,
+        433,
+        783,
+        675
+      ]
+    ],
+    "page_idx": 17
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Figure 4. Bond space. (A). Reaction centers and ordinary bonds. Red dots indicate reaction centers while blue dots indicate ordinary bonds. (B). Bond types. Different colors represent different bonds. (C). Bond energies. Different colors indicate different bond energy bins (kJ/mol). In terms of chemical symbols, C stands for carbon atoms, c is for carbon atoms in aromatic bonds, and C' is for carbon atoms in general rings.",
+    "footnote": [],
+    "bbox": [
+      [
+        153,
+        80,
+        828,
+        808
+      ]
+    ],
+    "page_idx": 19
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Figure 5. Electrical property distribution of chemical bonds. The X-axis denotes pairwise electrical property patterns of atom pairs forming bonds. Its left zone lists six patterns of same/similar electrical properties while its right zone lists four patterns of opposite electrical properties. The Y-axis indicates the frequencies of electrical property patterns. The member atoms of ordinary bonds usually have same or similar electrical properties (80.1%), whereas those in reaction centers tend to have opposite electrical properties (97.0%).",
+    "footnote": [],
+    "bbox": [
+      [
+        153,
+        85,
+        825,
+        323
+      ]
+    ],
+    "page_idx": 21
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_6.jpg",
+    "caption": "Figure 6. Leaving Group view. (A) Electrical property map. LGs having positive and negative electrical properties are rendered by red and blue. In total, \\(94.3\\%\\) of LG pairs have opposite electrical properties. (B) Reaction type map. LGs occurring in single reaction types, multiple reaction types ( \\(\\leq 3\\) ) and many types ( \\(\\geq 4\\) , reaction-common) are highlighted in different colors.",
+    "footnote": [],
+    "bbox": [
+      [
+        152,
+        90,
+        831,
+        707
+      ]
+    ],
+    "page_idx": 23
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_7.jpg",
+    "caption": "Figure 7. Predicted retrosynthetic routes and real chemical synthetic routes. A. Retrosynthesis prediction of Sonidegib. Sonidegib ('1') is split at the reaction center into intermediate molecules '2' and '3', where the highest",
+    "footnote": [],
+    "bbox": [
+      [
+        150,
+        365,
+        828,
+        859
+      ]
+    ],
+    "page_idx": 25
+  }
+]

preprint/preprint__0eedd8efcf3fab683389f91e77a829502b5c698130c328f60b9e1bf314f4cc29/preprint__0eedd8efcf3fab683389f91e77a829502b5c698130c328f60b9e1bf314f4cc29.mmd

@@ -0,0 +1,500 @@
+
+# Retro-MTGR: Molecule Retrosynthesis Prediction via Multi-Task Graph Representation Learning  
+
+Jian- Yu Shi jianyushi@nwpu.edu.cn  
+
+Northwestern Polytechnical University  
+
+Pengcheng Zhao  
+
+School of Life Sciences, Northwestern Polytechnical University  
+
+Xue- Xin Wei  
+
+School of Life Sciences, Northwestern Polytechnical University  
+
+Qiong Wang  
+
+School of Life Sciences, Northwestern Polytechnical University  
+
+Qi- Hao Wang  
+
+School of Chemistry and Chemical Engineering, Northwestern Polytechnical University  
+
+Jia- Ning Li  
+
+School of Life Sciences, Northwestern Polytechnical University  
+
+Jie Shang  
+
+School of Life Sciences, Northwestern Polytechnical University  
+
+Cheng Lu  
+
+Institute of Basic Research in Clinical Medicine China Academy of Chinese Medical Sciences  
+
+## Article  
+
+# Keywords:  
+
+Posted Date: September 6th, 2023  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 3205328/v1  
+
+License: © © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Additional Declarations: There is NO Competing Interest.
+
+<--- Page Split --->
+
+Version of Record: A version of this preprint was published at Nature Communications on January 18th, 2025. See the published version at https://doi.org/10.1038/s41467-025-56062-y.
+
+<--- Page Split --->
+
+# Retro-MTGR: Molecule Retrosynthesis Prediction via Multi-Task Graph Representation Learning  
+
+Peng- Cheng Zhao \(^{1}\) , Xue- Xin Wei \(^{1}\) , Qiong Wang \(^{1}\) , Qi- Hao Wang \(^{2}\) , Jia- Ning Li \(^{1}\) , Jie Shang \(^{1*}\) , Cheng Lu \(^{3*}\) , Jian- Yu Shi \(^{1*}\)  
+
+\(^{1}\) School of Life Sciences, Northwestern Polytechnical University, Xi'an 710072, China \(^{2}\) School of Chemistry and Chemical Engineering, Northwestern Polytechnical University, Xi'an 710072, China  
+
+\(^{3}\) Institute of Basic Research in Clinical Medicine China Academy of Chinese Medical Sciences, Beijing 100700, China  
+
+Corresponding author: jianyushi@nwpu.edu.cn; lv_cheng0816@163. com; shangjie03@nwpu.edu.cn  
+
+## Abstract  
+
+It is a vital bridging step to infer appropriate synthesis reaction routes (i.e., retrosynthesis) of newly- designed molecules. Unlike classical experience- based retrosynthesis approaches, artificial intelligence enables a cheap and fast retrosynthesis approach. Template- based models, limited in known synthesis templates, leverage substructure searching to infer candidate reaction centers (i.e., bonds). In contrast, both translation- based models (TransMs) and discriminative methods (DiscMs) are free to synthesis templates. TransM regards retrosynthesis as a translation from the target molecule to its reactants by generative algorithms. DiscM, directly inspired by chemical synthesis, performs reaction center recognition and leaving group identification in turn. Nevertheless, TransMs are redundant and weakly interpretable, while existing DiscMs neglect the associations between reaction centers and leaving groups. To address these issues, this paper elaborates a novel discriminative Multi- Task Graph Representation learning model of Retrosynthesis prediction (Retro- MTGR). It solves two major supervised discriminative tasks (i.e., the reaction center recognition and the leaving group identification respectively), and an auxiliary self- supervised task (i.e., atom embedding enhancer) simultaneously. The comparison with various state- of- the- art methods first demonstrates the superiority of Retro- MTGR. Then, the ablation studies reveal how its crucial components contribute to the prediction respectively, including the atom embedding enhancer, bond energies, and the leaving group co- occurrence graph. More importantly, comprehensive investigations validate its chemical interpretability by answering two questions: why a bond can be the reaction center or not, and what leaving groups are appropriate to given synthons. The answers demonstrate that Retro- MTGR can reflect five underlying chemical synthesis rules by characterizing molecule structures
+
+<--- Page Split --->
+
+alone. Finally, two case studies demonstrate that the inferred retrosynthesis routes by Retro- MTGR are significantly consistent with those achieved by performed chemical synthesis assays. It's anticipated that our Retro- MTGR can provide prior guidance for real retrosynthesis route planning. The code and data underlying this article are freely available at https://github.com/zpczaizheli/Retro- MTGR.  
+
+## 1 Introduction  
+
+By integrating artificial intelligence (AI) technologies<sup>1</sup>, modern drug design has exhibited marvelous achievements on diverse tasks (e.g., target screening<sup>2</sup>, molecule generation<sup>3</sup>, ADMET prediction<sup>4</sup>, etc.) with a significant reduction in cost and time<sup>5,6</sup>. Once the chemical structure of a small molecule is determined in silico, there is an important task, retrosynthesis, which finds available reactants to be synthesized into the drug- like molecule in reality<sup>7</sup>. Such a retrosynthesis process works as a bridge from in silico to in reality. Compared to the ordinary synthesis reaction, the retrosynthesis is its inverse inference process<sup>8,9</sup>. A complete route of retrosynthesis is composed of multiple steps of synthesis reactions. However, even inferring a single step of synthesis heavily relies on individual domain experiences of chemists under costly trial- and- error assays<sup>10</sup>.  
+
+In recent years, both the accumulation of chemical synthesis data and the blooming of deep learning methods boost the rapid development of computer- assisted synthesis processes (CASP) in retrosynthesis, which are roughly grouped into template- based, translation- based and discriminative methods.  
+
+Upon summarizing empirical rules of existing chemical syntheses (by usually RDKit<sup>11</sup>), template- based methods infer the single- step retrosynthesis of a newly given molecule by local structure similarity- based searching. For example, after determining possible reaction types of a target molecule, DHN, derived from gating neural networks, searches candidate templates among reaction type- specific templates<sup>12</sup>. GLN, a conditional graphical model upon graph neural networks, acquires candidate templates of the target molecule by subgraph pattern matching<sup>13</sup>. However, template- based methods cannot predict the retrosynthesis for target molecules having novel synthesis patterns outside the synthesis rules in the template library. In addition, it is tedious to update template libraries as new synthesis knowledge is discovered<sup>14</sup>.  
+
+In contrast, both translation- based and discriminative methods are template- free and can predict the retrosynthesis reaction without a pre- built template library. Translation- based methods generally regard the retrosynthesis process as a case of machine translation<sup>15</sup>, which learns a translation model from the target molecule to its reactants by generative algorithms (e.g., LSTM<sup>16</sup> and Transformer<sup>17</sup>). Existing
+
+<--- Page Split --->
+
+translation- based methods can be further categorized into sequence- to- sequence translation models (Seq2Seq) and graph- to- seq translation models (Graph2Seq).  
+
+(1) Seq2Seq models treat a target molecule as one string (e.g., SMILES) and its two reactants as another string (the concatenation of two reactant SMILES). The first Seq2Seq model retrosynthesis method utilizes attention-enhanced LSTMs to convert the target molecule to its reactants under an encoder-decoder architecture<sup>16</sup>. As the new super-star in natural language processing, Transformer is also applied to retrosynthesis prediction by treating each character in SMILE as a word in recent years<sup>18</sup>. However, these methods arise a new issue that generated reactants are probably invalid in terms of chemistry. To meet the chemical validity of generated reactants, SCROP designs an extra syntax post-checker (derived from RDKit) based on Transformer<sup>15</sup>. RetroTRAE treats molecule substructures (capturing local atomic environment) as words in the Transformer to guarantee the validity of generated reactants<sup>19</sup>. Although these Seq2Seq models have achieved inspiring retrosynthesis predictions, they ignore rich information hidden in molecule chemical structures (i.e., the topology between atoms and bonds)<sup>20</sup>.  
+
+(2) Graph2Seq models, representing target molecules as graphs, enriches the molecule representation and then map them into their reactant sequences under the auto-regressive generation framework<sup>8</sup>. Since the decoders of Graph2Seq models are similar, their contributions mainly focus on the design of encoders by graph neural networks (GNNs). For example, to encode target molecular graphs, G2GT designs a self-attention module enhanced by degrees of atoms and pairwise shortest distances between atoms<sup>21</sup>. Graph2SMILES utilizes a directed message-passing neural network (MPNN) to capture atom representations with the extra enhancement of global attention encoding<sup>22</sup>. By treating a chemical reaction as a queue of graph edits, MEGAN adopts a GNN-based encoder-decoder architecture and outputs reactants by a queue of leaned graph edits on the input molecule graph step by step<sup>23</sup>. Usually, these Graph2Seq models achieve better retrosynthesis prediction since they capture richer topology of molecule structures than Seq2Seq models.  
+
+However, both Seq2Seq and Graph2Seq models are still derived from generative models, which cannot provide well-interpretable results to chemists or pharmacologists in terms of chemical synthesis mechanisms. Moreover, dissimilar to the translation from one language to another language, translation- based retrosynthesis prediction has a redundant learning process due to highly overlapping structures between target molecules and their reactants.  
+
+Different from translation- based models, discriminative models are inspired by real
+
+<--- Page Split --->
+
+chemical synthesis. In brief, their prime step is to find the reaction center (e.g., the bond broken in retrosynthesis inference) where the molecule is split into two synthons (i.e., incomplete reactants) \(^{24}\) . Then, two synthons are attached to appropriate functional groups (i.e., leaving groups, LGs) to form reactants respectively. For example, Hasic et al. characterize local substructures of two bonding atoms as the bond representation by extended- connectivity fingerprints, and query candidate reactants by similarity search in a pre- built compound library \(^{25}\) . The G2Gs model encodes bonds by atom local topology and molecule global topology to find the reaction center by relational graph convolutional networks, and then builds a variational graph model to infer functional groups to be attached to synthons \(^{26}\) . However, current discriminative models treat the reaction center recognition and the LG identification as two separate steps, such that the association between the target molecule and its reactants is neglected.  
+
+In summary, existing translation- based generative models have weak interpretability and a redundant translation from target molecules and their reactants, while current discriminative models neglect the association between the reaction center recognition and the LG identification.  
+
+To address these issues, this paper elaborates a novel Multi- Task Graph Representation learning framework of Retrosynthesis prediction (Retro- MTGR), which is a discriminative model in essence. It solves three related tasks, including two major supervised discriminative tasks (i.e., the reaction center recognition and the LG identification respectively) and an auxiliary self- supervised task (i.e., atom embedding enhancer).  
+
+Overall, the main contributions of this work are as follows.  
+
+(1) This work proposes a novel retrosynthesis prediction model (Retro-MTGR), which consists of a reaction-center perceptron (RCP), a leaving group predictor (LGP), and an atom embedding enhancer (AEE). RCP characterizes the molecule topology to recognize the retrosynthesis reaction center, while AEE utilizes the redundancy between the product molecule and its synthons to boost atom embeddings for RCP. By leveraging the co-occurrence between LGs, LGP identifies appropriate LGs of synthons to provide complete reactants.  
+
+(2) Aiming to better chemical interpretability of retrosynthesis prediction, Retro-MTGR answers why a bond can be the reaction center or not from three aspects. First, a bond having high-breaking energy (>=360 kJ/mol) is usually an ordinary bond. Secondly, aromatic bonds (c-c) between carbon atoms are always ordinary bonds, while other types of bonds can be either reaction
+
+<--- Page Split --->
+
+centers or ordinary bonds. Last, a bond is the reaction center in a molecule if its member atoms tend to have opposite electrical properties (reflected by local substructures), otherwise an ordinary bond.  
+
+(3) Furthermore, Retro-MTGR answers what leaving groups are appropriate to given synthons as follows. First, LG pairs in reaction centers usually have opposite electrical properties, and occurrence-dominant LG pairs always consist of two simple groups (e.g., H, OH, or halogens). Secondly, individual LGs can be categorized into reaction-common and reaction-specific LGs. The former group is usually of simple LG and occurs in multiple types of synthesis reactions. The latter is of usually chemical substructure groups and occurs only in specific types of reactions. Also, either similar chemical properties or structures between LGs imply their potential substitution in synthesis reactions.  
+
+## 2. Methods  
+
+### 2.1 Problem formulation and model construction  
+
+Given a set of \(n\) chemical reactions \(R = \{r_1^i + r_2^i = c_i|i = 1,\dots,n\}\) , where the reactants \(r_1^i\) and \(r_2^i\) are two reactant molecules for the synthesis of target molecule \(c_i\) . The task is to find the retrosynthetic strategy for a newly-designed target molecule \(c\) (i.e., to recommend reactants \((r_1,r_2)\) for \(c\) ).  
+
+To perform a chemist- like retrosynthesis, we elaborate a Multi- Task Graph Representation learning framework of Retrosynthesis prediction (Retro- MTGR), which contains a Reaction- Center Perceptron (RCP) module, an Atom Embedding Enhancer (AEE) module and a Leaving Group Predictor (LGP). They account for two major tasks and one auxiliary task respectively. The first major task, implemented by RCP, is modeled as a binary discriminative problem, which recognizes the reaction center \(b_{u^*,v^*}\) among all the bonds \(\{b_{u,v}\}\) of the target molecule. Also, it breaks down \(b_{u^*,v^*}\) to obtain two synthons \(s_{u^*}\) and \(s_{v^*}\) , where \(u,v\) are two bonding atoms in \(c\) . To support the first major task in terms of atom embeddings, the auxiliary task (implemented by AEE) is modeled as a self- supervised contrast learning problem, which characterizes the structural commonality and difference between \(c\) and its synthons \((s_{u^*,s_{v^*}})\) . The second major task is modeled as a multi- class discriminative problem, which assigns appropriate leaving groups \((k_{u^*,k_{v^*}})\) to the synthons \((s_{u^*,s_{v^*}})\) to form complete reactants \((r_{u^*,r_{v^*}})\) under the enhancement of leaving group dependence.
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Figure 1. The framework of Retro-MTGR. Molecules in the form of graphs first are represented by an MPNN-based atom encoder to learn initial atom embeddings. The atom embedding enhancer (AEE) further boosts the atom embeddings by contrastive learning on molecules and their synthons w.r.t. molecule embeddings. The Reaction-Center Perceptron (RCP) leverages a bond-level readout on enhanced atom embeddings to learn bond embeddings, which are sequentially augmented by extra bond energies and then recognize reaction centers among bonds. After that, the leaving group predictor (LGP) learns LG embeddings based on a leaving group co-occurrence graph and measures the proximity between them and synthon embeddings (involving atoms and bonds in reaction centers) to predict appropriate LGs for given synthons. </center>  
+
+### 2.2 Reaction-Center Perceptron  
+
+Identifying the reaction center is the first step in the inference of retrosynthesis. Inspired by this chemical empirical approach, we primarily attempt to recognize the reaction center among all the bonds of a given target molecule. The reaction center is the bond broken in terms of retrosynthesis. Thus, the task of reaction center recognition can be naturally
+
+<--- Page Split --->
+
+modeled as a binary discriminative problem, which recognizes the reaction center \(b_{u^{*},v^{*}}\) among all the bonds \(\{b_{u,v}\}\) of the target molecule.  
+
+For this task, we design a reaction- center perceptron (RCP) module, which includes an atom encoder, a bond- level readout layer, and a multi- layer perceptron (MLP). The atom encoder is implemented by a two- layer message passing neural network (MPNN) to turn molecule graphs \(\mathcal{G} = (\mathcal{A},\mathcal{B})\) into atom embeddings \(\{\mathbf{a}_i\}\) , which are further refined by an Atom Embedding Enhancer (AEE). The bond- level readout layer generates bond embeddings \(\{\mathbf{b}_{u,v}\}\) , which are further boosted by concatenating with bond energy \(b_e\) . The MLP accounts for the discrimination of bonds by \(y = \mathcal{F}(\mathbf{b}_{u,v})\) , where \(y = 1\) if \(b_{u,v}\) is the reaction center, \(y = 0\) otherwise.  
+
+## Atom Encoder  
+
+According to chemical structure, each compound \(m\) is represented as a molecule graph \(\mathcal{G} = (\mathcal{A},\mathcal{B})\) , where \(\mathcal{A}\) is the set of its atoms \(\{a_i\}\) , \(\mathcal{B}\) is the set of its chemical bonds \(\{b_{ij}\}\) , and \(i,j = 1,2,\dots,|\mathcal{A}|\) . Let \(\mathbf{E}\in R^{N\times N}(N = |\mathcal{A}|)\) be its adjacency matrix, in which \(e_{ij} = 1\) , indicates the occurring bond \(\left(b_{ij}\in \mathcal{B}\right)\) between two atoms (i.e., \(a_i\) and \(a_j\) ) and \(e_{ij} = 0\) indicates no bond. Suppose that \(\mathbf{x}_i\in R^n\) is the initial feature vector of atom \(a_i\) , which is usually coded into a vector containing one- hot- shaped atom types, number of hydrogen atoms, and other attributes. Since \(\mathbf{x}\) is sparse, an extra multi- layer MLP maps it into its dense form \((\mathbf{x}_i\in R^p)\) to avoid the vanishing gradient problem.  
+
+Both \(\mathbf{E}\) and \(\mathbf{x}\) are input into a two- layer MPNN to generate atom embeddings \(\{\mathbf{a}_i\}\) for molecule \(c\) . The MPNN updates the embedding \(\mathbf{a}_i\) of each atom \(a_i\) by aggregating those of its neighboring atoms in a layer as follows,  
+
+\[\mathbf{a}_i^{t + 1} = \sigma \left(\mathbf{w}_1^t \left(\sum_{j\in \mathcal{N}(a_i)} (\mathbf{w}_j^t \mathbf{a}_j^t) + \mathbf{b}^t\right) + \mathbf{w}_2^t \mathbf{a}_i^t\right), t = \{1,2\} , \quad (1)\]  
+
+where \(\mathbf{a}_i^t\) denotes the embedding of atom \(i\) in the t- th layer of MPNN, \(\mathbf{a}_i^1 = \mathbf{x}\) , \(\mathcal{N}(a_i)\) denotes the neighbors of atom \(a_i\) in the molecule graph \(\mathcal{G}\) , \(\sigma (\cdot)\) is a non- linear activation function (e.g., \(ReLu\) ), all \(\{\mathbf{w}^t\}\) are layer- wise learnable weight matrices accounting for a linear transformation, and \(\mathbf{b}^t\) denotes a learnable bias.  
+
+## Perceptron  
+
+After passing through the MPNN, the initial atom feature \(\mathbf{x}\in R^n\) is turned to the updated atom embedding \(\mathbf{a}_i\in R^q\) . It is further refined by the AEE module, which characterizes the structural commonality and difference between the molecule and its synths. Meanwhile, the refined atom embeddings are utilized by the LGP module to help find appropriate leaving groups for the synths. All three tasks are associated together by shared atom embeddings. See Sections 2.3 and 2.4 for details.  
+
+Sequentially, the refined atom embeddings \(\{\mathbf{a}_i\}\) are then used to generate bond
+
+<--- Page Split --->
+
+embedding by the bond- level readout. Let \(b_{ij}\) be the bond connecting atoms \(a_{i}\) and \(a_{j}\) . Unlike the ordinary molecule- level readout (e.g., the combination of all atoms), the RCP model defines a bond- level readout function \(\mathcal{R}_{B}(a_{i},a_{j})\) , which is augmented by bond energy, to obtain the bond embedding \(\mathbf{b}_{ij}\) as follows,  
+
+\[\mathbf{b}_{ij} = \mathcal{R}_{B}(a_{i},a_{j}) = \left[(\mathbf{a}_{i} + \mathbf{a}_{j}); g_{ij}\right], \quad (2)\]  
+
+where \(g_{ij}\) is the theoretical bond energy, and ';' indicates the concatenation of \(g_{ij}\) and the atom embedding summation. As we observed, bond energy contributes to screen out the ordinary bonds having high bond energies. See also Section 3.4.1.  
+
+Last, RCP identifies the reaction center among all the bonds of \(m\) . Define \(\mathcal{C}\) as the set of bond flags \(\{c_{ij}\}\) w.r.t. molecule \(m\) , where \(c_{ij} = 1\) if \(b_{ij}\) is the reaction center (a positive sample), \(c_{ij} = 0\) otherwise (a negative sample). Based on the abovementioned bond embeddings \(\{\mathbf{b}_{ij}\}\) , it is naive to construct a two- layer MLP (denoted as \(\mathcal{F}_{b}\) ) as the classifier to achieve such a bond identification (i.e., \(c_{ij}^{*} = \mathcal{F}_{b}(\mathbf{b}_{ij})\) ). To train the model, the cross- entropy loss function over all the training molecules is defined as follows:  
+
+\[l_{bond} = -\frac{1}{|M|}\sum_{m = 1}^{|M|}\left(\frac{1}{|\mathcal{B}_{m}|}\sum_{b_{ij}\in \mathcal{B}_{m}}^{|\mathcal{B}_{m}|}\left(c_{ij}\log c_{ij}^{*} + \left(1 - c_{ij}\right)\log \left(1 - c_{ij}^{*}\right)\right)\right), \quad (3)\]  
+
+where \(\mathcal{M}\) is the set of all the training molecules, and \(\mathcal{B}_{m}\) is the bond set of \(m\) .  
+
+### 2.3 Atom Embedding Enhancer  
+
+As we observed, there is an amazing analog between the edge perturbation in graph contrast learning \(^{29}\) and the breaking of the reaction center in the retrosynthesis. Inspired by this observation, we designed an atom embedding enhancer (AEE) module based on graph contrastive learning.  
+
+For a target molecule \(m\) , we treat its two synthons \((s_{1}\) and \(s_{2}\) ) as a new perturbed molecule \(s\) , which is generated by removing the reaction center from \(m\) . We also collect another different molecule \(\bar{s}\) , which is randomly selected from other molecules or their perturbed molecules. Let \(\mathbf{h}_{m}\) , \(\mathbf{h}_{s}\) , and \(\mathbf{h}_{\bar{s}}\) be their embeddings respectively. These molecule embeddings are generated by an atom encoder and a molecule- level readout function \(\mathcal{R}_{M}(\cdot)\) . The former has the shared parameters with the atom encoder used in the RCP module. For a given molecule, \(\mathcal{R}_{M}(\cdot)\) aggregates the embeddings of its all atoms to generate the molecular- level embedding by an ordinary average pooling \(\mathbf{h} = \frac{1}{|\mathcal{A}|}\sum_{i = 1}^{|\mathcal{A}|}\mathbf{a}_{i}\) .  
+
+In terms of contrastive learning, the molecule embedding pair of \(\mathbf{h}_{m}\) and \(\mathbf{h}_{s}\) is regarded as a positive sample while the pair of \(\mathbf{h}_{m}\) and \(\mathbf{h}_{\bar{s}}\) is taken as a negative sample. Our goal is to train a contrastive learning model, which pushes \(\mathbf{h}_{m}\) and \(\mathbf{h}_{s}\) as near as possible (similar) while pushing \(\mathbf{h}_{m}\) and \(\mathbf{h}_{\bar{s}}\) as far as possible (different). For this
+
+<--- Page Split --->
+
+purpose, we design a contrastive loss function as follows:  
+
+\[l_{contrast} = -\frac{1}{|M|}\sum_{m = 1}^{|M|}\log \left(\frac{\exp(\mathbf{h}_{m},\mathbf{h}_{s}^{T})}{\exp(\mathbf{h}_{m},\mathbf{h}_{s}^{T}) + \exp(\mathbf{h}_{m},\mathbf{h}_{s}^{T})}\right). \quad (4)\]  
+
+Thus, the AEE module enables an ingenious utility of chemical structural commonness and differences between a molecule and its synthons to enhance atom embeddings for other tasks.  
+
+### 2.4 Leaving Group Predictor  
+
+Once synthons are decided based on the reaction center, chemists can obtain corresponding reactants by attaching appropriate leaving groups (LGs) to them. Thus, the task of LG recognition can be modeled as a multi- class classification. Inspired by chemists, we hold the idea that both the reaction center and the local substructures around reaction sites are crucial factors to determine LGs. Meanwhile, we consider the fact that LGs are not independent but are associated in the chemical synthesis sense (See also Section 3.4.2). Based on these considerations, we propose an elaborate leaving group predictor (LGP) based on multi- class classification to identify leaving groups.  
+
+Let \(\mathcal{K}\) be the list of all possible LGs, \(b_{u,v}\) be the reaction center of molecule \(m\) , \(a_{u},a_{v}\) be the reaction sites (i.e., atoms forming \(b_{u,v}\) ), \(s_{u},s_{v}\) be its synthons. Formally, \(s_{u}\) can be assigned with one or more LGs (i.e., \(\mathcal{K}_{u} = \mathcal{F}(s_{u})\subseteq \mathcal{K}\) ) to form its corresponding reactants \(r_{u}\) (i.e., \(\{r_{u}(i) = \mathcal{K}_{u}(i) + s_{u}\}\) , \(i = 1,\dots,|\mathcal{K}_{u}|\}\) ).  
+
+To implement our first idea, the reaction center \(b_{u,v}\) is represented by its bond embedding \(\mathbf{b}_{u,v}\) in RCP, while the local substructure around reaction site \(a_{u}\) is just represented by the atom embedding \(\mathbf{a}_{u}\) , which already aggregates its neighbors due to the MPNN in RCP. Thus, the embedding of the synthon \(s_{u}\) containing \(a_{u}\) can be defined as their concatenation \(\mathbf{s}_{u} = [\mathbf{a}_{u};\mathbf{b}_{uv}]\) . Similarly, we can define the embedding of \(s_{v}\) by \(\mathbf{s}_{v} = [\mathbf{a}_{v};\mathbf{b}_{uv}]\) .  
+
+To implement our second idea, we construct the leaving group co- occurrence graph (LGCoG) \(\mathcal{G}_{k} = \{\mathcal{K},\mathcal{E}\}\) , where \(\mathcal{K} = \{k_{i}|i = 1,\dots,|\mathcal{K}|\}\) denotes the set of nodes (leaving groups), and \(\mathcal{E} = \{e_{ij}\}\) denotes the set of weighted edges (normalized co- occurrences between leaving groups). Each LG is a small- size chemical substructure (e.g., - OH, - B(OH)2) or an atom/ion individual (e.g., - Cl, - Br, - H). The popular one- hot coding is used as initial node features \(\{\mathbf{k}_{i}^{0}\}\) . The edge building contains two steps as follows. First, the LG co- occurrence is calculated based on the training dataset, where the co- occurrence of two LGs is counted if they are involved in the same reaction. Define \(\mathbf{U} = \{u_{ij}\} \in \mathbb{R}^{|\mathcal{K}|\times |\mathcal{K}|}\) as the LG co- occurrence matrix, where \(u_{ij}\) denotes the pairwise co- occurrence counts between \(k_{i}\) and \(k_{j}\) . Then, a probability matrix \(\mathbf{P}\) can be calculated by \(\mathbf{U}\) . Therefore, \(p_{ij}\) is calculated as follows:
+
+<--- Page Split --->
+
+\[p_{ij} = \frac{u_{ij}}{\sum_{j = 1}^{k}\sum_{i = 1}^{k}u_{ij}}. \quad (5)\]  
+
+We set it as the weight of the edge from \(k_{j}\) to \(k_{i}\) (i.e., \(e_{ij} = p_{ij}\) ). Thus, the embedding \((\mathbf{k}_{i})\) of LG \(k_{i}\) can be represented by performing an MPNN on \(g_{k}\) as follows:  
+
+\[\mathbf{k}_{i}^{t + 1} = \sigma \left(\mathbf{w}_{1}^{t}\left(\sum_{j\in \mathcal{N}(k_{i})}\left(e_{ij}\mathbf{w}_{j}^{t}\mathbf{k}_{j}^{t}\right) + \mathbf{b}^{t}\right) + \mathbf{w}_{2}^{t}\mathbf{k}_{i}^{t}\right),t = \{1,2\} , \quad (6)\]  
+
+where \(j\in \mathcal{N}(k_{i})\) is the neighborhood of \(k_{i}\) , \(\mathbf{w}\) is the learnable transformation matrix, and \(\sigma (\cdot)\) ) is a non- linear activation function (i.e., ReLU).  
+
+After obtaining the synthon embeddings and the LG embeddings, we can directly perform discriminate the candidate LG to be attached to a synthon. As suggested by MLGL- MP (2022) \(^{30}\) , we measure the proximities between a given synthon \(s_{u}\) and a given LG \(k_{i}\) as follows:  
+
+\[\hat{y}_{ui} = \mathbf{s}_u(\mathbf{k}_i)^T. \quad (7)\]  
+
+The proximity \(\hat{y}_{ui}\) is the predicting score of the given synthon attaching the \(i\) - th LG among the LG set \(\mathcal{K}\) , and reflects how possibly \(s_{u}\) is attached to \(k_{i}\) .  
+
+However, such a direct proximity measure would be senseless since the synthon embedding space and the LG embedding space are of different vector spaces. To tackle this issue, we design an adapter to map \(\{\mathbf{s}_u\}\) into \(\{\mathbf{k}_i\}\) . The adapter can be implemented by an MLP containing an input layer, a hidden layer, and an output layer. Thus, the final compound representation feature is defined as \(\mathbf{s}_u^* = \mathrm{MLP}(\mathbf{s}_u) \in \mathbb{R}^s\) , where \(s\) is the dimension of \(k_{i}\) .  
+
+Last, the mean square error (MSE) loss function is used when training LGP as follows:  
+
+\[l_{group} = \frac{1}{2M*|\mathcal{K}|}\sum_{u = 1}^{2M}\sum_{i = 1}^{|\mathcal{K}|}(\hat{y}_{ui} - y_{ui})^2, \quad (8)\]  
+
+where \(y_{ui} \in \{0,1\}\) is the true label indicating whether or not a synthon \(s_{j}\) is attached by an LG \(k_{i}\) , \(\hat{y}_{ji}\) is the corresponding score output by LGP, \(M\) is the number of all the training molecules, and \(2M\) represents the number of their synthons.  
+
+### 2.4 Training Loss and Testing  
+
+To train the whole Retro- MTGR model, we combine the three abovementioned loss functions w.r.t. tasks into a linear joint as follows  
+
+\[Loss = w_{1}*l_{bond} + w_{2}*l_{contrast} + w_{3}*l_{group}, \quad (9)\]  
+
+where \(\sum_{i = 1}^{3}w_{i} = 1\) are normalized hyperparameters to adjust task weights.  
+
+Note that Retro- MTGR in the testing should remove the AEE module since it cannot be available in the scenario.  
+
+## 3. Result and discussion  
+
+### 3.1 Dataset and parameter settings  
+
+As popularly used in existing methods \(^{31}\) , we collected the benchmark dataset from the USPTO- 50K dataset, which was derived from an open- source patent database
+
+<--- Page Split --->
+
+containing 50,016 atom- mapped reactions \(^{32}\) . In this work, by discarding modification- like chemical reactions, we only extracted the chemical reactions where one target molecule is synthesized from two reactants. As a result, our dataset contains 30,565 reaction entries, which are divided into 7 categories according to reaction type (Table 1).  
+
+In the Atom Encoder, as suggested in existing methods \(^{4}\) , each atom was initially represented by a 28- dimensional (28- d) atom feature vector \((\mathbf{x}_i \in R^{28})\) , including Atom Type (23- d), the number of Hydrogens (1- d), the number of linking neighbors of atom (Degree, 1- d), Is Aromatic (1- d), Formal Charge (1- d), as Atomic Mass (1- d). See also Table 2 for details. Due to the one- hot coding of Atom Type, the initial atom representation \(\mathbf{x}_k\) is sparse. To avoid the vanishing gradient problem \(^{28}\) , an extra three- layer MLP maps it into its dense form \((\mathbf{x}_k \in R^{32})\) . We empirically set 64 and 32 neurons in its hidden layer and output layer respectively. Moreover, bonds are represented as a binary adjacent matrix \(\mathbf{E}\) , of which \(e_{ij} = 1\) indicates the occurring bond between two atoms \((a_i\) and \(a_j)\) , and no bond otherwise. Both \(\{\mathbf{x}_k\}\) and \(\mathbf{E}\) are input into a two- layer MPNN to obtain atom embeddings having the same dimensions as those of \(\{\mathbf{x}_k\}\) .  
+
+In the RCP module, the MLP accounting for reaction center identification contains also an input layer, a hidden layer, and an output layer. There are 33 neurons in the input layer, where 32 neurons are responsible for the resulting embeddings from the bond- level readout and the last one takes charge of the bond energy. The number of neurons is empirically set to 16. The unique neuron in the output layer accounts for the confidence score of being a reaction center.  
+
+In the LGP module, the nodes in the leaving group co- occurrence graph (LGCoG) are initially represented as \(n\) - dimensional one- hot coding vectors \(\{\mathbf{k}_i^0 \in \mathbb{R}^{|\mathcal{K}|}\}\) , where \(n = |\mathcal{K}|\) is the cardinality of the LG set. Then, they are mapped into LG embeddings \(\{\mathbf{k}_i \in \mathbb{R}^{|\mathcal{K}|}\}\) by another two- layer MPNN without dimensional change. On the other side, the adapter maps the synthon embedding space (the concatenation of 32- d atom embeddings and 33- d bond embeddings) into the LG embedding space. It is implemented by a three- layer MLP, which contains an input layer accounting for synthon embeddings \((\mathbf{s}_u \in \mathbb{R}^{65})\) , a hidden layer having empirically 128 neurons, and an output layer having \(n\) neurons respectively. Thus, \(\mathbf{s}_u^* = \mathrm{MLP}(\mathbf{s}_u) \in \mathbb{R}^n\) , where \(n\) (i.e., \(|\mathcal{K}|\) ) is scenario- specific since the type- known scenario and the type- unknown scenario have different types of LGs.
+
+<--- Page Split --->
+
+
+Table 1. Dataset overview   
+
+<table><tr><td>Reaction type</td><td>Reaction name</td><td>No. of examples</td></tr><tr><td>1</td><td>heteroatom alkylation and arylation</td><td>14188</td></tr><tr><td>2</td><td>acylation and related processes</td><td>10509</td></tr><tr><td>3</td><td>C-C bond formation</td><td>4378</td></tr><tr><td>4</td><td>protections</td><td>144</td></tr><tr><td>5</td><td>oxidations</td><td>142</td></tr><tr><td>6</td><td>functional group interconversion (FGI)</td><td>986</td></tr><tr><td>7</td><td>functional group addition (FGA)</td><td>218</td></tr><tr><td>Total</td><td>/</td><td>30565</td></tr></table>  
+
+Table 2.Atom attributes   
+
+<table><tr><td>Feature</td><td>Description</td><td>Dimension</td></tr><tr><td>Atom type</td><td>Cl, N, P, Br, B, S, I, F, C, O, ... (one-hot)</td><td>23</td></tr><tr><td>Number of H</td><td>Integer</td><td>1</td></tr><tr><td>Degree</td><td>Integer</td><td>1</td></tr><tr><td>Is Aromatic</td><td>True or False (binary)</td><td>1</td></tr><tr><td>Formal charge</td><td>Integer</td><td>1</td></tr><tr><td>Atomic Mass</td><td>Integer</td><td>1</td></tr></table>  
+
+### 3.2 Comparison with state-of-the-art methods  
+
+To evaluate the effectiveness of Retro- MTGR, we compared it with five state- of- the- art single- step retrosynthesis methods, including two sequence- to- sequence translation methods (including seq2seq \(^{16}\) and SCPOP \(^{15}\) ), two graph- to- seq translation methods (including MEGAN \(^{23}\) and Graph2SMILES \(^{22}\) ), and one discriminative method (i.e., G2Gs \(^{26}\) ). They are briefly summarized below.  
+
+- Seq2seq: It is the first sequence-2-sequence translation method, which utilizes attention-enhanced LSTMs to convert the target molecule to its reactants in the form of SMILES under an encoder-decoder architecture \(^{16}\) .- SCROP: It is a transformer-based method with the aid of an extra syntax post-checker to guarantee the chemical validity of generated reactants \(^{15}\) .- MEGAN: Its outputs reactants by a queue of leaned graph edits (chemical structure modification) on the input molecule graph step by step under a GNN-based auto-regressive architecture \(^{23}\) .- Graph2SMILES: It is also an auto-regressive model, which utilizes a directed MPNN to capture atom representations with the extra enhancement of global
+
+<--- Page Split --->
+
+attention encoding<sup>22</sup>.  
+
+- G2Gs: It is a two-step GNN-based discriminative model, which first encodes bonds to find the reaction center by relational GCNs, and then builds a variational graph model to infer leaving groups to be attached on synthons<sup>26</sup>.  
+
+For a fair comparison, we run ten- fold cross- validation (10- CV) as suggested by these methods<sup>33</sup>. In detail, the dataset was randomly and equally split into 10 subsets, of which each subset (10% samples) was taken as the testing set and the remaining subsets (90% samples) were taken as the training set. Such a 10- CV was repeated 50 times under different random seeds. The average performance over 50 rounds of cross- validations was reported to measure the retrosynthesis prediction of Retro- MTGR. Moreover, the top- k accuracy (e.g., Top- 1, Top- 3, and Top- 5) was adopted as the measuring metric in 10- CV. It is defined as the ratio of the number of correctly predicted target molecules to the total number of target molecules, where a target molecule is correctly predicted if its correct reactants are found among the top- k predicted candidates.  
+
+In addition, two retrosynthesis scenarios, Reaction Type Known (RTK) and Reaction Type Unknown (RTU), were considered, as these methods performed. In the first scenario RTK, we are required to perform a retrosynthesis for a molecule while being given its possible reaction type. In the second one RTU, we have no information about its potential reaction type. Usually, RTU is more practical but difficult than RTK. Besides, since the prediction in RTK is specific to reaction types, we reported only the average performance over those reaction types in Table 3 and listed the detailed results in Supplementary (Section 1).  
+
+The comparison results demonstrate that our Retro- MTGR achieves the best prediction and is significantly superior to other state- of- the- art methods over two testing scenarios. The results also validate that RTU is more difficult than RTK since extra type information in RTK helps the prediction. In detail, it achieves 69.1%, 89.2%, and 91.0% accuracies in the case of RTK while achieving 57.3%, 81.0%, and 86.5% respectively in the case of RTU in terms of Top- 1, Top- 3, and Top- 5. Remarkably, Retro- MTGR achieves 5\~7% improvements in the case of RTK while achieving 4\~12% improvements in RTU over Top- 1, Top- 3, and Top- 5 accuracies respectively.
+
+<--- Page Split --->
+
+
+Table3. Top-k accuracy for retrosynthesis prediction on USPTO.   
+
+<table><tr><td rowspan="3">Methods</td><td colspan="6">Top-k Accuracy (%)</td></tr><tr><td colspan="3">Reaction Type Unknown</td><td colspan="3">Reaction Type Known</td></tr><tr><td>1</td><td>3</td><td>5</td><td>1</td><td>3</td><td>5</td></tr><tr><td rowspan="2">Sequence-to-Sequence</td><td>Seq2seq</td><td>35.0±2.5</td><td>41.7±2.1</td><td>56.5±3.5</td><td>43.4±2.7</td><td>60.2±2.8</td></tr><tr><td>SCROP</td><td>44.6±3.5</td><td>63.3±2.2</td><td>66.8±3.2</td><td>58.1±2.5</td><td>75.6±3.1</td></tr><tr><td rowspan="2">Graph-to-Seq</td><td>MEGAN</td><td>46.8±5.9</td><td>72.7±5.1</td><td>75.4±5.0</td><td>57.7±6.0</td><td>83.1±5.6</td></tr><tr><td>Graph2Smiles</td><td>51.7±3.5</td><td>65.9±1.7</td><td>72.8±2.4</td><td>61.7±3.2</td><td>80.3±2.5</td></tr><tr><td rowspan="2">Discriminative</td><td>G2Gs</td><td>53.0±1.2</td><td>70.5±1.2</td><td>74.2±0.9</td><td>59.0±0.9</td><td>83.4±1.0</td></tr><tr><td>Retro-MTGR</td><td>57.3±1.4</td><td>81.0±0.8</td><td>86.5±0.7</td><td>69.1±0.6</td><td>89.2±1.1</td></tr></table>  
+
+### 3.3 Ablation studies  
+
+In this section, we investigated how crucial components of our Retro- MTGR contribute to the retrosynthesis prediction by ablation studies. We made three variants of our original model by masking one block of Retro- MTGR in turn. First, we removed the AEE module (denoted as w/o AEE). Secondly, we discarded bond energies in bond embeddings (denoted as w/o BE). Last, we deleted the leaving group co- occurrence graph (LGCoG) in the LGP module and used one- hot coding as the initial representations of leaving groups (denoted as w/o LGCoG).  
+
+As the ablation comparison illustrates, the superiority of Retro- MTGR to all its variants demonstrates that all of the AEE module, the bond energy, and the leaving group graph play significant roles in the retro- synthesis prediction in the case of both unknown and known reaction types (Figure 2). Specifically, the AEE module plays the most important role. For example, Retro- MTGR with the AEE module improves the Top- 1, Top- 3, and Top- 5 accuracies by \(5.9\%\) , \(8.2\%\) , and \(3.1\%\) respectively in the case of unknown reaction type. The result indicates that the AEE module can enhance bond embeddings in finding the reaction centers because it utilizes chemical structural commonness and differences between a molecule and its synthons.  
+
+The bond energy block also provides an untrivial contribution to bond embeddings. For example, Retro- MTGR with bond energies improves the Top- 1, Top- 3, and Top- 5 accuracies by \(1.6\%\) , \(4.9\%\) , and \(3.0\%\) respectively when reaction types are unknown. The underlying reason is that bonds having high bond energies are usually ordinary bonds but not reaction centers. See also Section 3.4.1 for details.  
+
+The LGCoG provides a significant contribution to leaving group identification. For instance, Retro- MTGR with the LGCoG improves the Top- 1, Top- 3, and Top- 5 accuracies by \(3.3\%\) , \(3.7\%\) , and \(1.8\%\) respectively when reaction types are unknown. The essential
+
+<--- Page Split --->
+
+reason for such improvements is that the captured LG dependences enrich LG representations when identifying LGs for synthons. Details can be found in Section 3.4.2.  
+
+In general, all of the AEE module, the bond energy, and the leaving group co- occurrence graph play indispensable roles in retrosynthesis prediction. More detailed investigations in the next section indicate why they work.  
+
+![](images/Figure_2.jpg)
+
+<center>Figure 2. Ablation comparison. Compared with the three variants (red, green and purple bars), Retro-MTGR (blue bars) achieves the best retrosynthesis prediction in terms of top-1, top-3, and top-5 in the case of both unknown and known reaction types. </center>  
+
+### 3.4 Retrosynthesis rule discovery  
+
+In this section, we shall attempt to uncover retrosynthesis rules by Retro- MTGR in two interpretable views, bond view and leaving group view. First, we shall investigate three questions to reveal why a bond can be the reaction center. Furthermore, we shall explore two questions to indicate what leaving groups are appropriate to given synthons.  
+
+#### 3.4.1 Bond view  
+
+To interpret why a bond can be the reaction center, we considered three bond- derived questions as follows.  
+
+## (1) Can bond energies determine the reaction center in a molecule alone?  
+
+For a chemical bond, its bond energy (i.e., the minimum energy to break it down) measures its stableness<sup>34</sup>. The larger, the stabler, the more difficult to be synthesized from the point of view of chemical retrosynthesis. Thus, we assume that the reaction center is of low bond- energy bond.  
+
+To validate it, we made a statistical distribution of bond energies across all the bonds of molecules in a histogram (Figure 3). Note that we considered only the theoretical breaking energy of each chemical bond, but not considering the influence of its neighboring bonds or near atoms. The bonds were sorted into 20 equally spaced bins along the axis of bond energy between the
+
+<--- Page Split --->
+
+minimum and maximum energy values (kJ/mol). Due to the number difference between reaction centers and ordinary bonds, the heights of bins (i.e., the number of bonds falling in the bins) were normalized by the total number of bonds for convenient comparison.  
+
+As illustrated, the majority of bonds fall into four bins, [270,315], [315, 360], [450- 495], and [495- 540]. Specifically, the bond energies of reaction centers are usually located in the lower range of bond energy (i.e., \(95.26\%\) having bond energy \(< 360 \mathrm{kJ / mol}\) ). In contrast, \(45.33\%\) of ordinary bonds have bond energy \(< 360 \mathrm{kJ / mol}\) and \(54.65\%\) have bond energy \(> = 360 \mathrm{kJ / mol}\) . Thus, a naïve decision can be made that bonds having bond energy \(> 360 \mathrm{kJ / mol}\) are usually ordinary bonds. Such a finding can be used to filter out the ordinary bonds having large breaking energies in the process of reaction center identification. This is why bond energies have a significant contribution to finding the reaction center. However, bond energies can NOT determine the reaction center in a molecule alone, since there are still a large number of ordinary bonds (45.33 %) overlapping with reaction centers in the case of having low breaking energies \(< 360 \mathrm{kJ / mol}\) . This issue can be further investigated by the answer to the second question.  
+
+![](images/Figure_3.jpg)
+
+<center>Figure 3. Breaking energy distribution of chemical bonds. The X-axis shows the bond energy (0-900) divided into 20 intervals. The Y-axis shows the frequency of chemical bonds falling in different bins. As illustrated, bonds having bond energy \(> 360 \mathrm{kJ / mol}\) are usually ordinary bonds. However, \(45.33\%\) of ordinary bonds overlap with reaction centers in the case of breaking energies \(< 360 \mathrm{kJ / mol}\) . </center>  
+
+(2) What is the underlying chemical rule captured by bond embeddings such that reaction centers can be distinguished from ordinary bonds?  
+
+One of the core contributions of our model (Retro- MTGR) is just the discrimination of reaction centers from ordinary bonds in the case of low bond energy. Since bond embedding representations (Formula 2) characterize bond features based on molecule graph topology, we
+
+<--- Page Split --->
+
+utilized them to exhibit the difference between reaction centers and ordinary bonds. Principal component analysis (PCA) was used to visualize bonds in 2- dimensional space, where each point represents a bond.  
+
+Such a bond space was rendered in three maps (Figure 4).  
+
+- The first map shows a clear separation between reaction centers (red points) and ordinary bonds (blue points), except for a small overlapping (Figure 4-a). Such a separation demonstrates that our model can characterize the difference between reaction centers and ordinary bonds well. More importantly, both reaction centers and ordinary can be split into communities, which are strongly specific to bond types.  
+
+- The second map indicates that bond communities are consistence with bond types (Figure 4-b). Some bonds are always ordinary bonds, such as c-c (an aromatic bond linking two carbon atoms). More importantly, it is remarkable that some bonds having the same types (e.g., C-C, C-O, and C-N.) may occur in different communities, which belong to reaction centers and ordinary bonds respectively. The underlying reason is investigated in the answer to the third question.  
+
+- The last map illustrates the distribution of bond energies in terms of energy bins (Figure 4-c). As observed, bonds with large breaking energies (two bins) are almost of ordinary bonds. More importantly, reaction centers and ordinary bonds having low breaking energies can be clearly distinguished. With the consideration of bond types, the bonds in the energy bin [270-315] mainly include C-N (a bond linking a carbon atom to a nitrogen atom) and C-Br (a bond linking a carbon atom to a bromine atom), while those in the energy bin [315-360] mainly include C-C (a bond linking two carbon atoms) and C-O (a bond linking a carbon atom to an oxygen atom).
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Figure 4. Bond space. (A). Reaction centers and ordinary bonds. Red dots indicate reaction centers while blue dots indicate ordinary bonds. (B). Bond types. Different colors represent different bonds. (C). Bond energies. Different colors indicate different bond energy bins (kJ/mol). In terms of chemical symbols, C stands for carbon atoms, c is for carbon atoms in aromatic bonds, and C' is for carbon atoms in general rings. </center>
+
+<--- Page Split --->
+
+Moreover, N stands for nitrogen atoms, n is for nitrogen atoms in aromatic bonds, and N' represents nitrogen atoms in general rings. In addition, O is for oxygen atoms, O' represents oxygen atoms in general rings, and S is for sulfur atoms. Four specific symbols, including '\\~', '\\~', '='', and '#', denotes aromatic, single, double, and triple bonds respectively.  
+
+## (3) Why can a bond be the reaction center in a molecule but cannot be in another one?  
+
+Considering that bond energy can NOT determine the reaction center in a molecule alone, our model (Retro- MTGR) leverages molecule graph topologies to capture the differences between reaction centers and ordinary bonds, even having the same bond types. We investigated what chemical rule hidden is captured by atom/bond embeddings. It is anticipated that the inherent law helps identify reaction centers and ordinary bonds, especially in the case of both common bond type and similar bond energy.  
+
+Our investigation was inspired by the chemical knowledge that the electrical property (denoted as \(p\) ) of an atom in a molecule is determined by the conjugation effect of the motion of its electrons as well as the union of its spatial neighboring atoms. When showing an attractive effect on electrons, the atom is considered an electron- withdrawing atom. On the contrary, it is called an electron- donating atom. Since it is difficult to quantify atom electrical properties due to complicated inter- atom influences, we first proposed a qualitative manner to label their strength and weakness. Then, by enumerating atom- centered substructures, we found 356 substructures (Section 2 in Supplementary), which are categorized into four groups in terms of electrical property strength. In detail, the atoms showing strong electron- withdrawing/donating properties are labeled as \(p + + |p - -\) respectively. Meanwhile, those atoms exhibiting weak electron- withdrawing/donating properties are marked as \(p + /p -\) respectively. As a result, the atom pairs forming bonds show ten possible pairs of electrical properties (e.g., \((p + + |p - - )\) , \((p + |p - )\) ) in total. Last, we counted the percentages of all types of electrical property pairs in the case of both reaction centers and ordinary bonds (Figure 5). The result illustrates that the atom pairs forming reaction centers have dominant opposite electrical properties (97.0% with ' \(p + |p - '\) , ' \(p + |p - - '\) , ' \(p + + |p - '\) , and ' \(p + + |p - - '\) pairs) while those pairs forming ordinary bonds have same or similar electrical properties (80.1%). In fact, some ordinary bonds having opposite electrical properties are also reaction centers in the deeper steps of retrosynthesis. See also Section 3.5 Case Study. Thus, we conclude that the pair of an electron- withdrawing atom and an electron- donating atom tends to form a reaction center.  
+
+In summary, our answers to these questions demonstrate that our Retro- MTGR can capture the underlying matching rule of why a bond can be the reaction center by embedding molecule topologies.
+
+<--- Page Split --->
+![](images/Figure_5.jpg)
+
+<center>Figure 5. Electrical property distribution of chemical bonds. The X-axis denotes pairwise electrical property patterns of atom pairs forming bonds. Its left zone lists six patterns of same/similar electrical properties while its right zone lists four patterns of opposite electrical properties. The Y-axis indicates the frequencies of electrical property patterns. The member atoms of ordinary bonds usually have same or similar electrical properties (80.1%), whereas those in reaction centers tend to have opposite electrical properties (97.0%). </center>  
+
+#### 3.4.2 Leaving Group view  
+
+To dig out what leaving groups are appropriate to given synthons, we leveraged the leaving group co- occurrence graph (LGCoG) to answer two LG- derived questions as follows.  
+
+#### (1) Whether are two leaving groups associated when accounting for a pair of synthons derived from the same molecule?  
+
+We sought its answer in two aspects. First, we followed the answer to the third question in the previous section that retrosynthesis site atoms in synthons always have opposite electrical properties. Analogously, when two leaving groups account for a pair of synthons derived from the same molecule, they are also supposed to have opposite electrical properties based on the electrical matching rule between a synthon and its LG. We validated this assumption by labeling the electrical properties of LGs (Figure 6- A) in a similar manner as that in Section 3.4.1. As counted, 94.3% of LG pairs (28823/30565) have opposite electrical properties.  
+
+Then, we considered the occurrence of LG pairs. The co- occurrence is indicated by the edge thickness in the graph (Figure 6- A). As calculated, we found that the LG pairs consisting of two simple groups (e.g., H, OH, or halogens) usually occur frequently. Especially, the pairs 'H, OH' (27.8%), 'H, Cl' (27.4%), 'H, Br' (12.9%), and 'H, I' (8.1%) are the most frequent LG pairs. Moreover, the LG pairs including a simple group and a chemical substructure occur in a low frequency, such as 'Br, CC1(C)OBOCl(C)C' (1.8%). Few LG pairs (0.386%) are composed of chemical substructures only, such as 'O=S(=O)(O)C(F)(F)F- B(OH)2, O=S(=O)(O)C(F)(F)F- CC1(C)OBOCl(C)C'. More details can be found in Supplementary
+
+<--- Page Split --->
+
+(Section 3).  
+
+Thus, two leaving groups accounting for a pair of synthons are associated.  
+
+## (2) Whether is a leaving group specific to a reaction type?  
+
+After taking a close look at the topology of the graph, we found a few LGs having many partner LGs and many LGs having only one partner. The former is always of simple groups (e.g., H, OH, or halogens), while the latter is usually chemical substructures. In addition, the remaining LGs have a small number of partners. To dig out the underlying reason, we made an investigation by counting LG occurrences according to reaction types (Section 4 in Supplementary).  
+
+The investigation reveals interesting knowledge that LGs can be split into two categories according to the number of reaction types they attending in.  
+
+The first category, named 'reaction- common' LGs, contains the LGs appearing in equal to or more than the half number of reaction types (i.e., \(> = 4\) ). They are usually of simple LGs, and occur frequently and have many matching partner LGs. In details, 'H', 'OH', 'Cl', and 'Br' appear in 7, 7, 6 and 5 categories, respectively. Especially, 'H' attends in almost all the reactions while occurring 27623 times and having 37 kinds of matching partner LGs (denoted by the node degree in the graph). Moreover, halogens including 'Cl', 'Br', and 'I', have many common partner LGs. The possible reason is that they have similar chemical properties, which suggests a potential replacement between them.  
+
+The second category, named 'reaction- specific' LGs, includes the LGs occurring only in specific types of reactions. They are of usually chemical substructure groups, such as '- CC', '- OCC(Cl)(Cl)Cl', and '- OCC(F)(F)F'. We found that reaction- specific LGs are surrounding reaction- common LGs or clustered together (Figure 6- B). In addition, some LGs involving in the same type of reaction have similar chemical structures, such as the pair of 'CCOC(=O)O' and 'CC(C)(C)OOC(=O)O' in Type 2 reaction (0.783) and the pair of 'B1OCC01' and 'B1OCCCO1' (0.754) in Type 3 reaction, where the similarity is calculated by MACCS fingerprints in terms of Jaccard similarity. Again, the chemical structure similarity of LGs implies their possible substitution, which is determined by the cost or the reaction condition available in synthesis reaction routes.  
+
+In general, the co- occurrence graph of leaving groups contains rich retrosynthesis information, including the inter- associations between LGs and their reaction specificity. Our Retro- MTGR can capture such rich information to enhance the retrosynthesis prediction. The answers in this section dig out why the graph contributes to the prediction.
+
+<--- Page Split --->
+![](images/Figure_6.jpg)
+
+<center>Figure 6. Leaving Group view. (A) Electrical property map. LGs having positive and negative electrical properties are rendered by red and blue. In total, \(94.3\%\) of LG pairs have opposite electrical properties. (B) Reaction type map. LGs occurring in single reaction types, multiple reaction types ( \(\leq 3\) ) and many types ( \(\geq 4\) , reaction-common) are highlighted in different colors. </center>  
+
+### 3.5 Case Study  
+
+To evaluate the ability of our Retro- MTGR in the real scenario of retrosynthesis prediction, we collected two drugs (i.e., Sonidegib \(^{35}\) and Acotiamide \(^{36}\) ), not included in our dataset, as the studying cases. We inferred their retrosynthesis routes by our Retro- MTGR and then validated
+
+<--- Page Split --->
+
+them by chemical assays respectively.  
+
+The two drugs we selected are briefly summarized as follows. The first drug, Sonidegib, is a Hedgehog signaling pathway inhibitor (via smoothened antagonism), which was developed as an anticancer agent by Novartis and approved by the FDA in 2015 for the treatment of basal cell carcinoma. Now it is commonly used for the treatment of locally advanced recurrent basal cell carcinoma (BCC) following surgery and radiation therapy, or in cases where surgery or radiation therapy are not appropriate (DrugBank ID: DB09143) \(^{35}\) . The second drug, Acotiamide, being an investigational drug, was designed and developed for the treatment of functional dyspepsia (FD) \(^{36}\) (DrugBank ID: DB12482). It works as a novel upper gastrointestinal (GI) motility modulator and stress regulator by improving upper gastrointestinal functions in both the stomach and esophagus.  
+
+Since Retro- MTG is a single- step retrosynthesis prediction model, we iteratively applied it to infer a complete retrosynthesis route for a given drug molecule. In the first iteration, Retro- MTG split the complete molecule into two synthons under the top- 1 criterion (the first candidate reaction center), which were further turned into smaller intermediate molecules by appending appropriate leave groups. Intermediate molecules were then split into smaller ones by Retro- MTG in a similar way unless the intermediate molecules are reactants, which can be easily bought in the market.  
+
+As shown in Figure 7, the predicted retrosynthesis route of Sonidegib illustrates that it (marked as '1') can be split into two intermediate molecules (marked as '2' and '3'). Further, they are split into two pairs of reactants respectively, where one pair is (marked as '4' and '5') while another is (marked as '6' and '7'). In each retrosynthesis step, the attention scores of bonds are labeled and the highest one (top- 1) is considered as the reaction center. During predicting the retrosynthesis route, both reaction centers and leaving groups are corrected predicted (Figure 7- A). Furthermore, according to the prediction, a series of chemical synthesis reactions starting from reactants ('4', '5', '6', and '7') were performed (Figure 7- B) to validate the predicted retrosynthesis route. Remarkably, due to the potential intra- reaction among '6' molecules triggered by their own Chlorine (- Cl) and amino group (- NH2), the expected reaction between '6' and '7' would generate less amount of '3' molecules. To guarantee a high production of '3' in the real synthesis, we converted '6' to '8' by a nitration reaction, which alters the amino group to a nitro group (- NO2). Then, the reaction of '8' and '7' produced '9', of which the nitro group was further converted back into the amino group in a reduction reaction to obtain '3' with the desired production. Meanwhile, the reaction of '4' and '5' generated '2'. Finally, we performed an amidation reaction by combining '2' and '3' to form the product molecule Sonidegib ('1').  
+
+Similarly, the retrosynthesis route of Acotiamide (marked as '10') is correctly predicted in terms of both reaction centers and leaving groups, and validated by chemical synthesis reactions
+
+<--- Page Split --->
+
+as well. Specifically, the first retrosynthesis step generates an intermediate molecule ('11') and a reactant ('12'). Then, the former is split into two reactants ('13' and '14') in the second step (Figure 7- C). Since there is also an intra-reaction issue in '14', a similar strategy was adopted to guarantee the final production of '10'. In brief, the carboxylic acid group (- OH) of '14' was altered to a methoxy group (- OCH3) so as to generate '15' by an esterification reaction. As the substitution of '14', '15' was put together with '13' to generate a new intermediate molecule '16', which was sequentially hydrolyzed to intermediate molecule 11 by both NaOH and HCl. Finally, a similar amidation reaction combining '11' and '12' was performed to form Acotiamide ('10') (Figure 7- D).  
+
+To summarize, Retro- MTGR shows that its inspiring retrosynthesis prediction is significantly consistent with chemical assays. Thus, it can provide clear guidance for retrosynthesis route planning with extra reaction conditions.  
+
+![](images/Figure_7.jpg)
+
+<center>Figure 7. Predicted retrosynthetic routes and real chemical synthetic routes. A. Retrosynthesis prediction of Sonidegib. Sonidegib ('1') is split at the reaction center into intermediate molecules '2' and '3', where the highest </center>
+
+<--- Page Split --->
+
+attention score generated by our Retro- MTGR is highlighted. Then, '2' and '3' are further decomposed into four reactants ('4', '5', '6' and '7') respectively according to their reaction centers. **B. Chemical synthesis of Sonidegib.** First, the Suzuki cross- coupling reaction between '4' and '5' was performed to obtain intermediate molecule '2'. Meanwhile, '6' is converted to '8' with the help of m- CPBA (m- Chloroperbenzoic Acid) for preventing its intra- reaction. Then, '8' is combined with '7' in the presence of DIEA (N, N- diisopropylethylamine) to form another intermediate molecule '9', which is further reduced by H₂ in the presence of Pd/C to generate '3'. In the last step, the coupling of compound '2' with '3' in the presence of HATU (2-(7- Azabenzotriazol- 1- yl)- N, N, N', N'- tetramethyluronium hexafluorophosphate) and DIEA (N,N- Diisopropylethylamine) in DMF (N,N- Dimethylformamide) generates Sonidegib ('1'). **C. Retrosynthesis prediction of Acotiamide.** Acotiamide ('10') is split at the reaction center into intermediate molecules '11' and reactant '12', where the highest attention score generated by our Retro- MTGR is highlighted as well. Then, '11' is further decomposed into two reactants ('13' and '14') respectively according to their reaction centers. **D. Chemical synthesis of Acotiamide.** First, reactant '14' is esterified with methanol to form '15'. Then, the coupling compound 13 with 15 in the presence of HATU and DIEA obtains the intermediate molecule '16', which is sequentially hydrolyzed to another intermediate molecule '11' by NaOH and HCl. Finally, the coupling between the intermediate molecule '11' and the reactant '12' in the presence of HATU and DIEA in DMF generates the molecule of Acotiamide ('10'). More details (e.g., reaction conditions and instruments) about the chemical synthesis reactions of these two drugs can be found in Supplementary (Section 5). Note: DIEA works as a catalyzer in the reduction reaction or a promoter in the aromatic nucleophilic substitution of N- hydrophosphoramide. HATU works as a popular condensation reagent in promoting amide bond formation by activating carboxyl groups. m- CPBA is a strong oxidizing agent widely used in organic synthesis. Pd/C is also a kind of popular catalyst in hydrogenation reduction.  
+
+## 4 Conclusions  
+
+Aiming at a well- interpretable discriminative model in terms of chemical synthesis mechanism, this paper elaborates a novel multi- task graph representation learning framework of retrosynthesis prediction (Retro- MTGR). Based on molecule graphs, three related tasks are considered in Retro- MTGR simultaneously, where two major supervised discriminative tasks account for recognizing reaction centers and identifying leaving groups respectively, and an auxiliary self- supervised task accounts for generating better atom embeddings.  
+
+The comparison with various state- of- the- art methods demonstrates the overall superiority of Retro- MTGR in terms of cross- validation prediction performance in both reaction- type unknown and reaction- type known scenarios. Furthermore, the ablation study demonstrates its contributions as follows. First, it significantly enhances atom embedding by leveraging chemical structural redundancy and differences between a molecule and its synthons. Then, it uncovers that bond energies can partially boost bond
+
+<--- Page Split --->
+
+embeddings due to the difference between ordinary bonds and reaction centers in the case of high bond energy. Last, it utilizes LG co-occurrences to enrich LG representations.  
+
+Furthermore, multiple comprehensive investigations validate the chemical synthesis interpretability of Retro- MTGR by answering two questions: why a bond can be the reaction center or not, and what leaving groups are appropriate to given synthons. The answers demonstrate that Retro- MTGR can capture and illustrate the underlying chemical synthesis rules as follows: (1) a bond having high- breaking energy (>=360 kJ/mol) is usually an ordinary bond; (2) aromatic bonds (c-c) between carbon atoms are always ordinary bonds, while other types of bonds can be either reaction centers or ordinary bonds; (3) a bond is the reaction center in a molecule if its member atoms tend to have opposite electrical properties, otherwise an ordinary bond;(4) accordingly, leaving group (LG) pairs in reaction centers usually have opposite electrical properties and occurrence- dominant LG pairs always consisting of two simple groups (e.g., H, OH, or halogens); (5) individual LGs can be categorized into reaction- common LGs (simple groups) and reaction- specific LGs (chemical substructures).  
+
+Finally, the practical ability of Retro- MTGR is evaluated by two novel drugs. The results reveal that the inferred retrosynthesis routes by Retro- MTGR are significantly consistent with those achieved by chemical synthesis assays.  
+
+In summary, our Retro- MTGR can provide prior guidance for retrosynthesis route planning. We believe that its extension with the integration of extra synthetic factors (e.g., reaction yield, conditions, and reagents) can be a complete retrosynthesis route planning in the coming future.  
+
+## Acknowledgements:  
+
+The authors would like to thank anonymous reviewers for suggestions that improved the paper.  
+
+## Funding  
+
+This work was supported by the National Nature Science Foundation of China [61872297], the Shaanxi Province Key R&D Program [2023- YBSF- 114], and the CAAI- Huawei Mind Spore Open Fund [CAAIXSJLJJ- 2022- 035A].  
+
+## References  
+
+1. Zhu, H. Big Data and Artificial Intelligence Modeling for Drug Discovery. Annu Rev Pharmacol Toxicol 60, 573-589 (2020).  
+2. Zhao, P., et al. Targets preliminary screening for the fresh natural drug molecule based on Cosine-correlation and similarity-comparison of local network. J Transl Med 20, 67 (2022).  
+3. Li, J.N., Yang, G., Zhao, P.C., Wei, X.X. & Shi, J.Y. CProMG: controllable protein-oriented
+
+<--- Page Split --->
+
+659 molecule generation with desired binding affinity and drug- like properties. 660 Bioinformatics 39, i326- i336 (2023). 661 4. Du, B.- X., Xu, Y., Yiu, S.- M., Yu, H. & Shi, J.- Y. MTGL- ADMET: A Novel Multi- task Graph 662 Learning Framework for ADMET Prediction Enhanced by Status- Theory and Maximum 663 Flow. in Research in Computational Molecular Biology: 27th Annual International 664 Conference, RECOMB 2023, Istanbul, Turkey, April 16- 19, 2023, Proceedings 85- 103 665 (Springer, 2023). 666 5. Gupta, R., et al. Artificial intelligence to deep learning: machine intelligence approach for 667 drug discovery. Mol Divers 25, 1315- 1360 (2021). 668 6. Vatansever, S., et al. Artificial intelligence and machine learning- aided drug discovery in 669 central nervous system diseases: State- of- the- arts and future directions. Med Res Rev 670 41, 1427- 1473 (2021). 671 7. Lee, A.A., et al. Molecular Transformer unifies reaction prediction and retrosynthesis 672 across pharma chemical space. Chem Commun (Camb) 55, 12152- 12155 (2019). 673 8. Dong, J., Zhao, M., Liu, Y., Su, Y. & Zeng, X. Deep learning in retrosynthesis planning: 674 datasets, models and tools. Brief Bioinform 23(2022). 675 9. Coley, C.W., Green, W.H. & Jensen, K.F. Machine Learning in Computer- Aided Synthesis 676 Planning. Acc Chem Res 51, 1281- 1289 (2018). 677 10. Yan, M. & Baran, P.S. Drug discovery: Fighting evolution with chemical synthesis. Nature 678 533, 326- 327 (2016). 679 11. Lovric, M., Molero, J.M. & Kern, R. PySpark and RDKit: Moving towards Big Data in 680 Cheminformatics. Mol Inform 38, e1800082 (2019). 681 12. Baylon, J.L., Cifone, N.A., Gulcher, J.R. & Chittenden, T.W. Enhancing Retrosynthetic 682 Reaction Prediction with Deep Learning Using Multiscale Reaction Classification. J Chem 683 Inf Model 59, 673- 688 (2019). 684 13. Dai, H., Li, C., Coley, C., Dai, B. & Song, L. Retrosynthesis prediction with conditional 685 graph logic network. Advances in Neural Information Processing Systems 32(2019). 686 14. Segler, M.H.S. & Waller, M.P. Modelling Chemical Reasoning to Predict and Invent 687 Reactions. Chemistry 23, 6118- 6128 (2017). 688 15. Zheng, S., Rao, J., Zhang, Z., Xu, J. & Yang, Y. Predicting Retrosynthetic Reactions Using 689 Self- Corrected Transformer Neural Networks. J Chem Inf Model 60, 47- 55 (2020). 690 16. Liu, B., et al. Retrosynthetic Reaction Prediction Using Neural Sequence- to- Sequence 691 Models. ACS Cent Sci 3, 1103- 1113 (2017). 692 17. Vaswani, A., et al. Attention is all you need. Advances in neural information processing 693 systems 30(2017). 694 18. Karpov, P., Godin, G. & Tetko, I.V. A transformer model for retrosynthesis. in International 695 Conference on Artificial Neural Networks 817- 830 (Springer, 2019). 696 19. Ucak, U.V., Ashrymamatov, I., Ko, J. & Lee, J. Retrosynthetic reaction pathway prediction 697 through neural machine translation of atomic environments. Nat Commun 13, 1186 698 (2022). 699 20. Kumar, A., DeGregorio, N. & Iyengar, S.S. Graph- Theory- Based Molecular Fragmentation 700 for Efficient and Accurate Potential Surface Calculations in Multiple Dimensions. J Chem 701 Theory Comput 17, 6671- 6690 (2021). 702 21. Lin, Z., Yin, S., Shi, L., Zhou, W. & Zhang, Y.J. G2GT: Retrosynthesis Prediction with
+
+<--- Page Split --->
+
+Graph- to- Graph Attention Neural Network and Self- Training. J Chem Inf Model 63, 1894- 1905 (2023).  
+
+Tu, Z. & Coley, C.W. Permutation Invariant Graph- to- Sequence Model for Template- Free Retrosynthesis and Reaction Prediction. J Chem Inf Model 62, 3503- 3513 (2022).  
+
+Sacha, M., et al. Molecule Edit Graph Attention Network: Modeling Chemical Reactions as Sequences of Graph Edits. J Chem Inf Model 61, 3273- 3284 (2021).  
+
+de Souza, R., Miranda, L.S.M. & Bornscheuer, U.T. A Retrosynthesis Approach for Biocatalysis in Organic Synthesis. Chemistry 23, 12040- 12063 (2017).  
+
+Hasic, H. & Ishida, T. Single- Step Retrosynthesis Prediction Based on the Identification of Potential Disconnection Sites Using Molecular Substructure Fingerprints. J Chem Inf Model 61, 641- 652 (2021).  
+
+Shi, C., Xu, M., Guo, H., Zhang, M. & Tang, J. A graph to graphs framework for retrosynthesis prediction. in International conference on machine learning 8818- 8827 (PMLR, 2020).  
+
+Wu, Z., et al. Mining Toxicity Information from Large Amounts of Toxicity Data. J Med Chem 64, 6924- 6936 (2021).  
+
+Li, X., Wang, Y. & Ruiz, R. A Survey on Sparse Learning Models for Feature Selection. IEEE Trans Cybern 52, 1642- 1660 (2022).  
+
+Wang, Y., Magar, R., Liang, C. & Barati Farimani, A. Improving Molecular Contrastive Learning via Faulty Negative Mitigation and Decomposed Fragment Contrast. J Chem Inf Model 62, 2713- 2725 (2022).  
+
+Du, B.X., et al. MLGL- MP: a Multi- Label Graph Learning framework enhanced by pathway interdependence for Metabolic Pathway prediction. Bioinformatics 38, 1325- 1332 (2022).  
+
+Schneider, N., Stiefl, N. & Landrum, G.A. What's What: The (Nearly) Definitive Guide to Reaction Role Assignment. J Chem Inf Model 56, 2336- 2346 (2016).  
+
+Lowe, D.M. University of Cambridge (2012).  
+
+van der Gaag, M., et al. The five- factor model of the Positive and Negative Syndrome Scale II: a ten- fold cross- validation of a revised model. Schizophr Res 85, 280- 287 (2006).  
+
+Cooper, M.M. & Klymkowsky, M.W. The trouble with chemical energy: why understanding bond energies requires an interdisciplinary systems approach. CBE Life Sci Educ 12, 306- 312 (2013).  
+
+Burness, C.B. & Scott, L.J. Sonidegib: A Review in Locally Advanced Basal Cell Carcinoma. Target Oncol 11, 239- 246 (2016).  
+
+Funaki, Y., et al. Effects of acotiamide on functional dyspepsia patients with heartburn who failed proton pump inhibitor treatment in Japanese patients: A randomized, double- blind, placebo- controlled crossover study. Neurogastroenterol Motil 32, e13749 (2020).
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+- Data1.xlsx- Data2.xlsx- Data3.xlsx- Supplementary.pdf
+
+<--- Page Split --->

preprint/preprint__0eedd8efcf3fab683389f91e77a829502b5c698130c328f60b9e1bf314f4cc29/preprint__0eedd8efcf3fab683389f91e77a829502b5c698130c328f60b9e1bf314f4cc29_det.mmd

@@ -0,0 +1,676 @@
+<|ref|>title<|/ref|><|det|>[[44, 108, 951, 177]]<|/det|>
+# Retro-MTGR: Molecule Retrosynthesis Prediction via Multi-Task Graph Representation Learning  
+
+<|ref|>text<|/ref|><|det|>[[44, 196, 290, 241]]<|/det|>
+Jian- Yu Shi jianyushi@nwpu.edu.cn  
+
+<|ref|>text<|/ref|><|det|>[[44, 268, 404, 287]]<|/det|>
+Northwestern Polytechnical University  
+
+<|ref|>text<|/ref|><|det|>[[44, 293, 195, 311]]<|/det|>
+Pengcheng Zhao  
+
+<|ref|>text<|/ref|><|det|>[[52, 315, 601, 334]]<|/det|>
+School of Life Sciences, Northwestern Polytechnical University  
+
+<|ref|>text<|/ref|><|det|>[[44, 340, 152, 357]]<|/det|>
+Xue- Xin Wei  
+
+<|ref|>text<|/ref|><|det|>[[52, 362, 601, 381]]<|/det|>
+School of Life Sciences, Northwestern Polytechnical University  
+
+<|ref|>text<|/ref|><|det|>[[44, 386, 152, 404]]<|/det|>
+Qiong Wang  
+
+<|ref|>text<|/ref|><|det|>[[52, 408, 601, 427]]<|/det|>
+School of Life Sciences, Northwestern Polytechnical University  
+
+<|ref|>text<|/ref|><|det|>[[44, 432, 160, 450]]<|/det|>
+Qi- Hao Wang  
+
+<|ref|>text<|/ref|><|det|>[[52, 455, 805, 474]]<|/det|>
+School of Chemistry and Chemical Engineering, Northwestern Polytechnical University  
+
+<|ref|>text<|/ref|><|det|>[[44, 479, 140, 497]]<|/det|>
+Jia- Ning Li  
+
+<|ref|>text<|/ref|><|det|>[[52, 501, 601, 520]]<|/det|>
+School of Life Sciences, Northwestern Polytechnical University  
+
+<|ref|>text<|/ref|><|det|>[[44, 526, 132, 544]]<|/det|>
+Jie Shang  
+
+<|ref|>text<|/ref|><|det|>[[52, 548, 601, 567]]<|/det|>
+School of Life Sciences, Northwestern Polytechnical University  
+
+<|ref|>text<|/ref|><|det|>[[44, 572, 128, 590]]<|/det|>
+Cheng Lu  
+
+<|ref|>text<|/ref|><|det|>[[52, 594, 861, 613]]<|/det|>
+Institute of Basic Research in Clinical Medicine China Academy of Chinese Medical Sciences  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 655, 104, 672]]<|/det|>
+## Article  
+
+<|ref|>title<|/ref|><|det|>[[44, 692, 135, 710]]<|/det|>
+# Keywords:  
+
+<|ref|>text<|/ref|><|det|>[[44, 729, 344, 748]]<|/det|>
+Posted Date: September 6th, 2023  
+
+<|ref|>text<|/ref|><|det|>[[44, 767, 475, 787]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 3205328/v1  
+
+<|ref|>text<|/ref|><|det|>[[42, 805, 914, 847]]<|/det|>
+License: © © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+<|ref|>text<|/ref|><|det|>[[42, 865, 535, 885]]<|/det|>
+Additional Declarations: There is NO Competing Interest.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[42, 45, 943, 88]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on January 18th, 2025. See the published version at https://doi.org/10.1038/s41467-025-56062-y.
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[181, 101, 815, 144]]<|/det|>
+# Retro-MTGR: Molecule Retrosynthesis Prediction via Multi-Task Graph Representation Learning  
+
+<|ref|>text<|/ref|><|det|>[[144, 151, 850, 191]]<|/det|>
+Peng- Cheng Zhao \(^{1}\) , Xue- Xin Wei \(^{1}\) , Qiong Wang \(^{1}\) , Qi- Hao Wang \(^{2}\) , Jia- Ning Li \(^{1}\) , Jie Shang \(^{1*}\) , Cheng Lu \(^{3*}\) , Jian- Yu Shi \(^{1*}\)  
+
+<|ref|>text<|/ref|><|det|>[[144, 198, 850, 260]]<|/det|>
+\(^{1}\) School of Life Sciences, Northwestern Polytechnical University, Xi'an 710072, China \(^{2}\) School of Chemistry and Chemical Engineering, Northwestern Polytechnical University, Xi'an 710072, China  
+
+<|ref|>text<|/ref|><|det|>[[144, 267, 850, 307]]<|/det|>
+\(^{3}\) Institute of Basic Research in Clinical Medicine China Academy of Chinese Medical Sciences, Beijing 100700, China  
+
+<|ref|>text<|/ref|><|det|>[[144, 313, 840, 330]]<|/det|>
+Corresponding author: jianyushi@nwpu.edu.cn; lv_cheng0816@163. com; shangjie03@nwpu.edu.cn  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 362, 222, 376]]<|/det|>
+## Abstract  
+
+<|ref|>text<|/ref|><|det|>[[140, 380, 850, 911]]<|/det|>
+It is a vital bridging step to infer appropriate synthesis reaction routes (i.e., retrosynthesis) of newly- designed molecules. Unlike classical experience- based retrosynthesis approaches, artificial intelligence enables a cheap and fast retrosynthesis approach. Template- based models, limited in known synthesis templates, leverage substructure searching to infer candidate reaction centers (i.e., bonds). In contrast, both translation- based models (TransMs) and discriminative methods (DiscMs) are free to synthesis templates. TransM regards retrosynthesis as a translation from the target molecule to its reactants by generative algorithms. DiscM, directly inspired by chemical synthesis, performs reaction center recognition and leaving group identification in turn. Nevertheless, TransMs are redundant and weakly interpretable, while existing DiscMs neglect the associations between reaction centers and leaving groups. To address these issues, this paper elaborates a novel discriminative Multi- Task Graph Representation learning model of Retrosynthesis prediction (Retro- MTGR). It solves two major supervised discriminative tasks (i.e., the reaction center recognition and the leaving group identification respectively), and an auxiliary self- supervised task (i.e., atom embedding enhancer) simultaneously. The comparison with various state- of- the- art methods first demonstrates the superiority of Retro- MTGR. Then, the ablation studies reveal how its crucial components contribute to the prediction respectively, including the atom embedding enhancer, bond energies, and the leaving group co- occurrence graph. More importantly, comprehensive investigations validate its chemical interpretability by answering two questions: why a bond can be the reaction center or not, and what leaving groups are appropriate to given synthons. The answers demonstrate that Retro- MTGR can reflect five underlying chemical synthesis rules by characterizing molecule structures
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[145, 88, 850, 197]]<|/det|>
+alone. Finally, two case studies demonstrate that the inferred retrosynthesis routes by Retro- MTGR are significantly consistent with those achieved by performed chemical synthesis assays. It's anticipated that our Retro- MTGR can provide prior guidance for real retrosynthesis route planning. The code and data underlying this article are freely available at https://github.com/zpczaizheli/Retro- MTGR.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 205, 270, 219]]<|/det|>
+## 1 Introduction  
+
+<|ref|>text<|/ref|><|det|>[[145, 227, 851, 451]]<|/det|>
+By integrating artificial intelligence (AI) technologies<sup>1</sup>, modern drug design has exhibited marvelous achievements on diverse tasks (e.g., target screening<sup>2</sup>, molecule generation<sup>3</sup>, ADMET prediction<sup>4</sup>, etc.) with a significant reduction in cost and time<sup>5,6</sup>. Once the chemical structure of a small molecule is determined in silico, there is an important task, retrosynthesis, which finds available reactants to be synthesized into the drug- like molecule in reality<sup>7</sup>. Such a retrosynthesis process works as a bridge from in silico to in reality. Compared to the ordinary synthesis reaction, the retrosynthesis is its inverse inference process<sup>8,9</sup>. A complete route of retrosynthesis is composed of multiple steps of synthesis reactions. However, even inferring a single step of synthesis heavily relies on individual domain experiences of chemists under costly trial- and- error assays<sup>10</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[146, 457, 850, 544]]<|/det|>
+In recent years, both the accumulation of chemical synthesis data and the blooming of deep learning methods boost the rapid development of computer- assisted synthesis processes (CASP) in retrosynthesis, which are roughly grouped into template- based, translation- based and discriminative methods.  
+
+<|ref|>text<|/ref|><|det|>[[145, 550, 851, 777]]<|/det|>
+Upon summarizing empirical rules of existing chemical syntheses (by usually RDKit<sup>11</sup>), template- based methods infer the single- step retrosynthesis of a newly given molecule by local structure similarity- based searching. For example, after determining possible reaction types of a target molecule, DHN, derived from gating neural networks, searches candidate templates among reaction type- specific templates<sup>12</sup>. GLN, a conditional graphical model upon graph neural networks, acquires candidate templates of the target molecule by subgraph pattern matching<sup>13</sup>. However, template- based methods cannot predict the retrosynthesis for target molecules having novel synthesis patterns outside the synthesis rules in the template library. In addition, it is tedious to update template libraries as new synthesis knowledge is discovered<sup>14</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[145, 783, 850, 892]]<|/det|>
+In contrast, both translation- based and discriminative methods are template- free and can predict the retrosynthesis reaction without a pre- built template library. Translation- based methods generally regard the retrosynthesis process as a case of machine translation<sup>15</sup>, which learns a translation model from the target molecule to its reactants by generative algorithms (e.g., LSTM<sup>16</sup> and Transformer<sup>17</sup>). Existing
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[145, 88, 850, 128]]<|/det|>
+translation- based methods can be further categorized into sequence- to- sequence translation models (Seq2Seq) and graph- to- seq translation models (Graph2Seq).  
+
+<|ref|>text<|/ref|><|det|>[[145, 134, 851, 428]]<|/det|>
+(1) Seq2Seq models treat a target molecule as one string (e.g., SMILES) and its two reactants as another string (the concatenation of two reactant SMILES). The first Seq2Seq model retrosynthesis method utilizes attention-enhanced LSTMs to convert the target molecule to its reactants under an encoder-decoder architecture<sup>16</sup>. As the new super-star in natural language processing, Transformer is also applied to retrosynthesis prediction by treating each character in SMILE as a word in recent years<sup>18</sup>. However, these methods arise a new issue that generated reactants are probably invalid in terms of chemistry. To meet the chemical validity of generated reactants, SCROP designs an extra syntax post-checker (derived from RDKit) based on Transformer<sup>15</sup>. RetroTRAE treats molecule substructures (capturing local atomic environment) as words in the Transformer to guarantee the validity of generated reactants<sup>19</sup>. Although these Seq2Seq models have achieved inspiring retrosynthesis predictions, they ignore rich information hidden in molecule chemical structures (i.e., the topology between atoms and bonds)<sup>20</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[145, 434, 851, 729]]<|/det|>
+(2) Graph2Seq models, representing target molecules as graphs, enriches the molecule representation and then map them into their reactant sequences under the auto-regressive generation framework<sup>8</sup>. Since the decoders of Graph2Seq models are similar, their contributions mainly focus on the design of encoders by graph neural networks (GNNs). For example, to encode target molecular graphs, G2GT designs a self-attention module enhanced by degrees of atoms and pairwise shortest distances between atoms<sup>21</sup>. Graph2SMILES utilizes a directed message-passing neural network (MPNN) to capture atom representations with the extra enhancement of global attention encoding<sup>22</sup>. By treating a chemical reaction as a queue of graph edits, MEGAN adopts a GNN-based encoder-decoder architecture and outputs reactants by a queue of leaned graph edits on the input molecule graph step by step<sup>23</sup>. Usually, these Graph2Seq models achieve better retrosynthesis prediction since they capture richer topology of molecule structures than Seq2Seq models.  
+
+<|ref|>text<|/ref|><|det|>[[145, 736, 850, 867]]<|/det|>
+However, both Seq2Seq and Graph2Seq models are still derived from generative models, which cannot provide well-interpretable results to chemists or pharmacologists in terms of chemical synthesis mechanisms. Moreover, dissimilar to the translation from one language to another language, translation- based retrosynthesis prediction has a redundant learning process due to highly overlapping structures between target molecules and their reactants.  
+
+<|ref|>text<|/ref|><|det|>[[180, 874, 848, 891]]<|/det|>
+Different from translation- based models, discriminative models are inspired by real
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[144, 88, 852, 360]]<|/det|>
+chemical synthesis. In brief, their prime step is to find the reaction center (e.g., the bond broken in retrosynthesis inference) where the molecule is split into two synthons (i.e., incomplete reactants) \(^{24}\) . Then, two synthons are attached to appropriate functional groups (i.e., leaving groups, LGs) to form reactants respectively. For example, Hasic et al. characterize local substructures of two bonding atoms as the bond representation by extended- connectivity fingerprints, and query candidate reactants by similarity search in a pre- built compound library \(^{25}\) . The G2Gs model encodes bonds by atom local topology and molecule global topology to find the reaction center by relational graph convolutional networks, and then builds a variational graph model to infer functional groups to be attached to synthons \(^{26}\) . However, current discriminative models treat the reaction center recognition and the LG identification as two separate steps, such that the association between the target molecule and its reactants is neglected.  
+
+<|ref|>text<|/ref|><|det|>[[147, 365, 850, 450]]<|/det|>
+In summary, existing translation- based generative models have weak interpretability and a redundant translation from target molecules and their reactants, while current discriminative models neglect the association between the reaction center recognition and the LG identification.  
+
+<|ref|>text<|/ref|><|det|>[[147, 457, 850, 590]]<|/det|>
+To address these issues, this paper elaborates a novel Multi- Task Graph Representation learning framework of Retrosynthesis prediction (Retro- MTGR), which is a discriminative model in essence. It solves three related tasks, including two major supervised discriminative tasks (i.e., the reaction center recognition and the LG identification respectively) and an auxiliary self- supervised task (i.e., atom embedding enhancer).  
+
+<|ref|>text<|/ref|><|det|>[[181, 597, 647, 614]]<|/det|>
+Overall, the main contributions of this work are as follows.  
+
+<|ref|>text<|/ref|><|det|>[[181, 620, 850, 777]]<|/det|>
+(1) This work proposes a novel retrosynthesis prediction model (Retro-MTGR), which consists of a reaction-center perceptron (RCP), a leaving group predictor (LGP), and an atom embedding enhancer (AEE). RCP characterizes the molecule topology to recognize the retrosynthesis reaction center, while AEE utilizes the redundancy between the product molecule and its synthons to boost atom embeddings for RCP. By leveraging the co-occurrence between LGs, LGP identifies appropriate LGs of synthons to provide complete reactants.  
+
+<|ref|>text<|/ref|><|det|>[[181, 783, 850, 892]]<|/det|>
+(2) Aiming to better chemical interpretability of retrosynthesis prediction, Retro-MTGR answers why a bond can be the reaction center or not from three aspects. First, a bond having high-breaking energy (>=360 kJ/mol) is usually an ordinary bond. Secondly, aromatic bonds (c-c) between carbon atoms are always ordinary bonds, while other types of bonds can be either reaction
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[210, 87, 849, 150]]<|/det|>
+centers or ordinary bonds. Last, a bond is the reaction center in a molecule if its member atoms tend to have opposite electrical properties (reflected by local substructures), otherwise an ordinary bond.  
+
+<|ref|>text<|/ref|><|det|>[[184, 157, 851, 359]]<|/det|>
+(3) Furthermore, Retro-MTGR answers what leaving groups are appropriate to given synthons as follows. First, LG pairs in reaction centers usually have opposite electrical properties, and occurrence-dominant LG pairs always consist of two simple groups (e.g., H, OH, or halogens). Secondly, individual LGs can be categorized into reaction-common and reaction-specific LGs. The former group is usually of simple LG and occurs in multiple types of synthesis reactions. The latter is of usually chemical substructure groups and occurs only in specific types of reactions. Also, either similar chemical properties or structures between LGs imply their potential substitution in synthesis reactions.  
+
+<|ref|>sub_title<|/ref|><|det|>[[147, 367, 235, 381]]<|/det|>
+## 2. Methods  
+
+<|ref|>sub_title<|/ref|><|det|>[[147, 390, 519, 405]]<|/det|>
+### 2.1 Problem formulation and model construction  
+
+<|ref|>text<|/ref|><|det|>[[147, 411, 850, 499]]<|/det|>
+Given a set of \(n\) chemical reactions \(R = \{r_1^i + r_2^i = c_i|i = 1,\dots,n\}\) , where the reactants \(r_1^i\) and \(r_2^i\) are two reactant molecules for the synthesis of target molecule \(c_i\) . The task is to find the retrosynthetic strategy for a newly-designed target molecule \(c\) (i.e., to recommend reactants \((r_1,r_2)\) for \(c\) ).  
+
+<|ref|>text<|/ref|><|det|>[[145, 504, 850, 824]]<|/det|>
+To perform a chemist- like retrosynthesis, we elaborate a Multi- Task Graph Representation learning framework of Retrosynthesis prediction (Retro- MTGR), which contains a Reaction- Center Perceptron (RCP) module, an Atom Embedding Enhancer (AEE) module and a Leaving Group Predictor (LGP). They account for two major tasks and one auxiliary task respectively. The first major task, implemented by RCP, is modeled as a binary discriminative problem, which recognizes the reaction center \(b_{u^*,v^*}\) among all the bonds \(\{b_{u,v}\}\) of the target molecule. Also, it breaks down \(b_{u^*,v^*}\) to obtain two synthons \(s_{u^*}\) and \(s_{v^*}\) , where \(u,v\) are two bonding atoms in \(c\) . To support the first major task in terms of atom embeddings, the auxiliary task (implemented by AEE) is modeled as a self- supervised contrast learning problem, which characterizes the structural commonality and difference between \(c\) and its synthons \((s_{u^*,s_{v^*}})\) . The second major task is modeled as a multi- class discriminative problem, which assigns appropriate leaving groups \((k_{u^*,k_{v^*}})\) to the synthons \((s_{u^*,s_{v^*}})\) to form complete reactants \((r_{u^*,r_{v^*}})\) under the enhancement of leaving group dependence.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[150, 90, 825, 560]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[147, 570, 832, 795]]<|/det|>
+<center>Figure 1. The framework of Retro-MTGR. Molecules in the form of graphs first are represented by an MPNN-based atom encoder to learn initial atom embeddings. The atom embedding enhancer (AEE) further boosts the atom embeddings by contrastive learning on molecules and their synthons w.r.t. molecule embeddings. The Reaction-Center Perceptron (RCP) leverages a bond-level readout on enhanced atom embeddings to learn bond embeddings, which are sequentially augmented by extra bond energies and then recognize reaction centers among bonds. After that, the leaving group predictor (LGP) learns LG embeddings based on a leaving group co-occurrence graph and measures the proximity between them and synthon embeddings (involving atoms and bonds in reaction centers) to predict appropriate LGs for given synthons. </center>  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 802, 390, 818]]<|/det|>
+### 2.2 Reaction-Center Perceptron  
+
+<|ref|>text<|/ref|><|det|>[[147, 825, 850, 910]]<|/det|>
+Identifying the reaction center is the first step in the inference of retrosynthesis. Inspired by this chemical empirical approach, we primarily attempt to recognize the reaction center among all the bonds of a given target molecule. The reaction center is the bond broken in terms of retrosynthesis. Thus, the task of reaction center recognition can be naturally
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 87, 848, 128]]<|/det|>
+modeled as a binary discriminative problem, which recognizes the reaction center \(b_{u^{*},v^{*}}\) among all the bonds \(\{b_{u,v}\}\) of the target molecule.  
+
+<|ref|>text<|/ref|><|det|>[[147, 134, 850, 313]]<|/det|>
+For this task, we design a reaction- center perceptron (RCP) module, which includes an atom encoder, a bond- level readout layer, and a multi- layer perceptron (MLP). The atom encoder is implemented by a two- layer message passing neural network (MPNN) to turn molecule graphs \(\mathcal{G} = (\mathcal{A},\mathcal{B})\) into atom embeddings \(\{\mathbf{a}_i\}\) , which are further refined by an Atom Embedding Enhancer (AEE). The bond- level readout layer generates bond embeddings \(\{\mathbf{b}_{u,v}\}\) , which are further boosted by concatenating with bond energy \(b_e\) . The MLP accounts for the discrimination of bonds by \(y = \mathcal{F}(\mathbf{b}_{u,v})\) , where \(y = 1\) if \(b_{u,v}\) is the reaction center, \(y = 0\) otherwise.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 320, 262, 334]]<|/det|>
+## Atom Encoder  
+
+<|ref|>text<|/ref|><|det|>[[147, 342, 851, 522]]<|/det|>
+According to chemical structure, each compound \(m\) is represented as a molecule graph \(\mathcal{G} = (\mathcal{A},\mathcal{B})\) , where \(\mathcal{A}\) is the set of its atoms \(\{a_i\}\) , \(\mathcal{B}\) is the set of its chemical bonds \(\{b_{ij}\}\) , and \(i,j = 1,2,\dots,|\mathcal{A}|\) . Let \(\mathbf{E}\in R^{N\times N}(N = |\mathcal{A}|)\) be its adjacency matrix, in which \(e_{ij} = 1\) , indicates the occurring bond \(\left(b_{ij}\in \mathcal{B}\right)\) between two atoms (i.e., \(a_i\) and \(a_j\) ) and \(e_{ij} = 0\) indicates no bond. Suppose that \(\mathbf{x}_i\in R^n\) is the initial feature vector of atom \(a_i\) , which is usually coded into a vector containing one- hot- shaped atom types, number of hydrogen atoms, and other attributes. Since \(\mathbf{x}\) is sparse, an extra multi- layer MLP maps it into its dense form \((\mathbf{x}_i\in R^p)\) to avoid the vanishing gradient problem.  
+
+<|ref|>text<|/ref|><|det|>[[147, 528, 850, 591]]<|/det|>
+Both \(\mathbf{E}\) and \(\mathbf{x}\) are input into a two- layer MPNN to generate atom embeddings \(\{\mathbf{a}_i\}\) for molecule \(c\) . The MPNN updates the embedding \(\mathbf{a}_i\) of each atom \(a_i\) by aggregating those of its neighboring atoms in a layer as follows,  
+
+<|ref|>equation<|/ref|><|det|>[[226, 594, 830, 631]]<|/det|>
+\[\mathbf{a}_i^{t + 1} = \sigma \left(\mathbf{w}_1^t \left(\sum_{j\in \mathcal{N}(a_i)} (\mathbf{w}_j^t \mathbf{a}_j^t) + \mathbf{b}^t\right) + \mathbf{w}_2^t \mathbf{a}_i^t\right), t = \{1,2\} , \quad (1)\]  
+
+<|ref|>text<|/ref|><|det|>[[147, 635, 850, 722]]<|/det|>
+where \(\mathbf{a}_i^t\) denotes the embedding of atom \(i\) in the t- th layer of MPNN, \(\mathbf{a}_i^1 = \mathbf{x}\) , \(\mathcal{N}(a_i)\) denotes the neighbors of atom \(a_i\) in the molecule graph \(\mathcal{G}\) , \(\sigma (\cdot)\) is a non- linear activation function (e.g., \(ReLu\) ), all \(\{\mathbf{w}^t\}\) are layer- wise learnable weight matrices accounting for a linear transformation, and \(\mathbf{b}^t\) denotes a learnable bias.  
+
+<|ref|>sub_title<|/ref|><|det|>[[147, 728, 242, 743]]<|/det|>
+## Perceptron  
+
+<|ref|>text<|/ref|><|det|>[[147, 750, 850, 882]]<|/det|>
+After passing through the MPNN, the initial atom feature \(\mathbf{x}\in R^n\) is turned to the updated atom embedding \(\mathbf{a}_i\in R^q\) . It is further refined by the AEE module, which characterizes the structural commonality and difference between the molecule and its synths. Meanwhile, the refined atom embeddings are utilized by the LGP module to help find appropriate leaving groups for the synths. All three tasks are associated together by shared atom embeddings. See Sections 2.3 and 2.4 for details.  
+
+<|ref|>text<|/ref|><|det|>[[180, 889, 848, 906]]<|/det|>
+Sequentially, the refined atom embeddings \(\{\mathbf{a}_i\}\) are then used to generate bond
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 86, 850, 174]]<|/det|>
+embedding by the bond- level readout. Let \(b_{ij}\) be the bond connecting atoms \(a_{i}\) and \(a_{j}\) . Unlike the ordinary molecule- level readout (e.g., the combination of all atoms), the RCP model defines a bond- level readout function \(\mathcal{R}_{B}(a_{i},a_{j})\) , which is augmented by bond energy, to obtain the bond embedding \(\mathbf{b}_{ij}\) as follows,  
+
+<|ref|>equation<|/ref|><|det|>[[323, 177, 829, 199]]<|/det|>
+\[\mathbf{b}_{ij} = \mathcal{R}_{B}(a_{i},a_{j}) = \left[(\mathbf{a}_{i} + \mathbf{a}_{j}); g_{ij}\right], \quad (2)\]  
+
+<|ref|>text<|/ref|><|det|>[[147, 202, 850, 267]]<|/det|>
+where \(g_{ij}\) is the theoretical bond energy, and ';' indicates the concatenation of \(g_{ij}\) and the atom embedding summation. As we observed, bond energy contributes to screen out the ordinary bonds having high bond energies. See also Section 3.4.1.  
+
+<|ref|>text<|/ref|><|det|>[[147, 272, 850, 406]]<|/det|>
+Last, RCP identifies the reaction center among all the bonds of \(m\) . Define \(\mathcal{C}\) as the set of bond flags \(\{c_{ij}\}\) w.r.t. molecule \(m\) , where \(c_{ij} = 1\) if \(b_{ij}\) is the reaction center (a positive sample), \(c_{ij} = 0\) otherwise (a negative sample). Based on the abovementioned bond embeddings \(\{\mathbf{b}_{ij}\}\) , it is naive to construct a two- layer MLP (denoted as \(\mathcal{F}_{b}\) ) as the classifier to achieve such a bond identification (i.e., \(c_{ij}^{*} = \mathcal{F}_{b}(\mathbf{b}_{ij})\) ). To train the model, the cross- entropy loss function over all the training molecules is defined as follows:  
+
+<|ref|>equation<|/ref|><|det|>[[177, 408, 829, 462]]<|/det|>
+\[l_{bond} = -\frac{1}{|M|}\sum_{m = 1}^{|M|}\left(\frac{1}{|\mathcal{B}_{m}|}\sum_{b_{ij}\in \mathcal{B}_{m}}^{|\mathcal{B}_{m}|}\left(c_{ij}\log c_{ij}^{*} + \left(1 - c_{ij}\right)\log \left(1 - c_{ij}^{*}\right)\right)\right), \quad (3)\]  
+
+<|ref|>text<|/ref|><|det|>[[147, 466, 773, 484]]<|/det|>
+where \(\mathcal{M}\) is the set of all the training molecules, and \(\mathcal{B}_{m}\) is the bond set of \(m\) .  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 490, 394, 506]]<|/det|>
+### 2.3 Atom Embedding Enhancer  
+
+<|ref|>text<|/ref|><|det|>[[147, 512, 851, 600]]<|/det|>
+As we observed, there is an amazing analog between the edge perturbation in graph contrast learning \(^{29}\) and the breaking of the reaction center in the retrosynthesis. Inspired by this observation, we designed an atom embedding enhancer (AEE) module based on graph contrastive learning.  
+
+<|ref|>text<|/ref|><|det|>[[147, 606, 851, 796]]<|/det|>
+For a target molecule \(m\) , we treat its two synthons \((s_{1}\) and \(s_{2}\) ) as a new perturbed molecule \(s\) , which is generated by removing the reaction center from \(m\) . We also collect another different molecule \(\bar{s}\) , which is randomly selected from other molecules or their perturbed molecules. Let \(\mathbf{h}_{m}\) , \(\mathbf{h}_{s}\) , and \(\mathbf{h}_{\bar{s}}\) be their embeddings respectively. These molecule embeddings are generated by an atom encoder and a molecule- level readout function \(\mathcal{R}_{M}(\cdot)\) . The former has the shared parameters with the atom encoder used in the RCP module. For a given molecule, \(\mathcal{R}_{M}(\cdot)\) aggregates the embeddings of its all atoms to generate the molecular- level embedding by an ordinary average pooling \(\mathbf{h} = \frac{1}{|\mathcal{A}|}\sum_{i = 1}^{|\mathcal{A}|}\mathbf{a}_{i}\) .  
+
+<|ref|>text<|/ref|><|det|>[[147, 804, 851, 893]]<|/det|>
+In terms of contrastive learning, the molecule embedding pair of \(\mathbf{h}_{m}\) and \(\mathbf{h}_{s}\) is regarded as a positive sample while the pair of \(\mathbf{h}_{m}\) and \(\mathbf{h}_{\bar{s}}\) is taken as a negative sample. Our goal is to train a contrastive learning model, which pushes \(\mathbf{h}_{m}\) and \(\mathbf{h}_{s}\) as near as possible (similar) while pushing \(\mathbf{h}_{m}\) and \(\mathbf{h}_{\bar{s}}\) as far as possible (different). For this
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[145, 88, 600, 105]]<|/det|>
+purpose, we design a contrastive loss function as follows:  
+
+<|ref|>equation<|/ref|><|det|>[[268, 110, 828, 142]]<|/det|>
+\[l_{contrast} = -\frac{1}{|M|}\sum_{m = 1}^{|M|}\log \left(\frac{\exp(\mathbf{h}_{m},\mathbf{h}_{s}^{T})}{\exp(\mathbf{h}_{m},\mathbf{h}_{s}^{T}) + \exp(\mathbf{h}_{m},\mathbf{h}_{s}^{T})}\right). \quad (4)\]  
+
+<|ref|>text<|/ref|><|det|>[[147, 148, 850, 210]]<|/det|>
+Thus, the AEE module enables an ingenious utility of chemical structural commonness and differences between a molecule and its synthons to enhance atom embeddings for other tasks.  
+
+<|ref|>sub_title<|/ref|><|det|>[[147, 217, 368, 233]]<|/det|>
+### 2.4 Leaving Group Predictor  
+
+<|ref|>text<|/ref|><|det|>[[147, 240, 851, 418]]<|/det|>
+Once synthons are decided based on the reaction center, chemists can obtain corresponding reactants by attaching appropriate leaving groups (LGs) to them. Thus, the task of LG recognition can be modeled as a multi- class classification. Inspired by chemists, we hold the idea that both the reaction center and the local substructures around reaction sites are crucial factors to determine LGs. Meanwhile, we consider the fact that LGs are not independent but are associated in the chemical synthesis sense (See also Section 3.4.2). Based on these considerations, we propose an elaborate leaving group predictor (LGP) based on multi- class classification to identify leaving groups.  
+
+<|ref|>text<|/ref|><|det|>[[147, 423, 850, 511]]<|/det|>
+Let \(\mathcal{K}\) be the list of all possible LGs, \(b_{u,v}\) be the reaction center of molecule \(m\) , \(a_{u},a_{v}\) be the reaction sites (i.e., atoms forming \(b_{u,v}\) ), \(s_{u},s_{v}\) be its synthons. Formally, \(s_{u}\) can be assigned with one or more LGs (i.e., \(\mathcal{K}_{u} = \mathcal{F}(s_{u})\subseteq \mathcal{K}\) ) to form its corresponding reactants \(r_{u}\) (i.e., \(\{r_{u}(i) = \mathcal{K}_{u}(i) + s_{u}\}\) , \(i = 1,\dots,|\mathcal{K}_{u}|\}\) ).  
+
+<|ref|>text<|/ref|><|det|>[[147, 516, 850, 652]]<|/det|>
+To implement our first idea, the reaction center \(b_{u,v}\) is represented by its bond embedding \(\mathbf{b}_{u,v}\) in RCP, while the local substructure around reaction site \(a_{u}\) is just represented by the atom embedding \(\mathbf{a}_{u}\) , which already aggregates its neighbors due to the MPNN in RCP. Thus, the embedding of the synthon \(s_{u}\) containing \(a_{u}\) can be defined as their concatenation \(\mathbf{s}_{u} = [\mathbf{a}_{u};\mathbf{b}_{uv}]\) . Similarly, we can define the embedding of \(s_{v}\) by \(\mathbf{s}_{v} = [\mathbf{a}_{v};\mathbf{b}_{uv}]\) .  
+
+<|ref|>text<|/ref|><|det|>[[147, 657, 851, 882]]<|/det|>
+To implement our second idea, we construct the leaving group co- occurrence graph (LGCoG) \(\mathcal{G}_{k} = \{\mathcal{K},\mathcal{E}\}\) , where \(\mathcal{K} = \{k_{i}|i = 1,\dots,|\mathcal{K}|\}\) denotes the set of nodes (leaving groups), and \(\mathcal{E} = \{e_{ij}\}\) denotes the set of weighted edges (normalized co- occurrences between leaving groups). Each LG is a small- size chemical substructure (e.g., - OH, - B(OH)2) or an atom/ion individual (e.g., - Cl, - Br, - H). The popular one- hot coding is used as initial node features \(\{\mathbf{k}_{i}^{0}\}\) . The edge building contains two steps as follows. First, the LG co- occurrence is calculated based on the training dataset, where the co- occurrence of two LGs is counted if they are involved in the same reaction. Define \(\mathbf{U} = \{u_{ij}\} \in \mathbb{R}^{|\mathcal{K}|\times |\mathcal{K}|}\) as the LG co- occurrence matrix, where \(u_{ij}\) denotes the pairwise co- occurrence counts between \(k_{i}\) and \(k_{j}\) . Then, a probability matrix \(\mathbf{P}\) can be calculated by \(\mathbf{U}\) . Therefore, \(p_{ij}\) is calculated as follows:
+
+<--- Page Split --->
+<|ref|>equation<|/ref|><|det|>[[378, 81, 829, 122]]<|/det|>
+\[p_{ij} = \frac{u_{ij}}{\sum_{j = 1}^{k}\sum_{i = 1}^{k}u_{ij}}. \quad (5)\]  
+
+<|ref|>text<|/ref|><|det|>[[147, 124, 849, 166]]<|/det|>
+We set it as the weight of the edge from \(k_{j}\) to \(k_{i}\) (i.e., \(e_{ij} = p_{ij}\) ). Thus, the embedding \((\mathbf{k}_{i})\) of LG \(k_{i}\) can be represented by performing an MPNN on \(g_{k}\) as follows:  
+
+<|ref|>equation<|/ref|><|det|>[[216, 167, 829, 207]]<|/det|>
+\[\mathbf{k}_{i}^{t + 1} = \sigma \left(\mathbf{w}_{1}^{t}\left(\sum_{j\in \mathcal{N}(k_{i})}\left(e_{ij}\mathbf{w}_{j}^{t}\mathbf{k}_{j}^{t}\right) + \mathbf{b}^{t}\right) + \mathbf{w}_{2}^{t}\mathbf{k}_{i}^{t}\right),t = \{1,2\} , \quad (6)\]  
+
+<|ref|>text<|/ref|><|det|>[[147, 208, 848, 249]]<|/det|>
+where \(j\in \mathcal{N}(k_{i})\) is the neighborhood of \(k_{i}\) , \(\mathbf{w}\) is the learnable transformation matrix, and \(\sigma (\cdot)\) ) is a non- linear activation function (i.e., ReLU).  
+
+<|ref|>text<|/ref|><|det|>[[147, 254, 850, 341]]<|/det|>
+After obtaining the synthon embeddings and the LG embeddings, we can directly perform discriminate the candidate LG to be attached to a synthon. As suggested by MLGL- MP (2022) \(^{30}\) , we measure the proximities between a given synthon \(s_{u}\) and a given LG \(k_{i}\) as follows:  
+
+<|ref|>equation<|/ref|><|det|>[[395, 345, 829, 364]]<|/det|>
+\[\hat{y}_{ui} = \mathbf{s}_u(\mathbf{k}_i)^T. \quad (7)\]  
+
+<|ref|>text<|/ref|><|det|>[[147, 369, 849, 410]]<|/det|>
+The proximity \(\hat{y}_{ui}\) is the predicting score of the given synthon attaching the \(i\) - th LG among the LG set \(\mathcal{K}\) , and reflects how possibly \(s_{u}\) is attached to \(k_{i}\) .  
+
+<|ref|>text<|/ref|><|det|>[[147, 415, 850, 527]]<|/det|>
+However, such a direct proximity measure would be senseless since the synthon embedding space and the LG embedding space are of different vector spaces. To tackle this issue, we design an adapter to map \(\{\mathbf{s}_u\}\) into \(\{\mathbf{k}_i\}\) . The adapter can be implemented by an MLP containing an input layer, a hidden layer, and an output layer. Thus, the final compound representation feature is defined as \(\mathbf{s}_u^* = \mathrm{MLP}(\mathbf{s}_u) \in \mathbb{R}^s\) , where \(s\) is the dimension of \(k_{i}\) .  
+
+<|ref|>text<|/ref|><|det|>[[180, 531, 800, 549]]<|/det|>
+Last, the mean square error (MSE) loss function is used when training LGP as follows:  
+
+<|ref|>equation<|/ref|><|det|>[[310, 554, 829, 583]]<|/det|>
+\[l_{group} = \frac{1}{2M*|\mathcal{K}|}\sum_{u = 1}^{2M}\sum_{i = 1}^{|\mathcal{K}|}(\hat{y}_{ui} - y_{ui})^2, \quad (8)\]  
+
+<|ref|>text<|/ref|><|det|>[[147, 590, 850, 655]]<|/det|>
+where \(y_{ui} \in \{0,1\}\) is the true label indicating whether or not a synthon \(s_{j}\) is attached by an LG \(k_{i}\) , \(\hat{y}_{ji}\) is the corresponding score output by LGP, \(M\) is the number of all the training molecules, and \(2M\) represents the number of their synthons.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 662, 373, 678]]<|/det|>
+### 2.4 Training Loss and Testing  
+
+<|ref|>text<|/ref|><|det|>[[147, 685, 849, 725]]<|/det|>
+To train the whole Retro- MTGR model, we combine the three abovementioned loss functions w.r.t. tasks into a linear joint as follows  
+
+<|ref|>equation<|/ref|><|det|>[[275, 730, 829, 749]]<|/det|>
+\[Loss = w_{1}*l_{bond} + w_{2}*l_{contrast} + w_{3}*l_{group}, \quad (9)\]  
+
+<|ref|>text<|/ref|><|det|>[[147, 753, 682, 772]]<|/det|>
+where \(\sum_{i = 1}^{3}w_{i} = 1\) are normalized hyperparameters to adjust task weights.  
+
+<|ref|>text<|/ref|><|det|>[[147, 777, 849, 816]]<|/det|>
+Note that Retro- MTGR in the testing should remove the AEE module since it cannot be available in the scenario.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 823, 330, 839]]<|/det|>
+## 3. Result and discussion  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 847, 412, 863]]<|/det|>
+### 3.1 Dataset and parameter settings  
+
+<|ref|>text<|/ref|><|det|>[[147, 869, 849, 909]]<|/det|>
+As popularly used in existing methods \(^{31}\) , we collected the benchmark dataset from the USPTO- 50K dataset, which was derived from an open- source patent database
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 88, 850, 175]]<|/det|>
+containing 50,016 atom- mapped reactions \(^{32}\) . In this work, by discarding modification- like chemical reactions, we only extracted the chemical reactions where one target molecule is synthesized from two reactants. As a result, our dataset contains 30,565 reaction entries, which are divided into 7 categories according to reaction type (Table 1).  
+
+<|ref|>text<|/ref|><|det|>[[147, 180, 851, 454]]<|/det|>
+In the Atom Encoder, as suggested in existing methods \(^{4}\) , each atom was initially represented by a 28- dimensional (28- d) atom feature vector \((\mathbf{x}_i \in R^{28})\) , including Atom Type (23- d), the number of Hydrogens (1- d), the number of linking neighbors of atom (Degree, 1- d), Is Aromatic (1- d), Formal Charge (1- d), as Atomic Mass (1- d). See also Table 2 for details. Due to the one- hot coding of Atom Type, the initial atom representation \(\mathbf{x}_k\) is sparse. To avoid the vanishing gradient problem \(^{28}\) , an extra three- layer MLP maps it into its dense form \((\mathbf{x}_k \in R^{32})\) . We empirically set 64 and 32 neurons in its hidden layer and output layer respectively. Moreover, bonds are represented as a binary adjacent matrix \(\mathbf{E}\) , of which \(e_{ij} = 1\) indicates the occurring bond between two atoms \((a_i\) and \(a_j)\) , and no bond otherwise. Both \(\{\mathbf{x}_k\}\) and \(\mathbf{E}\) are input into a two- layer MPNN to obtain atom embeddings having the same dimensions as those of \(\{\mathbf{x}_k\}\) .  
+
+<|ref|>text<|/ref|><|det|>[[147, 459, 850, 592]]<|/det|>
+In the RCP module, the MLP accounting for reaction center identification contains also an input layer, a hidden layer, and an output layer. There are 33 neurons in the input layer, where 32 neurons are responsible for the resulting embeddings from the bond- level readout and the last one takes charge of the bond energy. The number of neurons is empirically set to 16. The unique neuron in the output layer accounts for the confidence score of being a reaction center.  
+
+<|ref|>text<|/ref|><|det|>[[147, 597, 850, 823]]<|/det|>
+In the LGP module, the nodes in the leaving group co- occurrence graph (LGCoG) are initially represented as \(n\) - dimensional one- hot coding vectors \(\{\mathbf{k}_i^0 \in \mathbb{R}^{|\mathcal{K}|}\}\) , where \(n = |\mathcal{K}|\) is the cardinality of the LG set. Then, they are mapped into LG embeddings \(\{\mathbf{k}_i \in \mathbb{R}^{|\mathcal{K}|}\}\) by another two- layer MPNN without dimensional change. On the other side, the adapter maps the synthon embedding space (the concatenation of 32- d atom embeddings and 33- d bond embeddings) into the LG embedding space. It is implemented by a three- layer MLP, which contains an input layer accounting for synthon embeddings \((\mathbf{s}_u \in \mathbb{R}^{65})\) , a hidden layer having empirically 128 neurons, and an output layer having \(n\) neurons respectively. Thus, \(\mathbf{s}_u^* = \mathrm{MLP}(\mathbf{s}_u) \in \mathbb{R}^n\) , where \(n\) (i.e., \(|\mathcal{K}|\) ) is scenario- specific since the type- known scenario and the type- unknown scenario have different types of LGs.
+
+<--- Page Split --->
+<|ref|>table<|/ref|><|det|>[[147, 106, 848, 315]]<|/det|>
+<|ref|>table_caption<|/ref|><|det|>[[415, 90, 620, 103]]<|/det|>
+Table 1. Dataset overview   
+
+<table><tr><td>Reaction type</td><td>Reaction name</td><td>No. of examples</td></tr><tr><td>1</td><td>heteroatom alkylation and arylation</td><td>14188</td></tr><tr><td>2</td><td>acylation and related processes</td><td>10509</td></tr><tr><td>3</td><td>C-C bond formation</td><td>4378</td></tr><tr><td>4</td><td>protections</td><td>144</td></tr><tr><td>5</td><td>oxidations</td><td>142</td></tr><tr><td>6</td><td>functional group interconversion (FGI)</td><td>986</td></tr><tr><td>7</td><td>functional group addition (FGA)</td><td>218</td></tr><tr><td>Total</td><td>/</td><td>30565</td></tr></table>  
+
+<|ref|>table<|/ref|><|det|>[[147, 360, 848, 527]]<|/det|>
+<|ref|>table_caption<|/ref|><|det|>[[412, 345, 585, 359]]<|/det|>
+Table 2.Atom attributes   
+
+<table><tr><td>Feature</td><td>Description</td><td>Dimension</td></tr><tr><td>Atom type</td><td>Cl, N, P, Br, B, S, I, F, C, O, ... (one-hot)</td><td>23</td></tr><tr><td>Number of H</td><td>Integer</td><td>1</td></tr><tr><td>Degree</td><td>Integer</td><td>1</td></tr><tr><td>Is Aromatic</td><td>True or False (binary)</td><td>1</td></tr><tr><td>Formal charge</td><td>Integer</td><td>1</td></tr><tr><td>Atomic Mass</td><td>Integer</td><td>1</td></tr></table>  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 531, 495, 547]]<|/det|>
+### 3.2 Comparison with state-of-the-art methods  
+
+<|ref|>text<|/ref|><|det|>[[147, 554, 850, 662]]<|/det|>
+To evaluate the effectiveness of Retro- MTGR, we compared it with five state- of- the- art single- step retrosynthesis methods, including two sequence- to- sequence translation methods (including seq2seq \(^{16}\) and SCPOP \(^{15}\) ), two graph- to- seq translation methods (including MEGAN \(^{23}\) and Graph2SMILES \(^{22}\) ), and one discriminative method (i.e., G2Gs \(^{26}\) ). They are briefly summarized below.  
+
+<|ref|>text<|/ref|><|det|>[[184, 669, 850, 896]]<|/det|>
+- Seq2seq: It is the first sequence-2-sequence translation method, which utilizes attention-enhanced LSTMs to convert the target molecule to its reactants in the form of SMILES under an encoder-decoder architecture \(^{16}\) .- SCROP: It is a transformer-based method with the aid of an extra syntax post-checker to guarantee the chemical validity of generated reactants \(^{15}\) .- MEGAN: Its outputs reactants by a queue of leaned graph edits (chemical structure modification) on the input molecule graph step by step under a GNN-based auto-regressive architecture \(^{23}\) .- Graph2SMILES: It is also an auto-regressive model, which utilizes a directed MPNN to capture atom representations with the extra enhancement of global
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[219, 88, 384, 104]]<|/det|>
+attention encoding<sup>22</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[180, 111, 850, 175]]<|/det|>
+- G2Gs: It is a two-step GNN-based discriminative model, which first encodes bonds to find the reaction center by relational GCNs, and then builds a variational graph model to infer leaving groups to be attached on synthons<sup>26</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[145, 180, 852, 405]]<|/det|>
+For a fair comparison, we run ten- fold cross- validation (10- CV) as suggested by these methods<sup>33</sup>. In detail, the dataset was randomly and equally split into 10 subsets, of which each subset (10% samples) was taken as the testing set and the remaining subsets (90% samples) were taken as the training set. Such a 10- CV was repeated 50 times under different random seeds. The average performance over 50 rounds of cross- validations was reported to measure the retrosynthesis prediction of Retro- MTGR. Moreover, the top- k accuracy (e.g., Top- 1, Top- 3, and Top- 5) was adopted as the measuring metric in 10- CV. It is defined as the ratio of the number of correctly predicted target molecules to the total number of target molecules, where a target molecule is correctly predicted if its correct reactants are found among the top- k predicted candidates.  
+
+<|ref|>text<|/ref|><|det|>[[146, 410, 852, 567]]<|/det|>
+In addition, two retrosynthesis scenarios, Reaction Type Known (RTK) and Reaction Type Unknown (RTU), were considered, as these methods performed. In the first scenario RTK, we are required to perform a retrosynthesis for a molecule while being given its possible reaction type. In the second one RTU, we have no information about its potential reaction type. Usually, RTU is more practical but difficult than RTK. Besides, since the prediction in RTK is specific to reaction types, we reported only the average performance over those reaction types in Table 3 and listed the detailed results in Supplementary (Section 1).  
+
+<|ref|>text<|/ref|><|det|>[[146, 572, 852, 729]]<|/det|>
+The comparison results demonstrate that our Retro- MTGR achieves the best prediction and is significantly superior to other state- of- the- art methods over two testing scenarios. The results also validate that RTU is more difficult than RTK since extra type information in RTK helps the prediction. In detail, it achieves 69.1%, 89.2%, and 91.0% accuracies in the case of RTK while achieving 57.3%, 81.0%, and 86.5% respectively in the case of RTU in terms of Top- 1, Top- 3, and Top- 5. Remarkably, Retro- MTGR achieves 5\~7% improvements in the case of RTK while achieving 4\~12% improvements in RTU over Top- 1, Top- 3, and Top- 5 accuracies respectively.
+
+<--- Page Split --->
+<|ref|>table<|/ref|><|det|>[[135, 108, 844, 320]]<|/det|>
+<|ref|>table_caption<|/ref|><|det|>[[245, 90, 748, 105]]<|/det|>
+Table3. Top-k accuracy for retrosynthesis prediction on USPTO.   
+
+<table><tr><td rowspan="3">Methods</td><td colspan="6">Top-k Accuracy (%)</td></tr><tr><td colspan="3">Reaction Type Unknown</td><td colspan="3">Reaction Type Known</td></tr><tr><td>1</td><td>3</td><td>5</td><td>1</td><td>3</td><td>5</td></tr><tr><td rowspan="2">Sequence-to-Sequence</td><td>Seq2seq</td><td>35.0±2.5</td><td>41.7±2.1</td><td>56.5±3.5</td><td>43.4±2.7</td><td>60.2±2.8</td></tr><tr><td>SCROP</td><td>44.6±3.5</td><td>63.3±2.2</td><td>66.8±3.2</td><td>58.1±2.5</td><td>75.6±3.1</td></tr><tr><td rowspan="2">Graph-to-Seq</td><td>MEGAN</td><td>46.8±5.9</td><td>72.7±5.1</td><td>75.4±5.0</td><td>57.7±6.0</td><td>83.1±5.6</td></tr><tr><td>Graph2Smiles</td><td>51.7±3.5</td><td>65.9±1.7</td><td>72.8±2.4</td><td>61.7±3.2</td><td>80.3±2.5</td></tr><tr><td rowspan="2">Discriminative</td><td>G2Gs</td><td>53.0±1.2</td><td>70.5±1.2</td><td>74.2±0.9</td><td>59.0±0.9</td><td>83.4±1.0</td></tr><tr><td>Retro-MTGR</td><td>57.3±1.4</td><td>81.0±0.8</td><td>86.5±0.7</td><td>69.1±0.6</td><td>89.2±1.1</td></tr></table>  
+
+<|ref|>sub_title<|/ref|><|det|>[[145, 325, 298, 339]]<|/det|>
+### 3.3 Ablation studies  
+
+<|ref|>text<|/ref|><|det|>[[145, 348, 851, 504]]<|/det|>
+In this section, we investigated how crucial components of our Retro- MTGR contribute to the retrosynthesis prediction by ablation studies. We made three variants of our original model by masking one block of Retro- MTGR in turn. First, we removed the AEE module (denoted as w/o AEE). Secondly, we discarded bond energies in bond embeddings (denoted as w/o BE). Last, we deleted the leaving group co- occurrence graph (LGCoG) in the LGP module and used one- hot coding as the initial representations of leaving groups (denoted as w/o LGCoG).  
+
+<|ref|>text<|/ref|><|det|>[[145, 508, 851, 712]]<|/det|>
+As the ablation comparison illustrates, the superiority of Retro- MTGR to all its variants demonstrates that all of the AEE module, the bond energy, and the leaving group graph play significant roles in the retro- synthesis prediction in the case of both unknown and known reaction types (Figure 2). Specifically, the AEE module plays the most important role. For example, Retro- MTGR with the AEE module improves the Top- 1, Top- 3, and Top- 5 accuracies by \(5.9\%\) , \(8.2\%\) , and \(3.1\%\) respectively in the case of unknown reaction type. The result indicates that the AEE module can enhance bond embeddings in finding the reaction centers because it utilizes chemical structural commonness and differences between a molecule and its synthons.  
+
+<|ref|>text<|/ref|><|det|>[[145, 717, 850, 828]]<|/det|>
+The bond energy block also provides an untrivial contribution to bond embeddings. For example, Retro- MTGR with bond energies improves the Top- 1, Top- 3, and Top- 5 accuracies by \(1.6\%\) , \(4.9\%\) , and \(3.0\%\) respectively when reaction types are unknown. The underlying reason is that bonds having high bond energies are usually ordinary bonds but not reaction centers. See also Section 3.4.1 for details.  
+
+<|ref|>text<|/ref|><|det|>[[145, 833, 850, 898]]<|/det|>
+The LGCoG provides a significant contribution to leaving group identification. For instance, Retro- MTGR with the LGCoG improves the Top- 1, Top- 3, and Top- 5 accuracies by \(3.3\%\) , \(3.7\%\) , and \(1.8\%\) respectively when reaction types are unknown. The essential
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[145, 88, 850, 128]]<|/det|>
+reason for such improvements is that the captured LG dependences enrich LG representations when identifying LGs for synthons. Details can be found in Section 3.4.2.  
+
+<|ref|>text<|/ref|><|det|>[[147, 134, 850, 197]]<|/det|>
+In general, all of the AEE module, the bond energy, and the leaving group co- occurrence graph play indispensable roles in retrosynthesis prediction. More detailed investigations in the next section indicate why they work.  
+
+<|ref|>image<|/ref|><|det|>[[152, 205, 825, 456]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[147, 464, 832, 526]]<|/det|>
+<center>Figure 2. Ablation comparison. Compared with the three variants (red, green and purple bars), Retro-MTGR (blue bars) achieves the best retrosynthesis prediction in terms of top-1, top-3, and top-5 in the case of both unknown and known reaction types. </center>  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 534, 399, 549]]<|/det|>
+### 3.4 Retrosynthesis rule discovery  
+
+<|ref|>text<|/ref|><|det|>[[147, 557, 851, 643]]<|/det|>
+In this section, we shall attempt to uncover retrosynthesis rules by Retro- MTGR in two interpretable views, bond view and leaving group view. First, we shall investigate three questions to reveal why a bond can be the reaction center. Furthermore, we shall explore two questions to indicate what leaving groups are appropriate to given synthons.  
+
+<|ref|>title<|/ref|><|det|>[[148, 650, 270, 665]]<|/det|>
+#### 3.4.1 Bond view  
+
+<|ref|>text<|/ref|><|det|>[[148, 672, 850, 711]]<|/det|>
+To interpret why a bond can be the reaction center, we considered three bond- derived questions as follows.  
+
+<|ref|>sub_title<|/ref|><|det|>[[181, 718, 734, 735]]<|/det|>
+## (1) Can bond energies determine the reaction center in a molecule alone?  
+
+<|ref|>text<|/ref|><|det|>[[147, 741, 851, 803]]<|/det|>
+For a chemical bond, its bond energy (i.e., the minimum energy to break it down) measures its stableness<sup>34</sup>. The larger, the stabler, the more difficult to be synthesized from the point of view of chemical retrosynthesis. Thus, we assume that the reaction center is of low bond- energy bond.  
+
+<|ref|>text<|/ref|><|det|>[[147, 810, 851, 895]]<|/det|>
+To validate it, we made a statistical distribution of bond energies across all the bonds of molecules in a histogram (Figure 3). Note that we considered only the theoretical breaking energy of each chemical bond, but not considering the influence of its neighboring bonds or near atoms. The bonds were sorted into 20 equally spaced bins along the axis of bond energy between the
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[145, 88, 851, 151]]<|/det|>
+minimum and maximum energy values (kJ/mol). Due to the number difference between reaction centers and ordinary bonds, the heights of bins (i.e., the number of bonds falling in the bins) were normalized by the total number of bonds for convenient comparison.  
+
+<|ref|>text<|/ref|><|det|>[[145, 157, 855, 405]]<|/det|>
+As illustrated, the majority of bonds fall into four bins, [270,315], [315, 360], [450- 495], and [495- 540]. Specifically, the bond energies of reaction centers are usually located in the lower range of bond energy (i.e., \(95.26\%\) having bond energy \(< 360 \mathrm{kJ / mol}\) ). In contrast, \(45.33\%\) of ordinary bonds have bond energy \(< 360 \mathrm{kJ / mol}\) and \(54.65\%\) have bond energy \(> = 360 \mathrm{kJ / mol}\) . Thus, a naïve decision can be made that bonds having bond energy \(> 360 \mathrm{kJ / mol}\) are usually ordinary bonds. Such a finding can be used to filter out the ordinary bonds having large breaking energies in the process of reaction center identification. This is why bond energies have a significant contribution to finding the reaction center. However, bond energies can NOT determine the reaction center in a molecule alone, since there are still a large number of ordinary bonds (45.33 %) overlapping with reaction centers in the case of having low breaking energies \(< 360 \mathrm{kJ / mol}\) . This issue can be further investigated by the answer to the second question.  
+
+<|ref|>image<|/ref|><|det|>[[152, 433, 783, 675]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[147, 690, 833, 775]]<|/det|>
+<center>Figure 3. Breaking energy distribution of chemical bonds. The X-axis shows the bond energy (0-900) divided into 20 intervals. The Y-axis shows the frequency of chemical bonds falling in different bins. As illustrated, bonds having bond energy \(> 360 \mathrm{kJ / mol}\) are usually ordinary bonds. However, \(45.33\%\) of ordinary bonds overlap with reaction centers in the case of breaking energies \(< 360 \mathrm{kJ / mol}\) . </center>  
+
+<|ref|>text<|/ref|><|det|>[[147, 784, 850, 822]]<|/det|>
+(2) What is the underlying chemical rule captured by bond embeddings such that reaction centers can be distinguished from ordinary bonds?  
+
+<|ref|>text<|/ref|><|det|>[[147, 830, 850, 890]]<|/det|>
+One of the core contributions of our model (Retro- MTGR) is just the discrimination of reaction centers from ordinary bonds in the case of low bond energy. Since bond embedding representations (Formula 2) characterize bond features based on molecule graph topology, we
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[140, 88, 850, 150]]<|/det|>
+utilized them to exhibit the difference between reaction centers and ordinary bonds. Principal component analysis (PCA) was used to visualize bonds in 2- dimensional space, where each point represents a bond.  
+
+<|ref|>text<|/ref|><|det|>[[183, 157, 595, 174]]<|/det|>
+Such a bond space was rendered in three maps (Figure 4).  
+
+<|ref|>text<|/ref|><|det|>[[183, 180, 851, 290]]<|/det|>
+- The first map shows a clear separation between reaction centers (red points) and ordinary bonds (blue points), except for a small overlapping (Figure 4-a). Such a separation demonstrates that our model can characterize the difference between reaction centers and ordinary bonds well. More importantly, both reaction centers and ordinary can be split into communities, which are strongly specific to bond types.  
+
+<|ref|>text<|/ref|><|det|>[[183, 296, 851, 429]]<|/det|>
+- The second map indicates that bond communities are consistence with bond types (Figure 4-b). Some bonds are always ordinary bonds, such as c-c (an aromatic bond linking two carbon atoms). More importantly, it is remarkable that some bonds having the same types (e.g., C-C, C-O, and C-N.) may occur in different communities, which belong to reaction centers and ordinary bonds respectively. The underlying reason is investigated in the answer to the third question.  
+
+<|ref|>text<|/ref|><|det|>[[183, 435, 851, 615]]<|/det|>
+- The last map illustrates the distribution of bond energies in terms of energy bins (Figure 4-c). As observed, bonds with large breaking energies (two bins) are almost of ordinary bonds. More importantly, reaction centers and ordinary bonds having low breaking energies can be clearly distinguished. With the consideration of bond types, the bonds in the energy bin [270-315] mainly include C-N (a bond linking a carbon atom to a nitrogen atom) and C-Br (a bond linking a carbon atom to a bromine atom), while those in the energy bin [315-360] mainly include C-C (a bond linking two carbon atoms) and C-O (a bond linking a carbon atom to an oxygen atom).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[153, 80, 828, 808]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[147, 811, 835, 893]]<|/det|>
+<center>Figure 4. Bond space. (A). Reaction centers and ordinary bonds. Red dots indicate reaction centers while blue dots indicate ordinary bonds. (B). Bond types. Different colors represent different bonds. (C). Bond energies. Different colors indicate different bond energy bins (kJ/mol). In terms of chemical symbols, C stands for carbon atoms, c is for carbon atoms in aromatic bonds, and C' is for carbon atoms in general rings. </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 88, 837, 170]]<|/det|>
+Moreover, N stands for nitrogen atoms, n is for nitrogen atoms in aromatic bonds, and N' represents nitrogen atoms in general rings. In addition, O is for oxygen atoms, O' represents oxygen atoms in general rings, and S is for sulfur atoms. Four specific symbols, including '\\~', '\\~', '='', and '#', denotes aromatic, single, double, and triple bonds respectively.  
+
+<|ref|>sub_title<|/ref|><|det|>[[147, 177, 797, 194]]<|/det|>
+## (3) Why can a bond be the reaction center in a molecule but cannot be in another one?  
+
+<|ref|>text<|/ref|><|det|>[[147, 200, 851, 333]]<|/det|>
+Considering that bond energy can NOT determine the reaction center in a molecule alone, our model (Retro- MTGR) leverages molecule graph topologies to capture the differences between reaction centers and ordinary bonds, even having the same bond types. We investigated what chemical rule hidden is captured by atom/bond embeddings. It is anticipated that the inherent law helps identify reaction centers and ordinary bonds, especially in the case of both common bond type and similar bond energy.  
+
+<|ref|>text<|/ref|><|det|>[[145, 338, 852, 796]]<|/det|>
+Our investigation was inspired by the chemical knowledge that the electrical property (denoted as \(p\) ) of an atom in a molecule is determined by the conjugation effect of the motion of its electrons as well as the union of its spatial neighboring atoms. When showing an attractive effect on electrons, the atom is considered an electron- withdrawing atom. On the contrary, it is called an electron- donating atom. Since it is difficult to quantify atom electrical properties due to complicated inter- atom influences, we first proposed a qualitative manner to label their strength and weakness. Then, by enumerating atom- centered substructures, we found 356 substructures (Section 2 in Supplementary), which are categorized into four groups in terms of electrical property strength. In detail, the atoms showing strong electron- withdrawing/donating properties are labeled as \(p + + |p - -\) respectively. Meanwhile, those atoms exhibiting weak electron- withdrawing/donating properties are marked as \(p + /p -\) respectively. As a result, the atom pairs forming bonds show ten possible pairs of electrical properties (e.g., \((p + + |p - - )\) , \((p + |p - )\) ) in total. Last, we counted the percentages of all types of electrical property pairs in the case of both reaction centers and ordinary bonds (Figure 5). The result illustrates that the atom pairs forming reaction centers have dominant opposite electrical properties (97.0% with ' \(p + |p - '\) , ' \(p + |p - - '\) , ' \(p + + |p - '\) , and ' \(p + + |p - - '\) pairs) while those pairs forming ordinary bonds have same or similar electrical properties (80.1%). In fact, some ordinary bonds having opposite electrical properties are also reaction centers in the deeper steps of retrosynthesis. See also Section 3.5 Case Study. Thus, we conclude that the pair of an electron- withdrawing atom and an electron- donating atom tends to form a reaction center.  
+
+<|ref|>text<|/ref|><|det|>[[147, 802, 850, 865]]<|/det|>
+In summary, our answers to these questions demonstrate that our Retro- MTGR can capture the underlying matching rule of why a bond can be the reaction center by embedding molecule topologies.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[153, 85, 825, 323]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[147, 329, 833, 437]]<|/det|>
+<center>Figure 5. Electrical property distribution of chemical bonds. The X-axis denotes pairwise electrical property patterns of atom pairs forming bonds. Its left zone lists six patterns of same/similar electrical properties while its right zone lists four patterns of opposite electrical properties. The Y-axis indicates the frequencies of electrical property patterns. The member atoms of ordinary bonds usually have same or similar electrical properties (80.1%), whereas those in reaction centers tend to have opposite electrical properties (97.0%). </center>  
+
+<|ref|>title<|/ref|><|det|>[[149, 446, 344, 461]]<|/det|>
+#### 3.4.2 Leaving Group view  
+
+<|ref|>text<|/ref|><|det|>[[148, 468, 850, 508]]<|/det|>
+To dig out what leaving groups are appropriate to given synthons, we leveraged the leaving group co- occurrence graph (LGCoG) to answer two LG- derived questions as follows.  
+
+<|ref|>title<|/ref|><|det|>[[148, 515, 850, 554]]<|/det|>
+#### (1) Whether are two leaving groups associated when accounting for a pair of synthons derived from the same molecule?  
+
+<|ref|>text<|/ref|><|det|>[[147, 560, 852, 717]]<|/det|>
+We sought its answer in two aspects. First, we followed the answer to the third question in the previous section that retrosynthesis site atoms in synthons always have opposite electrical properties. Analogously, when two leaving groups account for a pair of synthons derived from the same molecule, they are also supposed to have opposite electrical properties based on the electrical matching rule between a synthon and its LG. We validated this assumption by labeling the electrical properties of LGs (Figure 6- A) in a similar manner as that in Section 3.4.1. As counted, 94.3% of LG pairs (28823/30565) have opposite electrical properties.  
+
+<|ref|>text<|/ref|><|det|>[[147, 722, 852, 902]]<|/det|>
+Then, we considered the occurrence of LG pairs. The co- occurrence is indicated by the edge thickness in the graph (Figure 6- A). As calculated, we found that the LG pairs consisting of two simple groups (e.g., H, OH, or halogens) usually occur frequently. Especially, the pairs 'H, OH' (27.8%), 'H, Cl' (27.4%), 'H, Br' (12.9%), and 'H, I' (8.1%) are the most frequent LG pairs. Moreover, the LG pairs including a simple group and a chemical substructure occur in a low frequency, such as 'Br, CC1(C)OBOCl(C)C' (1.8%). Few LG pairs (0.386%) are composed of chemical substructures only, such as 'O=S(=O)(O)C(F)(F)F- B(OH)2, O=S(=O)(O)C(F)(F)F- CC1(C)OBOCl(C)C'. More details can be found in Supplementary
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[145, 89, 232, 103]]<|/det|>
+(Section 3).  
+
+<|ref|>text<|/ref|><|det|>[[181, 111, 710, 128]]<|/det|>
+Thus, two leaving groups accounting for a pair of synthons are associated.  
+
+<|ref|>sub_title<|/ref|><|det|>[[147, 135, 583, 152]]<|/det|>
+## (2) Whether is a leaving group specific to a reaction type?  
+
+<|ref|>text<|/ref|><|det|>[[147, 158, 850, 267]]<|/det|>
+After taking a close look at the topology of the graph, we found a few LGs having many partner LGs and many LGs having only one partner. The former is always of simple groups (e.g., H, OH, or halogens), while the latter is usually chemical substructures. In addition, the remaining LGs have a small number of partners. To dig out the underlying reason, we made an investigation by counting LG occurrences according to reaction types (Section 4 in Supplementary).  
+
+<|ref|>text<|/ref|><|det|>[[147, 273, 850, 313]]<|/det|>
+The investigation reveals interesting knowledge that LGs can be split into two categories according to the number of reaction types they attending in.  
+
+<|ref|>text<|/ref|><|det|>[[147, 319, 851, 497]]<|/det|>
+The first category, named 'reaction- common' LGs, contains the LGs appearing in equal to or more than the half number of reaction types (i.e., \(> = 4\) ). They are usually of simple LGs, and occur frequently and have many matching partner LGs. In details, 'H', 'OH', 'Cl', and 'Br' appear in 7, 7, 6 and 5 categories, respectively. Especially, 'H' attends in almost all the reactions while occurring 27623 times and having 37 kinds of matching partner LGs (denoted by the node degree in the graph). Moreover, halogens including 'Cl', 'Br', and 'I', have many common partner LGs. The possible reason is that they have similar chemical properties, which suggests a potential replacement between them.  
+
+<|ref|>text<|/ref|><|det|>[[147, 504, 851, 728]]<|/det|>
+The second category, named 'reaction- specific' LGs, includes the LGs occurring only in specific types of reactions. They are of usually chemical substructure groups, such as '- CC', '- OCC(Cl)(Cl)Cl', and '- OCC(F)(F)F'. We found that reaction- specific LGs are surrounding reaction- common LGs or clustered together (Figure 6- B). In addition, some LGs involving in the same type of reaction have similar chemical structures, such as the pair of 'CCOC(=O)O' and 'CC(C)(C)OOC(=O)O' in Type 2 reaction (0.783) and the pair of 'B1OCC01' and 'B1OCCCO1' (0.754) in Type 3 reaction, where the similarity is calculated by MACCS fingerprints in terms of Jaccard similarity. Again, the chemical structure similarity of LGs implies their possible substitution, which is determined by the cost or the reaction condition available in synthesis reaction routes.  
+
+<|ref|>text<|/ref|><|det|>[[147, 736, 850, 844]]<|/det|>
+In general, the co- occurrence graph of leaving groups contains rich retrosynthesis information, including the inter- associations between LGs and their reaction specificity. Our Retro- MTGR can capture such rich information to enhance the retrosynthesis prediction. The answers in this section dig out why the graph contributes to the prediction.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[152, 90, 831, 707]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[147, 718, 839, 804]]<|/det|>
+<center>Figure 6. Leaving Group view. (A) Electrical property map. LGs having positive and negative electrical properties are rendered by red and blue. In total, \(94.3\%\) of LG pairs have opposite electrical properties. (B) Reaction type map. LGs occurring in single reaction types, multiple reaction types ( \(\leq 3\) ) and many types ( \(\geq 4\) , reaction-common) are highlighted in different colors. </center>  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 812, 262, 827]]<|/det|>
+### 3.5 Case Study  
+
+<|ref|>text<|/ref|><|det|>[[147, 834, 850, 898]]<|/det|>
+To evaluate the ability of our Retro- MTGR in the real scenario of retrosynthesis prediction, we collected two drugs (i.e., Sonidegib \(^{35}\) and Acotiamide \(^{36}\) ), not included in our dataset, as the studying cases. We inferred their retrosynthesis routes by our Retro- MTGR and then validated
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 88, 420, 104]]<|/det|>
+them by chemical assays respectively.  
+
+<|ref|>text<|/ref|><|det|>[[147, 111, 852, 313]]<|/det|>
+The two drugs we selected are briefly summarized as follows. The first drug, Sonidegib, is a Hedgehog signaling pathway inhibitor (via smoothened antagonism), which was developed as an anticancer agent by Novartis and approved by the FDA in 2015 for the treatment of basal cell carcinoma. Now it is commonly used for the treatment of locally advanced recurrent basal cell carcinoma (BCC) following surgery and radiation therapy, or in cases where surgery or radiation therapy are not appropriate (DrugBank ID: DB09143) \(^{35}\) . The second drug, Acotiamide, being an investigational drug, was designed and developed for the treatment of functional dyspepsia (FD) \(^{36}\) (DrugBank ID: DB12482). It works as a novel upper gastrointestinal (GI) motility modulator and stress regulator by improving upper gastrointestinal functions in both the stomach and esophagus.  
+
+<|ref|>text<|/ref|><|det|>[[147, 319, 852, 475]]<|/det|>
+Since Retro- MTG is a single- step retrosynthesis prediction model, we iteratively applied it to infer a complete retrosynthesis route for a given drug molecule. In the first iteration, Retro- MTG split the complete molecule into two synthons under the top- 1 criterion (the first candidate reaction center), which were further turned into smaller intermediate molecules by appending appropriate leave groups. Intermediate molecules were then split into smaller ones by Retro- MTG in a similar way unless the intermediate molecules are reactants, which can be easily bought in the market.  
+
+<|ref|>text<|/ref|><|det|>[[146, 482, 852, 846]]<|/det|>
+As shown in Figure 7, the predicted retrosynthesis route of Sonidegib illustrates that it (marked as '1') can be split into two intermediate molecules (marked as '2' and '3'). Further, they are split into two pairs of reactants respectively, where one pair is (marked as '4' and '5') while another is (marked as '6' and '7'). In each retrosynthesis step, the attention scores of bonds are labeled and the highest one (top- 1) is considered as the reaction center. During predicting the retrosynthesis route, both reaction centers and leaving groups are corrected predicted (Figure 7- A). Furthermore, according to the prediction, a series of chemical synthesis reactions starting from reactants ('4', '5', '6', and '7') were performed (Figure 7- B) to validate the predicted retrosynthesis route. Remarkably, due to the potential intra- reaction among '6' molecules triggered by their own Chlorine (- Cl) and amino group (- NH2), the expected reaction between '6' and '7' would generate less amount of '3' molecules. To guarantee a high production of '3' in the real synthesis, we converted '6' to '8' by a nitration reaction, which alters the amino group to a nitro group (- NO2). Then, the reaction of '8' and '7' produced '9', of which the nitro group was further converted back into the amino group in a reduction reaction to obtain '3' with the desired production. Meanwhile, the reaction of '4' and '5' generated '2'. Finally, we performed an amidation reaction by combining '2' and '3' to form the product molecule Sonidegib ('1').  
+
+<|ref|>text<|/ref|><|det|>[[147, 852, 850, 891]]<|/det|>
+Similarly, the retrosynthesis route of Acotiamide (marked as '10') is correctly predicted in terms of both reaction centers and leaving groups, and validated by chemical synthesis reactions
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[146, 88, 852, 290]]<|/det|>
+as well. Specifically, the first retrosynthesis step generates an intermediate molecule ('11') and a reactant ('12'). Then, the former is split into two reactants ('13' and '14') in the second step (Figure 7- C). Since there is also an intra-reaction issue in '14', a similar strategy was adopted to guarantee the final production of '10'. In brief, the carboxylic acid group (- OH) of '14' was altered to a methoxy group (- OCH3) so as to generate '15' by an esterification reaction. As the substitution of '14', '15' was put together with '13' to generate a new intermediate molecule '16', which was sequentially hydrolyzed to intermediate molecule 11 by both NaOH and HCl. Finally, a similar amidation reaction combining '11' and '12' was performed to form Acotiamide ('10') (Figure 7- D).  
+
+<|ref|>text<|/ref|><|det|>[[147, 296, 850, 358]]<|/det|>
+To summarize, Retro- MTGR shows that its inspiring retrosynthesis prediction is significantly consistent with chemical assays. Thus, it can provide clear guidance for retrosynthesis route planning with extra reaction conditions.  
+
+<|ref|>image<|/ref|><|det|>[[150, 365, 828, 859]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[147, 866, 850, 904]]<|/det|>
+<center>Figure 7. Predicted retrosynthetic routes and real chemical synthetic routes. A. Retrosynthesis prediction of Sonidegib. Sonidegib ('1') is split at the reaction center into intermediate molecules '2' and '3', where the highest </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[145, 87, 853, 570]]<|/det|>
+attention score generated by our Retro- MTGR is highlighted. Then, '2' and '3' are further decomposed into four reactants ('4', '5', '6' and '7') respectively according to their reaction centers. **B. Chemical synthesis of Sonidegib.** First, the Suzuki cross- coupling reaction between '4' and '5' was performed to obtain intermediate molecule '2'. Meanwhile, '6' is converted to '8' with the help of m- CPBA (m- Chloroperbenzoic Acid) for preventing its intra- reaction. Then, '8' is combined with '7' in the presence of DIEA (N, N- diisopropylethylamine) to form another intermediate molecule '9', which is further reduced by H₂ in the presence of Pd/C to generate '3'. In the last step, the coupling of compound '2' with '3' in the presence of HATU (2-(7- Azabenzotriazol- 1- yl)- N, N, N', N'- tetramethyluronium hexafluorophosphate) and DIEA (N,N- Diisopropylethylamine) in DMF (N,N- Dimethylformamide) generates Sonidegib ('1'). **C. Retrosynthesis prediction of Acotiamide.** Acotiamide ('10') is split at the reaction center into intermediate molecules '11' and reactant '12', where the highest attention score generated by our Retro- MTGR is highlighted as well. Then, '11' is further decomposed into two reactants ('13' and '14') respectively according to their reaction centers. **D. Chemical synthesis of Acotiamide.** First, reactant '14' is esterified with methanol to form '15'. Then, the coupling compound 13 with 15 in the presence of HATU and DIEA obtains the intermediate molecule '16', which is sequentially hydrolyzed to another intermediate molecule '11' by NaOH and HCl. Finally, the coupling between the intermediate molecule '11' and the reactant '12' in the presence of HATU and DIEA in DMF generates the molecule of Acotiamide ('10'). More details (e.g., reaction conditions and instruments) about the chemical synthesis reactions of these two drugs can be found in Supplementary (Section 5). Note: DIEA works as a catalyzer in the reduction reaction or a promoter in the aromatic nucleophilic substitution of N- hydrophosphoramide. HATU works as a popular condensation reagent in promoting amide bond formation by activating carboxyl groups. m- CPBA is a strong oxidizing agent widely used in organic synthesis. Pd/C is also a kind of popular catalyst in hydrogenation reduction.  
+
+<|ref|>sub_title<|/ref|><|det|>[[147, 576, 264, 590]]<|/det|>
+## 4 Conclusions  
+
+<|ref|>text<|/ref|><|det|>[[147, 598, 850, 752]]<|/det|>
+Aiming at a well- interpretable discriminative model in terms of chemical synthesis mechanism, this paper elaborates a novel multi- task graph representation learning framework of retrosynthesis prediction (Retro- MTGR). Based on molecule graphs, three related tasks are considered in Retro- MTGR simultaneously, where two major supervised discriminative tasks account for recognizing reaction centers and identifying leaving groups respectively, and an auxiliary self- supervised task accounts for generating better atom embeddings.  
+
+<|ref|>text<|/ref|><|det|>[[147, 759, 850, 892]]<|/det|>
+The comparison with various state- of- the- art methods demonstrates the overall superiority of Retro- MTGR in terms of cross- validation prediction performance in both reaction- type unknown and reaction- type known scenarios. Furthermore, the ablation study demonstrates its contributions as follows. First, it significantly enhances atom embedding by leveraging chemical structural redundancy and differences between a molecule and its synthons. Then, it uncovers that bond energies can partially boost bond
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[145, 88, 850, 128]]<|/det|>
+embeddings due to the difference between ordinary bonds and reaction centers in the case of high bond energy. Last, it utilizes LG co-occurrences to enrich LG representations.  
+
+<|ref|>text<|/ref|><|det|>[[145, 134, 852, 428]]<|/det|>
+Furthermore, multiple comprehensive investigations validate the chemical synthesis interpretability of Retro- MTGR by answering two questions: why a bond can be the reaction center or not, and what leaving groups are appropriate to given synthons. The answers demonstrate that Retro- MTGR can capture and illustrate the underlying chemical synthesis rules as follows: (1) a bond having high- breaking energy (>=360 kJ/mol) is usually an ordinary bond; (2) aromatic bonds (c-c) between carbon atoms are always ordinary bonds, while other types of bonds can be either reaction centers or ordinary bonds; (3) a bond is the reaction center in a molecule if its member atoms tend to have opposite electrical properties, otherwise an ordinary bond;(4) accordingly, leaving group (LG) pairs in reaction centers usually have opposite electrical properties and occurrence- dominant LG pairs always consisting of two simple groups (e.g., H, OH, or halogens); (5) individual LGs can be categorized into reaction- common LGs (simple groups) and reaction- specific LGs (chemical substructures).  
+
+<|ref|>text<|/ref|><|det|>[[147, 434, 850, 497]]<|/det|>
+Finally, the practical ability of Retro- MTGR is evaluated by two novel drugs. The results reveal that the inferred retrosynthesis routes by Retro- MTGR are significantly consistent with those achieved by chemical synthesis assays.  
+
+<|ref|>text<|/ref|><|det|>[[147, 504, 853, 589]]<|/det|>
+In summary, our Retro- MTGR can provide prior guidance for retrosynthesis route planning. We believe that its extension with the integration of extra synthetic factors (e.g., reaction yield, conditions, and reagents) can be a complete retrosynthesis route planning in the coming future.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 597, 314, 612]]<|/det|>
+## Acknowledgements:  
+
+<|ref|>text<|/ref|><|det|>[[147, 620, 850, 660]]<|/det|>
+The authors would like to thank anonymous reviewers for suggestions that improved the paper.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 668, 220, 683]]<|/det|>
+## Funding  
+
+<|ref|>text<|/ref|><|det|>[[147, 690, 850, 754]]<|/det|>
+This work was supported by the National Nature Science Foundation of China [61872297], the Shaanxi Province Key R&D Program [2023- YBSF- 114], and the CAAI- Huawei Mind Spore Open Fund [CAAIXSJLJJ- 2022- 035A].  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 784, 240, 798]]<|/det|>
+## References  
+
+<|ref|>text<|/ref|><|det|>[[147, 803, 852, 912]]<|/det|>
+1. Zhu, H. Big Data and Artificial Intelligence Modeling for Drug Discovery. Annu Rev Pharmacol Toxicol 60, 573-589 (2020).  
+2. Zhao, P., et al. Targets preliminary screening for the fresh natural drug molecule based on Cosine-correlation and similarity-comparison of local network. J Transl Med 20, 67 (2022).  
+3. Li, J.N., Yang, G., Zhao, P.C., Wei, X.X. & Shi, J.Y. CProMG: controllable protein-oriented
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[100, 78, 855, 900]]<|/det|>
+659 molecule generation with desired binding affinity and drug- like properties. 660 Bioinformatics 39, i326- i336 (2023). 661 4. Du, B.- X., Xu, Y., Yiu, S.- M., Yu, H. & Shi, J.- Y. MTGL- ADMET: A Novel Multi- task Graph 662 Learning Framework for ADMET Prediction Enhanced by Status- Theory and Maximum 663 Flow. in Research in Computational Molecular Biology: 27th Annual International 664 Conference, RECOMB 2023, Istanbul, Turkey, April 16- 19, 2023, Proceedings 85- 103 665 (Springer, 2023). 666 5. Gupta, R., et al. Artificial intelligence to deep learning: machine intelligence approach for 667 drug discovery. Mol Divers 25, 1315- 1360 (2021). 668 6. Vatansever, S., et al. Artificial intelligence and machine learning- aided drug discovery in 669 central nervous system diseases: State- of- the- arts and future directions. Med Res Rev 670 41, 1427- 1473 (2021). 671 7. Lee, A.A., et al. Molecular Transformer unifies reaction prediction and retrosynthesis 672 across pharma chemical space. Chem Commun (Camb) 55, 12152- 12155 (2019). 673 8. Dong, J., Zhao, M., Liu, Y., Su, Y. & Zeng, X. Deep learning in retrosynthesis planning: 674 datasets, models and tools. Brief Bioinform 23(2022). 675 9. Coley, C.W., Green, W.H. & Jensen, K.F. Machine Learning in Computer- Aided Synthesis 676 Planning. Acc Chem Res 51, 1281- 1289 (2018). 677 10. Yan, M. & Baran, P.S. Drug discovery: Fighting evolution with chemical synthesis. Nature 678 533, 326- 327 (2016). 679 11. Lovric, M., Molero, J.M. & Kern, R. PySpark and RDKit: Moving towards Big Data in 680 Cheminformatics. Mol Inform 38, e1800082 (2019). 681 12. Baylon, J.L., Cifone, N.A., Gulcher, J.R. & Chittenden, T.W. Enhancing Retrosynthetic 682 Reaction Prediction with Deep Learning Using Multiscale Reaction Classification. J Chem 683 Inf Model 59, 673- 688 (2019). 684 13. Dai, H., Li, C., Coley, C., Dai, B. & Song, L. Retrosynthesis prediction with conditional 685 graph logic network. Advances in Neural Information Processing Systems 32(2019). 686 14. Segler, M.H.S. & Waller, M.P. Modelling Chemical Reasoning to Predict and Invent 687 Reactions. Chemistry 23, 6118- 6128 (2017). 688 15. Zheng, S., Rao, J., Zhang, Z., Xu, J. & Yang, Y. Predicting Retrosynthetic Reactions Using 689 Self- Corrected Transformer Neural Networks. J Chem Inf Model 60, 47- 55 (2020). 690 16. Liu, B., et al. Retrosynthetic Reaction Prediction Using Neural Sequence- to- Sequence 691 Models. ACS Cent Sci 3, 1103- 1113 (2017). 692 17. Vaswani, A., et al. Attention is all you need. Advances in neural information processing 693 systems 30(2017). 694 18. Karpov, P., Godin, G. & Tetko, I.V. A transformer model for retrosynthesis. in International 695 Conference on Artificial Neural Networks 817- 830 (Springer, 2019). 696 19. Ucak, U.V., Ashrymamatov, I., Ko, J. & Lee, J. Retrosynthetic reaction pathway prediction 697 through neural machine translation of atomic environments. Nat Commun 13, 1186 698 (2022). 699 20. Kumar, A., DeGregorio, N. & Iyengar, S.S. Graph- Theory- Based Molecular Fragmentation 700 for Efficient and Accurate Potential Surface Calculations in Multiple Dimensions. J Chem 701 Theory Comput 17, 6671- 6690 (2021). 702 21. Lin, Z., Yin, S., Shi, L., Zhou, W. & Zhang, Y.J. G2GT: Retrosynthesis Prediction with
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[204, 85, 852, 120]]<|/det|>
+Graph- to- Graph Attention Neural Network and Self- Training. J Chem Inf Model 63, 1894- 1905 (2023).  
+
+<|ref|>text<|/ref|><|det|>[[204, 123, 850, 158]]<|/det|>
+Tu, Z. & Coley, C.W. Permutation Invariant Graph- to- Sequence Model for Template- Free Retrosynthesis and Reaction Prediction. J Chem Inf Model 62, 3503- 3513 (2022).  
+
+<|ref|>text<|/ref|><|det|>[[204, 160, 850, 195]]<|/det|>
+Sacha, M., et al. Molecule Edit Graph Attention Network: Modeling Chemical Reactions as Sequences of Graph Edits. J Chem Inf Model 61, 3273- 3284 (2021).  
+
+<|ref|>text<|/ref|><|det|>[[204, 197, 850, 232]]<|/det|>
+de Souza, R., Miranda, L.S.M. & Bornscheuer, U.T. A Retrosynthesis Approach for Biocatalysis in Organic Synthesis. Chemistry 23, 12040- 12063 (2017).  
+
+<|ref|>text<|/ref|><|det|>[[204, 234, 852, 287]]<|/det|>
+Hasic, H. & Ishida, T. Single- Step Retrosynthesis Prediction Based on the Identification of Potential Disconnection Sites Using Molecular Substructure Fingerprints. J Chem Inf Model 61, 641- 652 (2021).  
+
+<|ref|>text<|/ref|><|det|>[[204, 289, 850, 342]]<|/det|>
+Shi, C., Xu, M., Guo, H., Zhang, M. & Tang, J. A graph to graphs framework for retrosynthesis prediction. in International conference on machine learning 8818- 8827 (PMLR, 2020).  
+
+<|ref|>text<|/ref|><|det|>[[204, 345, 850, 380]]<|/det|>
+Wu, Z., et al. Mining Toxicity Information from Large Amounts of Toxicity Data. J Med Chem 64, 6924- 6936 (2021).  
+
+<|ref|>text<|/ref|><|det|>[[204, 382, 850, 417]]<|/det|>
+Li, X., Wang, Y. & Ruiz, R. A Survey on Sparse Learning Models for Feature Selection. IEEE Trans Cybern 52, 1642- 1660 (2022).  
+
+<|ref|>text<|/ref|><|det|>[[204, 419, 850, 472]]<|/det|>
+Wang, Y., Magar, R., Liang, C. & Barati Farimani, A. Improving Molecular Contrastive Learning via Faulty Negative Mitigation and Decomposed Fragment Contrast. J Chem Inf Model 62, 2713- 2725 (2022).  
+
+<|ref|>text<|/ref|><|det|>[[204, 475, 850, 509]]<|/det|>
+Du, B.X., et al. MLGL- MP: a Multi- Label Graph Learning framework enhanced by pathway interdependence for Metabolic Pathway prediction. Bioinformatics 38, 1325- 1332 (2022).  
+
+<|ref|>text<|/ref|><|det|>[[204, 512, 850, 546]]<|/det|>
+Schneider, N., Stiefl, N. & Landrum, G.A. What's What: The (Nearly) Definitive Guide to Reaction Role Assignment. J Chem Inf Model 56, 2336- 2346 (2016).  
+
+<|ref|>text<|/ref|><|det|>[[204, 549, 525, 565]]<|/det|>
+Lowe, D.M. University of Cambridge (2012).  
+
+<|ref|>text<|/ref|><|det|>[[204, 568, 850, 620]]<|/det|>
+van der Gaag, M., et al. The five- factor model of the Positive and Negative Syndrome Scale II: a ten- fold cross- validation of a revised model. Schizophr Res 85, 280- 287 (2006).  
+
+<|ref|>text<|/ref|><|det|>[[204, 623, 850, 675]]<|/det|>
+Cooper, M.M. & Klymkowsky, M.W. The trouble with chemical energy: why understanding bond energies requires an interdisciplinary systems approach. CBE Life Sci Educ 12, 306- 312 (2013).  
+
+<|ref|>text<|/ref|><|det|>[[204, 678, 848, 712]]<|/det|>
+Burness, C.B. & Scott, L.J. Sonidegib: A Review in Locally Advanced Basal Cell Carcinoma. Target Oncol 11, 239- 246 (2016).  
+
+<|ref|>text<|/ref|><|det|>[[204, 715, 850, 787]]<|/det|>
+Funaki, Y., et al. Effects of acotiamide on functional dyspepsia patients with heartburn who failed proton pump inhibitor treatment in Japanese patients: A randomized, double- blind, placebo- controlled crossover study. Neurogastroenterol Motil 32, e13749 (2020).
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[42, 42, 312, 70]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[42, 93, 768, 113]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[59, 131, 250, 231]]<|/det|>
+- Data1.xlsx- Data2.xlsx- Data3.xlsx- Supplementary.pdf
+
+<--- Page Split --->

preprint/preprint__0ef88a3d250cc02e121a31263f7d3eb45880dc07208a92e0c8644d22f4cb03f6/images_list.json

@@ -0,0 +1,122 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Fig. 1 Our synthetic concept of bowls from \\(p\\) -HBC.",
+    "footnote": [],
+    "bbox": [
+      [
+        550,
+        351,
+        874,
+        448
+      ]
+    ],
+    "page_idx": 3
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Fig. 2 The synthetic route for the superbowls. Reagents and conditions: (i) 6.6 equiv. \\(\\mathrm{TiCl_4}\\) , \\(\\alpha\\) -dichlorobenzene, (microwave), \\(180^{\\circ}\\mathrm{C}\\) , 3 h; (ii) 8 equiv. DDQ, \\(5\\%\\) TfOH, 1,2-dichloroethane, \\(50^{\\circ}\\mathrm{C}\\) , 6 h; (iii) 4 equiv. \\((\\mathrm{Bu}_3\\mathrm{Sn})_2\\mathrm{S}\\) or \\((\\mathrm{Bu}_3\\mathrm{Sn})_2\\mathrm{Se}\\) , 1 equiv. \\(\\mathrm{Pd(PPh}_3)_4\\) , toluene, \\(150^{\\circ}\\mathrm{C}\\) , Ar, 48 h.",
+    "footnote": [],
+    "bbox": [
+      [
+        105,
+        70,
+        895,
+        211
+      ]
+    ],
+    "page_idx": 4
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Fig. 3 \\(^1\\mathrm{H}\\) NMR characterization of 1a ( \\(\\mathrm{CDCl}_3\\) ) and geometry predicted by DFT calculations.",
+    "footnote": [],
+    "bbox": [
+      [
+        518,
+        247,
+        909,
+        430
+      ]
+    ],
+    "page_idx": 5
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Fig. 4 a,d, X-ray crystal structures of 1a (a) and 1b (d). b,e, The side view of 1a (b) and 1b (e) with diameters and depths. c,f, The top view of 1a (c) and 1b (f) together with POAV angles (blue) and mean bond lengths (red). Butyl groups in the side and top views are omitted for clarity. Thermal ellipsoids are shown at 30% probability.",
+    "footnote": [],
+    "bbox": [
+      [
+        135,
+        80,
+        861,
+        481
+      ]
+    ],
+    "page_idx": 6
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Fig. 5 a,b, \\(\\mathrm{NICS}(1)_{zz}\\) values (ppm) of \\(1\\mathbf{a}^{\\prime}\\) (a) and \\(1\\mathbf{b}^{\\prime}\\) (b) calculated at the GIAO B3LYP/6-311G(d,p) level of theory. c,d, Calculated AICD plots (isovalue \\(= 0.03\\) ) of \\(1\\mathbf{a}^{\\prime}\\) (c) and \\(1\\mathbf{b}^{\\prime}\\) (d). Only contributions from \\(\\pi\\) -electrons of the aromatic cores are considered. Red arrows indicate directions of induced ring current.",
+    "footnote": [],
+    "bbox": [
+      [
+        68,
+        78,
+        460,
+        370
+      ]
+    ],
+    "page_idx": 9
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_6.jpg",
+    "caption": "Fig. 6 a,d, Calculated electrostatic potentials of bowl concave for \\(1\\mathbf{a}^{\\prime}\\) (a) and \\(1\\mathbf{b}^{\\prime}\\) (d). b,e, Calculated electrostatic potentials of bowl convex for \\(1\\mathbf{a}^{\\prime}\\) (b) and \\(1\\mathbf{b}^{\\prime}\\) (e). c,f, Dipole moments of \\(1\\mathbf{a}^{\\prime}\\) (c) and \\(1\\mathbf{b}^{\\prime}\\) (f). Electrostatic potentials calculated on the 0.001 au. isodensity surface together with surface maxima (blue) and minima points (red) at the B3LYP/6-311G(d,p) level of theory.",
+    "footnote": [],
+    "bbox": [
+      [
+        127,
+        75,
+        875,
+        368
+      ]
+    ],
+    "page_idx": 10
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_7.jpg",
+    "caption": "Fig. 7 a,b, Inversion barriers of \\(1\\mathbf{a}^{\\prime}\\) (a) and \\(1\\mathbf{b}^{\\prime}\\) (b) calculated at M062x/6-31G(d) level of theory.",
+    "footnote": [],
+    "bbox": [
+      [
+        150,
+        457,
+        855,
+        638
+      ]
+    ],
+    "page_idx": 10
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_8.jpg",
+    "caption": "Fig. 8 a, UV-Vis absorption (solid curves) and emission spectra (dotted curves) of 1a (black) and 1b (red) in \\(\\mathrm{CH}_2\\mathrm{Cl}_2\\) at a concentration of \\(1.0\\times 10^{-5}\\mathrm{M}\\) . b, Cyclic voltammogram and differential pulse voltamogram of 1a and 1b in \\(\\mathrm{CH}_2\\mathrm{Cl}_2\\) (0.1 mol/L \\(n\\mathrm{-Bu}_4\\mathrm{NPF}_6\\) ) at a scan rate of 0.1 V/s. All potentials were calibrated versus an aqueous SCE by the addition of ferrocene as an internal standard taking \\(\\mathrm{E}_{1 / 2}(\\mathrm{Fc} / \\mathrm{Fc}^{+}) = 0.424\\mathrm{V}\\) vs. SCE [62]. c,d, Orbital correlation diagram and transition composition of the \\(\\mathrm{S}_0\\rightarrow (\\mathrm{S}_1,\\mathrm{S}_3,\\mathrm{S}_4)\\) excited states for 1a' (c) and 1b' (d) calculated at PBE/def2tzvp level of theory.",
+    "footnote": [],
+    "bbox": [
+      [
+        135,
+        77,
+        854,
+        504
+      ]
+    ],
+    "page_idx": 11
+  }
+]

preprint/preprint__0ef88a3d250cc02e121a31263f7d3eb45880dc07208a92e0c8644d22f4cb03f6/preprint__0ef88a3d250cc02e121a31263f7d3eb45880dc07208a92e0c8644d22f4cb03f6.mmd

@@ -0,0 +1,347 @@
+
+# Trichalcogenohexaindenocoronenes: Super Buckyblows Based on Coronene Core  
+
+Yixun SunShaanxi Normal UniversityXin WangShaanxi Normal UniversityMuhua ChenShaanxi Normal UniversityBo YangShaanxi Normal UniversityZiyi GuoShaanxi Normal UniversityMingyu XuShaanxi Normal UniversityYunjie ZhangShaanxi Normal UniversityHuaming SunShaanxi Normal UniversityJingshuang DangShaanxi Normal UniversityJuan FanShaanxi Normal UniversityJing LiShaanxi Normal UniversityJunfa Wei ( weijf@snnu.edu.cn )Shaanxi Normal University  
+
+Article  
+
+# Keywords:  
+
+Posted Date: September 29th, 2022  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 2095883/v1
+
+<--- Page Split --->
+
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License
+
+<--- Page Split --->
+
+# Trichalcogenohexaindenocoronenes: Super Buckylbowls Based  
+
+on Coronene Core  
+
+Yixun Sun \(^{1\dagger}\) , Xin Wang \(^{1\dagger}\) , Muhua Chen \(^{1\dagger}\) , Bo Yang \(^{1}\) , Ziyi Guo \(^{1}\) , Mingyu Xu \(^{1}\) , Yunjie Zhang \(^{1}\) , Huaming Sun \(^{1}\) , Jingshuang Dang \(^{1}\) , Juan Fan \(^{1}\) , Jing Li \(^{1*}\) and Junfa Wei \(^{1*}\) \(^{1}\) School of Chemistry and Chemical Engineering, Shaanxi Normal University, Xi'an, 710119, China.  
+
+\*Corresponding author(s). E- mail(s): li_jing@snnu.edu.cn; weijf@snnu.edu.cn; †These authors contributed equally to this work.  
+
+## Abstract  
+
+Synthesis of buckybowls, especially, the large sized ones, have remained a huge challenge yet due to the inherent high strain induced by curvature. Herein, we report two novel bowl- shaped polycyclic aromatics with three chalcogen (sulfur or selenium) atoms and three methylene groups embedded at the bay regions of hexa- peri- hexabenzocoronene and an expeditious three- step synthetic strategy for these superbowls, including an alold cyclotrimerization, a Scholl reaction, and a Stille reaction. The superbowls of this type feature a nanosized, compact, and \(C_{3v}\) symmetric architecture, composing of 19 fused rings, 48 constituent atoms. NMR spectroscopic and X- ray crystallographic studies confirmed their bowl- shaped geometries. The crystal structures revealed that they encompass 36 pyramidalized trigonal carbon atoms and have the bowl depths of 2.29 Å and 2.16 Å and diameters of 11.06 Å and 11.35 Å for the sulfur and selenium isologs. The curvature mainly distribute at the carbon atoms of the coronene frame and edge- to- convex packing predominates due to intermolecular C- H···π and chalcogen···π interactions in two instances. Variable temperature \(^{1\dagger}\) H NMR experiments and theoretical calculations demonstrated the bowls have considerably high inversion barriers. The optical and electrochemical properties were elucidated by UV/vis and fluorescence spectroscopy and cyclic voltammetry. Moreover, the aromaticity distribution and electrostatic potential characteristics as well as perpendicularly aligned convex- to- concave dipolomements were investigated by density functional theory calculations.  
+
+Bowl- shaped polycyclic aromatic hydrocarbons (PAHs), often referred as buckybowls or \(\pi\) - bowls, have emerged as attractive targets that captivate scientists from chemistry to materials science in light of their intriguing characteristics as well  
+
+as potential applications in a diverse of scientific fields [1- 10]. More significantly, some of the buckybowls could also serve as templates or seeds for the growth of single walled carbon nanotubes [11- 13] having a controlled chirality and diameter and
+
+<--- Page Split --->
+
+thus having uniform electronic properties useful in molecular electronic devices [14]. Owing to the huge inner strain of bowl- shaped hydrocarbons makes their synthesis a major challenge. Hitherto the most literature- known buckyballs were related to \(\mathrm{C}_{60}\) or \(\mathrm{C}_{70}\) fullerene fragments and their \(\pi\) - extended or heteroatoms doping derivatives [15- 29]. The syntheses of large- sized heteroatom- doped buckyballs derivatives remain limited.  
+
+In the past decades, our group has been involved in design and synthesis of new PAHs and, naturally, bowl shaped polyarenes have never escaped from our consideration. It is of course interesting to imagine that if each of all six bay regions of hexa- peri- hexabenzocoronene ( \(p\) - HBC, also known as "superbenzene") were bridged by one divalent group, it should formally constitute a large, compact, highly symmetric, zigzag rimed, and bowl- shaped architecture composed of 48 atoms and 19 rings in its bowl system (Fig. 1). Such unprecedented molecules feature a coronene core successively circumscribed by alternate hexagonal and pentagonal rings. Simple Chem3D simulation and normative density functional theory (DFT) calculation indicates a beautiful bowl- shaped geometry for these coronene- based polyarenes (i.e., X, \(\mathrm{Y} = \mathrm{C}\) , S, Se etc.). However, experimentally bringing these imagined molecules into existence by chemical synthesis represents a gigantic challenge. No precedent has  
+
+been known for bridging all six bay regions of \(p\) - HBCs to form decorated pentagonal rings, albeit methodologies for bridging the bays of triphenylene [27, 28] and perylene [24, 29, 30] has been established. Very recently, Mullen and Feng et al. disclosed an elegant synthesis of trisulfur annulated \(p\) - HBC derivatives in which only three five- membered rings form in bay regions, emphasizing the difficulty to construct such a curved structure with strain [31].  
+
+![](images/Figure_1.jpg)
+
+<center>Fig. 1 Our synthetic concept of bowls from \(p\) -HBC. </center>  
+
+To address this daunting challenge, we proposed a new synthetic concept based on the following philosophy: three of all six five- membered rings for these coronene- based bowls are provided by fluorene moieties; while the remaining three pentagonal rings should be constructed by inserting three chalcogen atoms in the bay positions preexisting chlorine atoms as closable functionalities with the concomitant introduction of curvature. Fig. 2 depicts our three- step synthetic route for constructing these highly strained geodesic polyarenes based on this strategy.  
+
+Our synthetic campaign was commenced with a aldol cyclotrimerization of 1- (9,9- dibutyl- 2,7- dichloro- 9H- fluoren- 4- yl)ethan- 1- one (2), which
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Fig. 2 The synthetic route for the superbowls. Reagents and conditions: (i) 6.6 equiv. \(\mathrm{TiCl_4}\) , \(\alpha\) -dichlorobenzene, (microwave), \(180^{\circ}\mathrm{C}\) , 3 h; (ii) 8 equiv. DDQ, \(5\%\) TfOH, 1,2-dichloroethane, \(50^{\circ}\mathrm{C}\) , 6 h; (iii) 4 equiv. \((\mathrm{Bu}_3\mathrm{Sn})_2\mathrm{S}\) or \((\mathrm{Bu}_3\mathrm{Sn})_2\mathrm{Se}\) , 1 equiv. \(\mathrm{Pd(PPh}_3)_4\) , toluene, \(150^{\circ}\mathrm{C}\) , Ar, 48 h. </center>  
+
+was prepared from commercially available 2,7- dichloro- 9H- fluorene in three steps (see the Supplementary Information for details). The initial attempts towards the cyclotrimerization of 2 using \(\mathrm{SiCl}_4\) [32] as catalyst in ethanol failed; only a trace amount of the desired product, 1,3,5- tris(9,9- dibutyl- 2,7- dichloro- 9H- fluoren- 4- yl)benzene (TFB, 3), was detected in the reaction mixture by MS spectroscopy. Fortunately, several explorations rewarded us with finding proper conditions to access the cyclotrimer 3 by performing the reaction in \(\mathrm{o}\) - dichlorobenzene ( \(\mathrm{o}\) - DCB) at \(180^{\circ}\mathrm{C}\) under microwave irradiation using \(\mathrm{TiCl}_4\) as catalyst [33, 34]. This venerable method, albeit under slightly harsh conditions, delivered the product 3 in \(52\%\) isolated yield after chromatographical purification. Noticeably, TFB has the carbon structure that constitutes the entire carbon scaffold of final bowls and the functionalities necessary for the formation of the five- membered rings.  
+
+Also fortunately and delightfully, the consequent Scholl reaction of 3 with DDQ/triflic acid in DCE at \(50^{\circ}\mathrm{C}\) , which is the crucial step toward the total synthesis of the designed bowls, afforded the hunted trifluorenocoronene (TFC, 4) in an isolated yield of \(37\%\) . Although not high, the achieved yield is quite reasonable if considering the high ring strain from three five- membered rings and steric hindrance of two chlorine atoms at each bay region in the product 4. The preinstalled \(n\) - butyl groups endow TFC with adequate solubility amenable to isolation and full spectroscopic characterization. All analytical data are consistent with the expected structure of TFC 4 (Supplementary Figs 52 and 53). We also briefly explored the improvement of Scholl reaction of 3 and found relatively satisfactory conditions, as denoted in Fig. 2.   
+
+With the requisite TFC (4) in hand, we then accomplished the total synthesis of the designed buckybowls via threefold heteroatom annulation at its dichlorinated bay positions using Stille type reaction, which has been established by Wang group in their synthesis of sulfur heterocyclic annulated perylene bisimide derivatives [35] with
+
+<--- Page Split --->
+
+some modifications. To our delight, we obtained successfully the hunted buckybowl, trithiahexain- denocoroneen 1a, as a yellowish powder in 58% isolated yield using \((\mathrm{Bu}_3\mathrm{Sn})_2\mathrm{S}\) as the sulfur donor and \(\mathrm{Pd(PPh_3)_4}\) as the catalyst at \(150^{\circ}\mathrm{C}\) in a sealed tube under Ar atmosphere. This is the first time that such a supersized buckybowl was achieved. The HR- MS spectrum showed the molecular peak at \(m / e\) 985.3936 consistent with the desired product \(\mathrm{(C_{60}H_{60}S_3 + H^+, 985.3930)}\) and also, the measured isotopic distribution was well coincident with that stimulated (Supplementary Fig. 56). The \(^1\mathrm{H}\) NMR spectrum (Fig. 3) presented only one sharp low field singlet at 8.06 ppm for the six equivalent aromatic protons; while the high field signals, which are well resolved, implied two inequivalent sets of \(n\) - butyl groups. These observations strongly suggest its 3- fold symmetry and the bowl- shaped conformation in solution since only if access was gained to the bowl structure, the aliphatic chains can be differentiated by their location at regions demarcated by the convex and concave faces of the bowl. The chemical shifts of butyl chains inside the concave follow an order of \(\mathrm{H}_a > \mathrm{H}_c > \mathrm{H}_d > \mathrm{H}_b\) , similar to 9,9- dibutyl- 9H- fluorene [36]. While butyl chains outward the bowl follow a different order of \(\mathrm{H}_a > \mathrm{H}_b > \mathrm{H}_c > \mathrm{H}_d\) . Interestingly, two distinctive signals, \(\mathrm{H}_b\) and \(\mathrm{H}_d\) , assignable to methylene and methyl protons in the butyl moieties inward bowl orientations manifest negative chemical shifts \((- 1.04\) to \(- 0.02\) ppm), which can be ascribed to the stronger shielding effect caused by the ring current of the bowl system. The \(^{13}\mathrm{C}\) NMR data further support the bowl topological character of 1a (Supplementary Fig. 55).  
+
+inward bowl orientations manifest negative chemical shifts \((- 1.04\) to \(- 0.02\) ppm), which can be ascribed to the stronger shielding effect caused by the ring current of the bowl system. The \(^{13}\mathrm{C}\) NMR data further support the bowl topological character of 1a (Supplementary Fig. 55).  
+
+![](images/Figure_3.jpg)
+
+<center>Fig. 3 \(^1\mathrm{H}\) NMR characterization of 1a ( \(\mathrm{CDCl}_3\) ) and geometry predicted by DFT calculations. </center>  
+
+The selenium version was also synthesized by bridging the bays with selenium atoms under the same conditions, but switching \((\mathrm{Bu}_3\mathrm{Sn})_2\mathrm{S}\) to \((\mathrm{Bu}_3\mathrm{Sn})_2\mathrm{Se}\) [37], as a yellowish powder in 39% isolated yield. Similarly, all spectroscopic data (Supplementary Figs 57 to 59) are well consistent with the expected structure of triselenohexainedenocoroneen 1b. It should be pointed out that its \(^1\mathrm{H}\) NMR and \(^{13}\mathrm{C}\) NMR spectra show virtually the same pattern to the sulfur analog. Furthermore, the corresponding \(\mathrm{CH}_2(\mathrm{b})\) and \(\mathrm{CH}_3(\mathrm{d})\) signals at higher field \((- 0.96\) to \(0.01\) ppm) due to the weaker shielding effect, reflecting that the bowl is flatter than its sulfur analog.
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Fig. 4 a,d, X-ray crystal structures of 1a (a) and 1b (d). b,e, The side view of 1a (b) and 1b (e) with diameters and depths. c,f, The top view of 1a (c) and 1b (f) together with POAV angles (blue) and mean bond lengths (red). Butyl groups in the side and top views are omitted for clarity. Thermal ellipsoids are shown at 30% probability. </center>  
+
+Single crystals of 1a and 1b suitable for X- ray diffraction studies were grown by slow vapor diffusion of methanol into solution of chloroform at room temperature. The X- ray structural analyses unambiguously confirm the anticipated bowl- shaped architecture (Figs 4(a) and 4(d)), with an approximate \(C_{3v}\) - like symmetry in both crystals; the deviations from ideal symmetry should be blamed to the crystal packing forces in crystalline state [38]. The maximum bowl depths and diameters are 2.29 Å and 11.06 Å for 1a while 2.16 Å and 11.35 Å for 1b, defined by the perpendicular distance between the plane of three peripheral  
+
+heteroatoms and the centroid of the hub and by the horizontal distance between saturated carbons and heteroatoms at opposite vertexes, respectively (Figs 4(b) and 4(e)). Deservedly, the selenium bowl is slightly shallower than its sulfur isolog due to its bigger atomic radius. It is worth mentioning that the depths and widths as well as other structural parameters of the bowls are closely aligned with those of DFT calculations (Supplementary Figs 25 and 31).  
+
+Both 1a and 1b crystallize in \(\mathrm{P2_1 / c}\) space group with a disordered chloroform molecule filling over each hub ring of concave side by \(\mathrm{C - H\cdots\pi}\)
+
+<--- Page Split --->
+
+or Cl··π interactions with the formation of chloroform in bowl supermolecular complex isologs (Supplementary Figs 1 and 11). Normal concaveconvex stacking fashion is restricted due to the presence of three pairs of \(n\) - butyl groups. Note that edge- to- convex predominant stacking manner stabilized with C- H··π and chalcogen··π interactions was observed (Supplementary Figs 3 and 13). Both bowls exhibit concave to concave packing motifs along \(a\) axis via six C- H··chalcogen hydrogen bonds between chalcogen atoms and hydrogen atoms \(\mathrm{(H_{d})}\) of methyl groups inward bowls (Supplementary Figs 4, 5, 14 and 15). Notably, 1a contains three crystallographically independent molecules while the selenium isolog 1b contains two crystallographically independent molecules in unit cell.  
+
+As shown in Figs 4(c) and 4(f), there are 4 circles, totally 36 different pyramidalized trigonal carbon atoms existed in these coronene- based superbowls, except for 18 atoms at the peripheral vertexes. The \(\pi\) - orbital axis vector (POAV) angle analyses [39, 40] show that the carbon atoms of the hub ring of two bowls are moderately curved with a mean POAV angle of \(5.40^{\circ}\) (1a) and of \(4.79^{\circ}\) (1b); the most curvature occurs at the carbon atoms making up the coronene rim, of which the mean POAV angles are \(6.26^{\circ}\) (1a) and \(5.85^{\circ}\) (1b) for carbon atoms fusion with cyclopentadiene rings. The carbon atoms linked to each hub ring are pyramidalized significantly with a mean  
+
+POAV angle of \(6.14^{\circ}\) (1a) and of \(5.61^{\circ}\) (1b), the secondary maximum value of those of the carbons in respective molecule. The curvature is even distributed among the peripheral carbons attached to heteroatom and to the saturated carbon atom of cyclopentadiene ring, of which POAV angles are \(3.91^{\circ}\) (1a) and \(3.55^{\circ}\) (1b) and \(3.73^{\circ}\) (1a) and \(3.60^{\circ}\) (1b), respectively (see the Supplementary Information for details).  
+
+The bond lengths, as denoted in Figs 4(c) and 4(f) (red colored data), are the symmetry- averaged values of all equivalent bonds around their respective \(C_{3}\) symmetric axis disregard deviations from perfect geometry. The hub ring in both molecules remains an almost regular hexagon with a negligible bond length alternation (BLA) of \(0.002 \mathrm{\AA}\) (1a) and \(0.001 \mathrm{\AA}\) (1b) and equalized yet elongated bond lengths of \(1.422\) and \(1.424 \mathrm{\AA}\) (1a) or \(1.421\) and \(1.422 \mathrm{\AA}\) (1b) in comparison with that in benzene \((1.40 \mathrm{\AA})\) , indicating their reduced aromatic bond character. The radial bonds joining the hub to the rim hexagons are elongated to \(1.450 \mathrm{\AA}\) (1a) and \(1.447 \mathrm{\AA}\) (1b), considerably longer than the others in the same ring and thus indicative of quasi- single bond character. These bonds are not components of aromatic rings and function only as geometric connections between the hub and outer rings. Accordingly, the six- membered rings containing the radial bonds deviate markedly from regular hexagon. Six rim benzene rings in both compounds are non- negligibly
+
+<--- Page Split --->
+
+irregular hexagons with the bond lengths spanning from 1.386 to 1.420 Å (1a) and from 1.389 to 1.420 Å (1b), falling within the scope of aromatic bond character. The C- S bond (1.779 Å) in 1a and C- Se bond (1.923 Å) in 1b are fairly longer than that in dibenzothiophene (1.744 Å) [41] and in dibenzoselenophene (1.895 Å) [42], indicating also declined aromatic bond character at the corresponding positions. These results lend support to that the peripheral and hub benzene rings possess more benzenoid character, consistent with the deductions from BLA analysis [43] based on the X- ray determined (Supplementary Fig. 9 and 19) and DFT calculated structural data (Supplementary Figs 26 and 32). In addition, the fluorene moieties show a C- C \(_{sat}\) . bond length of 1.559 Å (1a) and 1.560 Å (1b), longer than that in 9,9- dioctyl- 9H- fluorene (1.523 Å) [44], might due to the high inner strain of their bowl geometries.  
+
+The aromaticity distribution characteristics in the bowl systems were further evaluated by nucleus independent chemical shift (NICS) [45- 47] and anisotropy of the induced current density (ACID) [48] calculations on their unsubstituted analogs 1a' and 1b'. The calculated NICS(1) \(_{zz}\) values, as denoted in the corresponding rings (Figs 5(a) and 5(b)), suggest that the hub hexagons (ring a) and the six outer hexagons (e) hold a pronounced aromatic character while the hetero pentagons (d) and the hexagons (c) are  
+
+faintly aromatic. The hexagons (b), which possess positive average NICS(1) \(_{zz}\) value values, present a feeble antiaromatic character. Therefore, the aromaticity distribution in each bowl system is are nearly, yet not completely similar to that in \(p\) - HBC [49], as shown by the blue shaded benzene rings (Figs 5(a) and 5(b)). Such aromaticity distribution characteristics is also consistent with Clar sextet rule [50] because the largest number of aromatic sextets can be found in the resonance hybrid when aromatic sextets locate at rings (a and e) overwhelm rings (d and c)(Supplementary Figs 30 and 36)). ACID plots of 1a' and 1b' revealed that clockwise 6π- electron local current pathways are presented in the hub (a) and rim (e) benzene rings. On the contrary, no significant local currents were observed in rings (c and d) of neither bowl (Figs 5(c) and 5(d)). Rings (b) showed unnoticeable anticlockwise ring currents, reflecting their weak antiaromatic character. Remarkably, a clockwise global current was observed along 30π- electron pathway consisting of the lone pair electrons on S or Se atoms, which are caused by the superposition of these local currents in each molecule. Taken together, these outcomes lend further supports to the aromaticity distribution characteristics obtained by NICS calculations.  
+
+The electrostatic potential (ESP) [51, 52] of 1a' and 1b' were also calculated (Figs 6(a) and 6(b) for 1a', Figs 6(d) and 6(e) for 1b'). Unlike planar PAHs containing two identical faces, the
+
+<--- Page Split --->
+![](images/Figure_5.jpg)
+
+<center>Fig. 5 a,b, \(\mathrm{NICS}(1)_{zz}\) values (ppm) of \(1\mathbf{a}^{\prime}\) (a) and \(1\mathbf{b}^{\prime}\) (b) calculated at the GIAO B3LYP/6-311G(d,p) level of theory. c,d, Calculated AICD plots (isovalue \(= 0.03\) ) of \(1\mathbf{a}^{\prime}\) (c) and \(1\mathbf{b}^{\prime}\) (d). Only contributions from \(\pi\) -electrons of the aromatic cores are considered. Red arrows indicate directions of induced ring current. </center>  
+
+ESP maps of these bowl- shaped compounds illustrate that the convex faces display more negative values compared with their concave faces, corresponding to the directions of the intrinsic dipole moment from the convex surface to the concave surface (Figs 6(c) and 6(f)). The calculated dipole moments are 3.59 Debye ( \(1\mathbf{a}^{\prime}\) ) and 3.45 Debye ( \(1\mathbf{b}^{\prime}\) ), much higher than those of coranlunene (2.19 Debye) [53] and surmarrene (2.7 Debye) [54], likely due to the enlarged \(\pi\) - surface and depth of these supersized bowls. In addition, the incorporation of thiophene or selenophene subunits enhances their concave- convex polarization effect.  
+
+Bowl to bowl inversion of two bowls were investigated by variable temperature \(^1\mathrm{H}\) NMR  
+
+measurements in 1,2- dichlorobenzene- \(d_4\) from 30- 180 °C (Supplementary Figs 38 and 39) and no alteration on chemical shifts was observed, reflecting their inability to undergo bowl- inversion at the temperatures screened and thus their high inversion barriers. The DFT calculations for inversion barriers of the unsubstituted \(1\mathbf{a}^{\prime}\) and \(1\mathbf{b}^{\prime}\) at the M062x/6- 31g(d) level of theory (Fig. 7) revealed that the inversion would proceed via a planar transition state like coranlunene [55- 58] and sumanene [59, 60], with inversion barriers \((\Delta G^{\ddagger})\) of 70.2 kcal/mol and 51.0 kcal/mol for \(1\mathbf{a}^{\prime}\) and \(1\mathbf{b}^{\prime}\) , respectively (see the Supplementary Information for details). Such high energy barriers suggest that the bowl- to- bowl inversion process might be impossible at ordinary temperatures according to Eying equation [61] (Supplementary Fig. 41). So, their bowl- shaped \(\pi\) - systems are conformationally locked at mild temperature.  
+
+To elucidate the optophysical properties of both sulfur and selenium isologs, their UV- vis absorption and fluorescence spectra were recorded in dichloromethane (Fig. 8(a)). Both absorption spectra displayed a similar pattern with the maximum absorbance at around 400 nm and a faint hump extending up 510 nm. These observations implied that the displacement of sulfur by selenium only changes the relative absorptive intensity rather than the positions of bands. The spectroscopic similarity of two isologs stems from their similar structures in terms of geometrical and
+
+<--- Page Split --->
+![](images/Figure_6.jpg)
+
+<center>Fig. 6 a,d, Calculated electrostatic potentials of bowl concave for \(1\mathbf{a}^{\prime}\) (a) and \(1\mathbf{b}^{\prime}\) (d). b,e, Calculated electrostatic potentials of bowl convex for \(1\mathbf{a}^{\prime}\) (b) and \(1\mathbf{b}^{\prime}\) (e). c,f, Dipole moments of \(1\mathbf{a}^{\prime}\) (c) and \(1\mathbf{b}^{\prime}\) (f). Electrostatic potentials calculated on the 0.001 au. isodensity surface together with surface maxima (blue) and minima points (red) at the B3LYP/6-311G(d,p) level of theory. </center>  
+
+![](images/Figure_7.jpg)
+
+<center>Fig. 7 a,b, Inversion barriers of \(1\mathbf{a}^{\prime}\) (a) and \(1\mathbf{b}^{\prime}\) (b) calculated at M062x/6-31G(d) level of theory. </center>  
+
+electronic aspects, as revealed by X- ray structural analysis and DFT calculations described above. TD- DFT calculations at the PBE/def2tzvp level for models \(1\mathbf{a}^{\prime}\) and \(1\mathbf{b}^{\prime}\) (Figs 8(c) and Figs 8(d), see the Supplementary Information for details) revealed that the maximum absorbance of \(1\mathbf{a}\) is caused by the \(\mathrm{S}_0\rightarrow \mathrm{S}_3\) and \(\mathrm{S}_0\rightarrow \mathrm{S}_4\) transitions of \(1\mathbf{a}^{\prime}\) which contain those from frontier MOs to  
+
+nondegenerate LOMO+2 ( \(f = 0.4286\) ) while the lowest energy absorption band can be attribute to the \(\mathrm{S}_0\rightarrow \mathrm{S}_1\) transition ( \(f = 0.0005\) ). The low intensity of \(\mathrm{S}_0\rightarrow \mathrm{S}_1\) transition is related to the degeneracy of frontier molecular orbitals involved in the transition. The same rationale could be applied to explain the UV- vis spectrum of \(1\mathbf{b}\)
+
+<--- Page Split --->
+![](images/Figure_8.jpg)
+
+<center>Fig. 8 a, UV-Vis absorption (solid curves) and emission spectra (dotted curves) of 1a (black) and 1b (red) in \(\mathrm{CH}_2\mathrm{Cl}_2\) at a concentration of \(1.0\times 10^{-5}\mathrm{M}\) . b, Cyclic voltammogram and differential pulse voltamogram of 1a and 1b in \(\mathrm{CH}_2\mathrm{Cl}_2\) (0.1 mol/L \(n\mathrm{-Bu}_4\mathrm{NPF}_6\) ) at a scan rate of 0.1 V/s. All potentials were calibrated versus an aqueous SCE by the addition of ferrocene as an internal standard taking \(\mathrm{E}_{1 / 2}(\mathrm{Fc} / \mathrm{Fc}^{+}) = 0.424\mathrm{V}\) vs. SCE [62]. c,d, Orbital correlation diagram and transition composition of the \(\mathrm{S}_0\rightarrow (\mathrm{S}_1,\mathrm{S}_3,\mathrm{S}_4)\) excited states for 1a' (c) and 1b' (d) calculated at PBE/def2tzvp level of theory. </center>  
+
+with similar molecular orbital energy level distribution in \(\mathbf{1b'}\) . The optical energy gaps were estimated to be \(2.43\mathrm{eV}\) (1a) and \(2.39\mathrm{eV}\) (1b) from the onset wavelength of lowest energy absorption edge (Table S7). The fluorescence spectra of both isologs featured distinctive vibronic structures with multiple maxima at 503, 515, 523, and 543 nm. The sulfur isolog 1a displayed a weak fluorescence while and selenium isolog 1b was found to be almost non- emissive due to stronger heavy atom effect of selenium than sulfur. These results  
+
+are explicable to that the degeneration from \(\mathrm{S}_1\) are symmetry forbidden, thus leading to the excited state relaxation via different vibronic levels of \(\mathrm{S}_0\) . The absolute quantum yield and fluorescence lifetime of 1a were measured to be \(4\%\) and \(1.9\mathrm{ns}\) and the radiative \((k_f)\) and nonradiative \((k_{nr})\) decay rate constants were calculated from the singlet excited state to be \(2.3\times 10^7\mathrm{s}^{- 1}\) and \(5.0\times 10^8\mathrm{s}^{- 1}\) , respectively.  
+
+Cyclic and differential pulse voltametric studies (Fig. 8(b)) revealed that both compounds
+
+<--- Page Split --->
+
+underwent two reversible oxidation waves and an irreversible oxidation wave corresponding to the three sequential one- electron oxidation processes of three heteroatoms, indicating that the generation of the cation radical during the first two oxidation processes are stable to repeated cycles, the third oxidation process is unstable under the selected electrochemical conditions. The half- wave potentials \(\mathrm{E}_{ox}^{1 / 2}\) of two reversible oxidation processes locate at \(0.90\mathrm{V}\) and \(1.35\mathrm{V}\) (1a) and \(0.93\mathrm{V}\) and \(1.38\mathrm{V}\) (1b) against SCE. Note that the oxidation potential of selenium isolog 1b is lower than that of sulfur isolog 1a due to the smaller electronegativity of selenium atoms. Their HOMO/LUMO energy levels were estimated from the onsets of oxidation and reduction to be \(- 5.25 / - 2.50\) (1a) and \(- 5.24 / - 2.50\mathrm{eV}\) (1b), corresponding to the electrochemical energy gaps of \(2.75\mathrm{eV}\) and \(2.74\mathrm{eV}\) , respectively (Table S8).  
+
+Thermogravimetric analysis (TGA) experiments were conducted to determine the thermal stability of both compounds. The results showed that the onset decomposition temperatures of both 1a and 1b are above \(400^{\circ}\mathrm{C}\) (Supplementary Fig. 24), reflecting their thermal robustness.  
+
+In conclusion, we have paved a new synthetic strategy and, based on which, a concise 3- step approach for expeditiously access to large, compact, and symmetric buckybowls that underscore the following features: 1) a bowl- like architecture  
+
+nanosized, 19 rings, 48 atoms, and 36 pyramidalized trigonal carbon atoms; 2) a coronene core fully circumscribed by alternate hexagonal and pentagonal rings. Single- crystal X- ray crystallography and NMR spectroscopy clearly confirmed their bowl- shaped geometry in crystal and in solution. DFT calculations and variable temperature \(^1\mathrm{H}\) NMR experiments suggest their much high inversion barriers. Quantum chemical calculations also reveal their aromaticity distribution and electrostatic potential characteristics. Distinguishing from most known buckybowls, we present the first instance containing coronene core as the bowl bottom with complete edge topological structure. The success on the synthesis of these previously unknown superbowls delivered a new family of \(\pi\) - bowls and a new synthetic strategy and the most important of all, verified accessibility of bridging all bays of \(p\) - HBCs. We believed that it will promote more synthetic efforts towards their all- carbon parent, other hetero versions as well as the analogs based on other PAHs with bay regions.  
+
+## Data availability  
+
+X- ray crystallographic data for compounds 1a and 1b are freely available from the Cambridge Crystallographic Data Centre (CCDC 2205194 and 2205195, respectively).
+
+<--- Page Split --->
+
+## Methods  
+
+Preparation of 3. In a glove box, 1- (9,9- dibutyl- 2,7- dichloro- 9H- fluoren- 4- yl)ethan- 1- one 2 (390 mg, 1.0 mmol, 1.0 equiv.), \(\mathrm{TiCl}_4\) (0.7 mL, 6.6 mmol, 6.6 equiv.) and 5 mL dry o- dichlorobenzene were added to a \(10\mathrm{mL}\) microwave vial and the vial was capped. The vessel was removed from the glove box and placed into a microwave reactor where it was heated at \(180^{\circ}\mathrm{C}\) for \(3\mathrm{h}\) . After cooling down, the mixture was poured over concentrated hydrochloric acid/ice to quench the reaction and then extracted with \(\mathrm{CH}_2\mathrm{Cl}_2\) . The combined organic phase was washed with saturated aqueous \(\mathrm{NaHCO}_3\) and dried over \(\mathrm{Na}_2\mathrm{SO}_4\) . The organic solvent was removed at reduced pressure to give a yellow- brown solid. The above procedure was repeated 4 times and the crude material from each reaction was combined. The combined crude material was purified by column chromatography over silica gel (eluent: petroleum ether) to afford 1,3,5- Tris(9,9- dibutyl- 2,7- dichloro- 9H- fluoren- 4- yl)benzene 3 (970 mg, 52%) as an off- white powder; melting point (m.p.), \(147–148^{\circ}\mathrm{C}\) ; \(^1\mathrm{H}\) NMR (600 MHz, \(\mathrm{CDCl}_3\) ): \(\delta 7.67\) (d, \(J = 14.9\mathrm{Hz}\) , 3H), 7.32 (s, 6H), 7.29 (d, \(J = 1.9\mathrm{Hz}\) , 3H), 7.12 (d, \(J = 8.2\mathrm{Hz}\) , 3H), 6.89 (d, \(J = 8.0\mathrm{Hz}\) , 3H), 1.98- 1.91 (m, 12H), 1.13- 1.07 (m, 12H), 0.69 (t, \(J = 7.4\mathrm{Hz}\) , 18H), 0.60 (d, \(J = 7.5\mathrm{Hz}\) , 12H); \(^{13}\mathrm{C}\) NMR (150 MHz, \(\mathrm{CDCl}_3\) ): \(\delta 153.6\) , 153.2, 138.3, 137.3, 133.3, 132.9, 129.2, 128.9, 126.7, 123.3, 123.1, 122.6, 55.0, 40.3, 25.8, 22.9, 13.7; HRMS (APCI): \(m / z\) calcd for \(\mathrm{C}_{69}\mathrm{H}_{73}\mathrm{Cl}_6\) (M+H) \(^+\) : 1111.3838, found: 1111.3836.  
+
+129.2, 128.9, 126.7, 123.3, 123.1, 122.6, 55.0, 40.3, 25.8, 22.9, 13.7; HRMS (APCI): \(m / z\) calcd for \(\mathrm{C}_{69}\mathrm{H}_{73}\mathrm{Cl}_6\) (M+H) \(^+\) : 1111.3838, found: 1111.3836.  
+
+Preparation of 4. To a mixture of 3 (220 mg, 0.2 mmol, 1.0 equiv.) and DDQ (360 mg, 1.6 mmol, 8.0 equiv.) in 1,2- dichloroethane (20 mL) was added trifluoromethanesulfonic acid (1 mL) under argon atmosphere, and the mixture was stirred at \(50^{\circ}\mathrm{C}\) for \(6\mathrm{h}\) . After cooling to room temperature, the reaction mixture was quenched by adding saturated aqueous \(\mathrm{NaHCO}_3\) , and then the mixture was extracted with \(\mathrm{CH}_2\mathrm{Cl}_2\) . The combined organic layer was dried over \(\mathrm{Na}_2\mathrm{SO}_4\) , and the solvent was removed under reduced pressure. The crude product was purified by silica gel column chromatography (eluent: petroleum ether) to give the trifluorocoronene 4 (80 mg, 37%) as a yellow powder; m.p. \(>300^{\circ}\mathrm{C}\) ; \(^1\mathrm{H}\) NMR (600 MHz, \(\mathrm{Tol - d}_8\) ): \(\delta 8.28\) (s, 6H), 2.26 (dd, \(J = 9.8\) , 6.6 Hz, 12H), 1.16 (dd, \(J = 14.6\) , 7.3 Hz, 12H), 1.08- 1.02 (m, 12H), 0.66 (t, \(J = 7.3\mathrm{Hz}\) , 18H); \(^{13}\mathrm{C}\) NMR (150 MHz, \(\mathrm{Tol - d}_8\) ): \(\delta 151.7\) , 135.0, 133.9, 127.3, 126.9, 126.1, 122.0, 63.1, 38.8, 27.9, 23.9, 14.4; HRMS (APCI): \(m / z\) calcd for \(\mathrm{C}_{69}\mathrm{H}_{61}\mathrm{Cl}_6\) (M+H) \(^+\) : 1099.2908, found: 1099.2898.  
+
+Preparation of 1a and 1b. Inside the glove- box, to an oven- dried pressure vessel with a Teflon screw cap was added compound 4 (110 mg,
+
+<--- Page Split --->
+
+0.1 mmol, 1.0 equiv.), \(\mathrm{Bu_3SnSSnBu_3}\) (245 mg, 0.4 mmol, 4.0 equiv.) and \(\mathrm{Pd(PPh_3)_4}\) (115 mg, 0.1 mmol, 1.0 equiv.) and 5 mL dry and degassed toluene. After the vessel was resealed and moved out from the glovebox, the mixture was heated and stirred at \(150^{\circ}\mathrm{C}\) for \(48\mathrm{h}\) . On cooling to room temperature, the reaction mixture was quenched with saturated aqueous KF solution and extracted with \(\mathrm{CH_2Cl_2}\) . The combined organic phase was dried over \(\mathrm{Na_2SO_4}\) before being filtered and concentrated down to a solid under reduced pressure. The crude solid was adsorbed onto silica gel and subjected to silica gel column chromatography (eluent, petroleum ether) to afford trithiahexaindenocoronene 1a (57 mg, \(58\%\) yield) as a yellowish powder; m.p. \(>300^{\circ}\mathrm{C}\) ; Triselenohexaindenocoronene 1b was obtained under the similar procedure by using \(\mathrm{Bu_3SnSeSnBu_3}\) which was prepared according to literature procedure [39]. Characterization data of 1a: \(^1\mathrm{H}\) NMR (600 MHz, \(\mathrm{CDCl_3}\) ): \(\delta\) 8.06 (s, 6H), 2.70- 2.66 (m, 6H), 2.13- 2.09 (m, 6H), 2.05- 2.00 (m, 6H), 1.58- 1.56 (m, 6H), 1.05 (t, \(J = 7.4\mathrm{Hz}\) , 9H), 0.59 (dd, \(J = 14.7\) , 7.4 Hz, 9H), \(- 0.02\) (t, \(J = 7.3\mathrm{Hz}\) , 9H), \(- 1.01\mathrm{- }1.07\) (m, 6H). \(^{13}\mathrm{C}\) NMR (150 MHz, \(\mathrm{CDCl_3}\) ): \(\delta\) 153.0, 142.9, 139.8, 134.8, 130.8, 128.6, 119.6, 64.4, 42.1, 35.8, 28.7, 26.1, 23.6, 22.5, 14.3, 13.2; HRMS (APCI): \(m / z\) calcd for \(\mathrm{C_{69}H_{61}S_3(M + H)^+}\) : 985.3930, found: 985.3936. Characterization data of 1b: \(^1\mathrm{H}\) NMR (600 MHz, \(\mathrm{CDCl_3}\) ): \(\delta\) 8.21 (s, 6H),  
+
+\(2.71 - 2.66\) (m, 6H), 2.17- 2.12 (m, 6H), 2.02- 1.98 (m, 6H), 1.57- 1.54 (m, 6H), 1.04 (t, \(J = 7.4\mathrm{Hz}\) , 9H), 0.59 (dd, \(J = 14.7\) , 7.4 Hz, 6H), 0.01 (t, \(J = 7.3\mathrm{Hz}\) , 9H), \(- 0.93 - 0.98\) (m, 6H). \(^{13}\mathrm{C}\) NMR (150 MHz, \(\mathrm{CDCl_3}\) ): \(\delta\) 151.8, 141.9, 139.0, 136.0, 129.3, 128.2, 121.6, 64.4, 41.5, 36.5, 28.5, 25.8, 23.6, 22.4, 14.3, 13.2; MS (MALDI- TOF): m/z calcd for \(\mathrm{C_{69}H_{60}Se_3(M)^+}\) : 1126.1, found: 1126.4.  
+
+## Acknowledgments  
+
+J. W. thanks the National Science Foundation of China (NSFC) (Grant Nos. 21871169); J. L. thanks the National Science Foundation of China (NSFC) (Grant Nos. 21702131); Y. S. thanks the Fundamental Research Funds for the Central Universities (SNNU 2019TS040).  
+
+## Author contributions  
+
+Y. S., X. W. and M. C. conducted the experiments with help from B. Y., M. X. and Y. Z., Y. S. performed the most DFT calculations. Z. G. and J. D. performed the inversion barrier calculations. H. S. collected and processed X-ray diffraction data. J. F. collected high-resolution mass spectrometry. J. W. conceived the original idea, designed and supervised the whole studies. J. W., Y. S. and J. L. performed the data analysis and wrote the manuscript with feedback from others. All authors discussed the results and commented on the manuscript.
+
+<--- Page Split --->
+
+## Additional information  
+
+Supplementary information and chemical compound information are available in the online version of the paper. Reprints and permissions information is available online at www.nature.com/reprints. Correspondence and requests for materials should be addressed to J. L. and J. W.  
+
+## Competing financial interests  
+
+The authors declare no competing financial interests.  
+
+## References  
+
+[1] Wu, Y.- T. & Siegel, J. S. Aromatic molecular bowl hydrocarbons: synthetic derivatives, their structures, and physical properties. Chem. Rev. 106, 4843- 4867 (2006)  
+
+[2] Gao, G. et al. Rational functionalization of a \(\mathrm{C}_{70}\) buckybowl to enable a \(\mathrm{C}_{70}\) : buckybowl cocrystal for organic semiconductor applications. J. Am. Chem. Soc. 142, 2460- 2470 (2020)  
+
+[3] Saito, M., Shinokubo, H., & Sakurai, H. Figuration of bowl- shaped \(\pi\) - conjugated molecules: properties and functions. Mater. Chem. Front. 2, 635- 661 (2018)  
+
+[4] Fu, B. et al. A bowl- shaped sumanene derivative with dense convex- concave columnar packing for high- performance organic field- effect transistors. Chem. Commun. 53, 11407- 11409 (2017)  
+
+[5] Chen, R. et al. Corannulene derivatives with low LUMO levels and dense convex- concave packing for n- channel organic field- effect transistors. Chem. Commun. 51, 13768- 13771 (2015)  
+
+[6] Lu, R.- Q. et al. Thiophene- fused bowl- shaped polycyclic aromatics with a dibenzo[a,g]corannulene core for organic field- effect transistors. Chem. Commun. 51, 1681- 1684 (2015)  
+
+[7] Shi, K., Lei, T., Wang, X. Y., Wang, J. Y. & Pei, J. A bowl- shaped molecule for organic field- effect transistors: crystal engineering and charge transport switching by oxygen doping. Chem. Sci. 5, 1041- 1045 (2014)  
+
+[8] Gallego, M. et al. Electron transfer in a supramolecular associate of a fullerene fragment. Angew. Chem. Int. Ed. 53, 2170- 2175 (2014)  
+
+[9] Amaya, T. et al. Anisotropic electron transport properties in sumanene crystal. J. Am. Chem. Soc. 131, 408- 409 (2009)
+
+<--- Page Split --->
+
+[10] Miyajima, D. et al. Liquid crystalline corannulene responsive to electric field. J. Am. Chem. Soc. 131, 44- 45 (2009)  
+
+[11] Scott, L. T. et al. A short, rigid, structurally pure carbon nanotube by stepwise chemical synthesis. J. Am. Chem. Soc. 134, 107- 110 (2012)  
+
+[12] Fort, E. H., Donovan, P. M. & Scott, L. T. Diels- alder reactivity of polycyclic aromatic hydrocarbon bay regions: implications for metal- free growth of single- chirality carbon nanotubes. J. Am. Chem. Soc. 131, 16606- 16607 (2009)  
+
+[13] Sanchez- Valencia, J. R. et al. Controlled synthesis of single- chirality carbon nanotubes. Nature 512, 61- 64 (2014)  
+
+[14] Cao, Q. et al. Arrays of single- walled carbon nanotubes with full surface coverage for high- performance electronics. Nat. Nanotechnol. 8, 180- 186 (2013)  
+
+[15] Barth, W. E. & Lawton, R. G. Dibenzo[ghi,mno]fluoranthene. J. Am. Chem. Soc. 88, 380- 381 (1966)  
+
+[16] Rabideau, P. W. et al. Buckybowls: synthesis and ab initio calculated structure of the first semibuckminsterfullerene. J. Am. Chem. Soc. 116, 7891- 7892 (1994)  
+
+[17] Sakurai, H., Daiko, T. & Hirao, T. A synthesis of sumanene, a fullerene fragment. Science 301, 1878 (2003)  
+
+[18] Petrukhina, M. A., Andreini, K. W., Peng, L. Q. & Scott, L. T. Hemibuckminsterfullerene \(\mathrm{C_{30}H_{12}}\) : X- ray crystal structures of the parent hydrocarbon and of the two- dimensional organometallic network \(\{\mathrm{[Rh_2(O_2CCF_3)_4]_3\cdot(C_30H_{12})}\}\) . Angew. Chem. Int. Ed. 43, 5477- 5481 (2004)  
+
+[19] Steinberg, B. D. et al. Aromatic \(\pi\) - systems more curved than \(\mathrm{C_{60}}\) . The complete family of all indenocorannulenes synthesized by iterative microwave- assisted intramolecular arylatons. J. Am. Chem. Soc. 131, 10537- 10545 (2009)  
+
+[20] Wu, T. C., Hsin, H. J., Kuo, M. Y., Li, C. H. & Wu, Y. T. Synthesis and structural analysis of a highly curved buckybowl containing corannulene and sumanene fragments. J. Am. Chem. Soc. 133, 16319- 16321 (2011)  
+
+[21] Wu, T. C., Chen, M. K., Lee, Y. W., Kuo, M. Y. & Wu, Y. T. Bowl- shaped fragments of \(\mathrm{C_{70}}\) or higher fullerenes: synthesis, structural analysis, and inversion dynamics. Angew. Chem. Int. Ed. 52, 1289- 1293 (2013)  
+
+[22] Chen, M. K. et al. Highly curved bowl- shaped fragments of fullerenes: synthesis, structural
+
+<--- Page Split --->
+
+analysis, and physical properties. Chem. Eur. J. 20, 598- 608 (2014)  
+
+[23] Liu, J. Z. et al. Fused dibenzo[a,m]rubicene: a new bowl-shaped subunit of \(\mathrm{C}_{70}\) containing two pentagons. J. Am. Chem. Soc. 138, 8364- 8367 (2016)  
+
+[24] Zou, Y. et al. Dicyclopenta[4,3,2,1- ghi:4',3',2',1'- pqr]perylene: a bowl- shaped fragment of fullerene \(\mathrm{C}_{70}\) with global antiaromaticity. J. Am. Chem. Soc. 141, 7266- 7270 (2019)  
+
+[25] Yokoi, H., Hiraoka, Y., Hiroto, S., Sakamaki, D., Seki, S. & Shinokubo, H. Nitrogen- embedded buckybowl and its assembly with \(\mathrm{C}_{60}\) . Nat. Commun. 6, 8215 (2015)  
+
+[26] Ito, S., Tokimaru, Y. & Nozaki, K. Benzene- fused azacorannulene bearing an internal nitrogen atom. Angew. Chem. Int. Ed. 54, 7256- 7260 (2015)  
+
+[27] Li, X. X. et al. Non- pyrolytic, large- scale synthesis of trichalcogenasumanene: a two- step approach. Angew. Chem. Int. Ed. 53, 535- 538 (2014)  
+
+[28] Tan, Q. T., Zhou, D. D., Zhang, T., Liu, B. X. & Xu, B. Iodine- doped sumanene and its application for the synthesis of chalcogenasumanenes and silasumanenes. Chem. Commun. 53, 10279- 10282 (2017)  
+
+Commun. 53, 10279- 10282 (2017)  
+
+[29] Huang, J. L. et al. Dicyclopenta[ghi,pqr]perylene as a structural motif for bowl- shaped hydrocarbons: synthetic and conformational studies. Org. Lett. 21, 2504- 2508 (2019)  
+
+[30] Jiang, W. et al. Solution- processed, high- performance nanoribbon transistors based on dithioperylene. J. Am. Chem. Soc. 133, 1- 3 (2011)  
+
+[31] Tan, Y. Z. et al. Sulfur- annulated hexaperi- hexabenzocoronene decorated with phenylthio groups at the periphery. Angew. Chem. Int. Ed. 54, 2927- 2931 (2015)  
+
+[32] Elmorsy, S. S., Pelter, A. & Smith, K. The direct production of tri- and hexa- substituted benzenes from ketones under mild conditions. Tetrahedron Lett. 32, 4175- 4176 (1991)  
+
+[33] Anseems, R. B. M. & Scott, L. T. Circumtrindene: a geodesic dome of molecular dimensions. rational synthesis of 60% of \(\mathrm{C}_{60}\) . J. Am. Chem. Soc. 122, 2719- 2724 (2000)  
+
+[34] Scott, L. T. et al. A rational chemical synthesis of \(\mathrm{C}_{60}\) . Science 295, 1500- 1503 (2002)  
+
+[35] Qian, H. L., Liu, C. M., Wang, Z. H. & Zhu, D. B. S- heterocyclic annulated perylene bisimide: synthesis and co- crystal with
+
+<--- Page Split --->
+
+pyrene. Chem. Commun. 4587- 4589 (2006)  
+
+[36] Ito, Y., Shimada, T., Ha, J., Vacha, M. & Sato, H. Synthesis and characterization of a novel electroluminescent polymer based on a phenoxazine derivative. J. Polym. Sci. Pol. Chem. 44, 4338- 4345 (2006)  
+
+[37] Maeda, H. et al. One- pot synthesis of seleno- nureas and selenocarbamates via selenium of isocyanates with bis(dimethylaluminum) selenide. Tetrahedron Lett. 52, 415- 417 (2011)  
+
+[38] Sharafi, M. et al. Crystal- packing- driven enrichment of atropoisomers. Angew. Chem. Int. Ed. 56, 7097- 7101 (2017)  
+
+[39] Haddon, R. C. and Scott, L. T. \(\pi\) - Orbital conjugation and rehybridization in bridged annulenes and deformed molecules in general: \(\pi\) - orbital axis vector analysis. Pure Appl. Chem. 58, 137- 142 (1986)  
+
+[40] Haddon, R. C. Comment on the relationship of the pyramidalization angle at a conjugated carbon atom to the sigma bond angles. J. Phys. Chem. A 105, 4164- 4165 (2001)  
+
+[41] Subhrajyoti B., Rik, D., Anna M. K. & Kristof H. Deformation- induced phosphorescence shift in a 2D elastically flexible organic single crystal: role of chalcogen-centered weak interactions. Chem. Sci. DOI:  
+
+10.1039/d2sc03729g (2022)  
+
+[42] Klapoetke, T. M., Krumm, B. & Scherr, M. Halogenation of dibenzoselenophene and dibenzo[1,2]diselenine Z. Anorg. Allg. Chem. 636, 1955- 1961 (2010)  
+
+[43] Bourhill, G et al. Experimental demonstration of the dependence of the first hyperpolarizability of donor- acceptor- substituted polyenes on the ground- state polarization and bond length alternation. J. Am. Chem. Soc. 116, 2619- 2620 (1994)  
+
+[44] McFarlane, S., McDonald, R. & Veinot, J.G.C. 9,9- Di- \(n\) - octyl- 9H- fluorene J. Phys. Chem. A 105, 4164- 4165 (2001)  
+
+[45] Chen, Z. F., Wannere, C. S., Corminboeuf, C., Puchta, R. & Schleyer, P. R. Nucleus- independent chemical shifts (NICS) as an aromaticity criterion. Chem. Rev., 105, 3842- 3888 (2005)  
+
+[46] Schleyer, P. R., Maerker, C., Dransfeld, A., Jiao, H. & Eikema Hommes, N. J. R. Nucleus- independent chemical shifts: a simple and efficient aromaficity probe. J. Am. Chem. Soc. 118, 6317- 6318 (1996)  
+
+[47] Fallah- Bagher- Shaidaei, H., Wannere, C. S., Corminboeuf, C., Puchta, R. & Schleyer, P. v. R. Which NICS aromaticity index for planar
+
+<--- Page Split --->
+
+pi rings is best? Org. Lett. 8, 863- 866 (2006)  
+
+[48] Geuenich, D., Hess, K., Köhler, F. & Herges, R. Anisotropy of the induced current density (ACID), a general method to quantify and visualize electronic delocalization. Chem. Rev. 105, 3758- 3772 (2005)  
+
+[49] Moran, D., Stahl, F., Bettinger, H. F., Schaefer III, H. F. & Schaefer, P. v. R. Towards graphite: Magnetic properties of large polybenzoic hydrocarbons. J. Am. Chem. Soc. 125, 6746- 6752 (2003)  
+
+[50] Clar, E. The aromatic sextet. Wiley (1972)  
+
+[51] Lu, T. & Chen, F. W. Multiwfn: a multifunctional wavefunction analyzer. J. Comput. Chem. 33, 580- 592 (2012)  
+
+[52] Lu, T. & Chen, F. W. Quantitative analysis of molecular surface based on improved marching tetrahedra algorithm. J. Mol. Graph. Model. 38, 314- 323 (2012)  
+
+[53] Biglari, Z., & Rezaei, H. Investigation into molecular properties of pristine corannulene and its analogs with B/N- doping at hub location. Comput. Theor. Chem. 1190, 113002 (2020)  
+
+[54] Shoji, Y. et al. Hexathioalkyl sumanenes: an electron- donating buckybowl as a building block for supramolecular materials. Chem.  
+
+Sci. 8, 8405- 8410 (2017)  
+
+[55] Seiders, T. J., Baldridge, K. K., Grube, G. H. & Siegel. J. S. Structure/energy correlation of bowl depth and inversion barrier in corannulene derivatives: combined experimental and quantum mechanical analysis. J. Am. Chem. Soc. 123, 517- 525 (2001)  
+
+[56] Scott, L. T., Hashemi, M. M. & Bratcher, M. S. Corannulene bowl- to- bowl inversion is rapid at room temperature. J. Am. Chem. Soc. 114, 1920- 1921 (1992)  
+
+[57] Sygula, A., Abdourazak, A. H. & Rabideau, P. W. Cyclopentacorannulene: \(\pi\) - facial stereoselective deutero generation and determination of the bowl- to- bowl inversion barrier for a constrained buckybowl. J. Am. Chem. Soc. 118, 339- 343 (1996)  
+
+[58] Sakurai, H., Daiko, T., Sakane, H., Amaya, T. & Hirao, T. Structural elucidation of sumanene and generation of its benzylic anions. J. Am. Chem. Soc. 127, 11580- 11581 (2005)  
+
+[59] Amaya, T., Sakane, H., Muneishi, T. & Hirao. T. Bowl- to- bowl inversion of sumanene derivatives. Chem. Commun. 765- 767 (2008)  
+
+[60] Amaya, T. & Hirao, T. A molecular bowl sumanene. Chem. Commun. 47, 10524- 10535
+
+<--- Page Split --->
+
+(2011)
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+SupplementaryInformation.pdf
+
+<--- Page Split --->

preprint/preprint__0ef88a3d250cc02e121a31263f7d3eb45880dc07208a92e0c8644d22f4cb03f6/preprint__0ef88a3d250cc02e121a31263f7d3eb45880dc07208a92e0c8644d22f4cb03f6_det.mmd

@@ -0,0 +1,468 @@
+<|ref|>title<|/ref|><|det|>[[44, 108, 805, 175]]<|/det|>
+# Trichalcogenohexaindenocoronenes: Super Buckyblows Based on Coronene Core  
+
+<|ref|>text<|/ref|><|det|>[[42, 195, 293, 750]]<|/det|>
+Yixun SunShaanxi Normal UniversityXin WangShaanxi Normal UniversityMuhua ChenShaanxi Normal UniversityBo YangShaanxi Normal UniversityZiyi GuoShaanxi Normal UniversityMingyu XuShaanxi Normal UniversityYunjie ZhangShaanxi Normal UniversityHuaming SunShaanxi Normal UniversityJingshuang DangShaanxi Normal UniversityJuan FanShaanxi Normal UniversityJing LiShaanxi Normal UniversityJunfa Wei ( weijf@snnu.edu.cn )Shaanxi Normal University  
+
+<|ref|>text<|/ref|><|det|>[[42, 789, 100, 805]]<|/det|>
+Article  
+
+<|ref|>title<|/ref|><|det|>[[44, 826, 135, 844]]<|/det|>
+# Keywords:  
+
+<|ref|>text<|/ref|><|det|>[[44, 864, 352, 882]]<|/det|>
+Posted Date: September 29th, 2022  
+
+<|ref|>text<|/ref|><|det|>[[42, 902, 473, 920]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 2095883/v1
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[42, 45, 910, 87]]<|/det|>
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[110, 150, 890, 179]]<|/det|>
+# Trichalcogenohexaindenocoronenes: Super Buckylbowls Based  
+
+<|ref|>text<|/ref|><|det|>[[384, 202, 616, 226]]<|/det|>
+on Coronene Core  
+
+<|ref|>text<|/ref|><|det|>[[106, 250, 894, 340]]<|/det|>
+Yixun Sun \(^{1\dagger}\) , Xin Wang \(^{1\dagger}\) , Muhua Chen \(^{1\dagger}\) , Bo Yang \(^{1}\) , Ziyi Guo \(^{1}\) , Mingyu Xu \(^{1}\) , Yunjie Zhang \(^{1}\) , Huaming Sun \(^{1}\) , Jingshuang Dang \(^{1}\) , Juan Fan \(^{1}\) , Jing Li \(^{1*}\) and Junfa Wei \(^{1*}\) \(^{1}\) School of Chemistry and Chemical Engineering, Shaanxi Normal University, Xi'an, 710119, China.  
+
+<|ref|>text<|/ref|><|det|>[[160, 370, 838, 409]]<|/det|>
+\*Corresponding author(s). E- mail(s): li_jing@snnu.edu.cn; weijf@snnu.edu.cn; †These authors contributed equally to this work.  
+
+<|ref|>sub_title<|/ref|><|det|>[[462, 439, 537, 454]]<|/det|>
+## Abstract  
+
+<|ref|>text<|/ref|><|det|>[[127, 457, 872, 712]]<|/det|>
+Synthesis of buckybowls, especially, the large sized ones, have remained a huge challenge yet due to the inherent high strain induced by curvature. Herein, we report two novel bowl- shaped polycyclic aromatics with three chalcogen (sulfur or selenium) atoms and three methylene groups embedded at the bay regions of hexa- peri- hexabenzocoronene and an expeditious three- step synthetic strategy for these superbowls, including an alold cyclotrimerization, a Scholl reaction, and a Stille reaction. The superbowls of this type feature a nanosized, compact, and \(C_{3v}\) symmetric architecture, composing of 19 fused rings, 48 constituent atoms. NMR spectroscopic and X- ray crystallographic studies confirmed their bowl- shaped geometries. The crystal structures revealed that they encompass 36 pyramidalized trigonal carbon atoms and have the bowl depths of 2.29 Å and 2.16 Å and diameters of 11.06 Å and 11.35 Å for the sulfur and selenium isologs. The curvature mainly distribute at the carbon atoms of the coronene frame and edge- to- convex packing predominates due to intermolecular C- H···π and chalcogen···π interactions in two instances. Variable temperature \(^{1\dagger}\) H NMR experiments and theoretical calculations demonstrated the bowls have considerably high inversion barriers. The optical and electrochemical properties were elucidated by UV/vis and fluorescence spectroscopy and cyclic voltammetry. Moreover, the aromaticity distribution and electrostatic potential characteristics as well as perpendicularly aligned convex- to- concave dipolomements were investigated by density functional theory calculations.  
+
+<|ref|>text<|/ref|><|det|>[[84, 759, 481, 886]]<|/det|>
+Bowl- shaped polycyclic aromatic hydrocarbons (PAHs), often referred as buckybowls or \(\pi\) - bowls, have emerged as attractive targets that captivate scientists from chemistry to materials science in light of their intriguing characteristics as well  
+
+<|ref|>text<|/ref|><|det|>[[517, 760, 916, 886]]<|/det|>
+as potential applications in a diverse of scientific fields [1- 10]. More significantly, some of the buckybowls could also serve as templates or seeds for the growth of single walled carbon nanotubes [11- 13] having a controlled chirality and diameter and
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[80, 73, 479, 308]]<|/det|>
+thus having uniform electronic properties useful in molecular electronic devices [14]. Owing to the huge inner strain of bowl- shaped hydrocarbons makes their synthesis a major challenge. Hitherto the most literature- known buckyballs were related to \(\mathrm{C}_{60}\) or \(\mathrm{C}_{70}\) fullerene fragments and their \(\pi\) - extended or heteroatoms doping derivatives [15- 29]. The syntheses of large- sized heteroatom- doped buckyballs derivatives remain limited.  
+
+<|ref|>text<|/ref|><|det|>[[80, 315, 479, 888]]<|/det|>
+In the past decades, our group has been involved in design and synthesis of new PAHs and, naturally, bowl shaped polyarenes have never escaped from our consideration. It is of course interesting to imagine that if each of all six bay regions of hexa- peri- hexabenzocoronene ( \(p\) - HBC, also known as "superbenzene") were bridged by one divalent group, it should formally constitute a large, compact, highly symmetric, zigzag rimed, and bowl- shaped architecture composed of 48 atoms and 19 rings in its bowl system (Fig. 1). Such unprecedented molecules feature a coronene core successively circumscribed by alternate hexagonal and pentagonal rings. Simple Chem3D simulation and normative density functional theory (DFT) calculation indicates a beautiful bowl- shaped geometry for these coronene- based polyarenes (i.e., X, \(\mathrm{Y} = \mathrm{C}\) , S, Se etc.). However, experimentally bringing these imagined molecules into existence by chemical synthesis represents a gigantic challenge. No precedent has  
+
+<|ref|>text<|/ref|><|det|>[[515, 73, 913, 337]]<|/det|>
+been known for bridging all six bay regions of \(p\) - HBCs to form decorated pentagonal rings, albeit methodologies for bridging the bays of triphenylene [27, 28] and perylene [24, 29, 30] has been established. Very recently, Mullen and Feng et al. disclosed an elegant synthesis of trisulfur annulated \(p\) - HBC derivatives in which only three five- membered rings form in bay regions, emphasizing the difficulty to construct such a curved structure with strain [31].  
+
+<|ref|>image<|/ref|><|det|>[[550, 351, 874, 448]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[515, 450, 877, 467]]<|/det|>
+<center>Fig. 1 Our synthetic concept of bowls from \(p\) -HBC. </center>  
+
+<|ref|>text<|/ref|><|det|>[[515, 499, 913, 817]]<|/det|>
+To address this daunting challenge, we proposed a new synthetic concept based on the following philosophy: three of all six five- membered rings for these coronene- based bowls are provided by fluorene moieties; while the remaining three pentagonal rings should be constructed by inserting three chalcogen atoms in the bay positions preexisting chlorine atoms as closable functionalities with the concomitant introduction of curvature. Fig. 2 depicts our three- step synthetic route for constructing these highly strained geodesic polyarenes based on this strategy.  
+
+<|ref|>text<|/ref|><|det|>[[515, 825, 912, 900]]<|/det|>
+Our synthetic campaign was commenced with a aldol cyclotrimerization of 1- (9,9- dibutyl- 2,7- dichloro- 9H- fluoren- 4- yl)ethan- 1- one (2), which
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[105, 70, 895, 211]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[83, 220, 916, 264]]<|/det|>
+<center>Fig. 2 The synthetic route for the superbowls. Reagents and conditions: (i) 6.6 equiv. \(\mathrm{TiCl_4}\) , \(\alpha\) -dichlorobenzene, (microwave), \(180^{\circ}\mathrm{C}\) , 3 h; (ii) 8 equiv. DDQ, \(5\%\) TfOH, 1,2-dichloroethane, \(50^{\circ}\mathrm{C}\) , 6 h; (iii) 4 equiv. \((\mathrm{Bu}_3\mathrm{Sn})_2\mathrm{S}\) or \((\mathrm{Bu}_3\mathrm{Sn})_2\mathrm{Se}\) , 1 equiv. \(\mathrm{Pd(PPh}_3)_4\) , toluene, \(150^{\circ}\mathrm{C}\) , Ar, 48 h. </center>  
+
+<|ref|>text<|/ref|><|det|>[[82, 277, 483, 844]]<|/det|>
+was prepared from commercially available 2,7- dichloro- 9H- fluorene in three steps (see the Supplementary Information for details). The initial attempts towards the cyclotrimerization of 2 using \(\mathrm{SiCl}_4\) [32] as catalyst in ethanol failed; only a trace amount of the desired product, 1,3,5- tris(9,9- dibutyl- 2,7- dichloro- 9H- fluoren- 4- yl)benzene (TFB, 3), was detected in the reaction mixture by MS spectroscopy. Fortunately, several explorations rewarded us with finding proper conditions to access the cyclotrimer 3 by performing the reaction in \(\mathrm{o}\) - dichlorobenzene ( \(\mathrm{o}\) - DCB) at \(180^{\circ}\mathrm{C}\) under microwave irradiation using \(\mathrm{TiCl}_4\) as catalyst [33, 34]. This venerable method, albeit under slightly harsh conditions, delivered the product 3 in \(52\%\) isolated yield after chromatographical purification. Noticeably, TFB has the carbon structure that constitutes the entire carbon scaffold of final bowls and the functionalities necessary for the formation of the five- membered rings.  
+
+<|ref|>text<|/ref|><|det|>[[84, 852, 481, 897], [517, 280, 917, 707]]<|/det|>
+Also fortunately and delightfully, the consequent Scholl reaction of 3 with DDQ/triflic acid in DCE at \(50^{\circ}\mathrm{C}\) , which is the crucial step toward the total synthesis of the designed bowls, afforded the hunted trifluorenocoronene (TFC, 4) in an isolated yield of \(37\%\) . Although not high, the achieved yield is quite reasonable if considering the high ring strain from three five- membered rings and steric hindrance of two chlorine atoms at each bay region in the product 4. The preinstalled \(n\) - butyl groups endow TFC with adequate solubility amenable to isolation and full spectroscopic characterization. All analytical data are consistent with the expected structure of TFC 4 (Supplementary Figs 52 and 53). We also briefly explored the improvement of Scholl reaction of 3 and found relatively satisfactory conditions, as denoted in Fig. 2.   
+
+<|ref|>text<|/ref|><|det|>[[517, 715, 916, 896]]<|/det|>
+With the requisite TFC (4) in hand, we then accomplished the total synthesis of the designed buckybowls via threefold heteroatom annulation at its dichlorinated bay positions using Stille type reaction, which has been established by Wang group in their synthesis of sulfur heterocyclic annulated perylene bisimide derivatives [35] with
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[75, 70, 480, 890]]<|/det|>
+some modifications. To our delight, we obtained successfully the hunted buckybowl, trithiahexain- denocoroneen 1a, as a yellowish powder in 58% isolated yield using \((\mathrm{Bu}_3\mathrm{Sn})_2\mathrm{S}\) as the sulfur donor and \(\mathrm{Pd(PPh_3)_4}\) as the catalyst at \(150^{\circ}\mathrm{C}\) in a sealed tube under Ar atmosphere. This is the first time that such a supersized buckybowl was achieved. The HR- MS spectrum showed the molecular peak at \(m / e\) 985.3936 consistent with the desired product \(\mathrm{(C_{60}H_{60}S_3 + H^+, 985.3930)}\) and also, the measured isotopic distribution was well coincident with that stimulated (Supplementary Fig. 56). The \(^1\mathrm{H}\) NMR spectrum (Fig. 3) presented only one sharp low field singlet at 8.06 ppm for the six equivalent aromatic protons; while the high field signals, which are well resolved, implied two inequivalent sets of \(n\) - butyl groups. These observations strongly suggest its 3- fold symmetry and the bowl- shaped conformation in solution since only if access was gained to the bowl structure, the aliphatic chains can be differentiated by their location at regions demarcated by the convex and concave faces of the bowl. The chemical shifts of butyl chains inside the concave follow an order of \(\mathrm{H}_a > \mathrm{H}_c > \mathrm{H}_d > \mathrm{H}_b\) , similar to 9,9- dibutyl- 9H- fluorene [36]. While butyl chains outward the bowl follow a different order of \(\mathrm{H}_a > \mathrm{H}_b > \mathrm{H}_c > \mathrm{H}_d\) . Interestingly, two distinctive signals, \(\mathrm{H}_b\) and \(\mathrm{H}_d\) , assignable to methylene and methyl protons in the butyl moieties inward bowl orientations manifest negative chemical shifts \((- 1.04\) to \(- 0.02\) ppm), which can be ascribed to the stronger shielding effect caused by the ring current of the bowl system. The \(^{13}\mathrm{C}\) NMR data further support the bowl topological character of 1a (Supplementary Fig. 55).  
+
+<|ref|>text<|/ref|><|det|>[[514, 72, 913, 228]]<|/det|>
+inward bowl orientations manifest negative chemical shifts \((- 1.04\) to \(- 0.02\) ppm), which can be ascribed to the stronger shielding effect caused by the ring current of the bowl system. The \(^{13}\mathrm{C}\) NMR data further support the bowl topological character of 1a (Supplementary Fig. 55).  
+
+<|ref|>image<|/ref|><|det|>[[518, 247, 909, 430]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[514, 434, 912, 463]]<|/det|>
+<center>Fig. 3 \(^1\mathrm{H}\) NMR characterization of 1a ( \(\mathrm{CDCl}_3\) ) and geometry predicted by DFT calculations. </center>  
+
+<|ref|>text<|/ref|><|det|>[[514, 495, 913, 868]]<|/det|>
+The selenium version was also synthesized by bridging the bays with selenium atoms under the same conditions, but switching \((\mathrm{Bu}_3\mathrm{Sn})_2\mathrm{S}\) to \((\mathrm{Bu}_3\mathrm{Sn})_2\mathrm{Se}\) [37], as a yellowish powder in 39% isolated yield. Similarly, all spectroscopic data (Supplementary Figs 57 to 59) are well consistent with the expected structure of triselenohexainedenocoroneen 1b. It should be pointed out that its \(^1\mathrm{H}\) NMR and \(^{13}\mathrm{C}\) NMR spectra show virtually the same pattern to the sulfur analog. Furthermore, the corresponding \(\mathrm{CH}_2(\mathrm{b})\) and \(\mathrm{CH}_3(\mathrm{d})\) signals at higher field \((- 0.96\) to \(0.01\) ppm) due to the weaker shielding effect, reflecting that the bowl is flatter than its sulfur analog.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[135, 80, 861, 481]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[83, 486, 916, 530]]<|/det|>
+<center>Fig. 4 a,d, X-ray crystal structures of 1a (a) and 1b (d). b,e, The side view of 1a (b) and 1b (e) with diameters and depths. c,f, The top view of 1a (c) and 1b (f) together with POAV angles (blue) and mean bond lengths (red). Butyl groups in the side and top views are omitted for clarity. Thermal ellipsoids are shown at 30% probability. </center>  
+
+<|ref|>text<|/ref|><|det|>[[83, 548, 483, 892]]<|/det|>
+Single crystals of 1a and 1b suitable for X- ray diffraction studies were grown by slow vapor diffusion of methanol into solution of chloroform at room temperature. The X- ray structural analyses unambiguously confirm the anticipated bowl- shaped architecture (Figs 4(a) and 4(d)), with an approximate \(C_{3v}\) - like symmetry in both crystals; the deviations from ideal symmetry should be blamed to the crystal packing forces in crystalline state [38]. The maximum bowl depths and diameters are 2.29 Å and 11.06 Å for 1a while 2.16 Å and 11.35 Å for 1b, defined by the perpendicular distance between the plane of three peripheral  
+
+<|ref|>text<|/ref|><|det|>[[518, 548, 917, 811]]<|/det|>
+heteroatoms and the centroid of the hub and by the horizontal distance between saturated carbons and heteroatoms at opposite vertexes, respectively (Figs 4(b) and 4(e)). Deservedly, the selenium bowl is slightly shallower than its sulfur isolog due to its bigger atomic radius. It is worth mentioning that the depths and widths as well as other structural parameters of the bowls are closely aligned with those of DFT calculations (Supplementary Figs 25 and 31).  
+
+<|ref|>text<|/ref|><|det|>[[518, 820, 916, 892]]<|/det|>
+Both 1a and 1b crystallize in \(\mathrm{P2_1 / c}\) space group with a disordered chloroform molecule filling over each hub ring of concave side by \(\mathrm{C - H\cdots\pi}\)
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[77, 73, 480, 528]]<|/det|>
+or Cl··π interactions with the formation of chloroform in bowl supermolecular complex isologs (Supplementary Figs 1 and 11). Normal concaveconvex stacking fashion is restricted due to the presence of three pairs of \(n\) - butyl groups. Note that edge- to- convex predominant stacking manner stabilized with C- H··π and chalcogen··π interactions was observed (Supplementary Figs 3 and 13). Both bowls exhibit concave to concave packing motifs along \(a\) axis via six C- H··chalcogen hydrogen bonds between chalcogen atoms and hydrogen atoms \(\mathrm{(H_{d})}\) of methyl groups inward bowls (Supplementary Figs 4, 5, 14 and 15). Notably, 1a contains three crystallographically independent molecules while the selenium isolog 1b contains two crystallographically independent molecules in unit cell.  
+
+<|ref|>text<|/ref|><|det|>[[78, 533, 480, 906]]<|/det|>
+As shown in Figs 4(c) and 4(f), there are 4 circles, totally 36 different pyramidalized trigonal carbon atoms existed in these coronene- based superbowls, except for 18 atoms at the peripheral vertexes. The \(\pi\) - orbital axis vector (POAV) angle analyses [39, 40] show that the carbon atoms of the hub ring of two bowls are moderately curved with a mean POAV angle of \(5.40^{\circ}\) (1a) and of \(4.79^{\circ}\) (1b); the most curvature occurs at the carbon atoms making up the coronene rim, of which the mean POAV angles are \(6.26^{\circ}\) (1a) and \(5.85^{\circ}\) (1b) for carbon atoms fusion with cyclopentadiene rings. The carbon atoms linked to each hub ring are pyramidalized significantly with a mean  
+
+<|ref|>text<|/ref|><|det|>[[515, 73, 914, 308]]<|/det|>
+POAV angle of \(6.14^{\circ}\) (1a) and of \(5.61^{\circ}\) (1b), the secondary maximum value of those of the carbons in respective molecule. The curvature is even distributed among the peripheral carbons attached to heteroatom and to the saturated carbon atom of cyclopentadiene ring, of which POAV angles are \(3.91^{\circ}\) (1a) and \(3.55^{\circ}\) (1b) and \(3.73^{\circ}\) (1a) and \(3.60^{\circ}\) (1b), respectively (see the Supplementary Information for details).  
+
+<|ref|>text<|/ref|><|det|>[[515, 315, 914, 909]]<|/det|>
+The bond lengths, as denoted in Figs 4(c) and 4(f) (red colored data), are the symmetry- averaged values of all equivalent bonds around their respective \(C_{3}\) symmetric axis disregard deviations from perfect geometry. The hub ring in both molecules remains an almost regular hexagon with a negligible bond length alternation (BLA) of \(0.002 \mathrm{\AA}\) (1a) and \(0.001 \mathrm{\AA}\) (1b) and equalized yet elongated bond lengths of \(1.422\) and \(1.424 \mathrm{\AA}\) (1a) or \(1.421\) and \(1.422 \mathrm{\AA}\) (1b) in comparison with that in benzene \((1.40 \mathrm{\AA})\) , indicating their reduced aromatic bond character. The radial bonds joining the hub to the rim hexagons are elongated to \(1.450 \mathrm{\AA}\) (1a) and \(1.447 \mathrm{\AA}\) (1b), considerably longer than the others in the same ring and thus indicative of quasi- single bond character. These bonds are not components of aromatic rings and function only as geometric connections between the hub and outer rings. Accordingly, the six- membered rings containing the radial bonds deviate markedly from regular hexagon. Six rim benzene rings in both compounds are non- negligibly
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[83, 72, 485, 580]]<|/det|>
+irregular hexagons with the bond lengths spanning from 1.386 to 1.420 Å (1a) and from 1.389 to 1.420 Å (1b), falling within the scope of aromatic bond character. The C- S bond (1.779 Å) in 1a and C- Se bond (1.923 Å) in 1b are fairly longer than that in dibenzothiophene (1.744 Å) [41] and in dibenzoselenophene (1.895 Å) [42], indicating also declined aromatic bond character at the corresponding positions. These results lend support to that the peripheral and hub benzene rings possess more benzenoid character, consistent with the deductions from BLA analysis [43] based on the X- ray determined (Supplementary Fig. 9 and 19) and DFT calculated structural data (Supplementary Figs 26 and 32). In addition, the fluorene moieties show a C- C \(_{sat}\) . bond length of 1.559 Å (1a) and 1.560 Å (1b), longer than that in 9,9- dioctyl- 9H- fluorene (1.523 Å) [44], might due to the high inner strain of their bowl geometries.  
+
+<|ref|>text<|/ref|><|det|>[[83, 588, 483, 880]]<|/det|>
+The aromaticity distribution characteristics in the bowl systems were further evaluated by nucleus independent chemical shift (NICS) [45- 47] and anisotropy of the induced current density (ACID) [48] calculations on their unsubstituted analogs 1a' and 1b'. The calculated NICS(1) \(_{zz}\) values, as denoted in the corresponding rings (Figs 5(a) and 5(b)), suggest that the hub hexagons (ring a) and the six outer hexagons (e) hold a pronounced aromatic character while the hetero pentagons (d) and the hexagons (c) are  
+
+<|ref|>text<|/ref|><|det|>[[518, 70, 918, 799]]<|/det|>
+faintly aromatic. The hexagons (b), which possess positive average NICS(1) \(_{zz}\) value values, present a feeble antiaromatic character. Therefore, the aromaticity distribution in each bowl system is are nearly, yet not completely similar to that in \(p\) - HBC [49], as shown by the blue shaded benzene rings (Figs 5(a) and 5(b)). Such aromaticity distribution characteristics is also consistent with Clar sextet rule [50] because the largest number of aromatic sextets can be found in the resonance hybrid when aromatic sextets locate at rings (a and e) overwhelm rings (d and c)(Supplementary Figs 30 and 36)). ACID plots of 1a' and 1b' revealed that clockwise 6π- electron local current pathways are presented in the hub (a) and rim (e) benzene rings. On the contrary, no significant local currents were observed in rings (c and d) of neither bowl (Figs 5(c) and 5(d)). Rings (b) showed unnoticeable anticlockwise ring currents, reflecting their weak antiaromatic character. Remarkably, a clockwise global current was observed along 30π- electron pathway consisting of the lone pair electrons on S or Se atoms, which are caused by the superposition of these local currents in each molecule. Taken together, these outcomes lend further supports to the aromaticity distribution characteristics obtained by NICS calculations.  
+
+<|ref|>text<|/ref|><|det|>[[518, 806, 917, 907]]<|/det|>
+The electrostatic potential (ESP) [51, 52] of 1a' and 1b' were also calculated (Figs 6(a) and 6(b) for 1a', Figs 6(d) and 6(e) for 1b'). Unlike planar PAHs containing two identical faces, the
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[68, 78, 460, 370]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[80, 377, 479, 457]]<|/det|>
+<center>Fig. 5 a,b, \(\mathrm{NICS}(1)_{zz}\) values (ppm) of \(1\mathbf{a}^{\prime}\) (a) and \(1\mathbf{b}^{\prime}\) (b) calculated at the GIAO B3LYP/6-311G(d,p) level of theory. c,d, Calculated AICD plots (isovalue \(= 0.03\) ) of \(1\mathbf{a}^{\prime}\) (c) and \(1\mathbf{b}^{\prime}\) (d). Only contributions from \(\pi\) -electrons of the aromatic cores are considered. Red arrows indicate directions of induced ring current. </center>  
+
+<|ref|>text<|/ref|><|det|>[[80, 475, 479, 820]]<|/det|>
+ESP maps of these bowl- shaped compounds illustrate that the convex faces display more negative values compared with their concave faces, corresponding to the directions of the intrinsic dipole moment from the convex surface to the concave surface (Figs 6(c) and 6(f)). The calculated dipole moments are 3.59 Debye ( \(1\mathbf{a}^{\prime}\) ) and 3.45 Debye ( \(1\mathbf{b}^{\prime}\) ), much higher than those of coranlunene (2.19 Debye) [53] and surmarrene (2.7 Debye) [54], likely due to the enlarged \(\pi\) - surface and depth of these supersized bowls. In addition, the incorporation of thiophene or selenophene subunits enhances their concave- convex polarization effect.  
+
+<|ref|>text<|/ref|><|det|>[[80, 829, 479, 893]]<|/det|>
+Bowl to bowl inversion of two bowls were investigated by variable temperature \(^1\mathrm{H}\) NMR  
+
+<|ref|>text<|/ref|><|det|>[[515, 72, 914, 580]]<|/det|>
+measurements in 1,2- dichlorobenzene- \(d_4\) from 30- 180 °C (Supplementary Figs 38 and 39) and no alteration on chemical shifts was observed, reflecting their inability to undergo bowl- inversion at the temperatures screened and thus their high inversion barriers. The DFT calculations for inversion barriers of the unsubstituted \(1\mathbf{a}^{\prime}\) and \(1\mathbf{b}^{\prime}\) at the M062x/6- 31g(d) level of theory (Fig. 7) revealed that the inversion would proceed via a planar transition state like coranlunene [55- 58] and sumanene [59, 60], with inversion barriers \((\Delta G^{\ddagger})\) of 70.2 kcal/mol and 51.0 kcal/mol for \(1\mathbf{a}^{\prime}\) and \(1\mathbf{b}^{\prime}\) , respectively (see the Supplementary Information for details). Such high energy barriers suggest that the bowl- to- bowl inversion process might be impossible at ordinary temperatures according to Eying equation [61] (Supplementary Fig. 41). So, their bowl- shaped \(\pi\) - systems are conformationally locked at mild temperature.  
+
+<|ref|>text<|/ref|><|det|>[[515, 588, 914, 907]]<|/det|>
+To elucidate the optophysical properties of both sulfur and selenium isologs, their UV- vis absorption and fluorescence spectra were recorded in dichloromethane (Fig. 8(a)). Both absorption spectra displayed a similar pattern with the maximum absorbance at around 400 nm and a faint hump extending up 510 nm. These observations implied that the displacement of sulfur by selenium only changes the relative absorptive intensity rather than the positions of bands. The spectroscopic similarity of two isologs stems from their similar structures in terms of geometrical and
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[127, 75, 875, 368]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[83, 372, 916, 428]]<|/det|>
+<center>Fig. 6 a,d, Calculated electrostatic potentials of bowl concave for \(1\mathbf{a}^{\prime}\) (a) and \(1\mathbf{b}^{\prime}\) (d). b,e, Calculated electrostatic potentials of bowl convex for \(1\mathbf{a}^{\prime}\) (b) and \(1\mathbf{b}^{\prime}\) (e). c,f, Dipole moments of \(1\mathbf{a}^{\prime}\) (c) and \(1\mathbf{b}^{\prime}\) (f). Electrostatic potentials calculated on the 0.001 au. isodensity surface together with surface maxima (blue) and minima points (red) at the B3LYP/6-311G(d,p) level of theory. </center>  
+
+<|ref|>image<|/ref|><|det|>[[150, 457, 855, 638]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[83, 644, 770, 662]]<|/det|>
+<center>Fig. 7 a,b, Inversion barriers of \(1\mathbf{a}^{\prime}\) (a) and \(1\mathbf{b}^{\prime}\) (b) calculated at M062x/6-31G(d) level of theory. </center>  
+
+<|ref|>text<|/ref|><|det|>[[84, 679, 483, 888]]<|/det|>
+electronic aspects, as revealed by X- ray structural analysis and DFT calculations described above. TD- DFT calculations at the PBE/def2tzvp level for models \(1\mathbf{a}^{\prime}\) and \(1\mathbf{b}^{\prime}\) (Figs 8(c) and Figs 8(d), see the Supplementary Information for details) revealed that the maximum absorbance of \(1\mathbf{a}\) is caused by the \(\mathrm{S}_0\rightarrow \mathrm{S}_3\) and \(\mathrm{S}_0\rightarrow \mathrm{S}_4\) transitions of \(1\mathbf{a}^{\prime}\) which contain those from frontier MOs to  
+
+<|ref|>text<|/ref|><|det|>[[517, 680, 917, 861]]<|/det|>
+nondegenerate LOMO+2 ( \(f = 0.4286\) ) while the lowest energy absorption band can be attribute to the \(\mathrm{S}_0\rightarrow \mathrm{S}_1\) transition ( \(f = 0.0005\) ). The low intensity of \(\mathrm{S}_0\rightarrow \mathrm{S}_1\) transition is related to the degeneracy of frontier molecular orbitals involved in the transition. The same rationale could be applied to explain the UV- vis spectrum of \(1\mathbf{b}\)
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[135, 77, 854, 504]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[78, 509, 915, 595]]<|/det|>
+<center>Fig. 8 a, UV-Vis absorption (solid curves) and emission spectra (dotted curves) of 1a (black) and 1b (red) in \(\mathrm{CH}_2\mathrm{Cl}_2\) at a concentration of \(1.0\times 10^{-5}\mathrm{M}\) . b, Cyclic voltammogram and differential pulse voltamogram of 1a and 1b in \(\mathrm{CH}_2\mathrm{Cl}_2\) (0.1 mol/L \(n\mathrm{-Bu}_4\mathrm{NPF}_6\) ) at a scan rate of 0.1 V/s. All potentials were calibrated versus an aqueous SCE by the addition of ferrocene as an internal standard taking \(\mathrm{E}_{1 / 2}(\mathrm{Fc} / \mathrm{Fc}^{+}) = 0.424\mathrm{V}\) vs. SCE [62]. c,d, Orbital correlation diagram and transition composition of the \(\mathrm{S}_0\rightarrow (\mathrm{S}_1,\mathrm{S}_3,\mathrm{S}_4)\) excited states for 1a' (c) and 1b' (d) calculated at PBE/def2tzvp level of theory. </center>  
+
+<|ref|>text<|/ref|><|det|>[[78, 608, 479, 899]]<|/det|>
+with similar molecular orbital energy level distribution in \(\mathbf{1b'}\) . The optical energy gaps were estimated to be \(2.43\mathrm{eV}\) (1a) and \(2.39\mathrm{eV}\) (1b) from the onset wavelength of lowest energy absorption edge (Table S7). The fluorescence spectra of both isologs featured distinctive vibronic structures with multiple maxima at 503, 515, 523, and 543 nm. The sulfur isolog 1a displayed a weak fluorescence while and selenium isolog 1b was found to be almost non- emissive due to stronger heavy atom effect of selenium than sulfur. These results  
+
+<|ref|>text<|/ref|><|det|>[[515, 610, 914, 846]]<|/det|>
+are explicable to that the degeneration from \(\mathrm{S}_1\) are symmetry forbidden, thus leading to the excited state relaxation via different vibronic levels of \(\mathrm{S}_0\) . The absolute quantum yield and fluorescence lifetime of 1a were measured to be \(4\%\) and \(1.9\mathrm{ns}\) and the radiative \((k_f)\) and nonradiative \((k_{nr})\) decay rate constants were calculated from the singlet excited state to be \(2.3\times 10^7\mathrm{s}^{- 1}\) and \(5.0\times 10^8\mathrm{s}^{- 1}\) , respectively.  
+
+<|ref|>text<|/ref|><|det|>[[515, 854, 912, 900]]<|/det|>
+Cyclic and differential pulse voltametric studies (Fig. 8(b)) revealed that both compounds
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[83, 72, 485, 581]]<|/det|>
+underwent two reversible oxidation waves and an irreversible oxidation wave corresponding to the three sequential one- electron oxidation processes of three heteroatoms, indicating that the generation of the cation radical during the first two oxidation processes are stable to repeated cycles, the third oxidation process is unstable under the selected electrochemical conditions. The half- wave potentials \(\mathrm{E}_{ox}^{1 / 2}\) of two reversible oxidation processes locate at \(0.90\mathrm{V}\) and \(1.35\mathrm{V}\) (1a) and \(0.93\mathrm{V}\) and \(1.38\mathrm{V}\) (1b) against SCE. Note that the oxidation potential of selenium isolog 1b is lower than that of sulfur isolog 1a due to the smaller electronegativity of selenium atoms. Their HOMO/LUMO energy levels were estimated from the onsets of oxidation and reduction to be \(- 5.25 / - 2.50\) (1a) and \(- 5.24 / - 2.50\mathrm{eV}\) (1b), corresponding to the electrochemical energy gaps of \(2.75\mathrm{eV}\) and \(2.74\mathrm{eV}\) , respectively (Table S8).  
+
+<|ref|>text<|/ref|><|det|>[[84, 589, 483, 741]]<|/det|>
+Thermogravimetric analysis (TGA) experiments were conducted to determine the thermal stability of both compounds. The results showed that the onset decomposition temperatures of both 1a and 1b are above \(400^{\circ}\mathrm{C}\) (Supplementary Fig. 24), reflecting their thermal robustness.  
+
+<|ref|>text<|/ref|><|det|>[[84, 751, 483, 877]]<|/det|>
+In conclusion, we have paved a new synthetic strategy and, based on which, a concise 3- step approach for expeditiously access to large, compact, and symmetric buckybowls that underscore the following features: 1) a bowl- like architecture  
+
+<|ref|>text<|/ref|><|det|>[[517, 73, 917, 661]]<|/det|>
+nanosized, 19 rings, 48 atoms, and 36 pyramidalized trigonal carbon atoms; 2) a coronene core fully circumscribed by alternate hexagonal and pentagonal rings. Single- crystal X- ray crystallography and NMR spectroscopy clearly confirmed their bowl- shaped geometry in crystal and in solution. DFT calculations and variable temperature \(^1\mathrm{H}\) NMR experiments suggest their much high inversion barriers. Quantum chemical calculations also reveal their aromaticity distribution and electrostatic potential characteristics. Distinguishing from most known buckybowls, we present the first instance containing coronene core as the bowl bottom with complete edge topological structure. The success on the synthesis of these previously unknown superbowls delivered a new family of \(\pi\) - bowls and a new synthetic strategy and the most important of all, verified accessibility of bridging all bays of \(p\) - HBCs. We believed that it will promote more synthetic efforts towards their all- carbon parent, other hetero versions as well as the analogs based on other PAHs with bay regions.  
+
+<|ref|>sub_title<|/ref|><|det|>[[520, 692, 732, 714]]<|/det|>
+## Data availability  
+
+<|ref|>text<|/ref|><|det|>[[518, 735, 916, 834]]<|/det|>
+X- ray crystallographic data for compounds 1a and 1b are freely available from the Cambridge Crystallographic Data Centre (CCDC 2205194 and 2205195, respectively).
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[80, 70, 197, 90]]<|/det|>
+## Methods  
+
+<|ref|>text<|/ref|><|det|>[[78, 110, 480, 900]]<|/det|>
+Preparation of 3. In a glove box, 1- (9,9- dibutyl- 2,7- dichloro- 9H- fluoren- 4- yl)ethan- 1- one 2 (390 mg, 1.0 mmol, 1.0 equiv.), \(\mathrm{TiCl}_4\) (0.7 mL, 6.6 mmol, 6.6 equiv.) and 5 mL dry o- dichlorobenzene were added to a \(10\mathrm{mL}\) microwave vial and the vial was capped. The vessel was removed from the glove box and placed into a microwave reactor where it was heated at \(180^{\circ}\mathrm{C}\) for \(3\mathrm{h}\) . After cooling down, the mixture was poured over concentrated hydrochloric acid/ice to quench the reaction and then extracted with \(\mathrm{CH}_2\mathrm{Cl}_2\) . The combined organic phase was washed with saturated aqueous \(\mathrm{NaHCO}_3\) and dried over \(\mathrm{Na}_2\mathrm{SO}_4\) . The organic solvent was removed at reduced pressure to give a yellow- brown solid. The above procedure was repeated 4 times and the crude material from each reaction was combined. The combined crude material was purified by column chromatography over silica gel (eluent: petroleum ether) to afford 1,3,5- Tris(9,9- dibutyl- 2,7- dichloro- 9H- fluoren- 4- yl)benzene 3 (970 mg, 52%) as an off- white powder; melting point (m.p.), \(147–148^{\circ}\mathrm{C}\) ; \(^1\mathrm{H}\) NMR (600 MHz, \(\mathrm{CDCl}_3\) ): \(\delta 7.67\) (d, \(J = 14.9\mathrm{Hz}\) , 3H), 7.32 (s, 6H), 7.29 (d, \(J = 1.9\mathrm{Hz}\) , 3H), 7.12 (d, \(J = 8.2\mathrm{Hz}\) , 3H), 6.89 (d, \(J = 8.0\mathrm{Hz}\) , 3H), 1.98- 1.91 (m, 12H), 1.13- 1.07 (m, 12H), 0.69 (t, \(J = 7.4\mathrm{Hz}\) , 18H), 0.60 (d, \(J = 7.5\mathrm{Hz}\) , 12H); \(^{13}\mathrm{C}\) NMR (150 MHz, \(\mathrm{CDCl}_3\) ): \(\delta 153.6\) , 153.2, 138.3, 137.3, 133.3, 132.9, 129.2, 128.9, 126.7, 123.3, 123.1, 122.6, 55.0, 40.3, 25.8, 22.9, 13.7; HRMS (APCI): \(m / z\) calcd for \(\mathrm{C}_{69}\mathrm{H}_{73}\mathrm{Cl}_6\) (M+H) \(^+\) : 1111.3838, found: 1111.3836.  
+
+<|ref|>text<|/ref|><|det|>[[515, 73, 913, 149]]<|/det|>
+129.2, 128.9, 126.7, 123.3, 123.1, 122.6, 55.0, 40.3, 25.8, 22.9, 13.7; HRMS (APCI): \(m / z\) calcd for \(\mathrm{C}_{69}\mathrm{H}_{73}\mathrm{Cl}_6\) (M+H) \(^+\) : 1111.3838, found: 1111.3836.  
+
+<|ref|>text<|/ref|><|det|>[[515, 180, 914, 752]]<|/det|>
+Preparation of 4. To a mixture of 3 (220 mg, 0.2 mmol, 1.0 equiv.) and DDQ (360 mg, 1.6 mmol, 8.0 equiv.) in 1,2- dichloroethane (20 mL) was added trifluoromethanesulfonic acid (1 mL) under argon atmosphere, and the mixture was stirred at \(50^{\circ}\mathrm{C}\) for \(6\mathrm{h}\) . After cooling to room temperature, the reaction mixture was quenched by adding saturated aqueous \(\mathrm{NaHCO}_3\) , and then the mixture was extracted with \(\mathrm{CH}_2\mathrm{Cl}_2\) . The combined organic layer was dried over \(\mathrm{Na}_2\mathrm{SO}_4\) , and the solvent was removed under reduced pressure. The crude product was purified by silica gel column chromatography (eluent: petroleum ether) to give the trifluorocoronene 4 (80 mg, 37%) as a yellow powder; m.p. \(>300^{\circ}\mathrm{C}\) ; \(^1\mathrm{H}\) NMR (600 MHz, \(\mathrm{Tol - d}_8\) ): \(\delta 8.28\) (s, 6H), 2.26 (dd, \(J = 9.8\) , 6.6 Hz, 12H), 1.16 (dd, \(J = 14.6\) , 7.3 Hz, 12H), 1.08- 1.02 (m, 12H), 0.66 (t, \(J = 7.3\mathrm{Hz}\) , 18H); \(^{13}\mathrm{C}\) NMR (150 MHz, \(\mathrm{Tol - d}_8\) ): \(\delta 151.7\) , 135.0, 133.9, 127.3, 126.9, 126.1, 122.0, 63.1, 38.8, 27.9, 23.9, 14.4; HRMS (APCI): \(m / z\) calcd for \(\mathrm{C}_{69}\mathrm{H}_{61}\mathrm{Cl}_6\) (M+H) \(^+\) : 1099.2908, found: 1099.2898.  
+
+<|ref|>text<|/ref|><|det|>[[515, 805, 913, 877]]<|/det|>
+Preparation of 1a and 1b. Inside the glove- box, to an oven- dried pressure vessel with a Teflon screw cap was added compound 4 (110 mg,
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[80, 70, 485, 884]]<|/det|>
+0.1 mmol, 1.0 equiv.), \(\mathrm{Bu_3SnSSnBu_3}\) (245 mg, 0.4 mmol, 4.0 equiv.) and \(\mathrm{Pd(PPh_3)_4}\) (115 mg, 0.1 mmol, 1.0 equiv.) and 5 mL dry and degassed toluene. After the vessel was resealed and moved out from the glovebox, the mixture was heated and stirred at \(150^{\circ}\mathrm{C}\) for \(48\mathrm{h}\) . On cooling to room temperature, the reaction mixture was quenched with saturated aqueous KF solution and extracted with \(\mathrm{CH_2Cl_2}\) . The combined organic phase was dried over \(\mathrm{Na_2SO_4}\) before being filtered and concentrated down to a solid under reduced pressure. The crude solid was adsorbed onto silica gel and subjected to silica gel column chromatography (eluent, petroleum ether) to afford trithiahexaindenocoronene 1a (57 mg, \(58\%\) yield) as a yellowish powder; m.p. \(>300^{\circ}\mathrm{C}\) ; Triselenohexaindenocoronene 1b was obtained under the similar procedure by using \(\mathrm{Bu_3SnSeSnBu_3}\) which was prepared according to literature procedure [39]. Characterization data of 1a: \(^1\mathrm{H}\) NMR (600 MHz, \(\mathrm{CDCl_3}\) ): \(\delta\) 8.06 (s, 6H), 2.70- 2.66 (m, 6H), 2.13- 2.09 (m, 6H), 2.05- 2.00 (m, 6H), 1.58- 1.56 (m, 6H), 1.05 (t, \(J = 7.4\mathrm{Hz}\) , 9H), 0.59 (dd, \(J = 14.7\) , 7.4 Hz, 9H), \(- 0.02\) (t, \(J = 7.3\mathrm{Hz}\) , 9H), \(- 1.01\mathrm{- }1.07\) (m, 6H). \(^{13}\mathrm{C}\) NMR (150 MHz, \(\mathrm{CDCl_3}\) ): \(\delta\) 153.0, 142.9, 139.8, 134.8, 130.8, 128.6, 119.6, 64.4, 42.1, 35.8, 28.7, 26.1, 23.6, 22.5, 14.3, 13.2; HRMS (APCI): \(m / z\) calcd for \(\mathrm{C_{69}H_{61}S_3(M + H)^+}\) : 985.3930, found: 985.3936. Characterization data of 1b: \(^1\mathrm{H}\) NMR (600 MHz, \(\mathrm{CDCl_3}\) ): \(\delta\) 8.21 (s, 6H),  
+
+<|ref|>text<|/ref|><|det|>[[516, 72, 917, 283]]<|/det|>
+\(2.71 - 2.66\) (m, 6H), 2.17- 2.12 (m, 6H), 2.02- 1.98 (m, 6H), 1.57- 1.54 (m, 6H), 1.04 (t, \(J = 7.4\mathrm{Hz}\) , 9H), 0.59 (dd, \(J = 14.7\) , 7.4 Hz, 6H), 0.01 (t, \(J = 7.3\mathrm{Hz}\) , 9H), \(- 0.93 - 0.98\) (m, 6H). \(^{13}\mathrm{C}\) NMR (150 MHz, \(\mathrm{CDCl_3}\) ): \(\delta\) 151.8, 141.9, 139.0, 136.0, 129.3, 128.2, 121.6, 64.4, 41.5, 36.5, 28.5, 25.8, 23.6, 22.4, 14.3, 13.2; MS (MALDI- TOF): m/z calcd for \(\mathrm{C_{69}H_{60}Se_3(M)^+}\) : 1126.1, found: 1126.4.  
+
+<|ref|>sub_title<|/ref|><|det|>[[520, 310, 752, 333]]<|/det|>
+## Acknowledgments  
+
+<|ref|>text<|/ref|><|det|>[[518, 353, 916, 510]]<|/det|>
+J. W. thanks the National Science Foundation of China (NSFC) (Grant Nos. 21871169); J. L. thanks the National Science Foundation of China (NSFC) (Grant Nos. 21702131); Y. S. thanks the Fundamental Research Funds for the Central Universities (SNNU 2019TS040).  
+
+<|ref|>sub_title<|/ref|><|det|>[[520, 538, 795, 561]]<|/det|>
+## Author contributions  
+
+<|ref|>text<|/ref|><|det|>[[518, 580, 916, 900]]<|/det|>
+Y. S., X. W. and M. C. conducted the experiments with help from B. Y., M. X. and Y. Z., Y. S. performed the most DFT calculations. Z. G. and J. D. performed the inversion barrier calculations. H. S. collected and processed X-ray diffraction data. J. F. collected high-resolution mass spectrometry. J. W. conceived the original idea, designed and supervised the whole studies. J. W., Y. S. and J. L. performed the data analysis and wrote the manuscript with feedback from others. All authors discussed the results and commented on the manuscript.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[60, 68, 380, 92]]<|/det|>
+## Additional information  
+
+<|ref|>text<|/ref|><|det|>[[60, 110, 479, 293]]<|/det|>
+Supplementary information and chemical compound information are available in the online version of the paper. Reprints and permissions information is available online at www.nature.com/reprints. Correspondence and requests for materials should be addressed to J. L. and J. W.  
+
+<|ref|>sub_title<|/ref|><|det|>[[60, 324, 462, 348]]<|/det|>
+## Competing financial interests  
+
+<|ref|>text<|/ref|><|det|>[[60, 366, 476, 410]]<|/det|>
+The authors declare no competing financial interests.  
+
+<|ref|>sub_title<|/ref|><|det|>[[60, 441, 222, 463]]<|/det|>
+## References  
+
+<|ref|>text<|/ref|><|det|>[[80, 483, 477, 585]]<|/det|>
+[1] Wu, Y.- T. & Siegel, J. S. Aromatic molecular bowl hydrocarbons: synthetic derivatives, their structures, and physical properties. Chem. Rev. 106, 4843- 4867 (2006)  
+
+<|ref|>text<|/ref|><|det|>[[80, 606, 478, 736]]<|/det|>
+[2] Gao, G. et al. Rational functionalization of a \(\mathrm{C}_{70}\) buckybowl to enable a \(\mathrm{C}_{70}\) : buckybowl cocrystal for organic semiconductor applications. J. Am. Chem. Soc. 142, 2460- 2470 (2020)  
+
+<|ref|>text<|/ref|><|det|>[[80, 756, 478, 856]]<|/det|>
+[3] Saito, M., Shinokubo, H., & Sakurai, H. Figuration of bowl- shaped \(\pi\) - conjugated molecules: properties and functions. Mater. Chem. Front. 2, 635- 661 (2018)  
+
+<|ref|>text<|/ref|><|det|>[[525, 71, 913, 200]]<|/det|>
+[4] Fu, B. et al. A bowl- shaped sumanene derivative with dense convex- concave columnar packing for high- performance organic field- effect transistors. Chem. Commun. 53, 11407- 11409 (2017)  
+
+<|ref|>text<|/ref|><|det|>[[525, 220, 913, 348]]<|/det|>
+[5] Chen, R. et al. Corannulene derivatives with low LUMO levels and dense convex- concave packing for n- channel organic field- effect transistors. Chem. Commun. 51, 13768- 13771 (2015)  
+
+<|ref|>text<|/ref|><|det|>[[525, 368, 913, 498]]<|/det|>
+[6] Lu, R.- Q. et al. Thiophene- fused bowl- shaped polycyclic aromatics with a dibenzo[a,g]corannulene core for organic field- effect transistors. Chem. Commun. 51, 1681- 1684 (2015)  
+
+<|ref|>text<|/ref|><|det|>[[525, 517, 913, 675]]<|/det|>
+[7] Shi, K., Lei, T., Wang, X. Y., Wang, J. Y. & Pei, J. A bowl- shaped molecule for organic field- effect transistors: crystal engineering and charge transport switching by oxygen doping. Chem. Sci. 5, 1041- 1045 (2014)  
+
+<|ref|>text<|/ref|><|det|>[[525, 695, 913, 796]]<|/det|>
+[8] Gallego, M. et al. Electron transfer in a supramolecular associate of a fullerene fragment. Angew. Chem. Int. Ed. 53, 2170- 2175 (2014)  
+
+<|ref|>text<|/ref|><|det|>[[525, 817, 913, 890]]<|/det|>
+[9] Amaya, T. et al. Anisotropic electron transport properties in sumanene crystal. J. Am. Chem. Soc. 131, 408- 409 (2009)
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[84, 72, 481, 145]]<|/det|>
+[10] Miyajima, D. et al. Liquid crystalline corannulene responsive to electric field. J. Am. Chem. Soc. 131, 44- 45 (2009)  
+
+<|ref|>text<|/ref|><|det|>[[84, 166, 481, 268]]<|/det|>
+[11] Scott, L. T. et al. A short, rigid, structurally pure carbon nanotube by stepwise chemical synthesis. J. Am. Chem. Soc. 134, 107- 110 (2012)  
+
+<|ref|>text<|/ref|><|det|>[[84, 288, 481, 444]]<|/det|>
+[12] Fort, E. H., Donovan, P. M. & Scott, L. T. Diels- alder reactivity of polycyclic aromatic hydrocarbon bay regions: implications for metal- free growth of single- chirality carbon nanotubes. J. Am. Chem. Soc. 131, 16606- 16607 (2009)  
+
+<|ref|>text<|/ref|><|det|>[[84, 465, 480, 539]]<|/det|>
+[13] Sanchez- Valencia, J. R. et al. Controlled synthesis of single- chirality carbon nanotubes. Nature 512, 61- 64 (2014)  
+
+<|ref|>text<|/ref|><|det|>[[84, 560, 480, 660]]<|/det|>
+[14] Cao, Q. et al. Arrays of single- walled carbon nanotubes with full surface coverage for high- performance electronics. Nat. Nanotechnol. 8, 180- 186 (2013)  
+
+<|ref|>text<|/ref|><|det|>[[84, 681, 480, 754]]<|/det|>
+[15] Barth, W. E. & Lawton, R. G. Dibenzo[ghi,mno]fluoranthene. J. Am. Chem. Soc. 88, 380- 381 (1966)  
+
+<|ref|>text<|/ref|><|det|>[[84, 776, 480, 877]]<|/det|>
+[16] Rabideau, P. W. et al. Buckybowls: synthesis and ab initio calculated structure of the first semibuckminsterfullerene. J. Am. Chem. Soc. 116, 7891- 7892 (1994)  
+
+<|ref|>text<|/ref|><|det|>[[518, 73, 916, 145]]<|/det|>
+[17] Sakurai, H., Daiko, T. & Hirao, T. A synthesis of sumanene, a fullerene fragment. Science 301, 1878 (2003)  
+
+<|ref|>text<|/ref|><|det|>[[518, 166, 916, 350]]<|/det|>
+[18] Petrukhina, M. A., Andreini, K. W., Peng, L. Q. & Scott, L. T. Hemibuckminsterfullerene \(\mathrm{C_{30}H_{12}}\) : X- ray crystal structures of the parent hydrocarbon and of the two- dimensional organometallic network \(\{\mathrm{[Rh_2(O_2CCF_3)_4]_3\cdot(C_30H_{12})}\}\) . Angew. Chem. Int. Ed. 43, 5477- 5481 (2004)  
+
+<|ref|>text<|/ref|><|det|>[[518, 370, 916, 525]]<|/det|>
+[19] Steinberg, B. D. et al. Aromatic \(\pi\) - systems more curved than \(\mathrm{C_{60}}\) . The complete family of all indenocorannulenes synthesized by iterative microwave- assisted intramolecular arylatons. J. Am. Chem. Soc. 131, 10537- 10545 (2009)  
+
+<|ref|>text<|/ref|><|det|>[[518, 545, 916, 673]]<|/det|>
+[20] Wu, T. C., Hsin, H. J., Kuo, M. Y., Li, C. H. & Wu, Y. T. Synthesis and structural analysis of a highly curved buckybowl containing corannulene and sumanene fragments. J. Am. Chem. Soc. 133, 16319- 16321 (2011)  
+
+<|ref|>text<|/ref|><|det|>[[518, 694, 916, 820]]<|/det|>
+[21] Wu, T. C., Chen, M. K., Lee, Y. W., Kuo, M. Y. & Wu, Y. T. Bowl- shaped fragments of \(\mathrm{C_{70}}\) or higher fullerenes: synthesis, structural analysis, and inversion dynamics. Angew. Chem. Int. Ed. 52, 1289- 1293 (2013)  
+
+<|ref|>text<|/ref|><|det|>[[518, 843, 916, 889]]<|/det|>
+[22] Chen, M. K. et al. Highly curved bowl- shaped fragments of fullerenes: synthesis, structural
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 73, 477, 120]]<|/det|>
+analysis, and physical properties. Chem. Eur. J. 20, 598- 608 (2014)  
+
+<|ref|>text<|/ref|><|det|>[[80, 140, 479, 240]]<|/det|>
+[23] Liu, J. Z. et al. Fused dibenzo[a,m]rubicene: a new bowl-shaped subunit of \(\mathrm{C}_{70}\) containing two pentagons. J. Am. Chem. Soc. 138, 8364- 8367 (2016)  
+
+<|ref|>text<|/ref|><|det|>[[80, 260, 479, 391]]<|/det|>
+[24] Zou, Y. et al. Dicyclopenta[4,3,2,1- ghi:4',3',2',1'- pqr]perylene: a bowl- shaped fragment of fullerene \(\mathrm{C}_{70}\) with global antiaromaticity. J. Am. Chem. Soc. 141, 7266- 7270 (2019)  
+
+<|ref|>text<|/ref|><|det|>[[80, 411, 479, 512]]<|/det|>
+[25] Yokoi, H., Hiraoka, Y., Hiroto, S., Sakamaki, D., Seki, S. & Shinokubo, H. Nitrogen- embedded buckybowl and its assembly with \(\mathrm{C}_{60}\) . Nat. Commun. 6, 8215 (2015)  
+
+<|ref|>text<|/ref|><|det|>[[80, 532, 479, 632]]<|/det|>
+[26] Ito, S., Tokimaru, Y. & Nozaki, K. Benzene- fused azacorannulene bearing an internal nitrogen atom. Angew. Chem. Int. Ed. 54, 7256- 7260 (2015)  
+
+<|ref|>text<|/ref|><|det|>[[80, 653, 479, 752]]<|/det|>
+[27] Li, X. X. et al. Non- pyrolytic, large- scale synthesis of trichalcogenasumanene: a two- step approach. Angew. Chem. Int. Ed. 53, 535- 538 (2014)  
+
+<|ref|>text<|/ref|><|det|>[[80, 775, 479, 890]]<|/det|>
+[28] Tan, Q. T., Zhou, D. D., Zhang, T., Liu, B. X. & Xu, B. Iodine- doped sumanene and its application for the synthesis of chalcogenasumanenes and silasumanenes. Chem. Commun. 53, 10279- 10282 (2017)  
+
+<|ref|>text<|/ref|><|det|>[[553, 73, 829, 90]]<|/det|>
+Commun. 53, 10279- 10282 (2017)  
+
+<|ref|>text<|/ref|><|det|>[[516, 111, 913, 240]]<|/det|>
+[29] Huang, J. L. et al. Dicyclopenta[ghi,pqr]perylene as a structural motif for bowl- shaped hydrocarbons: synthetic and conformational studies. Org. Lett. 21, 2504- 2508 (2019)  
+
+<|ref|>text<|/ref|><|det|>[[516, 261, 913, 363]]<|/det|>
+[30] Jiang, W. et al. Solution- processed, high- performance nanoribbon transistors based on dithioperylene. J. Am. Chem. Soc. 133, 1- 3 (2011)  
+
+<|ref|>text<|/ref|><|det|>[[516, 383, 913, 483]]<|/det|>
+[31] Tan, Y. Z. et al. Sulfur- annulated hexaperi- hexabenzocoronene decorated with phenylthio groups at the periphery. Angew. Chem. Int. Ed. 54, 2927- 2931 (2015)  
+
+<|ref|>text<|/ref|><|det|>[[516, 504, 913, 604]]<|/det|>
+[32] Elmorsy, S. S., Pelter, A. & Smith, K. The direct production of tri- and hexa- substituted benzenes from ketones under mild conditions. Tetrahedron Lett. 32, 4175- 4176 (1991)  
+
+<|ref|>text<|/ref|><|det|>[[516, 625, 913, 727]]<|/det|>
+[33] Anseems, R. B. M. & Scott, L. T. Circumtrindene: a geodesic dome of molecular dimensions. rational synthesis of 60% of \(\mathrm{C}_{60}\) . J. Am. Chem. Soc. 122, 2719- 2724 (2000)  
+
+<|ref|>text<|/ref|><|det|>[[516, 748, 913, 794]]<|/det|>
+[34] Scott, L. T. et al. A rational chemical synthesis of \(\mathrm{C}_{60}\) . Science 295, 1500- 1503 (2002)  
+
+<|ref|>text<|/ref|><|det|>[[516, 816, 913, 890]]<|/det|>
+[35] Qian, H. L., Liu, C. M., Wang, Z. H. & Zhu, D. B. S- heterocyclic annulated perylene bisimide: synthesis and co- crystal with
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[120, 73, 468, 90]]<|/det|>
+pyrene. Chem. Commun. 4587- 4589 (2006)  
+
+<|ref|>text<|/ref|><|det|>[[84, 112, 483, 240]]<|/det|>
+[36] Ito, Y., Shimada, T., Ha, J., Vacha, M. & Sato, H. Synthesis and characterization of a novel electroluminescent polymer based on a phenoxazine derivative. J. Polym. Sci. Pol. Chem. 44, 4338- 4345 (2006)  
+
+<|ref|>text<|/ref|><|det|>[[84, 262, 481, 390]]<|/det|>
+[37] Maeda, H. et al. One- pot synthesis of seleno- nureas and selenocarbamates via selenium of isocyanates with bis(dimethylaluminum) selenide. Tetrahedron Lett. 52, 415- 417 (2011)  
+
+<|ref|>text<|/ref|><|det|>[[84, 411, 480, 485]]<|/det|>
+[38] Sharafi, M. et al. Crystal- packing- driven enrichment of atropoisomers. Angew. Chem. Int. Ed. 56, 7097- 7101 (2017)  
+
+<|ref|>text<|/ref|><|det|>[[84, 506, 481, 633]]<|/det|>
+[39] Haddon, R. C. and Scott, L. T. \(\pi\) - Orbital conjugation and rehybridization in bridged annulenes and deformed molecules in general: \(\pi\) - orbital axis vector analysis. Pure Appl. Chem. 58, 137- 142 (1986)  
+
+<|ref|>text<|/ref|><|det|>[[84, 655, 481, 756]]<|/det|>
+[40] Haddon, R. C. Comment on the relationship of the pyramidalization angle at a conjugated carbon atom to the sigma bond angles. J. Phys. Chem. A 105, 4164- 4165 (2001)  
+
+<|ref|>text<|/ref|><|det|>[[84, 778, 481, 904]]<|/det|>
+[41] Subhrajyoti B., Rik, D., Anna M. K. & Kristof H. Deformation- induced phosphorescence shift in a 2D elastically flexible organic single crystal: role of chalcogen-centered weak interactions. Chem. Sci. DOI:  
+
+<|ref|>text<|/ref|><|det|>[[555, 73, 773, 90]]<|/det|>
+10.1039/d2sc03729g (2022)  
+
+<|ref|>text<|/ref|><|det|>[[519, 111, 916, 213]]<|/det|>
+[42] Klapoetke, T. M., Krumm, B. & Scherr, M. Halogenation of dibenzoselenophene and dibenzo[1,2]diselenine Z. Anorg. Allg. Chem. 636, 1955- 1961 (2010)  
+
+<|ref|>text<|/ref|><|det|>[[519, 234, 916, 390]]<|/det|>
+[43] Bourhill, G et al. Experimental demonstration of the dependence of the first hyperpolarizability of donor- acceptor- substituted polyenes on the ground- state polarization and bond length alternation. J. Am. Chem. Soc. 116, 2619- 2620 (1994)  
+
+<|ref|>text<|/ref|><|det|>[[519, 410, 916, 485]]<|/det|>
+[44] McFarlane, S., McDonald, R. & Veinot, J.G.C. 9,9- Di- \(n\) - octyl- 9H- fluorene J. Phys. Chem. A 105, 4164- 4165 (2001)  
+
+<|ref|>text<|/ref|><|det|>[[519, 505, 916, 633]]<|/det|>
+[45] Chen, Z. F., Wannere, C. S., Corminboeuf, C., Puchta, R. & Schleyer, P. R. Nucleus- independent chemical shifts (NICS) as an aromaticity criterion. Chem. Rev., 105, 3842- 3888 (2005)  
+
+<|ref|>text<|/ref|><|det|>[[519, 654, 916, 784]]<|/det|>
+[46] Schleyer, P. R., Maerker, C., Dransfeld, A., Jiao, H. & Eikema Hommes, N. J. R. Nucleus- independent chemical shifts: a simple and efficient aromaficity probe. J. Am. Chem. Soc. 118, 6317- 6318 (1996)  
+
+<|ref|>text<|/ref|><|det|>[[519, 805, 916, 876]]<|/det|>
+[47] Fallah- Bagher- Shaidaei, H., Wannere, C. S., Corminboeuf, C., Puchta, R. & Schleyer, P. v. R. Which NICS aromaticity index for planar
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 73, 479, 92]]<|/det|>
+pi rings is best? Org. Lett. 8, 863- 866 (2006)  
+
+<|ref|>text<|/ref|><|det|>[[80, 111, 479, 241]]<|/det|>
+[48] Geuenich, D., Hess, K., Köhler, F. & Herges, R. Anisotropy of the induced current density (ACID), a general method to quantify and visualize electronic delocalization. Chem. Rev. 105, 3758- 3772 (2005)  
+
+<|ref|>text<|/ref|><|det|>[[80, 261, 479, 391]]<|/det|>
+[49] Moran, D., Stahl, F., Bettinger, H. F., Schaefer III, H. F. & Schaefer, P. v. R. Towards graphite: Magnetic properties of large polybenzoic hydrocarbons. J. Am. Chem. Soc. 125, 6746- 6752 (2003)  
+
+<|ref|>text<|/ref|><|det|>[[80, 411, 460, 430]]<|/det|>
+[50] Clar, E. The aromatic sextet. Wiley (1972)  
+
+<|ref|>text<|/ref|><|det|>[[80, 449, 479, 524]]<|/det|>
+[51] Lu, T. & Chen, F. W. Multiwfn: a multifunctional wavefunction analyzer. J. Comput. Chem. 33, 580- 592 (2012)  
+
+<|ref|>text<|/ref|><|det|>[[80, 544, 479, 644]]<|/det|>
+[52] Lu, T. & Chen, F. W. Quantitative analysis of molecular surface based on improved marching tetrahedra algorithm. J. Mol. Graph. Model. 38, 314- 323 (2012)  
+
+<|ref|>text<|/ref|><|det|>[[80, 666, 479, 797]]<|/det|>
+[53] Biglari, Z., & Rezaei, H. Investigation into molecular properties of pristine corannulene and its analogs with B/N- doping at hub location. Comput. Theor. Chem. 1190, 113002 (2020)  
+
+<|ref|>text<|/ref|><|det|>[[80, 818, 479, 890]]<|/det|>
+[54] Shoji, Y. et al. Hexathioalkyl sumanenes: an electron- donating buckybowl as a building block for supramolecular materials. Chem.  
+
+<|ref|>text<|/ref|><|det|>[[553, 73, 752, 90]]<|/det|>
+Sci. 8, 8405- 8410 (2017)  
+
+<|ref|>text<|/ref|><|det|>[[516, 111, 914, 268]]<|/det|>
+[55] Seiders, T. J., Baldridge, K. K., Grube, G. H. & Siegel. J. S. Structure/energy correlation of bowl depth and inversion barrier in corannulene derivatives: combined experimental and quantum mechanical analysis. J. Am. Chem. Soc. 123, 517- 525 (2001)  
+
+<|ref|>text<|/ref|><|det|>[[516, 288, 914, 390]]<|/det|>
+[56] Scott, L. T., Hashemi, M. M. & Bratcher, M. S. Corannulene bowl- to- bowl inversion is rapid at room temperature. J. Am. Chem. Soc. 114, 1920- 1921 (1992)  
+
+<|ref|>text<|/ref|><|det|>[[516, 410, 914, 565]]<|/det|>
+[57] Sygula, A., Abdourazak, A. H. & Rabideau, P. W. Cyclopentacorannulene: \(\pi\) - facial stereoselective deutero generation and determination of the bowl- to- bowl inversion barrier for a constrained buckybowl. J. Am. Chem. Soc. 118, 339- 343 (1996)  
+
+<|ref|>text<|/ref|><|det|>[[516, 585, 914, 714]]<|/det|>
+[58] Sakurai, H., Daiko, T., Sakane, H., Amaya, T. & Hirao, T. Structural elucidation of sumanene and generation of its benzylic anions. J. Am. Chem. Soc. 127, 11580- 11581 (2005)  
+
+<|ref|>text<|/ref|><|det|>[[516, 735, 914, 808]]<|/det|>
+[59] Amaya, T., Sakane, H., Muneishi, T. & Hirao. T. Bowl- to- bowl inversion of sumanene derivatives. Chem. Commun. 765- 767 (2008)  
+
+<|ref|>text<|/ref|><|det|>[[516, 830, 914, 875]]<|/det|>
+[60] Amaya, T. & Hirao, T. A molecular bowl sumanene. Chem. Commun. 47, 10524- 10535
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[120, 73, 175, 89]]<|/det|>
+(2011)
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[44, 42, 311, 70]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[44, 93, 765, 113]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[60, 130, 354, 150]]<|/det|>
+SupplementaryInformation.pdf
+
+<--- Page Split --->

preprint/preprint__0f042a5deb498dc8e8254b8bba7497dde5dc16c990ccc2c2976a7006ddcbf03f/images_list.json

@@ -0,0 +1,162 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_0.jpg",
+    "caption": "Age Associated Change in Cognition Across the Adult Lifespan",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 36
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_1.jpg",
+    "caption": "Cognitive Performance by Cohort",
+    "footnote": [],
+    "bbox": [
+      [
+        123,
+        450,
+        775,
+        600
+      ]
+    ],
+    "page_idx": 39
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_2.jpg",
+    "caption": "Objective Cognitive Performance and Subjective Cognitive Concerns Discriminate MCI",
+    "footnote": [],
+    "bbox": [
+      [
+        123,
+        640,
+        775,
+        830
+      ]
+    ],
+    "page_idx": 39
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_3.jpg",
+    "caption": "E Overall Study Adherence by Group - iPhone Device-Use Contributing Participants by Passive Data Donation",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 39
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_4.jpg",
+    "caption": "Overall, the enrollment rate across the United States was like the general population. In terms of absolute deviation, only a handful of states had noticeable over or under enrollment. California and North Carolina over enrolled by \\(1.3\\%\\) and \\(2.4\\%\\) respectively compared to their general population. Texas and New York were the largest under enrollers which under enrolled by \\(1.2\\%\\) each. The remainder of the states were all within \\(1\\%\\) of the reference levels. For a complete list of state-based enrollment and reference population statistics see Online Supplement Table 2.",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 40
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_5.jpg",
+    "caption": "App Structure",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 40
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_6.jpg",
+    "caption": "Educational Content",
+    "footnote": [],
+    "bbox": [
+      [
+        420,
+        240,
+        600,
+        407
+      ]
+    ],
+    "page_idx": 40
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_7.jpg",
+    "caption": "Burst Testing Onboarding",
+    "footnote": [],
+    "bbox": [
+      [
+        657,
+        240,
+        897,
+        407
+      ]
+    ],
+    "page_idx": 41
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_8.jpg",
+    "caption": "Testing Session",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 43
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_9.jpg",
+    "caption": "Burst Progress",
+    "footnote": [],
+    "bbox": [
+      [
+        727,
+        470,
+        900,
+        630
+      ]
+    ],
+    "page_idx": 44
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_10.jpg",
+    "caption": "Points Tab",
+    "footnote": [],
+    "bbox": [
+      [
+        123,
+        701,
+        300,
+        866
+      ]
+    ],
+    "page_idx": 45
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_11.jpg",
+    "caption": "Out of App Tasks",
+    "footnote": [],
+    "bbox": [
+      [
+        420,
+        701,
+        600,
+        866
+      ]
+    ],
+    "page_idx": 46
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_12.jpg",
+    "caption": "Stand Challenge",
+    "footnote": [],
+    "bbox": [
+      [
+        727,
+        701,
+        900,
+        866
+      ]
+    ],
+    "page_idx": 47
+  }
+]

preprint/preprint__0f379738303c091d25c79d22f8cc9636176b386206be7b6f97954732ccab96c7/images_list.json

@@ -0,0 +1,198 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Fig. 1. Clinical response to ICI across cancer types in patients with germline DNA replication repair deficiency. (a) Distribution of tumour types across 38 patients who developed 45 tumours. (b) Waterfall plot of all radiological responses in non-haematological malignancies. Values show the best fractional change in the 2 dimensions from baseline measurements as per RANO and RECIST criteria. Arrows point to representative T2-weighted FLAIR and T1-weighted contrast-enhanced MRI sequences in two patients showing flare and partial responses.",
+    "footnote": [],
+    "bbox": [
+      [
+        117,
+        110,
+        915,
+        401
+      ]
+    ],
+    "page_idx": 29
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Fig.2. Patient outcome and survival by tumour type. (a) Swimmer plot by patient and tumour type. (b) Kaplan-Meier (KM) estimates of overall survival for all patients (c) KM estimates of overall survival as per tumour type. Median survival for CNS tumours was 21.6 months. Median survival was not reached for non-CNS solid tumours. (d) KM estimates of progression free and overall survival for CNS tumours continuing ICI therapy. Note: prolonged median survival at 24 months (estimated 3-year OS=49.1%) despite initial radiological progression at a median of 9.9 months (estimated 3-year PFS=32%).",
+    "footnote": [],
+    "bbox": [
+      [
+        122,
+        93,
+        863,
+        430
+      ]
+    ],
+    "page_idx": 30
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Fig. 3. Genomic biomarkers, survival and response to PD1 blockade. (a) Onco-plot summarising the genomic features from 39 available paired tumour and germline exomes, and their clinical correlates. (b) Response and overall survival (OS) by single nucleotide variants (SNVs) per Mb. For survival analysis, median SNV burden was used. (c) SNVs as a function of MMRD (blue) and MMRD+PPD (orange) status (left), and response association with both SNV and RRD status. (c) Response and overall survival (OS) by microsatellite indels (MS-indels). For survival analysis, median MS-indel values were used. (d) Response and overall survival by total MS-indel count for MMRD and MMRD+PPD cancers separately. (e) Kaplan-Meier (KM) estimates using combined SNVs/Mb and MS-indel in all RRD cancers. (Abbreviations: MMRD: mismatch-repair deficiency; PPD: Polymerase-proofreading deficiency; MS-indel: microsatellite insertion/deletion.)",
+    "footnote": [],
+    "bbox": [
+      [
+        120,
+        88,
+        880,
+        450
+      ]
+    ],
+    "page_idx": 31
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Fig. 4. Genomic biomarkers, survival and response to PD1 blockade. (a) Onco-plot summarising the genomic features from 39 available paired tumour and germline exomes, and their clinical correlates. (b) Response and overall survival (OS) by single nucleotide variants (SNVs) per Mb. For survival analysis, median SNV burden was used. (c) SNVs as a function of MMRD (blue) and MMRD+PPD (orange) status (left), and response association with both SNV and RRD status. (c) Response and overall survival (OS) by microsatellite indels (MS-indels). For survival analysis, median MS-indel values were used. (d) Response and overall survival by total MS-indel count for MMRD and MMRD+PPD cancers separately. (e) Kaplan-Meier (KM) estimates using combined SNVs/Mb and MS-indel in all RRD cancers. (Abbreviations: MMRD: mismatch-repair deficiency; PPD: Polymerase-proofreading deficiency; MS-indel: microsatellite insertion/deletion.)",
+    "footnote": [],
+    "bbox": [
+      [
+        120,
+        88,
+        872,
+        425
+      ]
+    ],
+    "page_idx": 32
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Fig. 5. Tumour immune microenvironment, survival and response to PD-1 blockade. (a)",
+    "footnote": [],
+    "bbox": [
+      [
+        140,
+        116,
+        852,
+        485
+      ]
+    ],
+    "page_idx": 33
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_6.jpg",
+    "caption": "Fig.6. Characterization of the tumour flare response. (a-d) Analysis of 2 patients who had tumour-debulking prior to therapy and at the time of flare. (a, b) Total immune cell content in pre-therapy and at flare. (c, d) The corresponding CapTCR-sequencing and T-cell receptor clonotype analysis in these samples. (e, f) Immunohistochemistry for PD-L1 expression, and CD8-T-cell infiltration in the pre-therapy sample and at flare, as shown in the representative 20X images from the tumour sample in patient-1 (P33). (g) Representative flow cytometry plot showing activation of \\(\\mathrm{CD + }\\) T-cell (TIGIT and 4-1BB) from the blood sample of a patient before treatment initiation and at flare. (h) 41BB+ CD8+ T-cells in blood from responders without flare, non-responders and flare.",
+    "footnote": [],
+    "bbox": [
+      [
+        132,
+        87,
+        833,
+        682
+      ]
+    ],
+    "page_idx": 34
+  },
+  {
+    "type": "image",
+    "img_path": "images/Extended_Data_Figure_1.jpg",
+    "caption": "Extended Data Fig.1. KM estimates of progression-free and overall survival for (a) all tumours, (b) CNS tumours, (c) all solid tumours (CNS and non-CNS). (d) Progression free survival for non-CNS solid tumours versus CNS tumours \\((p = 0.01)\\) .",
+    "footnote": [],
+    "bbox": [
+      [
+        144,
+        123,
+        760,
+        485
+      ]
+    ],
+    "page_idx": 35
+  },
+  {
+    "type": "image",
+    "img_path": "images/Extended_Data_Figure_6.jpg",
+    "caption": "Extended Data Fig.6: CNS tumours: Immune microenvironment, survival and response to PD-1 blockade. Response and overall survival in CNS tumours by (a) PD-L1, (b) CD8, (c) CD3, (d) CD4, and (e) CD68 expression.",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 36
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Figure 1",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 37
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2",
+    "footnote": [],
+    "bbox": [
+      [
+        63,
+        66,
+        940,
+        500
+      ]
+    ],
+    "page_idx": 38
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Figure 3",
+    "footnote": [],
+    "bbox": [
+      [
+        50,
+        55,
+        943,
+        540
+      ]
+    ],
+    "page_idx": 39
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Figure 4",
+    "footnote": [],
+    "bbox": [
+      [
+        52,
+        54,
+        940,
+        490
+      ]
+    ],
+    "page_idx": 40
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Figure 5",
+    "footnote": [],
+    "bbox": [
+      [
+        50,
+        52,
+        940,
+        550
+      ]
+    ],
+    "page_idx": 41
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_6.jpg",
+    "caption": "Figure 6 Characterization of the tumour flare response. (a- d) Analysis of 2 patients who had tumour- debulking prior to therapy and at the time of flare. (a, b) Total immune cell content in pre- therapy and at flare. (c, d) The corresponding CapTCR- sequencing and T- cell receptor clonotype analysis in these samples. (e, f) Immunohistochemistry for PD- L1 expression, and CD8- T- cell infiltration in the pre- therapy sample and at flare, as shown in the representative 20X images from the tumour sample in patient- 1 (P33). (g)",
+    "footnote": [],
+    "bbox": [
+      [
+        66,
+        55,
+        844,
+        777
+      ]
+    ],
+    "page_idx": 42
+  }
+]

preprint/preprint__0f379738303c091d25c79d22f8cc9636176b386206be7b6f97954732ccab96c7/preprint__0f379738303c091d25c79d22f8cc9636176b386206be7b6f97954732ccab96c7.mmd

@@ -0,0 +1,652 @@
+
+# Mutation and Microsatellite Burden Predict Response to PD-1 Inhibition in Children with Germline DNA Replication Repair Deficiency  
+
+Uri Tabori ( \(\boxed{\bullet}\) uri.tabori@sickkids.ca) Hospital for Sick Children  
+
+Daniel Morgenstern Division of Haematology Oncology, The Hospital for Sick Children, Toronto, Canada https://orcid.org/0000- 0002- 4859- 1108  
+
+Sumedha Sudhaman Program in Genetics and Genome Biology, The Hospital for Sick Children, Toronto, Canada  
+
+Anirban Das The Hospital for Sick Children https://orcid.org/0000- 0001- 7653- 9529  
+
+Ailish Coblentz Department of Diagnostic Imaging, The Hospital for Sick Children, Toronto, Canada  
+
+Jill Chung Program in Genetics and Genome Biology, The Hospital for Sick Children, Toronto, Canada  
+
+Simone Stone Princess Margaret Cancer Centre, University Health Network, Toronto, Canada  
+
+Noor Alsafwani Division of Neuropathology, Department of Laboratory Medicine and Pathobiology, University of Toronto, Canada https://orcid.org/0000- 0002- 9449- 9708  
+
+Vanja Cabric Division of Haematology Oncology, The Hospital for Sick Children, Toronto, Canada  
+
+Liana Nobre Division of Haematology Oncology, The Hospital for Sick Children, Toronto, Canada  
+
+Vanessa Bianchi Program in Genetics and Genome Biology, The Hospital for Sick Children, Toronto, Canada  
+
+Melissa Edwards Program in Genetics and Genome Biology, The Hospital for Sick Children, Toronto, Canada  
+
+Lauren Sambira Division of Haematology Oncology, The Hospital for Sick Children, Toronto, Canada  
+
+Shlomi Constantin Tel Aviv University  
+
+Rina Dvir
+
+<--- Page Split --->
+
+Department of Pediatric Hematology- Oncology, Tel- Aviv Sourasky Medical Center, Tel- Aviv, IsraelMichal Yalon  
+
+The Chaim Sheba Medical Center  
+
+Gadi CampinoDepartment of Pediatric Hematology- Oncology, Sheba Medical Center, Ramat Gan, Israel  
+
+Shani CaspiDepartment of Pediatric Hematology- Oncology, Sheba Medical Center, Ramat Gan, Israel  
+
+Valerie LaroucheCentre Hospitalier Universitaire de Quebec- Université Laval  
+
+Alyssa ReddyDepartments of Neurology and Pediatrics, University of California, San Francisco, California, USA  
+
+Michael OsbornWomen's and Children's Hospital, North Adelaide, Australia  
+
+Gary MasonDepartment of Pediatrics, University of Pittsburgh School of Medicine, Pittsburgh, USA  
+
+Scott LindhorstNeuro- Oncology, Department of Neurosurgery, and Department of Medicine, Division of Hematology/Medical Oncology, Medical University of South Carolina, Charleston, USA  
+
+Annika BronsemaDepartment of Pediatric Hematology and Oncology, University Medical Center Hamburg- Eppendorf, Hamburg, Germany  
+
+Vanan MagimairajanDepartment of Pediatric Hematology- Oncology, Cancer Care Manitoba; Research Institute in Oncology and Hematology, University of Manitoba, Winnipeg, Canada  
+
+Enrico OpocherPediatric Hematology, Oncology and Stem Cell Transplant Division, Padua University Hospital, Padua, Italy  
+
+Rebecca De MolaOregon Health and Science University, Portland, Oregon, USA  
+
+Magnus SabelDepartment of Pediatrics, Institute of Clinical Sciences, Sahlgrenska Academy, University of Gothenburg & Queen Silvia Children's Hospital, Sahlgrenska University Hospital, Gothenburg, Swedenhttps://orcid.org/0000- 0002- 3072- 657X  
+
+Charlotta Frojd  
+
+Department of Pediatrics, Institute of Clinical Sciences, Sahlgrenska Academy, University of Gothenburg & Queen Silvia Children's Hospital, Sahlgrenska University Hospital, Gothenburg, Sweden  
+
+David SumeraerCharles University and University Hospital Motol  
+
+David Samuel
+
+<--- Page Split --->
+
+Valley Children's Hospital https://orcid.org/0000- 0001- 6850- 4932  
+
+Kristina Cole The Children's Hospital of Philadelphia  
+
+Stefano Chiaravalli Pediatric Unit, Fondazione IRCCS Istituto Nazionale dei Tumori, Milano, Italy  
+
+Maura Massimino Fondazione IRCCS Istituto Nazionale Tumori  
+
+Patrick Tomboc Department of Pediatrics, Ruby Memorial Hospital, West Virginia University, West Virginia, USA https://orcid.org/0000- 0002- 1255- 4719  
+
+David Ziegler Sydney Children's Hospital https://orcid.org/0000- 0001- 7451- 7916  
+
+Ben George Division of Hematology and Oncology, Department of Medicine, Medical College of Wisconsin, Milwaukee, Milwaukee, Wisconsin, USA  
+
+An Van Damme Department of Pediatric Hematology and Oncology, Saint Luc University Hospital, Université Catholique de Louvain, Brussels Belgium  
+
+Nobuko Hijaya Division of Hematology/ Oncology/ Stem Cell Transplantation, Columbia University Irving Medical Center, New York, USA  
+
+David Gass Atrium Health/ Levine Children's Hospital, Charlotte, USA  
+
+Rose McGee Cancer Predisposition Division, Oncology Department, St Jude Children's Research Hospital, Memphis, USA https://orcid.org/0000- 0001- 7392- 6515  
+
+Oz Mordechai Department of Pediatric Hematology Oncology, Rambam Health Care Campus, Haifa, Israel https://orcid.org/0000- 0002- 4340- 9290  
+
+Daniel Bowers UT Southwestern https://orcid.org/0000- 0002- 3947- 2481  
+
+Ted Laetsch Division of Oncology and Center for Childhood Cancer Research, Children's Hospital of Philadelphia, and Perelman School of Medicine at the University of Pennsylvania, Philadelphia, Pennsylvania, U  
+
+Alexander Lossos Department of Oncology, Leslie and Michael Gaffin Center for Neuro-Oncology, Hadassah- Hebrew University Medical Center, Jerusalem, Israel  
+
+Deborah Blumenthal
+
+<--- Page Split --->
+
+Neuro-Oncology Service, Tel-Aviv Medical Center, Sackler Faculty of Medicine, Tel Aviv University, Tel-Aviv, Israel  
+
+Tomasz Sarosiek  
+
+Lux Med Onkologia, Warsaw, Poland  
+
+Lee Yen  
+
+Department of Neurosurgery, Neurological Institute, Taipei Veterans General Hospital, Taipei, Taiwan  
+
+Jeffrey Knipstein  
+
+PRA Health Sciences https://orcid.org/0000- 0002- 0632- 2751  
+
+Anne Bendel  
+
+Children's of Minnesota https://orcid.org/0000- 0002- 2062- 0974  
+
+Lindsey Hoffman  
+
+Phoenix Children's Hospital, Phoenix, Arizona, USA  
+
+Sandra Luna- Fineman  
+
+Department of Pediatrics, Anschutz Medical Campus, Children's Hospital of Colorado, Colorado, USA  
+
+Stefanie Zimmermann  
+
+University Hospital Frankfurt, Paediatric Haematology and Oncology, Frankfurt, Germany  
+
+Isabelle Scheers  
+
+Pediatric Gastroenterology, Hepatology and Nutrition Unit, Cliniques Universitaires St Luc, Brussels, Belgium  
+
+Kim Nichols  
+
+St. Jude Children's Research Hospital https://orcid.org/0000- 0002- 5581- 6555  
+
+Michal Zapotocky  
+
+Department of Pediatric Hematology and Oncology, Second Faculty of Medicine, University Hospital Motol, Charles University, Prague, Czech Republic  
+
+Jordan Hansford  
+
+Royal Children's Hospital https://orcid.org/0000- 0001- 7733- 383X  
+
+John Maris  
+
+University of Pennsylvania and Children's Hospital of Philadelphia https://orcid.org/0000- 0002- 8088- 7929  
+
+Peter Dirks  
+
+Hospital for Sick Children https://orcid.org/0000- 0001- 5718- 6465  
+
+Michael Taylor  
+
+Sickkids Hospital  
+
+Abhaya Kulkarni  
+
+University of Toronto https://orcid.org/0000- 0002- 1706- 7004  
+
+Manohar Shroff  
+
+Department of Diagnostic Imaging, The Hospital for Sick Children, Toronto, Canada  
+
+Derek Tsang  
+
+Princess Margaret Cancer Centre, University Health Network https://orcid.org/0000- 0002- 9762- 6901
+
+<--- Page Split --->
+
+David Malkin The Hospital for Sick Children https://orcid.org/0000- 0001- 5752- 9763  
+
+Anita Villani Division of Haematology Oncology, The Hospital for Sick Children, Toronto, Canada  
+
+Melyssa Aronson Zane Cohen Centre for Digestive Diseases, Mount Sinai Hospital, Toronto, Canada  
+
+Carol Durno Zane Cohen Centre for Digestive Diseases, Mount Sinai Hospital, Toronto, Canada  
+
+Adam Shlien Hospital for Sick Children  
+
+Gad Getz Broad Institute https://orcid.org/0000- 0002- 0936- 0753  
+
+Yosef Maruvka Borad Institute  
+
+Pamela Ohashi Princess Margaret Cancer Centre  
+
+Cynthia Hawkins Hospital for Sick Children https://orcid.org/0000- 0003- 2618- 4402  
+
+Trevor Pugh Ontario Institute for Cancer Research  
+
+Eric Bouffet Hospital for Sick Children  
+
+## Article  
+
+Keywords: Mismatch repair, DNA polymerase, hypermutation, microsatellite instability, immune checkpoint inhibition, cancer, CMMR, Lynch, glioma  
+
+Posted Date: February 15th, 2021  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 155292/v1  
+
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Version of Record: A version of this preprint was published at Nature Medicine on January 6th, 2022. See the published version at https://doi.org/10.1038/s41591- 021- 01581- 6.
+
+<--- Page Split --->
+
+1 Mutation and Microsatellite Burden Predict Response to PD- 1 Inhibition in Children with Germline DNA Replication Repair Deficiency  
+
+4 Daniel Morgenstern \(^{1,6}\) , \(^{8}\) Sumedha Sudhaman \(^{2,49}\) , \(^{8}\) Anirban Das \(^{1,2,49,59}\) , Ailish Colbentz \(^{3}\) , Jil Chung \(^{2,49,62}\) , Simone Stone \(^{4}\) , Noor Alsafwani \(^{5,60}\) , Vanja Cabric \(^{1}\) , Liana Nobre \(^{1,49}\) , Vanessa Bianchi \(^{2,49}\) , Melissa Edwards \(^{2,49}\) , Lauren Sambira \(^{1}\) , Shlomi Constantinii \(^{7}\) , Rina Dvir \(^{8}\) , Michal Yalon- Oren \(^{9}\) , Gadi Abebe Campino \(^{9}\) , Shani Caspi \(^{9}\) , Valerie Larouche \(^{10}\) , Alyssa Reddy \(^{11}\) , Michael Osborn \(^{12}\) , Gary Mason \(^{13}\) , Scott Lindhorst \(^{14}\) , Annika Bronsema \(^{15}\) , Vanan Magimairajan \(^{16}\) , Enrico Opocher \(^{17}\) , Rebecca Loret De Mola \(^{18}\) , Magnus Sabel \(^{19}\) , Charlotta Frojd \(^{19}\) , David Sumeraer \(^{20}\) , David Samuel \(^{21}\) , Kristina Cole \(^{22}\) , Stefano Chiaravalli \(^{23}\) , Maura Massimino \(^{23}\) , Patrick Tomboc \(^{24}\) , David Ziegler \(^{25}\) , Ben George \(^{26}\) , An Van Damme \(^{27}\) , Nobuko Hijiya \(^{28}\) , David Gass \(^{29}\) , Rose McGee \(^{30}\) , Oz Mordechai \(^{31}\) , Daniel C. Bowers \(^{32}\) , Ted Laetsch \(^{33}\) , Alexander Lossos \(^{34}\) , Deborah T. Blumenthal \(^{35}\) , Tomasz Sarosiek \(^{36}\) , Lee Yi Yen \(^{37}\) , Jeffrey Knipstein \(^{38}\) , Anne Bendel \(^{39}\) , Lindsey Hoffman \(^{40}\) , Sandra Luna- Fineman \(^{41}\) , Stefanie Zimmermann \(^{42}\) , Isabelle Scheers \(^{43}\) , Kim E. Nichols \(^{44}\) , Michal Zapotocky \(^{20}\) , Jordan R. Hansford \(^{45}\) , John M. Maris \(^{33}\) , Peter Dirks \(^{46,47,49}\) , Michael D. Taylor \(^{46,47,49}\) , Abhaya V. Kulkarni \(^{46,48}\) , Manohar Shroff \(^{8}\) , Derek S. Tsang \(^{50}\) , Anita Villani \(^{1,6}\) , Melyssa Aronson \(^{51}\) , Carol Durno \(^{51}\) , Adam Shlien \(^{2,61}\) , David Malkin \(^{1,2,6}\) , Gad Getz \(^{52,53}\) , Yosef E. Maruvka \(^{54}\) , Pamela S. Ohashi \(^{4,55}\) , Cynthia Hawkins \(^{49,56,61}\) , Trevor J. Pugh \(^{57,58}\) , Eric Bouffet \(^{1,6}\) , \*Uri Tabori \(^{1,2,6,49}\) .  
+
+21 \(^{8}\) DM, SS and AD are joint first authors and contributed equally to the work.  
+
+Number of pages: 40  
+
+Figures: 6. Tables: 0  
+
+Extended Data: Figures: 9. Tables: 3  
+
+Word Count: Summary: 197; Text: 3500  
+
+Keywords: Mismatch repair, DNA polymerase, hypermutation, microsatellite instability, immune checkpoint inhibition, cancer, CMMR, Lynch, glioma  
+
+## Corresponding author:  
+
+\*Uri Tabori, MD  Staff Haematologist/Oncologist, Division of Haematology/Oncology.  The Hospital for Sick Children. 555 University Avenue, Toronto, ON, Canada, M5G 1X8.  Professor of Paediatrics and Medical Biophysics, University of Toronto.  Senior Scientist, The Arthur and Sonia Labatt Brain Tumour Research Centre.  Tel: (416) 813- 7654, ext. 201503, Fax: (416) 813- 5327  E- mail: uri.tabori@sickkids.ca
+
+<--- Page Split --->
+
+## Author affiliations:  
+
+1. Division of Haematology Oncology, The Hospital for Sick Children, Toronto, Canada  
+2. Program in Genetics and Genome Biology, The Hospital for Sick Children, Toronto, Canada  
+3. Department of Diagnostic Imaging, The Hospital for Sick Children, Toronto, Canada  
+4. Princess Margaret Cancer Centre, University Health Network, Toronto, Canada  
+5. Division of Neuropathology, Department of Laboratory Medicine and Pathobiology, University of Toronto, Canada  
+6. Department of Paediatrics, University of Toronto, Toronto, Canada  
+7. Department of Paediatric Neurosurgery, Dana Children's Hospital, Tel-Aviv, Israel  
+8. Department of Pediatric Haematology-Oncology, Tel-Aviv Sourasky Medical Centre, Tel-Aviv, Israel  
+9. Department of Paediatric Haematology-Oncology, Sheba Medical Centre, Ramat Gan, Israel  
+10. Haematology/Oncology Centre Hospitalier Universitaire de Quebec, Quebec, Canada  
+11. Departments of Neurology and Paediatrics, University of California, San Francisco, California, USA  
+12. Women's and Children's Hospital, North Adelaide, Australia  
+13. Department of Paediatrics, University of Pittsburgh School of Medicine, Pittsburgh, USA  
+14. Neuro-Oncology, Department of Neurosurgery, and Department of Medicine, Division of Haematology/Medical Oncology, Medical University of South Carolina, Charleston, USA  
+15. Department of Paediatric Haematology and Oncology, University Medical Centre Hamburg-Eppendorf, Hamburg, Germany  
+16. Department of Paediatric Haematology-Oncology, Cancer Care Manitoba; Research Institute in Oncology and Haematology, University of Manitoba, Winnipeg, Canada  
+17. Paediatric Haematology, Oncology and Stem Cell Transplant Division, Padua University Hospital, Padua, Italy  
+18. Oregon Health and Science University, Portland, Oregon, USA  
+19. Department of Paediatrics, Institute of Clinical Sciences, Sahlgrenska Academy, University of Gothenburg & Queen Silvia Children's Hospital, Sahlgrenska University Hospital, Gothenburg, Sweden  
+20. Department of Paediatric Haematology and Oncology, Second Faculty of Medicine, University Hospital Motol, Charles University, Prague, Czech Republic  
+21. Department of Paediatric Oncology, Valley Children's Hospital, Madera, California, USA  
+22. Division of Oncology and Centre for Childhood Cancer Research, Children's Hospital of Philadelphia, Department of Paediatrics, University of Pennsylvania School of Medicine, Pennsylvania, USA  
+23. Paediatric Unit, Fondazione IRCCS Istituto Nazionale dei Tumori, Milano, Italy  
+24. Department of Paediatrics, Ruby Memorial Hospital, West Virginia University, West Virginia, USA  
+25. Kids Cancer Centre, Sydney Children's Hospital, Randwick, New South Wales, and School of Women's and Children's Health, UNSW Sydney, Sydney, NSW, Australia  
+26. Division of Haematology and Oncology, Department of Medicine, Medical College of Wisconsin, Milwaukee, Milwaukee, Wisconsin, USA  
+27. Department of Paediatric Haematology and Oncology, Saint Luc University Hospital, Université Catholique de Louvain, Brussels Belgium  
+28. Division of Haematology/Oncology/Stem Cell Transplantation, Columbia University Irving Medical Centre, New York, USA  
+29. Atrium Health/Levine Children's Hospital, Charlotte, USA  
+30. Cancer Predisposition Division, Oncology Department, St Jude Children's Research Hospital, Memphis, USA  
+31. Department of Paediatric Haematology Oncology, Rambam Health Care Campus, Haifa, Israel  
+32. Department of Paediatrics, The University of Texas Southwestern Medical School, Dallas, USA  
+33. Division of Oncology and Centre for Childhood Cancer Research, Children's Hospital of Philadelphia, and Perelman School of Medicine at the University of Pennsylvania, Philadelphia, Pennsylvania, USA  
+34. Department of Oncology, Leslie and Michael Gaffin Centre for Neuro-Oncology, Hadassah-Hebrew University Medical Centre, Jerusalem, Israel  
+35. Neuro-Oncology Service, Tel-Aviv Medical Centre, Sackler Faculty of Medicine, Tel Aviv University, Tel-Aviv, Israel  
+36. Lux Med Onkologia, Warsaw, Poland  
+37. Department of Neurosurgery, Neurological Institute, Taipei Veterans General Hospital, Taipei, Taiwan  
+38. Division of Paediatric Haematology/Oncology/BMT, Medical College of Wisconsin, Milwaukee, Wisconsin, USA  
+39. Department of Paediatric Haematology-Oncology, Children's Hospitals and Clinics of Minnesota, Minnesota, USA  
+40. Phoenix Children's Hospital, Phoenix, Arizona, USA  
+41. Department of Paediatrics, Anschutz Medical Campus, Children's Hospital of Colorado, Colorado, USA
+
+<--- Page Split --->
+
+97 42. University Hospital Frankfurt, Paediatric Haematology and Oncology, Frankfurt, Germany98 43. Paediatric Gastroenterology, Hepatology and Nutrition Unit, Cliniques Universitaires St Luc, Brussels, Belgium100 44. Division of Cancer Predisposition, St Jude Children's Research Hospital, Memphis, USA101 45. Children's Cancer Centre, Royal Children's Hospital, Murdoch Children's Research Institute; Department of Paediatrics, University of Melbourne, Parkville, Victoria, Australia102 46. Division of Neurosurgery, The Hospital for Sick Children, Toronto, ON, Canada103 47. Developmental and Stem Cell Biology Program, The Hospital for Sick Children, Toronto, Canada104 48. Child Health Evaluative Sciences, Research Institute, The Hospital for Sick Children, Toronto, Canada105 49. The Arthur and Sonia Labatt Brain Tumour Research Centre, The Hospital for Sick Children, Toronto, Canada106 50. Radiation Medicine Program, Princess Margaret Cancer Centre, Toronto, Canada107 51. Zane Cohen Centre for Digestive Diseases, Mount Sinai Hospital, Toronto, Canada108 52. Massachusetts General Hospital Centre for Cancer Research, Charlestown, Massachusetts, USA.109 53. Broad Institute of Harvard and MIT, Cambridge, Massachusetts, USA110 54. Biotechnology and Food Engineering. Technion. Haifa. Israel.111 55. Department of Immunology, University of Toronto, Canada112 56. Department of Paediatric Laboratory Medicine, The Hospital for Sick Children, Toronto, Canada113 57. Ontario Institute for Cancer Research, Princess Margaret Cancer Centre, Toronto, Canada114 58. Department of Medical Biophysics, University of Toronto, Toronto, Canada115 59. Department of Paediatric Haematology/ Oncology, Tata Medical Centre, Kolkata, India116 60. Department of Pathology, College of Medicine, Imam Abdulrahman Bin Faisal University (IAU), Dammam, Saudi Arabia117 61. Department of Laboratory Medicine and Pathobiology, Faculty of Medicine, University of Toronto, Toronto, Canada118 62. Institute of Medical Science, Faculty of Medicine, University of Toronto, Toronto, Canada
+
+<--- Page Split --->
+
+## Abstract  
+
+122 Abstract123 Cancers arising from germline DNA mismatch- repair or polymerase- proofreading deficiencies124 (MMRD and PPD) in children harbour the highest mutational and microsatellite125 insertion/deletion (MS- indel) burden in humans and are lethal due to inherent resistance to126 chemo- irradiation. Although immune checkpoint inhibitors (ICI) have failed to benefit children127 in previous studies, we hypothesized that hypermutation caused by MMRD and PPD will128 improve outcomes following ICI in these patients. ICI treatment of 45 progressive/recurrent129 tumours from 38 patients revealed durable objective responses in the majority, culminating in130 3- year survival of \(41.4\%\) . High mutation burden predicted response for ultra- hypermutant131 cancers ( \(>100\) mutations/Mb) enriched for combined MMRD+PPD, while MS- indels predicted132 response in MMRD tumours with lower mutation burden (10- 100 mutations/Mb). Further, both133 mechanisms were associated with increased immune infiltration even in "immunologically-134 cold" tumours such as gliomas, contributing to the favorable response. Pseudo- progression135 (flare) was common and associated with immune activation in both the tumour136 microenvironment and systemically. Further, patients with flare continuing ICI treatment137 achieved durable responses. Our study demonstrates improved survival for patients with138 tumours not previously known to respond to ICI, including CNS and synchronous cancers, and139 identifies the dual roles of mutation burden and MS- indels in predicting sustained responses to140 immunotherapy.
+
+<--- Page Split --->
+
+## Introduction  
+
+Accurate DNA replication in eukaryotic cells is ensured by the DNA polymerases Pol \(\delta\) and Pol \(\epsilon\) , which control base incorporation and proofreading, and the mismatch repair (MMR) system that undertakes post- replication surveillance. Germline and somatic mutations in POLD1 and POLE (termed polymerase- proofreading deficiency: PPD), or the MMR genes (MLH1, MSH2, MSH6, PMS2) result in DNA replication repair deficiency (RRD). RRD is a major driver of hypermutation and microsatellite instability (MSI) in several adult and pediatric cancers. Both germline PPD, and monoallelic germline pathogenic variants in the MMR genes (Lynch syndrome) lead to adult- onset gastrointestinal and genitourinary cancers. In contrast, biallelic loss of MMR function in the germline causes constitutional mismatch repair deficiency (CMMRD) syndrome, a highly penetrant and aggressive cancer- predisposing condition. Affected individuals typically develop cancers at a young age, most commonly malignant gliomas, gastrointestinal and haematological malignancies. These cancers are frequently chemo- resistant, resulting in poor survival for affected patients. Indeed, individuals with CMMRD rarely survive beyond early adulthood. The burden of CMMRD cancers is significant in areas with high consanguinity, including many developing countries and among indigenous populations.  
+
+RRD cancers are universally hypermutant due to the continuous acquisition of multiple somatic mutations. Tumour mutation burden (TMB) of cancers driven by germline RRD is 100 to 1000 times higher than MMR- intact pediatric cancers. Further, many MMRD tumours acquire a secondary somatic mutation in POLD1/POLE leading to combined MMR+PPD, characterised by ultra- hypermutation (>100 mutations/Mb). As a result, RRD cancers harbour the highest TMB observed among all human cancers. Hypermutant cancers such as melanoma and lung cancer, which are driven by ultraviolet light and smoking, respectively, respond to immune checkpoint inhibitors (ICI) targeting programmed death 1 (PD- 1) signalling. However, despite the dramatic anti- tumour effects reported in several hypermutant adult cancers, these responses are sustained in only a subset of patients. The role of TMB in determining the nature and duration of response is not well- established. Other studies have also raised questions regarding the role of TMB and PD- ligand 1 (PD- L1) expression as robust biomarkers of response to ICI. In contrast, MMR- deficient colorectal carcinomas are responsive to ICI due to excess MSI suggesting that biomarkers like TMB, MSI or PD- L1 expression may not be individually sufficient to drive immune responses following ICI across diverse cancer types.
+
+<--- Page Split --->
+
+Most cancers, including hypermutant brain tumours are considered ‘immunologically cold’ and are unresponsive to ICI.\(^{30}\) Importantly, ICI did not result in significant responses in multiple large pediatric clinical trials and is considered ineffective in the management of solid tumours in childhood and adolescence.\(^{24,31- 34}\) Finally, for all solid tumours receiving immunotherapy, the distinction between true tumour progression and an inflammatory pseudo-progression is a major challenge and a barrier to effective therapy.\(^{35- 38}\)  
+
+Despite the lack of response to ICI observed in children in previous studies, we hypothesized that cancers originating from germline RRD may benefit from ICI due to their excess mutational load.\(^{12}\) We further postulated that cancers driven by MMRD- only, PPD, or combined RRD (MMRD+PPD) will respectively exert their own unique mutational spectrum, driving local and systemic immune reactions, which would help shed light on the mechanisms of both response and pseudo- progression following ICI.  
+
+To address these hypotheses, we conducted a large, registry- based study, leveraging systematically- collected data gathered both retrospectively and prospectively through the International RRD Consortium.\(^{2,7,10,2,39}\) This enabled us to evaluate outcomes and predictors of response to anti- PD- 1 therapy in children with cancers driven by germline RRD. Uniquely, this also provided us the opportunity to investigate the efficacy of ICI in individuals with synchronous malignancies who are otherwise excluded from conventional clinical trials.  
+
+## Results  
+
+Thirty- eight patients who developed 45 cancers were treated with PD- 1 inhibitors between May 2015 and March 2019 and followed by the International RRD Consortium registry study group (Methods and Extended Data Table S1). While guidelines for treatment were provided by the consortium, and regular communication was organized with the treating team for ongoing radiological monitoring, response assessment, and management of adverse effects (Methods), the patients were not treated prospectively as part of a clinical trial. The PD- 1 inhibitor used was either nivolumab (n=34, 75%) or pembrolizumab (n=11, 25%), as per availability and choice of the treating team. All patients had germline RRD, diagnosed as CMMRD (n=28, 74%), Lynch (n=8, 21%), or PPD (n=2, 5%) syndromes (Extended Data Table S1). Median age at treatment was 12.1 years (range: 3.1- 28.1) for patients with CMMRD, and 15.7 years (range: 8.5- 43.4) for those with Lynch syndrome (p=0.07). Seven cancer types were included and classified into 3 major groups: central nervous system (CNS) tumours (n=31, 69%; disseminated: 2, 6%), non- CNS solid tumours (n=11, 24%; disseminated: 7, 64%), and haematological malignancies (n=3, 7%) (Fig.1a). The majority (n=43, 93%) of cancers were
+
+<--- Page Split --->
+
+progressive/recurrent after failure of first- line therapy (Extended Data Table S1). Three patients with gastrointestinal cancers received ICI directly following surgery: two of whom had synchronous CNS tumours and one who had metastatic disease. Data cutoff for outcomes was October 2019.  
+
+Responses and/or stable disease were observed in 25/45 (55.5%) tumours, with most of the responses (n=20; 80%) being sustained at a median follow- up of 1.87 years at the time of this analysis. Central radiological review (RANO and RECIST criteria; Methods) \(^{40,41}\) revealed complete response in 6 (17%), partial response in 9 (25%), stable disease in 7 (19%), and progressive disease in 14 (39%) (Fig.1b,c). Among the 7 patients with synchronous malignancies, responses in both tumours were seen in one patient, and at least in one tumour in 4 patients (Fig.2a). The three patients with haematological malignancies had progressed at a median time of 4.5 months after starting ICI therapy.  
+
+Of note, 12 (27%) tumours exhibited early radiological findings of edema and enhancement, suggestive of peri- tumoural inflammation or tumour progression (Fig.1b,d and 2a). This phenomenon occurred at a median of 34 days (range 7- 74) from treatment initiation and has been termed tumour 'flare.' \(^{42}\) These patients presented with acute clinical deterioration with headache, bone or abdominal pain, depending on the location of their tumours. Eight patients (6 with CNS and 2 with non- CNS solid tumours) stopped therapy and died. Importantly, 4 patients (3 with CNS and 1 with non- CNS solid tumour) who continued to receive ICI with adequate supportive care subsequently demonstrated objective responses. Because this suggested pseudo- progression, \(^{35,38,43}\) these tumours were studied in more detail.  
+
+Estimated 3- year overall survival (OS) was 41.4% (95% CI: 38.5, 44.2) (Fig.2b), with 18 (47%) patients being alive at the time of last follow- up (Fig.2a). This is noteworthy considering the refractory nature of their cancers. Analysis by cancer type revealed that non- CNS solid tumours had a significantly better survival compared to CNS tumours (Fig.2c and Extended Data Fig.1d; p=0.01). Nevertheless, the OS of 39.3%, (95% CI; 36.3, 42.3) and PFS of 26.9% (95% CI; 23.2, 30.6) for patients with recurrent/progressive CNS tumours is a dramatic improvement compared to the historically rapidly fatal outcomes (Fig.2c,d and Extended Data Fig.1c) \(^{7,9,12,44}\) . All patients with non- CNS solid tumours continuing ICI had durable responses and are alive at a median follow- up of 2.6 years (range: 0.38- 3.5). Remarkably, 13 patients with CNS tumours continued to survive for an additional 9.6 months (median; range: 1.5- 27) after radiological progression (Fig.2a,d and Extended Data Table S1). Collectively, these data suggest that late and continued responses to immunotherapy possibly due to the dynamic clonal evolution and obligatory mutation accumulation in RRD cancers \(^{10}\) .
+
+<--- Page Split --->
+
+## Molecular determinants of response to immunotherapy  
+
+To better understand the molecular determinants of response of RRD cancers to ICI, biopsy specimens and blood samples were collected before and during therapy from the patients for centralized analysis (Methods). Whole exome analysis of tumours (n=39, Fig.3) revealed high variability in the number of single nucleotide variants (SNV), including ultra- hypermutation (median: 233.8 mutations/Mb, range: 3.4- 912), which was associated with tumour genotype. MMRD- only cancers (n=16) had significantly fewer SNVs (median: 15.8 mutations/Mb) than MMRD+PPD cancers (n=23, median 391.4 mutations/Mb; p <0.0001; Fig.4b). This correlated with germline status, as 21 (67.7%) cancers in CMMRD patients harboured MMRD+PPD (median: 398.98 mutations/Mb), while all cancers in individuals with Lynch syndrome lacked somatic PPD (median: 21.76 mutations/mb; p=0.03). Both cancers originating from germline PPD, one colorectal carcinoma (P13; ICI.29) and one glioblastoma (P17; ICI.33) had acquired somatic MMRD resulting in ultra- hypermutation (Fig.2a and 3).  
+
+Analysis of COSMIC signatures, which reflect the imprints of the underlying RRD mutational processes,45,46 revealed that signatures 6 and 26 were enriched in MMRD- only cancers, and that all tumours in patients with Lynch syndrome exhibited signature 6. In contrast, signatures 14 and 15 were frequent in cancers in patients with CMMRD, and signature 14 was enriched in tumours with MMRD+PPD, highlighting the unique and potential diagnostic role of these signatures in determining germline predisposition in RRD cancers.45  
+
+As questions exist regarding the contribution of tumour- intrinsic characteristics such as mutation load and MS- indels to ICI response, we sought to determine whether response is modulated by a dynamic interaction between these factors.47 TMB correlated significantly with both response and survival. This was true for tumours with higher total SNVs/Mb (p=0.003, Fig.4a), as well as non- synonymous variants, synonymous variants and total indels (Extended Data Fig.2a- d). Both response and survival were also associated with a higher tumour neoantigen load (Extended Data Fig.2e). Remarkably, response and survival were predicted by RRD status. Cancers with MMRD+PPD (median: 392.7 mutations/Mb) had higher likelihood of response compared with cancers with MMRD alone (median: 15.7 mutations/Mb) (p<0.0001) (Fig.4b).  
+
+Recent work has shown that response in MMRD tumours is related to the degree of genome- wide MS- indels.3,48,49 Across our entire cohort, total MS- indels, calculated by MS- mutect (Methods),50 were not predictive of tumour response, but still predicted improved patient survival (Fig.4c). We did not find a significant association between high MS- indel and
+
+<--- Page Split --->
+
+TMB (p=0.47), thereby suggesting independent roles for both carcinogenic mechanisms for likely different tumour subsets. We therefore hypothesized that MS-indels are immunogenic in MMRD- only cancers, where the contribution of TMB is less dominant. To test this, we analyzed MMRD- only and MMRD+PPD tumours separately (Fig.4d). In MMRD- only cancers, total MS-indels were higher in responders than in non- responders, and were significantly associated with survival (p=0.025, Fig.4d). This was not observed in MMRD+PPD cancers with high TMB (Fig.4d). Further, in MMRD- only cancers, high TMB failed to predict response (p=0.52). Combining the prediction models for both these types of RRD cancers (MMRD- only and MMRD+PPD), high SNVs and total MS-indels together strongly predicted improved survival (Fig.4e, p=0.0024).  
+
+To determine whether SNVs and MS- indels drive response within a more homogeneous cancer type, we interrogated both these genomic signatures in CNS tumours. Both components contributed independently to the response and survival (Extended Data Fig.3c,g). In multivariate analysis (Methods), high mutation burden remained the most significant predictor for response (p=0.01) and OS (p=0.05) (Extended Data Table S2 and S3), highlighting the important contribution of a combined MMRD+PPD status in determining response to ICI. Collectively, these data suggest dual roles for TMB and MS- indels in determining immunotherapy responses in RRD cancers.  
+
+## Immune microenvironment and response to therapy in childhood RRD cancers  
+
+Next, we examined whether the RRD subgroups affect the tumour micro- environment and response to therapy. We tested multiple immune markers using immunohistochemistry for immune cell infiltration (CD3, CD4, CD8 and CD68) and checkpoint ligand (PD- L1) expression (Methods, Fig.5 and Extended Data Fig.4). All immune markers were scored blindly by our central pathologist (CH, Methods).  
+
+PD- L1 expression correlated with both response and improved survival (p=0.04, Fig.5a). Overall, increased lymphocytic infiltration within the tumour microenvironment correlated with response (Extended Data Fig.4a,b). Specifically, high CD8- T- cell infiltration predicted both response and improved OS (p=0.0002, Fig.5b). All non- CNS solid tumours including those with MMRD- only, harboured high MSI, exhibited high CD8+T- cell infiltration (Extended Data Fig.5a,b), and responded to ICI. (Extended Data Fig.5a-c). This corroborates previous reports in which MMR- deficient gastrointestinal tumours had high CD8+ T- cell infiltration. High T- cell infiltration was also observed in CNS tumours which are traditionally considered an ‘immune- privileged’ site. CNS tumours with a high mutation burden and
+
+<--- Page Split --->
+
+MMRD+PPD not only had increased CD8+ T- cell infiltration, but also significantly higher expression of PD- L1 (p<0.04; Fig.5c). This suggests that in the setting of ultra- hypermutation driven by combined MMRD+PPD and high genomic MS- indels, the increased activation of the immune microenvironment can explain the remarkable responses seen in the CNS tumours. Indeed, in these CNS tumours, high CD8+T- cell infiltration was associated with improved OS (Extended Data Fig.6a,b). Collectively, tumours exhibiting high expression of immune markers and a favorable genomic profile (high SNVs or MS- indels) had a 3- year OS of \(87.8\%\) (95% CI; 84, 91.5) as compared to \(33.2\%\) (95% CI; 27.2, 39.2) for tumours lacking these biomarkers (p=0.005, Fig.5d).  
+
+## Tumour flare is an immune reaction to therapy  
+
+To determine etiology of the flare phenomenon, we first analyzed the genomic and immune markers of these tumours. Cancers developing flare had pre- treatment genomic and immune characteristics similar to the responders without flare (Methods; Extended Data Fig.7a- d). We then compared the pre- and on- therapy tumours in two patients who had further surgical debulking at flare. Transcriptomic analysis and immune inference using single state deconvolution (Methods) revealed an increase in the overall immune cell expression at flare (Fig.6a,b). Notably, transcriptome signaling revealed that activated CD8+ T- cells were significantly increased in both samples following ICI. Using T- cell receptor clonotype analysis (TCR, Fig.6c,d), we observed a dramatic increase of T- cell repertoire in the post- treatment samples as compared to their baseline. One sample demonstrated an increase in both clonality and diversity of the T- cell population, with some original clones expanding during flare (Fig.6c), whereas for the other sample, there was reduction in diversity but significantly heightened clonality in the T- cell population (Fig6d). Additionally, both increased CD8+ T- cell infiltration and PD- L1 expression were observed in both tumours during flare when compared to their pretreatment samples (Fig.6e,f and Extended Data Fig.8a,b). These observations suggest a pre- existing (specific) immune response and further ICI- driven (non- specific and specific) intra- tumoral immune expansion at flare.  
+
+We next investigated whether this immune activation could be observed systemically. We performed serial flow cytometry analysis of blood samples from multiple patients, prior to treatment initiation and within the first 90 days of initiation of ICI (Methods). Flare was associated with an expansion of peripheral CD8+T- cells expressing Ki67 (a marker of proliferation) and 4- 1BB (a member of the TNF receptor family) (Extended Data Fig.9a- d and Fig.6g). Detection of Ki67+ CD8+ T- cells has been previously reported to be associated with
+
+<--- Page Split --->
+
+superior response to PD- 1 blockade. \(^{53}\) 4- 1BB is well- known as a marker of T- cell activation and has co- stimulatory activity for activated T- cells. \(^{54,55}\) Flare was associated with higher proportion of 4- 1BB+ CD8+ T- cells in the peripheral blood as compared to non- responders, as well as responders without flare (Fig.6h). Studies have previously shown that 4- 1BB+ CD8+ T- cells correlate with response to PD- 1 blockade. \(^{56}\) Importantly, we did not find an increase in the 4- 1BB+ CD4+ T- cell population (Extended Data Fig.9e,f), supporting our hypothesis that CD8+ T- cells expressing 4- 1BB reflect the expansion of a tumour- specific response. Taken together with the immune profile of the tumour microenvironment, we conclude that flare following PD- 1 blockade is associated with proliferation of tumour- antigen reactive T- cells.  
+
+## Discussion  
+
+Our study reveals dramatic responses to PD- 1 blockade and an associated improved survival for relapsed/refractory hypermutant cancers in children and young adults with germline DNA replication repair deficiency. Several insights can be derived from the sustained responses in different tumour types, and the contributions of TMB, MS- indels and the microenvironment to both response and flare.  
+
+The unique responses to ICI in children with germline RRD are different from previous observations by several groups. First, our data starkly contrast with the lack of ICI efficacy in childhood cancers in general \(^{32 - 34}\) , as well as specifically for progressive paediatric brain tumours \(^{57}\) . The lack of response in paediatric cancers, which is independent of PD- L1 expression \(^{24}\) or systemic immune activation \(^{24,33}\) , has been attributed to the low mutation burden \(^{2,10,12}\) , a different immunomodulatory role of the gut microbiome \(^{58}\) at a very young age \(^{24,32,34}\) , low expression of major histocompatibility complex \(^{59}\) and the predominance of macrophages in the tumour micro- environment \(^{60,61}\) . Some of these causes may need to be reexamined in view of our data.  
+
+Second, more than half of RRD paediatric CNS tumours had partial to complete objective responses, resulting in a median survival of 2 years when ICI was continued. This is remarkable, as historically, rapid progression with a median post- relapse survival of merely 2.6 months has been reported in children with RRD high- grade glioma. \(^{62}\) In contrast, hypermutant gliomas in adults fail to respond to ICI. These are different from RRD gliomas as they harbour relatively lower TMB with late acquisition of sub- clonal mutations during tumorigenesis possibly due to treatment- related secondary MMR deficiency, which is associated with less MS- indels, and lack of robust neoantigens. \(^{63 - 65}\) This is accompanied by multi- layered immunosuppression with a non- inflamed tumour microenvironment dominated
+
+<--- Page Split --->
+
+by a myeloid infiltrate, exhausted T- cell phenotype, and secretion of suppressive cytokines/molecules and hostile physical factors (hypoxia, acidosis and nutrient deprivation) \(^{63,64}\) .  
+
+Third, all patients who continued ICI therapy for non- CNS solid tumours including disseminated cancers responded and are alive at a median follow- up of 2.6 years. This compares favourably with MMR- deficient cancers in older patients (median age: 60 years, versus 12.3 years in our study), in whom late failures were noted, resulting in \(50 - 55\%\) survival at 2 years. \(^{26,27,66,67}\)  
+
+Collectively, the dramatic responses and favourable outcome observed in childhood RRD cancers can be explained by several key biological features. First, the earlier onset \(^{10}\) , combined with significantly higher mutations \(^{2}\) , MS- indels \(^{3}\) , and neoantigen burden drive CD8+ T- cell activation which is especially robust in children and adolescents when compared to older patients \(^{68 - 70}\) . Second, the additional loss of the polymerase- proofreading mechanism confers genomic mutational signatures (both MS- signatures \(^{3}\) and COSMIC signatures 10, 20, and 14) which may play unique immunogenic role in determining response and survival. Third, the exceptionally high rate of obligatory and continuous accumulation of mutations in combined MRRD+PPD cancers \(^{10}\) likely confers ongoing immunogenicity, contributing to immune- surveillance leading to both the durable and the delayed responses observed in our cohort.  
+
+Additional insights from this study include the ongoing and delayed responses observed in tumours that progressed on ICI therapy. This resulted in the difference between progression- free (post- ICI) and overall survival for patients who continued treatment after a second progression while on ICI therapy (Fig.2d and Extended Data Fig.1b,d). Additionally, responses in a tissue- agnostic manner were seen in a patient with synchronous tumours exhibiting favourable genomic and immune biomarkers for response (P01). As cancer immune surveillance is different than the irreversible resistance which occurs upon progression after chemo- radiation approaches, these data support the exploration of neoadjuvant \(^{71,72}\) , maintenance, and combinatorial uses of ICI in these patients, to limit toxicities and improve effectiveness of first- line strategies.  
+
+There is increasing evidence that responses to ICI cannot be predicted by single biomarker. \(^{73,74,19,75}\) Our data confirm that this is true in RRD cancers which are driven by dysfunction in both SNV and MS- indel repair. Although initially TMB was thought to be the sole contributor to immune response in hypermutant cancers \(^{22}\) , indels and MS- indels have been suggested to be the main drivers of response to ICI in MMR- deficient cancers \(^{3,49,76}\) . We add a
+
+<--- Page Split --->
+
+new dimension to this concept by revealing that in cancers driven by MMRD- only, TMB is relatively lower, and MS- indels drive response, while in MMRD+PPD cancers, the role of MS- indels is attenuated, and TMB is the major driver of response. The dual roles of both mutational mechanisms also affect the microenvironment, with upregulation of PD- L1 and infiltration of CD8+ T- cells. Importantly, the combination of both genomic mechanisms and immune markers are powerful predictors of survival in RRD cancers and should be incorporated as combined biomarkers in future clinical trials.  
+
+Paradoxically, this hyperactivation of the immune microenvironment can be detrimental, because tumour flare, which indicates both specific (pre- existing) and non- specific (new) expansion of TCR clones, can be misinterpreted as tumour progression and lead to premature treatment abandonment. Flare is quite common in RRD cancers, especially when tumour burden is high.<sup>12</sup> These flare responses are very different from hypermutant adult glioblastomas originating as a result previous chemo- radiotherapy<sup>65</sup> where hyper- progression is reported.<sup>37,77</sup> Our data support the development of novel functional imaging techniques<sup>78</sup> and minimally invasive biomarkers<sup>79,80</sup> to better predict tumour response, and innovative strategies to modulate the immune response. This is important, as continuation of immunotherapy can result in late responses and long- term survival for patients developing tumour flare.  
+
+Despite the limitations of a registry study, this is the first description of a large cohort of children and young adults with previously recurrent/progressive fatal germline DNA replication repair deficient cancers demonstrating impressive responses to PD- 1 blockade.<sup>81</sup> Although longer follow- up is required to determine whether immunotherapy can be a curative strategy for RRD cancers, the sustained responses and lack of late relapses in a significant number of patients in this cohort are encouraging. This study also sheds light on the complex interplay between the tumour genomic status, microenvironment, and the systemic immune response, especially in the context of extreme mutation and MS- indel burdens. Future trials should prospectively analyse the roles of germline versus somatic RRD, and the components of the RRD machinery, to identify patients who are likely to derive maximal benefit from anti- PD- 1 immunotherapy. Lastly, our study highlights the impact of studying a genetic cancer syndrome to understand general cancer processes<sup>82</sup> and deriving direct therapeutic implications for patients.
+
+<--- Page Split --->
+
+Acknowledgements: This research is supported by a Stand Up to Cancer- Bristol- Meyers Squibb Catalyst Research Grant (Grant Number: SU2C- AACR- CT07- 17), which is administered by the American Association for Cancer Research, the scientific partner of SU2C. This work is also supported by an Enabling Studies Program grant from BioCanRx - Canada's Immunotherapy Network (a Network Centre of Excellence), the Canadian Institutes for Health Research (CIHR) grant (PJT- 156006), and the CIHR Joint Canada- Israel Health Research Program. Additional financial support was provided by Meagan's Walk Foundation, LivWise Foundation, the Zane Cohen Center, BRAINchild Foundation, St. Baldrick's Foundation International Scholar Award (with generous support from the Team Campbell Foundation), and Guglietti We Love You Connie Foundation.  
+
+## Author contributions:  
+
+DM, UT and EB planned the study. AD and UT wrote the manuscript. AC and MS reported the radiology. NA and CH reported the pathology. SS, AS and TP were responsible for the bioinformatics analysis. JC, YM and GG were responsible for the microsatellite indel analysis. SS and PO were responsible for the flow cytometry studies. VB and ME were involved in patient enrolment and sample coordination. Those involved in patient management included EB, DM, VC, LFN, LS, SC, MO, RD, GAC, SC, VL, AR, MO, GM, SL, AB, VM, EO, RLDM, AD, MS, CF, DS, DS, KC, SC, MM, PT, DZ, BG, AVD, NH, DG, RM, KEN, MZ, OM, DM, TL, AL, DM, TS, LYY, JK AB, LH, SLF, SZ, IS, DS, PD, MT, AK, DT, DM, AV, MA, CD, MA and UT. All authors have reviewed and agreed to the contents of the manuscript.  
+
+Competing Interests: None
+
+<--- Page Split --->
+
+## ONLINE METHODS  
+
+## Study design and patients  
+
+Patients included in this retrospective registry study were identified through the International Replication Repair Deficiency Consortium (IRRDC), based at the Hospital for Sick Children (SickKids), Toronto, that has enrolled 200 patients from 45 countries since 2007. Patients with confirmed or suspected RRD were eligible to participate. Institutional ethics board approval was obtained. Consent was obtained from patients and families for participation in the study. This included submission of clinical and imaging data, as well as tissue and blood samples for centralized analysis. Germline diagnosis of CMMRD, Lynch or PPD were confirmed by the genetic counsellor from the consortium. This was done based on the results of the family history, next- generation panel sequencing of germline samples for MMR and POL genes (performed locally or centrally at CLIA- approved laboratories), and immunohistochemical (IHC) staining pattern of the tumour and normal tissues (performed at the Hospital for Sick children, Toronto). Thirty- eight patients with 45 cancers who had received treatment with anti- PD1- directed immune checkpoint inhibitor (ICI) therapy were identified and are reported here. Guidelines for ICI treatment were provided by the consortium. Regular communication was organized with the treating team to discuss clinical responses, along with ongoing radiological monitoring, as well as management of adverse effects of ICI. Monitoring for toxicity and stopping rules were as per the recommendations provided by an ongoing clinical trial (NCT02992964). However, patients were not treated prospectively as part of a clinical trial. Ultimately, the choice of the agent and the nuances of therapy remained at the discretion of the treating team. Blood samples were collected prospectively throughout the course of ICI therapy at specified time- points.  
+
+Clinical records were reviewed to obtain details regarding patient demographics, cancer diagnosis, date of initiation and completion of ICI, choice of ICI agent, and survival outcomes (including date of disease progression and/or patient death). Imaging at baseline and following initiation of ICI were obtained for central review of objective tumour response. For the remaining cases, response (or otherwise) was determined by the assessment of the local treating team. Available scans were centrally reviewed by a radiologist blind to the clinically determined response, and tumour measurements were documented according to the RANO criteria.40 Briefly, the best tumour response was determined as the percentage change in the product of bi- perpendicular dimensions from baseline on the contrast- enhanced T1 images. Complete response (CR) was defined as the disappearance of all enhancing disease (measurable and non- measurable) sustained for at least 4 weeks, with stable or improved non
+
+<--- Page Split --->
+
+enhancing FLAIR/T2W lesions, no new lesions and clinical stability. Partial response (PR) was defined as \(50\%\) or more decrease of all measurable enhancing lesions sustained for at least 4 weeks, with no progression of non- measurable disease, stable or improved non- enhancing FLAIR/T2W lesions, no new lesions and clinical stability. Stable disease (SD) was defined as images that did not qualify for complete response, partial response, or progression, with stable non- enhancing FLAIR/T2W lesions and clinical stability. Progressive disease (PD) was defined as \(25\%\) or more increase in enhancing lesions, with increase (significant) in non- enhancing FLAIR/T2W lesions, not attributable to other non- tumour causes any new lesions, and clinical deterioration not attributable to other causes. For patients with non- CNS solid tumours, the revised RECIST (version 1.1) was used. \(^{41}\) CR was defined as complete disappearance, PR as at least \(30\%\) decrease in sum of the diameters, PD as at least \(20\%\) increase in the sum of the diameters of the target lesion, and SD as lack of sufficient change to be classified as CR/PR/PD. Patients with objective radiological response (CR/PR) and/ or stable disease (SD) were labelled as 'responders.' Among those with progressive disease, patients experiencing rapid early clinical and/or radiological deterioration (with \(>100\%\) increase in tumour size within 90 days of starting ICI therapy) were defined as demonstrating a 'flare' response and were studied in more detail. For patients able to continue ICI, subsequent imaging was reviewed to confirm response or progression. \(^{43}\) Those with sustained clinical and/or radiological progression despite continuation of ICI treatment were classified as 'non- responders.' For biomarker prediction analyses, 'responders' also included those with an initial 'flare,' who continued on ICI and demonstrated delayed responses.  
+
+## Whole exome sequencing (WES) and analysis  
+
+Genomic DNA from 39 tumours, along with matched germline blood samples, was extracted using the PaxGene Blood DNA Extraction Kit (Cat No./ID: 761133) for blood samples, Qiagen DNeasy Blood & Tissue Kits (Cat No./ID: 69504) for frozen tissue, MasterPure Complete DNA and RNA Purification Kit (Epicentre #MC85200) for paraffin embedded tissue). WES was performed at The Centre for Applied Genomics (TCAG), SickKids, using SureSelect Agilent All Exon v5 kit, followed by sequencing (150X) on Illumina HiSeq 2500. The software bcl2fastq2 v2.17 was used to generate raw fastq files. Alignment to the hg38 reference genome, followed by pre- processing and QC was adapted from the GATK standard pipeline, using BWA- MEM 0.7.12 (alignment), BAMQC, Picard 2.6.0 (QC). Somatic variant calling was done post- alignment, using processed bam files from tumour and matched normal samples, to call both single nucleotide variants (SNVs) and
+
+<--- Page Split --->
+
+insertion deletion (indel) variants. A consensus vcf file of shared variants across 2 or more of 4 variant callers (Mutect v1.1.5, GATK v3.6/Mutect2, Strelka v1.0.14, and Varscan2 Somatic v2.4.2) was generated for SNVs and indels separately, using VCFtools 0.1.15, and these vcfs were annotated using VEP v83. The tumour mutation burden (SNVs per megabase) from WES data was calculated by counting total number of somatic SNVs divided by total number of callable bases in megabases ( \(\sim 50\mathrm{Mb}\) ). DeconstructSigs \(^{83}\) was used to determine COSMIC signatures \(^{46}\) in the mutation spectrum within a tri-nucleotide context for each sample. All bioinformatics analyses were performed on the SickKids High Performance Cluster (HPF) and the UHN High Performance Cluster for Health (H4H).  
+
+## Neoantigen calling  
+
+The Mutect2 vcfs generated for each tumour (described above) were used as input along with bioinformatically generated HLA- typing (using a consensus of HLAminer, and HLAVBSeq) of germline WES fastqs using MuPeXI, \(^{84}\) to get a list of strong binding candidate neoantigens per HLA- type.  
+
+## Microsatellite (MS)-indel calling  
+
+Microsatellite (MS)- indel callingMicrosatellite indels were called on the bam files of tumour and matched normal samples, using an in- house pipeline using MSMuTect. The detailed methods for this algorithm have been previously reported. \(^{50}\) Briefly, repeats of five or more nucleotides were considered to be MS loci, and using the PHOBOS algorithm and the lobSTR approach, tumour and normal BAM files were aligned with their 5' and 3' flanking sequences. Each MS- locus allele was estimated using the empirical noise model, which is the probability of observing a read with a microsatellite (MS) length k and motif m, where the true length of the allele is j with the motif m. This was used to call the MS alleles with the highest likelihood of being the true allele at each MS- locus. The MS alleles of each tumour and matched normal pair were called individually, which were compared to identify the mutations on the tumour MS- loci. The Akaike Information Criterion (AIC) score was assigned to both the tumour and normal models, and a threshold score that was determined by using simulated data was applied to make the final MS- indel call.  
+
+## Immunohistochemistry  
+
+Four- micron thick sections of formalin- fixed paraffin- embedded (FFPE) surgical specimens were stained using an automated stainer (Dako OMNIS) with the following primary
+
+<--- Page Split --->
+
+antibodies: PD- L1 (clone28- 8, Abcam, 1/500), CD68 (Clone c8/144B, Dako OMNIS), CD8 (Clone c8/144B, Dako OMNIS), CD3 (polyclonal rabbit, Dako OMNIS), and CD4 (Clone SP35, Sigma- Aldrich, 1/50). Quantitative evaluation of the immunohistochemical stains was performed by examining each section using at least five to seven different high- power fields with the most abundant tumour- infiltrating lymphocyte areas. The tumour was considered PD- L1 positive if \(\geq 1\%\) of tumour cells exhibited a circumferential and/or partial linear plasma membrane PD- L1 staining of tumour cells at any intensity. \(^{85}\) The percentage of infiltrating immune system cells was estimated by manual eyeballing as none, mild, moderate, and severe (0=none, \(< 10\% =\) mild, 10- \(50\% =\) moderate, \(>50\% =\) severe). For downstream analyses, infiltration higher than the median values of the continuous data of immune infiltrates was used to classify tumours as 'high' or 'low' infiltration for each marker.  
+
+## Immune inference analysis  
+
+RNA was extracted as per standard kit protocol from tumours biopsied at both, baseline and time of flare, in 2 patients (P31 and P33) and submitted for total r- RNA depletion RNAsequencing on HiSeq 2500, at TCAG. Following sequencing, the raw data were analyzed using STAR aligner followed by RSEM expression analysis to generate a gene expression matrix for each sample. This was then run through CIBERSORT, \(^{86}\) a single state deconvolution algorithm, to generate immune inference data for 22 immune cell subtypes. The immune inference results from CIBERSORT were plotted using "ComplexHeatmap" \(^{87}\) package on R 3.5.  
+
+## T-cell receptor rearrangement repertoire profiling  
+
+Genomic DNA was extracted (methods as above) from tumours biopsied at both baseline and at time of flare in 2 patients (P31 and P33). These were transferred to the Pugh laboratory at the University Health Network (UHN) in Toronto, where library preparation and capTCRseq \(^{88}\) hybrid capture were performed. Following library preparation, the samples were sequenced first on a MiSeq for QC purposes and then 300ng of each sample, pooled in a ratio of 1:1:1, was processed for a 3- step capture using target hybrid capture panel. \(^{88}\) Post- capture QC was performed on a MiSeq, followed by sequencing of up to a depth of \(\sim 2\) millions reads on the NextSeq. Post- sequencing the raw data were analyzed using MiXCR version 2.1.12, 'iNext', 'immunarch' R packages and Pugh Lab customized functions on R version 3.5 to look at T- cell receptor rearrangements in the form of unique clonotypes (VDJ rearranged sequences) for T- cell receptors alpha, beta, gamma and delta. As the total read depth varied across the
+
+<--- Page Split --->
+
+cohort, affecting the total successfully aligned reads, all raw fastq reads were down- sampled to \(\sim 1\) million reads. QC parameters of percent aligned reads, reads used in clonotypes, final clonotype count and the total number of clonotypes per 1000 reads were considered.  
+
+## Flow cytometry  
+
+Viable frozen peripheral blood mononuclear cells were incubated with Fc block (BD Biosciences) prior to staining for surface markers (anti- CD3 - clone UCHT1, anti- CD4 - clone RPA- T4, anti- CD8 - clone RPA- T8, anti- 4- 1BB- clone 4B4- 1, anti- TIGIT - clone MBSA43, anti- Ki67 - clone 20Raj1) and viability dye (eBioscience). Cells were fixed and permeabilized for intercellular staining with the Foxp3 transcription factor staining buffer set (BD). Flow cytometry voltages were set using Rainbow beads (Spherotech) with the same setting between experiments. Samples were acquired on a BD LSR Fortessa flow cytometer and data were analyzed using the FlowJo software.  
+
+## Statistical analyses and reproducibility  
+
+Overall survival (OS) and event- free survival (EFS) was estimated using Kaplan- Meier statistics and determined from the date of initiation of ICI therapy. Patients without an event were censored at the date of last known contact. Uniquely in this population, several patients had multiple separate synchronous malignancies and therefore in these analyses, survival is presented for each individual cancer/tumour in addition to analyses per patient. For example, for the tumour- wise analysis, a patient experiencing an event related to one cancer diagnosis, would be shown as censored (rather than an event) for a second synchronous cancer diagnosis. Multivariate analysis was performed using binary logistic regression analysis for independent variables for predicting response. Cox regression analysis was performed for predictors of overall survival. Variables used in multivariate analysis included age, gender, site, ICI agent, higher mutations (SNVs/Mb) and higher MS- indels (than the median). Statistical analyses were performed with SPSS version 20, R version 3.5 and Python version 2.7. Statistical significance was calculated using Welch's unequal variances t- test and the Wilcoxon- Mann- Whitney test, for parametric and nonparametric data, respectively. All p- values were 2- sided, with a cut- off of 0.05 for significance. Plots generated were edited for aesthetics using Adobe Illustrator version 23.0.1.
+
+<--- Page Split --->
+
+633 1. Cortez, D. Replication-Coupled DNA Repair. Molecular cell 74, 866-876 (2019). 634 2. Campbell, B.B., et al. Comprehensive Analysis of Hypermutation in Human Cancer. Cell 171, 1042-1056. e1010 (2017). 635 3. Chung, J., et al. DNA polymerase and mismatch repair exert distinct microsatellite instability signatures in normal and malignant human cells. Cancer discovery (2020). 636 4. Mur, P., et al. Role of POLE and POLD1 in familial cancer. Genetics in medicine : official journal of the American College of Medical Genetics (2020). 637 5. Seppälä, T.T., et al. European guidelines from the EHTG and ESCP for Lynch syndrome: an updated third edition of the Mallorca guidelines based on gene and gender. The British journal of surgery (2020). 638 6. Lynch, H.T., Snyder, C.L., Shaw, T.G., Heinen, C.D. & Hitchins, M.P. Milestones of Lynch syndrome: 1895-2015. Nature reviews. Cancer 15, 181-194 (2015). 639 7. Tabori, U., et al. Clinical Management and Tumour Surveillance Recommendations of Inherited Mismatch Repair Deficiency in Childhood. Clinical cancer research : an official journal of the American Association for Cancer Research 23, e32-e37 (2017). 640 8. Ripperger, T. & Schlegelberger, B. Acute lymphoblastic leukemia and lymphoma in the context of constitutional mismatch repair deficiency syndrome. European journal of medical genetics 59, 133-142 (2016). 641 9. Amayiri, N., et al. High frequency of mismatch repair deficiency among pediatric high grade gliomas in Jordan. International journal of cancer 138, 380-385 (2016). 642 10. Shlien, A., et al. Combined hereditary and somatic mutations of replication error repair genes result in rapid onset of ultra-hypermutated cancers. Nat Genet 47, 257-262 (2015). 643 11. Finocchiaro, G., Langella, T., Corbetta, C. & Pellegatta, S. Hypermutations in gliomas: a potential immunotherapy target. Discovery medicine 23, 113-120 (2017). 644 12. Bouffet, E., et al. Immune Checkpoint Inhibition for Hypermutant Glioblastoma Multiforme Resulting From Germline Biallelic Mismatch Repair Deficiency. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 34, 2206-2211 (2016). 645 13. Wei, S.C., Duffy, C.R. & Allison, J.P. Fundamental Mechanisms of Immune Checkpoint Blockade Therapy. Cancer discovery 8, 1069-1086 (2018). 646 14. Liu, D., et al. Integrative molecular and clinical modeling of clinical outcomes to PD1 blockade in patients with metastatic melanoma. Nature medicine 25, 1916-1927 (2019). 647 15. Kammerer-Jacquet, S.F., et al. Targeting the PD-1/PD-L1 Pathway in Renal Cell Carcinoma. International journal of molecular sciences 20(2019). 648 16. Garon, E.B., et al. Pembrolizumab for the treatment of non-small-cell lung cancer. N Engl J Med 372, 2018-2028 (2015). 649 17. Topalian, S.L., et al. Five-Year Survival and Correlates Among Patients With Advanced Melanoma, Renal Cell Carcinoma, or Non-Small Cell Lung Cancer Treated With Nivolumab. JAMA oncology (2019). 650 18. Hellmann, M.D., et al. Nivolumab plus Ipilimumab in Advanced Non-Small-Cell Lung Cancer. The New England journal of medicine 381, 2020-2031 (2019). 651 19. Gurjao, C., Tsukrov, D., Imakaev, M., Luquette, L.J. & Mirny, L.A. Limited evidence of tumour mutational burden as a biomarker of response to immunotherapy. bioRxiv, 2020.2009.2003.260265 (2020). 652 20. Marabelle, A., et al. Association of tumour mutational burden with outcomes in patients with advanced solid tumours treated with pembrolizumab: prospective biomarker analysis of the multicohort, open-label, phase 2 KEYNOTE-158 study. The Lancet. Oncology 21, 1353-1365 (2020).
+
+<--- Page Split --->
+
+682 21. Bersanelli, M. Tumour mutational burden as a driver for treatment choice in resistant tumours (and beyond). The Lancet. Oncology 21, 1255- 1257 (2020).683 22. Yarchoan, M., Hopkins, A. & Jaffee, E.M. Tumour Mutational Burden and Response 685 Rate to PD- 1 Inhibition. The New England journal of medicine 377, 2500- 2501 (2017).686 23. Cristescu, R., et al. Pan- tumour genomic biomarkers for PD- 1 checkpoint blockade- 687 based immunotherapy. Science (New York, N.Y.) 362(2018).688 24. Geoerger, B., et al. Pembrolizumab in paediatric patients with advanced melanoma or 689 a PD- L1- positive, advanced, relapsed, or refractory solid tumour or lymphoma 690 (KEYNOTE- 051): interim analysis of an open- label, single- arm, phase 1- 2 trial. The 691 Lancet. Oncology 21, 121- 133 (2020).692 25. Le, D.T., et al. PD- 1 Blockade in Tumours with Mismatch- Repair Deficiency. The New 693 England journal of medicine 372, 2509- 2520 (2015).694 26. Le, D.T., et al. Phase II Open- Label Study of Pembrolizumab in Treatment- Refractory, 695 Microsatellite Instability- High/Mismatch Repair- Deficient Metastatic Colorectal 696 Cancer: KEYNOTE- 164. J Clin Oncol 38, 11- 19 (2020).697 27. André, T., et al. Pembrolizumab in Microsatellite- Instability- High Advanced 698 Colorectal Cancer. The New England journal of medicine 383, 2207- 2218 (2020).699 28. Llosa, N.J., et al. The vigorous immune microenvironment of microsatellite instable 700 colon cancer is balanced by multiple counter- inhibitory checkpoints. Cancer discovery 701 5, 43- 51 (2015).702 29. Schrock, A.B., et al. Tumour mutational burden is predictive of response to immune 703 checkpoint inhibitors in MSI- high metastatic colorectal cancer. Ann Oncol 30, 1096- 1103 (2019).704 30. Sampson, J.H., Gunn, M.D., Fecci, P.E. & Ashley, D.M. Brain immunology and 705 immunotherapy in brain tumours. Nature reviews. Cancer 20, 12- 25 (2020).706 31. Pearson, A.D.J., et al. ACCELERATE and European Medicines Agency Paediatric 707 Strategy Forum for medicinal product development of checkpoint inhibitors for use in 708 combination therapy in paediatric patients. European journal of cancer (Oxford, 709 England : 1990) 127, 52- 66 (2020).710 32. Geoerger, B., et al. Atezolizumab for children and young adults with previously treated 711 solid tumours, non- Hodgkin lymphoma, and Hodgkin lymphoma (iMATRIX): a 712 multicentre phase 1- 2 study. The Lancet. Oncology 21, 134- 144 (2020).713 33. Merchant, M.S., et al. Phase I Clinical Trial of Ipilimumab in Pediatric Patients with 714 Advanced Solid Tumours. Clinical cancer research : an official journal of the 715 American Association for Cancer Research 22, 1364- 1370 (2016).716 34. Davis, K.L., et al. Nivolumab in children and young adults with relapsed or refractory 717 solid tumours or lymphoma (ADVL1412): a multicentre, open- label, single- arm, phase 718 1- 2 trial. The Lancet. Oncology 21, 541- 550 (2020).719 35. Martin- Romano, P., et al. Evidence of pseudoprogression in patients treated with 720 PD1/PDL1 antibodies across tumour types. Cancer medicine 9, 2643- 2652 (2020).721 36. Okada, H., et al. Immunotherapy response assessment in neuro- oncology: a report of 722 the RANO working group. The Lancet. Oncology 16, e534- e542 (2015).723 37. Sasaki, A., et al. Predictive factors for hyperprogressive disease during nivolumab as 724 anti- PD1 treatment in patients with advanced gastric cancer. Gastric cancer : official 725 journal of the International Gastric Cancer Association and the Japanese Gastric 726 Cancer Association 22, 793- 802 (2019).727 38. Ellingson, B.M., Chung, C., Pope, W.B., Boxerman, J.L. & Kaufmann, T.J. 728 Pseudoprogression, radionecrosis, inflammation or true tumour progression? 729 challenges associated with glioblastoma response assessment in an evolving therapeutic 730 landscape. Journal of neuro- oncology 134, 495- 504 (2017).
+
+<--- Page Split --->
+
+39. Bakry, D., et al. Genetic and clinical determinants of constitutional mismatch repair deficiency syndrome: report from the constitutional mismatch repair deficiency consortium. European journal of cancer (Oxford, England : 1990) 50, 987-996 (2014).
+40. Wen, P.Y., et al. Updated response assessment criteria for high-grade gliomas: response assessment in neuro-oncology working group. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 28, 1963-1972 (2010).
+41. Eisenhauer, E.A., et al. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). European journal of cancer (Oxford, England : 1990) 45, 228-247 (2009).
+42. Yoshida, T., Furuta, H. & Hida, T. Risk of tumour flare after nivolumab treatment in patients with irradiated field recurrence. Medical oncology (Northwood, London, England) 34, 34 (2017).
+43. Wen, P.Y., et al. Response Assessment in Neuro-Oncology Clinical Trials. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 35, 2439-2449 (2017).
+44. Guerrini-Rousseau, L., et al. Constitutional mismatch repair deficiency-associated brain tumours: report from the European C4CMMRD consortium. Neuro-oncology advances 1, vdz033 (2019).
+45. Alexandrov, L.B., et al. The repertoire of mutational signatures in human cancer. Nature 578, 94-101 (2020).
+46. Alexandrov, L.B., et al. Signatures of mutational processes in human cancer. Nature 500, 415-421 (2013).
+47. Chen, J.A., Ma, W., Yuan, J. & Li, T. Translational Biomarkers and Rationale Strategies to Overcome Resistance to Immune Checkpoint Inhibitors in Solid Tumours. Cancer treatment and research 180, 251-279 (2020).
+48. Le, D.T., et al. Mismatch repair deficiency predicts response of solid tumours to PD-1 blockade. Science (New York, N.Y.) 357, 409-413 (2017).
+49. Mandal, R., et al. Genetic diversity of tumours with mismatch repair deficiency influences anti-PD-1 immunotherapy response. Science (New York, N.Y.) 364, 485-491 (2019).
+50. Maruvka, Y.E., et al. Analysis of somatic microsatellite indels identifies driver events in human tumours. Nature biotechnology 35, 951-959 (2017).
+51. Gang, W., et al. Strategy to targeting the immune resistance and novel therapy in colorectal cancer. Cancer medicine 7, 1578-1603 (2018).
+52. Carson, M.J., Doose, J.M., Melchior, B., Schmid, C.D. & Ploix, C.C. CNS immune privilege: hiding in plain sight. Immunol Rev 213, 48-65 (2006).
+53. Huang, A.C., et al. T-cell invigoration to tumour burden ratio associated with anti-PD-1 response. Nature 545, 60-65 (2017).
+54. Watts, T.H. TNF/TNFR family members in costimulation of T cell responses. Annual review of immunology 23, 23-68 (2005).
+55. Chester, C., Sanmamed, M.F., Wang, J. & Melero, I. Immunotherapy targeting 4-1BB: mechanistic rationale, clinical results, and future strategies. Blood 131, 49-57 (2018).
+56. Clouthier, D.L., et al. An interim report on the investigator-initiated phase 2 study of pembrolizumab immunological response evaluation (INSPIRE). J Immunother Cancer 7, 72 (2019).
+57. Blumenthal, D.T., et al. Pembrolizumab: first experience with recurrent primary central nervous system (CNS) tumours. Journal of neuro-oncology 129, 453-460 (2016).
+58. Routy, B., et al. Gut microbiome influences efficacy of PD-1-based immunotherapy against epithelial tumours. Science (New York, N.Y.) 359, 91-97 (2018).
+
+<--- Page Split --->
+
+781 59. Haworth, K.B., et al. Going back to class I: MHC and immunotherapies for childhood cancer. Pediatric blood & cancer 62, 571- 576 (2015). 782 60. Vakkila, J., Jaffe, R., Michelow, M. & Lotze, M.T. Pediatric cancers are infiltrated predominantly by macrophages and contain a paucity of dendritic cells: a major nosologic difference with adult tumours. Clinical cancer research : an official journal of the American Association for Cancer Research 12, 2049- 2054 (2006). 786 61. Majzner, R.G., et al. Assessment of programmed death- ligand 1 expression and tumour- associated immune cells in pediatric cancer tissues. Cancer 123, 3807- 3815 (2017). 789 62. Amayiri, N., et al. High frequency of mismatch repair deficiency among pediatric high grade gliomas in Jordan. Int J Cancer (2015). 792 63. Khasraw, M., Reardon, D.A., Weller, M. & Sampson, J.H. PD- 1 inhibitors: Do they have a future in the treatment of glioblastoma? Clinical cancer research : an official journal of the American Association for Cancer Research (2020). 794 64. Khasraw, M., Walsh, K.M., Heimberger, A.B. & Ashley, D.M. What is the Burden of Proof for Tumour Mutational Burden in gliomas? Neuro- Oncology (2020). 797 65. Touat, M., et al. Mechanisms and therapeutic implications of hypermutation in gliomas. Nature 580, 517- 523 (2020). 799 66. Azad, N.S., et al. Nivolumab Is Effective in Mismatch Repair- Deficient Noncolorectal Cancers: Results From Arm Z1D- A Subprotocol of the NCI- MATCH (EAY131) Study. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 38, 214- 222 (2020). 802 67. Marabelle, A., et al. Efficacy of Pembrolizumab in Patients With Noncolorectal High Microsatellite Instability/Mismatch Repair- Deficient Cancer: Results From the Phase II KEYNOTE- 158 Study. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 38, 1- 10 (2020). 806 68. Zhao, J., et al. Immune and genomic correlates of response to anti- PD- 1 immunotherapy in glioblastoma. Nature medicine 25, 462- 469 (2019). 809 69. Thomas, R., Wang, W. & Su, D.M. Contributions of Age- Related Thymic Involution to Immunosenescence and Inflammaging. Immun Ageing 17, 2 (2020). 811 70. Simon, A.K., Hollander, G.A. & McMichael, A. Evolution of the immune system in humans from infancy to old age. Proc Biol Sci 282, 20143085 (2015). 813 71. Cloughesy, T.F., et al. Neoadjuvant anti- PD- 1 immunotherapy promotes a survival benefit with intratumoural and systemic immune responses in recurrent glioblastoma. Nature medicine 25, 477- 486 (2019). 816 72. Schalper, K.A., et al. Neoadjuvant nivolumab modifies the tumour immune microenvironment in resectable glioblastoma. Nature medicine 25, 470- 476 (2019). 817 73. Blank, C.U., Haanen, J.B., Ribas, A. & Schumacher, T.N. CANCER IMMUNOLOGY. The "cancer immunogram". Science (New York, N.Y.) 352, 658- 660 (2016). 820 74. Bruni, D., Angell, H.K. & Galon, J. The immune contexture and Immunoscore in cancer prognosis and therapeutic efficacy. Nature reviews. Cancer 20, 662- 680 (2020). 821 75. Bai, R., Lv, Z., Xu, D. & Cui, J. Predictive biomarkers for cancer immunotherapy with immune checkpoint inhibitors. Biomark Res 8, 34 (2020). 822 76. Turajlic, S., et al. Insertion- and- deletion- derived tumour- specific neoantigens and the immunogenic phenotype: a pan- cancer analysis. The Lancet. Oncology 18, 1009- 1021 (2017). 823 77. Ellingson, B.M., et al. Estimated clinical efficacy and radiographic response characteristics of PD1 inhibition in newly diagnosed and recurrent glioblastoma in clinical practice: A report from the iRANO Working Group. Journal of Clinical Oncology 38, 2521- 2521 (2020).
+
+<--- Page Split --->
+
+831 78. Antonios, J.P., et al. Detection of immune responses after immunotherapy in glioblastoma using PET and MRI. Proceedings of the National Academy of Sciences of the United States of America 114, 10220- 10225 (2017). 833 79. Kim, K.H., et al. The First- week Proliferative Response of Peripheral Blood PD- 1(+CD8(+) T Cells Predicts the Response to Anti- PD- 1 Therapy in Solid Tumours. Clinical cancer research : an official journal of the American Association for Cancer Research 25, 2144- 2154 (2019). 837 80. Georgiadis, A., et al. Noninvasive Detection of Microsatellite Instability and High Tumour Mutation Burden in Cancer Patients Treated with PD- 1 Blockade. Clinical cancer research : an official journal of the American Association for Cancer Research 25, 7024- 7034 (2019). 841 81. Pasello, G., et al. Real world data in the era of Immune Checkpoint Inhibitors (ICIs): Increasing evidence and future applications in lung cancer. Cancer treatment reviews 87, 102031 (2020). 845 82. Tomasetti, C., Li, L. & Vogelstein, B. Stem cell divisions, somatic mutations, cancer etiology, and cancer prevention. Science (New York, N.Y.) 355, 1330- 1334 (2017). 847 83. Rosenthal, R., McGranahan, N., Herrero, J., Taylor, B.S. & Swanton, C. DeconstructSigs: delineating mutational processes in single tumours distinguishes DNA repair deficiencies and patterns of carcinoma evolution. Genome biology 17, 31 (2016). 851 84. Bjerregaard, A.M., Nielsen, M., Hadrup, S.R., Szallasi, Z. & Eklund, A.C. MuPeXI: prediction of neo- epitopes from tumour sequencing data. Cancer immunology, immunotherapy : CII 66, 1123- 1130 (2017). 853 85. Ionescu, D.N., Downes, M.R., Christofides, A. & Tsao, M.S. Harmonization of PD- L1 testing in oncology: a Canadian pathology perspective. Current oncology (Toronto, Ont.) 25, e209- e216 (2018). 857 86. Newman, A.M., et al. Robust enumeration of cell subsets from tissue expression profiles. Nature methods 12, 453- 457 (2015). 859 87. Gu, Z., Eils, R. & Schlesner, M. Complex heatmaps reveal patterns and correlations in multidimensional genomic data. Bioinformatics (Oxford, England) 32, 2847- 2849 (2016). 862 88. Mulder, D.T., et al. CapTCR- seq: hybrid capture for T- cell receptor repertoire profiling. Blood advances 2, 3506- 3514 (2018).
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Fig. 1. Clinical response to ICI across cancer types in patients with germline DNA replication repair deficiency. (a) Distribution of tumour types across 38 patients who developed 45 tumours. (b) Waterfall plot of all radiological responses in non-haematological malignancies. Values show the best fractional change in the 2 dimensions from baseline measurements as per RANO and RECIST criteria. Arrows point to representative T2-weighted FLAIR and T1-weighted contrast-enhanced MRI sequences in two patients showing flare and partial responses. </center>
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Fig.2. Patient outcome and survival by tumour type. (a) Swimmer plot by patient and tumour type. (b) Kaplan-Meier (KM) estimates of overall survival for all patients (c) KM estimates of overall survival as per tumour type. Median survival for CNS tumours was 21.6 months. Median survival was not reached for non-CNS solid tumours. (d) KM estimates of progression free and overall survival for CNS tumours continuing ICI therapy. Note: prolonged median survival at 24 months (estimated 3-year OS=49.1%) despite initial radiological progression at a median of 9.9 months (estimated 3-year PFS=32%). </center>
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Fig. 3. Genomic biomarkers, survival and response to PD1 blockade. (a) Onco-plot summarising the genomic features from 39 available paired tumour and germline exomes, and their clinical correlates. (b) Response and overall survival (OS) by single nucleotide variants (SNVs) per Mb. For survival analysis, median SNV burden was used. (c) SNVs as a function of MMRD (blue) and MMRD+PPD (orange) status (left), and response association with both SNV and RRD status. (c) Response and overall survival (OS) by microsatellite indels (MS-indels). For survival analysis, median MS-indel values were used. (d) Response and overall survival by total MS-indel count for MMRD and MMRD+PPD cancers separately. (e) Kaplan-Meier (KM) estimates using combined SNVs/Mb and MS-indel in all RRD cancers. (Abbreviations: MMRD: mismatch-repair deficiency; PPD: Polymerase-proofreading deficiency; MS-indel: microsatellite insertion/deletion.) </center>
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Fig. 4. Genomic biomarkers, survival and response to PD1 blockade. (a) Onco-plot summarising the genomic features from 39 available paired tumour and germline exomes, and their clinical correlates. (b) Response and overall survival (OS) by single nucleotide variants (SNVs) per Mb. For survival analysis, median SNV burden was used. (c) SNVs as a function of MMRD (blue) and MMRD+PPD (orange) status (left), and response association with both SNV and RRD status. (c) Response and overall survival (OS) by microsatellite indels (MS-indels). For survival analysis, median MS-indel values were used. (d) Response and overall survival by total MS-indel count for MMRD and MMRD+PPD cancers separately. (e) Kaplan-Meier (KM) estimates using combined SNVs/Mb and MS-indel in all RRD cancers. (Abbreviations: MMRD: mismatch-repair deficiency; PPD: Polymerase-proofreading deficiency; MS-indel: microsatellite insertion/deletion.) </center>
+
+<--- Page Split --->
+![](images/Figure_5.jpg)
+
+<center>Fig. 5. Tumour immune microenvironment, survival and response to PD-1 blockade. (a) </center>  
+
+PD- L1 expression, response and survival in all RRD cancers. Cut- off is \(\geq 1\%\) of cells (Methods). (b) CD8 expression, response and survival for RRD cancers. Cut- off is \(\geq 3\%\) of cells (higher than the median; Methods). For both (a) and (b), the histology depicts glioblastoma at 20X magnification. (c) Association of immune markers with SNV and RRD status. (d) Combined immune (PD- L1 and CD8 expression) and genomic (TMB and MS- indels) and overall survival in RRD cancers.
+
+<--- Page Split --->
+![](images/Figure_6.jpg)
+
+<center>Fig.6. Characterization of the tumour flare response. (a-d) Analysis of 2 patients who had tumour-debulking prior to therapy and at the time of flare. (a, b) Total immune cell content in pre-therapy and at flare. (c, d) The corresponding CapTCR-sequencing and T-cell receptor clonotype analysis in these samples. (e, f) Immunohistochemistry for PD-L1 expression, and CD8-T-cell infiltration in the pre-therapy sample and at flare, as shown in the representative 20X images from the tumour sample in patient-1 (P33). (g) Representative flow cytometry plot showing activation of \(\mathrm{CD + }\) T-cell (TIGIT and 4-1BB) from the blood sample of a patient before treatment initiation and at flare. (h) 41BB+ CD8+ T-cells in blood from responders without flare, non-responders and flare. </center>
+
+<--- Page Split --->
+![](images/Extended_Data_Figure_1.jpg)
+
+<center>Extended Data Fig.1. KM estimates of progression-free and overall survival for (a) all tumours, (b) CNS tumours, (c) all solid tumours (CNS and non-CNS). (d) Progression free survival for non-CNS solid tumours versus CNS tumours \((p = 0.01)\) . </center>
+
+<--- Page Split --->
+![](images/Extended_Data_Figure_6.jpg)
+  
+
+Extended Data Fig.2: Genomic biomarkers, survival and response to PD- 1 blockade. Response and overall survival by (a) total synonymous variants, (b) non- synonymous variants, (c) total indels, (d) total mutations/Mb, and (e) neoantigens.
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+  
+
+Extended Data Fig.3: CNS tumours: Genomic biomarkers, survival and response to PD- 1 blockade. Response and overall survival for CNS tumours by (a) synonymous variants, (b) non- synonymous variants, (c) SNVs/Mb, (d) total indels, (e) total mutations/ Mb, (f) neoantigens, and (g) MS- indels.
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+  
+
+Extended Data Fig.4: Tumour immune microenvironment, survival and response to PD- 1 blockade. Response and overall survival by (a) CD3, (b) CD4, and (c) CD8- positive T- cell infiltration in the pre- ICI tumour specimens.
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+  
+
+943  
+
+Extended Data Fig.5: Total CD8 cells and response to PD- 1 blockade. (a) Responders had higher levels of CD8 T- cell infiltration as compared to non- responders in all tumours. Green- gastrointestinal, blue- CNS tumours. (b) Representative images from non- CNS solid tumours, all of which demonstrated high CD8 and (c) PD- L1 expression.
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Extended Data Fig.6: CNS tumours: Immune microenvironment, survival and response to PD-1 blockade. Response and overall survival in CNS tumours by (a) PD-L1, (b) CD8, (c) CD3, (d) CD4, and (e) CD68 expression.</center>
+
+<--- Page Split --->
+![](images/Figure_5.jpg)
+  
+
+Extended Data Fig.7: Genomic and immune markers of response in tumours exhibiting flare. Responders without flare, non- responders and flare stratified by (a) SNVs/mb, (b) MS- indels, (c) PD- L1 and (d) CD8 expression.
+
+<--- Page Split --->
+![](images/Figure_6.jpg)
+  
+
+957  
+
+958 Extended Data Fig.8: Representative images (20X) from patient P31 demonstrating (a) CD8 and (b) PD- L1 expression in pre- ICI and flare samples.
+
+<--- Page Split --->
+![PLACEHOLDER_43_0]
+  
+
+Extended Data Fig.9: Immune activation in patients experiencing tumour flare. Flow cytometry dot-plots from two additional patients (P39, P40) comparing pre-ICI and flare samples, showing (a, b) TIGIT and 4-1BB expressing CD8+ T-cells, (c, d) Ki67 expressing CD8+ T-cells, and (e, f) TIGIT and 4-1BB expressing CD4+ T-cells.
+
+<--- Page Split --->
+
+# Extended Data Table S2:  
+
+Binary logistic regression analysis for response to ICI in RRD cancers   
+
+<table><tr><td>Variables</td><td>Odds ratio</td><td>95% CI</td><td>p-value</td></tr><tr><td>Age</td><td>1.19</td><td>0.8, 1.6</td><td>0.26</td></tr><tr><td>Male vs female gender</td><td>0.14</td><td>0, 6.1</td><td>0.30</td></tr><tr><td>CNS vs non-CNS solid tumours</td><td>1.15</td><td>0.0, -</td><td>0.99</td></tr><tr><td>Nivolumab vs Pembrolizumab</td><td>0.01</td><td>0, 3.6</td><td>0.12</td></tr><tr><td>High SNV/Mb (&amp;gt;median)</td><td>0.03</td><td>0, 0.5</td><td>0.01</td></tr><tr><td>High MS-indel (&amp;gt;median)</td><td>0.08</td><td>0, 2.1</td><td>0.13</td></tr></table>  
+
+# Extended Data Table S3:  
+
+Cox regression analysis for OS following ICI in patients with RRD cancers   
+
+<table><tr><td>Variables</td><td>Odds ratio</td><td>95% CI</td><td>p-value</td></tr><tr><td>Age</td><td>0.88</td><td>0.7, 1.1</td><td>0.33</td></tr><tr><td>Male vs female gender</td><td>1.20</td><td>0.2, 6.1</td><td>0.83</td></tr><tr><td>CNS vs non-CNS solid tumours</td><td>0</td><td>0.0, -</td><td>0.96</td></tr><tr><td>Nivolumab vs Pembrolizumab</td><td>1.20</td><td>0.1, 19.8</td><td>0.89</td></tr><tr><td>High SNV/Mb (&amp;gt;median)</td><td>5.57</td><td>0.9, 32.4</td><td>0.05</td></tr><tr><td>High MS-indel (&amp;gt;median)</td><td>3.15</td><td>0.6, 16.4</td><td>0.17</td></tr></table>
+
+<--- Page Split --->
+
+## Figures  
+
+![PLACEHOLDER_45_0]
+
+<center>Figure 1 </center>  
+
+Clinical response to ICI across cancer types in patients with germline DNA replication repair deficiency. (a) Distribution of tumour types across 38 patients who developed 45 tumours. (b) Waterfall plot of all radiological responses in non- haematological malignancies. Values show the best fractional change in the 2 dimensions from baseline measurements as per RANO and RECIST criteria. Arrows point to representative T2- weighted FLAIR and T1- weighted contrast- enhanced MRI sequences in two patients showing flare and partial responses.
+
+<--- Page Split --->
+![PLACEHOLDER_46_0]
+
+<center>Figure 2 </center>  
+
+Patient outcome and survival by tumour type. (a) Swimmer plot by patient and tumour type. (b) Kaplan- Meier (KM) estimates of overall survival for all patients (c) KM estimates of overall survival as per tumour type. Median survival for CNS tumours was 21.6 months. Median survival was not reached for non- CNS solid tumours. (d) KM estimates of progression free and overall survival for CNS tumours continuing ICI therapy. Note: prolonged median survival at 24 months (estimated 3- year OS=49.1%) despite initial radiological progression at a median of 9.9 months (estimated 3- year PFS=32%).
+
+<--- Page Split --->
+![PLACEHOLDER_47_0]
+
+<center>Figure 3 </center>  
+
+Genomic biomarkers, survival and response to PD1 blockade. (a) Onco- plot summarising the genomic features from 39 available paired tumour and germline exomes, and their clinical correlates. (b) Response and overall survival (OS) by single nucleotide variants (SNVs) per Mb. For survival analysis, median SNV burden was used. (c) SNVs as a function of MMRD (blue) and MMRD+PPD (orange) status (left), and response association with both SNV and RRD status. (c) Response and overall survival (OS) by microsatellite indels (MS indels). For survival analysis, median MS- indel values were used. (d) Response and overall survival by total MS- indel count for MMRD and MMRD+PPD cancers separately. (e) Kaplan- Meier (KM) estimates using combined SNVs/Mb and MS- indel in all RRD cancers. (Abbreviations: MMRD: mismatch- repair deficiency; PPD: Polymerase- proofreading deficiency; MS- indel: microsatellite insertion/ deletion.)
+
+<--- Page Split --->
+![PLACEHOLDER_48_0]
+
+<center>Figure 4 </center>  
+
+Genomic biomarkers, survival and response to PD1 blockade. (a) Onco- plot summarising the genomic features from 39 available paired tumour and germline exomes, and their clinical correlates. (b) Response and overall survival (OS) by single nucleotide variants (SNVs) per Mb. For survival analysis, median SNV burden was used. (c) SNVs as a function of MMRD (blue) and MMRD+PPD (orange) status (left), and response association with both SNV and RRD status. (c) Response and overall survival (OS) by microsatellite indels (MS indels). For survival analysis, median MS- indel values were used. (d) Response and overall survival by total MS- indel count for MMRD and MMRD+PPD cancers separately. (e) Kaplan- Meier (KM) estimates using combined SNVs/Mb and MS- indel in all RRD cancers. (Abbreviations: MMRD: mismatch- repair deficiency; PPD: Polymerase- proofreading deficiency; MS- indel: microsatellite insertion/ deletion.)
+
+<--- Page Split --->
+![PLACEHOLDER_49_0]
+
+<center>Figure 5 </center>  
+
+Tumour immune microenvironment, survival and response to PD- 1 blockade. (a) PD- L1 expression, response and survival in all RRD cancers. Cut- off is \(\geq 1\%\) of cells (Methods). (b) CD8 expression, response and survival for RRD cancers. Cut- off is \(\geq 3\%\) of cells (higher than the median; Methods). For both (a) and (b), the histology depicts glioblastoma at 20X magnification. (c) Association of immune markers with SNV and RRD status. (d) Combined immune (PD- L1 and CD8 expression) and genomic (TMB and MS indels) and overall survival in RRD cancers.
+
+<--- Page Split --->
+![PLACEHOLDER_50_0]
+
+<center>Figure 6 </center>  
+
+Characterization of the tumour flare response. (a- d) Analysis of 2 patients who had tumour- debulking prior to therapy and at the time of flare. (a, b) Total immune cell content in pre- therapy and at flare. (c, d) The corresponding CapTCR- sequencing and T- cell receptor clonotype analysis in these samples. (e, f) Immunohistochemistry for PD- L1 expression, and CD8- T- cell infiltration in the pre- therapy sample and at flare, as shown in the representative 20X images from the tumour sample in patient- 1 (P33). (g)
+
+<--- Page Split --->
+
+Representative flow cytometry plot showing activation of CD+ T-cell (TIGIT and 4- 1BB) from the blood sample of a patient before treatment initiation and at flare. (h) 41BB+ CD8+ T-cells in blood from responders without flare, non- responders and flare.  
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+ExtendedDataTableS1FINAL.pdf
+
+<--- Page Split --->

preprint/preprint__0f38253e04174167d6d03bbdbbc7f1a3288595a100abf195116c81c963341bb8/images_list.json

@@ -0,0 +1,17 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Figure 1",
+    "footnote": [],
+    "bbox": [
+      [
+        42,
+        90,
+        875,
+        670
+      ]
+    ],
+    "page_idx": 28
+  }
+]

preprint/preprint__0f38253e04174167d6d03bbdbbc7f1a3288595a100abf195116c81c963341bb8/preprint__0f38253e04174167d6d03bbdbbc7f1a3288595a100abf195116c81c963341bb8.mmd

@@ -0,0 +1,392 @@
+
+# Microspectroscopic visualization of how biochar elevates the soil organic carbon ceiling  
+
+Zhe (Han) Weng  University of Queensland  Lukas Van Zwieten ( \(\square\) lukas.van.zwieten@dpi.nsw.gov.au )  NSW Department of Primary Industries  https://orcid.org/0000- 0002- 8832- 360X  
+
+Michael Rose  NSW Department of Primary Industries  
+
+Bhupinder Pal Singh  NSW Department of Primary Industries  
+
+Ehsan Tavakkoli  NSW Department of Primary Industries  
+
+Stephen Joseph  University of New South Wales  
+
+Lynne Macdonald  CSIRO  
+
+Stephen Kimber  NSW Department of Primary Industries  
+
+Stephen Morris  NSW Department of Primary Industries  
+
+Terry James Rose  Southern Cross University  
+
+Bräulio Archanjo  Instituto Nacional de Metrologia, Qualidade e Tecnologia  
+
+Caixian Tang  La Trobe University  
+
+Ashley Franks  La Trobe University  https://orcid.org/0000- 0003- 1664- 6060  
+
+Hui Diao  The University of Qld  
+
+Peter Kopittke  The University of Queensland  https://orcid.org/0000- 0003- 4948- 1880  
+
+Annette Cowie  
+
+NSW Department of Primary Industries / University of New England  https://orcid.org/0000- 0002- 3858- 959X
+
+<--- Page Split --->
+
+## Article  
+
+Keywords: soil carbon, soil organic carbon, biochar  
+
+Posted Date: September 10th, 2021  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 860309/v1  
+
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Version of Record: A version of this preprint was published at Nature Communications on September 2nd, 2022. See the published version at https://doi.org/10.1038/s41467- 022- 32819- 7.
+
+<--- Page Split --->
+
+# Microspectroscopic visualization of how biochar elevates the soil organic carbon ceiling  
+
+Zhe (Han) Weng<sup>1,2,3,4</sup>, Lukas Van Zwieten<sup>1,5\*</sup>, Michael T. Rose<sup>1</sup>, Bhupinder Pal Singh<sup>6</sup>, Ehsan Tavakkoli<sup>7</sup>, Stephen Joseph<sup>8</sup>, Lynne M. Macdonald<sup>9</sup>, Stephen Kimber<sup>1</sup>, Stephen Morris<sup>1</sup>, Terry J. Rose<sup>5</sup>, Braulio S. Archanjo<sup>10</sup>, Caixian Tang<sup>3</sup>, Ashley Franks<sup>11,12</sup>, Hui Diao<sup>13</sup>, Peter M. Kopittke<sup>4</sup>, Annette Cowie<sup>2,14</sup>  
+
+<sup>1</sup>NSW Department of Primary Industries, Wollongbar Primary Industries Institute, Wollongbar, NSW 2477, Australia  <sup>2</sup>School of Environmental and Rural Sciences, University of New England, Armidale, NSW 2351, Australia  <sup>3</sup>Department of Animal, Plant & Soil Sciences, Centre for AgriBioscience, La Trobe University, Melbourne, Vic 3086, Australia  <sup>4</sup>School of Agriculture and Food Sciences, The University of Queensland, St. Lucia, Queensland 4072, Australia  <sup>5</sup>Southern Cross University, East Lismore, NSW 2480, Australia  <sup>6</sup>NSW Department of Primary Industries, Elizabeth Macarthur Agricultural Institute, Woodbridge Rd, Menangle, NSW 2568, Australia  <sup>7</sup>NSW Department of Primary Industries, Wagga Wagga Agriculture Institute, Wagga Wagga, NSW 2650, Australia  <sup>8</sup>University of New South Wales, Sydney, NSW 2052, Australia  <sup>9</sup>CSIRO Agriculture & Food, Waite campus, Glen Osmond, SA 5064, Australia  <sup>10</sup>Divisão de Metrologia de Materiais - DIMAT, Instituto Nacional de Metrologia, Normalização e Qualidade Industrial - INMETRO, Duque de Caxias, RJ, 25250-020, Brazil  <sup>11</sup>Department of Physiology, Anatomy and Microbiology, La Trobe University, Melbourne, Vic 3086, Australia  <sup>12</sup>Centre for Future Landscapes, La Trobe University, Melbourne, Vic 3086, Australia  <sup>13</sup>Centre for Microscopy and Microanalysis, The University of Queensland, QLD, 4072, Australia  <sup>14</sup>NSW Department of Primary Industries/ University of New England, Armidale, NSW 2351, Australia  \*e- mail: lukas.van.zwieten@dpi.nsw.gov.au
+
+<--- Page Split --->
+
+1 The soil carbon saturation concept suggests an upper limit to store soil organic carbon (SOC), set by the mechanisms that protect soil organic matter from decomposition. Biochar has the capacity to protect new C including rhizodeposits and microbial necromass. However, the decadal scale mechanisms by which biochar influences the molecular diversity, spatial heterogeneity, and temporal changes of SOC persistence remain unresolved. Here we show that the soil C saturation ceiling of a Ferralsol under subtropical pasture could be elevated by \(2\mathrm{Mg}\) (new) C ha \(^{- 1}\) by the application of Eucalyptus saligna biochar 8.2 years after the first application. Using one, two-, and three- dimensional analyses, significant increases were observed in the spatial distribution of root- derived \(^{13}\mathrm{C}\) in microaggregates (53- 250 \(\mu \mathrm{m}\) , 11 %) and new C protected in mineral fractions (<53 \(\mu \mathrm{m}\) , 5 %). Microbial C- use efficiency was concomitantly improved by lowering specific enzyme activities, contributing to the decreased mineralization of native SOC by 18 %. We provide evidence that the global SOC ceiling can be elevated using biochar in Ferralsols by 0.01- 0.1 \(\mathrm{Pg}\) new C yr \(^{- 1}\) .  
+
+## 13 Main  
+
+14 Human activities risk releasing 260 \(\mathrm{Pg}\) of carbon (C) as carbon dioxide ( \(\mathrm{CO_2}\) ) globally that is irrecoverable on a timescale relevant to avoiding profound climate impacts \(^{1,2}\) . Agricultural soils contribute an average of \(2\mathrm{MgC}\) lost ha \(^{- 1}\) yr \(^{- 1}\) globally \(^{3 - 5}\) . It has been estimated that 122 \(\mathrm{Mg}\) soil organic C (SOC) ha \(^{- 1}\) to 1 m depth has been lost over 1 Mha of land converted to tropical grasslands \(^{6}\) , with 40 % of this area occurring on Ferralsols \(^{7}\) . To meet the Paris Agreement of limiting global warming to below \(2^{\circ}\mathrm{C}\) , the Intergovernmental Panel on Climate Change has shown that \(\mathrm{CO_2}\) removal (CDR) techniques are urgently needed \(^{8,9}\) .  
+
+21 Soil C management \(^{4 - 6}\) and the application of biochar \(^{10}\) are appealing \(\mathrm{CDR}s^{9,11}\) as they also improve soil health, sustain agricultural productivity \(^{12,13}\) , and increase resilience of ecosystem services \(^{14,15}\) . Protecting and rebuilding soil C could draw down \(5.5\mathrm{PgCO_2yr^{- 1}}\) , representing 25 % of the potential of natural climate solutions to deliver CDR through conservation, restoration, and improved land
+
+<--- Page Split --->
+
+25 management practices<sup>6</sup>. However, there are biophysical and socio-economic barriers to CDR with SOC management<sup>4,6,16,17</sup>.  
+
+27 Biochar is recognized as a CDR because of its persistence<sup>9,11</sup> in the environment. The pyrolysis of biomass can deliver bioenergy outcomes, as well as agronomic and non- \(\text{CO}_2\) greenhouse gas benefits through use of biochar as a soil amendment<sup>18- 22</sup>. Biochar systems generally show life cycle climate change impacts of emissions reduction in the range of 0.4 - 1.2 Mg \(\text{CO}_2\text{e Mg}^{- 1}\) dry feedstock<sup>23</sup>.  
+
+31 Organo- mineral interactions can increase SOC persistence, in some cases over a millennial timescale<sup>24,25</sup>. There is a general understanding of the effect of microbial activity<sup>26</sup> and mineral protection<sup>27</sup> on SOC storage. However, there are knowledge gaps on the contribution of molecular diversity of organic compounds, fine- scale spatial heterogeneity, and temporal variability in soil conditions. Such composition- space- time interactions influence the accessibility of decomposer communities to the substrate<sup>28,29</sup>. Here, we propose a mechanism by which biochar acts as a biocatalyst to accelerate the formation of organo- mineral and organo- organic interfaces in microaggregates (53- 250 \(\mu \text{m}\) ) and mineral protection of SOC (Fig. 1). Biochar can sorb root- derived C (rhizodeposits) and form biofilms on its surfaces. The very fine layer of soil minerals that subsequently builds on the surfaces of biochar as it ages in soil<sup>30- 32</sup> protects rhizodeposits from microbial metabolism<sup>33,34</sup>, and at the same time incorporates microbial necromass<sup>35- 38</sup>. This coating can desorb from the surface during aggregate turnover or in response to a change in soil conditions such as pH, redox and moisture<sup>39</sup>. The rhizodeposits and microbial necromass are then captured in microaggregates<sup>35,40,41</sup> (e.g. <250 \(\mu \text{m}\) ). A new coating can then form in its place. These processes repeat, building rhizodeposits in soil over time (Fig. 1). We examine these processes in detail to quantify the potential of biochar to elevate the SOC storage ceiling. To do this, we applied Eucalyptus saligna biochar (550°C) to a historic field site established in 2006<sup>41</sup> (Ferralsol under managed subtropical pasture). The mechanisms (Fig. 1) that we tested included the negative priming via higher microbial C- use efficiency and restricted access to substrates, and enhanced mineral protection via
+
+<--- Page Split --->
+
+catalytic biochar surfaces. We demonstrate the importance of fine- scale spatial heterogeneity and temporal variability of diverse C functional groups in association with mineral fractions for building and protecting rhizodeposits over a decade.  
+
+## Elevating SOC storage capacity  
+
+We hypothesize that biochar enhances the protective mechanisms for soil organic matter (SOM), and that a greater C storage capacity can therefore be obtained through strategic applications of biochar. The field site was converted to managed pasture from subtropical forest 100 years ago. This led to a loss of \(17\%\) of the original soil C stock compared to the adjacent native rainforest (data not shown). To quantify elevated C storage capacity, we measured soil C stocks in the managed pasture over 9.5 years (Table S1) from four treatments: (1) biochar applied to a part of the historical biochar plots 8.2 years after the trial was established ("recent + historical"); (2) biochar applied to a part of the control plots 8.2 years after the trial was established ("recent"); (3) biochar applied 8.2 years previously ("historical"); and (4) nil biochar plots ("control").  
+
+All field plots were managed via annual fertilizer application at the start of winter, coinciding with over sowing by annual ryegrass (Methods). The total soil C stock in the unamended pasture soil (control) did not change over 9.5 years<sup>41,42,43</sup> (Fig. 2a; \(P > 0.05\) ) and remained at \(35 \text{Mg C ha}^{- 1}\) in the 0- 100 mm layer when sampled at 8.2 and 9.5 years after the field trial was set- up. The original application of Eucalyptus saligna biochar (550°C) in 2006 resulted in a rapid increase in soil C to \(40 \text{Mg C ha}^{- 1}\) (10 Mg biochar ha<sup>- 1</sup>, 76% C, 7.6 Mg biochar- C) and SOC continued to increase, plateauing at \(50 \text{Mg C ha}^{- 1}\) at 8.2 years ("historical"). When recent biochar amendment was incorporated into the plots at the same dose after 8.2 years following the original application ("recent + historical"), the C stock immediately increased from 50 to \(56 \text{Mg C ha}^{- 1}\) (Fig. 2a) because of the C added from biochar. The C stock then continued to increase further to \(58 \text{Mg C ha}^{- 1}\) between 8.2 and 9.5 years. This additional \(2 \text{Mg C ha}^{- 1}\) was accumulated because of the increased protection mechanisms for new C provided by the biochar, thus elevating the C storage ceiling (Fig. 2a).
+
+<--- Page Split --->
+
+The additional 2 Mg new C ha \(^{- 1}\) (Fig. 2a) can be explained by negative priming of SOC mineralization (Fig. 2b). Comparing the net cumulative SOC mineralization where there was a recent application of biochar to the control (defined as priming in this study), the recent biochar amendment to the historical plots ("recent + historical") lowered SOC mineralization (i.e. negative priming) by 89 g \(\mathrm{CO_2}\) - C \(\mathrm{m^{- 2}}\) compared with the recent biochar amendment to the control ("recent") which lowered SOC mineralization by 55 g \(\mathrm{CO_2}\) - C \(\mathrm{m^{- 2}}\) ( \(P< 0.05\) , Fig. 2b). Recent biochar initiated a small (3 g \(\mathrm{CO_2}\) - C \(\mathrm{m^{- 2}}\) ) positive priming effect, whilst recent + historical biochar immediately triggered negative priming (Fig. 2b). Neither recent + historical or recent had an impact on total respiration or root respiration compared to the control ( \(P > 0.05\) , Figs. S1 & S2). As a portion of the total \(\mathrm{CO_2}\) flux, root respiration remained relatively consistent (30- 32 %) and was unaffected by treatments ( \(P > 0.05\) , Table S2).  
+
+The additional 2 Mg C ha \(^{- 1}\) that accumulated between 8.2 and 9.5 years because of the recent application of biochar to the historical biochar plots ("recent + historical") accrued through the stabilization of rhizodeposits and microbial necromass (Fig. 3). Recent + historical biochar had a similar proportion of total recovered \(^{13}\mathrm{C}\) (58 ± 5.7 %, Fig. 3a) compared to the historical biochar plots ("historical") (60 ± 9.8 %), with this being around 18 % greater than the control (42 ± 7.3 %) and the recent biochar amendment ("recent") (45 ± 4.5 %; Fig. 3b) after the pulse-labelling event at 9.5 years ( \(P< 0.05\) ; Table S3). The increase in belowground \(^{13}\mathrm{C}\) recovery could be largely explained by an increase in mineral- protected soil organic matter (M- SOM) associated \(^{13}\mathrm{C}\) (14 %, \(P< 0.05\) , Table S4). Initially, recent + historical biochar nearly doubled the \(^{13}\mathrm{C}\) retention in the occluded particulate organic matter (O- POM) fractions (5 mg \(^{13}\mathrm{C} \mathrm{m}^{- 2}\) ) of microaggregates (< 250 μm) and M- SOM fractions (14 mg \(^{13}\mathrm{C} \mathrm{m}^{- 2}\) ) of macroaggregates (250- 2000 μm) at 8.9 years compared to the recent biochar (Figs. S3a & b). The root- derived \(^{13}\mathrm{C}\) from rhizodeposition was accumulated gradually into O- POM in macroaggregates and M- SOM at 9.2 years (Figs. S3c & d), which was in turn transformed into M- SOM fractions in micro- and macroaggregates by 9.5 years (Figs. S3 e & f).  
+
+Microbial contribution and responses to stabilization of rhizodeposits
+
+<--- Page Split --->
+
+To determine the microbial contribution to the increased SOC storage capacity, we quantified catabolic enzyme activities, metabolic quotient of native SOC (bulk soil) and rhizodeposition ( \(^{13}\mathrm{C}\) content), and specific enzyme activity (the ratio of enzyme activity- to- microbial biomass) comparing a recent biochar amendment to the control ("recent") and to the historical biochar plots ("recent + historical"). Microbial biomass was increased by \(11\%\) in the recent + historical biochar compared to the recent biochar between 8.9 and 9.5 years (Table S5a). This might result from the stimulation of microbial co- metabolism \(^{31}\) of biochar- C, root- C and SOC which induced initial positive priming in the recent biochar amendment (Fig. 2b). The recent + historical biochar had increased substrate- induced respiration for citric, oxalic and malic acids compared to the recent biochar (Microresp, Fig. S4; Table S6). No differences were detected for 12 other substrates. This greater respiration induced by carboxylic acids (e.g. root exudates) may partially explain the higher metabolic quotient associated with bulk SOC and rhizodeposition in the recent biochar cf. the recent + historical biochar (Table S5b & c). The recent + historical biochar might result in higher substrate- use efficiency which supports an earlier establishment of negative priming compared to the recent biochar (Fig. 2b). It was previously shown that the recent biochar significantly increased bacterial diversity and the relative abundance of nitrifiers and bacteria consuming biochar C after one year, but the soil bacterial communities from the recent + historical plots did not differ from the control \(^{33}\) . This suggests that the microbial accessibility to SOM might be limited in the recent + historical biochar plots, whereas in the recent biochar, the soil microorganisms had to cope with changes in C- substrate type and availability \(^{45}\) .  
+
+The ratio of enzyme activities to total microbial biomass was lower in both recent + historical biochar and recent biochar compared to the control (Table S7) despite no difference in enzyme activities (Table S8). This suggests that for a given amount of microbial biomass, less enzymes were produced in the biochar- amended soil. A low ratio of extracellular enzymes to microbial biomass can slow down the degradation of native SOC \(^{46}\) . This is consistent with increased microbial C- use efficiency (Tables S4b & c), which may indirectly contribute to the negative priming. It has been suggested that the presence of opportunistic microbes that meet their energy and nutrient demands by exploiting the
+
+<--- Page Split --->
+
+catalytic activities of decomposers could lower the specific enzyme activity<sup>46</sup>. The spatial arrangement between microbes and substrates is critical to this process.  
+
+## Spatial examination of organo-mineral interactions  
+
+To better understand the process of negative priming following biochar application, we examined whether the stabilization of rhizodeposits may be facilitated via protection with the abundant Fe and Al in soil. Root exudates can interact with Fe oxides in the soil to promote the formation organomineral complexes. Aluminum may also protect root- C from biodegradation. The development of organo- mineral complexes was assessed using three- dimensional focused ion beam (FIB) coupled with scanning electron microscopy (SEM) with energy dispersive X- ray spectroscopy (3D- FIB- SEM- EDS) on intact representative soil aggregates from the recent + historical biochar (Fig. 3c) where C retention co- located with clay minerals, but, not in the recent biochar (Fig. 3d).  
+
+To provide a deeper insight into the molecular diversity of organic compounds and the temporal and lateral arrangement with respect to organo- mineral interfaces, we conducted in situ spectromicroscopic analysis of free water- stable microaggregates (53- 250 μm) and organo- mineral fractions (<53 μm) of the recent + historical biochar and recent biochar, and of field- extracted biochars at micro- and nanoscale. The C functional groups, examined using synchrotron- based soft X- ray (SXR) analyses, from the microaggregates (53- 250 μm) were dominated by quinones (284.1 eV), aromatic C (285.2 eV, 1s- π\* transitions of conjugated C=C), and aliphatic C (287.3 eV) (Fig. 4a). For the mineral- protected fractions (<53 μm), two prominent features were the low intensity of quinones and high intensity of aliphatic C (Fig. 4a). The dominant peaks of aliphatic, amide and carboxylic C (287- 289 eV) are the direct consequence of deposition of microbial metabolites or debris, exopolysaccharides, and root exudates onto mineral surfaces<sup>27,47,48</sup>. These align with the micro- spatial maps produced from the synchrotron- based infrared (IR) microspectroscopy (Fig. 4b). The close correlation between clay minerals and microbial metabolites on the biochar surface supports SOC stabilization, and highlights the importance of clay minerals for the protection of SOC and the control of microbial activity.
+
+<--- Page Split --->
+
+The correlation between C forms and clay minerals, confirmed by the IR maps, was further examined on a nanometer scale by directly analyzing the chemical composition of the organo- mineral coatings at the surface and in the pores of field- extracted biochars. Greater intensities of quinones (284.1 eV) and carboxyl C- OOH (288.6 eV) were observed in the 9.5- year aged biochar compared with the 1- year aged biochar (Fig. 5a). A high magnification image of the area where the fungi were located inside a biochar fragment shows a high concentration of irregular pores and a coating of organic material (Figs. 5b & c). Fungi can mine nutrients from minerals by exuding acids<sup>49,50</sup> which may cause the observed microporosity of organo- mineral- biochar interfaces (Figs. 5b, d, f, h & Fig. S5). Energy dispersive X- ray spectroscopy (EDS) analysis showed that complex changes had occurred on the surface of the biochar over the one- year period (Figs. 5c, 5g). Positively charged nanoparticulate minerals rich in Al, Si, Ca, P, Fe and other cations were attracted to the surface of the negatively charged areas on the biochar. These positively- charged minerals and elements subsequently attracted negatively- charged organic molecules with detectable concentrations of C=C, C-OH, C-N/C=N, C=O, COOH functional groups, quinone bonds and anions thus initiating a process whereby porous clusters are formed on the biochar surface (Figs. 5e, 5i). Similarly, exudates from plants and microorganisms can be deposited around mineral surfaces on the biochar, and cations and minerals can be attracted to these organic molecules. Recent biochar amendment to historical biochar plots would provide unoccupied surfaces and pores in the soil to increase sorption capacity for root exudates<sup>51</sup>, which would then serve as binding agents to further enhance aggregate formation<sup>52</sup>. As these clusters are built up, they may also be detached from the biochar either through fluctuating redox conditions and interaction with microbes or perturbation caused by soil invertebrates<sup>30</sup>.  
+
+These biochar micro- sites have a high concentration of free radicals with labile easily- mineralizable organic C and/or inorganics dissolved from the biochar (Table S9). Colloidal biochar particles, leachate, dissolved native OM and rhizodeposits may be further stabilized separately or held together via cation bridging with \(\mathrm{Ca^{2 + }}\) , or with Al and Fe oxyhydroxides<sup>53,54,55</sup> and organo- organic interactions at the nanometer scale<sup>56</sup> (Figs. 5e, 5i). These processes may be encouraged by oxidation of the biochar
+
+<--- Page Split --->
+
+surface as it ages in soil<sup>36,37</sup>. This is supported by our LC- OCD results where dissolved hydrophobic C fractions and building blocks (medium molecular weight) were greater in the recent + historical biochar compared to the recent biochar (Fig. 5j; Table S10). The analysis of the surface of the 9.5- year aged biochar by C- edge EELS and XPS indicated that most of the oxidized C species were formed in the organo- mineral coating. The concentrations of the different functional groups appear to be influenced by the presence of nanophase Fe, Si and Al oxides<sup>36</sup> (Fig. 5i).  
+
+## Global impact of elevating the soil carbon ceiling  
+
+Building SOC is a global priority<sup>9</sup>, and our results showed that the SOC storage ceiling can be elevated through single or multiple applications of biochars. We observed a plateau in rhizodeposit accumulation rate over 9.5 years in the historical biochar plots (Fig. 2a; \(\gamma = 4.24\ln (x) + 17.6\) ; \(R^2 = 0.95\) ) which implies that the system was approaching a new (16 % higher) equilibrium for SOC storage, ten years after the initial application. We showed that a strategic application of 10 Mg biochar \(ha^{- 1}\) after 8.2 years raised the SOC storage ceiling by a further 2 Mg C \(ha^{- 1}\) . In summary, this Rhodic Ferralsol under the managed pasture had a C storage capacity of 35 Mg C \(ha^{- 1}\) in the surface soil, which increased to 44 Mg C \(ha^{- 1}\) one year following the application of biochar, which further increased to 50 Mg C \(ha^{- 1}\) after nearly a decade. The C storage ceiling was further raised to 59 Mg C \(ha^{- 1}\) where biochar was applied to the historically amended field plots. Of this C storage, 7.6 Mg C \(ha^{- 1}\) was attributed to the addition of C from biochar, while 2 Mg C \(ha^{- 1}\) was attributed to the stabilization of new C.  
+
+The long- term stabilization of rhizodeposits by biochar after 9.5 years has significant implications for elevating the SOC ceiling. Plants release \(\sim 50\%\) of photosynthetically fixed C into the soil, which is available for microbial growth<sup>57- 59</sup>. Global grasslands annually contribute 0.04 Pg C to SOC<sup>6</sup>. The retention and stabilization of belowground C by biochar could play an important part in a natural climate solution for tropical grasslands, which occupy 0.7 Gha of land with an estimated global C content of 30 Pg C. Here we showed a 16 % increase in retention of new C in the recent + historical plots compared with the control via the stabilization of root- derived <sup>13</sup>C in microaggregates (53- 250
+
+<--- Page Split --->
+
+\(\mu \mathrm{m}, 11 \%\) ) and microbial products and rhizodeposits protected in mineral fractions (<53 \(\mu \mathrm{m}, 5 \%\) ) (Fig. 3). To demonstrate the potential of raising the C saturation ceiling, we extrapolated this \(16 \%\) increase in stabilization based on the global projected biochar production. We estimated the strategic application of biochar can increase SOC storage capacity in Ferralsols by 0.01- 0.1 \(\mathrm{Pg C yr^{- 1}}\) worldwide (Supplementary information). This mechanism would increase the global mitigation potential using biochar as a soil amendment, estimated at 1.3 \(\mathrm{Pg C yr^{- 1}}\) (central average; Woolf et al. 2010), by another \(0.8 - 7.7 \%\) .  
+
+In our study, we raised the SOC storage capacity in a subtropical pasture with a strategic application of a Eucalyptus saligna biochar (550°C) 8.2 years following the original biochar application. Of importance to building soil C stocks, the strategic application of biochar in the aged plots resulted in \(16 \%\) more new C (i.e. microbial products and rhizodeposits) being stored as soil C in both microaggregates (53- 250 \(\mu \mathrm{m}\) ) and mineral fractions (<53 \(\mu \mathrm{m}\) ). Microbial C- use efficiency was improved by lowering the specific enzyme activity and slowing down degradation of SOC and rhizodeposits (negative priming). Our in situ spectromicroscopic analyses suggest that the catalytic biochar surfaces accelerated the micro- and nanoscale heterogeneity and temporal variability for new C storage.  
+
+## Methods  
+
+## Field site details  
+
+The field experiment was situated at the Wollongbar Primary Industries Institute (28°49'S, 153°23'E, elevation: 140 m), Wollongbar, New South Wales, Australia. The classification and properties of the soil can be found in Weng et al. (2015). Briefly, the Rhodic Ferralsol is fine- textured and Fe- rich mineral soil dominated by kaolinite, gibbsite and goethite. The 100- mm topsoil had a \(\mathrm{pH_{CaCl2}}\) of 4.5 with total C of \(35 \mathrm{g kg^{- 1}}\) , total Fe \(84 \mathrm{g kg^{- 1}}\) , and total Al \(67 \mathrm{g kg^{- 1}}\) . Details of the initial field site setup in 2006 can be found in Slavich et al. (2013). The treatments (n=3) included (1) Eucalyptus saligna biochar incorporated into the topsoil (0- 100 mm) at \(10 \mathrm{th a^{- 1}}\) (eq. \(1 \% \mathrm{w / w}\) , bulk density of \(1 \mathrm{g cm^{- 3}}\) ) plus NPK
+
+<--- Page Split --->
+
+fertilizer and (2) NPK only (control). The biochar was derived from a single source of above- ground biomass of mature Eucalyptus saligna and pyrolyzed at \(550^{\circ}C\) with a residence time of 30 mins (Pacific Pyrolysis, NSW, Australia). The physicochemical properties of the biochar can be found in Slavich et al. (2013). A tetraploid annual ryegrass (Lolium multiflorum) was broadcast at a seeding rate of \(35kg\) ha \(^{- 1}\) and repeated annually. Urea was applied at \(46kgNha^{- 1}\) on six occasions ( \(276kgha^{- 1}\) in total) between winter and spring each year following manual cuts of the pasture to simulate grazing. Basal nutrients were applied annually at sowing (Slavich et al. 2013).  
+
+In April 2014, the control (NPK only) and 8.2- year- old biochar (Eucalyptus saligna) plots were superimposed with nine subplots \((0.5m\times 0.5m)\) (Weng et al. 2017). Four treatments were: (1) Recent biochar to historical biochar plots (biochar applied to a part of the historical biochar plots 8.2 years after the trial was established; "recent + historical"); (2) Recent biochar amendment (biochar applied to a part of the control plots 8.2 years after the trial was established; "recent"); (3) Historical biochar amendment (biochar applied 8.2 years previously, "historical"); and (4) Control (nil biochar plots). There was one subplot per field replicate and a total of three field replicates. The biochar added to the control and aged plots was taken from the biochar 'batch' applied in 2006, which had been air- dried and archived in sealed 200 L steel containers at room temperature. The details of the subplot set up and installation of belowground respiration collars can be found in Weng et al. (2015). Bulk density of the biochar was \(0.332gcm^{- 1}\) measured using a method described in Quin et al. (2014). The weight of soil- biochar mixture in the topsoil (100- mm) was determined based on the bulk density assessed in each treatment. Before application, the biochar was sieved to \(< 2mm\) . The soil/biochar mixture was carefully packed into the subplots. The control subplots were also excavated and repacked to a bulk density of \(1gcm^{- 3}\) . A root signature sand collar (50 mm diameter) was installed in each of the control subplots to measure the \(\delta^{13}C\) signature of root respiration. It was packed with acid- washed sand and planted with ryegrass (i.e. a down- sized version of the soil plus root respiration collar). Similarly, a biochar+root signature sand collar was packed with a biochar- sand mixture (1 % w/w) in each of the recent + historical biochar subplots. To maintain the root growth into the collars,
+
+<--- Page Split --->
+
+NPK fertilizers were applied at the same dose as in Slavich et al. (2013). This study employed the same pasture management regime as Slavich et al. (2013), in terms of ryegrass sowing, NPK applications and weed control.  
+
+## Periodic \(^{13}\mathrm{C}\) pulse labelling to quantify SOC mineralization and root respiration  
+
+To understand how plant- biochar- soil interactions affect SOC priming, \(^{13}\mathrm{CO}_{2}\) pulse labelling campaigns were conducted on three occasions: 12 June 2014, 01 August 2014 and 30 July 2015. The procedure of the pulse labelling experiment and a detailed description of quantification of SOC mineralization and root respiration using three- pool C partitioning can be found in Weng et al. (2015, 2018).  
+
+Soil sampling was carried out at 8.9, 9.2 and 9.5 years following the first biochar application in 2006. Intact soil cores (40 mm in diameter) were sampled to 80 mm depth within each subplot but outside the respiration collar area to reduce disturbance. The sampled areas were avoided in the ensuing sampling events. The sample was mixed evenly and analyzed for pH, total soil organic C, and microbial biomass carbon (MBC). Soil pH was measured on the samples prepared for the enzyme assay (1:5 w/w ratio in distilled water with constant stirring using a vortex) using an IntelliCAL PHC101 pH probe on a Hach HQ40d portable meter (Loveland, Colorado, USA). The analytical procedures for SOC and MBC can be found in Weng et al. 2015. The remaining soil was stored at - 20°C.  
+
+At the 15- month soil sampling event, fresh soil was also taken from the soil respiration collars (i.e. unplanted) and soil+root respiration collars (i.e. planted) to quantify the effect of plant- biochar- soil interactions on catabolic enzyme activity, substrate- induced respiration, and MBC. The MBC was analyzed using the chloroform fumigation method (Van Zwieten et al. 2010). Metabolic quotient of total C or rhizodeposits was then quantified as the ratio of respiration (native SOC or \(^{13}\mathrm{C}\) - labelled root respiration) over the total MBC. The metabolic quotient has been used as an indicator of C- use efficiency (Fang et al., 2018). Detailed calculations for determining rhizodeposit- derived respiration are given in Weng et al. (2015).
+
+<--- Page Split --->
+
+## SOC priming in the plant-biochar-soil systems  
+
+The rhizosphere priming of native SOC from the biochar- plant- soil interactions was quantified using a three- pool C partitioning model. Specialized respiration collars were used to isolate soil plus root respiration from shoot respiration<sup>41,42</sup>. Native SOC mineralization was separated using the \(\delta^{13}\mathrm{C}\) signature of biochar plus root (biochar plus root sand collars, Supplementary Information) and total respiration (biochar plus soil plus root collars) after pulse labelling. Moisture content was maintained between 60- 80% field capacity in the root collars to minimize potential C isotopic fractionation during photosynthesis caused by water stress<sup>44</sup>. The \(\delta^{13}\mathrm{C}\) signatures of extracted field- aged biochar and fresh biochar (the same biochar archived in a sealed container for 8.2 years) were both \(- 25 \pm 0.1\%\) . Any interactive effect of biochar and root on the \(\delta^{13}\mathrm{C}\) signature of soil would be surpassed by a greater level of \(\delta^{13}\mathrm{C}\) enrichment of the root component compared with any isotopic signature contribution from soil and biochar to the \(\delta^{13}\mathrm{C}\) signature of the total respiration. A sensitivity analysis of C source partitioning was performed to assess the impact of plant- biochar ( \(C_3\) - dominated)- soil interactions on \(\delta^{13}\mathrm{C}\) signatures of soil (a mixture of \(C_3\) and \(C_4\) pools). Errors generated from isotopic partitioning were propagated using the first order Tyler series approximations of the variances of native SOC mineralization.  
+
+The recovery of \(^{13}\mathrm{C}\) in various SOC pools at time t (i.e. \(\mathrm{A}^{13}\mathrm{C}_{i,t}\) , in \(\%\) ) was calculated by dividing the amount of \(^{13}\mathrm{C}\) (g m<sup>2</sup>) in a specific C pool (i.e. C<sub>i</sub>) by the initial amount of total added \(^{13}\mathrm{CO}_{2}\) (g m<sup>2</sup>) at each labelling event (i.e. \(^{13}\mathrm{C}_{\mathrm{added}}\) ):  
+
+\[\mathrm{A}^{13}\mathrm{C}_{i,t} = (^{13}\mathrm{C}_{\mathrm{excess},t}\times \mathrm{C}_{i}) / ^{13}\mathrm{C}_{\mathrm{added}}\times 100\]  
+
+where; represents soil plus root respiration, root biomass, soil aggregates or its associated fractions, \(^{13}\mathrm{C}_{\mathrm{excess},t}\) indicates the increment of the \(^{13}\mathrm{C}\) atom \(\%\) of an individual C pool from its natural abundance level at a specific sampling time, t.
+
+<--- Page Split --->
+
+299 The mineralization of native SOC \((C_{S})\) was calculated using the \(^{13}\mathrm{C}\) signature of biochar+root \((^{13}\mathrm{C_{B + R}}\) , sand collar) from the plant- biochar- soil systems after pulse labelling:  
+
+\[C_{S}(\%) = 100^{*}(\delta^{13}C_{T} - \delta^{13}C_{B + R}) / (\delta^{13}C_{S} - \delta^{13}C_{B + R}) \quad (1)\]  
+
+where \(\delta^{13}C_{T}\) : \(\delta^{13}C\) signature of the total respiration from the planted system after pulse labelling; \(\delta^{13}C_{S}\) : \(\delta^{13}C\) signature of the soil- derived \(\mathrm{CO_2}\) - C evolved from the unplanted control soil without pulse labelling.  
+
+The percentage of soil- derived \(\mathrm{CO_2}\) - C in the total respiration from the planted control soil \((C_{S}(\%)\) was determined (Weng et al. 2015):  
+
+\[C_{S}(\%) = 100^{*}(\delta^{13}C_{T} - \delta^{13}C_{R}) / (\delta^{13}C_{S} - \delta^{13}C_{R}) \quad (2)\]  
+
+where \(\delta^{13}C_{T}\) : \(\delta^{13}C\) signature of the total respiration from the planted control; \(\delta^{13}C_{S}\) : the \(\delta^{13}C\)  
+
+signature of the unplanted control soil; \(\delta^{13}C_{R}\) : the \(\delta^{13}C\) signature of roots, which was determined from root respiration from the root sand collar as described in Weng et al. (2017).  
+
+Similarly, the percentage of soil- derived \(\mathrm{CO_2}\) - C in the total respiration from the unplanted biochar- amended soil \((C_{S^{\prime}}(\%)\) ) was determined:  
+
+\[C_{S^{\prime}}(\%) = 100^{*}(\delta^{13}C_{T^{\prime}} - \delta^{13}C_{B}) / (\delta^{13}C_{S} - \delta^{13}C_{B}) \quad (3)\]  
+
+where \(\delta^{13}C_{T^{\prime}}\) : the \(\delta^{13}C\) signature of the total respiration from the unplanted biochar soil. \(\delta^{13}C_{S}\) : the \(\delta^{13}C\) signature of the unplanted control soil; \(\delta^{13}C_{B}\) : the \(\delta^{13}C\) signature of either fresh (- 25.02 ± 0.13 \(\%\) ) or aged biochar (- 25.04 ± 0.11 \(\%\) ). Biochars were recovered by hand from field soil samples, thoroughly rinsed with distilled water on a 100 μm sieve and oven- dried at 50°C for 24 h.  
+
+Rhizosphere priming was calculated in two systems:  
+
+i. Unamended system (Planted vs. Unplanted)
+
+<--- Page Split --->
+
+- SOC planted, unamended: soil C mineralization in the planted control calculated by \(^{13}\mathrm{C}\) -enriched root end-member  
+
+- SOC unplanted, unamended: soil C mineralization in the unplanted control  
+
+Rhizosphere priming in the control soil:  
+
+\(\Delta \mathrm{SOC}_{\mathrm{unamended}} = (\mathrm{SOC}_{\mathrm{planted, unamended}}) - (\mathrm{SOC}_{\mathrm{unplanted, unamended}})\)  
+
+ii. Biochar-amended system (Planted vs. Unplanted)  
+
+- SOC planted, amended: soil C mineralization in the planted biochar soil partitioned from \(^{13}\mathrm{C}\) -  
+
+enriched 'Biochar+Root' end-member  
+
+- SOC unplanted, amended: soil C mineralization in the unplanted biochar soil partitioned from  
+
+biochar end members  
+
+Rhizosphere priming in the biochar system:  
+
+\(\Delta \mathrm{SOC}_{\mathrm{amended}} = (\mathrm{SOC}_{\mathrm{planted, amended}}) - (\mathrm{SOC}_{\mathrm{unplanted, amended}})\)  
+
+SOC priming was the difference in native SOC mineralization between the biochar-amended and  
+
+control soils:  
+
+\(\Delta \mathrm{SOC} = (\mathrm{C}_{\mathrm{s}}(\%)^{*}\mathrm{C}_{\mathrm{Tplanted}} - \mathrm{C}_{\mathrm{s}'}(\%)^{*}\mathrm{C}_{\mathrm{Tunplanted}}) / 100\)  
+
+where \(\mathrm{C}_{\mathrm{Tplanted}}\) and \(\mathrm{C}_{\mathrm{Tunplanted}}\) are the total respiration in planted and unplanted systems either with biochar amendment or the control.  
+
+Calculated \(^{13}\mathrm{C}\) atom \(\% (\%)\) :  
+
+\(^{13}\mathrm{C}\) atom \(\% = [(\delta^{13}\mathrm{C} + 1000)^{*}R_{\mathrm{PDB}}]^{*}100 / [(\delta^{13}\mathrm{C} + 1000)^{*}R_{\mathrm{PRB}} + 1]\)  
+
+where \(\mathrm{R}_{\mathrm{PDB}} = 0.01118\)
+
+<--- Page Split --->
+
+## Sensitivity analysis of isotopic partitioning  
+
+Because of the uncertainty of the direction of biochar- induced priming of soil carbon and/or rhizodeposits, the contribution of biochar on the \(^{13}\mathrm{C}\) endmember of \((\delta^{13}\mathrm{C}_5)\) was assessed. Therefore, three alternative scenarios of three- pool C partitioning were evaluated:  
+
+1) dominant positive priming of new C from the \(\mathsf{C}_3\) pasture, where \(\delta^{13}\mathrm{C}_5 = -27\%\) (i.e. the upper boundary, grey dashed line, Fig. 2b);  
+
+2) equal native SOC priming and rhizosphere priming, hence, the same \(^{13}\mathrm{C}\) signatures of soil+root in the biochar and control plots, where \(\delta^{13}\mathrm{C}_5 = \delta^{13}\mathrm{C}_5\) (i.e. solid lines in Fig. 2b);  
+
+3) dominant positive priming of the native \(\mathsf{C}_4\) -dominant SOC, where \(\delta^{13}\mathrm{C}_{S + R'} = -13\%\) (i.e. the lower boundary, grey dashed line, Fig. 2b).  
+
+The boundary conditions were calculated from the published \(^{13}\mathrm{C}\) signatures for Scenarios 1 and 3 (Farquhar et al. 1989). The \(95\%\) confidence intervals were the combination of the lowest and highest scenarios (n = 3). First order Tyler series of the variances of the percentage of soil respiration, \(\mathsf{C}_5(\%)\) , were approximated to propagate errors from isotopic partitioning (Derrien et al. 2014).  
+
+\[\sigma^2\mathrm{C}_5(\%) = (\sigma^2\delta^{13}\mathrm{C}_T - \sigma^2\delta^{13}\mathrm{C}_5) / (\delta^{13}\mathrm{C}_T - \delta^{13}\mathrm{C}_5)^2 \quad (6)\]  
+
+## Enzyme activity and substrate-induced respiration  
+
+The determination of catabolic enzyme activities using a soil suspension method is described in Weng et al. (2017). Six treatments were derived from the control, the recent + historical biochar and recent biochar plots in both the unplanted (i.e. soil respiration collar) and planted (i.e. soil+ root respiration collar) systems (Table S5, Weng et al., 2017). After 7- d incubation at \(40\%\) water- holding capacity (WHC), the activities of four C- degrading enzymes: \(\beta\) - glucosidase, xylosidase, cellulase, and N- acetyl- glucosaminidase, in the soils were analysed using a fluorescent microplate reader (BMG labtech FLUOstar Omega). Specific C enzyme activity was obtained by dividing the activity of individual
+
+<--- Page Split --->
+
+enzymes over the total MBC at each sampling time. These ratios provided an indication of the C- turnover efficiency of the soil microbial community (Kaiser et al., 2015). Substrate- induced respiration was used to measure Community level physiological profiles using the MicroResp™ method (Campbell et al., 2003) with minor modifications. Fresh soil samples, packed in 96- deepwell plates (around 0.5 g per well), were prepared in the same manner as the enzyme experiment (i.e. incubation conditions). Each treatment per field replicate was sub- replicated eight times for measurement. The experimental protocol is detailed in Weng et al. (2017). Fifteen C substrates (Table S6) were selected to represent a broad range of soil and root exudates (Campbell et al. 2003; Chapman et al. 2007).  
+
+## Aggregate size and density fractionation  
+
+Aggregate size (dry sieving) and density fractionation was conducted based on the method described by Weng et al. (2018). No large macroaggregates ( \(>2000 \mu m\) ) was found in this study. Macroaggregates (250- 2000 \(\mu m\) ) and microaggregates ( \(< 250 \mu m\) ) were fractioned into free POM (F- POM, \(\rho < 1.6 \text{kg m}^{- 3}\) ), occluded POM (O- POM, \(>53 \mu m\) ), and mineral- protected soil organic matter (M- SOM, combining silt- and clay- bound SOM, \(< 53 \mu m\) ).  
+
+## Belowground \(^{13}\mathrm{C}\) pools  
+
+The C and N content, and \(\delta^{13}\mathrm{C}\) signatures of bulk soil, aggregates and fractions were measured using a PDZ Europa ANCA- GSL elemental analyzer interfaced to a PDZ Europa 20- 20 isotope ratio mass spectrometer (Sercon Ltd., Cheshire, UK) (Weng et al., 2015).  
+
+## 3D-FIB-SEM-EDS  
+
+The serial section and EDS Mapping of the soil particle was prepared in a FEI SCIOS focused ion beam/scanning electron microscope (FIB/SEM) DualBeam system. The SCIOS FIB/SEM DualBeam system has a vertical mounted SEM column and an ion column sitting at an angle of 52 degrees with respect to the electron column. The particle was located with the aid of the electron beam. Before
+
+<--- Page Split --->
+
+any milling, one micrometre thick platinum layer was deposited on the sample surface covering the area of interest to prevent it from damage caused by the ion bombardment in the following steps. The smooth finish of the Pt layer would also help to reduce the curtaining effect during the following milling procedure. The serial sectioning of the volume was carried out at 3nA and 30kV ion beam current and the EDS mapping were collected at 5kV and 6.4nA electron beam current. The voxel size of the SEM images is \(84 \text{nm} (\times) \times 84 \text{nm} (\times) \times 1000 \text{nm} (\text{z}, \text{slicing thickness}).\)  
+
+## Synchrotron soft X-ray  
+
+Synchrotron-based soft X- ray (SXR) analysis was performed at the SXR Spectroscopy beamline (14ID) at the Australian Synchrotron on the microaggregate (53- 250 μm) and mineral fractions (<53 μm) from 1) recent biochar- amended plots, and 2) the historically biochar- amended plots; and then biochar recovered from the soil, that is: i) 1- year (aged) and ii) 9.5- year (aged) biochar. The samples were ground to fine powder and mounted on double sided carbon tape affixed to a stainless steel ruler.  
+
+The SXR spectra were collected at an angle of \(100^{\circ}\) to the beam over a photon energy range of 275- 325 eV with a step size of 0.1 eV. The energy was calibrated using a graphite standard in the beamline which was collected simultaneously with the \(I_0\) and sample SXR spectra. The double normalization and a pre- and post- edge linear subtraction (background) were conducted using the Athena software (Stöhr 2013).  
+
+## Synchrotron IR  
+
+For infrared microspectroscopy, approximately \(\sim 30\) free water- stable microaggregates (53- 250 μm) and mineral fractions (<53 μm) were hand- picked on a glass fibre filter paper and humidified gently over 18 hours (Lehmann et al. 2017; Hernandez- Soriano et al. 2018). The aggregates and fractions were frozen at \(- 20^{\circ}C\) before being cryo- ultramicrotomed at 200 nm using a diamond knife. No embedding media was used. The multiple sections per sample (n > 2) were directly collected on CaF2 windows (IR transparent).
+
+<--- Page Split --->
+
+The sections on \(\mathsf{CaF}_2\) were directly scanned at the IR beamline at the Australian Synchrotron using a Bruker Hyperion 3000 infrared microscope and a V80v Fourier transform infrared spectrometer. The detail of the microscope was described in Hernandez- Soriano et al. (2018). The spectral maps were produced in transmission mode from 64 scans with a resolution of \(4 \mathsf{cm}^{- 1}\) , step size of \(5 \mu \mathsf{m}\) . Multiple maps were acquired for each treatment to represent the heterogeneity of the sample.  
+
+Maps were processed using the software OPUS 8.2 (Bruker Optik GmbH, Germany), targeting absorbance at \(3630 \mathsf{cm}^{- 1}\) (O-H groups of clays), \(2920 \mathsf{cm}^{- 1}\) (aliphatic- C), \(1600 \mathsf{cm}^{- 1}\) (aromatic- C), and \(1035 \mathsf{cm}^{- 1}\) (polysaccharides- C)(Hernandez- Soriano et al. 2018). The area of these four absorbance peaks was integrated to the map. A linear regression was conducted to assess the correlation between clay content and the selected C functional groups.  
+
+## STEM-EDS-EELS and XPS  
+
+Forty biochar particles were extracted from the soil samples per plot and were examined using a Zeiss Sigma Scanning electron microscope. Detailed analysis of 5 particles was carried using a Bruker X- ray Dispersive analyser (EDS). A Cs- corrected FEI Titan 80/300 scanning transmission electron microscope (STEM) working at \(80 \mathsf{keV}\) , equipped with a Gatan imaging filter Tridiem and an EDX analyzer was utilised to determine the structure and composition of the organo- mineral clusters that had formed on the surface of the aged biochar. Twenty biochar particles were sonicated in ethanol and then a sample of this was placed on a lacey carbon grind as described by Archanjo et al (2017). Detailed examination of 2 clusters was carried out using energy electron loss spectroscopy (EELS) and EDS. X- ray photoelectron spectroscopy (XPS) examination of both whole and crushed \((< 0.5 \mathsf{mm})\) 1- year aged particles of biochar was undertaken. Carbon 1s photoelectron peak was decomposed in five components: C1- C5 (Table S9). The first one (C1) centered in \(284.6 \mathsf{eV}\) , typical of electrons in carbon \(\mathsf{sp}^2\) bounds (C=C), i.e., delocalized \(\mathsf{sp}^2\) electrons. For this component, an asymmetrical line shape was used to fit. The asymmetry of the C1 component, known as a "defect peak", is related to the localized \(\mathsf{sp}^2\) electrons. Electrons of C- C or C- H bounds typically appear with a binding energy shift of \(0.9 \mathsf{eV}\) in
+
+<--- Page Split --->
+
+relation to sp² delocalized electrons, causing a broadening in the first component. The component C2, centered in 286.2±0.2 eV may be attributed to C- OH (phenol or hydroxyl groups), ether (C- O- C) or pyrrolic groups (C- N). Some authors also attribute this component to Csp³ free radicals. The component C3 is attributed to carbonyl groups (C=O) centered in 287.5±0.4 eV, the component C4 is attributed to the carboxyl groups (COOH) centered in 289.1 ± 0.3 eV, and the last one, the component C5 in 291.4 eV is attributed to the shake- up satellite peak, characteristic of π→π* transition of electrons delocalized sp².  
+
+## The concentration of DOC and its fractions, measured by LC-OCD  
+
+Dissolved organic carbon (DOC) in a water solution was analysed using liquid chromatography – organic carbon detection (LC- OCD). Two major fractions were: chromatographic organic carbon (CDOC) and hydrophobic organic carbon (HOC). CDOC (hydrophilic fraction) can be categorized into five fractions as a factor of retention time and molecular weight: i) biopolymers, ii) persistent C- like substances, iii) building blocks, iv) low molecular weight acids and v) low molecular weight neutrals. Samples were extracted in distilled water with a ratio of 1:10 (w/v). The solutions were regularly stirred at 50 °C for 24 hours before filtration to differentiate solid and liquid phases.  
+
+## Calculation and statistical analysis  
+
+The cumulative SOC, biochar- C mineralization, and root respiration over 466 d were calculated as the area of a linear interpolation across all measurement points. All statistical analyses were conducted within the R environment (R development core team 2012). When significant F- tests were obtained (P= 0.05), means were separated using a least significant difference (LSD) test at the 0.05 probability.  
+
+## Calculations of global implication for increasing soil carbon sink  
+
+Projection for wood biochar production is estimated at 4.8 - 8.3 Pg, based on the total annual production of 5.5 to 9.5 Pg biochar by 2100 (Lehmann et al. 2011) with 87% of the feedstock as
+
+<--- Page Split --->
+
+wood (Jirka and Tomlinson 2013). Using the same application rate in this current study (10 t ha<sup>-1</sup>), all 
+
+wood biochar is assumed to be applied to 0.5 - 0.8 Gha in 2100, accounting for up to 100 % of 
+
+tropical Ferralsol and 36 % of tropical grasslands (Lal 2004). The global C sequestration rate in 
+
+grasslands is reported between \(1.3 \times 10^{-10}\) and \(7.6 \times 10^{-10}\) Pg C ha<sup>-1</sup> yr<sup>-1</sup> (Minasny et al. 2017). The 
+
+range of C sequestration in grasslands before biochar amendment in 2100 would be around 0.07- 
+
+0.61 Pg C (i.e. 0.5 or 0.8 Gha at \(1.3 \times 10^{-10}\) and \(7.6 \times 10^{-10}\) Pg C ha<sup>-1</sup> yr<sup>-1</sup>). We found a 16 % increase in 
+
+retention of new C in the recent + historical plots compared with the control. This would lead to an 
+
+additional soil C sink potential of 0.01-0.1 Pg C (i.e. 16 % of 0.07 or 0.61 Pg C). 
+
+## Figure Captions 
+
+**Fig. 1 Conceptual diagram of the formation of organo-mineral coatings on catalytic biochar** **surfaces over time in a Rhodic Ferralsol.** Biochar can act as a bio-catalyst to accelerate formation of organo-mineral microaggregates (53-250 μm) and mineral-protected soil organic matter (<53 μm) on its surfaces and induce negative priming of soil organic carbon. Microbes, fungal hyphae and root hairs can further mine minerals within pores via exudation and dissolution. Microbial necromass covered with minerals is then incorporated into the organo-mineral (<250 μm) and organo-organic (<100 nm) interfaces. Following wetting-drying and plant growth cycles, organo mineral and organo-organic aggregates break-off from organic matter because of weak bonding. Once these aggregates break off, new mineral layers can form. 
+
+**Fig. 2 Belowground carbon dynamics in the longest continuous biochar field experiment. a,** Changes in total soil organic carbon (SOC, Mg C ha<sup>-1</sup>) in the control and biochar-amended soils over 9.5 years (n=3, LSD = 1.1). Total SOC was measured in the 0-100 soil layer on an equivalent mass basis using Dumas combustion. **b,** Rhizosphere priming as difference in cumulative SOC mineralization between planted and unplanted “recent” biochar amended soil or soil with the “recent + historical” biochar. “Recent” biochar is biochar applied to a part of the control plots 8.2 years after the trial was established (closed triangles). The “recent + historical” amendment is biochar applied to a part of the historical biochar plots 8.2 years after the trial was established (open triangles). Confidence intervals (95%) of “recent” biochar and “recent + historical” biochar amendments are plotted in dashed lines and normalized against the mean squares across all treatments at each sampling event (n=3). For biochar amendment, the CI was based on a sensitivity analysis (Online Method Section), which considers the extreme scenarios of contrasting SOC pools (C3 vs. C4 dominated) by differences in δ<sup>13</sup>C signatures. The six arrows represent nitrogen fertilizer additions. 
+
+**Fig. 3 Allocation and retention of rhizodeposits (¹³C-enriched) and three-dimensional elemental distribution in a biochar-amended Ferralsol at 9.5 years. a, 8.2 years after the first application, the biochar-amended soil received a recent dose of biochar at 10 Mg ha-1 to the historical plots (“recent + historical”). The same total amount (190 mg ¹³C m⁻²) was supplied in each treatment plot (n=3). b, Recent biochar (“recent”) was mixed in top 100 mm of soil at 10 Mg ha⁻¹ one year before measurement. c, 3D FIB-SEM-EDS of an intact soil aggregate (30 μm × 25 μm × 24 μm) from the “recent + historical” biochar plots. d, 3D FIB-SEM-EDS of an intact soil aggregate (30 μm × 20 μm ×
+
+<--- Page Split --->
+
+30 \(\mu \mathrm{m}\) ) from the "recent" biochar plots. Soil sampling was conducted before and 15 days after labelling. The total recovery of \(^{13}\mathrm{C}\) labelling is given, including soil + root respiration, root biomass, free- and occluded particulate matter and mineral fractions.  
+
+Fig. 4 Synchrotron- based spectromicroscopic analysis of microaggregates (53- 250 \(\mu \mathrm{m}\) ) and mineral fractions (<53 \(\mu \mathrm{m}\) ) in the unamended control and historical biochar- amended plots with recent biochar addition. a, Average SXR spectra of microaggregates (53- 250 \(\mu \mathrm{m}\) ) and mineral fractions (<53 \(\mu \mathrm{m}\) ) with the "recent + historical" and "recent" biochar amendments (n=9, CV% < 3%). b, Semi- thin (200 nm) sections of free water- stable microaggregates (53- 250 \(\mu \mathrm{m}\) ) and mineral fractions (<53 \(\mu \mathrm{m}\) ) isolated from the Ferralsol with the "recent + historical" and "recent" biochar amendments analysed using synchrotron- based IR- microspectroscopy. Spectral maps showing the distribution of polysaccharide- C (1035 cm \(^{-1}\) ), aromatic- C (1600 cm \(^{-1}\) ), aliphatic- C (2920 cm \(^{-1}\) ), and mineral- OH (3650 cm \(^{-1}\) ) were obtained from 64 co- added scans (4 cm \(^{-1}\) resolution), lateral resolution 5 \(\mu \mathrm{m}\) (bars: 50 \(\mu \mathrm{m}\) ). The signal intensity for each molecular group varied according to the colour scale shown. The images on the left are optical micrographs of the semi- thin sections.  
+
+Fig 5. In situ spectromicroscopic analysis of the organo- mineral coating on biochar surfaces and pores over time. a, Average SXR spectra of field- extracted "recent" (1- yr aged) and "historical" (9.5- yr aged) biochars (n=9, CV% < 3%). b, high magnification secondary electron image of a pore where fungi exist. c, EDS spectrum of the area in b. d, STEM- HAADF image of organo- mineral clusters on the "recent" biochar surface; e, its EELS spectra. f, High resolution image of the surface of the organomineral layer inside the pore of the "historical" biochar. g, EDS spectrum of the area in f. h, HAADF image of a deposit attached to the surface of the "historical" biochar. i, its EELS spectra. j, DOC of bulk soils from the "recent + historical" and "recent" biochar amendments analysed with LCOCD. The hydrophilic fraction is further sub- divided into five categories, i: biopolymer, ii: persistent C, iii: building blocks, iv: low molecular weight acids and v: low molecular weight neutrals.  
+
+## References  
+
+1 Goldstein, A. et al. Protecting irrecoverable carbon in Earth's ecosystems. Nature Climate Change, 1- 9 (2020).  2 Cavicchioli, R. et al. Scientists' warning to humanity: microorganisms and climate change. Nature Reviews Microbiology 17, 569- 586 (2019).  3 Ogle, S. M., Breidt, F. J. & Paustian, K. Agricultural management impacts on soil organic carbon storage under moist and dry climatic conditions of temperate and tropical regions. Biogeochemistry 72, 87- 121 (2005).  4 Minasny, B. et al. Soil carbon 4 per mille. Geoderma 292, 59- 86 (2017).  5 Arneth, A. et al. in Climate Change and Land: an IPCC special report on climate change, desertification, land degradation, sustainable land management, food security, and greenhouse gas fluxes in terrestrial ecosystems 1- 98 (Intergovernmental Panel on Climate Change (IPCC), 2019).  6 Bossio, D. et al. The role of soil carbon in natural climate solutions. Nature Sustainability 3, 391- 398 (2020).  7 Stocking, M. A. Tropical soils and food security: the next 50 years. Science 302, 1356- 1359 (2003).  8 Pachauri, R. K. et al. Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. (2014).
+
+<--- Page Split --->
+
+546 9 Allen, M. et al. Global Warming of \(1.5^{\circ}C\) . Summary for Policymakers. (2018). 547 10 Woolf, D., Amonette, J. E., Street- Perrott, F. A., Lehmann, J. & Joseph, S. Sustainable biochar to mitigate global climate change. Nature Communications 1, 56 (2010). 548 11 Masson- Delmotte, V. et al. Global warming of 1.5 C. An IPCC Special Report on the impacts of global warming of 1 (2018). 549 12 Lal, R. Managing soils and ecosystems for mitigating anthropogenic carbon emissions and advancing global food security. BioScience 60, 708- 721, doi:10.1525/bio.2010.60.9.8 (2010). 550 13 Lal, R. Soil carbon sequestration to mitigate climate change. Geoderma 123, 1- 22, doi:10.1016/j.geoderma.2004.01.032 (2004). 551 14 Lehmann, J., Bossio, D. A., Kögel- Knabner, I. & Rillig, M. C. The concept and future prospects of soil health. Nature Reviews Earth & Environment, 1- 10 (2020). 552 15 Amelung, W. et al. Towards a global- scale soil climate mitigation strategy. Nature Communications 11, 1- 10 (2020). 553 16 Chenu, C. et al. Increasing organic stocks in agricultural soils: Knowledge gaps and potential innovations. Soil and Tillage Research 188, 41- 52 (2019). 554 17 Amundson, R. & Biardeau, L. Opinion: Soil carbon sequestration is an elusive climate mitigation tool. Proceedings of the National Academy of Sciences 115, 11652- 11656 (2018). 555 18 Ji, C., Cheng, K., Nayak, D. & Pan, G. Environmental and economic assessment of crop residue competitive utilization for biochar, briquette fuel and combined heat and power generation. Journal of Cleaner Production 192, 916- 923 (2018). 556 19 Woolf, D., Solomon, D. & Lehmann, J. Land restoration in food security programmes: synergies with climate change mitigation. Climate Policy 18, 1260- 1270 (2018). 557 20 Woolf, D., Lehmann, J. & Lee, D. R. Optimal bioenergy power generation for climate change mitigation with or without carbon sequestration. Nature Communications 7, 1- 11 (2016). 558 21 Woolf, D., Amonette, J. E., Street- Perrott, F. A., Lehmann, J. & Joseph, S. Sustainable biochar to mitigate global climate change. Nature communications 1, 1- 9 (2010). 559 22 Xu, X. et al. Greenhouse gas mitigation potential in crop production with biochar soil amendment—a carbon footprint assessment for cross- site field experiments from China. Gcb Bioenergy 11, 592- 605 (2019). 560 23 Cowie, A. et al. Biochar, carbon accounting and climate change. (En: Biochar for environmental management: science, technology and ..., 2015). 561 24 Torn, M. S., Trumbore, S. E., Chadwick, O. A., Vitousek, P. M. & Hendricks, D. M. Mineral control of soil organic carbon storage and turnover. Nature 389, 170- 173, doi:10.1038/38260 (1997). 562 25 Hemingway, J. D. et al. Mineral protection regulates long- term global preservation of natural organic carbon. Nature 570, 228- 231 (2019). 563 26 Woolf, D. & Lehmann, J. Microbial models with minimal mineral protection can explain long- term soil organic carbon persistence. Scientific reports 9, 1- 8 (2019). 564 27 Kopittke, P. M. et al. Soil organic matter is stabilized by organo- mineral associations through two key processes: The role of the carbon to nitrogen ratio. Geoderma 357, 113974 (2020). 565 28 Lehmann, J. & Kleber, M. The contentious nature of soil organic matter. Nature 528, 60- 68, doi:10.1038/nature16069 (2015). 566 29 Lehmann, J. et al. Persistence of soil organic carbon caused by functional complexity. Nature Geoscience 13, 529- 534 (2020). 567 30 Joseph, S. et al. An investigation into the reactions of biochar in soil. Soil Research 48, 501- 515 (2010). 568 31 Lin, Y., Munroe, P., Joseph, S., Kimber, S. & Van Zwieten, L. Nanoscale organo- mineral reactions of biochars in ferrosol: an investigation using microscopy. Plant and Soil 357, 369- 380, doi:10.1007/s11104- 012- 1169- 8 (2012).
+
+<--- Page Split --->
+
+595 32 Archanjo, B. et al. Nanoscale analyses of the surface structure and composition of biochars extracted from field trials or after co- composting using advanced analytical electron microscopy. Geoderma 294, 70- 79 (2017). 597 33 Nguyen, T. T. N. et al. The effects of short term, long term and reapplication of biochar on soil bacteria. Science of the Total Environment 636, 142- 151 (2018). 600 34 Ding, F. et al. A meta- analysis and critical evaluation of influencing factors on soil carbon priming following biochar amendment. Journal of soils and sediments 18, 1507- 1517 (2018). 602 35 Liang, C., Amelung, W., Lehmann, J., & Kastner, M. Quantitative assessment of microbial necromass contribution to soil organic matter. Global change biology 25, 3578- 3590 (2019). 603 36 Lehmann, J. et al. Biochar effects on soil biota - A review. Soil Biology & Biochemistry 43, 1812- 1836, doi:10.1016/j.soilbio.2011.04.022 (2011). 605 37 Singh, B. P. & Cowie, A. L. Long- term influence of biochar on native organic carbon mineralisation in a low- carbon clayey soil. Scientific Reports 4, doi:10.1038/srep03687 (2014). 608 38 Li, X., Wang, T., Chang, S. X., Jiang, X., & Song, Y. Biochar increases soil microbial biomass but has variable effects on microbial diversity: A meta- analysis. Science of The Total Environment 749, 141593 (2020). 611 39 Husson, O. Redox potential (Eh) and pH as drivers of soil/plant/microorganism systems: a transdisciplinary overview pointing to integrative opportunities for agronomy. Plant and Soil 362, 389- 417 (2013). 614 40 Weng, Z. H. et al. The accumulation of rhizodeposits in organo- mineral fractions promoted biochar- induced negative priming of native soil organic carbon in Ferralsol. Soil Biology and Biochemistry 118, 91- 96 (2018). 617 41 Weng, Z. et al. Biochar built soil carbon over a decade by stabilizing rhizodeposits. Nature Clim. Change 7, 371- 376, doi:10.1038/nclimate3276  
+
+http://www.nature.com/nclimate/journal/v7/n5/abs/nclimate3276. html#supplementary- information (2017). 621 42 Weng, Z. H. et al. Plant- biochar interactions drive the negative priming of soil organic carbon in an annual ryegrass field system. Soil Biology and Biochemistry 90, 111- 121 (2015). 623 43 Slavich, P. et al. Contrasting effects of manure and green waste biochars on the properties of an acidic ferralsol and productivity of a subtropical pasture. Plant and Soil 366, 213- 227 (2013). 626 44 Farquhar, G., Hubick, K., Condon, A. & Richards, R. in Stable isotopes in ecological research 21- 40 (Springer, 1989). 628 45 Graber, E. R. et al. Biochar impact on development and productivity of pepper and tomato grown in fertigated soilless media. Plant and Soil 337, 481- 496, doi:10.1007/s11104- 010- 0544- 6 (2010). 631 46 Kaiser, C., Franklin, O., Richter, A. & Dieckmann, U. Social dynamics within decomposer communities lead to nitrogen retention and organic matter build- up in soils. Nature Communications 6 (2015). 634 47 Lehmann, J. et al. Spatial complexity of soil organic matter forms at nanometre scales. Nature Geoscience 1, 238- 242 (2008). 636 48 Lehmann, J. & Solomon, D. in Developments in soil science Vol. 34 289- 312 (Elsevier, 2010). 637 49 Lybrand, R. A. et al. A coupled microscopy approach to assess the nano- landscape of weathering. Scientific reports 9, 1- 14 (2019). 638 50 Van Hees, P., Jones, D., Jentschke, G. & Godbold, D. Mobilization of aluminium, iron and silicon by Picea abies and ectomycorrhizas in a forest soil. European Journal of Soil Science 55, 101- 112 (2004). 641 51 Kasozi, G. N., Zimmerman, A. R., Nkedi- Kizza, P. & Gao, B. Catechol and humic acid sorption onto a range of laboratory- produced black carbons (biochars). Environmental Science & Technology 44, 6189- 6195 (2010).
+
+<--- Page Split --->
+
+645 52 Rasse, D. P., Rumpel, C. & Dignac, M.- F. Is soil carbon mostly root carbon? Mechanisms for a 646 specific stabilisation. Plant and Soil 269, 341- 356 (2005). 647 53 Violante, A., Barberis, E., Pigna, M. & Boero, V. Factors affecting the formation, nature, and 648 properties of iron precipitation products at the soil- root interface. Journal of Plant Nutrition 649 26, 1889- 1908 (2003). 650 54 Glaser, B., Balashov, E., Haumaier, L., Guggenberger, G. & Zech, W. Black carbon in density 651 fractions of anthropogenic soils of the Brazilian Amazon region. Organic Geochemistry 31, 652 669- 678 (2000). 653 55 Czimczik, C. I. & Masiello, C. A. Controls on black carbon storage in soils. Global 654 Biogeochemical Cycles 21 (2007). 655 56 Possinger, A. R., Zachman, M. J., Enders, A., Levin, B. D., Muller, D. A., Kourkoutis, L. F., & 656 Lehmann, J. Organo- organic and organo- mineral interfaces in soil at the nanometer scale. 657 Nature Communications 11, 1- 11 (2020). 658 57 Huang, P.- M., Wang, M.- K. & Chiu, C.- Y. Soil mineral- organic matter- microbe interactions: 659 impacts on biogeochemical processes and biodiversity in soils. Pedobiologia 49, 609- 635 660 (2005). 661 Keluweit, M. et al. Mineral protection of soil carbon counteracted by root exudates. Nature 662 Climate Change 5, 588- 595 (2015). 663 59 Tang, J. & Riley, W. J. Weaker soil carbon- climate feedbacks resulting from microbial and 664 abiotic interactions. Nature Climate Change 5, 56- 60 (2015). 665 60 Schmidt, M. W. et al. Persistence of soil organic matter as an ecosystem property. Nature 478, 466- 59 (2011).  
+
+## Acknowledgements  
+
+The authors thank the Australian Government, Department of Agriculture and Water Resources for supporting the National Biochar Initiatives (2009- 2012, 2012- 2014) which co- funded this research. We are particularly grateful to Dr. Peter Slavich, as one of the key founders of this long- term field experiment for providing insightful comments on the initial draft. Part of this research was undertaken on the Soft X- ray spectroscopy beamline and the Infrared microscopy beamline at the Australian Synchrotron, part of ANSTO (grant numbers AS1_SXR_15754 and AS1_IRM_15940). We thank the beamline scientists, Drs Bruce Cowie and Lars Thomsen, for their technical support on the soft x ray analysis and Drs Mark Tobin, Annaleise Klein and Jitraporn (Pimm) Vongsvivut, for their technical support on the infrared microscopy analysis. Part of the research is funded by La Trobe University's Research Focus Area in Securing Food, Water and the Environment (Grant Ready: SFWE RFA 2000004295; Collaboration Ready: SFWE RFA 2000004349). We also appreciate the technical support from Scott Petty and Josh Rust for maintaining this field experiment over the past decade, and laboratory support from Nichole Morris. We also thank Dr Carlos Achete from INMETRO, Brazil and Dr
+
+<--- Page Split --->
+
+Bin Gong from the University of New South Wales, Australia, for performing XPS analysis of biochars and soils, Dr Sarasadat Taherymoosavi from the University of New South Wales, Australia, for technical assistance in LC- OCD analysis. We acknowledge the intellectual contribution from Prof Johannes Lehmann for discussions on the potential mechanisms of biochar- induced stabilization of rhizodeposits.  
+
+## Author contributions  
+
+ZW drafted and wrote the manuscript, experimental design, set- up and conducted experiments, and data collection and analysis; LVZ, BPS and LMM wrote the manuscript, aided in experimental design, critical revision of the article; SJ, ET, BSA and MTR collected and analyzed data, critical revision of the article; TJT, CT, AF, PMK, SK, SM and AC provided critical revision of the article. All authors provided final approval of the revision to be published.  
+
+Correspondence and requests for materials should be addressed to LVZ via email: lukas.van.zwieten@dpi.nsw.gov.au
+
+<--- Page Split --->
+
+## Figures  
+
+![](images/Figure_1.jpg)
+
+<center>Figure 1 </center>  
+
+Conceptual diagram of the formation of organo- mineral coatings on catalytic biochar surfaces over time in a Rhodic Ferralsol. Biochar can act as a bio- catalyst to accelerate formation of organo- mineral microaggregates (53- 250 \(\mu \mathrm{m}\) ) and mineral- protected soil organic matter ( \(< 53 \mu \mathrm{m}\) ) on its surfaces and induce negative priming of soil organic carbon. Microbes, fungal hyphae and root hairs can further mine minerals within pores via exudation and dissolution. Microbial necromass covered with minerals is then incorporated into the organo- mineral ( \(< 250 \mu \mathrm{m}\) ) and organo- organic ( \(< 100 \mathrm{nm}\) ) interfaces. Following wetting- drying and plant growth cycles, organo mineral and organo- organic aggregates break- off from organic matter because of weak bonding. Once these aggregates break off, new mineral layers can form.
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+- Videoabstract3DFIBSEMEDXofrhizodepositsretentioninaggregate.mp4- SupplementaryInformation.pdf
+
+<--- Page Split --->

preprint/preprint__0f38253e04174167d6d03bbdbbc7f1a3288595a100abf195116c81c963341bb8/preprint__0f38253e04174167d6d03bbdbbc7f1a3288595a100abf195116c81c963341bb8_det.mmd

@@ -0,0 +1,526 @@
+<|ref|>title<|/ref|><|det|>[[44, 108, 896, 177]]<|/det|>
+# Microspectroscopic visualization of how biochar elevates the soil organic carbon ceiling  
+
+<|ref|>text<|/ref|><|det|>[[44, 196, 753, 285]]<|/det|>
+Zhe (Han) Weng  University of Queensland  Lukas Van Zwieten ( \(\square\) lukas.van.zwieten@dpi.nsw.gov.au )  NSW Department of Primary Industries  https://orcid.org/0000- 0002- 8832- 360X  
+
+<|ref|>text<|/ref|><|det|>[[44, 290, 394, 330]]<|/det|>
+Michael Rose  NSW Department of Primary Industries  
+
+<|ref|>text<|/ref|><|det|>[[44, 336, 393, 376]]<|/det|>
+Bhupinder Pal Singh  NSW Department of Primary Industries  
+
+<|ref|>text<|/ref|><|det|>[[44, 382, 393, 422]]<|/det|>
+Ehsan Tavakkoli  NSW Department of Primary Industries  
+
+<|ref|>text<|/ref|><|det|>[[44, 428, 325, 468]]<|/det|>
+Stephen Joseph  University of New South Wales  
+
+<|ref|>text<|/ref|><|det|>[[44, 474, 202, 513]]<|/det|>
+Lynne Macdonald  CSIRO  
+
+<|ref|>text<|/ref|><|det|>[[44, 520, 393, 560]]<|/det|>
+Stephen Kimber  NSW Department of Primary Industries  
+
+<|ref|>text<|/ref|><|det|>[[44, 566, 393, 606]]<|/det|>
+Stephen Morris  NSW Department of Primary Industries  
+
+<|ref|>text<|/ref|><|det|>[[44, 612, 281, 652]]<|/det|>
+Terry James Rose  Southern Cross University  
+
+<|ref|>text<|/ref|><|det|>[[44, 658, 548, 699]]<|/det|>
+Bräulio Archanjo  Instituto Nacional de Metrologia, Qualidade e Tecnologia  
+
+<|ref|>text<|/ref|><|det|>[[44, 705, 225, 744]]<|/det|>
+Caixian Tang  La Trobe University  
+
+<|ref|>text<|/ref|><|det|>[[44, 751, 582, 791]]<|/det|>
+Ashley Franks  La Trobe University  https://orcid.org/0000- 0003- 1664- 6060  
+
+<|ref|>text<|/ref|><|det|>[[44, 797, 240, 837]]<|/det|>
+Hui Diao  The University of Qld  
+
+<|ref|>text<|/ref|><|det|>[[44, 844, 670, 884]]<|/det|>
+Peter Kopittke  The University of Queensland  https://orcid.org/0000- 0003- 4948- 1880  
+
+<|ref|>text<|/ref|><|det|>[[44, 890, 168, 907]]<|/det|>
+Annette Cowie  
+
+<|ref|>text<|/ref|><|det|>[[44, 911, 950, 951]]<|/det|>
+NSW Department of Primary Industries / University of New England  https://orcid.org/0000- 0002- 3858- 959X
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[44, 80, 102, 98]]<|/det|>
+## Article  
+
+<|ref|>text<|/ref|><|det|>[[44, 117, 485, 137]]<|/det|>
+Keywords: soil carbon, soil organic carbon, biochar  
+
+<|ref|>text<|/ref|><|det|>[[44, 155, 350, 175]]<|/det|>
+Posted Date: September 10th, 2021  
+
+<|ref|>text<|/ref|><|det|>[[44, 194, 463, 213]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 860309/v1  
+
+<|ref|>text<|/ref|><|det|>[[42, 231, 910, 273]]<|/det|>
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+<|ref|>text<|/ref|><|det|>[[42, 308, 955, 351]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on September 2nd, 2022. See the published version at https://doi.org/10.1038/s41467- 022- 32819- 7.
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[118, 85, 879, 131]]<|/det|>
+# Microspectroscopic visualization of how biochar elevates the soil organic carbon ceiling  
+
+<|ref|>text<|/ref|><|det|>[[118, 203, 881, 271]]<|/det|>
+Zhe (Han) Weng<sup>1,2,3,4</sup>, Lukas Van Zwieten<sup>1,5\*</sup>, Michael T. Rose<sup>1</sup>, Bhupinder Pal Singh<sup>6</sup>, Ehsan Tavakkoli<sup>7</sup>, Stephen Joseph<sup>8</sup>, Lynne M. Macdonald<sup>9</sup>, Stephen Kimber<sup>1</sup>, Stephen Morris<sup>1</sup>, Terry J. Rose<sup>5</sup>, Braulio S. Archanjo<sup>10</sup>, Caixian Tang<sup>3</sup>, Ashley Franks<sup>11,12</sup>, Hui Diao<sup>13</sup>, Peter M. Kopittke<sup>4</sup>, Annette Cowie<sup>2,14</sup>  
+
+<|ref|>text<|/ref|><|det|>[[115, 350, 884, 710]]<|/det|>
+<sup>1</sup>NSW Department of Primary Industries, Wollongbar Primary Industries Institute, Wollongbar, NSW 2477, Australia  <sup>2</sup>School of Environmental and Rural Sciences, University of New England, Armidale, NSW 2351, Australia  <sup>3</sup>Department of Animal, Plant & Soil Sciences, Centre for AgriBioscience, La Trobe University, Melbourne, Vic 3086, Australia  <sup>4</sup>School of Agriculture and Food Sciences, The University of Queensland, St. Lucia, Queensland 4072, Australia  <sup>5</sup>Southern Cross University, East Lismore, NSW 2480, Australia  <sup>6</sup>NSW Department of Primary Industries, Elizabeth Macarthur Agricultural Institute, Woodbridge Rd, Menangle, NSW 2568, Australia  <sup>7</sup>NSW Department of Primary Industries, Wagga Wagga Agriculture Institute, Wagga Wagga, NSW 2650, Australia  <sup>8</sup>University of New South Wales, Sydney, NSW 2052, Australia  <sup>9</sup>CSIRO Agriculture & Food, Waite campus, Glen Osmond, SA 5064, Australia  <sup>10</sup>Divisão de Metrologia de Materiais - DIMAT, Instituto Nacional de Metrologia, Normalização e Qualidade Industrial - INMETRO, Duque de Caxias, RJ, 25250-020, Brazil  <sup>11</sup>Department of Physiology, Anatomy and Microbiology, La Trobe University, Melbourne, Vic 3086, Australia  <sup>12</sup>Centre for Future Landscapes, La Trobe University, Melbourne, Vic 3086, Australia  <sup>13</sup>Centre for Microscopy and Microanalysis, The University of Queensland, QLD, 4072, Australia  <sup>14</sup>NSW Department of Primary Industries/ University of New England, Armidale, NSW 2351, Australia  \*e- mail: lukas.van.zwieten@dpi.nsw.gov.au
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 82, 883, 455]]<|/det|>
+1 The soil carbon saturation concept suggests an upper limit to store soil organic carbon (SOC), set by the mechanisms that protect soil organic matter from decomposition. Biochar has the capacity to protect new C including rhizodeposits and microbial necromass. However, the decadal scale mechanisms by which biochar influences the molecular diversity, spatial heterogeneity, and temporal changes of SOC persistence remain unresolved. Here we show that the soil C saturation ceiling of a Ferralsol under subtropical pasture could be elevated by \(2\mathrm{Mg}\) (new) C ha \(^{- 1}\) by the application of Eucalyptus saligna biochar 8.2 years after the first application. Using one, two-, and three- dimensional analyses, significant increases were observed in the spatial distribution of root- derived \(^{13}\mathrm{C}\) in microaggregates (53- 250 \(\mu \mathrm{m}\) , 11 %) and new C protected in mineral fractions (<53 \(\mu \mathrm{m}\) , 5 %). Microbial C- use efficiency was concomitantly improved by lowering specific enzyme activities, contributing to the decreased mineralization of native SOC by 18 %. We provide evidence that the global SOC ceiling can be elevated using biochar in Ferralsols by 0.01- 0.1 \(\mathrm{Pg}\) new C yr \(^{- 1}\) .  
+
+<|ref|>sub_title<|/ref|><|det|>[[70, 481, 161, 496]]<|/det|>
+## 13 Main  
+
+<|ref|>text<|/ref|><|det|>[[70, 525, 883, 736]]<|/det|>
+14 Human activities risk releasing 260 \(\mathrm{Pg}\) of carbon (C) as carbon dioxide ( \(\mathrm{CO_2}\) ) globally that is irrecoverable on a timescale relevant to avoiding profound climate impacts \(^{1,2}\) . Agricultural soils contribute an average of \(2\mathrm{MgC}\) lost ha \(^{- 1}\) yr \(^{- 1}\) globally \(^{3 - 5}\) . It has been estimated that 122 \(\mathrm{Mg}\) soil organic C (SOC) ha \(^{- 1}\) to 1 m depth has been lost over 1 Mha of land converted to tropical grasslands \(^{6}\) , with 40 % of this area occurring on Ferralsols \(^{7}\) . To meet the Paris Agreement of limiting global warming to below \(2^{\circ}\mathrm{C}\) , the Intergovernmental Panel on Climate Change has shown that \(\mathrm{CO_2}\) removal (CDR) techniques are urgently needed \(^{8,9}\) .  
+
+<|ref|>text<|/ref|><|det|>[[70, 762, 883, 877]]<|/det|>
+21 Soil C management \(^{4 - 6}\) and the application of biochar \(^{10}\) are appealing \(\mathrm{CDR}s^{9,11}\) as they also improve soil health, sustain agricultural productivity \(^{12,13}\) , and increase resilience of ecosystem services \(^{14,15}\) . Protecting and rebuilding soil C could draw down \(5.5\mathrm{PgCO_2yr^{- 1}}\) , representing 25 % of the potential of natural climate solutions to deliver CDR through conservation, restoration, and improved land
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[67, 83, 883, 135]]<|/det|>
+25 management practices<sup>6</sup>. However, there are biophysical and socio-economic barriers to CDR with SOC management<sup>4,6,16,17</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[67, 160, 883, 275]]<|/det|>
+27 Biochar is recognized as a CDR because of its persistence<sup>9,11</sup> in the environment. The pyrolysis of biomass can deliver bioenergy outcomes, as well as agronomic and non- \(\text{CO}_2\) greenhouse gas benefits through use of biochar as a soil amendment<sup>18- 22</sup>. Biochar systems generally show life cycle climate change impacts of emissions reduction in the range of 0.4 - 1.2 Mg \(\text{CO}_2\text{e Mg}^{- 1}\) dry feedstock<sup>23</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[66, 298, 884, 900]]<|/det|>
+31 Organo- mineral interactions can increase SOC persistence, in some cases over a millennial timescale<sup>24,25</sup>. There is a general understanding of the effect of microbial activity<sup>26</sup> and mineral protection<sup>27</sup> on SOC storage. However, there are knowledge gaps on the contribution of molecular diversity of organic compounds, fine- scale spatial heterogeneity, and temporal variability in soil conditions. Such composition- space- time interactions influence the accessibility of decomposer communities to the substrate<sup>28,29</sup>. Here, we propose a mechanism by which biochar acts as a biocatalyst to accelerate the formation of organo- mineral and organo- organic interfaces in microaggregates (53- 250 \(\mu \text{m}\) ) and mineral protection of SOC (Fig. 1). Biochar can sorb root- derived C (rhizodeposits) and form biofilms on its surfaces. The very fine layer of soil minerals that subsequently builds on the surfaces of biochar as it ages in soil<sup>30- 32</sup> protects rhizodeposits from microbial metabolism<sup>33,34</sup>, and at the same time incorporates microbial necromass<sup>35- 38</sup>. This coating can desorb from the surface during aggregate turnover or in response to a change in soil conditions such as pH, redox and moisture<sup>39</sup>. The rhizodeposits and microbial necromass are then captured in microaggregates<sup>35,40,41</sup> (e.g. <250 \(\mu \text{m}\) ). A new coating can then form in its place. These processes repeat, building rhizodeposits in soil over time (Fig. 1). We examine these processes in detail to quantify the potential of biochar to elevate the SOC storage ceiling. To do this, we applied Eucalyptus saligna biochar (550°C) to a historic field site established in 2006<sup>41</sup> (Ferralsol under managed subtropical pasture). The mechanisms (Fig. 1) that we tested included the negative priming via higher microbial C- use efficiency and restricted access to substrates, and enhanced mineral protection via
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 84, 881, 164]]<|/det|>
+catalytic biochar surfaces. We demonstrate the importance of fine- scale spatial heterogeneity and temporal variability of diverse C functional groups in association with mineral fractions for building and protecting rhizodeposits over a decade.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 195, 356, 211]]<|/det|>
+## Elevating SOC storage capacity  
+
+<|ref|>text<|/ref|><|det|>[[115, 240, 883, 514]]<|/det|>
+We hypothesize that biochar enhances the protective mechanisms for soil organic matter (SOM), and that a greater C storage capacity can therefore be obtained through strategic applications of biochar. The field site was converted to managed pasture from subtropical forest 100 years ago. This led to a loss of \(17\%\) of the original soil C stock compared to the adjacent native rainforest (data not shown). To quantify elevated C storage capacity, we measured soil C stocks in the managed pasture over 9.5 years (Table S1) from four treatments: (1) biochar applied to a part of the historical biochar plots 8.2 years after the trial was established ("recent + historical"); (2) biochar applied to a part of the control plots 8.2 years after the trial was established ("recent"); (3) biochar applied 8.2 years previously ("historical"); and (4) nil biochar plots ("control").  
+
+<|ref|>text<|/ref|><|det|>[[115, 540, 883, 911]]<|/det|>
+All field plots were managed via annual fertilizer application at the start of winter, coinciding with over sowing by annual ryegrass (Methods). The total soil C stock in the unamended pasture soil (control) did not change over 9.5 years<sup>41,42,43</sup> (Fig. 2a; \(P > 0.05\) ) and remained at \(35 \text{Mg C ha}^{- 1}\) in the 0- 100 mm layer when sampled at 8.2 and 9.5 years after the field trial was set- up. The original application of Eucalyptus saligna biochar (550°C) in 2006 resulted in a rapid increase in soil C to \(40 \text{Mg C ha}^{- 1}\) (10 Mg biochar ha<sup>- 1</sup>, 76% C, 7.6 Mg biochar- C) and SOC continued to increase, plateauing at \(50 \text{Mg C ha}^{- 1}\) at 8.2 years ("historical"). When recent biochar amendment was incorporated into the plots at the same dose after 8.2 years following the original application ("recent + historical"), the C stock immediately increased from 50 to \(56 \text{Mg C ha}^{- 1}\) (Fig. 2a) because of the C added from biochar. The C stock then continued to increase further to \(58 \text{Mg C ha}^{- 1}\) between 8.2 and 9.5 years. This additional \(2 \text{Mg C ha}^{- 1}\) was accumulated because of the increased protection mechanisms for new C provided by the biochar, thus elevating the C storage ceiling (Fig. 2a).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 84, 883, 390]]<|/det|>
+The additional 2 Mg new C ha \(^{- 1}\) (Fig. 2a) can be explained by negative priming of SOC mineralization (Fig. 2b). Comparing the net cumulative SOC mineralization where there was a recent application of biochar to the control (defined as priming in this study), the recent biochar amendment to the historical plots ("recent + historical") lowered SOC mineralization (i.e. negative priming) by 89 g \(\mathrm{CO_2}\) - C \(\mathrm{m^{- 2}}\) compared with the recent biochar amendment to the control ("recent") which lowered SOC mineralization by 55 g \(\mathrm{CO_2}\) - C \(\mathrm{m^{- 2}}\) ( \(P< 0.05\) , Fig. 2b). Recent biochar initiated a small (3 g \(\mathrm{CO_2}\) - C \(\mathrm{m^{- 2}}\) ) positive priming effect, whilst recent + historical biochar immediately triggered negative priming (Fig. 2b). Neither recent + historical or recent had an impact on total respiration or root respiration compared to the control ( \(P > 0.05\) , Figs. S1 & S2). As a portion of the total \(\mathrm{CO_2}\) flux, root respiration remained relatively consistent (30- 32 %) and was unaffected by treatments ( \(P > 0.05\) , Table S2).  
+
+<|ref|>text<|/ref|><|det|>[[111, 415, 883, 850]]<|/det|>
+The additional 2 Mg C ha \(^{- 1}\) that accumulated between 8.2 and 9.5 years because of the recent application of biochar to the historical biochar plots ("recent + historical") accrued through the stabilization of rhizodeposits and microbial necromass (Fig. 3). Recent + historical biochar had a similar proportion of total recovered \(^{13}\mathrm{C}\) (58 ± 5.7 %, Fig. 3a) compared to the historical biochar plots ("historical") (60 ± 9.8 %), with this being around 18 % greater than the control (42 ± 7.3 %) and the recent biochar amendment ("recent") (45 ± 4.5 %; Fig. 3b) after the pulse-labelling event at 9.5 years ( \(P< 0.05\) ; Table S3). The increase in belowground \(^{13}\mathrm{C}\) recovery could be largely explained by an increase in mineral- protected soil organic matter (M- SOM) associated \(^{13}\mathrm{C}\) (14 %, \(P< 0.05\) , Table S4). Initially, recent + historical biochar nearly doubled the \(^{13}\mathrm{C}\) retention in the occluded particulate organic matter (O- POM) fractions (5 mg \(^{13}\mathrm{C} \mathrm{m}^{- 2}\) ) of microaggregates (< 250 μm) and M- SOM fractions (14 mg \(^{13}\mathrm{C} \mathrm{m}^{- 2}\) ) of macroaggregates (250- 2000 μm) at 8.9 years compared to the recent biochar (Figs. S3a & b). The root- derived \(^{13}\mathrm{C}\) from rhizodeposition was accumulated gradually into O- POM in macroaggregates and M- SOM at 9.2 years (Figs. S3c & d), which was in turn transformed into M- SOM fractions in micro- and macroaggregates by 9.5 years (Figs. S3 e & f).  
+
+<|ref|>text<|/ref|><|det|>[[68, 877, 654, 895]]<|/det|>
+Microbial contribution and responses to stabilization of rhizodeposits
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 80, 883, 680]]<|/det|>
+To determine the microbial contribution to the increased SOC storage capacity, we quantified catabolic enzyme activities, metabolic quotient of native SOC (bulk soil) and rhizodeposition ( \(^{13}\mathrm{C}\) content), and specific enzyme activity (the ratio of enzyme activity- to- microbial biomass) comparing a recent biochar amendment to the control ("recent") and to the historical biochar plots ("recent + historical"). Microbial biomass was increased by \(11\%\) in the recent + historical biochar compared to the recent biochar between 8.9 and 9.5 years (Table S5a). This might result from the stimulation of microbial co- metabolism \(^{31}\) of biochar- C, root- C and SOC which induced initial positive priming in the recent biochar amendment (Fig. 2b). The recent + historical biochar had increased substrate- induced respiration for citric, oxalic and malic acids compared to the recent biochar (Microresp, Fig. S4; Table S6). No differences were detected for 12 other substrates. This greater respiration induced by carboxylic acids (e.g. root exudates) may partially explain the higher metabolic quotient associated with bulk SOC and rhizodeposition in the recent biochar cf. the recent + historical biochar (Table S5b & c). The recent + historical biochar might result in higher substrate- use efficiency which supports an earlier establishment of negative priming compared to the recent biochar (Fig. 2b). It was previously shown that the recent biochar significantly increased bacterial diversity and the relative abundance of nitrifiers and bacteria consuming biochar C after one year, but the soil bacterial communities from the recent + historical plots did not differ from the control \(^{33}\) . This suggests that the microbial accessibility to SOM might be limited in the recent + historical biochar plots, whereas in the recent biochar, the soil microorganisms had to cope with changes in C- substrate type and availability \(^{45}\) .  
+
+<|ref|>text<|/ref|><|det|>[[115, 703, 883, 915]]<|/det|>
+The ratio of enzyme activities to total microbial biomass was lower in both recent + historical biochar and recent biochar compared to the control (Table S7) despite no difference in enzyme activities (Table S8). This suggests that for a given amount of microbial biomass, less enzymes were produced in the biochar- amended soil. A low ratio of extracellular enzymes to microbial biomass can slow down the degradation of native SOC \(^{46}\) . This is consistent with increased microbial C- use efficiency (Tables S4b & c), which may indirectly contribute to the negative priming. It has been suggested that the presence of opportunistic microbes that meet their energy and nutrient demands by exploiting the
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 84, 880, 133]]<|/det|>
+catalytic activities of decomposers could lower the specific enzyme activity<sup>46</sup>. The spatial arrangement between microbes and substrates is critical to this process.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 162, 513, 179]]<|/det|>
+## Spatial examination of organo-mineral interactions  
+
+<|ref|>text<|/ref|><|det|>[[115, 206, 883, 450]]<|/det|>
+To better understand the process of negative priming following biochar application, we examined whether the stabilization of rhizodeposits may be facilitated via protection with the abundant Fe and Al in soil. Root exudates can interact with Fe oxides in the soil to promote the formation organomineral complexes. Aluminum may also protect root- C from biodegradation. The development of organo- mineral complexes was assessed using three- dimensional focused ion beam (FIB) coupled with scanning electron microscopy (SEM) with energy dispersive X- ray spectroscopy (3D- FIB- SEM- EDS) on intact representative soil aggregates from the recent + historical biochar (Fig. 3c) where C retention co- located with clay minerals, but, not in the recent biochar (Fig. 3d).  
+
+<|ref|>text<|/ref|><|det|>[[115, 476, 883, 910]]<|/det|>
+To provide a deeper insight into the molecular diversity of organic compounds and the temporal and lateral arrangement with respect to organo- mineral interfaces, we conducted in situ spectromicroscopic analysis of free water- stable microaggregates (53- 250 μm) and organo- mineral fractions (<53 μm) of the recent + historical biochar and recent biochar, and of field- extracted biochars at micro- and nanoscale. The C functional groups, examined using synchrotron- based soft X- ray (SXR) analyses, from the microaggregates (53- 250 μm) were dominated by quinones (284.1 eV), aromatic C (285.2 eV, 1s- π\* transitions of conjugated C=C), and aliphatic C (287.3 eV) (Fig. 4a). For the mineral- protected fractions (<53 μm), two prominent features were the low intensity of quinones and high intensity of aliphatic C (Fig. 4a). The dominant peaks of aliphatic, amide and carboxylic C (287- 289 eV) are the direct consequence of deposition of microbial metabolites or debris, exopolysaccharides, and root exudates onto mineral surfaces<sup>27,47,48</sup>. These align with the micro- spatial maps produced from the synchrotron- based infrared (IR) microspectroscopy (Fig. 4b). The close correlation between clay minerals and microbial metabolites on the biochar surface supports SOC stabilization, and highlights the importance of clay minerals for the protection of SOC and the control of microbial activity.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 80, 884, 745]]<|/det|>
+The correlation between C forms and clay minerals, confirmed by the IR maps, was further examined on a nanometer scale by directly analyzing the chemical composition of the organo- mineral coatings at the surface and in the pores of field- extracted biochars. Greater intensities of quinones (284.1 eV) and carboxyl C- OOH (288.6 eV) were observed in the 9.5- year aged biochar compared with the 1- year aged biochar (Fig. 5a). A high magnification image of the area where the fungi were located inside a biochar fragment shows a high concentration of irregular pores and a coating of organic material (Figs. 5b & c). Fungi can mine nutrients from minerals by exuding acids<sup>49,50</sup> which may cause the observed microporosity of organo- mineral- biochar interfaces (Figs. 5b, d, f, h & Fig. S5). Energy dispersive X- ray spectroscopy (EDS) analysis showed that complex changes had occurred on the surface of the biochar over the one- year period (Figs. 5c, 5g). Positively charged nanoparticulate minerals rich in Al, Si, Ca, P, Fe and other cations were attracted to the surface of the negatively charged areas on the biochar. These positively- charged minerals and elements subsequently attracted negatively- charged organic molecules with detectable concentrations of C=C, C-OH, C-N/C=N, C=O, COOH functional groups, quinone bonds and anions thus initiating a process whereby porous clusters are formed on the biochar surface (Figs. 5e, 5i). Similarly, exudates from plants and microorganisms can be deposited around mineral surfaces on the biochar, and cations and minerals can be attracted to these organic molecules. Recent biochar amendment to historical biochar plots would provide unoccupied surfaces and pores in the soil to increase sorption capacity for root exudates<sup>51</sup>, which would then serve as binding agents to further enhance aggregate formation<sup>52</sup>. As these clusters are built up, they may also be detached from the biochar either through fluctuating redox conditions and interaction with microbes or perturbation caused by soil invertebrates<sup>30</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[111, 767, 883, 914]]<|/det|>
+These biochar micro- sites have a high concentration of free radicals with labile easily- mineralizable organic C and/or inorganics dissolved from the biochar (Table S9). Colloidal biochar particles, leachate, dissolved native OM and rhizodeposits may be further stabilized separately or held together via cation bridging with \(\mathrm{Ca^{2 + }}\) , or with Al and Fe oxyhydroxides<sup>53,54,55</sup> and organo- organic interactions at the nanometer scale<sup>56</sup> (Figs. 5e, 5i). These processes may be encouraged by oxidation of the biochar
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 882, 261]]<|/det|>
+surface as it ages in soil<sup>36,37</sup>. This is supported by our LC- OCD results where dissolved hydrophobic C fractions and building blocks (medium molecular weight) were greater in the recent + historical biochar compared to the recent biochar (Fig. 5j; Table S10). The analysis of the surface of the 9.5- year aged biochar by C- edge EELS and XPS indicated that most of the oxidized C species were formed in the organo- mineral coating. The concentrations of the different functional groups appear to be influenced by the presence of nanophase Fe, Si and Al oxides<sup>36</sup> (Fig. 5i).  
+
+<|ref|>sub_title<|/ref|><|det|>[[116, 290, 495, 306]]<|/det|>
+## Global impact of elevating the soil carbon ceiling  
+
+<|ref|>text<|/ref|><|det|>[[115, 333, 882, 673]]<|/det|>
+Building SOC is a global priority<sup>9</sup>, and our results showed that the SOC storage ceiling can be elevated through single or multiple applications of biochars. We observed a plateau in rhizodeposit accumulation rate over 9.5 years in the historical biochar plots (Fig. 2a; \(\gamma = 4.24\ln (x) + 17.6\) ; \(R^2 = 0.95\) ) which implies that the system was approaching a new (16 % higher) equilibrium for SOC storage, ten years after the initial application. We showed that a strategic application of 10 Mg biochar \(ha^{- 1}\) after 8.2 years raised the SOC storage ceiling by a further 2 Mg C \(ha^{- 1}\) . In summary, this Rhodic Ferralsol under the managed pasture had a C storage capacity of 35 Mg C \(ha^{- 1}\) in the surface soil, which increased to 44 Mg C \(ha^{- 1}\) one year following the application of biochar, which further increased to 50 Mg C \(ha^{- 1}\) after nearly a decade. The C storage ceiling was further raised to 59 Mg C \(ha^{- 1}\) where biochar was applied to the historically amended field plots. Of this C storage, 7.6 Mg C \(ha^{- 1}\) was attributed to the addition of C from biochar, while 2 Mg C \(ha^{- 1}\) was attributed to the stabilization of new C.  
+
+<|ref|>text<|/ref|><|det|>[[115, 700, 882, 910]]<|/det|>
+The long- term stabilization of rhizodeposits by biochar after 9.5 years has significant implications for elevating the SOC ceiling. Plants release \(\sim 50\%\) of photosynthetically fixed C into the soil, which is available for microbial growth<sup>57- 59</sup>. Global grasslands annually contribute 0.04 Pg C to SOC<sup>6</sup>. The retention and stabilization of belowground C by biochar could play an important part in a natural climate solution for tropical grasslands, which occupy 0.7 Gha of land with an estimated global C content of 30 Pg C. Here we showed a 16 % increase in retention of new C in the recent + historical plots compared with the control via the stabilization of root- derived <sup>13</sup>C in microaggregates (53- 250
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 83, 881, 293]]<|/det|>
+\(\mu \mathrm{m}, 11 \%\) ) and microbial products and rhizodeposits protected in mineral fractions (<53 \(\mu \mathrm{m}, 5 \%\) ) (Fig. 3). To demonstrate the potential of raising the C saturation ceiling, we extrapolated this \(16 \%\) increase in stabilization based on the global projected biochar production. We estimated the strategic application of biochar can increase SOC storage capacity in Ferralsols by 0.01- 0.1 \(\mathrm{Pg C yr^{- 1}}\) worldwide (Supplementary information). This mechanism would increase the global mitigation potential using biochar as a soil amendment, estimated at 1.3 \(\mathrm{Pg C yr^{- 1}}\) (central average; Woolf et al. 2010), by another \(0.8 - 7.7 \%\) .  
+
+<|ref|>text<|/ref|><|det|>[[113, 321, 883, 562]]<|/det|>
+In our study, we raised the SOC storage capacity in a subtropical pasture with a strategic application of a Eucalyptus saligna biochar (550°C) 8.2 years following the original biochar application. Of importance to building soil C stocks, the strategic application of biochar in the aged plots resulted in \(16 \%\) more new C (i.e. microbial products and rhizodeposits) being stored as soil C in both microaggregates (53- 250 \(\mu \mathrm{m}\) ) and mineral fractions (<53 \(\mu \mathrm{m}\) ). Microbial C- use efficiency was improved by lowering the specific enzyme activity and slowing down degradation of SOC and rhizodeposits (negative priming). Our in situ spectromicroscopic analyses suggest that the catalytic biochar surfaces accelerated the micro- and nanoscale heterogeneity and temporal variability for new C storage.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 587, 191, 602]]<|/det|>
+## Methods  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 632, 245, 648]]<|/det|>
+## Field site details  
+
+<|ref|>text<|/ref|><|det|>[[113, 678, 883, 888]]<|/det|>
+The field experiment was situated at the Wollongbar Primary Industries Institute (28°49'S, 153°23'E, elevation: 140 m), Wollongbar, New South Wales, Australia. The classification and properties of the soil can be found in Weng et al. (2015). Briefly, the Rhodic Ferralsol is fine- textured and Fe- rich mineral soil dominated by kaolinite, gibbsite and goethite. The 100- mm topsoil had a \(\mathrm{pH_{CaCl2}}\) of 4.5 with total C of \(35 \mathrm{g kg^{- 1}}\) , total Fe \(84 \mathrm{g kg^{- 1}}\) , and total Al \(67 \mathrm{g kg^{- 1}}\) . Details of the initial field site setup in 2006 can be found in Slavich et al. (2013). The treatments (n=3) included (1) Eucalyptus saligna biochar incorporated into the topsoil (0- 100 mm) at \(10 \mathrm{th a^{- 1}}\) (eq. \(1 \% \mathrm{w / w}\) , bulk density of \(1 \mathrm{g cm^{- 3}}\) ) plus NPK
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 883, 295]]<|/det|>
+fertilizer and (2) NPK only (control). The biochar was derived from a single source of above- ground biomass of mature Eucalyptus saligna and pyrolyzed at \(550^{\circ}C\) with a residence time of 30 mins (Pacific Pyrolysis, NSW, Australia). The physicochemical properties of the biochar can be found in Slavich et al. (2013). A tetraploid annual ryegrass (Lolium multiflorum) was broadcast at a seeding rate of \(35kg\) ha \(^{- 1}\) and repeated annually. Urea was applied at \(46kgNha^{- 1}\) on six occasions ( \(276kgha^{- 1}\) in total) between winter and spring each year following manual cuts of the pasture to simulate grazing. Basal nutrients were applied annually at sowing (Slavich et al. 2013).  
+
+<|ref|>text<|/ref|><|det|>[[115, 319, 883, 919]]<|/det|>
+In April 2014, the control (NPK only) and 8.2- year- old biochar (Eucalyptus saligna) plots were superimposed with nine subplots \((0.5m\times 0.5m)\) (Weng et al. 2017). Four treatments were: (1) Recent biochar to historical biochar plots (biochar applied to a part of the historical biochar plots 8.2 years after the trial was established; "recent + historical"); (2) Recent biochar amendment (biochar applied to a part of the control plots 8.2 years after the trial was established; "recent"); (3) Historical biochar amendment (biochar applied 8.2 years previously, "historical"); and (4) Control (nil biochar plots). There was one subplot per field replicate and a total of three field replicates. The biochar added to the control and aged plots was taken from the biochar 'batch' applied in 2006, which had been air- dried and archived in sealed 200 L steel containers at room temperature. The details of the subplot set up and installation of belowground respiration collars can be found in Weng et al. (2015). Bulk density of the biochar was \(0.332gcm^{- 1}\) measured using a method described in Quin et al. (2014). The weight of soil- biochar mixture in the topsoil (100- mm) was determined based on the bulk density assessed in each treatment. Before application, the biochar was sieved to \(< 2mm\) . The soil/biochar mixture was carefully packed into the subplots. The control subplots were also excavated and repacked to a bulk density of \(1gcm^{- 3}\) . A root signature sand collar (50 mm diameter) was installed in each of the control subplots to measure the \(\delta^{13}C\) signature of root respiration. It was packed with acid- washed sand and planted with ryegrass (i.e. a down- sized version of the soil plus root respiration collar). Similarly, a biochar+root signature sand collar was packed with a biochar- sand mixture (1 % w/w) in each of the recent + historical biochar subplots. To maintain the root growth into the collars,
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 84, 881, 164]]<|/det|>
+NPK fertilizers were applied at the same dose as in Slavich et al. (2013). This study employed the same pasture management regime as Slavich et al. (2013), in terms of ryegrass sowing, NPK applications and weed control.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 194, 722, 211]]<|/det|>
+## Periodic \(^{13}\mathrm{C}\) pulse labelling to quantify SOC mineralization and root respiration  
+
+<|ref|>text<|/ref|><|det|>[[115, 240, 882, 354]]<|/det|>
+To understand how plant- biochar- soil interactions affect SOC priming, \(^{13}\mathrm{CO}_{2}\) pulse labelling campaigns were conducted on three occasions: 12 June 2014, 01 August 2014 and 30 July 2015. The procedure of the pulse labelling experiment and a detailed description of quantification of SOC mineralization and root respiration using three- pool C partitioning can be found in Weng et al. (2015, 2018).  
+
+<|ref|>text<|/ref|><|det|>[[115, 380, 882, 622]]<|/det|>
+Soil sampling was carried out at 8.9, 9.2 and 9.5 years following the first biochar application in 2006. Intact soil cores (40 mm in diameter) were sampled to 80 mm depth within each subplot but outside the respiration collar area to reduce disturbance. The sampled areas were avoided in the ensuing sampling events. The sample was mixed evenly and analyzed for pH, total soil organic C, and microbial biomass carbon (MBC). Soil pH was measured on the samples prepared for the enzyme assay (1:5 w/w ratio in distilled water with constant stirring using a vortex) using an IntelliCAL PHC101 pH probe on a Hach HQ40d portable meter (Loveland, Colorado, USA). The analytical procedures for SOC and MBC can be found in Weng et al. 2015. The remaining soil was stored at - 20°C.  
+
+<|ref|>text<|/ref|><|det|>[[115, 650, 882, 891]]<|/det|>
+At the 15- month soil sampling event, fresh soil was also taken from the soil respiration collars (i.e. unplanted) and soil+root respiration collars (i.e. planted) to quantify the effect of plant- biochar- soil interactions on catabolic enzyme activity, substrate- induced respiration, and MBC. The MBC was analyzed using the chloroform fumigation method (Van Zwieten et al. 2010). Metabolic quotient of total C or rhizodeposits was then quantified as the ratio of respiration (native SOC or \(^{13}\mathrm{C}\) - labelled root respiration) over the total MBC. The metabolic quotient has been used as an indicator of C- use efficiency (Fang et al., 2018). Detailed calculations for determining rhizodeposit- derived respiration are given in Weng et al. (2015).
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 85, 470, 101]]<|/det|>
+## SOC priming in the plant-biochar-soil systems  
+
+<|ref|>text<|/ref|><|det|>[[115, 128, 883, 595]]<|/det|>
+The rhizosphere priming of native SOC from the biochar- plant- soil interactions was quantified using a three- pool C partitioning model. Specialized respiration collars were used to isolate soil plus root respiration from shoot respiration<sup>41,42</sup>. Native SOC mineralization was separated using the \(\delta^{13}\mathrm{C}\) signature of biochar plus root (biochar plus root sand collars, Supplementary Information) and total respiration (biochar plus soil plus root collars) after pulse labelling. Moisture content was maintained between 60- 80% field capacity in the root collars to minimize potential C isotopic fractionation during photosynthesis caused by water stress<sup>44</sup>. The \(\delta^{13}\mathrm{C}\) signatures of extracted field- aged biochar and fresh biochar (the same biochar archived in a sealed container for 8.2 years) were both \(- 25 \pm 0.1\%\) . Any interactive effect of biochar and root on the \(\delta^{13}\mathrm{C}\) signature of soil would be surpassed by a greater level of \(\delta^{13}\mathrm{C}\) enrichment of the root component compared with any isotopic signature contribution from soil and biochar to the \(\delta^{13}\mathrm{C}\) signature of the total respiration. A sensitivity analysis of C source partitioning was performed to assess the impact of plant- biochar ( \(C_3\) - dominated)- soil interactions on \(\delta^{13}\mathrm{C}\) signatures of soil (a mixture of \(C_3\) and \(C_4\) pools). Errors generated from isotopic partitioning were propagated using the first order Tyler series approximations of the variances of native SOC mineralization.  
+
+<|ref|>text<|/ref|><|det|>[[116, 622, 881, 705]]<|/det|>
+The recovery of \(^{13}\mathrm{C}\) in various SOC pools at time t (i.e. \(\mathrm{A}^{13}\mathrm{C}_{i,t}\) , in \(\%\) ) was calculated by dividing the amount of \(^{13}\mathrm{C}\) (g m<sup>2</sup>) in a specific C pool (i.e. C<sub>i</sub>) by the initial amount of total added \(^{13}\mathrm{CO}_{2}\) (g m<sup>2</sup>) at each labelling event (i.e. \(^{13}\mathrm{C}_{\mathrm{added}}\) ):  
+
+<|ref|>equation<|/ref|><|det|>[[116, 732, 401, 751]]<|/det|>
+\[\mathrm{A}^{13}\mathrm{C}_{i,t} = (^{13}\mathrm{C}_{\mathrm{excess},t}\times \mathrm{C}_{i}) / ^{13}\mathrm{C}_{\mathrm{added}}\times 100\]  
+
+<|ref|>text<|/ref|><|det|>[[116, 778, 881, 859]]<|/det|>
+where; represents soil plus root respiration, root biomass, soil aggregates or its associated fractions, \(^{13}\mathrm{C}_{\mathrm{excess},t}\) indicates the increment of the \(^{13}\mathrm{C}\) atom \(\%\) of an individual C pool from its natural abundance level at a specific sampling time, t.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[60, 82, 864, 135]]<|/det|>
+299 The mineralization of native SOC \((C_{S})\) was calculated using the \(^{13}\mathrm{C}\) signature of biochar+root \((^{13}\mathrm{C_{B + R}}\) , sand collar) from the plant- biochar- soil systems after pulse labelling:  
+
+<|ref|>equation<|/ref|><|det|>[[115, 160, 868, 181]]<|/det|>
+\[C_{S}(\%) = 100^{*}(\delta^{13}C_{T} - \delta^{13}C_{B + R}) / (\delta^{13}C_{S} - \delta^{13}C_{B + R}) \quad (1)\]  
+
+<|ref|>text<|/ref|><|det|>[[115, 206, 870, 290]]<|/det|>
+where \(\delta^{13}C_{T}\) : \(\delta^{13}C\) signature of the total respiration from the planted system after pulse labelling; \(\delta^{13}C_{S}\) : \(\delta^{13}C\) signature of the soil- derived \(\mathrm{CO_2}\) - C evolved from the unplanted control soil without pulse labelling.  
+
+<|ref|>text<|/ref|><|det|>[[115, 317, 857, 367]]<|/det|>
+The percentage of soil- derived \(\mathrm{CO_2}\) - C in the total respiration from the planted control soil \((C_{S}(\%)\) was determined (Weng et al. 2015):  
+
+<|ref|>equation<|/ref|><|det|>[[115, 393, 870, 414]]<|/det|>
+\[C_{S}(\%) = 100^{*}(\delta^{13}C_{T} - \delta^{13}C_{R}) / (\delta^{13}C_{S} - \delta^{13}C_{R}) \quad (2)\]  
+
+<|ref|>text<|/ref|><|det|>[[115, 439, 810, 480]]<|/det|>
+where \(\delta^{13}C_{T}\) : \(\delta^{13}C\) signature of the total respiration from the planted control; \(\delta^{13}C_{S}\) : the \(\delta^{13}C\)  
+
+<|ref|>text<|/ref|><|det|>[[115, 475, 844, 525]]<|/det|>
+signature of the unplanted control soil; \(\delta^{13}C_{R}\) : the \(\delta^{13}C\) signature of roots, which was determined from root respiration from the root sand collar as described in Weng et al. (2017).  
+
+<|ref|>text<|/ref|><|det|>[[115, 551, 860, 601]]<|/det|>
+Similarly, the percentage of soil- derived \(\mathrm{CO_2}\) - C in the total respiration from the unplanted biochar- amended soil \((C_{S^{\prime}}(\%)\) ) was determined:  
+
+<|ref|>equation<|/ref|><|det|>[[115, 627, 870, 649]]<|/det|>
+\[C_{S^{\prime}}(\%) = 100^{*}(\delta^{13}C_{T^{\prime}} - \delta^{13}C_{B}) / (\delta^{13}C_{S} - \delta^{13}C_{B}) \quad (3)\]  
+
+<|ref|>text<|/ref|><|det|>[[115, 674, 860, 787]]<|/det|>
+where \(\delta^{13}C_{T^{\prime}}\) : the \(\delta^{13}C\) signature of the total respiration from the unplanted biochar soil. \(\delta^{13}C_{S}\) : the \(\delta^{13}C\) signature of the unplanted control soil; \(\delta^{13}C_{B}\) : the \(\delta^{13}C\) signature of either fresh (- 25.02 ± 0.13 \(\%\) ) or aged biochar (- 25.04 ± 0.11 \(\%\) ). Biochars were recovered by hand from field soil samples, thoroughly rinsed with distilled water on a 100 μm sieve and oven- dried at 50°C for 24 h.  
+
+<|ref|>text<|/ref|><|det|>[[115, 817, 512, 834]]<|/det|>
+Rhizosphere priming was calculated in two systems:  
+
+<|ref|>text<|/ref|><|det|>[[115, 864, 513, 881]]<|/det|>
+i. Unamended system (Planted vs. Unplanted)
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[55, 84, 856, 131]]<|/det|>
+- SOC planted, unamended: soil C mineralization in the planted control calculated by \(^{13}\mathrm{C}\) -enriched root end-member  
+
+<|ref|>text<|/ref|><|det|>[[55, 161, 686, 179]]<|/det|>
+- SOC unplanted, unamended: soil C mineralization in the unplanted control  
+
+<|ref|>text<|/ref|><|det|>[[147, 208, 446, 224]]<|/det|>
+Rhizosphere priming in the control soil:  
+
+<|ref|>text<|/ref|><|det|>[[147, 254, 592, 271]]<|/det|>
+\(\Delta \mathrm{SOC}_{\mathrm{unamended}} = (\mathrm{SOC}_{\mathrm{planted, unamended}}) - (\mathrm{SOC}_{\mathrm{unplanted, unamended}})\)  
+
+<|ref|>text<|/ref|><|det|>[[140, 300, 553, 317]]<|/det|>
+ii. Biochar-amended system (Planted vs. Unplanted)  
+
+<|ref|>text<|/ref|><|det|>[[147, 346, 833, 364]]<|/det|>
+- SOC planted, amended: soil C mineralization in the planted biochar soil partitioned from \(^{13}\mathrm{C}\) -  
+
+<|ref|>text<|/ref|><|det|>[[147, 379, 435, 395]]<|/det|>
+enriched 'Biochar+Root' end-member  
+
+<|ref|>text<|/ref|><|det|>[[147, 424, 833, 441]]<|/det|>
+- SOC unplanted, amended: soil C mineralization in the unplanted biochar soil partitioned from  
+
+<|ref|>text<|/ref|><|det|>[[147, 457, 317, 473]]<|/det|>
+biochar end members  
+
+<|ref|>text<|/ref|><|det|>[[147, 503, 476, 520]]<|/det|>
+Rhizosphere priming in the biochar system:  
+
+<|ref|>text<|/ref|><|det|>[[147, 550, 555, 566]]<|/det|>
+\(\Delta \mathrm{SOC}_{\mathrm{amended}} = (\mathrm{SOC}_{\mathrm{planted, amended}}) - (\mathrm{SOC}_{\mathrm{unplanted, amended}})\)  
+
+<|ref|>text<|/ref|><|det|>[[55, 594, 844, 612]]<|/det|>
+SOC priming was the difference in native SOC mineralization between the biochar-amended and  
+
+<|ref|>text<|/ref|><|det|>[[55, 628, 217, 644]]<|/det|>
+control soils:  
+
+<|ref|>text<|/ref|><|det|>[[55, 673, 473, 691]]<|/det|>
+\(\Delta \mathrm{SOC} = (\mathrm{C}_{\mathrm{s}}(\%)^{*}\mathrm{C}_{\mathrm{Tplanted}} - \mathrm{C}_{\mathrm{s}'}(\%)^{*}\mathrm{C}_{\mathrm{Tunplanted}}) / 100\)  
+
+<|ref|>text<|/ref|><|det|>[[55, 719, 860, 767]]<|/det|>
+where \(\mathrm{C}_{\mathrm{Tplanted}}\) and \(\mathrm{C}_{\mathrm{Tunplanted}}\) are the total respiration in planted and unplanted systems either with biochar amendment or the control.  
+
+<|ref|>text<|/ref|><|det|>[[55, 796, 328, 813]]<|/det|>
+Calculated \(^{13}\mathrm{C}\) atom \(\% (\%)\) :  
+
+<|ref|>text<|/ref|><|det|>[[55, 842, 860, 860]]<|/det|>
+\(^{13}\mathrm{C}\) atom \(\% = [(\delta^{13}\mathrm{C} + 1000)^{*}R_{\mathrm{PDB}}]^{*}100 / [(\delta^{13}\mathrm{C} + 1000)^{*}R_{\mathrm{PRB}} + 1]\)  
+
+<|ref|>text<|/ref|><|det|>[[55, 889, 285, 905]]<|/det|>
+where \(\mathrm{R}_{\mathrm{PDB}} = 0.01118\)
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 85, 445, 101]]<|/det|>
+## Sensitivity analysis of isotopic partitioning  
+
+<|ref|>text<|/ref|><|det|>[[115, 128, 875, 212]]<|/det|>
+Because of the uncertainty of the direction of biochar- induced priming of soil carbon and/or rhizodeposits, the contribution of biochar on the \(^{13}\mathrm{C}\) endmember of \((\delta^{13}\mathrm{C}_5)\) was assessed. Therefore, three alternative scenarios of three- pool C partitioning were evaluated:  
+
+<|ref|>text<|/ref|><|det|>[[115, 240, 835, 289]]<|/det|>
+1) dominant positive priming of new C from the \(\mathsf{C}_3\) pasture, where \(\delta^{13}\mathrm{C}_5 = -27\%\) (i.e. the upper boundary, grey dashed line, Fig. 2b);  
+
+<|ref|>text<|/ref|><|det|>[[115, 316, 868, 366]]<|/det|>
+2) equal native SOC priming and rhizosphere priming, hence, the same \(^{13}\mathrm{C}\) signatures of soil+root in the biochar and control plots, where \(\delta^{13}\mathrm{C}_5 = \delta^{13}\mathrm{C}_5\) (i.e. solid lines in Fig. 2b);  
+
+<|ref|>text<|/ref|><|det|>[[115, 394, 857, 444]]<|/det|>
+3) dominant positive priming of the native \(\mathsf{C}_4\) -dominant SOC, where \(\delta^{13}\mathrm{C}_{S + R'} = -13\%\) (i.e. the lower boundary, grey dashed line, Fig. 2b).  
+
+<|ref|>text<|/ref|><|det|>[[115, 472, 870, 588]]<|/det|>
+The boundary conditions were calculated from the published \(^{13}\mathrm{C}\) signatures for Scenarios 1 and 3 (Farquhar et al. 1989). The \(95\%\) confidence intervals were the combination of the lowest and highest scenarios (n = 3). First order Tyler series of the variances of the percentage of soil respiration, \(\mathsf{C}_5(\%)\) , were approximated to propagate errors from isotopic partitioning (Derrien et al. 2014).  
+
+<|ref|>equation<|/ref|><|det|>[[115, 614, 864, 633]]<|/det|>
+\[\sigma^2\mathrm{C}_5(\%) = (\sigma^2\delta^{13}\mathrm{C}_T - \sigma^2\delta^{13}\mathrm{C}_5) / (\delta^{13}\mathrm{C}_T - \delta^{13}\mathrm{C}_5)^2 \quad (6)\]  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 660, 505, 677]]<|/det|>
+## Enzyme activity and substrate-induced respiration  
+
+<|ref|>text<|/ref|><|det|>[[115, 704, 883, 914]]<|/det|>
+The determination of catabolic enzyme activities using a soil suspension method is described in Weng et al. (2017). Six treatments were derived from the control, the recent + historical biochar and recent biochar plots in both the unplanted (i.e. soil respiration collar) and planted (i.e. soil+ root respiration collar) systems (Table S5, Weng et al., 2017). After 7- d incubation at \(40\%\) water- holding capacity (WHC), the activities of four C- degrading enzymes: \(\beta\) - glucosidase, xylosidase, cellulase, and N- acetyl- glucosaminidase, in the soils were analysed using a fluorescent microplate reader (BMG labtech FLUOstar Omega). Specific C enzyme activity was obtained by dividing the activity of individual
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 883, 325]]<|/det|>
+enzymes over the total MBC at each sampling time. These ratios provided an indication of the C- turnover efficiency of the soil microbial community (Kaiser et al., 2015). Substrate- induced respiration was used to measure Community level physiological profiles using the MicroResp™ method (Campbell et al., 2003) with minor modifications. Fresh soil samples, packed in 96- deepwell plates (around 0.5 g per well), were prepared in the same manner as the enzyme experiment (i.e. incubation conditions). Each treatment per field replicate was sub- replicated eight times for measurement. The experimental protocol is detailed in Weng et al. (2017). Fifteen C substrates (Table S6) were selected to represent a broad range of soil and root exudates (Campbell et al. 2003; Chapman et al. 2007).  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 353, 430, 370]]<|/det|>
+## Aggregate size and density fractionation  
+
+<|ref|>text<|/ref|><|det|>[[115, 399, 882, 545]]<|/det|>
+Aggregate size (dry sieving) and density fractionation was conducted based on the method described by Weng et al. (2018). No large macroaggregates ( \(>2000 \mu m\) ) was found in this study. Macroaggregates (250- 2000 \(\mu m\) ) and microaggregates ( \(< 250 \mu m\) ) were fractioned into free POM (F- POM, \(\rho < 1.6 \text{kg m}^{- 3}\) ), occluded POM (O- POM, \(>53 \mu m\) ), and mineral- protected soil organic matter (M- SOM, combining silt- and clay- bound SOM, \(< 53 \mu m\) ).  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 574, 296, 590]]<|/det|>
+## Belowground \(^{13}\mathrm{C}\) pools  
+
+<|ref|>text<|/ref|><|det|>[[115, 619, 881, 701]]<|/det|>
+The C and N content, and \(\delta^{13}\mathrm{C}\) signatures of bulk soil, aggregates and fractions were measured using a PDZ Europa ANCA- GSL elemental analyzer interfaced to a PDZ Europa 20- 20 isotope ratio mass spectrometer (Sercon Ltd., Cheshire, UK) (Weng et al., 2015).  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 730, 247, 745]]<|/det|>
+## 3D-FIB-SEM-EDS  
+
+<|ref|>text<|/ref|><|det|>[[115, 775, 882, 888]]<|/det|>
+The serial section and EDS Mapping of the soil particle was prepared in a FEI SCIOS focused ion beam/scanning electron microscope (FIB/SEM) DualBeam system. The SCIOS FIB/SEM DualBeam system has a vertical mounted SEM column and an ion column sitting at an angle of 52 degrees with respect to the electron column. The particle was located with the aid of the electron beam. Before
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 882, 262]]<|/det|>
+any milling, one micrometre thick platinum layer was deposited on the sample surface covering the area of interest to prevent it from damage caused by the ion bombardment in the following steps. The smooth finish of the Pt layer would also help to reduce the curtaining effect during the following milling procedure. The serial sectioning of the volume was carried out at 3nA and 30kV ion beam current and the EDS mapping were collected at 5kV and 6.4nA electron beam current. The voxel size of the SEM images is \(84 \text{nm} (\times) \times 84 \text{nm} (\times) \times 1000 \text{nm} (\text{z}, \text{slicing thickness}).\)  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 290, 294, 306]]<|/det|>
+## Synchrotron soft X-ray  
+
+<|ref|>text<|/ref|><|det|>[[115, 335, 882, 483]]<|/det|>
+Synchrotron-based soft X- ray (SXR) analysis was performed at the SXR Spectroscopy beamline (14ID) at the Australian Synchrotron on the microaggregate (53- 250 μm) and mineral fractions (<53 μm) from 1) recent biochar- amended plots, and 2) the historically biochar- amended plots; and then biochar recovered from the soil, that is: i) 1- year (aged) and ii) 9.5- year (aged) biochar. The samples were ground to fine powder and mounted on double sided carbon tape affixed to a stainless steel ruler.  
+
+<|ref|>text<|/ref|><|det|>[[115, 508, 882, 655]]<|/det|>
+The SXR spectra were collected at an angle of \(100^{\circ}\) to the beam over a photon energy range of 275- 325 eV with a step size of 0.1 eV. The energy was calibrated using a graphite standard in the beamline which was collected simultaneously with the \(I_0\) and sample SXR spectra. The double normalization and a pre- and post- edge linear subtraction (background) were conducted using the Athena software (Stöhr 2013).  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 684, 235, 700]]<|/det|>
+## Synchrotron IR  
+
+<|ref|>text<|/ref|><|det|>[[115, 729, 882, 907]]<|/det|>
+For infrared microspectroscopy, approximately \(\sim 30\) free water- stable microaggregates (53- 250 μm) and mineral fractions (<53 μm) were hand- picked on a glass fibre filter paper and humidified gently over 18 hours (Lehmann et al. 2017; Hernandez- Soriano et al. 2018). The aggregates and fractions were frozen at \(- 20^{\circ}C\) before being cryo- ultramicrotomed at 200 nm using a diamond knife. No embedding media was used. The multiple sections per sample (n > 2) were directly collected on CaF2 windows (IR transparent).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 882, 229]]<|/det|>
+The sections on \(\mathsf{CaF}_2\) were directly scanned at the IR beamline at the Australian Synchrotron using a Bruker Hyperion 3000 infrared microscope and a V80v Fourier transform infrared spectrometer. The detail of the microscope was described in Hernandez- Soriano et al. (2018). The spectral maps were produced in transmission mode from 64 scans with a resolution of \(4 \mathsf{cm}^{- 1}\) , step size of \(5 \mu \mathsf{m}\) . Multiple maps were acquired for each treatment to represent the heterogeneity of the sample.  
+
+<|ref|>text<|/ref|><|det|>[[115, 256, 882, 403]]<|/det|>
+Maps were processed using the software OPUS 8.2 (Bruker Optik GmbH, Germany), targeting absorbance at \(3630 \mathsf{cm}^{- 1}\) (O-H groups of clays), \(2920 \mathsf{cm}^{- 1}\) (aliphatic- C), \(1600 \mathsf{cm}^{- 1}\) (aromatic- C), and \(1035 \mathsf{cm}^{- 1}\) (polysaccharides- C)(Hernandez- Soriano et al. 2018). The area of these four absorbance peaks was integrated to the map. A linear regression was conducted to assess the correlation between clay content and the selected C functional groups.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 431, 306, 448]]<|/det|>
+## STEM-EDS-EELS and XPS  
+
+<|ref|>text<|/ref|><|det|>[[115, 475, 882, 911]]<|/det|>
+Forty biochar particles were extracted from the soil samples per plot and were examined using a Zeiss Sigma Scanning electron microscope. Detailed analysis of 5 particles was carried using a Bruker X- ray Dispersive analyser (EDS). A Cs- corrected FEI Titan 80/300 scanning transmission electron microscope (STEM) working at \(80 \mathsf{keV}\) , equipped with a Gatan imaging filter Tridiem and an EDX analyzer was utilised to determine the structure and composition of the organo- mineral clusters that had formed on the surface of the aged biochar. Twenty biochar particles were sonicated in ethanol and then a sample of this was placed on a lacey carbon grind as described by Archanjo et al (2017). Detailed examination of 2 clusters was carried out using energy electron loss spectroscopy (EELS) and EDS. X- ray photoelectron spectroscopy (XPS) examination of both whole and crushed \((< 0.5 \mathsf{mm})\) 1- year aged particles of biochar was undertaken. Carbon 1s photoelectron peak was decomposed in five components: C1- C5 (Table S9). The first one (C1) centered in \(284.6 \mathsf{eV}\) , typical of electrons in carbon \(\mathsf{sp}^2\) bounds (C=C), i.e., delocalized \(\mathsf{sp}^2\) electrons. For this component, an asymmetrical line shape was used to fit. The asymmetry of the C1 component, known as a "defect peak", is related to the localized \(\mathsf{sp}^2\) electrons. Electrons of C- C or C- H bounds typically appear with a binding energy shift of \(0.9 \mathsf{eV}\) in
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 883, 293]]<|/det|>
+relation to sp² delocalized electrons, causing a broadening in the first component. The component C2, centered in 286.2±0.2 eV may be attributed to C- OH (phenol or hydroxyl groups), ether (C- O- C) or pyrrolic groups (C- N). Some authors also attribute this component to Csp³ free radicals. The component C3 is attributed to carbonyl groups (C=O) centered in 287.5±0.4 eV, the component C4 is attributed to the carboxyl groups (COOH) centered in 289.1 ± 0.3 eV, and the last one, the component C5 in 291.4 eV is attributed to the shake- up satellite peak, characteristic of π→π* transition of electrons delocalized sp².  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 322, 617, 339]]<|/det|>
+## The concentration of DOC and its fractions, measured by LC-OCD  
+
+<|ref|>text<|/ref|><|det|>[[115, 368, 883, 577]]<|/det|>
+Dissolved organic carbon (DOC) in a water solution was analysed using liquid chromatography – organic carbon detection (LC- OCD). Two major fractions were: chromatographic organic carbon (CDOC) and hydrophobic organic carbon (HOC). CDOC (hydrophilic fraction) can be categorized into five fractions as a factor of retention time and molecular weight: i) biopolymers, ii) persistent C- like substances, iii) building blocks, iv) low molecular weight acids and v) low molecular weight neutrals. Samples were extracted in distilled water with a ratio of 1:10 (w/v). The solutions were regularly stirred at 50 °C for 24 hours before filtration to differentiate solid and liquid phases.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 606, 382, 622]]<|/det|>
+## Calculation and statistical analysis  
+
+<|ref|>text<|/ref|><|det|>[[115, 652, 883, 765]]<|/det|>
+The cumulative SOC, biochar- C mineralization, and root respiration over 466 d were calculated as the area of a linear interpolation across all measurement points. All statistical analyses were conducted within the R environment (R development core team 2012). When significant F- tests were obtained (P= 0.05), means were separated using a least significant difference (LSD) test at the 0.05 probability.  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 794, 607, 811]]<|/det|>
+## Calculations of global implication for increasing soil carbon sink  
+
+<|ref|>text<|/ref|><|det|>[[115, 840, 840, 887]]<|/det|>
+Projection for wood biochar production is estimated at 4.8 - 8.3 Pg, based on the total annual production of 5.5 to 9.5 Pg biochar by 2100 (Lehmann et al. 2011) with 87% of the feedstock as
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 84, 875, 101]]<|/det|>
+wood (Jirka and Tomlinson 2013). Using the same application rate in this current study (10 t ha<sup>-1</sup>), all 
+
+<|ref|>text<|/ref|><|det|>[[115, 117, 825, 132]]<|/det|>
+wood biochar is assumed to be applied to 0.5 - 0.8 Gha in 2100, accounting for up to 100 % of 
+
+<|ref|>text<|/ref|><|det|>[[115, 148, 825, 163]]<|/det|>
+tropical Ferralsol and 36 % of tropical grasslands (Lal 2004). The global C sequestration rate in 
+
+<|ref|>text<|/ref|><|det|>[[115, 180, 850, 195]]<|/det|>
+grasslands is reported between \(1.3 \times 10^{-10}\) and \(7.6 \times 10^{-10}\) Pg C ha<sup>-1</sup> yr<sup>-1</sup> (Minasny et al. 2017). The 
+
+<|ref|>text<|/ref|><|det|>[[115, 212, 852, 226]]<|/det|>
+range of C sequestration in grasslands before biochar amendment in 2100 would be around 0.07- 
+
+<|ref|>text<|/ref|><|det|>[[115, 243, 867, 259]]<|/det|>
+0.61 Pg C (i.e. 0.5 or 0.8 Gha at \(1.3 \times 10^{-10}\) and \(7.6 \times 10^{-10}\) Pg C ha<sup>-1</sup> yr<sup>-1</sup>). We found a 16 % increase in 
+
+<|ref|>text<|/ref|><|det|>[[115, 276, 864, 290]]<|/det|>
+retention of new C in the recent + historical plots compared with the control. This would lead to an 
+
+<|ref|>text<|/ref|><|det|>[[115, 308, 700, 323]]<|/det|>
+additional soil C sink potential of 0.01-0.1 Pg C (i.e. 16 % of 0.07 or 0.61 Pg C). 
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 355, 242, 369]]<|/det|>
+## Figure Captions 
+
+<|ref|>text<|/ref|><|det|>[[115, 397, 878, 538]]<|/det|>
+**Fig. 1 Conceptual diagram of the formation of organo-mineral coatings on catalytic biochar** **surfaces over time in a Rhodic Ferralsol.** Biochar can act as a bio-catalyst to accelerate formation of organo-mineral microaggregates (53-250 μm) and mineral-protected soil organic matter (<53 μm) on its surfaces and induce negative priming of soil organic carbon. Microbes, fungal hyphae and root hairs can further mine minerals within pores via exudation and dissolution. Microbial necromass covered with minerals is then incorporated into the organo-mineral (<250 μm) and organo-organic (<100 nm) interfaces. Following wetting-drying and plant growth cycles, organo mineral and organo-organic aggregates break-off from organic matter because of weak bonding. Once these aggregates break off, new mineral layers can form. 
+
+<|ref|>text<|/ref|><|det|>[[115, 558, 868, 777]]<|/det|>
+**Fig. 2 Belowground carbon dynamics in the longest continuous biochar field experiment. a,** Changes in total soil organic carbon (SOC, Mg C ha<sup>-1</sup>) in the control and biochar-amended soils over 9.5 years (n=3, LSD = 1.1). Total SOC was measured in the 0-100 soil layer on an equivalent mass basis using Dumas combustion. **b,** Rhizosphere priming as difference in cumulative SOC mineralization between planted and unplanted “recent” biochar amended soil or soil with the “recent + historical” biochar. “Recent” biochar is biochar applied to a part of the control plots 8.2 years after the trial was established (closed triangles). The “recent + historical” amendment is biochar applied to a part of the historical biochar plots 8.2 years after the trial was established (open triangles). Confidence intervals (95%) of “recent” biochar and “recent + historical” biochar amendments are plotted in dashed lines and normalized against the mean squares across all treatments at each sampling event (n=3). For biochar amendment, the CI was based on a sensitivity analysis (Online Method Section), which considers the extreme scenarios of contrasting SOC pools (C3 vs. C4 dominated) by differences in δ<sup>13</sup>C signatures. The six arrows represent nitrogen fertilizer additions. 
+
+<|ref|>text<|/ref|><|det|>[[115, 795, 878, 906]]<|/det|>
+**Fig. 3 Allocation and retention of rhizodeposits (¹³C-enriched) and three-dimensional elemental distribution in a biochar-amended Ferralsol at 9.5 years. a, 8.2 years after the first application, the biochar-amended soil received a recent dose of biochar at 10 Mg ha-1 to the historical plots (“recent + historical”). The same total amount (190 mg ¹³C m⁻²) was supplied in each treatment plot (n=3). b, Recent biochar (“recent”) was mixed in top 100 mm of soil at 10 Mg ha⁻¹ one year before measurement. c, 3D FIB-SEM-EDS of an intact soil aggregate (30 μm × 25 μm × 24 μm) from the “recent + historical” biochar plots. d, 3D FIB-SEM-EDS of an intact soil aggregate (30 μm × 20 μm ×
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 84, 858, 132]]<|/det|>
+30 \(\mu \mathrm{m}\) ) from the "recent" biochar plots. Soil sampling was conducted before and 15 days after labelling. The total recovery of \(^{13}\mathrm{C}\) labelling is given, including soil + root respiration, root biomass, free- and occluded particulate matter and mineral fractions.  
+
+<|ref|>text<|/ref|><|det|>[[115, 148, 878, 325]]<|/det|>
+Fig. 4 Synchrotron- based spectromicroscopic analysis of microaggregates (53- 250 \(\mu \mathrm{m}\) ) and mineral fractions (<53 \(\mu \mathrm{m}\) ) in the unamended control and historical biochar- amended plots with recent biochar addition. a, Average SXR spectra of microaggregates (53- 250 \(\mu \mathrm{m}\) ) and mineral fractions (<53 \(\mu \mathrm{m}\) ) with the "recent + historical" and "recent" biochar amendments (n=9, CV% < 3%). b, Semi- thin (200 nm) sections of free water- stable microaggregates (53- 250 \(\mu \mathrm{m}\) ) and mineral fractions (<53 \(\mu \mathrm{m}\) ) isolated from the Ferralsol with the "recent + historical" and "recent" biochar amendments analysed using synchrotron- based IR- microspectroscopy. Spectral maps showing the distribution of polysaccharide- C (1035 cm \(^{-1}\) ), aromatic- C (1600 cm \(^{-1}\) ), aliphatic- C (2920 cm \(^{-1}\) ), and mineral- OH (3650 cm \(^{-1}\) ) were obtained from 64 co- added scans (4 cm \(^{-1}\) resolution), lateral resolution 5 \(\mu \mathrm{m}\) (bars: 50 \(\mu \mathrm{m}\) ). The signal intensity for each molecular group varied according to the colour scale shown. The images on the left are optical micrographs of the semi- thin sections.  
+
+<|ref|>text<|/ref|><|det|>[[115, 339, 872, 500]]<|/det|>
+Fig 5. In situ spectromicroscopic analysis of the organo- mineral coating on biochar surfaces and pores over time. a, Average SXR spectra of field- extracted "recent" (1- yr aged) and "historical" (9.5- yr aged) biochars (n=9, CV% < 3%). b, high magnification secondary electron image of a pore where fungi exist. c, EDS spectrum of the area in b. d, STEM- HAADF image of organo- mineral clusters on the "recent" biochar surface; e, its EELS spectra. f, High resolution image of the surface of the organomineral layer inside the pore of the "historical" biochar. g, EDS spectrum of the area in f. h, HAADF image of a deposit attached to the surface of the "historical" biochar. i, its EELS spectra. j, DOC of bulk soils from the "recent + historical" and "recent" biochar amendments analysed with LCOCD. The hydrophilic fraction is further sub- divided into five categories, i: biopolymer, ii: persistent C, iii: building blocks, iv: low molecular weight acids and v: low molecular weight neutrals.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 560, 205, 574]]<|/det|>
+## References  
+
+<|ref|>text<|/ref|><|det|>[[112, 600, 881, 905]]<|/det|>
+1 Goldstein, A. et al. Protecting irrecoverable carbon in Earth's ecosystems. Nature Climate Change, 1- 9 (2020).  2 Cavicchioli, R. et al. Scientists' warning to humanity: microorganisms and climate change. Nature Reviews Microbiology 17, 569- 586 (2019).  3 Ogle, S. M., Breidt, F. J. & Paustian, K. Agricultural management impacts on soil organic carbon storage under moist and dry climatic conditions of temperate and tropical regions. Biogeochemistry 72, 87- 121 (2005).  4 Minasny, B. et al. Soil carbon 4 per mille. Geoderma 292, 59- 86 (2017).  5 Arneth, A. et al. in Climate Change and Land: an IPCC special report on climate change, desertification, land degradation, sustainable land management, food security, and greenhouse gas fluxes in terrestrial ecosystems 1- 98 (Intergovernmental Panel on Climate Change (IPCC), 2019).  6 Bossio, D. et al. The role of soil carbon in natural climate solutions. Nature Sustainability 3, 391- 398 (2020).  7 Stocking, M. A. Tropical soils and food security: the next 50 years. Science 302, 1356- 1359 (2003).  8 Pachauri, R. K. et al. Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. (2014).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[57, 87, 884, 870]]<|/det|>
+546 9 Allen, M. et al. Global Warming of \(1.5^{\circ}C\) . Summary for Policymakers. (2018). 547 10 Woolf, D., Amonette, J. E., Street- Perrott, F. A., Lehmann, J. & Joseph, S. Sustainable biochar to mitigate global climate change. Nature Communications 1, 56 (2010). 548 11 Masson- Delmotte, V. et al. Global warming of 1.5 C. An IPCC Special Report on the impacts of global warming of 1 (2018). 549 12 Lal, R. Managing soils and ecosystems for mitigating anthropogenic carbon emissions and advancing global food security. BioScience 60, 708- 721, doi:10.1525/bio.2010.60.9.8 (2010). 550 13 Lal, R. Soil carbon sequestration to mitigate climate change. Geoderma 123, 1- 22, doi:10.1016/j.geoderma.2004.01.032 (2004). 551 14 Lehmann, J., Bossio, D. A., Kögel- Knabner, I. & Rillig, M. C. The concept and future prospects of soil health. Nature Reviews Earth & Environment, 1- 10 (2020). 552 15 Amelung, W. et al. Towards a global- scale soil climate mitigation strategy. Nature Communications 11, 1- 10 (2020). 553 16 Chenu, C. et al. Increasing organic stocks in agricultural soils: Knowledge gaps and potential innovations. Soil and Tillage Research 188, 41- 52 (2019). 554 17 Amundson, R. & Biardeau, L. Opinion: Soil carbon sequestration is an elusive climate mitigation tool. Proceedings of the National Academy of Sciences 115, 11652- 11656 (2018). 555 18 Ji, C., Cheng, K., Nayak, D. & Pan, G. Environmental and economic assessment of crop residue competitive utilization for biochar, briquette fuel and combined heat and power generation. Journal of Cleaner Production 192, 916- 923 (2018). 556 19 Woolf, D., Solomon, D. & Lehmann, J. Land restoration in food security programmes: synergies with climate change mitigation. Climate Policy 18, 1260- 1270 (2018). 557 20 Woolf, D., Lehmann, J. & Lee, D. R. Optimal bioenergy power generation for climate change mitigation with or without carbon sequestration. Nature Communications 7, 1- 11 (2016). 558 21 Woolf, D., Amonette, J. E., Street- Perrott, F. A., Lehmann, J. & Joseph, S. Sustainable biochar to mitigate global climate change. Nature communications 1, 1- 9 (2010). 559 22 Xu, X. et al. Greenhouse gas mitigation potential in crop production with biochar soil amendment—a carbon footprint assessment for cross- site field experiments from China. Gcb Bioenergy 11, 592- 605 (2019). 560 23 Cowie, A. et al. Biochar, carbon accounting and climate change. (En: Biochar for environmental management: science, technology and ..., 2015). 561 24 Torn, M. S., Trumbore, S. E., Chadwick, O. A., Vitousek, P. M. & Hendricks, D. M. Mineral control of soil organic carbon storage and turnover. Nature 389, 170- 173, doi:10.1038/38260 (1997). 562 25 Hemingway, J. D. et al. Mineral protection regulates long- term global preservation of natural organic carbon. Nature 570, 228- 231 (2019). 563 26 Woolf, D. & Lehmann, J. Microbial models with minimal mineral protection can explain long- term soil organic carbon persistence. Scientific reports 9, 1- 8 (2019). 564 27 Kopittke, P. M. et al. Soil organic matter is stabilized by organo- mineral associations through two key processes: The role of the carbon to nitrogen ratio. Geoderma 357, 113974 (2020). 565 28 Lehmann, J. & Kleber, M. The contentious nature of soil organic matter. Nature 528, 60- 68, doi:10.1038/nature16069 (2015). 566 29 Lehmann, J. et al. Persistence of soil organic carbon caused by functional complexity. Nature Geoscience 13, 529- 534 (2020). 567 30 Joseph, S. et al. An investigation into the reactions of biochar in soil. Soil Research 48, 501- 515 (2010). 568 31 Lin, Y., Munroe, P., Joseph, S., Kimber, S. & Van Zwieten, L. Nanoscale organo- mineral reactions of biochars in ferrosol: an investigation using microscopy. Plant and Soil 357, 369- 380, doi:10.1007/s11104- 012- 1169- 8 (2012).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[57, 85, 884, 470]]<|/det|>
+595 32 Archanjo, B. et al. Nanoscale analyses of the surface structure and composition of biochars extracted from field trials or after co- composting using advanced analytical electron microscopy. Geoderma 294, 70- 79 (2017). 597 33 Nguyen, T. T. N. et al. The effects of short term, long term and reapplication of biochar on soil bacteria. Science of the Total Environment 636, 142- 151 (2018). 600 34 Ding, F. et al. A meta- analysis and critical evaluation of influencing factors on soil carbon priming following biochar amendment. Journal of soils and sediments 18, 1507- 1517 (2018). 602 35 Liang, C., Amelung, W., Lehmann, J., & Kastner, M. Quantitative assessment of microbial necromass contribution to soil organic matter. Global change biology 25, 3578- 3590 (2019). 603 36 Lehmann, J. et al. Biochar effects on soil biota - A review. Soil Biology & Biochemistry 43, 1812- 1836, doi:10.1016/j.soilbio.2011.04.022 (2011). 605 37 Singh, B. P. & Cowie, A. L. Long- term influence of biochar on native organic carbon mineralisation in a low- carbon clayey soil. Scientific Reports 4, doi:10.1038/srep03687 (2014). 608 38 Li, X., Wang, T., Chang, S. X., Jiang, X., & Song, Y. Biochar increases soil microbial biomass but has variable effects on microbial diversity: A meta- analysis. Science of The Total Environment 749, 141593 (2020). 611 39 Husson, O. Redox potential (Eh) and pH as drivers of soil/plant/microorganism systems: a transdisciplinary overview pointing to integrative opportunities for agronomy. Plant and Soil 362, 389- 417 (2013). 614 40 Weng, Z. H. et al. The accumulation of rhizodeposits in organo- mineral fractions promoted biochar- induced negative priming of native soil organic carbon in Ferralsol. Soil Biology and Biochemistry 118, 91- 96 (2018). 617 41 Weng, Z. et al. Biochar built soil carbon over a decade by stabilizing rhizodeposits. Nature Clim. Change 7, 371- 376, doi:10.1038/nclimate3276  
+
+<|ref|>text<|/ref|><|det|>[[58, 475, 884, 892]]<|/det|>
+http://www.nature.com/nclimate/journal/v7/n5/abs/nclimate3276. html#supplementary- information (2017). 621 42 Weng, Z. H. et al. Plant- biochar interactions drive the negative priming of soil organic carbon in an annual ryegrass field system. Soil Biology and Biochemistry 90, 111- 121 (2015). 623 43 Slavich, P. et al. Contrasting effects of manure and green waste biochars on the properties of an acidic ferralsol and productivity of a subtropical pasture. Plant and Soil 366, 213- 227 (2013). 626 44 Farquhar, G., Hubick, K., Condon, A. & Richards, R. in Stable isotopes in ecological research 21- 40 (Springer, 1989). 628 45 Graber, E. R. et al. Biochar impact on development and productivity of pepper and tomato grown in fertigated soilless media. Plant and Soil 337, 481- 496, doi:10.1007/s11104- 010- 0544- 6 (2010). 631 46 Kaiser, C., Franklin, O., Richter, A. & Dieckmann, U. Social dynamics within decomposer communities lead to nitrogen retention and organic matter build- up in soils. Nature Communications 6 (2015). 634 47 Lehmann, J. et al. Spatial complexity of soil organic matter forms at nanometre scales. Nature Geoscience 1, 238- 242 (2008). 636 48 Lehmann, J. & Solomon, D. in Developments in soil science Vol. 34 289- 312 (Elsevier, 2010). 637 49 Lybrand, R. A. et al. A coupled microscopy approach to assess the nano- landscape of weathering. Scientific reports 9, 1- 14 (2019). 638 50 Van Hees, P., Jones, D., Jentschke, G. & Godbold, D. Mobilization of aluminium, iron and silicon by Picea abies and ectomycorrhizas in a forest soil. European Journal of Soil Science 55, 101- 112 (2004). 641 51 Kasozi, G. N., Zimmerman, A. R., Nkedi- Kizza, P. & Gao, B. Catechol and humic acid sorption onto a range of laboratory- produced black carbons (biochars). Environmental Science & Technology 44, 6189- 6195 (2010).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[60, 85, 884, 435]]<|/det|>
+645 52 Rasse, D. P., Rumpel, C. & Dignac, M.- F. Is soil carbon mostly root carbon? Mechanisms for a 646 specific stabilisation. Plant and Soil 269, 341- 356 (2005). 647 53 Violante, A., Barberis, E., Pigna, M. & Boero, V. Factors affecting the formation, nature, and 648 properties of iron precipitation products at the soil- root interface. Journal of Plant Nutrition 649 26, 1889- 1908 (2003). 650 54 Glaser, B., Balashov, E., Haumaier, L., Guggenberger, G. & Zech, W. Black carbon in density 651 fractions of anthropogenic soils of the Brazilian Amazon region. Organic Geochemistry 31, 652 669- 678 (2000). 653 55 Czimczik, C. I. & Masiello, C. A. Controls on black carbon storage in soils. Global 654 Biogeochemical Cycles 21 (2007). 655 56 Possinger, A. R., Zachman, M. J., Enders, A., Levin, B. D., Muller, D. A., Kourkoutis, L. F., & 656 Lehmann, J. Organo- organic and organo- mineral interfaces in soil at the nanometer scale. 657 Nature Communications 11, 1- 11 (2020). 658 57 Huang, P.- M., Wang, M.- K. & Chiu, C.- Y. Soil mineral- organic matter- microbe interactions: 659 impacts on biogeochemical processes and biodiversity in soils. Pedobiologia 49, 609- 635 660 (2005). 661 Keluweit, M. et al. Mineral protection of soil carbon counteracted by root exudates. Nature 662 Climate Change 5, 588- 595 (2015). 663 59 Tang, J. & Riley, W. J. Weaker soil carbon- climate feedbacks resulting from microbial and 664 abiotic interactions. Nature Climate Change 5, 56- 60 (2015). 665 60 Schmidt, M. W. et al. Persistence of soil organic matter as an ecosystem property. Nature 478, 466- 59 (2011).  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 449, 272, 465]]<|/det|>
+## Acknowledgements  
+
+<|ref|>text<|/ref|><|det|>[[110, 494, 884, 896]]<|/det|>
+The authors thank the Australian Government, Department of Agriculture and Water Resources for supporting the National Biochar Initiatives (2009- 2012, 2012- 2014) which co- funded this research. We are particularly grateful to Dr. Peter Slavich, as one of the key founders of this long- term field experiment for providing insightful comments on the initial draft. Part of this research was undertaken on the Soft X- ray spectroscopy beamline and the Infrared microscopy beamline at the Australian Synchrotron, part of ANSTO (grant numbers AS1_SXR_15754 and AS1_IRM_15940). We thank the beamline scientists, Drs Bruce Cowie and Lars Thomsen, for their technical support on the soft x ray analysis and Drs Mark Tobin, Annaleise Klein and Jitraporn (Pimm) Vongsvivut, for their technical support on the infrared microscopy analysis. Part of the research is funded by La Trobe University's Research Focus Area in Securing Food, Water and the Environment (Grant Ready: SFWE RFA 2000004295; Collaboration Ready: SFWE RFA 2000004349). We also appreciate the technical support from Scott Petty and Josh Rust for maintaining this field experiment over the past decade, and laboratory support from Nichole Morris. We also thank Dr Carlos Achete from INMETRO, Brazil and Dr
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[60, 82, 883, 230]]<|/det|>
+Bin Gong from the University of New South Wales, Australia, for performing XPS analysis of biochars and soils, Dr Sarasadat Taherymoosavi from the University of New South Wales, Australia, for technical assistance in LC- OCD analysis. We acknowledge the intellectual contribution from Prof Johannes Lehmann for discussions on the potential mechanisms of biochar- induced stabilization of rhizodeposits.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 254, 283, 270]]<|/det|>
+## Author contributions  
+
+<|ref|>text<|/ref|><|det|>[[115, 284, 880, 430]]<|/det|>
+ZW drafted and wrote the manuscript, experimental design, set- up and conducted experiments, and data collection and analysis; LVZ, BPS and LMM wrote the manuscript, aided in experimental design, critical revision of the article; SJ, ET, BSA and MTR collected and analyzed data, critical revision of the article; TJT, CT, AF, PMK, SK, SM and AC provided critical revision of the article. All authors provided final approval of the revision to be published.  
+
+<|ref|>text<|/ref|><|det|>[[60, 457, 880, 508]]<|/det|>
+Correspondence and requests for materials should be addressed to LVZ via email: lukas.van.zwieten@dpi.nsw.gov.au
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[44, 43, 142, 68]]<|/det|>
+## Figures  
+
+<|ref|>image<|/ref|><|det|>[[42, 90, 875, 670]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[44, 697, 115, 716]]<|/det|>
+<center>Figure 1 </center>  
+
+<|ref|>text<|/ref|><|det|>[[41, 737, 949, 919]]<|/det|>
+Conceptual diagram of the formation of organo- mineral coatings on catalytic biochar surfaces over time in a Rhodic Ferralsol. Biochar can act as a bio- catalyst to accelerate formation of organo- mineral microaggregates (53- 250 \(\mu \mathrm{m}\) ) and mineral- protected soil organic matter ( \(< 53 \mu \mathrm{m}\) ) on its surfaces and induce negative priming of soil organic carbon. Microbes, fungal hyphae and root hairs can further mine minerals within pores via exudation and dissolution. Microbial necromass covered with minerals is then incorporated into the organo- mineral ( \(< 250 \mu \mathrm{m}\) ) and organo- organic ( \(< 100 \mathrm{nm}\) ) interfaces. Following wetting- drying and plant growth cycles, organo mineral and organo- organic aggregates break- off from organic matter because of weak bonding. Once these aggregates break off, new mineral layers can form.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[44, 42, 311, 70]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[44, 93, 765, 113]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[60, 130, 694, 177]]<|/det|>
+- Videoabstract3DFIBSEMEDXofrhizodepositsretentioninaggregate.mp4- SupplementaryInformation.pdf
+
+<--- Page Split --->

preprint/preprint__0f43ab377913d883885fc5b793f1a6d8dbf89b2a4f2bd4ee445349870e3187e4/images_list.json

@@ -0,0 +1,107 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Fig. 1. Senescent \\(\\beta\\) -cell–derived miR-503-322 promoted pancreatitis in mice. (A) qPCR analysis of Pri-miR-503 expression in islets and acini of 20-week-old control (WT) and \\(\\beta\\) -cell specific miR-503 transgenic (βTG) male mice. \\(n = 4\\) . (B) qPCR analysis of miR-503 expression in acini of WT and βTG mice. \\(n = 3\\) . (C) qPCR analysis of Pri-miR-503 expression in pancreas, islets and acini of 12 weeks and 1.5 years old male mice. \\(n = 5\\) . (D) qPCR analysis of miR-503 and miR-322 expression in islets and acini of 12 weeks and 1.5 years old male mice, respectively. \\(n = 5\\) . (E) Schematic flow diagram of sponge \\(\\beta\\) -cell miR-503-322 and induced pancreatitis in aged male mice. The 1.4-years C57BL/6J male mice were randomly divided into two groups. The control group (Ctr, \\(n = 4\\) ) and the experimental group (SP, \\(n = 5\\) ) were respectively injected with ctr-AAV and miR-503-322 sponge-AAV through pancreatic ductal Infusion. Two months later, AP was induced by intraperitoneal injection (ip.) of caerulein (50 \\(\\mu \\mathrm{g / kg}\\) , hourly for 6 consecutive times), and pancreatitis parameters were detected 2 hours",
+    "footnote": [],
+    "bbox": [
+      [
+        123,
+        111,
+        848,
+        690
+      ]
+    ],
+    "page_idx": 30
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Fig. 2. \\(\\beta\\) cells secreted nano-vesicular miR-503-322 to enter acinar cells. (A) Experimental scheme: PKH67-labeled nano-vesicles derived from islets of WT (WT-βNVs) or βTG islets (βTG-βNVs) were co-incubated with fresh acini for 8 hours. Representative confocal images of PKH67, phalloidin and nuclei in primary acini and quantitation of relative PKH67 fluorescence intensity. (B) Experimental scheme: PKH67-labeled WT-βNVs or βTG-βNVs were infusion into the C57 male mouse pancreas via pancreatic ductal. The pancreas was harvested after 12 hours and stained with frozen sections for amylase and then visualized. Representative confocal images of PKH67, amylase and nuclei of pancreas and quantitation of relative PKH67 fluorescence intensity. (C) qPCR analysis of miR-503 expression in acini of received βNVs. n=3. (D-G) Pancreatitis parameters assay after initial caerulein (50 μg/kg) injection 7 hours in 12-week-old control (WT) and β-cell specific miR-503-322 knock-in (βKI) female mice. H&E and F4/80 immunohistochemistry (IHC) of pancreatic sections (D), quantitation of the number of F4/80 positive cells in pancreatic sections under 200x microscopic view (E), pancreatic histological scores (F) and level of serum amylase (G). Arrows indicate the macrophages. Data are presented as Mean ± SEM. n=5 mice/group. Data were analyzed using one-way ANOVA followed by Bonferroni's post hoc (G) or unpaired two-tailed Student's \\(t\\) test (A, B, C, E and F). NS, Not Significant; \\(*P < 0.05\\) ; \\(**P < 0.01\\) ; \\(***P < 0.001\\) .",
+    "footnote": [],
+    "bbox": [
+      [
+        120,
+        88,
+        884,
+        560
+      ]
+    ],
+    "page_idx": 32
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Fig. 3. Direct elevation of miR-503-322 in acinar cells triggers both acute and chronic pancreatitis. (A) qPCR analysis for pancreatic Pri-miR-503 expression in 8-week-old control (WT), PKI heterozygous (PKI/WT) and PKI homozygous (PKI/KI) male mice. n=5. (B) Weight",
+    "footnote": [],
+    "bbox": [
+      [
+        125,
+        85,
+        880,
+        844
+      ]
+    ],
+    "page_idx": 33
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Fig. 4. MiR-503-322 knockout alleviated caerulein-induced acute pancreatitis. (A) Schematic of caerulien-induced AP on 12-week WT and KO male mice. \\(n = 5\\) . (B-E) Pancreatic weights after calibration with body weight (B), serum amylase (C) and lipase (D) levels, histological score of the pancreas (E) after PBS or caerulein treatment groups. (F and G) Representative sections of pancreatic H&E (F) and immunofluorescence staining of F4/80 (green) after PBS or caerulein treatment 7 hours in WT and KO male mice. Quantitation of the number of F4/80 positive cells in pancreatic sections under 600x microscopic view (G). Arrows indicate the macrophages. Data are presented as Mean ± SEM. Data were analyzed using one-way ANOVA followed by Bonferroni's post hoc. \\(n = 4 - 5\\) mice/group. NS, Not Significant; \\(*P < 0.05\\) ; \\(**P < 0.01\\) ; \\(***P < 0.001\\) .",
+    "footnote": [],
+    "bbox": [
+      [
+        125,
+        85,
+        875,
+        560
+      ]
+    ],
+    "page_idx": 35
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Fig. 5. MiR-503-322 promotes pancreatitis by inhibiting zymogen secretion and acinar-cell proliferation. (A) Extraction of fresh acinar cells from 8-week-old WT and PKI/WT male mice, in vitro stimulation with different concentrations of caerulein for 30 minutes and determination of amylase content in the supernatant. See Materials and Methods for details. n=3. (B and C) Amylase levels after calibration of total content release from acinar cells of 12-week WT and KO male mice (B) and 12-week and 1.5-year C57BL/6J male mice (C) after 30 min of stimulation with caerulein. n=3. (D) After 48 hours of induction by tamoxifen in WT or EKI mice, pancreatic acini were isolated and incubated with or without caerulein (0.01 μM) for",
+    "footnote": [],
+    "bbox": [
+      [
+        115,
+        88,
+        864,
+        750
+      ]
+    ],
+    "page_idx": 36
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_6.jpg",
+    "caption": "Fig. 6. MKNK1 is a target of miR-503-322 and acinar cell-specific restoration of it reverses the phenotype of pancreatitis in mice. (A)The MKNK1 network was predicted based on the common signature from the Ingenuity database overlaid with microarray data from miR-503-overexpressing mouse pancreatic \\(\\beta\\) cell line MIN6 cells with a 1.5-fold change cutoff compared with negative control cells. (B) WT and EKI male mice at 5 days after tamoxifen induction; Male C57BL/6J at 12-week and 1.5-year; male WT and KO at 12-week after AP induced and \\(\\beta\\) -cell specific sponge of miR-503-322 in control and experimental mice pancreatic protein western blotting. \\(n = 3 - 5\\) . (C) Experimental scheme: 8-week-old WT male mice were injected intraperitoneally (ip.) with control AAV and EKI male mice were injected with control (Ctr-AAV) and MKNK1-AAV, respectively, one month later tamoxifen was induced for 3 consecutive days and tested at day 7. \\(n = 5\\) . (D and E) Immunofluorescence staining of Flag and MKNK1 of pancreas sections (D) from each group of mice at 13 weeks and western blotting of pancreatic proteins (E). (F) Gain of body weight, serum amylase and lipase level",
+    "footnote": [],
+    "bbox": [
+      [
+        113,
+        87,
+        875,
+        655
+      ]
+    ],
+    "page_idx": 38
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_7.jpg",
+    "caption": "Fig. 7. Evidence of miR-503 and MKNK1 in aging-associated pancreatitis changes in the Chinese population. (A) Representative images of H&E and Masson staining of pancreatic sections from the young adult (YA) and the elderly adult (EA); quantitation of collagen volume fraction. The dashed area indicates acini. n=10. (B) Representative images of immunofluorescence staining of PCNA (green) in pancreatic sections and counted the number of PCNA-positive cells. n=10. Arrows indicate proliferating acinar cells. (C) In situ hybridization of miR-503 (40 nM) in young and elderly pancreatic sections. Scramble-RNA was negative reference (40 nM) and U6 was positive reference (0.1 nM). The dotted line indicate pancreatic islet and solid line is exocrine. (D) Representative images of immunofluorescence staining of",
+    "footnote": [],
+    "bbox": [
+      [
+        118,
+        80,
+        830,
+        730
+      ]
+    ],
+    "page_idx": 40
+  }
+]

preprint/preprint__0f43ab377913d883885fc5b793f1a6d8dbf89b2a4f2bd4ee445349870e3187e4/preprint__0f43ab377913d883885fc5b793f1a6d8dbf89b2a4f2bd4ee445349870e3187e4.mmd

@@ -0,0 +1,371 @@
+
+# Endocrine-exocrine miR-503-322 drives aging-associated pancreatitis via targeting MKNK1 in acinar cells  
+
+Yunxia Zhu zhuyx@njmu.edu.cn  
+
+Nanjing Medical University https://orcid.org/0000- 0002- 4597- 4445  
+
+Kerong Liu Nanjing Medical University  
+
+Tingting Lv Nanjing Medical University  
+
+Wei Tang Nanjing Medical University  
+
+Yan Zhang Children's Hospital of Nanjing Medical University  
+
+Xiao Xiao Nanjing Medical University  
+
+Yating Li Nanjing Medical University  
+
+Xiaoai Chang Nanjing Medical University  
+
+Shusen Wang Tianjin First Central Hospital https://orcid.org/0000- 0002- 2323- 6564  
+
+Stephen Pandol Cedars- Sinai https://orcid.org/0000- 0003- 0818- 6017  
+
+Ling Li Southeast University  
+
+Xiao Han Nanjing Medical University https://orcid.org/0000- 0002- 6467- 1802  
+
+Article  
+
+Keywords:
+
+<--- Page Split --->
+
+Posted Date: June 21st, 2024  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 4521626/v1  
+
+License: © © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Additional Declarations: There is NO Competing Interest.  
+
+Version of Record: A version of this preprint was published at Nature Communications on March 17th, 2025. See the published version at https://doi.org/10.1038/s41467- 025- 57615- x.
+
+<--- Page Split --->
+
+1 Endocrine-exocrine miR- 503- 322 drives aging-associated pancreatitis via targeting2 MKNK1 in acinar cells3 Kerong Liu,<sup>1</sup> Tingting Lv,<sup>1</sup> Wei Tang,<sup>5</sup> Yan Zhang,<sup>6</sup> Xiao Xiao,<sup>1</sup> Yating Li,<sup>1</sup> Xiaoai Chang,<sup>1</sup>4 Shusen Wang,<sup>4</sup> Stephen J Pandol,<sup>3,</sup>* Ling Li,<sup>2,</sup>* Xiao Han,<sup>1,</sup>* and Yunxia Zhu<sup>1,</sup>*5 <sup>1</sup>Key Laboratory of Human Functional Genomics of Jiangsu Province, Biochemistry and6 Molecular Biology, Nanjing Medical University, Nanjing, Jiangsu 211166, China.7 <sup>2</sup>Department of Endocrinology, Zhongda Hospital, School of Medicine, Southeast University,8 Nanjing, 210009, China.9 <sup>3</sup>Division of Gastroenterology, Department of Medicine, Cedars-Sinai Medical Center, Los10 Angeles, CA, United States.11 <sup>4</sup>Organ Transplant Center, Tianjin First Central Hospital, Nankai University, Tianjin 300192,12 China.13 <sup>5</sup>Department of Endocrinology, Geriatric Hospital of Nanjing Medical University, Nanjing,14 Jiangsu 210024, China.15 <sup>6</sup>Children's Hospital of Nanjing Medical University, Nanjing, Jiangsu 210093, China.16 <sup>*</sup>Correspondence to: Yunxia Zhu, zhuyx@njmu.edu.cn, Phone: +86 25 86869426; Fax: +86 17 25 86869425; Xiao Han, hanxiao@njmu.edu.cn; Ling Li, lingli@seu.edu.cn or Stephen J18 Pandol, Stephen.Pandol@cshs.org
+
+<--- Page Split --->
+
+## Abstract  
+
+Aging is the major risk factor for chronic pancreatitis and severity determinant for its acute attack, yet the underlying cause is unclear. Here, we demonstrate that senescent \(\beta\) - cells of endocrine pancreas decide the onset and severity of chronic and acute pancreatitis. During physiological aging, senescent \(\beta\) - cells increase the expression of miR- 503- 322 which is secreted as nano- vesicles to enter exocrine acinar cells, driving a causal and reversible role on aging- associated pancreatitis. Mechanistically, miR- 503- 322 represses MKNK1 to inhibit acinar- cell secretion and proliferation, thereby causing autodigestion and repairing damage of exocrine pancreas. In the elderly population, serum miR- 503 concentration is negatively correlated with amylase, prone to chronic pancreatitis due to increased miR- 503 and decreased MKNK1 in the elderly pancreas. Our findings highlight the miR- 503- 322- MKNK1 axis mediating the endocrine- exocrine regulatory pathway specifically in aged mice and humans. Modulating this axis may provide potential preventive and therapeutic strategies for aging- associated pancreatitis.
+
+<--- Page Split --->
+
+## Introduction  
+
+Pancreatitis is one of the most common causes of hospitalization worldwide and represents higher prevalence in the elderly. \(^{1 - 3}\) Chronic inflammation accumulates during natural aging has been identified responsible for the onset of many diseases, including pancreatitis and type 2 diabetes mellitus (T2DM). \(^{4}\) Recent clinical data showed that the incidence of pancreatitis increases in patients with T2DM, \(^{5 - 7}\) indicating the endocrine part of the pancreas participant in pancreatitis formation. However, the underlined mechanisms remain elusive.  
+
+The endocrine pancreatic islets have a well- recognized anatomical and physiological integration with the exocrine pancreas and regulate its function. \(^{8}\) An involvement of the islet- acinar axis (IAA) has been suggested in the islet- acinar portal system for the physiological regulation of acinar cell function by islet peptides. \(^{9,10}\) A recent study found that islet \(\beta\) - cell- derived cholecystokinin (CCK) acts on acinar cells via the IAA to promote the progression of pancreatic ductal adenocarcinoma (PDAC), \(^{11}\) suggesting that endocrine islet \(\beta\) - cells can crosstalk with acinar cells. In addition, \(\beta\) - cell inflammation exacerbates pancreatitis through chemokine signaling. \(^{12,13}\) These findings suggest that factors secreted abnormally by pancreatic \(\beta\) - cells play a key role in the development of pancreatitis. One possibility is that abnormal secretion of microRNAs (miRNAs) may be involved.  
+
+Pancreatic \(\beta\) - cells are known to mediate intercellular communication through the secretion of extracellular vesicles (EVs) rich in miRNAs, resulting in reduced insulin sensitivity and secretion capacity in a paracrine or distal manner and elevated blood glucose levels. \(^{14}\) However, a regulatory role for miRNAs carried by EVs derived from \(\beta\) - cells has not been established for pancreatitis. We have previously demonstrated that senescent \(\beta\) - cells released
+
+<--- Page Split --->
+
+miR- 503- 322 as exosomes (\~45 nm, also called nano- vesicles) which were transported into peripheral target organs to cause insulin resistance, thereby leading to the onset of T2DM. \(^{15}\) Serendipitously, overexpression of miR- 503 in \(\beta\) cells caused pancreatitis- like changes with age, suggesting that miR- 503 secreted by endocrine \(\beta\) - cells may be important in regulating exocrine functions including pancreatitis.  
+
+The X- linked miR- 503, clustered with miR- 322 has been investigated and shown to play an important role in modulating cell proliferation, cell differentiation, and tissue remodeling. \(^{16}\) In the present study, we found that during natural aging, primary miR- 503- 322 (Pri- miR- 503) was transcribed in the endocrine islets while mature miR- 503 and miR- 322 could be detected in both endocrine and exocrine pancreas. Increased levels of miR- 503- 322 in senescent acinar cells were derived from \(\beta\) - cells and intra- acinar miR- 503- 322 promoted pancreatitis by targeting MAP kinase- interacting kinases (MKNK1). The regulation mode was also conserved in aged population, adding further evidence for endocrine- exocrine crosstalk in regulating pancreatitis and providing novel therapeutic targets for the prevention and treatment of aging- associated pancreatitis.  
+
+## Results  
+
+## Senescent \(\beta\) -cell- derived miR-503-322 promoted pancreatitis in mice  
+
+Our previous study showed that \(\beta\) - cell- specific miR- 503 transgenic (βTG) mice suffered from insulin resistance and \(\beta\) - cell dysfunction, leading to T2DM. \(^{15}\) Coincidentally, we noted that the βTG mice also showed chronic pancreatitis- like changes with advanced age, including diffuse expansion of the interlobar septae, fat accumulation, and fibrosis (Fig. S1A and B). Adult βTG
+
+<--- Page Split --->
+
+mice were also showed significant exacerbation of caerulein- induced AP attack, as evidenced by pancreatic edema, macrophage infiltration, and more severe histologic scorings compared with the WT mice (Fig. S1C- E).  
+
+To understand the role of \(\beta\) - cell miR- 503 on the development of pancreatitis, the expressing distribution of miR- 503 in \(\beta\) TG mice was detected. We found that pri- miR- 503 was significantly increased in islets but not in acini, while the mature miR- 503 was increased in both islets and acini (Fig. 1A and B), suggesting \(\beta\) - cell miR- 503 entering acinar cells. The same expression profiles of pri- miR- 503 and mature miR- 503 and miR- 322 was also observed in aged mice (Fig. 1C and D), making us think about the contribution of miR- 503- 322 to pancreatitis in older age. Consistent with our hypothesis, aged mice showed a more severe form of caerulein- induced AP compared to younger mice (Fig. S2), which could be significantly improved by blocking \(\beta\) - cell miR- 503- 322 levels. An insulin2 promoter- driven sponge- AAV (SP- AAV) specifically expressed in \(\beta\) cells resulted in decreased expression levels of miR- 503- 322 in pancreas (Fig. 1E- G). Meanwhile, caerulein- induced AP measured by serum amylase and lipase levels, pancreatic edema, histologic scorings and macrophage infiltration were significantly ameliorated in aged mice infected with SP- AAV (Fig. 1H- J). These findings indicate that increased levels of miR- 503- 322 in senescent \(\beta\) cells contribute pancreatitis severity associated with older age.  
+
+## \(\beta\) cells secreted nano-vesicular miR-503-322 to enter acinar cells  
+
+We previously verified that \(\beta\) - cell- derived nano- vesicles (βNVs) were secreted from insulin granules and were trafficked into liver and adipose tissues via circulation. \(^{15}\) Whether βNVs
+
+<--- Page Split --->
+
+entered acinar cells was unknown. Here, we shown that acinar cells engulf \(\beta \mathrm{NV}\) s both in vitro and in vivo (Fig. 2A and B); and acinar cell that received \(\beta \mathrm{NV}\) s purified from \(\beta \mathrm{TG}\) islets ( \(\beta \mathrm{TG}\) - \(\beta \mathrm{NV}\) s) had significantly greater levels of miR- 503 than acinar cells receiving \(\beta \mathrm{NV}\) s from wildtype islets (WT- \(\beta \mathrm{NV}\) s, Fig. 2C). However, acinar cells indiscriminately engulfed WT- \(\beta \mathrm{NV}\) s and \(\beta \mathrm{TG}\) - \(\beta \mathrm{NV}\) s in both in vitro and in vivo models.  
+
+To avoid the influence of insulin resistance and hyperglycemia in \(\beta \mathrm{TG}\) mice, \(^{15}\) we constructed RIP2- cre;miR- 503- 322 KI ( \(\beta \mathrm{KI}\) ) mice which were barely diabetic (Fig. S3A- C). \(\beta \mathrm{KI}\) mice also exhibited an exacerbation of caerulein- induced AP compared to littermate controls (Fig. 2D- G), confirming the effects of \(\beta\) - cell miR- 503- 322 to the onset of pancreatitis. Thus, we concluded that \(\beta\) - cell derived nano- vesicles enter acinar cells and drive pancreatitis at a miR- 503- 322- dependent manner in mice.  
+
+# Direct elevation of miR-503-322 in acinar cells triggers both acute and chronic pancreatitis  
+
+Next, we sought to investigate the effects of miR- 503- 322 under inducible global elevation condition by using CAG- creER;miR- 503- 322 KI (CKI). After tamoxifen induction for 3 times, pri- miR- 503 expression levels were significantly elevated in pancreas, skeletal muscle and other metabolic tissues (Fig. S4A and B). Surprisingly, CKI mice started to lose weight and activity, and all committed to death post- induction for 6 days due to severe AP, as observed by significantly increased serum amylase and lipase levels, abdominal infiltration of neutrophils and macrophages, and pancreatic saponification, necrosis and histological analysis (Fig. S4C- I). However, no concomitant histological changes were observed in other
+
+<--- Page Split --->
+
+major abdominal organs (Fig. S4J). Severe AP induced systemically inflammatory responses were shown by inverted serum ratios of neutrophils and lymphocytes, and elevated serum levels of C- reactive protein (Fig. S4K- M). These results validate that the global overexpression of miR- 503- 322 promotes severe AP, indicating the specificity of the miR- 503- 322 for pancreas damage.  
+
+To rule out the contribution of other tissues, Pdx1- cre; miR- 503- 322 KI (heterozygous PKI/WT and homozygous PKI/KI) mice were used to yield high pancreatic- specific expression of miR- 503- 322. The pancreatic Pri- miR- 503 expression was increased in the heterozygous PKI mice (PKI/WT) compared to wildtype controls and was further increased in the homozygous mice (PKI/KI) (Fig. 3A). The PKI/KI mice showed an unexpected weight loss at about 6 weeks of age, while the PKI/WT mice showed no change during natural growth (Fig. 3B and Fig. S5A). The most prominent features of chronic pancreatitis (CP), including pancreatic atrophy, fibrosis, tubular complexes, and inflammatory infiltration were observed in PKI/WT mice, with more severe CP and gross changes in the homozygous PKI/KI mice (Fig. 3C- E, Fig. S 5B- D). Accordingly, PKI/KI mice could not survive for 12 weeks (Fig. 3F).  
+
+PDX1 is a master regulator in pancreas organogenesis while the maturation and identity preservation of islet \(\beta\) - cells and \(\delta\) - cells. \(^{17,18}\) To avoid development defect, inducible acini- specific miR- 503- 322 (Elastase- CreER; miR- 503- 322 KI, EKI) mice were also constructed and overexpression verified post- induction for 3 days (Fig. 3G and H). After tamoxifen injection, the EKI mice showed significantly increased indicators of AP, including macrophage infiltration, tissue damage, and necrosis (Fig. 3I- J, Fig. S5E and F), and had a 20% mortality rate (Fig. 3K). Those mice that survived developed histology of CP one- month post- induction,
+
+<--- Page Split --->
+
+manifested as pancreatic atrophy (Fig. S5G), fibrosis, fat replacement, and acinar- to- ductal metaplasia (ADM, Fig. 3L and M), while a return to normal levels of serum amylase and lipase (Fig. S5E and F). As shown in Fig. S5H- J, EKI female mice presented an AP phenotype similar to that of male mice 5 days after tamoxifen induction.  
+
+The above findings demonstrate that global, pancreatic, and acinar cell- specific overexpression of miR- 503- 322 can directly trigger (severe) acute and chronic pancreatitis in a dose- and tissue- dependent manner.  
+
+# MiR-503-322 knockout alleviated caerulein-induced acute pancreatitis  
+
+The possibility that ablation of miR- 503- 322 could alleviate AP was investigated by the global deletion of miR- 503- 322 (KO) (Fig. S6A). The KO mice were viable and fertile, with normal body weight (Fig. S6B). Histology of the pancreas revealed normal pancreatic morphology (Fig. S6C and D). Challenging the KO and WT mice with caerulein or PBS and assessing for AP severity revealed markedly lower pancreatic edema and amylase and lipase levels in the KO group (Fig. 4A- D). Histological examination revealed reduced pancreatic acinar cell damage, less interstitial expansion (indication of edema), and diminished macrophage infiltration in KO mice during the acute AP phase (Fig. 4E- G). Together, these data demonstrate that the deletion of miR- 503- 322 can significantly alleviate caerulein- induced acute pancreatitis.  
+
+# MiR-503-322 promotes pancreatitis by inhibiting zymogen secretion and acinar-cell proliferation
+
+<--- Page Split --->
+
+Next, we sought to identify the mechanisms by which miR- 503- 322 promotes the development of pancreatitis. Transmission Electron Microscope (TEM) images from the pancreas of the PKI/WT mice revealed an increased number of zymogen granules (Fig. S7A). However, the significantly lower transcript levels of pancreatic enzyme- related genes implied that this did not represent an increased production of zymogen in the acinar cells (Fig. S7B) but was possibly an indication of a secretion defect.  
+
+Therefore, we isolated acini and assessed their secretory ability in response to caerulein. The amylase release was significantly lower from the PKI cells than from the WT cells (Fig. 5A). The acinar cells from aged mice showed a similar response to that of the PKI cells, with a reduced secretion of pancreatic enzymes (Fig. 5B), in agreement with the results of previous studies. \(^{19}\) By contrast, the primary acinar cells from the KO mice showed enhanced amylase secretion (Fig. 5C). The defect of enzyme secretion was attributed to the loss of cytoskeleton modulation from tip to basolateral membranes of acinar cells responding to caerulein (Fig. 5D).  
+
+Enzyme secretion defect may cause trypsinogen activation. We observed that trypsinogen activation in acini was visualized by using rhodamine 110 (BZiPAR) which revealed a clear enrichment of green fluorescence in PKI cells (Fig. 5E), and serum trypsin activity was enhanced in the PKI mice (Fig. 5F). These findings indicate that miR- 503- 322 inhibits pancreatic enzyme secretion and promotes the intracellular accumulation of zymogen. Subsequent zymogen activation in situ may promote pancreas damage of miR- 503- 322 elevated mice.  
+
+Activation of trypsinogen by lysosomal enzymes after fusion of the lysosome is the classical mode of pancreatic enzyme activation during AP. \(^{20,21}\) TEM images of the pancreas
+
+<--- Page Split --->
+
+from PKI mice show morphological signs of this activation, including numerous autophagy vacuoles in the cytoplasm and abundant zymogen granules varying in size and electron density and sometimes fused together to form irregular "lakes" (Fig. S7C). These phenomena suggest a classical activation of intracellular zymogen in the lysosomes of acinar cells that highly express miR- 503- 322. We verified this by inducing pancreatitis in WT and EKI mice by administration of chloroquine, which destroys the acidic environment in autophagic lysosomes (Fig. S7D). The AP phenotype was alleviated in EKI mice treated with chloroquine, as evidenced by a smaller weight loss, reduced serum amylase and lipase levels, and less tissue damage compared to saline- treated control mice, despite a similar pancreas weight (Fig. S7E- J).  
+
+AP significantly stimulates the proliferation of acinar cells almost immediately at the point of injury. Not surprisingly, immunofluorescence staining for PCNA revealed a reduction in the numbers of proliferating acinar cells in the mice expressing high levels of the miR- 503- 322, and an increased proliferation of mesenchymal cells (Fig. 5G- I). Conversely, ablation of miR- 503- 322 enhanced acinar- cell proliferation during the repair phase of caerulein- induced AP (Fig. 5G and J). We also conducted a similar test in aged mice and again observed a significant decrease in acinar- cell proliferation similar to that seen in the high miR- 503- 322 expression model mice (Fig. 5G and K).  
+
+Taken together, these data suggest that miR- 503- 322 suppresses zymogen secretion to initiate acute pancreatitis. Meanwhile, miR- 503- 322 also inhibits regenerative proliferation of acinar cells to promote the formation of chronic pancreatitis.
+
+<--- Page Split --->
+
+# MKNK1 is a target of miR-503-322 and acinar cell-specific restoration of it reverses the phenotype of pancreatitis in mice  
+
+We previously used unbiased proteomics to identify target genes of miR- 503 in regulating peripheral insulin resistance and \(\beta\) - cell dysfunction. \(^{15}\) By analyzing the same proteomics data combined with Targetscan software analysis, five genes (MKNK1, CCNE1, IGF1R, PI3KR1 and INSR) were potential targets (Fig. S8A). After extensively searching and reading literature, we found that the MAP kinase- interacting kinases (MKNK1), mostly expressed in exocrine pancreas might contribute to miR- 503- 322- caused pancreatitis. MKNK1 plays an indispensable role in physiological exocrine secretory response. \(^{23}\) Consistent with published data, phosphorylation of MKNK1 and its downstream eIF4E was increased 4 hr after the first caerulein injection and gradually recovered (Fig. S8B- E). MKNK1 was redistributed to the basolateral region after caerulein administration, assisting acinar- cell secretion (Fig. S8F). Previous studies showed that ablation of MKNK1 results in exacerbation of pancreatitis caused by caerulein due to defects of zymogen secretion and acinar- cell proliferative in mice, \(^{23}\) making us pursue the role of MKNK1 as a target gene of miR- 503- 322.  
+
+Our proteomics analysis showed a decrease in MKNK1 after miR- 503 elevation. Dual- luciferase assay confirmed the regulatory role of miR- 503- 322 on the 3'UTR of Mknk1 gene (Figure 6 A and Fig. S9A and B). Next, immunohistochemistry staining of pancreas sections revealed clear suppression of MKNK1 protein amount in the three miR- 503- 322 overexpressing mouse model. (Fig. S9C), while upregulation of MKNK1 was induced by caerulein in KO mice (Fig. S9D). The protein levels of MKNK1 and its associated P- MKNK1/P- eIF4E signaling were significantly reduced in pancreas of miR- 503- 322 overexpressing model
+
+<--- Page Split --->
+
+mice and aged mice, and by contrast increased in pancreas of miR- 503- 322 knockout mice and aged mice with \(\beta\) - cell specific blocking miR- 503- 322 (Figure 6B and Fig. S9E- G). Taken together, these findings suggest that miR- 503- 322 targets MKNK1- elF4E pathway to inhibit zymogen secretion and acinar- cell proliferation, thereby leading to acute and chronic pancreatitis.  
+
+Next, we tested whether reconstitution of MKNK1 in pancreas could reverse pancreatitis of EKI mice following the schematic diagram (Fig. 6C). We generated an AAV, serotype pancreas (MKNK1- AAV) that directs specific MKNK1 overexpression in the exocrine pancreas. As shown in Fig. S9H, MKNK1 was highly expressed in the acini, but not in the islets of MKNK1- AAV mice. Restoration of MKNK1 also rescued the miR- 503- 322- suppressive protein levels of phos- MKNK1 and phos- elF4E in the EKI pancreas (Fig. 6D and E). Consequently, MKNK1- AAV infected EKI mice showed lessened AP phenotypes compared to Ctr- AAV infected EKI mice. In detail, the loss of body weight, increased serum levels of amylase and lipase, increased number of macrophage infiltration, and tissue damage in Ctr- AAV infected EKI mice were largely reduced in MKNK1- AAV infected EKI mice (Fig. 6F- H).  
+
+On the other hand, inhibition of MKNK1 by a verified inhibitor, CGP 57380 further exacerbated caerulein- caused AP phenotypes, and totally erased miR- 503- 322 knockout driven protective effects (Fig. S10A- G). These results from acinar- cell MKNK1 reconstitution and specific MKNK1 inhibitor support our view that the deficiency of MKNK1 in acini is primarily responsible for the pancreatitis observed in miR- 503- 322 elevated mice.  
+
+Evidence of miR- 503 and MKNK1 in aging- associated pancreatitis changes in the
+
+<--- Page Split --->
+
+## Chinese population  
+
+As the expression of miR- 503 is specifically increased in senescent \(\beta\) cells in mice, we also considered its change in humans. Pancreas sections from elderly adults (EA) showed chronic pancreatitis- like changes, including atrophy of the acinar cells, interstitial expansion, and a marked increase in fibrosis (Fig. 7A), as well as a significant reduction in the proportion of proliferating acinar cells (Fig. 7B), compared to that from young adults (YA). Intriguingly, miRNA in situ hybridization showed greater expression of miR- 503 in islets than in acini in pancreatic sections from EA group (Fig. 7C), whereas expression of miR- 503 was almost undetectable in YA group (Fig. 7C). The expression of MKNK1 was significantly downregulated in the acini from the EA pancreas compared to that from the YA pancreas (Fig. 7D). Moreover, the co- localization of MKNK1 and AMY1 in the young acini was dislocated in the elderly acini (Fig. 7D), indicating an activation of MKNK1 in the EA. Thus, the increased level of miR- 503 in the acini may come from the islet \(\beta\) cells and contribute to the decreased but activated MKNK1 protein in the elderly of Chinese population.  
+
+Numerous studies have reported that exocrine pancreas function is impaired in both healthy and diabetic older adults independent of gastrointestinal disease, judged by serum levels of amylase and maximum bicarbonate concentration. \(^{24,25}\) Consistently, we observed a significantly decreased level of serum amylase in the elderly adult with T2DM (EA+DM) compared to that in YA, moreover, the elderly adults also showed a decreased amylase level (Fig. 7E). Further analysis showed that serum concentration of exosomal miR- 503 was elevated in the elderly compared to that in the young adults and was further elevated in the EA+DM (Fig. 7F). The human subjects displayed negative associations of serum amylase
+
+<--- Page Split --->
+
+levels with both age and serum concentrations of exosomal miR- 503 (Fig. 7G and H). These results support the pancreatic exocrine insufficiency in the elderly and diabetic patients and point out serum concentrations of exosomal miR- 503 as molecular marker of aging- associated pancreatitis in the Chinese population.  
+
+## Discussion  
+
+In this study, we demonstrated that miR- 503- 322 derived from endocrine \(\beta\) - cells promotes aging- associated pancreatitis by targeting MKNK1 in exocrine acinar- cells. miR- 503- 322, which is produced by senescent \(\beta\) - cells, had an in- situ effect in acinar cells that inhibits zymogen secretion and regenerative proliferation. Thus, the miR- 503- 322- MKNK1 axis caused pancreas autodigestion and repairing damage, leading to the onset of acute and chronic pancreatitis in mice. This discovery provides an epigenetic mechanism for pancreatitis and adds to the existing evidence of crosstalk between pancreatic endocrine and exocrine.  
+
+In the past, several studies have described an impairment of exocrine function in T2DM. However, unlike the observations made for CP, these patients showed no obvious symptoms of exocrine disease, such as abdominal pain, and no ductal changes. As a result, a 2016 study proposed the term diabetic exocrine pancreatic disease (DEP) to describe this entity. Several hypotheses have been proposed to explain the features of DEP, including localized insulin deficiency, loss of regulatory functions of islet hormones, pancreatic fibrosis and exocrine atrophy due to vascular pathology, and impaired enteropancreatic reflexes due to neuropathy. However, none of these concepts are sufficient to explain all the pathological findings. Our previous results showed a significant upregulation of islet miR- 503
+
+<--- Page Split --->
+
+expression in patients with T2DM. \(^{15}\) Suggested by our current investigation, the expressed miR- 503 can then enter and accumulate in the exocrine acini, where it triggers damage to some of the acinar cells and causes chronic pancreatitis- like changes in the exocrine pancreas due to repeated pancreas damage.  
+
+Most studies report higher overall morbidity and mortality from pancreatitis in the elderly, \(^{1}\) and several explanations for this phenomenon have been put forward. \(^{32,33}\) Histologically, focal fibrosis also appears to be common in the pancreas of the elderly. \(^{34,35}\) This is consistent with the observations in human pancreatic sections in the present study. Clinical studies have indicated that pancreatic exocrine function is impaired in healthy older individuals without any gastrointestinal disease. \(^{36}\) However, few studies have linked exocrine CP- like changes in healthy older adults and patients with T2DM, and the underlying mechanisms remain to be explored. Our previous findings showed that miR- 503 is significantly upregulated in the islets of both diabetic patients and the elderly, \(^{15}\) and the present results confirmed that islet- derived miR- 503- 322 promotes both acute and chronic damages in the exocrine pancreas and increases mouse mortality with acute and high miR- 503- 322 expression (CKI and PKI/KI mice). Therefore, miR- 503- 322 may be a common pathogenic factor that can explain the higher morbidity and mortality from pancreatitis in the elderly.  
+
+In healthy adults, miR- 503- 322 is expressed mainly in lung, heart, and skeletal muscle progenitor cells. \(^{37}\) Upregulation of miR- 503- 322 occurs in aging acinar cells and is likely to arise from pancreatic \(\beta\) - cells, based on our present observations. Our evidence for this is that blocking miR- 503- 322 in islet \(\beta\) - cells of aging mice alleviated caerulein- induced pancreatitis. Our previous findings revealed that miR- 503 from pancreatic islet \(\beta\) - cells reaches the liver and
+
+<--- Page Split --->
+
+adipose tissue in the form of exosomes, which are known to transport biologically active proteins and miRNAs in their active forms to neighboring cells or distant organs. \(^{38 - 40}\) Thus, the involvement of exosomes in inter-organ and intra-organ crosstalk has been increasingly studied. \(^{41,42}\) Exosomes derived from mesenchymal stem cells have been reported as a new treatment for AP by delivering mitochondria and anti-inflammatory factors. \(^{43,44}\) In addition, senescent \(\beta\) cells have been reported to secrete senescence- associated secretory phenotypes that are rich in EVs and cause dysfunction of adjacent cells through paracrine effects. \(^{45}\) The reported anatomical characteristics of an IAA permits the access of high concentrations of islet- derived miR- 503- 322 to exocrine cells. Indeed, a recent study has determined that islet CCK can promote Kras- driven PDAC development of an endocrine exchange signal other than insulin, \(^{11}\) supporting the existence of endocrine- exocrine crosstalk via IAA. Therefore, we hypothesize that islet- derived miR- 503- 322 is transferred via exosomes and the IAA into acinar cells.  
+
+Our results support MKNK1 as an miR- 503- 322 target gene for the development of pancreatitis. However, MKNK1 knockout mice showed normal pancreatic histology, \(^{23}\) which was inconsistent with the phenotype of AP induced by miR- 503- 322. This normal histologic may reflect the presence of other compensatory pathways in MKNK1- knockout mice as the use of global mouse model. Indeed, the knockout of MKNK1 adds to the growing list of proteins that have a protective role during AP, \(^{46}\) whereas the acute induction of miR- 503- 322 lacks an effective compensatory mechanism. Alternatively, other target genes of miR- 503- 322 co- regulating the development of pancreatitis may exist. Moreover, exosomes carrying miR- 503- 322 may function through inter- organ crosstalk to regulate the severity of AP as shown in CKI
+
+<--- Page Split --->
+
+mice. In addition, tamoxifen administration occasionally causes pancreatitis also reminded us that effect of tamoxifen itself cannot be ignored, although it was added to the control group. \(^{47}\) The mechanisms involved in these possibilities need to be unraveled in further studies.  
+
+In conclusion, we demonstrate the role and a mechanism of action for pancreatic endocrine- derived miR- 503- 322 in promoting pancreatitis in the elderly. Blocking miR- 503- 322 in \(\beta\) - cells of aged mice showed good inhibitory effects on pancreatitis, revealing miR- 503- 322 as a potential therapeutic target for elderly patients with pancreatitis.  
+
+## Methods  
+
+## Human biospecimen acquisition  
+
+For human pancreas sections study, conducted in Organ Transplant Center, Tianjin First Central Hospital, Nankai University, Tianjin, China. A total of 20 healthy individuals were recruited, of these, 10 were young adult (YA, 18- 25 years old) and 10 were the elderly adult (EA, 60- 73 years old). The detailed information of donors was listed in Table. S1. Informed consent was obtained from all patients, and the research protocol was reviewed and approved by the research ethics committee of Tianjin First Central Hospital (No. 2018N112KY).  
+
+For blood sample collection, conducted in the Department of Endocrinology, Geriatric Hospital of Nanjing Medical University, Nanjing, China, 160 individuals were recruited, including 65 YA (18- 37 years old), 65 EA (65- 85 years old), and 30 EA with T2DM (EA+DM). Fasting blood samples, collected from all participants, were centrifuged at 3,000 rpm for 20 min to separate sera and blood cells, the sera were used for miR- 503 concentration analysis. Detailed information of donors including age, gender, fasting blood glucose levels, and history
+
+<--- Page Split --->
+
+of prior diseases were listed in Table S2. The study was approved by the research ethics committee of Nanjing Medical University (2022006), and all the volunteers gave written informed consent.  
+
+## Animal studies  
+
+Animal studies were approved by the Research Animal Care Committee of Nanjing Medical University (IACUC- 1707023 and IACUC- 2004040). Generation of the mouse miR- 503- 322 knock- in mouse (H11- CAG- LSL- miR- 503- 322 Cas9- KI) by CRISPR/Cas9 was outsourced to GemPharmatech Co, Ltd. The mice were created on the C57BL/6J genetic background. The gRNA (5'- CTGAGCCAACAGTGGTAGTA - 3') to the Hipp11 (H11) locus, the donor vector containing the "CAG- loxP- Stop- loxP- mouse miR- 503- 322- polyA" cassette, and Cas9 mRNA were co- injected into fertilized mouse eggs to generate targeted conditional knock- in offspring. Rat insulin 2 promoter (RIP2)- Cre (JAX:003573), CAG- CreER (JAX:004453) and PDX1- Cre (JAX:014647) mice were obtained from the Jackson Laboratory. Elastase (ELA)- CreER mice were obtained from Dr. Xianghui Fu (Professor of the West China Hospital, Sichuan University). We then crossed KI mice with CAG- CreER, PDX1- Cre, ELA- CreER, and RIP2- Cre mice respectively, to obtained global inducible (CKI), pancreas- specific (PKI), acinar cell- specific inducible (EKI), and islet \(\beta\) - cell- specific ( \(\beta\) KI) overexpression miR- 503- 322 mice. Details on each animal strain were listed in Table S3. EKI or CKI and their litter control mice were injected intraperitoneally with tamoxifen solution, 100 mg/kg, in corn oil, for three consecutive days to induce miR- 503- 322 overexpression in acinar cells or the whole body, respectively. The control groups used their respective littermates and were genotyped as KI- positive and Cre- negative. All experimental mice were heterozygous except for PKI mice, which included both
+
+<--- Page Split --->
+
+homozygotes and heterozygotes. MiR- 503 transgenic mice (βTG) and miR- 503- 322 global deletion mice (KO) were also generated by GemPharmatech Co, Ltd. Refer to our previous findings for the exact construction workflow. \(^{15}\) Aged C57BL6/J mice were purchased from GemPharmatech Co, Ltd.  
+
+The animals were randomly allocated to experimental groups, at least 4 per group, not according to genotype to minimize potential confounding factors. Male mice were mostly used in this study, and female mice were also involved to rule out the sex bias, as described in the Fig. legends. Mice were housed in a temperature- and humidity- controlled environment (23- 25°C, 12- h light/dark cycle, 60- 70% humidity) in a specific pathogen- free facility at Nanjing Medical University and provided with free access to commercial rodent chow and tap water. Health was monitored at least weekly by weight, food and water intake, and general assessment of animal activity, panting, and fur condition. Mice were euthanized by \(\mathrm{CO_2}\) asphyxiation when met euthanasia criteria.  
+
+Adult animals of both genders were used in tamoxifen induction studies. Collected blood serum was used to measure amylase and lipase. The pancreatic tissue was collected and immediately embedded in optimum cutting temperature compound for hematoxylin and eosin staining, evaluation of necrosis, and immunohistochemistry. Necrosis and acinar cell damage quantified by morphometry as described. \(^{48}\) Tissue damage was quantified using scoring system as describe by Schmidt et al. \(^{49}\)  
+
+## Pancreatic Acinar Cell Experiments  
+
+Mouse pancreatic acini were isolated using the standard collagenase digestion protocol, as previously described. \(^{50}\) Acini were isolated and left to recover for 30 min at 37°C before
+
+<--- Page Split --->
+
+stimulation with the indicated concentrations of caerulein (MCE, Shanghai, China) to assess the secretory capacity. The supernatant for amylase activity was analyzed with a commercial kit (JianCheng Bioengineering Institute, Nanjing, China) and the percentage of amylase secretion was calculated. To visualize trypsinogen activation in acinar cells, freshly prepared acini were loaded with active trypsin enzyme substrate BZiPAR (10 μM) (Invitrogen, America) and incubated for 30 min. Images were captured and analyzed by a confocal laser scanning microscope (Olympus FV1200). The image fluorescence intensity was analyzed with ImageJ software.  
+
+## Induction of Murine Pancreatitis  
+
+Caerulein was solubilized in phosphate- buffered saline at a final concentration of 15 mg/mL. Experimental mice were challenged with caerulein (50 mg/kg, intraperitoneal injection, once an hr, 6 times) to induce AP. Control animals received an equal amount of saline. The parameters of AP were assessed 2 hrs after the last caerulein treatment. Edema, serum lipase (ElabScience, Wuhan, China), amylase and trypsin activity (JianCheng Bioengineering Institute, Nanjing, China) were analyzed as parameters of pancreatitis.  
+
+## Histopathology, Immunohistochemistry and Immunofluorescence  
+
+Mice were euthanized by \(\mathrm{CO_2}\) asphyxiation and tissue was dissected, rinsed in PBS and fixed overnight in \(4\%\) paraformaldehyde (Servicebio). Paraffin embedding, serial sectioning, H&E and Masson staining of all samples were commissioned from Servicebio Technologies. After dewaxing and antigen retrieval, the pancreatic paraffinic sections were incubated with primary antibodies overnight at \(4^{\circ}\mathrm{C}\) . According fluorescent- conjugated secondary antibodies (Proteintech) were used for multiple labeling and the nuclei were stained with Hoechst 33342
+
+<--- Page Split --->
+
+(5 μg/mL) (Sigma-Aldrich). Fluorescent images were visualized by a confocal laser scanning microscope (Olympus FV1200). Immunohistochemistry staining was labeled with DAB substrate system (BCA Kit) (Gene Tech) and positive labeled cells were captured by a light microscope (Leica, Germany). Quantification was done with at least three mice per group, three sections per mouse (50 μm apart), and at least 10 microscopic fields per section. The antibodies were listed in Table S4.  
+
+## Pancreatic Ductal Infusion of Adeno-associated Viral (AAV) Vectors  
+
+Pancreatic ductal infusion was performed following the standard surgical protocol, as we previously described.51 Serotype pancreas of PAAV- CMV- MCS- EF1- mNeonGreen- WPRE (Ctr- AAV) and PAAV- CMV- MKNK1- flag- EF1- mNeonGreen- WPRE (MKNK1- AAV) were provided by the company of OBIOTechnology Co, Ltd. Serotype 2/8 under insulin2 promoter of HBAAV2/8- insulin2- zsGreen (Ctr) and HBAAV2/8- insulin2- mmu- miR- 503/322- 5p- sponge- zsGreen (SP) were provided by the company of Hanheng Biotechnology Co, Ltd. AAV titer of \(10^{11} / \mathrm{mL}\) in PBS, 100 μL total volume in 20 g body weight mice was infused at a rate of 6 μL/min. After infusion and suture, surgical mice were placed on a heated pad (37°C) until full recovery. Ketoprofen (Sigma, k1751) at a dose of 5 mg/kg once per day was given continuously for 3 d for post- surgery analgesia.  
+
+## Locked Nucleic Acid (LNA)-Based in situ Hybridization of miR-503  
+
+Locked nucleic acid- based in situ assay was introduced to detect miR- 503 in human pancreas sections. Double- labeled with carboxyfluorescein (FAM), LNA enhanced probes including U6 snRNA control probe, negative scramble- miR control and has- miR- 503 were constructed by QIAGEN. The assay was performed according to the manufacturer's protocol.52 In short,
+
+<--- Page Split --->
+
+sample slides were deparaffinized in xylene and ethanol solutions at room temperature (15- 25°C) and digested with Proteinase K reagent for 10 min at 37°C. After washing, each sample was reacted with 50 μL of hybridization mix (1 nM LNA U6 snRNA probe, 40 nM double- FAM LNA miR- 503 probe and scramble- miR) in a programmed hybridizer for 1 hr. After strictly washing and blocking, the samples were incubated with anti- FAM reagent for 1 hr and labeled with alkaline phosphatase substrates for 2 hrs. The nuclei were labeled with Nuclear Fast Red. All sample slices were visualized by light microscopy.  
+
+## Islet-derived Exosomes Isolation and Fluorescence Labelling  
+
+Freshly isolated islets were cultured in serum- exosome- free medium (11.1 mM glucose) for 7 days, with the medium replaced and collected every 24 hr. The medium was first centrifuged at 700 g for 5 min and at 10,000 g for 1 hr to remove cell debris. Next, the collected supernatant was centrifuged as described previously to obtain exosomes of good purity. For cell imaging, exosomes were labelled with PKH67 (Sigma- Aldrich) for 1 hr and then washed three times with PBS. PKH67- labelled exosomes (100 μg/35 mm culture dish) were resuspended in PBS and then incubated with freshly isolated acini for 8 h. The acini were then stained with phalloidin (MCE) for 15 min and Hoechst 33342 for 8 min. Images were taken and analyzed by a confocal laser scanning microscope (Olympus FV1200).  
+
+## Plasmid Construction and Dual-Luciferase Reporter Assay  
+
+The wild- type and mutant 3' UTR- luciferase constructs containing miR- 503- 322 binding site of mouse Mknk1 were generated by annealing and cloning the short sequences into pMIR- REPORT Luciferase miRNA Expression Reporter Vector (Ambion) between the Spel and HindIII sites. Primer sequences were listed in Table S5. Luciferase activities were measured
+
+<--- Page Split --->
+
+using the Dual- Glo Luciferase Assay System (Promega, America) on a TD- 20/20 Luminometer (Turner BioSystems, America) according to the manufacturer's protocols.  
+
+## Quantitative RT-PCR  
+
+Total RNA was extracted from cells and tissues using Trizol reagent (Invitrogen). cDNA was synthesized from total RNA using a ReverTra Ace Kit (TOYOBO, Japan). qPCR of Pri- miRNA and miRNA were performed using the THUNDERBIRD probe qPCR Mix (TOYOBO, Japan), and SYBR Green qPCR Master Mix (Vazyme, China) for mRNA on Roche LightCycle480 II Sequence Detection System (Roche, Switzerland). Primers of qPCR for pri- miRNA and miRNA were purchased from Thermofisher Co., Ltd, other primer sequences were available in Table S5.  
+
+## Western Blot Analysis  
+
+Cells or tissues were lysed with ice- cold RIPA buffer (Thermo Fisher Scientific), supplemented with 0.5 mM EDTA and Halt protease/phosphatase inhibitor cocktail (Thermo Fisher Scientific), rotated at \(4^{\circ}C\) for 15- 30 minutes to mix, and centrifuged at maximum speed for 15 min to collect whole cell lysates. Protein concentration was measured with the BCA protein assay (Takara). Thirty mg of total protein per sample was loaded into \(4\% - 12\%\) NuPAGE Tris- Bis (Thermo Fisher Scientific) gradient gels and separated by SDS- PAGE. Proteins were transferred to PVDF membranes (Millipore Billerica) and blocked with \(5\%\) milk. Beta- actin and \(\alpha\) - tubulin were used as loading controls. Primary antibodies were detected with HRP- conjugated (Sigma- Aldrich) secondary antibodies for chemiluminescent detection (Perkin Elmer ECL). Protein quantification was performed by Image J (NIH Image). Key reagents and antibodies were listed in Table S6.
+
+<--- Page Split --->
+
+## Statistical Analysis  
+
+All results were expressed as mean \(\pm\) SEM. Results were analyzed with GraphPad Prism software (version 8.3.0, San Diego, CA, USA). Two- tailed unpaired Student's \(t\) test was used for the comparison of two sets of Gaussian distributed data. Gaussian distributions were tested using the Kolmogorov- Smirnov test. Data conforming to the Gaussian distribution were analyzed with parametric tests, while data with non- Gaussian distribution were analyzed with nonparametric tests. To compare the difference between two groups, Student \(t\) tests were performed. For non- Gaussian distributed data, the Mann- Whitney \(U\) rank sum test was used for one- way analysis. Unless otherwise stated, differences in means between multiple groups of Gaussian- distributed data were analyzed by ordinary one- way analysis of variance followed by Bonferroni's multiple comparisons, and for non Gaussian- distributed data, Kruskal- Wallis test was used. Two- way ANOVA followed by Bonferroni's multiple comparisons was used for two- way analysis. Linear regression analysis was performed using GraphPad Prism software. Pearson correlation analysis was used to test for correlations. In all analyses, \(P < 0.05\) was considered statistically significant.  
+
+## References  
+
+1 Baeza- Zapata, A. A., Garcia- Compean, D., Jaquez- Quintana, J. O. & Collaborators. Acute Pancreatitis in Elderly Patients. Gastroenterology 161, 1736- 1740, doi:10.1053/j.gastro.2021.06.081 (2021).  2 Kayar, Y., Dertli, R. & Konur, S. Clinical outcomes of acute pancreatitis in elderly patients: An experience of single tertiary center. Pancreatology 20, 1296- 1301, doi:10.1016/j.pan.2020.06.006 (2020).  3 Satis, H., Kayahan, N., Sargin, Z. G., Karatas, A. & Celiker, D. Evaluation of the clinical course and prognostic indices of acute pancreatitis in elderly patients: a prospective study. Acta Gastro- Enterologica Belgica 83, 413- 417 (2020).  4 Grunewald, M. et al. Counteracting age- related VEGF signaling insufficiency promotes healthy aging and extends life span. Science 373, doi:10.1126/science.abc8479 (2021).
+
+<--- Page Split --->
+
+5 Investigators, F. S. Effects of long- term fenofibrate therapy on cardiovascular events in 9795 people with type 2 diabetes mellitus (the FIELD study): randomised controlled trial (vol 366, pg 1849, 2005). Lancet 368, 1420- 1420 (2006). 6 Girman, C. J. et al. Patients with type 2 diabetes mellitus have higher risk for acute pancreatitis compared with those without diabetes. Diabetes Obes Metab 12, 766- 771, doi:10.1111/j.1463- 1326.2010.01231. x (2010). 7 Gonzalez- Perez, A., Schlienger, R. G. & Rodriguez, L. A. Acute pancreatitis in association with type 2 diabetes and antidiabetic drugs: a population- based cohort study. Diabetes Care 33, 2580- 2585, doi:10.2337/dc10- 0842 (2010). 8 Mizumoto, R. & Sakoguchi, T. Relationship between endocrine and exocrine glands of the pancreas. Nihon rinsho. Japanese journal of clinical medicine Suppl, 2250- 2251 (1978). 9 Williams, J. A. & Goldfine, I. D. The insulin- pancreatic acinar axis. Diabetes 34, 980- 986, doi:10.2337/diabetes.34.10.980 (1985). 10 Barreto, S. G., Carati, C. J., Toouli, J. & Saccone, G. T. The islet- acinar axis of the pancreas: more than just insulin. Am J Physiol Gastrointest Liver Physiol 299, G10- 22, doi:10.1152/ajpg.00077.2010 (2010). 11 Chung, K. M. et al. Endocrine- Exocrine Signaling Drives Obesity- Associated Pancreatic Ductal Adenocarcinoma. Cell 181, 832- +, doi:10.1016/j.cell.2020.03.062 (2020). 12 Schludi, B. et al. Islet inflammation and ductal proliferation may be linked to increased pancreatitis risk in type 2 diabetes. JCI Insight 2, doi:10.1172/jci.insight.92282 (2017). 13 Li, X. et al. Islet alpha- cell Inflammation Induced By NF- kappaB inducing kinase (NIK) Leads to Hypoglycemia, Pancreatitis, Growth Retardation, and Postnatal Death in Mice. Theranostics 8, 5960- 5971, doi:10.7150/thno.28960 (2018). 14 Li, J. et al. Pancreatic beta cells control glucose homeostasis via the secretion of exosomal miR- 29 family. J Extracell Vesicles 10, e12055, doi:10.1002/jev2.12055 (2021). 15 Zhou, Y. et al. \(\beta\) - Cell miRNA- 503- 5p Induced by Hypomethylation and Inflammation Promotes Insulin Resistance and \(\beta\) - Cell Decompensation. Diabetes 73, 57- 74, doi:10.2337/db22- 1044 (2024). 16 Liang, R. et al. H19X- encoded miR- 322(424)/miR- 503 regulates muscle mass by targeting translation initiation factors. J Cachexia Sarcopenia Muscle 12, 2174- 2186, doi:10.1002/jcsm.12827 (2021). 17 Hermann, P. C. et al. Nicotine promotes initiation and progression of KRAS- induced pancreatic cancer via Gata6- dependent dedifferentiation of acinar cells in mice. Gastroenterology 147, 1119- 1133 e1114, doi:10.1053/j.gastro.2014.08.002 (2014). 18 Weidemann, B. J. et al. Repression of latent NF- kappaB enhancers by PDX1 regulates beta cell functional heterogeneity. Cell Metab 36, 90- 102 e107, doi:10.1016/j.cmet.2023.11.018 (2024). 19 Jiang, Z. E. et al. Age- associated changes in pancreatic exocrine secretion of the isolated perfused rat pancreas. Lab Anim Res 29, 19- 26, doi:10.5625/lar.2013.29.1.19 (2013). 20 Chvanov, M. et al. Intracellular rupture, exocytosis and actin interaction of endocytic vacuoles in pancreatic acinar cells: initiating events in acute pancreatitis. Journal of Physiology- London 596, 2547- 2564, doi:10.1113/jp275879 (2018). 21 Sendler, M. et al. Cathepsin B- Mediated Activation of Trypsinogen in Endocytosing Macrophages Increases Severity of Pancreatitis in Mice. Gastroenterology 154, 704- 718
+
+<--- Page Split --->
+
+e710, doi:10.1053/j.gastro.2017.10.018 (2018).22 Desai, B. M. et al. Preexisting pancreatic acinar cells contribute to acinar cell, but not islet beta cell, regeneration. J Clin Invest 117, 971- 977, doi:10.1172/JCI29988 (2007).23 Cendrowski, J. et al. Mnk1 is a novel acinar cell- specific kinase required for exocrine pancreatic secretion and response to pancreatitis in mice. Gut 64, 937- 947, doi:10.1136/gutjnl- 2013- 306068 (2015).24 Hardt, P. D. et al. Pancreatic exocrine function in patients with type 1 and type 2 diabetes mellitus. Acta Diabetol 37, 105- 110 (2000).25 Yilmaztepe, A., Ulukaya, E., Ersoy, C., Yilmaz, M. & Tokullugil, H. A. Investigation of fecal pancreatic elastase- 1 levels in type 2 diabetic patients. The Turkish Journal of Gastroenterology : the Official Journal of Turkish Society of Gastroenterology 16, 75- 80 (2005).26 Hardt, P. D. et al. Chronic pancreatitis and diabetes mellitus. A retrospective analysis of 156 ERCP investigations in patients with insulin- dependent and non- insulin- dependent diabetes mellitus. Pancreatology 2, 30- 33, doi:10.1159/000049445 (2002).27 Majumder, S. et al. Diabetes Mellitus is Associated With an Exocrine Pancreatopathy (EP) That is Distinct From Chronic Pancreatitis (CP). Pancreas 44, 1395- 1395 (2015).28 Zechner, D. et al. Diabetes aggravates acute pancreatitis and inhibits pancreas regeneration in mice. Diabetologia 55, 1526- 1534, doi:10.1007/s00125- 012- 2479- 3 (2012).29 Zechner, D. et al. Diabetes increases pancreatic fibrosis during chronic inflammation. Exp Biol Med (Maywood) 239, 670- 676, doi:10.1177/1535370214527890 (2014).30 Rupnik, M. S. & Hara, M. Local dialogues between the endocrine and exocrine cells in the pancreas. Diabetes, doi:10.2337/db23- 0760 (2024).31 Whitcomb, D. C., Buchner, A. M. & Forsmark, C. E. AGA Clinical Practice Update on the Epidemiology, Evaluation, and Management of Exocrine Pancreatic Insufficiency: Expert Review. Gastroenterology 165, 1292- 1301, doi:10.1053/j.gastro.2023.07.007 (2023).32 Kudoh, A., Katagai, H., Takazawa, T. & Matsuki, A. Plasma proinflammatory cytokine response to surgical stress in elderly patients. Cytokine 15, 270- 273, doi:10.1006/cyto.2001.0927 (2001).33 Barbeiro, D. F., Koike, M. K., Coelho, A. M. M., da Silva, F. P. & Machado, M. C. C. Intestinal barrier dysfunction and increased COX- 2 gene expression in the gut of elderly rats with acute pancreatitis. Pancreatology 16, 52- 56, doi:10.1016/j.pan.2015.10.012 (2016).34 Kim, C. I. Clinicopathological study of pancreatic fibrosis in diabetes mellitus. Medical journal of Osaka University 28, 23- 31 (1977).35 Lohr, J. M., Panic, N., Vujasinovic, M. & Verbeke, C. S. The ageing pancreas: a systematic review of the evidence and analysis of the consequences. J Intern Med 283, 446- 460, doi:10.1111/jaim.12745 (2018).36 Sato, T. et al. Age- related changes in normal adult pancreas: MR imaging evaluation. Eur J Radiol 81, 2093- 2098, doi:10.1016/j.ejrad.2011.07.014 (2012).37 Shen, X. et al. miR- 322/- 503 cluster is expressed in the earliest cardiac progenitor cells and drives cardiomyocyte specification. Proc Natl Acad Sci U S A 113, 9551- 9556, doi:10.1073/pnas.1608256113 (2016).
+
+<--- Page Split --->
+
+38 Wang, G. et al. Tumour extracellular vesicles and particles induce liver metabolic dysfunction. Nature 618, 374- 382, doi:10.1038/s41586- 023- 06114- 4 (2023).39 Li, K. et al. Salivary Extracellular MicroRNAs for Early Detection and Prognostication of Esophageal Cancer: A Clinical Study. Gastroenterology 165, 932- 945 e939, doi:10.1053/j.gastro.2023.06.021 (2023).40 Jiang, T. Y. et al. PTEN Deficiency Facilitates Exosome Secretion and Metastasis in Cholangiocarcinoma by Impairing TFEB- mediated Lysosome Biogenesis. Gastroenterology 164, 424- 438, doi:10.1053/j.gastro.2022.11.025 (2023).41 Wang, J. et al. Extracellular vesicles mediate the communication of adipose tissue with brain and promote cognitive impairment associated with insulin resistance. Cell Metab 34, 1264- 1279 e1268, doi:10.1016/j.cmet.2022.08.004 (2022).42 Cao, M. et al. Cancer- cell- secreted extracellular vesicles suppress insulin secretion through miR- 122 to impair systemic glucose homeostasis and contribute to tumour growth. Nat Cell Biol 24, 954- 967, doi:10.1038/s41556- 022- 00919- 7 (2022).43 Li, S. et al. Hair follicle- MSC- derived small extracellular vesicles as a novel remedy for acute pancreatitis. J Control Release 352, 1104- 1115, doi:10.1016/j.jconrel.2022.11.029 (2022).44 Hu, Z. et al. MSCs Deliver Hypoxia- Treated Mitochondria Reprogramming Acinar Metabolism to Alleviate Severe Acute Pancreatitis Injury. Adv Sci (Weinh) 10, e2207691, doi:10.1002/advs.202207691 (2023).45 Ryaboshapkina, M. et al. Characterization of the Secretome, Transcriptome, and Proteome of Human beta Cell Line EndoC- betaH1. Mol Cell Proteomics 21, 100229, doi:10.1016/j.mcpro.2022.100229 (2022).46 Saluja, A., Dudeja, V., Dawra, R. & Sah, R. P. Early Intra- Acinar Events in Pathogenesis of Pancreatitis. Gastroenterology 156, 1979- 1993, doi:10.1053/j.gastro.2019.01.268 (2019).47 Kim, Y. A. et al. Severe acute pancreatitis due to tamoxifen- induced hypertriglyceridemia with diabetes mellitus. Chinese Journal of Cancer Research 26, 341- 344, doi:10.3978/j.issn.1000- 9604.2014.05.01 (2014).48 Bhatia, M. et al. Role of substance P and the neurokinin 1 receptor in acute pancreatitis and pancreatitis- associated lung injury. Proceedings of the National Academy of Sciences of the United States of America 95, 4760- 4765, doi:10.1073/pnas.95.8.4760 (1998).49 Schmidt, J. et al. A better model of acute pancreatitis for evaluating therapy. Ann Surg 215, 44- 56 (1992).50 Grosfils, K., Metioui, M., Tiouli, M. & Dehaye, J. P. Isolation of rat pancreatic acini with crude collagenase and permeabilization of these acini with streptolysin O. Research communications in chemical pathology and pharmacology 79, 99- 115 (1993).51 Xiao, X. et al. Pancreatic cell tracing, lineage tagging and targeted genetic manipulations in multiple cell types using pancreatic ductal infusion of adeno- associated viral vectors and/or cell- tagging dyes. Nat Protoc 9, 2719- 2724, doi:10.1038/nprot.2014.183 (2014).52 Obernosterer, G., Martinez, J. & Alenius, M. Locked nucleic acid- based in situ detection of microRNAs in mouse tissue sections. Nature Protocols 2, 1508- 1514, doi:10.1038/nprot.2007.153 (2007).
+
+<--- Page Split --->
+
+## Acknowledgments  
+
+We are indebted to Scribendi Inc. (Chatham, ON, Canada) for proofreading the manuscript. This work was supported by grants from the National Natural Science Foundation of China (82070843 and 82270844 to Y- X.Z; 82330027 to X.H).  
+
+## Author contributions  
+
+Conceptualization: Y- X.Z., X.H. and L.L.; Methodology: K- R.L., T- T.L. Y.Z. and X.X; Investigation: Y.Z., W.T. and S- S.W.; Visualization: K- R.L. and Y- X.Z.; Supervision: Y.Z., Y- T.L., and X- A.C.; Writing—original draft: K- R.L. and Y- X.Z.; Writing—review & editing: K- R.L., X.H., S- J.P. and Y- X.Z. All authors discussed the results and commented on the manuscript.  
+
+## Competing interests  
+
+Authors declare that they have no competing interests.  
+
+## Data and materials availability:  
+
+All data are available in the main text or the supplementary materials. All other supporting data in this study are available from the Lead Contact (Yunxia Zhu, zhuyx@njmu.edu.cn) on request.
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Fig. 1. Senescent \(\beta\) -cell–derived miR-503-322 promoted pancreatitis in mice. (A) qPCR analysis of Pri-miR-503 expression in islets and acini of 20-week-old control (WT) and \(\beta\) -cell specific miR-503 transgenic (βTG) male mice. \(n = 4\) . (B) qPCR analysis of miR-503 expression in acini of WT and βTG mice. \(n = 3\) . (C) qPCR analysis of Pri-miR-503 expression in pancreas, islets and acini of 12 weeks and 1.5 years old male mice. \(n = 5\) . (D) qPCR analysis of miR-503 and miR-322 expression in islets and acini of 12 weeks and 1.5 years old male mice, respectively. \(n = 5\) . (E) Schematic flow diagram of sponge \(\beta\) -cell miR-503-322 and induced pancreatitis in aged male mice. The 1.4-years C57BL/6J male mice were randomly divided into two groups. The control group (Ctr, \(n = 4\) ) and the experimental group (SP, \(n = 5\) ) were respectively injected with ctr-AAV and miR-503-322 sponge-AAV through pancreatic ductal Infusion. Two months later, AP was induced by intraperitoneal injection (ip.) of caerulein (50 \(\mu \mathrm{g / kg}\) , hourly for 6 consecutive times), and pancreatitis parameters were detected 2 hours </center>
+
+<--- Page Split --->
+
+after the last injection. (F) Representative sections of ZsGreen (green), insulin (red) and nucleus (blue) immunofluorescence co- staining in pancreas of mice 1 month after AAV injection. (G) qPCR analysis of pancreatic miR- 503 and miR- 322 in the Ctr and SP groups 2 months after AAV injection. (H- J) Pancreatic weights after calibration with body weight (H), serum amylase and lipase (I), representative histologic sections of H&E and F4/80 immunohistochemistry (IHC) of pancreas, pancreatic histological scores and quantitation of the number of F4/80 positive cells in pancreatic sections under 200x microscopic view (J) in the Ctr and SP groups after caerulein (50 μg/kg) induced. Data are presented as Mean ± SEM. Data were analyzed using one- way ANOVA followed by Bonferroni's post hoc (A, C, D and I) or unpaired two- tailed Student's t test (B, G, H and J). NS, Not Significant; \(*P < 0.05\) ; \(**P < 0.01\) ; \(***P < 0.001\) .
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Fig. 2. \(\beta\) cells secreted nano-vesicular miR-503-322 to enter acinar cells. (A) Experimental scheme: PKH67-labeled nano-vesicles derived from islets of WT (WT-βNVs) or βTG islets (βTG-βNVs) were co-incubated with fresh acini for 8 hours. Representative confocal images of PKH67, phalloidin and nuclei in primary acini and quantitation of relative PKH67 fluorescence intensity. (B) Experimental scheme: PKH67-labeled WT-βNVs or βTG-βNVs were infusion into the C57 male mouse pancreas via pancreatic ductal. The pancreas was harvested after 12 hours and stained with frozen sections for amylase and then visualized. Representative confocal images of PKH67, amylase and nuclei of pancreas and quantitation of relative PKH67 fluorescence intensity. (C) qPCR analysis of miR-503 expression in acini of received βNVs. n=3. (D-G) Pancreatitis parameters assay after initial caerulein (50 μg/kg) injection 7 hours in 12-week-old control (WT) and β-cell specific miR-503-322 knock-in (βKI) female mice. H&E and F4/80 immunohistochemistry (IHC) of pancreatic sections (D), quantitation of the number of F4/80 positive cells in pancreatic sections under 200x microscopic view (E), pancreatic histological scores (F) and level of serum amylase (G). Arrows indicate the macrophages. Data are presented as Mean ± SEM. n=5 mice/group. Data were analyzed using one-way ANOVA followed by Bonferroni's post hoc (G) or unpaired two-tailed Student's \(t\) test (A, B, C, E and F). NS, Not Significant; \(*P < 0.05\) ; \(**P < 0.01\) ; \(***P < 0.001\) . </center>
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Fig. 3. Direct elevation of miR-503-322 in acinar cells triggers both acute and chronic pancreatitis. (A) qPCR analysis for pancreatic Pri-miR-503 expression in 8-week-old control (WT), PKI heterozygous (PKI/WT) and PKI homozygous (PKI/KI) male mice. n=5. (B) Weight </center>
+
+<--- Page Split --->
+
+monitoring in WT, PKI/WT and PKI/KI male mice. n=4. (C) Photograph of the pancreas of 8- week- old WT, PKI/WT and PKI/KI male mice. Lines indicate the pancreas. (D) Representative sections of H&E and masson dyeing of 8- week- old PKI male mice. Representative sections of F4/80 (red) and CK19 (red) immunofluorescence staining in pancreas of 8- week- old PKI male mice. (E) Pancreatic weights after calibration with body weight in 8- week- old PKI male mice. n≥3. (F) Survival curves for WT, PKI/WT and PKI/KI male mice. (G) Schematic of acinar cell specific miR- 503- 322 knock- in (EKI) mice: 6- 8 weeks WT and EKI male or female mice were injected intraperitoneally (ip.) with tamoxifen solution, 100 mg/kg, in corn oil, for three consecutive days and sacrificed three days after the last tamoxifen injection. (H) qPCR analysis of Pri- miR- 503 in acinar cells of WT and EKI male mice after 3 times tamoxifen induced. n=4. (I) Representative sections of pancreas of H&E, receptor- interacting serine-threonine kinase 3 (RIPK3) immunohistochemistry and immunofluorescence staining of F4/80 after first tamoxifen injection 5 days in WT and EKI male mice. Arrows indicate the macrophages. (J) Pancreatic histological scores, quantitation of average optical density of RIPK3 and the number of F4/80 positive cells in pancreatic sections under 600x microscopic view for Fig. I. n=5. (K) Survival curves for WT and EKI male and female mice. (L) Pancreatic H&E of WT and EKI female mice after first tamoxifen injection 28 days. (M) Representative sections of pancreatic masson dyeing from EKI male mice after first tamoxifen injection 28 days. ADM, Acinar- to- ductal metaplasia. Data are presented as Mean ± SEM. Data were analyzed using one- way ANOVA followed by Bonferroni's post hoc (A and E), GEE followed by Tukey's post hoc test (B), unpaired two- tailed Student's t test (H and J) or Survival curve analyses (F and K). NS, Not Significant; \(*P< 0.05\) ; \(**P< 0.01\) ; \(***P< 0.001\) .
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Fig. 4. MiR-503-322 knockout alleviated caerulein-induced acute pancreatitis. (A) Schematic of caerulien-induced AP on 12-week WT and KO male mice. \(n = 5\) . (B-E) Pancreatic weights after calibration with body weight (B), serum amylase (C) and lipase (D) levels, histological score of the pancreas (E) after PBS or caerulein treatment groups. (F and G) Representative sections of pancreatic H&E (F) and immunofluorescence staining of F4/80 (green) after PBS or caerulein treatment 7 hours in WT and KO male mice. Quantitation of the number of F4/80 positive cells in pancreatic sections under 600x microscopic view (G). Arrows indicate the macrophages. Data are presented as Mean ± SEM. Data were analyzed using one-way ANOVA followed by Bonferroni's post hoc. \(n = 4 - 5\) mice/group. NS, Not Significant; \(*P < 0.05\) ; \(**P < 0.01\) ; \(***P < 0.001\) . </center>
+
+<--- Page Split --->
+![](images/Figure_5.jpg)
+
+<center>Fig. 5. MiR-503-322 promotes pancreatitis by inhibiting zymogen secretion and acinar-cell proliferation. (A) Extraction of fresh acinar cells from 8-week-old WT and PKI/WT male mice, in vitro stimulation with different concentrations of caerulein for 30 minutes and determination of amylase content in the supernatant. See Materials and Methods for details. n=3. (B and C) Amylase levels after calibration of total content release from acinar cells of 12-week WT and KO male mice (B) and 12-week and 1.5-year C57BL/6J male mice (C) after 30 min of stimulation with caerulein. n=3. (D) After 48 hours of induction by tamoxifen in WT or EKI mice, pancreatic acini were isolated and incubated with or without caerulein (0.01 μM) for </center>
+
+<--- Page Split --->
+
+30 minutes. At the indicated times, cells were harvested, stained for F-actin with phalloidin (red) and nuclei (blue), and evaluated by laser confocal scanning microscopy. Representative fluorescence micrographs of untreated acini (CTR) and acini pretreated by caerulin stimulation (CER). (E) Representative confocal images of WT and PKI/WT male mice acini after incubation with BziPAR for 30 min at \(37^{\circ}C\) and quantification of fluorescence intensity. (F) Detection of serum trypsin activity levels in 16-week WT and PKI/WT male mice and EKI after Tamoxifen injection 5 days. n=5. (G) Quantitation of the number of proliferating acinar cells of 8-week WT and PKI/WT male mice; 28 days after tamoxifen induction in WT and EKI male mice; 4 days after caerulein-induced AP in 12-week WT and KO male mice and 12-week and 1.5-year-old male mice. n=3-5 mice per group. (H-J) Representative sections of immunofluorescence staining of amylase (red) and PCNA (green) in pancreatic sections from 8-week WT and PKI/WT male mice (H), 28 days after tamoxifen induction in WT and EKI male mice (I), 4 days after caerulein-induced AP in 12-week WT and KO male mice (J). Arrows indicate proliferating acinar cells; asterisks are proliferating interstitial cells. (K) Representative sections of immunofluorescence staining of PCNA (green) in pancreatic sections from 12-week and 1.5-year-old male mice. PCNA, proliferating cell nuclear antigen. Arrows indicate proliferating acinar cells. Data are presented as Mean ± SEM. Data were analyzed using GEE followed by Tukey's post hoc test (A, B and C) or unpaired two-tailed Student's t test (F and G). \(^{*}P< 0.05\) ; \(^{**}P< 0.01\) ; \(^{***}P< 0.001\) .
+
+<--- Page Split --->
+![](images/Figure_6.jpg)
+
+<center>Fig. 6. MKNK1 is a target of miR-503-322 and acinar cell-specific restoration of it reverses the phenotype of pancreatitis in mice. (A)The MKNK1 network was predicted based on the common signature from the Ingenuity database overlaid with microarray data from miR-503-overexpressing mouse pancreatic \(\beta\) cell line MIN6 cells with a 1.5-fold change cutoff compared with negative control cells. (B) WT and EKI male mice at 5 days after tamoxifen induction; Male C57BL/6J at 12-week and 1.5-year; male WT and KO at 12-week after AP induced and \(\beta\) -cell specific sponge of miR-503-322 in control and experimental mice pancreatic protein western blotting. \(n = 3 - 5\) . (C) Experimental scheme: 8-week-old WT male mice were injected intraperitoneally (ip.) with control AAV and EKI male mice were injected with control (Ctr-AAV) and MKNK1-AAV, respectively, one month later tamoxifen was induced for 3 consecutive days and tested at day 7. \(n = 5\) . (D and E) Immunofluorescence staining of Flag and MKNK1 of pancreas sections (D) from each group of mice at 13 weeks and western blotting of pancreatic proteins (E). (F) Gain of body weight, serum amylase and lipase level </center>
+
+<--- Page Split --->
+
+797 were monitored during tamoxifen induction. (G) Representative images of H&E and F4/80 798 immunohistochemistry of pancreas in each group. Arrows indicate the macrophages. (H) 799 Quantitation of the number of F4/80 positive cells in each group and the pancreatic histological 800 score. Data are presented as Mean ± SEM. Data were analyzed using one- way ANOVA 801 followed by Bonferroni's post hoc (F and H). NS, Not Significant; \(*P < 0.05; **P < 0.01; ***P < 0.001\) . 802 803
+
+<--- Page Split --->
+![](images/Figure_7.jpg)
+
+<center>Fig. 7. Evidence of miR-503 and MKNK1 in aging-associated pancreatitis changes in the Chinese population. (A) Representative images of H&E and Masson staining of pancreatic sections from the young adult (YA) and the elderly adult (EA); quantitation of collagen volume fraction. The dashed area indicates acini. n=10. (B) Representative images of immunofluorescence staining of PCNA (green) in pancreatic sections and counted the number of PCNA-positive cells. n=10. Arrows indicate proliferating acinar cells. (C) In situ hybridization of miR-503 (40 nM) in young and elderly pancreatic sections. Scramble-RNA was negative reference (40 nM) and U6 was positive reference (0.1 nM). The dotted line indicate pancreatic islet and solid line is exocrine. (D) Representative images of immunofluorescence staining of </center>
+
+<--- Page Split --->
+
+MKNK1 (green) and amylase (red) in pancreatic sections from young and elderly people and quantitation of amylase and MKNK1 mean fluorescence intensity. \(\mathrm{n = 10}\) . (E) Serum amylase assay of the young adult (YA), the elderly adult (EA) and the elderly adult with diabetes (EA+DM). YA group, \(\mathrm{n = 65}\) . EA group, \(\mathrm{n = 65}\) . EA+DM, \(\mathrm{n = 30}\) . (F) MiR- 503 concentration in human serum of YA, EA and EA+DM. YA group, \(\mathrm{n = 45}\) . EA group, \(\mathrm{n = 45}\) . EA+DM, \(\mathrm{n = 30}\) . (G) Correlation analysis of amylase levels of human serum and age. Each point represents one people ( \(\mathrm{n} = 160\) ). Correlation coefficient (R) and p value from simple linear regression are shown. (H) Correlation analysis of miR- 503 concentration in human serum and serum amylase levels. Each point represents one people ( \(\mathrm{n = 120}\) ). Correlation coefficient (R) and P value from simple linear regression are shown. Data are presented as Mean \(\pm\) SEM. Data were analyzed using one- way ANOVA followed by Bonferroni's post hoc (E and F) or unpaired two- tailed Student's t test (A, B and D) or Correlation analysis (G and H). \(^{**}P< 0.01\) ; \(^{***}P< 0.001\) .
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+SupplementaryMaterials.pdf
+
+<--- Page Split --->

preprint/preprint__0f43ab377913d883885fc5b793f1a6d8dbf89b2a4f2bd4ee445349870e3187e4/preprint__0f43ab377913d883885fc5b793f1a6d8dbf89b2a4f2bd4ee445349870e3187e4_det.mmd

@@ -0,0 +1,467 @@
+<|ref|>title<|/ref|><|det|>[[44, 108, 864, 207]]<|/det|>
+# Endocrine-exocrine miR-503-322 drives aging-associated pancreatitis via targeting MKNK1 in acinar cells  
+
+<|ref|>text<|/ref|><|det|>[[44, 230, 245, 275]]<|/det|>
+Yunxia Zhu zhuyx@njmu.edu.cn  
+
+<|ref|>text<|/ref|><|det|>[[44, 301, 650, 321]]<|/det|>
+Nanjing Medical University https://orcid.org/0000- 0002- 4597- 4445  
+
+<|ref|>text<|/ref|><|det|>[[44, 327, 290, 368]]<|/det|>
+Kerong Liu Nanjing Medical University  
+
+<|ref|>text<|/ref|><|det|>[[44, 374, 290, 415]]<|/det|>
+Tingting Lv Nanjing Medical University  
+
+<|ref|>text<|/ref|><|det|>[[44, 420, 290, 461]]<|/det|>
+Wei Tang Nanjing Medical University  
+
+<|ref|>text<|/ref|><|det|>[[44, 466, 480, 507]]<|/det|>
+Yan Zhang Children's Hospital of Nanjing Medical University  
+
+<|ref|>text<|/ref|><|det|>[[44, 512, 290, 553]]<|/det|>
+Xiao Xiao Nanjing Medical University  
+
+<|ref|>text<|/ref|><|det|>[[44, 559, 290, 599]]<|/det|>
+Yating Li Nanjing Medical University  
+
+<|ref|>text<|/ref|><|det|>[[44, 604, 290, 645]]<|/det|>
+Xiaoai Chang Nanjing Medical University  
+
+<|ref|>text<|/ref|><|det|>[[44, 650, 666, 691]]<|/det|>
+Shusen Wang Tianjin First Central Hospital https://orcid.org/0000- 0002- 2323- 6564  
+
+<|ref|>text<|/ref|><|det|>[[44, 696, 528, 737]]<|/det|>
+Stephen Pandol Cedars- Sinai https://orcid.org/0000- 0003- 0818- 6017  
+
+<|ref|>text<|/ref|><|det|>[[44, 743, 238, 784]]<|/det|>
+Ling Li Southeast University  
+
+<|ref|>text<|/ref|><|det|>[[44, 790, 650, 831]]<|/det|>
+Xiao Han Nanjing Medical University https://orcid.org/0000- 0002- 6467- 1802  
+
+<|ref|>text<|/ref|><|det|>[[44, 872, 102, 890]]<|/det|>
+Article  
+
+<|ref|>text<|/ref|><|det|>[[44, 910, 135, 928]]<|/det|>
+Keywords:
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[43, 46, 301, 64]]<|/det|>
+Posted Date: June 21st, 2024  
+
+<|ref|>text<|/ref|><|det|>[[42, 84, 474, 103]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 4521626/v1  
+
+<|ref|>text<|/ref|><|det|>[[42, 120, 916, 164]]<|/det|>
+License: © © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+<|ref|>text<|/ref|><|det|>[[42, 181, 535, 201]]<|/det|>
+Additional Declarations: There is NO Competing Interest.  
+
+<|ref|>text<|/ref|><|det|>[[42, 236, 930, 280]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on March 17th, 2025. See the published version at https://doi.org/10.1038/s41467- 025- 57615- x.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[72, 92, 884, 750]]<|/det|>
+1 Endocrine-exocrine miR- 503- 322 drives aging-associated pancreatitis via targeting2 MKNK1 in acinar cells3 Kerong Liu,<sup>1</sup> Tingting Lv,<sup>1</sup> Wei Tang,<sup>5</sup> Yan Zhang,<sup>6</sup> Xiao Xiao,<sup>1</sup> Yating Li,<sup>1</sup> Xiaoai Chang,<sup>1</sup>4 Shusen Wang,<sup>4</sup> Stephen J Pandol,<sup>3,</sup>* Ling Li,<sup>2,</sup>* Xiao Han,<sup>1,</sup>* and Yunxia Zhu<sup>1,</sup>*5 <sup>1</sup>Key Laboratory of Human Functional Genomics of Jiangsu Province, Biochemistry and6 Molecular Biology, Nanjing Medical University, Nanjing, Jiangsu 211166, China.7 <sup>2</sup>Department of Endocrinology, Zhongda Hospital, School of Medicine, Southeast University,8 Nanjing, 210009, China.9 <sup>3</sup>Division of Gastroenterology, Department of Medicine, Cedars-Sinai Medical Center, Los10 Angeles, CA, United States.11 <sup>4</sup>Organ Transplant Center, Tianjin First Central Hospital, Nankai University, Tianjin 300192,12 China.13 <sup>5</sup>Department of Endocrinology, Geriatric Hospital of Nanjing Medical University, Nanjing,14 Jiangsu 210024, China.15 <sup>6</sup>Children's Hospital of Nanjing Medical University, Nanjing, Jiangsu 210093, China.16 <sup>*</sup>Correspondence to: Yunxia Zhu, zhuyx@njmu.edu.cn, Phone: +86 25 86869426; Fax: +86 17 25 86869425; Xiao Han, hanxiao@njmu.edu.cn; Ling Li, lingli@seu.edu.cn or Stephen J18 Pandol, Stephen.Pandol@cshs.org
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 95, 197, 110]]<|/det|>
+## Abstract  
+
+<|ref|>text<|/ref|><|det|>[[111, 128, 884, 595]]<|/det|>
+Aging is the major risk factor for chronic pancreatitis and severity determinant for its acute attack, yet the underlying cause is unclear. Here, we demonstrate that senescent \(\beta\) - cells of endocrine pancreas decide the onset and severity of chronic and acute pancreatitis. During physiological aging, senescent \(\beta\) - cells increase the expression of miR- 503- 322 which is secreted as nano- vesicles to enter exocrine acinar cells, driving a causal and reversible role on aging- associated pancreatitis. Mechanistically, miR- 503- 322 represses MKNK1 to inhibit acinar- cell secretion and proliferation, thereby causing autodigestion and repairing damage of exocrine pancreas. In the elderly population, serum miR- 503 concentration is negatively correlated with amylase, prone to chronic pancreatitis due to increased miR- 503 and decreased MKNK1 in the elderly pancreas. Our findings highlight the miR- 503- 322- MKNK1 axis mediating the endocrine- exocrine regulatory pathway specifically in aged mice and humans. Modulating this axis may provide potential preventive and therapeutic strategies for aging- associated pancreatitis.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[117, 95, 230, 111]]<|/det|>
+## Introduction  
+
+<|ref|>text<|/ref|><|det|>[[115, 130, 883, 335]]<|/det|>
+Pancreatitis is one of the most common causes of hospitalization worldwide and represents higher prevalence in the elderly. \(^{1 - 3}\) Chronic inflammation accumulates during natural aging has been identified responsible for the onset of many diseases, including pancreatitis and type 2 diabetes mellitus (T2DM). \(^{4}\) Recent clinical data showed that the incidence of pancreatitis increases in patients with T2DM, \(^{5 - 7}\) indicating the endocrine part of the pancreas participant in pancreatitis formation. However, the underlined mechanisms remain elusive.  
+
+<|ref|>text<|/ref|><|det|>[[115, 353, 883, 705]]<|/det|>
+The endocrine pancreatic islets have a well- recognized anatomical and physiological integration with the exocrine pancreas and regulate its function. \(^{8}\) An involvement of the islet- acinar axis (IAA) has been suggested in the islet- acinar portal system for the physiological regulation of acinar cell function by islet peptides. \(^{9,10}\) A recent study found that islet \(\beta\) - cell- derived cholecystokinin (CCK) acts on acinar cells via the IAA to promote the progression of pancreatic ductal adenocarcinoma (PDAC), \(^{11}\) suggesting that endocrine islet \(\beta\) - cells can crosstalk with acinar cells. In addition, \(\beta\) - cell inflammation exacerbates pancreatitis through chemokine signaling. \(^{12,13}\) These findings suggest that factors secreted abnormally by pancreatic \(\beta\) - cells play a key role in the development of pancreatitis. One possibility is that abnormal secretion of microRNAs (miRNAs) may be involved.  
+
+<|ref|>text<|/ref|><|det|>[[115, 723, 883, 890]]<|/det|>
+Pancreatic \(\beta\) - cells are known to mediate intercellular communication through the secretion of extracellular vesicles (EVs) rich in miRNAs, resulting in reduced insulin sensitivity and secretion capacity in a paracrine or distal manner and elevated blood glucose levels. \(^{14}\) However, a regulatory role for miRNAs carried by EVs derived from \(\beta\) - cells has not been established for pancreatitis. We have previously demonstrated that senescent \(\beta\) - cells released
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 94, 881, 260]]<|/det|>
+miR- 503- 322 as exosomes (\~45 nm, also called nano- vesicles) which were transported into peripheral target organs to cause insulin resistance, thereby leading to the onset of T2DM. \(^{15}\) Serendipitously, overexpression of miR- 503 in \(\beta\) cells caused pancreatitis- like changes with age, suggesting that miR- 503 secreted by endocrine \(\beta\) - cells may be important in regulating exocrine functions including pancreatitis.  
+
+<|ref|>text<|/ref|><|det|>[[115, 279, 883, 630]]<|/det|>
+The X- linked miR- 503, clustered with miR- 322 has been investigated and shown to play an important role in modulating cell proliferation, cell differentiation, and tissue remodeling. \(^{16}\) In the present study, we found that during natural aging, primary miR- 503- 322 (Pri- miR- 503) was transcribed in the endocrine islets while mature miR- 503 and miR- 322 could be detected in both endocrine and exocrine pancreas. Increased levels of miR- 503- 322 in senescent acinar cells were derived from \(\beta\) - cells and intra- acinar miR- 503- 322 promoted pancreatitis by targeting MAP kinase- interacting kinases (MKNK1). The regulation mode was also conserved in aged population, adding further evidence for endocrine- exocrine crosstalk in regulating pancreatitis and providing novel therapeutic targets for the prevention and treatment of aging- associated pancreatitis.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 690, 189, 705]]<|/det|>
+## Results  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 724, 725, 742]]<|/det|>
+## Senescent \(\beta\) -cell- derived miR-503-322 promoted pancreatitis in mice  
+
+<|ref|>text<|/ref|><|det|>[[115, 760, 883, 891]]<|/det|>
+Our previous study showed that \(\beta\) - cell- specific miR- 503 transgenic (βTG) mice suffered from insulin resistance and \(\beta\) - cell dysfunction, leading to T2DM. \(^{15}\) Coincidentally, we noted that the βTG mice also showed chronic pancreatitis- like changes with advanced age, including diffuse expansion of the interlobar septae, fat accumulation, and fibrosis (Fig. S1A and B). Adult βTG
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 93, 881, 187]]<|/det|>
+mice were also showed significant exacerbation of caerulein- induced AP attack, as evidenced by pancreatic edema, macrophage infiltration, and more severe histologic scorings compared with the WT mice (Fig. S1C- E).  
+
+<|ref|>text<|/ref|><|det|>[[112, 202, 883, 744]]<|/det|>
+To understand the role of \(\beta\) - cell miR- 503 on the development of pancreatitis, the expressing distribution of miR- 503 in \(\beta\) TG mice was detected. We found that pri- miR- 503 was significantly increased in islets but not in acini, while the mature miR- 503 was increased in both islets and acini (Fig. 1A and B), suggesting \(\beta\) - cell miR- 503 entering acinar cells. The same expression profiles of pri- miR- 503 and mature miR- 503 and miR- 322 was also observed in aged mice (Fig. 1C and D), making us think about the contribution of miR- 503- 322 to pancreatitis in older age. Consistent with our hypothesis, aged mice showed a more severe form of caerulein- induced AP compared to younger mice (Fig. S2), which could be significantly improved by blocking \(\beta\) - cell miR- 503- 322 levels. An insulin2 promoter- driven sponge- AAV (SP- AAV) specifically expressed in \(\beta\) cells resulted in decreased expression levels of miR- 503- 322 in pancreas (Fig. 1E- G). Meanwhile, caerulein- induced AP measured by serum amylase and lipase levels, pancreatic edema, histologic scorings and macrophage infiltration were significantly ameliorated in aged mice infected with SP- AAV (Fig. 1H- J). These findings indicate that increased levels of miR- 503- 322 in senescent \(\beta\) cells contribute pancreatitis severity associated with older age.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 799, 688, 817]]<|/det|>
+## \(\beta\) cells secreted nano-vesicular miR-503-322 to enter acinar cells  
+
+<|ref|>text<|/ref|><|det|>[[58, 834, 881, 890]]<|/det|>
+We previously verified that \(\beta\) - cell- derived nano- vesicles (βNVs) were secreted from insulin granules and were trafficked into liver and adipose tissues via circulation. \(^{15}\) Whether βNVs
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 93, 881, 260]]<|/det|>
+entered acinar cells was unknown. Here, we shown that acinar cells engulf \(\beta \mathrm{NV}\) s both in vitro and in vivo (Fig. 2A and B); and acinar cell that received \(\beta \mathrm{NV}\) s purified from \(\beta \mathrm{TG}\) islets ( \(\beta \mathrm{TG}\) - \(\beta \mathrm{NV}\) s) had significantly greater levels of miR- 503 than acinar cells receiving \(\beta \mathrm{NV}\) s from wildtype islets (WT- \(\beta \mathrm{NV}\) s, Fig. 2C). However, acinar cells indiscriminately engulfed WT- \(\beta \mathrm{NV}\) s and \(\beta \mathrm{TG}\) - \(\beta \mathrm{NV}\) s in both in vitro and in vivo models.  
+
+<|ref|>text<|/ref|><|det|>[[115, 279, 883, 483]]<|/det|>
+To avoid the influence of insulin resistance and hyperglycemia in \(\beta \mathrm{TG}\) mice, \(^{15}\) we constructed RIP2- cre;miR- 503- 322 KI ( \(\beta \mathrm{KI}\) ) mice which were barely diabetic (Fig. S3A- C). \(\beta \mathrm{KI}\) mice also exhibited an exacerbation of caerulein- induced AP compared to littermate controls (Fig. 2D- G), confirming the effects of \(\beta\) - cell miR- 503- 322 to the onset of pancreatitis. Thus, we concluded that \(\beta\) - cell derived nano- vesicles enter acinar cells and drive pancreatitis at a miR- 503- 322- dependent manner in mice.  
+
+<|ref|>title<|/ref|><|det|>[[117, 538, 880, 592]]<|/det|>
+# Direct elevation of miR-503-322 in acinar cells triggers both acute and chronic pancreatitis  
+
+<|ref|>text<|/ref|><|det|>[[115, 610, 883, 891]]<|/det|>
+Next, we sought to investigate the effects of miR- 503- 322 under inducible global elevation condition by using CAG- creER;miR- 503- 322 KI (CKI). After tamoxifen induction for 3 times, pri- miR- 503 expression levels were significantly elevated in pancreas, skeletal muscle and other metabolic tissues (Fig. S4A and B). Surprisingly, CKI mice started to lose weight and activity, and all committed to death post- induction for 6 days due to severe AP, as observed by significantly increased serum amylase and lipase levels, abdominal infiltration of neutrophils and macrophages, and pancreatic saponification, necrosis and histological analysis (Fig. S4C- I). However, no concomitant histological changes were observed in other
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 93, 883, 260]]<|/det|>
+major abdominal organs (Fig. S4J). Severe AP induced systemically inflammatory responses were shown by inverted serum ratios of neutrophils and lymphocytes, and elevated serum levels of C- reactive protein (Fig. S4K- M). These results validate that the global overexpression of miR- 503- 322 promotes severe AP, indicating the specificity of the miR- 503- 322 for pancreas damage.  
+
+<|ref|>text<|/ref|><|det|>[[115, 279, 883, 632]]<|/det|>
+To rule out the contribution of other tissues, Pdx1- cre; miR- 503- 322 KI (heterozygous PKI/WT and homozygous PKI/KI) mice were used to yield high pancreatic- specific expression of miR- 503- 322. The pancreatic Pri- miR- 503 expression was increased in the heterozygous PKI mice (PKI/WT) compared to wildtype controls and was further increased in the homozygous mice (PKI/KI) (Fig. 3A). The PKI/KI mice showed an unexpected weight loss at about 6 weeks of age, while the PKI/WT mice showed no change during natural growth (Fig. 3B and Fig. S5A). The most prominent features of chronic pancreatitis (CP), including pancreatic atrophy, fibrosis, tubular complexes, and inflammatory infiltration were observed in PKI/WT mice, with more severe CP and gross changes in the homozygous PKI/KI mice (Fig. 3C- E, Fig. S 5B- D). Accordingly, PKI/KI mice could not survive for 12 weeks (Fig. 3F).  
+
+<|ref|>text<|/ref|><|det|>[[115, 650, 883, 891]]<|/det|>
+PDX1 is a master regulator in pancreas organogenesis while the maturation and identity preservation of islet \(\beta\) - cells and \(\delta\) - cells. \(^{17,18}\) To avoid development defect, inducible acini- specific miR- 503- 322 (Elastase- CreER; miR- 503- 322 KI, EKI) mice were also constructed and overexpression verified post- induction for 3 days (Fig. 3G and H). After tamoxifen injection, the EKI mice showed significantly increased indicators of AP, including macrophage infiltration, tissue damage, and necrosis (Fig. 3I- J, Fig. S5E and F), and had a 20% mortality rate (Fig. 3K). Those mice that survived developed histology of CP one- month post- induction,
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 93, 881, 223]]<|/det|>
+manifested as pancreatic atrophy (Fig. S5G), fibrosis, fat replacement, and acinar- to- ductal metaplasia (ADM, Fig. 3L and M), while a return to normal levels of serum amylase and lipase (Fig. S5E and F). As shown in Fig. S5H- J, EKI female mice presented an AP phenotype similar to that of male mice 5 days after tamoxifen induction.  
+
+<|ref|>text<|/ref|><|det|>[[115, 242, 881, 335]]<|/det|>
+The above findings demonstrate that global, pancreatic, and acinar cell- specific overexpression of miR- 503- 322 can directly trigger (severe) acute and chronic pancreatitis in a dose- and tissue- dependent manner.  
+
+<|ref|>title<|/ref|><|det|>[[115, 391, 730, 409]]<|/det|>
+# MiR-503-322 knockout alleviated caerulein-induced acute pancreatitis  
+
+<|ref|>text<|/ref|><|det|>[[115, 425, 883, 780]]<|/det|>
+The possibility that ablation of miR- 503- 322 could alleviate AP was investigated by the global deletion of miR- 503- 322 (KO) (Fig. S6A). The KO mice were viable and fertile, with normal body weight (Fig. S6B). Histology of the pancreas revealed normal pancreatic morphology (Fig. S6C and D). Challenging the KO and WT mice with caerulein or PBS and assessing for AP severity revealed markedly lower pancreatic edema and amylase and lipase levels in the KO group (Fig. 4A- D). Histological examination revealed reduced pancreatic acinar cell damage, less interstitial expansion (indication of edema), and diminished macrophage infiltration in KO mice during the acute AP phase (Fig. 4E- G). Together, these data demonstrate that the deletion of miR- 503- 322 can significantly alleviate caerulein- induced acute pancreatitis.  
+
+<|ref|>title<|/ref|><|det|>[[115, 835, 880, 890]]<|/det|>
+# MiR-503-322 promotes pancreatitis by inhibiting zymogen secretion and acinar-cell proliferation
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 93, 882, 298]]<|/det|>
+Next, we sought to identify the mechanisms by which miR- 503- 322 promotes the development of pancreatitis. Transmission Electron Microscope (TEM) images from the pancreas of the PKI/WT mice revealed an increased number of zymogen granules (Fig. S7A). However, the significantly lower transcript levels of pancreatic enzyme- related genes implied that this did not represent an increased production of zymogen in the acinar cells (Fig. S7B) but was possibly an indication of a secretion defect.  
+
+<|ref|>text<|/ref|><|det|>[[115, 315, 882, 560]]<|/det|>
+Therefore, we isolated acini and assessed their secretory ability in response to caerulein. The amylase release was significantly lower from the PKI cells than from the WT cells (Fig. 5A). The acinar cells from aged mice showed a similar response to that of the PKI cells, with a reduced secretion of pancreatic enzymes (Fig. 5B), in agreement with the results of previous studies. \(^{19}\) By contrast, the primary acinar cells from the KO mice showed enhanced amylase secretion (Fig. 5C). The defect of enzyme secretion was attributed to the loss of cytoskeleton modulation from tip to basolateral membranes of acinar cells responding to caerulein (Fig. 5D).  
+
+<|ref|>text<|/ref|><|det|>[[115, 576, 882, 817]]<|/det|>
+Enzyme secretion defect may cause trypsinogen activation. We observed that trypsinogen activation in acini was visualized by using rhodamine 110 (BZiPAR) which revealed a clear enrichment of green fluorescence in PKI cells (Fig. 5E), and serum trypsin activity was enhanced in the PKI mice (Fig. 5F). These findings indicate that miR- 503- 322 inhibits pancreatic enzyme secretion and promotes the intracellular accumulation of zymogen. Subsequent zymogen activation in situ may promote pancreas damage of miR- 503- 322 elevated mice.  
+
+<|ref|>text<|/ref|><|det|>[[115, 834, 881, 890]]<|/det|>
+Activation of trypsinogen by lysosomal enzymes after fusion of the lysosome is the classical mode of pancreatic enzyme activation during AP. \(^{20,21}\) TEM images of the pancreas
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 92, 883, 448]]<|/det|>
+from PKI mice show morphological signs of this activation, including numerous autophagy vacuoles in the cytoplasm and abundant zymogen granules varying in size and electron density and sometimes fused together to form irregular "lakes" (Fig. S7C). These phenomena suggest a classical activation of intracellular zymogen in the lysosomes of acinar cells that highly express miR- 503- 322. We verified this by inducing pancreatitis in WT and EKI mice by administration of chloroquine, which destroys the acidic environment in autophagic lysosomes (Fig. S7D). The AP phenotype was alleviated in EKI mice treated with chloroquine, as evidenced by a smaller weight loss, reduced serum amylase and lipase levels, and less tissue damage compared to saline- treated control mice, despite a similar pancreas weight (Fig. S7E- J).  
+
+<|ref|>text<|/ref|><|det|>[[115, 464, 884, 742]]<|/det|>
+AP significantly stimulates the proliferation of acinar cells almost immediately at the point of injury. Not surprisingly, immunofluorescence staining for PCNA revealed a reduction in the numbers of proliferating acinar cells in the mice expressing high levels of the miR- 503- 322, and an increased proliferation of mesenchymal cells (Fig. 5G- I). Conversely, ablation of miR- 503- 322 enhanced acinar- cell proliferation during the repair phase of caerulein- induced AP (Fig. 5G and J). We also conducted a similar test in aged mice and again observed a significant decrease in acinar- cell proliferation similar to that seen in the high miR- 503- 322 expression model mice (Fig. 5G and K).  
+
+<|ref|>text<|/ref|><|det|>[[115, 760, 883, 853]]<|/det|>
+Taken together, these data suggest that miR- 503- 322 suppresses zymogen secretion to initiate acute pancreatitis. Meanwhile, miR- 503- 322 also inhibits regenerative proliferation of acinar cells to promote the formation of chronic pancreatitis.
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[115, 94, 880, 150]]<|/det|>
+# MKNK1 is a target of miR-503-322 and acinar cell-specific restoration of it reverses the phenotype of pancreatitis in mice  
+
+<|ref|>text<|/ref|><|det|>[[115, 165, 883, 632]]<|/det|>
+We previously used unbiased proteomics to identify target genes of miR- 503 in regulating peripheral insulin resistance and \(\beta\) - cell dysfunction. \(^{15}\) By analyzing the same proteomics data combined with Targetscan software analysis, five genes (MKNK1, CCNE1, IGF1R, PI3KR1 and INSR) were potential targets (Fig. S8A). After extensively searching and reading literature, we found that the MAP kinase- interacting kinases (MKNK1), mostly expressed in exocrine pancreas might contribute to miR- 503- 322- caused pancreatitis. MKNK1 plays an indispensable role in physiological exocrine secretory response. \(^{23}\) Consistent with published data, phosphorylation of MKNK1 and its downstream eIF4E was increased 4 hr after the first caerulein injection and gradually recovered (Fig. S8B- E). MKNK1 was redistributed to the basolateral region after caerulein administration, assisting acinar- cell secretion (Fig. S8F). Previous studies showed that ablation of MKNK1 results in exacerbation of pancreatitis caused by caerulein due to defects of zymogen secretion and acinar- cell proliferative in mice, \(^{23}\) making us pursue the role of MKNK1 as a target gene of miR- 503- 322.  
+
+<|ref|>text<|/ref|><|det|>[[115, 650, 883, 890]]<|/det|>
+Our proteomics analysis showed a decrease in MKNK1 after miR- 503 elevation. Dual- luciferase assay confirmed the regulatory role of miR- 503- 322 on the 3'UTR of Mknk1 gene (Figure 6 A and Fig. S9A and B). Next, immunohistochemistry staining of pancreas sections revealed clear suppression of MKNK1 protein amount in the three miR- 503- 322 overexpressing mouse model. (Fig. S9C), while upregulation of MKNK1 was induced by caerulein in KO mice (Fig. S9D). The protein levels of MKNK1 and its associated P- MKNK1/P- eIF4E signaling were significantly reduced in pancreas of miR- 503- 322 overexpressing model
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 93, 883, 260]]<|/det|>
+mice and aged mice, and by contrast increased in pancreas of miR- 503- 322 knockout mice and aged mice with \(\beta\) - cell specific blocking miR- 503- 322 (Figure 6B and Fig. S9E- G). Taken together, these findings suggest that miR- 503- 322 targets MKNK1- elF4E pathway to inhibit zymogen secretion and acinar- cell proliferation, thereby leading to acute and chronic pancreatitis.  
+
+<|ref|>text<|/ref|><|det|>[[115, 279, 884, 632]]<|/det|>
+Next, we tested whether reconstitution of MKNK1 in pancreas could reverse pancreatitis of EKI mice following the schematic diagram (Fig. 6C). We generated an AAV, serotype pancreas (MKNK1- AAV) that directs specific MKNK1 overexpression in the exocrine pancreas. As shown in Fig. S9H, MKNK1 was highly expressed in the acini, but not in the islets of MKNK1- AAV mice. Restoration of MKNK1 also rescued the miR- 503- 322- suppressive protein levels of phos- MKNK1 and phos- elF4E in the EKI pancreas (Fig. 6D and E). Consequently, MKNK1- AAV infected EKI mice showed lessened AP phenotypes compared to Ctr- AAV infected EKI mice. In detail, the loss of body weight, increased serum levels of amylase and lipase, increased number of macrophage infiltration, and tissue damage in Ctr- AAV infected EKI mice were largely reduced in MKNK1- AAV infected EKI mice (Fig. 6F- H).  
+
+<|ref|>text<|/ref|><|det|>[[115, 650, 883, 816]]<|/det|>
+On the other hand, inhibition of MKNK1 by a verified inhibitor, CGP 57380 further exacerbated caerulein- caused AP phenotypes, and totally erased miR- 503- 322 knockout driven protective effects (Fig. S10A- G). These results from acinar- cell MKNK1 reconstitution and specific MKNK1 inhibitor support our view that the deficiency of MKNK1 in acini is primarily responsible for the pancreatitis observed in miR- 503- 322 elevated mice.  
+
+<|ref|>text<|/ref|><|det|>[[115, 872, 880, 891]]<|/det|>
+Evidence of miR- 503 and MKNK1 in aging- associated pancreatitis changes in the
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[118, 94, 293, 111]]<|/det|>
+## Chinese population  
+
+<|ref|>text<|/ref|><|det|>[[115, 130, 883, 595]]<|/det|>
+As the expression of miR- 503 is specifically increased in senescent \(\beta\) cells in mice, we also considered its change in humans. Pancreas sections from elderly adults (EA) showed chronic pancreatitis- like changes, including atrophy of the acinar cells, interstitial expansion, and a marked increase in fibrosis (Fig. 7A), as well as a significant reduction in the proportion of proliferating acinar cells (Fig. 7B), compared to that from young adults (YA). Intriguingly, miRNA in situ hybridization showed greater expression of miR- 503 in islets than in acini in pancreatic sections from EA group (Fig. 7C), whereas expression of miR- 503 was almost undetectable in YA group (Fig. 7C). The expression of MKNK1 was significantly downregulated in the acini from the EA pancreas compared to that from the YA pancreas (Fig. 7D). Moreover, the co- localization of MKNK1 and AMY1 in the young acini was dislocated in the elderly acini (Fig. 7D), indicating an activation of MKNK1 in the EA. Thus, the increased level of miR- 503 in the acini may come from the islet \(\beta\) cells and contribute to the decreased but activated MKNK1 protein in the elderly of Chinese population.  
+
+<|ref|>text<|/ref|><|det|>[[115, 612, 883, 890]]<|/det|>
+Numerous studies have reported that exocrine pancreas function is impaired in both healthy and diabetic older adults independent of gastrointestinal disease, judged by serum levels of amylase and maximum bicarbonate concentration. \(^{24,25}\) Consistently, we observed a significantly decreased level of serum amylase in the elderly adult with T2DM (EA+DM) compared to that in YA, moreover, the elderly adults also showed a decreased amylase level (Fig. 7E). Further analysis showed that serum concentration of exosomal miR- 503 was elevated in the elderly compared to that in the young adults and was further elevated in the EA+DM (Fig. 7F). The human subjects displayed negative associations of serum amylase
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 93, 882, 223]]<|/det|>
+levels with both age and serum concentrations of exosomal miR- 503 (Fig. 7G and H). These results support the pancreatic exocrine insufficiency in the elderly and diabetic patients and point out serum concentrations of exosomal miR- 503 as molecular marker of aging- associated pancreatitis in the Chinese population.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 279, 220, 296]]<|/det|>
+## Discussion  
+
+<|ref|>text<|/ref|><|det|>[[115, 315, 883, 556]]<|/det|>
+In this study, we demonstrated that miR- 503- 322 derived from endocrine \(\beta\) - cells promotes aging- associated pancreatitis by targeting MKNK1 in exocrine acinar- cells. miR- 503- 322, which is produced by senescent \(\beta\) - cells, had an in- situ effect in acinar cells that inhibits zymogen secretion and regenerative proliferation. Thus, the miR- 503- 322- MKNK1 axis caused pancreas autodigestion and repairing damage, leading to the onset of acute and chronic pancreatitis in mice. This discovery provides an epigenetic mechanism for pancreatitis and adds to the existing evidence of crosstalk between pancreatic endocrine and exocrine.  
+
+<|ref|>text<|/ref|><|det|>[[115, 575, 883, 890]]<|/det|>
+In the past, several studies have described an impairment of exocrine function in T2DM. However, unlike the observations made for CP, these patients showed no obvious symptoms of exocrine disease, such as abdominal pain, and no ductal changes. As a result, a 2016 study proposed the term diabetic exocrine pancreatic disease (DEP) to describe this entity. Several hypotheses have been proposed to explain the features of DEP, including localized insulin deficiency, loss of regulatory functions of islet hormones, pancreatic fibrosis and exocrine atrophy due to vascular pathology, and impaired enteropancreatic reflexes due to neuropathy. However, none of these concepts are sufficient to explain all the pathological findings. Our previous results showed a significant upregulation of islet miR- 503
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 92, 882, 224]]<|/det|>
+expression in patients with T2DM. \(^{15}\) Suggested by our current investigation, the expressed miR- 503 can then enter and accumulate in the exocrine acini, where it triggers damage to some of the acinar cells and causes chronic pancreatitis- like changes in the exocrine pancreas due to repeated pancreas damage.  
+
+<|ref|>text<|/ref|><|det|>[[115, 241, 883, 708]]<|/det|>
+Most studies report higher overall morbidity and mortality from pancreatitis in the elderly, \(^{1}\) and several explanations for this phenomenon have been put forward. \(^{32,33}\) Histologically, focal fibrosis also appears to be common in the pancreas of the elderly. \(^{34,35}\) This is consistent with the observations in human pancreatic sections in the present study. Clinical studies have indicated that pancreatic exocrine function is impaired in healthy older individuals without any gastrointestinal disease. \(^{36}\) However, few studies have linked exocrine CP- like changes in healthy older adults and patients with T2DM, and the underlying mechanisms remain to be explored. Our previous findings showed that miR- 503 is significantly upregulated in the islets of both diabetic patients and the elderly, \(^{15}\) and the present results confirmed that islet- derived miR- 503- 322 promotes both acute and chronic damages in the exocrine pancreas and increases mouse mortality with acute and high miR- 503- 322 expression (CKI and PKI/KI mice). Therefore, miR- 503- 322 may be a common pathogenic factor that can explain the higher morbidity and mortality from pancreatitis in the elderly.  
+
+<|ref|>text<|/ref|><|det|>[[115, 724, 883, 890]]<|/det|>
+In healthy adults, miR- 503- 322 is expressed mainly in lung, heart, and skeletal muscle progenitor cells. \(^{37}\) Upregulation of miR- 503- 322 occurs in aging acinar cells and is likely to arise from pancreatic \(\beta\) - cells, based on our present observations. Our evidence for this is that blocking miR- 503- 322 in islet \(\beta\) - cells of aging mice alleviated caerulein- induced pancreatitis. Our previous findings revealed that miR- 503 from pancreatic islet \(\beta\) - cells reaches the liver and
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 93, 883, 560]]<|/det|>
+adipose tissue in the form of exosomes, which are known to transport biologically active proteins and miRNAs in their active forms to neighboring cells or distant organs. \(^{38 - 40}\) Thus, the involvement of exosomes in inter-organ and intra-organ crosstalk has been increasingly studied. \(^{41,42}\) Exosomes derived from mesenchymal stem cells have been reported as a new treatment for AP by delivering mitochondria and anti-inflammatory factors. \(^{43,44}\) In addition, senescent \(\beta\) cells have been reported to secrete senescence- associated secretory phenotypes that are rich in EVs and cause dysfunction of adjacent cells through paracrine effects. \(^{45}\) The reported anatomical characteristics of an IAA permits the access of high concentrations of islet- derived miR- 503- 322 to exocrine cells. Indeed, a recent study has determined that islet CCK can promote Kras- driven PDAC development of an endocrine exchange signal other than insulin, \(^{11}\) supporting the existence of endocrine- exocrine crosstalk via IAA. Therefore, we hypothesize that islet- derived miR- 503- 322 is transferred via exosomes and the IAA into acinar cells.  
+
+<|ref|>text<|/ref|><|det|>[[115, 576, 883, 891]]<|/det|>
+Our results support MKNK1 as an miR- 503- 322 target gene for the development of pancreatitis. However, MKNK1 knockout mice showed normal pancreatic histology, \(^{23}\) which was inconsistent with the phenotype of AP induced by miR- 503- 322. This normal histologic may reflect the presence of other compensatory pathways in MKNK1- knockout mice as the use of global mouse model. Indeed, the knockout of MKNK1 adds to the growing list of proteins that have a protective role during AP, \(^{46}\) whereas the acute induction of miR- 503- 322 lacks an effective compensatory mechanism. Alternatively, other target genes of miR- 503- 322 co- regulating the development of pancreatitis may exist. Moreover, exosomes carrying miR- 503- 322 may function through inter- organ crosstalk to regulate the severity of AP as shown in CKI
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 93, 881, 186]]<|/det|>
+mice. In addition, tamoxifen administration occasionally causes pancreatitis also reminded us that effect of tamoxifen itself cannot be ignored, although it was added to the control group. \(^{47}\) The mechanisms involved in these possibilities need to be unraveled in further studies.  
+
+<|ref|>text<|/ref|><|det|>[[115, 205, 883, 335]]<|/det|>
+In conclusion, we demonstrate the role and a mechanism of action for pancreatic endocrine- derived miR- 503- 322 in promoting pancreatitis in the elderly. Blocking miR- 503- 322 in \(\beta\) - cells of aged mice showed good inhibitory effects on pancreatitis, revealing miR- 503- 322 as a potential therapeutic target for elderly patients with pancreatitis.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 392, 198, 408]]<|/det|>
+## Methods  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 428, 405, 446]]<|/det|>
+## Human biospecimen acquisition  
+
+<|ref|>text<|/ref|><|det|>[[115, 463, 883, 668]]<|/det|>
+For human pancreas sections study, conducted in Organ Transplant Center, Tianjin First Central Hospital, Nankai University, Tianjin, China. A total of 20 healthy individuals were recruited, of these, 10 were young adult (YA, 18- 25 years old) and 10 were the elderly adult (EA, 60- 73 years old). The detailed information of donors was listed in Table. S1. Informed consent was obtained from all patients, and the research protocol was reviewed and approved by the research ethics committee of Tianjin First Central Hospital (No. 2018N112KY).  
+
+<|ref|>text<|/ref|><|det|>[[115, 686, 883, 890]]<|/det|>
+For blood sample collection, conducted in the Department of Endocrinology, Geriatric Hospital of Nanjing Medical University, Nanjing, China, 160 individuals were recruited, including 65 YA (18- 37 years old), 65 EA (65- 85 years old), and 30 EA with T2DM (EA+DM). Fasting blood samples, collected from all participants, were centrifuged at 3,000 rpm for 20 min to separate sera and blood cells, the sera were used for miR- 503 concentration analysis. Detailed information of donors including age, gender, fasting blood glucose levels, and history
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 93, 881, 186]]<|/det|>
+of prior diseases were listed in Table S2. The study was approved by the research ethics committee of Nanjing Medical University (2022006), and all the volunteers gave written informed consent.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 206, 253, 222]]<|/det|>
+## Animal studies  
+
+<|ref|>text<|/ref|><|det|>[[115, 238, 883, 896]]<|/det|>
+Animal studies were approved by the Research Animal Care Committee of Nanjing Medical University (IACUC- 1707023 and IACUC- 2004040). Generation of the mouse miR- 503- 322 knock- in mouse (H11- CAG- LSL- miR- 503- 322 Cas9- KI) by CRISPR/Cas9 was outsourced to GemPharmatech Co, Ltd. The mice were created on the C57BL/6J genetic background. The gRNA (5'- CTGAGCCAACAGTGGTAGTA - 3') to the Hipp11 (H11) locus, the donor vector containing the "CAG- loxP- Stop- loxP- mouse miR- 503- 322- polyA" cassette, and Cas9 mRNA were co- injected into fertilized mouse eggs to generate targeted conditional knock- in offspring. Rat insulin 2 promoter (RIP2)- Cre (JAX:003573), CAG- CreER (JAX:004453) and PDX1- Cre (JAX:014647) mice were obtained from the Jackson Laboratory. Elastase (ELA)- CreER mice were obtained from Dr. Xianghui Fu (Professor of the West China Hospital, Sichuan University). We then crossed KI mice with CAG- CreER, PDX1- Cre, ELA- CreER, and RIP2- Cre mice respectively, to obtained global inducible (CKI), pancreas- specific (PKI), acinar cell- specific inducible (EKI), and islet \(\beta\) - cell- specific ( \(\beta\) KI) overexpression miR- 503- 322 mice. Details on each animal strain were listed in Table S3. EKI or CKI and their litter control mice were injected intraperitoneally with tamoxifen solution, 100 mg/kg, in corn oil, for three consecutive days to induce miR- 503- 322 overexpression in acinar cells or the whole body, respectively. The control groups used their respective littermates and were genotyped as KI- positive and Cre- negative. All experimental mice were heterozygous except for PKI mice, which included both
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 93, 881, 223]]<|/det|>
+homozygotes and heterozygotes. MiR- 503 transgenic mice (βTG) and miR- 503- 322 global deletion mice (KO) were also generated by GemPharmatech Co, Ltd. Refer to our previous findings for the exact construction workflow. \(^{15}\) Aged C57BL6/J mice were purchased from GemPharmatech Co, Ltd.  
+
+<|ref|>text<|/ref|><|det|>[[115, 242, 883, 558]]<|/det|>
+The animals were randomly allocated to experimental groups, at least 4 per group, not according to genotype to minimize potential confounding factors. Male mice were mostly used in this study, and female mice were also involved to rule out the sex bias, as described in the Fig. legends. Mice were housed in a temperature- and humidity- controlled environment (23- 25°C, 12- h light/dark cycle, 60- 70% humidity) in a specific pathogen- free facility at Nanjing Medical University and provided with free access to commercial rodent chow and tap water. Health was monitored at least weekly by weight, food and water intake, and general assessment of animal activity, panting, and fur condition. Mice were euthanized by \(\mathrm{CO_2}\) asphyxiation when met euthanasia criteria.  
+
+<|ref|>text<|/ref|><|det|>[[115, 576, 883, 780]]<|/det|>
+Adult animals of both genders were used in tamoxifen induction studies. Collected blood serum was used to measure amylase and lipase. The pancreatic tissue was collected and immediately embedded in optimum cutting temperature compound for hematoxylin and eosin staining, evaluation of necrosis, and immunohistochemistry. Necrosis and acinar cell damage quantified by morphometry as described. \(^{48}\) Tissue damage was quantified using scoring system as describe by Schmidt et al. \(^{49}\)  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 799, 432, 816]]<|/det|>
+## Pancreatic Acinar Cell Experiments  
+
+<|ref|>text<|/ref|><|det|>[[115, 835, 881, 889]]<|/det|>
+Mouse pancreatic acini were isolated using the standard collagenase digestion protocol, as previously described. \(^{50}\) Acini were isolated and left to recover for 30 min at 37°C before
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 93, 883, 372]]<|/det|>
+stimulation with the indicated concentrations of caerulein (MCE, Shanghai, China) to assess the secretory capacity. The supernatant for amylase activity was analyzed with a commercial kit (JianCheng Bioengineering Institute, Nanjing, China) and the percentage of amylase secretion was calculated. To visualize trypsinogen activation in acinar cells, freshly prepared acini were loaded with active trypsin enzyme substrate BZiPAR (10 μM) (Invitrogen, America) and incubated for 30 min. Images were captured and analyzed by a confocal laser scanning microscope (Olympus FV1200). The image fluorescence intensity was analyzed with ImageJ software.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 391, 404, 409]]<|/det|>
+## Induction of Murine Pancreatitis  
+
+<|ref|>text<|/ref|><|det|>[[115, 426, 883, 632]]<|/det|>
+Caerulein was solubilized in phosphate- buffered saline at a final concentration of 15 mg/mL. Experimental mice were challenged with caerulein (50 mg/kg, intraperitoneal injection, once an hr, 6 times) to induce AP. Control animals received an equal amount of saline. The parameters of AP were assessed 2 hrs after the last caerulein treatment. Edema, serum lipase (ElabScience, Wuhan, China), amylase and trypsin activity (JianCheng Bioengineering Institute, Nanjing, China) were analyzed as parameters of pancreatitis.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 650, 700, 669]]<|/det|>
+## Histopathology, Immunohistochemistry and Immunofluorescence  
+
+<|ref|>text<|/ref|><|det|>[[115, 685, 883, 891]]<|/det|>
+Mice were euthanized by \(\mathrm{CO_2}\) asphyxiation and tissue was dissected, rinsed in PBS and fixed overnight in \(4\%\) paraformaldehyde (Servicebio). Paraffin embedding, serial sectioning, H&E and Masson staining of all samples were commissioned from Servicebio Technologies. After dewaxing and antigen retrieval, the pancreatic paraffinic sections were incubated with primary antibodies overnight at \(4^{\circ}\mathrm{C}\) . According fluorescent- conjugated secondary antibodies (Proteintech) were used for multiple labeling and the nuclei were stained with Hoechst 33342
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 93, 881, 298]]<|/det|>
+(5 μg/mL) (Sigma-Aldrich). Fluorescent images were visualized by a confocal laser scanning microscope (Olympus FV1200). Immunohistochemistry staining was labeled with DAB substrate system (BCA Kit) (Gene Tech) and positive labeled cells were captured by a light microscope (Leica, Germany). Quantification was done with at least three mice per group, three sections per mouse (50 μm apart), and at least 10 microscopic fields per section. The antibodies were listed in Table S4.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 315, 715, 335]]<|/det|>
+## Pancreatic Ductal Infusion of Adeno-associated Viral (AAV) Vectors  
+
+<|ref|>text<|/ref|><|det|>[[115, 351, 883, 707]]<|/det|>
+Pancreatic ductal infusion was performed following the standard surgical protocol, as we previously described.51 Serotype pancreas of PAAV- CMV- MCS- EF1- mNeonGreen- WPRE (Ctr- AAV) and PAAV- CMV- MKNK1- flag- EF1- mNeonGreen- WPRE (MKNK1- AAV) were provided by the company of OBIOTechnology Co, Ltd. Serotype 2/8 under insulin2 promoter of HBAAV2/8- insulin2- zsGreen (Ctr) and HBAAV2/8- insulin2- mmu- miR- 503/322- 5p- sponge- zsGreen (SP) were provided by the company of Hanheng Biotechnology Co, Ltd. AAV titer of \(10^{11} / \mathrm{mL}\) in PBS, 100 μL total volume in 20 g body weight mice was infused at a rate of 6 μL/min. After infusion and suture, surgical mice were placed on a heated pad (37°C) until full recovery. Ketoprofen (Sigma, k1751) at a dose of 5 mg/kg once per day was given continuously for 3 d for post- surgery analgesia.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 723, 702, 743]]<|/det|>
+## Locked Nucleic Acid (LNA)-Based in situ Hybridization of miR-503  
+
+<|ref|>text<|/ref|><|det|>[[115, 760, 883, 890]]<|/det|>
+Locked nucleic acid- based in situ assay was introduced to detect miR- 503 in human pancreas sections. Double- labeled with carboxyfluorescein (FAM), LNA enhanced probes including U6 snRNA control probe, negative scramble- miR control and has- miR- 503 were constructed by QIAGEN. The assay was performed according to the manufacturer's protocol.52 In short,
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 94, 881, 336]]<|/det|>
+sample slides were deparaffinized in xylene and ethanol solutions at room temperature (15- 25°C) and digested with Proteinase K reagent for 10 min at 37°C. After washing, each sample was reacted with 50 μL of hybridization mix (1 nM LNA U6 snRNA probe, 40 nM double- FAM LNA miR- 503 probe and scramble- miR) in a programmed hybridizer for 1 hr. After strictly washing and blocking, the samples were incubated with anti- FAM reagent for 1 hr and labeled with alkaline phosphatase substrates for 2 hrs. The nuclei were labeled with Nuclear Fast Red. All sample slices were visualized by light microscopy.  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 353, 658, 372]]<|/det|>
+## Islet-derived Exosomes Isolation and Fluorescence Labelling  
+
+<|ref|>text<|/ref|><|det|>[[115, 390, 883, 706]]<|/det|>
+Freshly isolated islets were cultured in serum- exosome- free medium (11.1 mM glucose) for 7 days, with the medium replaced and collected every 24 hr. The medium was first centrifuged at 700 g for 5 min and at 10,000 g for 1 hr to remove cell debris. Next, the collected supernatant was centrifuged as described previously to obtain exosomes of good purity. For cell imaging, exosomes were labelled with PKH67 (Sigma- Aldrich) for 1 hr and then washed three times with PBS. PKH67- labelled exosomes (100 μg/35 mm culture dish) were resuspended in PBS and then incubated with freshly isolated acini for 8 h. The acini were then stained with phalloidin (MCE) for 15 min and Hoechst 33342 for 8 min. Images were taken and analyzed by a confocal laser scanning microscope (Olympus FV1200).  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 723, 635, 741]]<|/det|>
+## Plasmid Construction and Dual-Luciferase Reporter Assay  
+
+<|ref|>text<|/ref|><|det|>[[115, 760, 883, 891]]<|/det|>
+The wild- type and mutant 3' UTR- luciferase constructs containing miR- 503- 322 binding site of mouse Mknk1 were generated by annealing and cloning the short sequences into pMIR- REPORT Luciferase miRNA Expression Reporter Vector (Ambion) between the Spel and HindIII sites. Primer sequences were listed in Table S5. Luciferase activities were measured
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 94, 880, 150]]<|/det|>
+using the Dual- Glo Luciferase Assay System (Promega, America) on a TD- 20/20 Luminometer (Turner BioSystems, America) according to the manufacturer's protocols.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 168, 304, 186]]<|/det|>
+## Quantitative RT-PCR  
+
+<|ref|>text<|/ref|><|det|>[[115, 203, 883, 445]]<|/det|>
+Total RNA was extracted from cells and tissues using Trizol reagent (Invitrogen). cDNA was synthesized from total RNA using a ReverTra Ace Kit (TOYOBO, Japan). qPCR of Pri- miRNA and miRNA were performed using the THUNDERBIRD probe qPCR Mix (TOYOBO, Japan), and SYBR Green qPCR Master Mix (Vazyme, China) for mRNA on Roche LightCycle480 II Sequence Detection System (Roche, Switzerland). Primers of qPCR for pri- miRNA and miRNA were purchased from Thermofisher Co., Ltd, other primer sequences were available in Table S5.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 465, 316, 482]]<|/det|>
+## Western Blot Analysis  
+
+<|ref|>text<|/ref|><|det|>[[115, 500, 884, 890]]<|/det|>
+Cells or tissues were lysed with ice- cold RIPA buffer (Thermo Fisher Scientific), supplemented with 0.5 mM EDTA and Halt protease/phosphatase inhibitor cocktail (Thermo Fisher Scientific), rotated at \(4^{\circ}C\) for 15- 30 minutes to mix, and centrifuged at maximum speed for 15 min to collect whole cell lysates. Protein concentration was measured with the BCA protein assay (Takara). Thirty mg of total protein per sample was loaded into \(4\% - 12\%\) NuPAGE Tris- Bis (Thermo Fisher Scientific) gradient gels and separated by SDS- PAGE. Proteins were transferred to PVDF membranes (Millipore Billerica) and blocked with \(5\%\) milk. Beta- actin and \(\alpha\) - tubulin were used as loading controls. Primary antibodies were detected with HRP- conjugated (Sigma- Aldrich) secondary antibodies for chemiluminescent detection (Perkin Elmer ECL). Protein quantification was performed by Image J (NIH Image). Key reagents and antibodies were listed in Table S6.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[118, 94, 290, 111]]<|/det|>
+## Statistical Analysis  
+
+<|ref|>text<|/ref|><|det|>[[115, 128, 883, 633]]<|/det|>
+All results were expressed as mean \(\pm\) SEM. Results were analyzed with GraphPad Prism software (version 8.3.0, San Diego, CA, USA). Two- tailed unpaired Student's \(t\) test was used for the comparison of two sets of Gaussian distributed data. Gaussian distributions were tested using the Kolmogorov- Smirnov test. Data conforming to the Gaussian distribution were analyzed with parametric tests, while data with non- Gaussian distribution were analyzed with nonparametric tests. To compare the difference between two groups, Student \(t\) tests were performed. For non- Gaussian distributed data, the Mann- Whitney \(U\) rank sum test was used for one- way analysis. Unless otherwise stated, differences in means between multiple groups of Gaussian- distributed data were analyzed by ordinary one- way analysis of variance followed by Bonferroni's multiple comparisons, and for non Gaussian- distributed data, Kruskal- Wallis test was used. Two- way ANOVA followed by Bonferroni's multiple comparisons was used for two- way analysis. Linear regression analysis was performed using GraphPad Prism software. Pearson correlation analysis was used to test for correlations. In all analyses, \(P < 0.05\) was considered statistically significant.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 680, 221, 694]]<|/det|>
+## References  
+
+<|ref|>text<|/ref|><|det|>[[115, 697, 884, 900]]<|/det|>
+1 Baeza- Zapata, A. A., Garcia- Compean, D., Jaquez- Quintana, J. O. & Collaborators. Acute Pancreatitis in Elderly Patients. Gastroenterology 161, 1736- 1740, doi:10.1053/j.gastro.2021.06.081 (2021).  2 Kayar, Y., Dertli, R. & Konur, S. Clinical outcomes of acute pancreatitis in elderly patients: An experience of single tertiary center. Pancreatology 20, 1296- 1301, doi:10.1016/j.pan.2020.06.006 (2020).  3 Satis, H., Kayahan, N., Sargin, Z. G., Karatas, A. & Celiker, D. Evaluation of the clinical course and prognostic indices of acute pancreatitis in elderly patients: a prospective study. Acta Gastro- Enterologica Belgica 83, 413- 417 (2020).  4 Grunewald, M. et al. Counteracting age- related VEGF signaling insufficiency promotes healthy aging and extends life span. Science 373, doi:10.1126/science.abc8479 (2021).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[60, 78, 884, 901]]<|/det|>
+5 Investigators, F. S. Effects of long- term fenofibrate therapy on cardiovascular events in 9795 people with type 2 diabetes mellitus (the FIELD study): randomised controlled trial (vol 366, pg 1849, 2005). Lancet 368, 1420- 1420 (2006). 6 Girman, C. J. et al. Patients with type 2 diabetes mellitus have higher risk for acute pancreatitis compared with those without diabetes. Diabetes Obes Metab 12, 766- 771, doi:10.1111/j.1463- 1326.2010.01231. x (2010). 7 Gonzalez- Perez, A., Schlienger, R. G. & Rodriguez, L. A. Acute pancreatitis in association with type 2 diabetes and antidiabetic drugs: a population- based cohort study. Diabetes Care 33, 2580- 2585, doi:10.2337/dc10- 0842 (2010). 8 Mizumoto, R. & Sakoguchi, T. Relationship between endocrine and exocrine glands of the pancreas. Nihon rinsho. Japanese journal of clinical medicine Suppl, 2250- 2251 (1978). 9 Williams, J. A. & Goldfine, I. D. The insulin- pancreatic acinar axis. Diabetes 34, 980- 986, doi:10.2337/diabetes.34.10.980 (1985). 10 Barreto, S. G., Carati, C. J., Toouli, J. & Saccone, G. T. The islet- acinar axis of the pancreas: more than just insulin. Am J Physiol Gastrointest Liver Physiol 299, G10- 22, doi:10.1152/ajpg.00077.2010 (2010). 11 Chung, K. M. et al. Endocrine- Exocrine Signaling Drives Obesity- Associated Pancreatic Ductal Adenocarcinoma. Cell 181, 832- +, doi:10.1016/j.cell.2020.03.062 (2020). 12 Schludi, B. et al. Islet inflammation and ductal proliferation may be linked to increased pancreatitis risk in type 2 diabetes. JCI Insight 2, doi:10.1172/jci.insight.92282 (2017). 13 Li, X. et al. Islet alpha- cell Inflammation Induced By NF- kappaB inducing kinase (NIK) Leads to Hypoglycemia, Pancreatitis, Growth Retardation, and Postnatal Death in Mice. Theranostics 8, 5960- 5971, doi:10.7150/thno.28960 (2018). 14 Li, J. et al. Pancreatic beta cells control glucose homeostasis via the secretion of exosomal miR- 29 family. J Extracell Vesicles 10, e12055, doi:10.1002/jev2.12055 (2021). 15 Zhou, Y. et al. \(\beta\) - Cell miRNA- 503- 5p Induced by Hypomethylation and Inflammation Promotes Insulin Resistance and \(\beta\) - Cell Decompensation. Diabetes 73, 57- 74, doi:10.2337/db22- 1044 (2024). 16 Liang, R. et al. H19X- encoded miR- 322(424)/miR- 503 regulates muscle mass by targeting translation initiation factors. J Cachexia Sarcopenia Muscle 12, 2174- 2186, doi:10.1002/jcsm.12827 (2021). 17 Hermann, P. C. et al. Nicotine promotes initiation and progression of KRAS- induced pancreatic cancer via Gata6- dependent dedifferentiation of acinar cells in mice. Gastroenterology 147, 1119- 1133 e1114, doi:10.1053/j.gastro.2014.08.002 (2014). 18 Weidemann, B. J. et al. Repression of latent NF- kappaB enhancers by PDX1 regulates beta cell functional heterogeneity. Cell Metab 36, 90- 102 e107, doi:10.1016/j.cmet.2023.11.018 (2024). 19 Jiang, Z. E. et al. Age- associated changes in pancreatic exocrine secretion of the isolated perfused rat pancreas. Lab Anim Res 29, 19- 26, doi:10.5625/lar.2013.29.1.19 (2013). 20 Chvanov, M. et al. Intracellular rupture, exocytosis and actin interaction of endocytic vacuoles in pancreatic acinar cells: initiating events in acute pancreatitis. Journal of Physiology- London 596, 2547- 2564, doi:10.1113/jp275879 (2018). 21 Sendler, M. et al. Cathepsin B- Mediated Activation of Trypsinogen in Endocytosing Macrophages Increases Severity of Pancreatitis in Mice. Gastroenterology 154, 704- 718
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[60, 78, 884, 900]]<|/det|>
+e710, doi:10.1053/j.gastro.2017.10.018 (2018).22 Desai, B. M. et al. Preexisting pancreatic acinar cells contribute to acinar cell, but not islet beta cell, regeneration. J Clin Invest 117, 971- 977, doi:10.1172/JCI29988 (2007).23 Cendrowski, J. et al. Mnk1 is a novel acinar cell- specific kinase required for exocrine pancreatic secretion and response to pancreatitis in mice. Gut 64, 937- 947, doi:10.1136/gutjnl- 2013- 306068 (2015).24 Hardt, P. D. et al. Pancreatic exocrine function in patients with type 1 and type 2 diabetes mellitus. Acta Diabetol 37, 105- 110 (2000).25 Yilmaztepe, A., Ulukaya, E., Ersoy, C., Yilmaz, M. & Tokullugil, H. A. Investigation of fecal pancreatic elastase- 1 levels in type 2 diabetic patients. The Turkish Journal of Gastroenterology : the Official Journal of Turkish Society of Gastroenterology 16, 75- 80 (2005).26 Hardt, P. D. et al. Chronic pancreatitis and diabetes mellitus. A retrospective analysis of 156 ERCP investigations in patients with insulin- dependent and non- insulin- dependent diabetes mellitus. Pancreatology 2, 30- 33, doi:10.1159/000049445 (2002).27 Majumder, S. et al. Diabetes Mellitus is Associated With an Exocrine Pancreatopathy (EP) That is Distinct From Chronic Pancreatitis (CP). Pancreas 44, 1395- 1395 (2015).28 Zechner, D. et al. Diabetes aggravates acute pancreatitis and inhibits pancreas regeneration in mice. Diabetologia 55, 1526- 1534, doi:10.1007/s00125- 012- 2479- 3 (2012).29 Zechner, D. et al. Diabetes increases pancreatic fibrosis during chronic inflammation. Exp Biol Med (Maywood) 239, 670- 676, doi:10.1177/1535370214527890 (2014).30 Rupnik, M. S. & Hara, M. Local dialogues between the endocrine and exocrine cells in the pancreas. Diabetes, doi:10.2337/db23- 0760 (2024).31 Whitcomb, D. C., Buchner, A. M. & Forsmark, C. E. AGA Clinical Practice Update on the Epidemiology, Evaluation, and Management of Exocrine Pancreatic Insufficiency: Expert Review. Gastroenterology 165, 1292- 1301, doi:10.1053/j.gastro.2023.07.007 (2023).32 Kudoh, A., Katagai, H., Takazawa, T. & Matsuki, A. Plasma proinflammatory cytokine response to surgical stress in elderly patients. Cytokine 15, 270- 273, doi:10.1006/cyto.2001.0927 (2001).33 Barbeiro, D. F., Koike, M. K., Coelho, A. M. M., da Silva, F. P. & Machado, M. C. C. Intestinal barrier dysfunction and increased COX- 2 gene expression in the gut of elderly rats with acute pancreatitis. Pancreatology 16, 52- 56, doi:10.1016/j.pan.2015.10.012 (2016).34 Kim, C. I. Clinicopathological study of pancreatic fibrosis in diabetes mellitus. Medical journal of Osaka University 28, 23- 31 (1977).35 Lohr, J. M., Panic, N., Vujasinovic, M. & Verbeke, C. S. The ageing pancreas: a systematic review of the evidence and analysis of the consequences. J Intern Med 283, 446- 460, doi:10.1111/jaim.12745 (2018).36 Sato, T. et al. Age- related changes in normal adult pancreas: MR imaging evaluation. Eur J Radiol 81, 2093- 2098, doi:10.1016/j.ejrad.2011.07.014 (2012).37 Shen, X. et al. miR- 322/- 503 cluster is expressed in the earliest cardiac progenitor cells and drives cardiomyocyte specification. Proc Natl Acad Sci U S A 113, 9551- 9556, doi:10.1073/pnas.1608256113 (2016).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[60, 87, 884, 880]]<|/det|>
+38 Wang, G. et al. Tumour extracellular vesicles and particles induce liver metabolic dysfunction. Nature 618, 374- 382, doi:10.1038/s41586- 023- 06114- 4 (2023).39 Li, K. et al. Salivary Extracellular MicroRNAs for Early Detection and Prognostication of Esophageal Cancer: A Clinical Study. Gastroenterology 165, 932- 945 e939, doi:10.1053/j.gastro.2023.06.021 (2023).40 Jiang, T. Y. et al. PTEN Deficiency Facilitates Exosome Secretion and Metastasis in Cholangiocarcinoma by Impairing TFEB- mediated Lysosome Biogenesis. Gastroenterology 164, 424- 438, doi:10.1053/j.gastro.2022.11.025 (2023).41 Wang, J. et al. Extracellular vesicles mediate the communication of adipose tissue with brain and promote cognitive impairment associated with insulin resistance. Cell Metab 34, 1264- 1279 e1268, doi:10.1016/j.cmet.2022.08.004 (2022).42 Cao, M. et al. Cancer- cell- secreted extracellular vesicles suppress insulin secretion through miR- 122 to impair systemic glucose homeostasis and contribute to tumour growth. Nat Cell Biol 24, 954- 967, doi:10.1038/s41556- 022- 00919- 7 (2022).43 Li, S. et al. Hair follicle- MSC- derived small extracellular vesicles as a novel remedy for acute pancreatitis. J Control Release 352, 1104- 1115, doi:10.1016/j.jconrel.2022.11.029 (2022).44 Hu, Z. et al. MSCs Deliver Hypoxia- Treated Mitochondria Reprogramming Acinar Metabolism to Alleviate Severe Acute Pancreatitis Injury. Adv Sci (Weinh) 10, e2207691, doi:10.1002/advs.202207691 (2023).45 Ryaboshapkina, M. et al. Characterization of the Secretome, Transcriptome, and Proteome of Human beta Cell Line EndoC- betaH1. Mol Cell Proteomics 21, 100229, doi:10.1016/j.mcpro.2022.100229 (2022).46 Saluja, A., Dudeja, V., Dawra, R. & Sah, R. P. Early Intra- Acinar Events in Pathogenesis of Pancreatitis. Gastroenterology 156, 1979- 1993, doi:10.1053/j.gastro.2019.01.268 (2019).47 Kim, Y. A. et al. Severe acute pancreatitis due to tamoxifen- induced hypertriglyceridemia with diabetes mellitus. Chinese Journal of Cancer Research 26, 341- 344, doi:10.3978/j.issn.1000- 9604.2014.05.01 (2014).48 Bhatia, M. et al. Role of substance P and the neurokinin 1 receptor in acute pancreatitis and pancreatitis- associated lung injury. Proceedings of the National Academy of Sciences of the United States of America 95, 4760- 4765, doi:10.1073/pnas.95.8.4760 (1998).49 Schmidt, J. et al. A better model of acute pancreatitis for evaluating therapy. Ann Surg 215, 44- 56 (1992).50 Grosfils, K., Metioui, M., Tiouli, M. & Dehaye, J. P. Isolation of rat pancreatic acini with crude collagenase and permeabilization of these acini with streptolysin O. Research communications in chemical pathology and pharmacology 79, 99- 115 (1993).51 Xiao, X. et al. Pancreatic cell tracing, lineage tagging and targeted genetic manipulations in multiple cell types using pancreatic ductal infusion of adeno- associated viral vectors and/or cell- tagging dyes. Nat Protoc 9, 2719- 2724, doi:10.1038/nprot.2014.183 (2014).52 Obernosterer, G., Martinez, J. & Alenius, M. Locked nucleic acid- based in situ detection of microRNAs in mouse tissue sections. Nature Protocols 2, 1508- 1514, doi:10.1038/nprot.2007.153 (2007).
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[118, 95, 285, 111]]<|/det|>
+## Acknowledgments  
+
+<|ref|>text<|/ref|><|det|>[[115, 130, 880, 225]]<|/det|>
+We are indebted to Scribendi Inc. (Chatham, ON, Canada) for proofreading the manuscript. This work was supported by grants from the National Natural Science Foundation of China (82070843 and 82270844 to Y- X.Z; 82330027 to X.H).  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 243, 305, 260]]<|/det|>
+## Author contributions  
+
+<|ref|>text<|/ref|><|det|>[[115, 278, 883, 410]]<|/det|>
+Conceptualization: Y- X.Z., X.H. and L.L.; Methodology: K- R.L., T- T.L. Y.Z. and X.X; Investigation: Y.Z., W.T. and S- S.W.; Visualization: K- R.L. and Y- X.Z.; Supervision: Y.Z., Y- T.L., and X- A.C.; Writing—original draft: K- R.L. and Y- X.Z.; Writing—review & editing: K- R.L., X.H., S- J.P. and Y- X.Z. All authors discussed the results and commented on the manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 427, 300, 444]]<|/det|>
+## Competing interests  
+
+<|ref|>text<|/ref|><|det|>[[118, 464, 567, 483]]<|/det|>
+Authors declare that they have no competing interests.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 502, 391, 520]]<|/det|>
+## Data and materials availability:  
+
+<|ref|>text<|/ref|><|det|>[[115, 538, 882, 632]]<|/det|>
+All data are available in the main text or the supplementary materials. All other supporting data in this study are available from the Lead Contact (Yunxia Zhu, zhuyx@njmu.edu.cn) on request.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[123, 111, 848, 690]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 688, 881, 910]]<|/det|>
+<center>Fig. 1. Senescent \(\beta\) -cell–derived miR-503-322 promoted pancreatitis in mice. (A) qPCR analysis of Pri-miR-503 expression in islets and acini of 20-week-old control (WT) and \(\beta\) -cell specific miR-503 transgenic (βTG) male mice. \(n = 4\) . (B) qPCR analysis of miR-503 expression in acini of WT and βTG mice. \(n = 3\) . (C) qPCR analysis of Pri-miR-503 expression in pancreas, islets and acini of 12 weeks and 1.5 years old male mice. \(n = 5\) . (D) qPCR analysis of miR-503 and miR-322 expression in islets and acini of 12 weeks and 1.5 years old male mice, respectively. \(n = 5\) . (E) Schematic flow diagram of sponge \(\beta\) -cell miR-503-322 and induced pancreatitis in aged male mice. The 1.4-years C57BL/6J male mice were randomly divided into two groups. The control group (Ctr, \(n = 4\) ) and the experimental group (SP, \(n = 5\) ) were respectively injected with ctr-AAV and miR-503-322 sponge-AAV through pancreatic ductal Infusion. Two months later, AP was induced by intraperitoneal injection (ip.) of caerulein (50 \(\mu \mathrm{g / kg}\) , hourly for 6 consecutive times), and pancreatitis parameters were detected 2 hours </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 84, 881, 290]]<|/det|>
+after the last injection. (F) Representative sections of ZsGreen (green), insulin (red) and nucleus (blue) immunofluorescence co- staining in pancreas of mice 1 month after AAV injection. (G) qPCR analysis of pancreatic miR- 503 and miR- 322 in the Ctr and SP groups 2 months after AAV injection. (H- J) Pancreatic weights after calibration with body weight (H), serum amylase and lipase (I), representative histologic sections of H&E and F4/80 immunohistochemistry (IHC) of pancreas, pancreatic histological scores and quantitation of the number of F4/80 positive cells in pancreatic sections under 200x microscopic view (J) in the Ctr and SP groups after caerulein (50 μg/kg) induced. Data are presented as Mean ± SEM. Data were analyzed using one- way ANOVA followed by Bonferroni's post hoc (A, C, D and I) or unpaired two- tailed Student's t test (B, G, H and J). NS, Not Significant; \(*P < 0.05\) ; \(**P < 0.01\) ; \(***P < 0.001\) .
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[120, 88, 884, 560]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 567, 881, 899]]<|/det|>
+<center>Fig. 2. \(\beta\) cells secreted nano-vesicular miR-503-322 to enter acinar cells. (A) Experimental scheme: PKH67-labeled nano-vesicles derived from islets of WT (WT-βNVs) or βTG islets (βTG-βNVs) were co-incubated with fresh acini for 8 hours. Representative confocal images of PKH67, phalloidin and nuclei in primary acini and quantitation of relative PKH67 fluorescence intensity. (B) Experimental scheme: PKH67-labeled WT-βNVs or βTG-βNVs were infusion into the C57 male mouse pancreas via pancreatic ductal. The pancreas was harvested after 12 hours and stained with frozen sections for amylase and then visualized. Representative confocal images of PKH67, amylase and nuclei of pancreas and quantitation of relative PKH67 fluorescence intensity. (C) qPCR analysis of miR-503 expression in acini of received βNVs. n=3. (D-G) Pancreatitis parameters assay after initial caerulein (50 μg/kg) injection 7 hours in 12-week-old control (WT) and β-cell specific miR-503-322 knock-in (βKI) female mice. H&E and F4/80 immunohistochemistry (IHC) of pancreatic sections (D), quantitation of the number of F4/80 positive cells in pancreatic sections under 200x microscopic view (E), pancreatic histological scores (F) and level of serum amylase (G). Arrows indicate the macrophages. Data are presented as Mean ± SEM. n=5 mice/group. Data were analyzed using one-way ANOVA followed by Bonferroni's post hoc (G) or unpaired two-tailed Student's \(t\) test (A, B, C, E and F). NS, Not Significant; \(*P < 0.05\) ; \(**P < 0.01\) ; \(***P < 0.001\) . </center>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[125, 85, 880, 844]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 846, 880, 901]]<|/det|>
+<center>Fig. 3. Direct elevation of miR-503-322 in acinar cells triggers both acute and chronic pancreatitis. (A) qPCR analysis for pancreatic Pri-miR-503 expression in 8-week-old control (WT), PKI heterozygous (PKI/WT) and PKI homozygous (PKI/KI) male mice. n=5. (B) Weight </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 85, 881, 490]]<|/det|>
+monitoring in WT, PKI/WT and PKI/KI male mice. n=4. (C) Photograph of the pancreas of 8- week- old WT, PKI/WT and PKI/KI male mice. Lines indicate the pancreas. (D) Representative sections of H&E and masson dyeing of 8- week- old PKI male mice. Representative sections of F4/80 (red) and CK19 (red) immunofluorescence staining in pancreas of 8- week- old PKI male mice. (E) Pancreatic weights after calibration with body weight in 8- week- old PKI male mice. n≥3. (F) Survival curves for WT, PKI/WT and PKI/KI male mice. (G) Schematic of acinar cell specific miR- 503- 322 knock- in (EKI) mice: 6- 8 weeks WT and EKI male or female mice were injected intraperitoneally (ip.) with tamoxifen solution, 100 mg/kg, in corn oil, for three consecutive days and sacrificed three days after the last tamoxifen injection. (H) qPCR analysis of Pri- miR- 503 in acinar cells of WT and EKI male mice after 3 times tamoxifen induced. n=4. (I) Representative sections of pancreas of H&E, receptor- interacting serine-threonine kinase 3 (RIPK3) immunohistochemistry and immunofluorescence staining of F4/80 after first tamoxifen injection 5 days in WT and EKI male mice. Arrows indicate the macrophages. (J) Pancreatic histological scores, quantitation of average optical density of RIPK3 and the number of F4/80 positive cells in pancreatic sections under 600x microscopic view for Fig. I. n=5. (K) Survival curves for WT and EKI male and female mice. (L) Pancreatic H&E of WT and EKI female mice after first tamoxifen injection 28 days. (M) Representative sections of pancreatic masson dyeing from EKI male mice after first tamoxifen injection 28 days. ADM, Acinar- to- ductal metaplasia. Data are presented as Mean ± SEM. Data were analyzed using one- way ANOVA followed by Bonferroni's post hoc (A and E), GEE followed by Tukey's post hoc test (B), unpaired two- tailed Student's t test (H and J) or Survival curve analyses (F and K). NS, Not Significant; \(*P< 0.05\) ; \(**P< 0.01\) ; \(***P< 0.001\) .
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[125, 85, 875, 560]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 567, 881, 750]]<|/det|>
+<center>Fig. 4. MiR-503-322 knockout alleviated caerulein-induced acute pancreatitis. (A) Schematic of caerulien-induced AP on 12-week WT and KO male mice. \(n = 5\) . (B-E) Pancreatic weights after calibration with body weight (B), serum amylase (C) and lipase (D) levels, histological score of the pancreas (E) after PBS or caerulein treatment groups. (F and G) Representative sections of pancreatic H&E (F) and immunofluorescence staining of F4/80 (green) after PBS or caerulein treatment 7 hours in WT and KO male mice. Quantitation of the number of F4/80 positive cells in pancreatic sections under 600x microscopic view (G). Arrows indicate the macrophages. Data are presented as Mean ± SEM. Data were analyzed using one-way ANOVA followed by Bonferroni's post hoc. \(n = 4 - 5\) mice/group. NS, Not Significant; \(*P < 0.05\) ; \(**P < 0.01\) ; \(***P < 0.001\) . </center>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[115, 88, 864, 750]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 752, 881, 900]]<|/det|>
+<center>Fig. 5. MiR-503-322 promotes pancreatitis by inhibiting zymogen secretion and acinar-cell proliferation. (A) Extraction of fresh acinar cells from 8-week-old WT and PKI/WT male mice, in vitro stimulation with different concentrations of caerulein for 30 minutes and determination of amylase content in the supernatant. See Materials and Methods for details. n=3. (B and C) Amylase levels after calibration of total content release from acinar cells of 12-week WT and KO male mice (B) and 12-week and 1.5-year C57BL/6J male mice (C) after 30 min of stimulation with caerulein. n=3. (D) After 48 hours of induction by tamoxifen in WT or EKI mice, pancreatic acini were isolated and incubated with or without caerulein (0.01 μM) for </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[63, 85, 881, 437]]<|/det|>
+30 minutes. At the indicated times, cells were harvested, stained for F-actin with phalloidin (red) and nuclei (blue), and evaluated by laser confocal scanning microscopy. Representative fluorescence micrographs of untreated acini (CTR) and acini pretreated by caerulin stimulation (CER). (E) Representative confocal images of WT and PKI/WT male mice acini after incubation with BziPAR for 30 min at \(37^{\circ}C\) and quantification of fluorescence intensity. (F) Detection of serum trypsin activity levels in 16-week WT and PKI/WT male mice and EKI after Tamoxifen injection 5 days. n=5. (G) Quantitation of the number of proliferating acinar cells of 8-week WT and PKI/WT male mice; 28 days after tamoxifen induction in WT and EKI male mice; 4 days after caerulein-induced AP in 12-week WT and KO male mice and 12-week and 1.5-year-old male mice. n=3-5 mice per group. (H-J) Representative sections of immunofluorescence staining of amylase (red) and PCNA (green) in pancreatic sections from 8-week WT and PKI/WT male mice (H), 28 days after tamoxifen induction in WT and EKI male mice (I), 4 days after caerulein-induced AP in 12-week WT and KO male mice (J). Arrows indicate proliferating acinar cells; asterisks are proliferating interstitial cells. (K) Representative sections of immunofluorescence staining of PCNA (green) in pancreatic sections from 12-week and 1.5-year-old male mice. PCNA, proliferating cell nuclear antigen. Arrows indicate proliferating acinar cells. Data are presented as Mean ± SEM. Data were analyzed using GEE followed by Tukey's post hoc test (A, B and C) or unpaired two-tailed Student's t test (F and G). \(^{*}P< 0.05\) ; \(^{**}P< 0.01\) ; \(^{***}P< 0.001\) .
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[113, 87, 875, 655]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 660, 881, 901]]<|/det|>
+<center>Fig. 6. MKNK1 is a target of miR-503-322 and acinar cell-specific restoration of it reverses the phenotype of pancreatitis in mice. (A)The MKNK1 network was predicted based on the common signature from the Ingenuity database overlaid with microarray data from miR-503-overexpressing mouse pancreatic \(\beta\) cell line MIN6 cells with a 1.5-fold change cutoff compared with negative control cells. (B) WT and EKI male mice at 5 days after tamoxifen induction; Male C57BL/6J at 12-week and 1.5-year; male WT and KO at 12-week after AP induced and \(\beta\) -cell specific sponge of miR-503-322 in control and experimental mice pancreatic protein western blotting. \(n = 3 - 5\) . (C) Experimental scheme: 8-week-old WT male mice were injected intraperitoneally (ip.) with control AAV and EKI male mice were injected with control (Ctr-AAV) and MKNK1-AAV, respectively, one month later tamoxifen was induced for 3 consecutive days and tested at day 7. \(n = 5\) . (D and E) Immunofluorescence staining of Flag and MKNK1 of pancreas sections (D) from each group of mice at 13 weeks and western blotting of pancreatic proteins (E). (F) Gain of body weight, serum amylase and lipase level </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[60, 85, 883, 201]]<|/det|>
+797 were monitored during tamoxifen induction. (G) Representative images of H&E and F4/80 798 immunohistochemistry of pancreas in each group. Arrows indicate the macrophages. (H) 799 Quantitation of the number of F4/80 positive cells in each group and the pancreatic histological 800 score. Data are presented as Mean ± SEM. Data were analyzed using one- way ANOVA 801 followed by Bonferroni's post hoc (F and H). NS, Not Significant; \(*P < 0.05; **P < 0.01; ***P < 0.001\) . 802 803
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[118, 80, 830, 730]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 734, 881, 900]]<|/det|>
+<center>Fig. 7. Evidence of miR-503 and MKNK1 in aging-associated pancreatitis changes in the Chinese population. (A) Representative images of H&E and Masson staining of pancreatic sections from the young adult (YA) and the elderly adult (EA); quantitation of collagen volume fraction. The dashed area indicates acini. n=10. (B) Representative images of immunofluorescence staining of PCNA (green) in pancreatic sections and counted the number of PCNA-positive cells. n=10. Arrows indicate proliferating acinar cells. (C) In situ hybridization of miR-503 (40 nM) in young and elderly pancreatic sections. Scramble-RNA was negative reference (40 nM) and U6 was positive reference (0.1 nM). The dotted line indicate pancreatic islet and solid line is exocrine. (D) Representative images of immunofluorescence staining of </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[60, 85, 881, 308]]<|/det|>
+MKNK1 (green) and amylase (red) in pancreatic sections from young and elderly people and quantitation of amylase and MKNK1 mean fluorescence intensity. \(\mathrm{n = 10}\) . (E) Serum amylase assay of the young adult (YA), the elderly adult (EA) and the elderly adult with diabetes (EA+DM). YA group, \(\mathrm{n = 65}\) . EA group, \(\mathrm{n = 65}\) . EA+DM, \(\mathrm{n = 30}\) . (F) MiR- 503 concentration in human serum of YA, EA and EA+DM. YA group, \(\mathrm{n = 45}\) . EA group, \(\mathrm{n = 45}\) . EA+DM, \(\mathrm{n = 30}\) . (G) Correlation analysis of amylase levels of human serum and age. Each point represents one people ( \(\mathrm{n} = 160\) ). Correlation coefficient (R) and p value from simple linear regression are shown. (H) Correlation analysis of miR- 503 concentration in human serum and serum amylase levels. Each point represents one people ( \(\mathrm{n = 120}\) ). Correlation coefficient (R) and P value from simple linear regression are shown. Data are presented as Mean \(\pm\) SEM. Data were analyzed using one- way ANOVA followed by Bonferroni's post hoc (E and F) or unpaired two- tailed Student's t test (A, B and D) or Correlation analysis (G and H). \(^{**}P< 0.01\) ; \(^{***}P< 0.001\) .
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[43, 42, 312, 70]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[43, 93, 768, 113]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[60, 130, 333, 150]]<|/det|>
+SupplementaryMaterials.pdf
+
+<--- Page Split --->

preprint/preprint__0f731c80040fa9beeef1b271f00ca5c000eaea0664e712d70f287e8bda9d7779/images_list.json

@@ -0,0 +1,145 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "FIG. 1: Overview of Simulated Annealing Design using Dimer Likelihood Estimation (SADDLE). Given a set of DNA target sequences \\(\\mathrm{T}_i\\) , with \\(i\\) between 1 and \\(N\\) , the goal is to design a total of \\(2N\\) PCR primers that can effectively amplify all DNA targets, while generating an acceptably low amount of primer dimer species. Steps 4 and 5 can be repeated a large number of times in order to improve (decrease) the Loss function value on the final primer set S. Multiple implementations, hyper-parameters, and parameters can be selected for each SADDLE step that can impact performance and speed.",
+    "footnote": [],
+    "bbox": [
+      [
+        108,
+        58,
+        876,
+        456
+      ]
+    ],
+    "page_idx": 2
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "FIG. 2: Implementation and experimental evaluation of a multiplex primer design algorithm based on the SADDLE framework. (a) Method for generating candidate primer sequences for a DNA target T. (b) Implementation of Badness function that can be rapidly evaluated using hash tables. (c) List of cancer genes selected as target sequences for a 96-plex primer set design. See Supplementary Excel spreadsheet for target selection details. (d) Loss function of primer sets \\(S(g)\\) across optimization generations \\(g\\) . The Loss function value decreases through the optimization and approaches a local minima after roughly 400 generations. We selected three different primer sets, constructed at generations 0, 200, and 400 for experimental evaluation; these are respectively called PS1, PS2, and PS3 for the remainder of this paper. (e) Capillary electrophoresis (Agilent Bioanalyzer 2100) analysis of amplicon products of PS1, PS2, and PS3. Here, 10 ng of the NA18562 human genomic DNA (Coriell) was used as input, and the median primer concentration was 45 nM in the PCR reaction. 17 cycles of PCR were performed using Vent (exo-) DNA polymerase (selected for its improved amplification ability for G/C-rich sequences). To facilitate more in-depth analysis by high-throughput sequencing (NGS), adapters/indexes were ligated to the amplicon products. No size selection was performed, in order to accurately reflect the fraction of primer dimer species following multiplex PCR. The on-target amplicons are expected to have an average length of roughly 250 nt.",
+    "footnote": [],
+    "bbox": [
+      [
+        58,
+        50,
+        490,
+        360
+      ]
+    ],
+    "page_idx": 5
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "FIG. 3: Experimental NGS results for SADDLE-designed primer sets. (a) Distribution of reads observed in NGS library constructed using PS1, PS2, and PS3. On Target reads are defined as those that aligned to the intended amplicons; Dimer reads are defined as those whose insert lengths are smaller than the sum of the two primer lengths; all other reads were classified as Non-specific. The vast majority of Non-specific reads align to other regions of the human genome, via a non-cognate pair of forward and reverse primers. The fraction of NGS reads mapped to Dimers dramatically decreases from PS1 to PS2 to PS3. (b) Distribution of NGS reads in the 3 primer set libraries. (c) Distribution of observed primer dimers, based on aligned reads. Because forward primers (FP) can also form primer dimers with other forward primers, we aligned the first and last 25 nucleotides of each NGS read to the merged set of FP and rPs, with primers 1 through 96 in the diagram showing fPs and primers 97 through 192 showing rPs. For clarity of visualization, the log number of reads of observed primer dimers are displayed via both coloration and circle size. (d) Performance of the PS3 primer set of formalin-fixed paraffin-embedded (FFPE) tissue samples from deidentified lung cancer patients. Because the NGS libraries for these 5 samples differed slightly in total reads, here we plotted the distribution of reads normalized to 1 million reads. (e) The observed primer dimer species and their corresponding NGS reads were relatively similar between cell line genomic DNA and FFPE samples. (f) Demonstration of a 384-plex primer set designed by SADDLE (768 primers). The main diagonal shows On Target reads. Only about \\(1\\%\\) of all reads were primer dimers (Supplementary Section S6).",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 5
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "FIG. 4: Evaluation of prediction accuracy of the Badness function for individual primer dimer candidates. (a) Comparison of observed vs. predicted primer dimers for all possible pairs of primers in the PS1 library. The horizontal orange line shows the mean on-target reads for the 96 intended amplicons. By changing the Badness Threshold, different tradeoffs of dimer prediction sensitivity and specificity can be achieved. For the current Reads Threshold and Badness Threshold, we observe \\(92.5\\%\\) sensitivity and \\(90.3\\%\\) specificity. (b) Receiver Operator Characteristic (ROC) curve for dimer prediction sensitivity and specificity achieved by changing the Badness Threshold. (c) The Area Under the ROC (AUROC) value depends on the Reads Threshold, with a highest achievable AUROC of about 0.98. (d) The top 5 dimer species experimentally observed to form with highest number of aligned NGS reads. (e) The top 5 potential dimer species predicted to form based on the Badness function. Note that only the \\(\\# 4\\) species are present are both list in panels (d) and (e). (f) Distribution of observed NGS reads and predicted Badness for all possible primer dimers.",
+    "footnote": [],
+    "bbox": [
+      [
+        60,
+        50,
+        930,
+        555
+      ]
+    ],
+    "page_idx": 6
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "FIG. 5: Highly multiplexed qPCR detection of gene fusions using SADDLE-designed primer sets. (a) Complementary DNA (cDNA) prepared through reverse transcription of RNA can have known target sequences at the exon breakpoints. Although it is trivial to design a single-plex qPCR assay to detect a single known fusion, such as BCR-ABL1 [20], we are not aware of any reports of highly multiplexed qPCR assays to simultaneously detect \\(\\geq 10\\) different gene fusion cDNA species. For this assay, we designed a 60 primer set (46 forward, 14 reverse) that together can amplify 56 distinct gene fusion types commonly observed in non-small-cell lung cancer [21]. (b) Summary of observed qPCR cycle threshold (Ct) values for the 56 reactions, each with 1 of the 56 synthetic fusion DNA species across 6 genes (ALK, ROS1, RET, and NTRK1/NTRK2/NTRK3), each with 1700 copies. WT indicates wildtype commercial cDNA, and NTC indicates no template control. See Supplementary Section S8 for additional details and experimental results, including Sanger sequencing traces of each reaction product. (c) Example qPCR trace showing detection of the fusion DNA sequence joining NACC2 exon 4 to NTRK2 exon 13. (d) Clinical sample results on cDNA reverse transcribed from RNA from extracellular vesicles. Samples 3, 4, and 7 tested positive for a gene fusion, and sequence alignment of the Sanger sequencing results (right panels) show the exact identifies of the fusions.",
+    "footnote": [],
+    "bbox": [
+      [
+        105,
+        50,
+        888,
+        315
+      ]
+    ],
+    "page_idx": 7
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Figure 1",
+    "footnote": [],
+    "bbox": [
+      [
+        60,
+        115,
+        920,
+        560
+      ]
+    ],
+    "page_idx": 8
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2",
+    "footnote": [],
+    "bbox": [
+      [
+        58,
+        66,
+        941,
+        377
+      ]
+    ],
+    "page_idx": 12
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Figure 3",
+    "footnote": [],
+    "bbox": [
+      [
+        55,
+        65,
+        940,
+        430
+      ]
+    ],
+    "page_idx": 13
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Figure 4 Evaluation of prediction accuracy of the Badness function for individual primer dimer candidates. (a) Comparison of observed vs. predicted primer dimers for all possible pairs of primers in the PS1 library. The horizontal orange line shows the mean on- target reads for the 96 intended amplicons. By changing the Badness Threshold, different tradeoffs of dimer prediction sensitivity and specificity can be achieved. For the current Reads Threshold and Badness Threshold, we observe \\(92.5\\%\\) sensitivity and \\(90.3\\%\\) specificity. (b) Receiver Operator Characteristic (ROC) curve for dimer prediction sensitivity and specificity achieved by changing the Badness Threshold. (c) The Area Under the ROC (AUROC) value depends on the Reads Threshold, with a highest achievable AUROC of about 0.98. (d) The top 5 dimer species experimentally observed to form with highest number of aligned NGS reads. (e) The top 5 potential dimer species predicted to form based on the Badness function. Note that only the #4 species are present are both list in panels (d) and (e). (f) Distribution of observed NGS reads and predicted Badness for all possible primer dimers.",
+    "footnote": [],
+    "bbox": [
+      [
+        60,
+        60,
+        936,
+        570
+      ]
+    ],
+    "page_idx": 14
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Figure 5",
+    "footnote": [],
+    "bbox": [
+      [
+        65,
+        50,
+        916,
+        345
+      ]
+    ],
+    "page_idx": 15
+  }
+]

preprint/preprint__0f731c80040fa9beeef1b271f00ca5c000eaea0664e712d70f287e8bda9d7779/preprint__0f731c80040fa9beeef1b271f00ca5c000eaea0664e712d70f287e8bda9d7779.mmd

@@ -0,0 +1,302 @@
+
+# Designing Highly Multiplex PCR Primer Sets with Simulated Annealing Design using Dimer Likelihood Estimation (SADDLE)  
+
+David Zhang (dyz1@rice.edu)  
+
+Rice University  
+
+Nina Xie Rice University  
+
+Michael Wang Rice University https://orcid.org/0000- 0001- 7009- 6958  
+
+Ping Song Rice University  
+
+Yifan Wang Nuprobe China  
+
+Yuxia Yang Nuprobe China  
+
+Junfeng Luo  
+
+Nuprobe China  
+
+## Article  
+
+Keywords: PCR primers, DNA sequences  
+
+Posted Date: May 13th, 2021  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 491811/v1  
+
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Version of Record: A version of this preprint was published at Nature Communications on April 11th, 2022. See the published version at https://doi.org/10.1038/s41467- 022- 29500- 4.
+
+<--- Page Split --->
+
+# Designing Highly Multiplex PCR Primer Sets with Simulated Annealing Design using Dimer Likelihood Estimation (SADDLE)  
+
+Nina G. Xie\*, \(^{1}\) Michael X. Wang\*, \(^{1}\) Ping Song, \(^{1}\) Yifan Wang, \(^{2}\) Yuxia Yang, \(^{2}\) Junfeng Luo, \(^{2}\) and David Yu Zhang \(^{1 + 1}\)  
+
+\(^{1}\) Department of Bioengineering, Rice University, Houston, TX \(^{2}\) NuProbe China, Shanghai, China(Dated: May 3, 2021)  
+
+The design of highly multiplex PCR primers to amplify and enrich many different DNA sequences is increasing in biomedical importance as new mutations and pathogens are identified. One major challenge in the design of highly multiplex PCR primer sets is the large number of potential primer dimer species that grows quadratically with the number of primers to be designed. Simultaneously, there are exponentially many choices for multiplex primer sequence selection, resulting in systematic evaluation approaches being computationally intractable. Here, we present and experimentally validate Simulated Annealing Design using Dimer Likelihood Estimation (SADDLE), a stochastic algorithm for design of highly multiplex PCR primer sets that minimize primer dimer formation. Our approach uses a rapidly computable Loss function to approximate the degree of primer dimer formation within a primer set, and randomly swaps primers in the set with alternative candidates using a simulated annealing algorithm. In a 96- plex PCR primer set (192 primers), we show that we can reduce the fraction of primer dimers from \(90.7\%\) in a naively designed PCR primer set to \(4.9\%\) in our optimized primer set. Running the optimized 96- plex primer set on FFPE DNA samples from cancer patients, we likewise observe a low fraction of primer dimer reads. Even when scaling to 384- plex (768 primers), the PCR primer set designed by our algorithm maintains low primer dimer fraction. In addition to NGS, we also show that our SADDLE- designed multiplex primer sets can be used in qPCR settings to allow highly multiplexed detection of gene fusions in cDNA, with a single- tube assay comprising 60 primers detecting 56 distinct gene fusions recurrently observed in lung cancer.  
+
+The advance of high throughput sequencing has uncovered a large number of biomedically relevant DNA sequences, from driver mutations in cancer to new bacterial/viral pathogen DNA sequences to microbiome metagenomic profiles that affect mental disorders on the gut- brain axis [1- 4]. For discovery applications, "shotgun" whole genome sequencing (WGS) is the preferred approach to identify novel DNA sequences of interest [5]. However, the human genome comprises over 3 billion nucleotides, and despite the lowering costs of high- throughput sequencing, it is not practical today to perform WGS to high depths necessary for identification of subclonal mutations, such as somatic mutations in cancer. For routine detection of disease- relevant DNA variants in known genes of interest, targeted sequencing or direct qPCR approaches are typically used [6, 7]. Of the two dominant methods today for target enrichment, multiplex PCR tends to have shorter workflows and require less DNA input than hybrid- capture probes [8]. However, multiplex PCR struggles to scale to large panels covering hundreds of genes, due to the nonlinear increase of primer dimer species that reduce NGS mapping rates and increase effective cost[10].  
+
+Currently, multiplex PCR methods for NGS target enrichment (e.g. Ampliseq [8]) primarily rely on (1) enzymatic digestion of primer dimer species at mismatch bubbles [9] and (2) DNA size selection to preferentially remove short amplicon species likely to be primer dimers. However, both steps are labor- intensive and does not allow full removal of all primer dimer species. In contrast, relatively little systematic work have been reported on computational approaches to minimizing the formation of primer dimers in the first place. The development of a robust multiplex primer set design algorithm that produces highly multiplexed primer sets with minimal primer dimer formation could allow further scaling of multiplex PCR target enrichment to even larger NGS panels when combined with enzymatic and size selection methods. Alternatively, it can simplify the workflow of moderate size NGS and qPCR assays by removing the need for strict contamination control from open- tube steps.   
+
+There are two primary challenges in designing highly multiplexed PCR primer sets: First, for an \(N\) - plex PCR primer set comprising \(2N\) primers, there are \(\binom{2N}{2}\) possible simple primer dimer interactions. For \(N = 50\) , this corresponds to \(\binom{100}{2} = 4950\) times as many potential primer dimer bindings as for a single- plex PCR primer set. Second, there are typically \(M > 10\) reasonable candidate choices for each primer when considering specific gene targets and amplicon length constraints, resulting in \(M^{2N}\) possible N- plex primer sets. For \(M = 20\) and \(N = 50\) , the number of possible primer sets is \(20^{100} \approx 1.3 \cdot 10^{130}\) , billions of times larger than the number of atoms in the universe. Thus, it is computationally intractable to evaluate all possible multiplex primer sets. Simultaneously, primer dimer formation emerges from the interactions of two or more primers in the primer set, so changing the sequence of a primer to mitigate one primer dimer interaction may result in the appearance of another more serious primer dimer. In the language of numerical optimization, multiplex primer design is high dimensional problem with a highly non- convex fitness landscape. Consequently, standard convex optimization algorithms (e.g. gradient descent) will not be effective.  
+
+Here, we present Simulated Annealing Design using Dimer Likelihood Estimation (SADDLE), an algorithmic framework
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>FIG. 1: Overview of Simulated Annealing Design using Dimer Likelihood Estimation (SADDLE). Given a set of DNA target sequences \(\mathrm{T}_i\) , with \(i\) between 1 and \(N\) , the goal is to design a total of \(2N\) PCR primers that can effectively amplify all DNA targets, while generating an acceptably low amount of primer dimer species. Steps 4 and 5 can be repeated a large number of times in order to improve (decrease) the Loss function value on the final primer set S. Multiple implementations, hyper-parameters, and parameters can be selected for each SADDLE step that can impact performance and speed. </center>  
+
+for designing highly multiplex PCR primer sets. Within this framework, we present an example multiplex primer design algorithm, comprising an algorithm for primer candidate generation and a rapidly computable Loss function for estimating primer dimers. Using the SADDLE, we designed and experimentally tested multiplex primer sets comprising 192 primers (96- plex) and 784 primers (384- plex), and show low primer dimer formation through NGS experiments. Building upon this success, we built a single- tube 60 primer qPCR and Sanger assay to detect and identify 56 gene fusions with clinical actionability for non- small cell lung cancer.  
+
+## Results  
+
+## Simulated Annealing Design using Dimer Loss Estimation (SADDLE).  
+
+There are 6 main steps in SADDLE, as illustrated in Fig. 1:  
+
+1. Generation of forward primer (fP) and reverse primer (rP) candidates for each gene target. 
+2. Selection of an initial primer set \(\mathrm{S}_0\) from the primer candidates. 
+3. Evaluation of the Loss function \(\mathrm{L}(\cdot)\) on the initial primer set \(\mathrm{S}_0\) .  
+
+4. Generate a temporary primer set \(\mathrm{T}\) based on set \(\mathrm{S}_g\) (primer set S from generation G) by randomly changing 1 or more primers.  
+
+5. Evaluate \(\mathrm{L}(\mathrm{T})\) , and set \(\mathrm{S}_{g + 1}\) to either \(\mathrm{S}_g\) (no change) or \(\mathrm{T}\) , depending probabilistically on the relative values of \(\mathrm{L}(\mathrm{S}_g)\) and \(\mathrm{L}(\mathrm{T})\) .  
+
+6. Repeat steps 4 to 5 until an acceptable primer set \(\mathrm{S}_{\mathrm{final}}\) is constructed.  
+
+The above abstract framework provides a basis for many potential multiplex primer design algorithms, depending on the specific details of primer candidate generation, form of Loss function, temporary set \(\mathrm{T}\) generation, and the dynamic probability of setting \(\mathrm{S}_{g + 1}\) to \(\mathrm{T}\) . Below, we describe our specific implementation of SADDLE, based on our accumulated understanding of primer design principles and primer dimer formation mechanisms. Given the infinite possibilities for function forms and hyper- parameters, we did not systematically evaluate or optimize at the high- level. Lower- level parameters, such as standard free energy \((\Delta G^{\circ})\) ranges for primers, were experimentally optimized and these are described below.
+
+<--- Page Split --->
+
+1. Primer candidate generation. We begin our implementation of primer candidate generation through the selection of one or more "pivot" nucleotides on human genomic DNA around which we design the forward primer (Fig. 2a). The pivot nucleotides are the ones that must be included in the amplicon insert, and for example could be the hotspot region of a gene that is frequently mutated. From the pivot nucleotides and a constraint on the maximum length of the amplicons (e.g. determined by the read length of NGS), we can systematically generate a series of different proto-primers with \(3'\) end just outside the pivot nucleotides. The proto-primers have a large range of different lengths and binding energies to their complementary sequences, and will next be trimmed at the \(3'\) end to generate the primer candidates (Fig. 2a).  
+
+From our past experiences and preliminary optimization experiments, primers that hybridize to their cognate templates with \(\Delta G^{\circ} \approx - 11.5 \mathrm{kcal / mol}\) have the best tradeoff between amplification efficiency/uniformity and nonspecific hybridization. Shorter primers may not bind consistently with high efficiency to their templates, resulting in variability in amplification efficiency and non-uniformity of amplicon on-target reads in the NGS library. Longer primers have increased likelihood of binding to other loci in human genome, and can result in non-specific amplicons. Based on this \(\Delta G^{\circ}\) goal, we next systematically constructed primer candidates from the proto-primers by truncating nucleotides from the \(3'\) end until the primer candidate has \(\Delta G^{\circ}\) between - 10.5 kcal/mol and - 12.5 kcal/mol. Due to the granularity of \(\Delta G^{\circ}\) for base stacks, some proto-primers with the same \(5'\) end will result in multiple primer candidates (e.g. with \(\Delta G^{\circ} = - 10.9 \mathrm{kcal / mol}\) and - 12.0 kcal/mol). Optionally, one could implement additional filters here to remove undesirable primer candidates, such as based on G/C content. For our demonstration panels, we restrict the G/C content of primer candidates to be between 0.25 and 0.75, removing primer candidates with G/C content outside this range.  
+
+In the implementation of SADDLE, primer candidates can be treated as individual primers or as primer pairs. Our specific implementation treats primers as pairs, so we next combinatorially generate all candidate primer pairs for an amplicon, in order to better constrain the distribution of amplicon lengths Any candidate primer pairs that generate amplicons with length exceeding our maximum amplicon length or below our minimum are removed.  
+
+2. Initial primer set \(\mathbf{S}_0\) selection. We randomly selected a primer pair candidate for every amplicon that we wish to design, and collectively the selected primers are known as the initial primer set \(\mathbf{S}_0\) .  
+
+3. Evaluation of Loss function \(\mathbf{L}(\cdot)\) on \(\mathbf{S}_0\) . The Loss function \(\mathbf{L}(\cdot)\) is a rapidly computable function that aims to approximate the severity of primer dimer formation by a primer set \(\mathbf{S}\) . \(\mathbf{L}(\cdot)\) sums the potential primer dimer interactions between every pair of primers in the primer set. To prevent confusion, we refer to the predicted formation of dimers for a  
+
+particular pair of primers to be the Badness. Mathematically,  
+
+\[\mathrm{L}(\cdot) = \sum_{a = 1}^{2N}\sum_{b = a}^{2N}\mathrm{Badness}(a,b) \quad (1)\]  
+
+One can imagine the Badness function to be proportional to the amount of primer dimers formed by two primers. In an optimized primer set with a relatively low concentration of primer dimers compared to the concentration of on- target amplicons, the amount of primer dimers formed between primer \(a\) and \(b\) should not significantly impact the amount of dimers formed between \(a\) and \(c\) , so the Loss function being defined as the sum of the component Badness functions is justified.  
+
+The Badness function in our implementation is defined as follows (Fig. 2b):  
+
+\[\mathrm{Badness}(a,b) = \sum \frac{1}{(i + 1)(j + 1)}\cdot 2^{\mathrm{Length}}\cdot 2^{\mathrm{numGC}} \quad (2)\]  
+
+The sum in the Badness definition is over all subsequences between primer \(a\) and \(b\) with at least 4 nt of continuous complementarity. The lower limit of 4 nt is both based on our preliminary experimental studies in qPCR showing that up to 3 nt of complementarity at the \(3'\) ends of two primers will not result in significant primer dimers even in no- template control (NTC) reactions.  
+
+For each complementary subsequence, its Length and its number of GC nucleotides in complementary subsequence (numGC) contribute exponentially to Badness. Thus, the exponential components of Badness roughly reflect the partition function of the complementarity interaction, with G/C base pairs roughly twice as strong as A/T base pairs. We chose to use these simplistic parameters, rather literature base stacking thermodynamics parameters [11], because there is significant uncertainty in the effective salinity of the PCR reaction buffer, and because our previous studies on DNA thermodynamics suggests that previously reported \(\Delta H^{\circ}\) and \(\Delta S^{\circ}\) parameters do not extrapolate well to higher temperatures [12].  
+
+The distances of the complementary subsequence to the \(3'\) ends of primers \(a\) and \(b\) , denoted as \(i\) and \(j\) , are known to significantly affect the likelihood of primer dimer formation. In our preliminary qPCR experiments, we observed that a primer pair with 10 nt of complementarity at the \(5'\) end will not result in observable primer dimer formation, but a primer pair with 5 nt of complementarity at the \(3'\) would. Depending on whether the specific DNA polymerase used, different \(i\) - and \(j\) - based attenuation of Badness may be optimal for minimizing primer dimers. Because high- fidelity DNA polymerases with \(3' > 5'\) exonuclease activity can remove mismatched \(3'\) nucleotides, the optimal \(i\) - and \(j\) - based attenuation should be significantly weaker for high- fidelity DNA polymerases.  
+
+The evaluation of the Badness function is single largest component of software runtime cost, due to the large number of times the Badness function will be evaluated. For a primer that is 25 nt in length, there are 22 subsequences of length 4, 21 subsequences of length 5, etc. Evaluation of Badness for a
+
+<--- Page Split --->
+
+single primer pair would thus have time complexity of \(\mathrm{O}(P^{3})\) where \(P\) is the length of each primer. However, due to the additive nature of subsequence Badness components to the overall Badness function, we implement more rapid Badness evaluation using a hash table [13]. The time complexity to set up the hash table is \(\mathrm{O}(P^{2})\) for all subsequences of the first primer of a minimum length, and the time complexity to evaluate the Badness by stepping through all subsequences of the second primer is also \(\mathrm{O}(P^{2})\) . Consequently, the overall time complexity of evaluating the Loss function is \(\mathrm{O}(N^{2}\cdot P^{2})\) for all primers in \(\mathrm{S}_{0}\) .  
+
+4. Generate temporary primer set T based on \(\mathbf{S}_{g}\) . Step 4 begins the recursive optimization process. Based on the current primer set \(\mathbf{S}_{g}\) at generation \(g\) , we first randomly select one primer pair to "mutate." For that primer pair, we randomly select a different primer pair from the list of all candidate primer pairs generated in Step 1. Temporary primer set T is thus generated by combining this new primer pair with all remaining primers in set \(\mathbf{S}_{g}\) . Optionally, multiple primer pairs can be replaced simultaneously in this step to allow faster and more efficient exploration of the primer set space. In our preliminary in silico evaluations, we found that simultaneously mutating multiple primer pairs generally caused a slowdown of the optimization process.  
+
+5. Evaluate \(\mathbf{L}(\mathbf{T})\) and set \(\mathbf{S}_{g + 1}\) to be either T or \(\mathbf{S}_{g}\) . The Loss of temporary primer set T can be evaluated significantly faster than the initial evaluation of \(\mathrm{L}(\mathrm{S}_{0})\) , because many of the Badness components are the same as in \(\mathbf{S}_{g}\) :  
+
+\[\mathrm{L}(\mathrm{T}) = \mathrm{L}(\mathrm{S}_{g}) - \sum_{b}(\mathrm{Badness}(a_{\mathrm{old}},b) - \mathrm{Badness}(a_{\mathrm{new}},b))\]  
+
+Thus, the time complexity of this step is only \(\mathrm{O}(N\cdot P^{2})\) , rather than \(\mathrm{O}(N^{2}\cdot P^{2})\) .  
+
+We next compare the value of \(\mathrm{L}(\mathrm{T})\) vs. \(\mathrm{L}(\mathrm{S}_{g})\) . If \(\mathrm{L}(\mathrm{T})\) is smaller than \(\mathrm{L}(\mathrm{S}_{g})\) , then the primer pair change was an improvement and accepted, so \(\mathrm{S}_{g + 1}\) is set to T. If \(\mathrm{L}(\mathrm{T})\) is larger than \(\mathrm{L}(\mathrm{S}_{g})\) , the change was detrimental, but we will still accept the change with a certain probability, as part of the simulated annealing algorithm [14]. To clarify, "simulated annealing" here refers to a specific computer science algorithm, and not a literal simulation of a physical DNA thermal annealing process. If we never accept any detrimental primer pair changes, then the approach degenerates to become a stochastic gradient descent approach. In preliminary in silico evaluations, we confirmed that stochastic gradient descent produces final primer sets with significantly worse Loss, because it becomes too easy to get stuck in a local Loss minima.  
+
+The probability of accepting a detrimental change depends on both the magnitude of the detriment \((\mathrm{L}(\mathrm{T}) - \mathrm{L}(\mathrm{S}_{g}))\) and the generation \(g\) of the optimization. Worse changes with higher \(\mathrm{L}(\mathrm{T})\) are accepted with lower probability, and later generations of the optimization (higher \(g\) ) are less tolerant of detrimental changes. In our implementations, the probability of setting \(\mathrm{S}_{g + 1}\) to be T when \(\mathrm{L}(\mathrm{T})\) is greater than \(\mathrm{L}(\mathrm{S}_{g})\) are  
+
+as follows:  
+
+\[p = e^{(\mathrm{L}(\mathrm{S}_{g}) - \mathrm{L}(\mathrm{T})) / C(g)}\quad \mathrm{if~}g< g_{t} \quad (3)\]  
+
+\(C(g)\) is a function that is monotonically non- increasing in \(g\) , indicating decreasing tolerance to detrimental changes at later generations. The parameter \(g_{t}\) indicates the generation in which simulated annealing terminates, and we switch over to stochastic gradient descent.  
+
+6. Repeat steps 4 and 5. Steps 4 and 5 are repeated until either a pre-determined generation \(g\) , or until \(\mathrm{L}(\mathrm{S}_{g})\) is below a pre-determined threshold \(\mathrm{L}_{t}\) . In our implementation, we typically run the optimization to about \(1.5\cdot g_{t}\) to ensure we reach local minima. To further improve the overall quality of the generated primer set, we recommend running multiple SADDLE optimization processes with different starting conditions (initial primer sets) and selecting the best final primer set.  
+
+Design and Experimental Evaluation of a 96- plex Primer Set. We first used SADDLE to optimize the design of a 96- plex primer set, each amplicon target one arbitrarily selected exon of a different cancer- related gene [15] (Fig. 2c). Fig. 2d shows the calculated value of \(\mathrm{L}(\mathrm{S}_{g})\) at different generations \(g\) , and is representative of our typical optimization trajectory. We selected the designed primer sets at three different optimization generations for experimental testing: PS1 (initial unoptimized primer set), PS2 (primer set with intermediate Loss optimization), and PS3 (primer set with saturating Loss optimization). The primer set Loss decreased roughly 24- fold from PS1 to PS3; after 40,000 generations, only very marginal improvements were observed. We chose the primer set at 40,000 generations as PS3, rather than the one at 60,000 generations, because we know that our Loss function is an imperfect predictor of primer dimers. Overtraining on an imperfect Loss function can lead to worse experimental results.  
+
+We applied each of the three primer sets individually to human genomic DNA (10ng NA18562, sheared to a mean length of approximately 150 nt) and amplified for 17 cycles. We next constructed NGS libraries from the amplicons generated using PS1, PS2, and PS3, using a standard adaptor ligation protocol (Supplementary Section S1). After library preparation, capillary electrophoresis results show a clear increase of amplicons of the expected length from PS1 to PS2 to PS3 (Fig. 2e). In the NGS data analysis workflow, after the first step of adapter trimming, we separated NGS reads into three major species: on- target amplicons, dimers, and non- specific amplicons (Supplementary Section S2). On- target amplicons are the NGS reads that were successfully aligned to the intended amplicon sequences using Bowtie2[19]. The remaining NGS reads were aligned separately to each forward and reverse primer sequence. Reads with insert length shorter than the sum of the two aligned primers are classified as Dimers, and reads with insert length longer than the sum of the two aligned primers are classified as Non- specific amplicons (amplifying unintended regions of the human genome).
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>FIG. 2: Implementation and experimental evaluation of a multiplex primer design algorithm based on the SADDLE framework. (a) Method for generating candidate primer sequences for a DNA target T. (b) Implementation of Badness function that can be rapidly evaluated using hash tables. (c) List of cancer genes selected as target sequences for a 96-plex primer set design. See Supplementary Excel spreadsheet for target selection details. (d) Loss function of primer sets \(S(g)\) across optimization generations \(g\) . The Loss function value decreases through the optimization and approaches a local minima after roughly 400 generations. We selected three different primer sets, constructed at generations 0, 200, and 400 for experimental evaluation; these are respectively called PS1, PS2, and PS3 for the remainder of this paper. (e) Capillary electrophoresis (Agilent Bioanalyzer 2100) analysis of amplicon products of PS1, PS2, and PS3. Here, 10 ng of the NA18562 human genomic DNA (Coriell) was used as input, and the median primer concentration was 45 nM in the PCR reaction. 17 cycles of PCR were performed using Vent (exo-) DNA polymerase (selected for its improved amplification ability for G/C-rich sequences). To facilitate more in-depth analysis by high-throughput sequencing (NGS), adapters/indexes were ligated to the amplicon products. No size selection was performed, in order to accurately reflect the fraction of primer dimer species following multiplex PCR. The on-target amplicons are expected to have an average length of roughly 250 nt. </center>  
+
+The amounts of these three species in the three primer set libraries are shown in Fig. 3a. Going from the PS1 to the PS3 library, the fraction of primer dimers dropped significantly, from \(90.7\%\) in the PS1 library to \(39.6\%\) in the PS2 library and then to \(4.9\%\) in the PS3 library. However, even with the decrease of dimers from the PS2 library to the PS3 library, the proportion of non- specific amplicons in these two libraries remained about the same. This is reasonable because the SADDLE Loss function was designed only minimizes primer Dimers, and does not consider likelihood of Non- specific amplicon formation. The distribution of amplicon length in NGS reads is consistent with the capillary electrophoresis results in 3 libraries (Fig. 3b, Supplementary section S3).  
+
+We next tested the PS3 primer set on 5 formalin- fixed, paraffin- embedded (FFPE) clinical tissue samples (1 breast cancer, 2 lung cancer, and 2 colorectal cancer samples, see also Supplementary Section S5). The beeswarm plot of the observed reads (Fig. 3d) show high consistency across the different samples, and are also consistent with our results from sheared genomic DNA. The identities and quantities of primer dimers formed, likewise, are similar between FFPE DNA samples and genomic DNA (Fig. 3e).  
+
+To demonstrate the scalability of SADDLE, we next designed and tested a 384 amplicon panel comprising 768  
+
+![](images/Figure_3.jpg)
+  
+
+primers. Due to the high cost of primer synthesis for this large panel, we only experimentally tested the final primer set design. Surprisingly, the observed Dimer fraction was only \(1\%\) for this library, using an input of 40 ng sheared NA18562 genomic DNA (Fig. 3f). Roughly \(56\%\) of the reads were Non- specific amplicons, resulting in a NGS library on- target rate of \(43\%\) (Supplementary section S6).  
+
+Accuracy of the Dimerization Prediction. We constructed the SADDLE Badness function based on our understanding of the mechanisms of primer dimer formation, but we know that this Badness function is imperfect both because our understanding of primer dimer formation is imperfect and because it is computationally too expensive to implement many classes of potentially more accurate Badness functions. Accurate assessment of how good or bad the current Badness function is at predicting Dimers, however, is critical to further incremental improvement in multiplex PCR primer design using SADDLE.  
+
+Through the course of SADDLE optimization, we expect that the Dimer prediction accuracy will get worse in later optimization generations, because we are selecting for primer sets with low expected Badness that will include false negatives. Experiments and analysis of PS1, PS2, and PS3 confirm this
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>FIG. 3: Experimental NGS results for SADDLE-designed primer sets. (a) Distribution of reads observed in NGS library constructed using PS1, PS2, and PS3. On Target reads are defined as those that aligned to the intended amplicons; Dimer reads are defined as those whose insert lengths are smaller than the sum of the two primer lengths; all other reads were classified as Non-specific. The vast majority of Non-specific reads align to other regions of the human genome, via a non-cognate pair of forward and reverse primers. The fraction of NGS reads mapped to Dimers dramatically decreases from PS1 to PS2 to PS3. (b) Distribution of NGS reads in the 3 primer set libraries. (c) Distribution of observed primer dimers, based on aligned reads. Because forward primers (FP) can also form primer dimers with other forward primers, we aligned the first and last 25 nucleotides of each NGS read to the merged set of FP and rPs, with primers 1 through 96 in the diagram showing fPs and primers 97 through 192 showing rPs. For clarity of visualization, the log number of reads of observed primer dimers are displayed via both coloration and circle size. (d) Performance of the PS3 primer set of formalin-fixed paraffin-embedded (FFPE) tissue samples from deidentified lung cancer patients. Because the NGS libraries for these 5 samples differed slightly in total reads, here we plotted the distribution of reads normalized to 1 million reads. (e) The observed primer dimer species and their corresponding NGS reads were relatively similar between cell line genomic DNA and FFPE samples. (f) Demonstration of a 384-plex primer set designed by SADDLE (768 primers). The main diagonal shows On Target reads. Only about \(1\%\) of all reads were primer dimers (Supplementary Section S6). </center>  
+
+understanding (Supplementary Section S7). The Dimer reads for each pair of primers from PS1 are plotted against the predicted Badness in Fig. 4a.  
+
+To facilitate discussions of Badness function accuracy in terms of sensitivity and specificity, we set two separate thresholds: the Reads Threshold (horizontal orange line) and the Badness Threshold (vertical dotted purple line). The plotted Reads Threshold in Fig. 4a corresponds to the mean on- target read depth, and the Badness Threshold plotted correspond to the value that maximizes prediction sensitivity plus specificity. For these Threshold values, we observe a sensitivity \(92.5\%\) ( \(\frac{62}{67}\) ) and a specificity of \(90.3\%\) ( \(\frac{33226}{36797}\) ). By adjusting the Badness Threshold value, we can change the tradeoff between sensitivity and specificity, resulting in a Receiver Operator Characteristic (ROC) curve (Fig. 4b). The area under the ROC curve (AUROC) is 0.9577, indicating very high Dimer prediction accuracy by the Badness function. When the Read Threshold is adjusted higher, the AUROC also increases (Fig. 4c), but the positive predictive value (PPV)  
+
+decreases.  
+
+We next examined the top 5 most dominant Dimer reads in the library (Fig. 4d) and compared them to the top 5 predicted dimer reads based on the Badness function (Fig. 4e). It is noteworthy that only 1 of the two different top 5 lists overlap. The other 4 predicted dimers did not contribute significantly experimentally, and the other 4 observed dimers were not predicted to have high risk for dimer formation. At a glance, it appears we over- weighted the possibility of forming primer dimers in which the \(3'\) - most nucleotide in unpaired, and we may need to adjust the Badness function to allow a stronger attenuation of Badness based on distance from the \(3'\) end. Additionally, it appears that the Badness function may be not scaled optimally, as the log10(Badness) ranges between 0 and 3.5, whereas the log10(Dimers) ranges between 0 and 5 (Fig. 4f). This may mean that the current algorithm over- weights weak potential dimers, at the expense of insufficiently avoiding strong predicted primer dimers.  
+
+Beyond the above observations, it is not clear why some
+
+<--- Page Split --->
+![](images/Figure_5.jpg)
+
+<center>FIG. 4: Evaluation of prediction accuracy of the Badness function for individual primer dimer candidates. (a) Comparison of observed vs. predicted primer dimers for all possible pairs of primers in the PS1 library. The horizontal orange line shows the mean on-target reads for the 96 intended amplicons. By changing the Badness Threshold, different tradeoffs of dimer prediction sensitivity and specificity can be achieved. For the current Reads Threshold and Badness Threshold, we observe \(92.5\%\) sensitivity and \(90.3\%\) specificity. (b) Receiver Operator Characteristic (ROC) curve for dimer prediction sensitivity and specificity achieved by changing the Badness Threshold. (c) The Area Under the ROC (AUROC) value depends on the Reads Threshold, with a highest achievable AUROC of about 0.98. (d) The top 5 dimer species experimentally observed to form with highest number of aligned NGS reads. (e) The top 5 potential dimer species predicted to form based on the Badness function. Note that only the \(\# 4\) species are present are both list in panels (d) and (e). (f) Distribution of observed NGS reads and predicted Badness for all possible primer dimers. </center>  
+
+dimers are observed at much higher reads experimentally than others. For example, the top observed dimer only has a 5 nt overlap at the \(3'\) end, compared to a 7 nt overlap at the \(3'\) end for the rank 4 dimer. This is not consistent with our understanding of DNA hybridization and polymerase extension kinetics, and implies that we may not be able to generate a perfect Badness function even ignoring computational resource constraints.  
+
+Gene Fusion Detection with qPCR and Sanger Sequencing. Gene fusions are therapeutic targets and attractive diagnostic biomarkers to guide treatment [22- 25]. Currently, gene fusions are detected either in single- plex by qPCR for known high- frequency fusions (e.g. BCR- ABL1),  
+
+or by NGS. A highly multiplexed qPCR assay that can detect tens of potential gene fusions relevant to a particular disease could greatly increase the accessibility of gene fusion testing.  
+
+Here, we used SADDLE to design a set of 60 primers to detect 56 actionable gene fusions for non- small cell lung cancer (NSCLC) across 6 genes (ALK, ROS1, RET, NRTK1, NTRK2, and NRTK3). The number of primers are lower than 56:2 because the same exon can be fused with multiple partner genes or exons. We detect the fusions in complementary DNA (cDNA) reverse transcribed from RNA, in order to limit the complexity and length of the detection targets. For each fusion of interest, the primer set includes a forward primer (fP) targets the upstream partner gene and a reverse primer
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>FIG. 5: Highly multiplexed qPCR detection of gene fusions using SADDLE-designed primer sets. (a) Complementary DNA (cDNA) prepared through reverse transcription of RNA can have known target sequences at the exon breakpoints. Although it is trivial to design a single-plex qPCR assay to detect a single known fusion, such as BCR-ABL1 [20], we are not aware of any reports of highly multiplexed qPCR assays to simultaneously detect \(\geq 10\) different gene fusion cDNA species. For this assay, we designed a 60 primer set (46 forward, 14 reverse) that together can amplify 56 distinct gene fusion types commonly observed in non-small-cell lung cancer [21]. (b) Summary of observed qPCR cycle threshold (Ct) values for the 56 reactions, each with 1 of the 56 synthetic fusion DNA species across 6 genes (ALK, ROS1, RET, and NTRK1/NTRK2/NTRK3), each with 1700 copies. WT indicates wildtype commercial cDNA, and NTC indicates no template control. See Supplementary Section S8 for additional details and experimental results, including Sanger sequencing traces of each reaction product. (c) Example qPCR trace showing detection of the fusion DNA sequence joining NACC2 exon 4 to NTRK2 exon 13. (d) Clinical sample results on cDNA reverse transcribed from RNA from extracellular vesicles. Samples 3, 4, and 7 tested positive for a gene fusion, and sequence alignment of the Sanger sequencing results (right panels) show the exact identifies of the fusions. </center>  
+
+(rP) targets the downstream partner gene (Fig. 5a).  
+
+We first tested the multiplex PCR panel against synthetic samples bearing the gene fusions of interest (Fig. 5bc). In all cases, the positive samples were clearly distinguishable by cycle threshold (Ct) value against both commercial wildtype cDNA (WT) and the no- template control (NTC), with all \(\Delta \mathrm{Ct}\) values above 10. We also tested the panel on synthetic gene fusion samples with a variant allele frequency (VAF) of \(1\%\) (Supplementary Section S8). The \(1\%\) VAF samples were constructed by mixing synthetic gBlocks that contained a single fusion (the variant) with human cDNA (the wildtype).  
+
+Finally, we applied the gene fusion qPCR panel to clinical cDNA samples extracted from extracellular vesicles in blood plasma from NSCLC patients (Fig. 5d). Of the 10 clinical samples analyzed, 3 were called positive for gene fusions. To identify the exact gene fusion in these samples, we performed Sanger sequencing on the amplicons from the positive samples. Two samples were identified with EML4 exon20- ALK exon20, and one was identified with EML4 exon 15- ALK exon 20.  
+
+## Discussion  
+
+In this study, we designed a multiplex PCR primer design algorithm SADDLE targeting numerous genomic regions in a single tube. We presented experimental validation of primer sets on a 96- plex cancer- related exons panel, demonstrating that the SADDLE was capable in selecting better primers by reducing dimerization in a multiplex PCR reac  
+
+tion. The dimer rate decreased going from the \(90.7\%\) in a naively designed PS1 to \(39.6\%\) in an intermediate PS2 and to \(4.9\%\) in an optimized PS3, resulting in an increased on- target rate as well as greater uniformity of on- target amplicons. In another 384- plex panel targeting random- selected SNPs in the human genome, the NGS library using the optimized primer set showed a dimer rate of \(1\%\) . SADDLE can reduce reagent costs and enable the amplification of hundreds of target templates simultaneously without wasting NGS reads. Importantly, library preparation using optimized primer sets generated by SADDLE does not depend on labor- intensive enzymatic cleavage or size selection steps to remove dimers.  
+
+The improvement of NGS library on- target rates through the reduction of primer dimers can allow significantly larger targeted panels to be possible using multiplex PCR library preparation. Because multiplex PCR generally requires less input DNA and are faster than ligation- based library preparation approaches, due to the low yields of end repair and ligation, we envision that SADDLE- designed primers can be useful for a variety of research and clinical applications where DNA sample quantities are limited and/or where rapid turnaround is needed. For example, in oncology tissue biopsies obtained through fine needle aspirates and core biopsies are frequently insufficient for standard NGS analysis, and cell- free DNA from peripheral blood plasma likewise are limited and impose sensitivity limitations to ligation- based approaches. Furthermore, in reproductive medicine, samples
+
+<--- Page Split --->
+
+from amniocentesis and preimplantation genetic screening (PGS) and preimplantation genetic diagnosis (PGD) are also very limited, and require rapid turnaround for molecular diagnostics due to the time- sensitivity of clinical decisions.  
+
+Through our analyses of predicted vs. observed dimers, we found that the parameters in the Loss function used in SADDLE could be adjusted to optimize dimer prediction performance, particularly in the \(3^{\prime}\) distance attenuation. However, with the current SADDLE algorithm, Non- specific amplicons now appear to dominate off- target rates, rather than Dimers. Thus, to further scale- up the panels that can be designed by SADDLE, it will be necessary to construct and optimize new Loss functions that penalize primer sets based on predicted off- target genomic amplification. Modification of the Loss function to minimize Non- specific amplicon formation would require significantly more work, as it requires consideration of the expected sample genome sequence. Whereas the current Loss function is "universal" in improving multiplex PCR primer set designs, a Loss function that considers Non- specific amplicons would inherently be suboptimal for primer dimer minimization. A Loss function predicting Non- specific amplification must also consider external factors, including the average length of the DNA molecules in the sample and nonpathogenic genomic polymorphisms.  
+
+In medical and research applications where the cost of NGS cannot be economically justified, qPCR assays will likely be the dominant tool for study of genomic variants. In qPCR, even single- plex primers can form significant dimers if poorly designed with Ct values below 30. Multiplex qPCR thus typically requires significant empirical optimization, even at around 4- plex [26]. SADDLE allowed us to successfully design a 60- primer qPCR panel targeting 56 gene fusions, and exact fusion identities can be determined through affordable Sanger sequencing. Thus, we envision that SADDLE can revolutionize the use of qPCR for highly multiplex molecular diagnostics.  
+
+Acknowledgements. This work was funded by NIH grant R01CA203964 to DYZ.  
+
+Author contributions. NGX and DYZ conceived the project. NGX performed the NGS experiments. NGX, MXW, and PS analyzed the NGS data. YW, YY, and JL designed and performed the qPCR- and Sanger- based gene fusion experiments. NGX and YW analyzed the qPCR and Sanger data. NGX and DYZ wrote the manuscript with input from all authors.  
+
+Additional information. Correspondence may be addressed to DYZ (dyz1@rice.edu). There is a patent pending on the Multiplex Primer Design Algorithm presented in this manuscript, and this patent has been exclusively licensed to Nuprobe Global. NGX, MXW, and PS declare a competing interest in the form of consulting for Nuprobe USA. DYZ declares a competing interest in the form of consulting for and significant equity ownership in Nuprobe Global, Torus Biosystems, and Pana Bio.  
+
+Data Availability. The reference and sample- specific  
+
+gDNA sequence data are available from the NCBI Nucleotide database, the Ensembl database, the COSMIC database, and the Foundation Medicine gene list. All other data supporting the findings of this study are available within the paper and its Supplementary Information files.  
+
+Code Availability. The MATLAB code used for multiplex PCR primer algorithm and the MATLAB code and Python code for NGS data processing are available at https://github.com/NinaGXie/SADDLE.  
+
+[1] Razavi, P. et al. High- intensity sequencing reveals the sources of plasma circulating cell- free DNA variants. Nat Med 25, 1928- 1937 (2019). [2] Cohen, J. D. et al. Detection and localization of surgically resectable cancers with a multi- analyte blood test. Science 359, 926- 930 (2018). [3] Claesson, M. J., Clooney, A. G., & O'toole, P. W. A clinician's guide to microbiome analysis. Nat Rev Gastro Hepat, 14(10), 585 (2017). [4] Mamanova, L. et al. Target- enrichment strategies for next- generation sequencing. Nat Methods 7, 111- 118 (2010). [5] Bailey, J. A., Gu, Z., Clark, R. A., Reinert, K., Samonte, R. V., Schwartz, S., ... & Eichler, E. E. Recent segmental duplications in the human genome. Science, 297(5583), 1003- 1007 (2002). [6] Goodwin, S., McPherson, J. D. & McCombie, W. R. Coming of age: ten years of next- generation sequencing technologies. Nat Rev Genet 17 333- 351 (2016). [7] Elnifro, E. M., Ashshi, A. M., Cooper, R. J., & Klapper, P. E. Multiplex PCR: optimization and application in diagnostic virology. Clin Microbiol Rev, 13(4), 559- 570 (2000). [8] Sun, J. M. et al. Small- cell lung cancer detection in never- smokers: clinical characteristics and multigene mutation profiling using targeted next- generation sequencing. Ann Oncol 26, 161- 166 (2015). [9] Leamon, J., Andersen, M., & Thornton, M. U.S. Patent No. 9,957,558. Washington, DC: U.S. Patent and Trademark Office (2018). [10] Khodakov, D., Wang, C., & Zhang, D. Y. Diagnostics based on nucleic acid sequence variant profiling: PCR, hybridization, and NGS approaches. Adv Drug Deliver Rev, 105, 3- 19 (2016). [11] SantaLucia J Jr, Hicks D. The thermodynamics of DNA structural motifs. Annu Rev Biophys Biomol Struct, 33:415- 440. doi:10.1146/annurev.biophys.32.110601.141800 (2004). [12] Bae, J. H., Fang, J. Z., & Zhang, D. Y. High- throughput methods for measuring DNA thermodynamics. Nucleic Acids Res, 48(15), e89- e89 (2020). [13] Maurer, W. D., & Lewis, T. G. Hash table methods. ACM Comput Surv, 7(1), 5- 19 (1975). [14] Kirkpatrick, S., Gelatt, C. D., & Vecchi, M. P. Optimization by simulated annealing. Science, 220(4598), 671- 680 (1983). [15] Foundation One® Current Gene List. https://www.foundationmedicineasia.com/content/dam/rfm/apac_v2- en/FOne_Current_Gene_List.pdf (2014) [16] Abyzov, A., Urban, A. E., Snyder, M. & Gerstein, M. CNVnator: an approach to discover, genotype, and characterize typical and atypical CNVs from family and population genome sequencing. Genome Res 21, 974- 984 (2011). [17] Corcoran, R. B., & Chabner, B. A. Application of cell- free DNA analysis to cancer treatment. New Engl J Med, 379(18), 1754- 1765 (2018). [18] Ye, J., Coulouris, G., Zaretskaya, I., Cutcutache, I., Rozen, S., & Madden, T. L. Primer- BLAST: a tool to design target- specific primers for polymerase chain reaction. BMC bioinformatics 13, 134 (2012). [19] Langmead, B. & Salzberg, S. L. Fast gapped- read alignment with Bowtie 2. Nat Methods 9 357- 359 (2012). [20] White, H., Deprez, L., Corbisier, P., Hall, V., Lin, F., Mazoua, S., ... & Emons, H. A certified plasmid reference material for the standardisation of BCR- ABL1 mRNA quantification by real- time quantitative PCR. Leukemia, 29(2), 369- 376 (2015). [21] Tang, Z., Zhang, J., Lu, X., Wang, W., Chen, H., Robinson, M. K., ... & Medeiros, L. J. Coexistent genetic alterations involving ALK, RET, ROS1 or MET in 15 cases of lung adenocarcinoma. Modern Pathol 31(2), 307- 312 (2018). [22] Mertens, F., Johansson, B., Fioretos, T. & Mitelman, F. The emerging complexity of gene fusions in cancer. Nat Rev Cancer 15, 371- 381 (2015). [23] Powers M. P. The ever- changing world of gene fusions in cancer: a secondary gene fusion and progression. Oncogene 38(47), 7197- 7199 (2019). [24] Latysheva, N. S. & Babu, M. M. Discovering and understanding oncogenic gene fusions through data intensive computational approaches. Nucleic Acids Res 44 4487- 4503 (2016).
+
+<--- Page Split --->
+
+[25] Heyer, E. E., Deveson, I. W., Wooi, D., Selinger, C. I., Lyons, R. J., Hayes, V. M., ... & Blackburn, J. Diagnosis of fusion genes using targeted RNA sequencing. Nat Commun, 10(1), 1- 12 (2019).[26] Shen, Z., Qu, W., Wang, W., Lu, Y., Wu, Y., Li, Z., ... & Zhang, C. MPprimer: a program for reliable multiplex PCR primer design. BMC bioinformatics, 11(1), 1- 7 (2010).
+
+<--- Page Split --->
+
+## Methods Results  
+
+Oligonucleotides and repository samples. All the primers and synthetic DNA templates were purchased from Integrated DNA Technologies. Primers were purchased as standard desalted DNA oligonucleotides, and synthetic templates as desalted double strand fragments (gBlocks). DNA oligonucleotides solutions were stored at \(4^{\circ}\mathrm{C}\) . Human cell- line gDNA sample NA18562 (Correll Biorepository) was stored at \(- 20^{\circ}\mathrm{C}\) . The gDNA was mixed with synthetic DNA templates at various ratios to create samples containing different proportions of a specific variant sequence. Dilution of gDNA samples and synthetic DNA templates were made in 1x TE buffer with \(0.1\%\) Tween 20 (Sigma Aldrich).  
+
+Multiplex PCR protocol. Multiplex PCR was performed on a T100 Thermocycler or a C1000 Thermocycler (Bio- Rad). The total volume of each reaction was 50ul. DNA sample input ranged from 10 ng to 100 ng per tube. PCR reagents including vent (exo- ) polymerase, ThermoPol Reaction Buffer (10X), and dNTP (New England Biolabs) were used for enzymatic amplification. Thermal cycling started with a 3 min incubation step at \(95^{\circ}\mathrm{C}\) for polymerase activation, followed by 17 cycles of 30 s at \(95^{\circ}\mathrm{C}\) for DNA denaturing, 3 min at \(60^{\circ}\mathrm{C}\) for annealing, and 30 s at \(72^{\circ}\mathrm{C}\) for extension, followed by a final extension of 5 min at \(72^{\circ}\mathrm{C}\) .  
+
+End repair protocol. Multiplex PCR product was end- repaired using NEBNext \((\widehat{\mathbb{R}})\) UltraTM II End Repair/dA- Tailing Module (New England Biolabs). Each reaction was a mixture of \(3\mu \mathrm{l}\) NEBNext Ultra II End Prep Enzyme Mix, \(7\mu \mathrm{l}\) NEBNext Ultra II End Prep Reaction Buffer, \(20\mu \mathrm{l}\) multiplex PCR products, and \(30\mu \mathrm{l}\) H2O. End repair was performed on a Eppendorf Mastercycler. Thermal cycling started with the incubation at \(20^{\circ}\mathrm{C}\) for 30 min and \(65^{\circ}\mathrm{C}\) for 30 min, with the heated lid set to \(80^{\circ}\mathrm{C}\) .  
+
+Adapter ligation. End repair mixture was ligated with adapters using NEBNext \((\widehat{\mathbb{R}})\) UltraTM II Ligation Module (New England Biolabs). Each reaction was a mixture of \(30\mu \mathrm{l}\) NEBNext Ultra II Ligation Master Mix, \(1\mu \mathrm{l}\) NEBNext Ligation Enhancer, \(2.5\mu \mathrm{l}\) NEBNext Adaptor for Illumina, and \(60\mu \mathrm{l}\) previous End repair mixture. Ligation was performed on a Mastercycler from Eppendorf. Thermocycling started with the incubation at \(20^{\circ}\mathrm{C}\) for 15 min with the heated lid off; after adding \(3\mu \mathrm{l}\) USERTM enzyme to the ligation mixture, the reaction was incubated at \(37^{\circ}\mathrm{C}\) for 15 min with the heated lid set to \(55^{\circ}\mathrm{C}\) .  
+
+Index quantitative PCR. Following adapter ligation, Index qPCR was performed on CFX96 Touch Deep Well Real- Time PCR Detection system (Bio- Rad). Quantification of  
+
+Index quantitative PCR. Following adapter ligation, Index qPCR was performed on CFX96 Touch Deep Well Real- Time PCR Detection system (Bio- Rad). Quantification of different libraries was performed simultaneously in each well. Each reaction was a \(10\mu \mathrm{l}\) mixture, with \(1\mu \mathrm{l}\) i5 index, \(1\mu \mathrm{l}\) i7 index, \(1\mu \mathrm{l}\) ligation products, \(2\mu \mathrm{l}\) Milli- Q, and \(5\mu \mathrm{l}\) PowerUp SYBR Green Master Mix. Experiment was performed following a thermal cycling protocol with a 3 min incubation step at \(95^{\circ}\mathrm{C}\) for polymerase activation, followed by 40 cycles of 10 s at \(95^{\circ}\mathrm{C}\) for DNA denaturing and 30 s at \(60^{\circ}\mathrm{C}\) for annealing and extension. Ct values were obtained directly from the CFX96 system.  
+
+Index PCR. Index PCR was performed on a T100 Thermocycler or a C1000 Thermocycler (Bio- Rad). Index primers used were NEBNext \((\widehat{\mathbb{R}})\) Multiplex Oligos for Illumina \((\widehat{\mathbb{R}})\) (New England Biolabs). Each reaction was a mixture of \(2\mu \mathrm{l}\) each i5 and i7 index primers, \(5\mu \mathrm{l}\) ligation products, and PCR reagents including vent (exo- ) polymerase, ThermoPol Reaction Buffer (10X), and dNTP. The volume of each reaction was \(52\mu \mathrm{l}\) . Thermal cycling started with a 3 min incubation step at \(95^{\circ}\mathrm{C}\) for polymerase activation, followed by various cycles of 30 s at \(95^{\circ}\mathrm{C}\) for DNA denaturing and 30 s at \(60^{\circ}\mathrm{C}\) for annealing, and 30 s at \(72^{\circ}\mathrm{C}\) for extension, followed by a final extension of 5 min at \(72^{\circ}\) .  
+
+Column Purification. Multiplex PCR products and ligation products were all purified using DNA Clean & Concentrator Kits (ZYMO Research). The volume of DNA- binding buffer was \(250\mu \mathrm{l}\) for multiplex PCR products clean- up, and \(482.5\mu \mathrm{l}\) for ligation products clean- up; \(25\mu \mathrm{l}\) Milli- Q water was used as elution buffer for each reaction.  
+
+Beads Purification. Index PCR product was purified using AMPure XP beads (Beckman Coulter). For each \(50\mu \mathrm{l}\) reaction mixture, \(90\mu \mathrm{l}\) of beads was added; \(40\mu \mathrm{l}\) Milli- Q water was used as elution buffer.  
+
+Library quantitation. All the libraries were quantified using the QubitTM dsDNA HS Assay Kit (ThermoFisher Scientific).  
+
+Bioanalyzer. Sizes of PCR products and libraries were measured using Bioanalyzer High Sensitivity DNA Assay (Agilent), and DNA chips were run on the Agilent 2100 Bioanalyzer system.  
+
+Next- Generation Sequencing. All the libraries were loaded on a Miseq Reagent V2 for obtaining pair- end reads and were sequenced on a Miseq (Illumina).  
+
+Sanger sequencing. PCR products were purified and prepared using a BigDye Terminator v1.1 Cycle Sequencing Kit (Thermo Fisher Scientific) and were sequenced on a Thermo Fisher Scientific 3500 Series Genetic Analyzer.
+
+<--- Page Split --->
+
+## Figures  
+
+![](images/Figure_2.jpg)
+
+<center>Figure 1 </center>  
+
+Overview of Simulated Annealing Design using Dimer Likelihood Estimation (SADDLE). Given a set of DNA target sequences Ti, with i between 1 and N, the goal is to design a total of 2N PCR primers that can effectively amplify all DNA targets, while generating an acceptably low amount of primer dimer species. Steps 4 and 5 can be repeated a large number of times in order to improve (decrease) the Loss function value on the final primer set S. Multiple implementations, hyper- parameters, and parameters can be selected for each SADDLE step that can impact performance and speed.
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Figure 2 </center>  
+
+Implementation and experimental evaluation of a multiplex primer design algorithm based on the SADDLE framework. (a) Method for generating candidate primer sequences for a DNA target T. (b) Implementation of Badness function that can be rapidly evaluated using hash tables. (c) List of cancer genes selected as target sequences for a 96- plex primer set design. See Supplementary Excel spreadsheet for target selection details. (d) Loss function of primer sets S(g) across optimization generations g. The Loss function value decreases through the optimization and approaches a local minima after roughly 400 generations. We selected three different primer sets, constructed at generations 0, 200, and 400 for experimental evaluation; these are respectively called PS1, PS2, and PS3 for the remainder of this paper. (e) Capillary electrophoresis (Agilent Bioanalyzer 2100) analysis of amplicon products of PS1, PS2, and PS3. Here, 10 ng of the NA18562 human genomic DNA (Coriell) was used as input, and the median primer concentration was 45 nM in the PCR reaction. 17 cycles of PCR were performed using Vent (exo- ) DNA polymerase (selected for its improved amplification ability for G/C- rich sequences). To facilitate more in- depth analysis by high- throughput sequencing (NGS), adapters/indexes were ligated to the amplicon products. No size selection was performed, in order to accurately reflect the fraction of primer dimer species following multiplex PCR. The on- target amplicons are expected to have an average length of roughly 250 nt.
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Figure 3 </center>  
+
+Experimental NGS results for SADDLE- designed primer sets. (a) Distribution of reads observed in NGS library constructed using PS1, PS2, and PS3. On Target reads are defined as those that aligned to the intended amplicons; Dimer reads are defined as those whose insert lengths are smaller than the sum of the two primer lengths; all other reads were classified as Non- specific. The vast majority of Non- specific reads align to other regions of the human genome, via a non- cognate pair of forward and reverse primers. The fraction of NGS reads mapped to Dimers dramatically decreases from PS1 to PS2 to PS3. (b) Distribution of NGS reads in the 3 primer set libraries. (c) Distribution of observed primer dimers, based on aligned reads. Because forward primers (fP) can also form primer dimers with other forward primers, we aligned the first and last 25 nucleotides of each NGS read to the merged set of fP and rPs, with primers 1 through 96 in the diagram showing fPs and primers 97 through 192 showing rPs. For clarity of visualization, the log number of reads of observed primer dimers are displayed via both coloration and circle size. (d) Performance of the PS3 primer set of formalin- fixed paraffin- embedded (FFPE) tissue samples from deidentified lung cancer patients. Because the NGS libraries for these 5 samples differed slightly in total reads, here we plotted the distribution of reads normalized to 1 million reads. (e) The observed primer dimer species and their corresponding NGS reads were relatively similar between cell line genomic DNA and FFPE samples. (f) Demonstration of a 384-plex primer set designed by SADDLE (768 primers). The main diagonal shows On Target reads. Only about 1% of all reads were primer dimers (Supplementary Section S6).
+
+<--- Page Split --->
+![](images/Figure_5.jpg)
+
+<center>Figure 4 </center>  
+
+Evaluation of prediction accuracy of the Badness function for individual primer dimer candidates. (a) Comparison of observed vs. predicted primer dimers for all possible pairs of primers in the PS1 library. The horizontal orange line shows the mean on- target reads for the 96 intended amplicons. By changing the Badness Threshold, different tradeoffs of dimer prediction sensitivity and specificity can be achieved. For the current Reads Threshold and Badness Threshold, we observe \(92.5\%\) sensitivity and \(90.3\%\) specificity. (b) Receiver Operator Characteristic (ROC) curve for dimer prediction sensitivity and specificity achieved by changing the Badness Threshold. (c) The Area Under the ROC (AUROC) value depends on the Reads Threshold, with a highest achievable AUROC of about 0.98. (d) The top 5 dimer species experimentally observed to form with highest number of aligned NGS reads. (e) The top 5 potential dimer species predicted to form based on the Badness function. Note that only the #4 species are present are both list in panels (d) and (e). (f) Distribution of observed NGS reads and predicted Badness for all possible primer dimers.
+
+<--- Page Split --->
+![PLACEHOLDER_16_0]
+
+<center>Figure 5 </center>  
+
+Highly multiplexed qPCR detection of gene fusions using SADDLE- designed primer sets. (a) Complementary DNA (cDNA) prepared through reverse transcription of RNA can have known target sequences at the exon breakpoints. Although it is trivial to design a single- plex qPCR assay to detect a single known fusion, such as BCR- ABL1 [20], we are not aware of any reports of highly multiplexed qPCR assays to simultaneously detect \(\boxed{10}\) different gene fusion cDNA species. For this assay, we designed a 60 primer set (46 forward, 14 reverse) that together can amplify 56 distinct gene fusion types commonly observed in non- small- cell lung cancer [21]. (b) Summary of observed qPCR cycle threshold (Ct) values for the 56 reactions, each with 1 of the 56 synthetic fusion DNA species across 6 genes (ALK, ROS1, RET, and NTRK1/NTRK2/NTRK3), each with 1700 copies. WT indicates wildtype commercial cDNA, and NTC indicates no template control. See Supplementary Section S8 for additional details and experimental results, including Sanger sequencing traces of each reaction product. (c) Example qPCR trace showing detection of the fusion DNA sequence joining NACC2 exon 4 to NTRK2 exon 13. (d) Clinical sample results on cDNA reverse transcribed from RNA from extracellular vesicles. Samples 3, 4, and 7 tested positive for a gene fusion, and sequence alignment of the Sanger sequencing results (right panels) show the exact identifies of the fusions.  
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+- SADDLESequence.xlsx- SADDLESequence.xlsx- SADDLESupv4.pdf- SADDLESupv4.pdf
+
+<--- Page Split --->

preprint/preprint__0f731c80040fa9beeef1b271f00ca5c000eaea0664e712d70f287e8bda9d7779/preprint__0f731c80040fa9beeef1b271f00ca5c000eaea0664e712d70f287e8bda9d7779_det.mmd

@@ -0,0 +1,408 @@
+<|ref|>title<|/ref|><|det|>[[44, 106, 950, 210]]<|/det|>
+# Designing Highly Multiplex PCR Primer Sets with Simulated Annealing Design using Dimer Likelihood Estimation (SADDLE)  
+
+<|ref|>text<|/ref|><|det|>[[44, 229, 344, 250]]<|/det|>
+David Zhang (dyz1@rice.edu)  
+
+<|ref|>text<|/ref|><|det|>[[50, 253, 185, 271]]<|/det|>
+Rice University  
+
+<|ref|>text<|/ref|><|det|>[[44, 278, 185, 316]]<|/det|>
+Nina Xie Rice University  
+
+<|ref|>text<|/ref|><|det|>[[44, 323, 185, 361]]<|/det|>
+Michael Wang Rice University https://orcid.org/0000- 0001- 7009- 6958  
+
+<|ref|>text<|/ref|><|det|>[[44, 367, 185, 404]]<|/det|>
+Ping Song Rice University  
+
+<|ref|>text<|/ref|><|det|>[[44, 412, 185, 450]]<|/det|>
+Yifan Wang Nuprobe China  
+
+<|ref|>text<|/ref|><|det|>[[44, 457, 185, 495]]<|/det|>
+Yuxia Yang Nuprobe China  
+
+<|ref|>text<|/ref|><|det|>[[44, 502, 185, 520]]<|/det|>
+Junfeng Luo  
+
+<|ref|>text<|/ref|><|det|>[[50, 525, 185, 543]]<|/det|>
+Nuprobe China  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 589, 102, 606]]<|/det|>
+## Article  
+
+<|ref|>text<|/ref|><|det|>[[44, 626, 393, 646]]<|/det|>
+Keywords: PCR primers, DNA sequences  
+
+<|ref|>text<|/ref|><|det|>[[44, 665, 295, 685]]<|/det|>
+Posted Date: May 13th, 2021  
+
+<|ref|>text<|/ref|><|det|>[[44, 703, 463, 723]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 491811/v1  
+
+<|ref|>text<|/ref|><|det|>[[44, 740, 910, 784]]<|/det|>
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+<|ref|>text<|/ref|><|det|>[[42, 819, 910, 863]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on April 11th, 2022. See the published version at https://doi.org/10.1038/s41467- 022- 29500- 4.
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[133, 49, 866, 86]]<|/det|>
+# Designing Highly Multiplex PCR Primer Sets with Simulated Annealing Design using Dimer Likelihood Estimation (SADDLE)  
+
+<|ref|>text<|/ref|><|det|>[[78, 98, 918, 115]]<|/det|>
+Nina G. Xie\*, \(^{1}\) Michael X. Wang\*, \(^{1}\) Ping Song, \(^{1}\) Yifan Wang, \(^{2}\) Yuxia Yang, \(^{2}\) Junfeng Luo, \(^{2}\) and David Yu Zhang \(^{1 + 1}\)  
+
+<|ref|>text<|/ref|><|det|>[[280, 117, 716, 158]]<|/det|>
+\(^{1}\) Department of Bioengineering, Rice University, Houston, TX \(^{2}\) NuProbe China, Shanghai, China(Dated: May 3, 2021)  
+
+<|ref|>text<|/ref|><|det|>[[172, 167, 825, 417]]<|/det|>
+The design of highly multiplex PCR primers to amplify and enrich many different DNA sequences is increasing in biomedical importance as new mutations and pathogens are identified. One major challenge in the design of highly multiplex PCR primer sets is the large number of potential primer dimer species that grows quadratically with the number of primers to be designed. Simultaneously, there are exponentially many choices for multiplex primer sequence selection, resulting in systematic evaluation approaches being computationally intractable. Here, we present and experimentally validate Simulated Annealing Design using Dimer Likelihood Estimation (SADDLE), a stochastic algorithm for design of highly multiplex PCR primer sets that minimize primer dimer formation. Our approach uses a rapidly computable Loss function to approximate the degree of primer dimer formation within a primer set, and randomly swaps primers in the set with alternative candidates using a simulated annealing algorithm. In a 96- plex PCR primer set (192 primers), we show that we can reduce the fraction of primer dimers from \(90.7\%\) in a naively designed PCR primer set to \(4.9\%\) in our optimized primer set. Running the optimized 96- plex primer set on FFPE DNA samples from cancer patients, we likewise observe a low fraction of primer dimer reads. Even when scaling to 384- plex (768 primers), the PCR primer set designed by our algorithm maintains low primer dimer fraction. In addition to NGS, we also show that our SADDLE- designed multiplex primer sets can be used in qPCR settings to allow highly multiplexed detection of gene fusions in cDNA, with a single- tube assay comprising 60 primers detecting 56 distinct gene fusions recurrently observed in lung cancer.  
+
+<|ref|>text<|/ref|><|det|>[[57, 441, 485, 760]]<|/det|>
+The advance of high throughput sequencing has uncovered a large number of biomedically relevant DNA sequences, from driver mutations in cancer to new bacterial/viral pathogen DNA sequences to microbiome metagenomic profiles that affect mental disorders on the gut- brain axis [1- 4]. For discovery applications, "shotgun" whole genome sequencing (WGS) is the preferred approach to identify novel DNA sequences of interest [5]. However, the human genome comprises over 3 billion nucleotides, and despite the lowering costs of high- throughput sequencing, it is not practical today to perform WGS to high depths necessary for identification of subclonal mutations, such as somatic mutations in cancer. For routine detection of disease- relevant DNA variants in known genes of interest, targeted sequencing or direct qPCR approaches are typically used [6, 7]. Of the two dominant methods today for target enrichment, multiplex PCR tends to have shorter workflows and require less DNA input than hybrid- capture probes [8]. However, multiplex PCR struggles to scale to large panels covering hundreds of genes, due to the nonlinear increase of primer dimer species that reduce NGS mapping rates and increase effective cost[10].  
+
+<|ref|>text<|/ref|><|det|>[[57, 768, 485, 919], [512, 442, 940, 548]]<|/det|>
+Currently, multiplex PCR methods for NGS target enrichment (e.g. Ampliseq [8]) primarily rely on (1) enzymatic digestion of primer dimer species at mismatch bubbles [9] and (2) DNA size selection to preferentially remove short amplicon species likely to be primer dimers. However, both steps are labor- intensive and does not allow full removal of all primer dimer species. In contrast, relatively little systematic work have been reported on computational approaches to minimizing the formation of primer dimers in the first place. The development of a robust multiplex primer set design algorithm that produces highly multiplexed primer sets with minimal primer dimer formation could allow further scaling of multiplex PCR target enrichment to even larger NGS panels when combined with enzymatic and size selection methods. Alternatively, it can simplify the workflow of moderate size NGS and qPCR assays by removing the need for strict contamination control from open- tube steps.   
+
+<|ref|>text<|/ref|><|det|>[[512, 554, 940, 884]]<|/det|>
+There are two primary challenges in designing highly multiplexed PCR primer sets: First, for an \(N\) - plex PCR primer set comprising \(2N\) primers, there are \(\binom{2N}{2}\) possible simple primer dimer interactions. For \(N = 50\) , this corresponds to \(\binom{100}{2} = 4950\) times as many potential primer dimer bindings as for a single- plex PCR primer set. Second, there are typically \(M > 10\) reasonable candidate choices for each primer when considering specific gene targets and amplicon length constraints, resulting in \(M^{2N}\) possible N- plex primer sets. For \(M = 20\) and \(N = 50\) , the number of possible primer sets is \(20^{100} \approx 1.3 \cdot 10^{130}\) , billions of times larger than the number of atoms in the universe. Thus, it is computationally intractable to evaluate all possible multiplex primer sets. Simultaneously, primer dimer formation emerges from the interactions of two or more primers in the primer set, so changing the sequence of a primer to mitigate one primer dimer interaction may result in the appearance of another more serious primer dimer. In the language of numerical optimization, multiplex primer design is high dimensional problem with a highly non- convex fitness landscape. Consequently, standard convex optimization algorithms (e.g. gradient descent) will not be effective.  
+
+<|ref|>text<|/ref|><|det|>[[512, 889, 940, 918]]<|/det|>
+Here, we present Simulated Annealing Design using Dimer Likelihood Estimation (SADDLE), an algorithmic framework
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[108, 58, 876, 456]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[55, 472, 943, 538]]<|/det|>
+<center>FIG. 1: Overview of Simulated Annealing Design using Dimer Likelihood Estimation (SADDLE). Given a set of DNA target sequences \(\mathrm{T}_i\) , with \(i\) between 1 and \(N\) , the goal is to design a total of \(2N\) PCR primers that can effectively amplify all DNA targets, while generating an acceptably low amount of primer dimer species. Steps 4 and 5 can be repeated a large number of times in order to improve (decrease) the Loss function value on the final primer set S. Multiple implementations, hyper-parameters, and parameters can be selected for each SADDLE step that can impact performance and speed. </center>  
+
+<|ref|>text<|/ref|><|det|>[[57, 552, 485, 718]]<|/det|>
+for designing highly multiplex PCR primer sets. Within this framework, we present an example multiplex primer design algorithm, comprising an algorithm for primer candidate generation and a rapidly computable Loss function for estimating primer dimers. Using the SADDLE, we designed and experimentally tested multiplex primer sets comprising 192 primers (96- plex) and 784 primers (384- plex), and show low primer dimer formation through NGS experiments. Building upon this success, we built a single- tube 60 primer qPCR and Sanger assay to detect and identify 56 gene fusions with clinical actionability for non- small cell lung cancer.  
+
+<|ref|>sub_title<|/ref|><|det|>[[57, 729, 129, 744]]<|/det|>
+## Results  
+
+<|ref|>sub_title<|/ref|><|det|>[[57, 752, 485, 782]]<|/det|>
+## Simulated Annealing Design using Dimer Loss Estimation (SADDLE).  
+
+<|ref|>text<|/ref|><|det|>[[75, 783, 485, 799]]<|/det|>
+There are 6 main steps in SADDLE, as illustrated in Fig. 1:  
+
+<|ref|>text<|/ref|><|det|>[[75, 807, 485, 915]]<|/det|>
+1. Generation of forward primer (fP) and reverse primer (rP) candidates for each gene target. 
+2. Selection of an initial primer set \(\mathrm{S}_0\) from the primer candidates. 
+3. Evaluation of the Loss function \(\mathrm{L}(\cdot)\) on the initial primer set \(\mathrm{S}_0\) .  
+
+<|ref|>text<|/ref|><|det|>[[536, 553, 941, 598]]<|/det|>
+4. Generate a temporary primer set \(\mathrm{T}\) based on set \(\mathrm{S}_g\) (primer set S from generation G) by randomly changing 1 or more primers.  
+
+<|ref|>text<|/ref|><|det|>[[536, 614, 941, 660]]<|/det|>
+5. Evaluate \(\mathrm{L}(\mathrm{T})\) , and set \(\mathrm{S}_{g + 1}\) to either \(\mathrm{S}_g\) (no change) or \(\mathrm{T}\) , depending probabilistically on the relative values of \(\mathrm{L}(\mathrm{S}_g)\) and \(\mathrm{L}(\mathrm{T})\) .  
+
+<|ref|>text<|/ref|><|det|>[[536, 676, 940, 705]]<|/det|>
+6. Repeat steps 4 to 5 until an acceptable primer set \(\mathrm{S}_{\mathrm{final}}\) is constructed.  
+
+<|ref|>text<|/ref|><|det|>[[512, 721, 941, 916]]<|/det|>
+The above abstract framework provides a basis for many potential multiplex primer design algorithms, depending on the specific details of primer candidate generation, form of Loss function, temporary set \(\mathrm{T}\) generation, and the dynamic probability of setting \(\mathrm{S}_{g + 1}\) to \(\mathrm{T}\) . Below, we describe our specific implementation of SADDLE, based on our accumulated understanding of primer design principles and primer dimer formation mechanisms. Given the infinite possibilities for function forms and hyper- parameters, we did not systematically evaluate or optimize at the high- level. Lower- level parameters, such as standard free energy \((\Delta G^{\circ})\) ranges for primers, were experimentally optimized and these are described below.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[57, 51, 486, 278]]<|/det|>
+1. Primer candidate generation. We begin our implementation of primer candidate generation through the selection of one or more "pivot" nucleotides on human genomic DNA around which we design the forward primer (Fig. 2a). The pivot nucleotides are the ones that must be included in the amplicon insert, and for example could be the hotspot region of a gene that is frequently mutated. From the pivot nucleotides and a constraint on the maximum length of the amplicons (e.g. determined by the read length of NGS), we can systematically generate a series of different proto-primers with \(3'\) end just outside the pivot nucleotides. The proto-primers have a large range of different lengths and binding energies to their complementary sequences, and will next be trimmed at the \(3'\) end to generate the primer candidates (Fig. 2a).  
+
+<|ref|>text<|/ref|><|det|>[[57, 288, 486, 619]]<|/det|>
+From our past experiences and preliminary optimization experiments, primers that hybridize to their cognate templates with \(\Delta G^{\circ} \approx - 11.5 \mathrm{kcal / mol}\) have the best tradeoff between amplification efficiency/uniformity and nonspecific hybridization. Shorter primers may not bind consistently with high efficiency to their templates, resulting in variability in amplification efficiency and non-uniformity of amplicon on-target reads in the NGS library. Longer primers have increased likelihood of binding to other loci in human genome, and can result in non-specific amplicons. Based on this \(\Delta G^{\circ}\) goal, we next systematically constructed primer candidates from the proto-primers by truncating nucleotides from the \(3'\) end until the primer candidate has \(\Delta G^{\circ}\) between - 10.5 kcal/mol and - 12.5 kcal/mol. Due to the granularity of \(\Delta G^{\circ}\) for base stacks, some proto-primers with the same \(5'\) end will result in multiple primer candidates (e.g. with \(\Delta G^{\circ} = - 10.9 \mathrm{kcal / mol}\) and - 12.0 kcal/mol). Optionally, one could implement additional filters here to remove undesirable primer candidates, such as based on G/C content. For our demonstration panels, we restrict the G/C content of primer candidates to be between 0.25 and 0.75, removing primer candidates with G/C content outside this range.  
+
+<|ref|>text<|/ref|><|det|>[[57, 628, 485, 747]]<|/det|>
+In the implementation of SADDLE, primer candidates can be treated as individual primers or as primer pairs. Our specific implementation treats primers as pairs, so we next combinatorially generate all candidate primer pairs for an amplicon, in order to better constrain the distribution of amplicon lengths Any candidate primer pairs that generate amplicons with length exceeding our maximum amplicon length or below our minimum are removed.  
+
+<|ref|>text<|/ref|><|det|>[[57, 757, 485, 817]]<|/det|>
+2. Initial primer set \(\mathbf{S}_0\) selection. We randomly selected a primer pair candidate for every amplicon that we wish to design, and collectively the selected primers are known as the initial primer set \(\mathbf{S}_0\) .  
+
+<|ref|>text<|/ref|><|det|>[[57, 827, 485, 917]]<|/det|>
+3. Evaluation of Loss function \(\mathbf{L}(\cdot)\) on \(\mathbf{S}_0\) . The Loss function \(\mathbf{L}(\cdot)\) is a rapidly computable function that aims to approximate the severity of primer dimer formation by a primer set \(\mathbf{S}\) . \(\mathbf{L}(\cdot)\) sums the potential primer dimer interactions between every pair of primers in the primer set. To prevent confusion, we refer to the predicted formation of dimers for a  
+
+<|ref|>text<|/ref|><|det|>[[512, 52, 941, 67]]<|/det|>
+particular pair of primers to be the Badness. Mathematically,  
+
+<|ref|>equation<|/ref|><|det|>[[616, 74, 940, 119]]<|/det|>
+\[\mathrm{L}(\cdot) = \sum_{a = 1}^{2N}\sum_{b = a}^{2N}\mathrm{Badness}(a,b) \quad (1)\]  
+
+<|ref|>text<|/ref|><|det|>[[513, 128, 941, 263]]<|/det|>
+One can imagine the Badness function to be proportional to the amount of primer dimers formed by two primers. In an optimized primer set with a relatively low concentration of primer dimers compared to the concentration of on- target amplicons, the amount of primer dimers formed between primer \(a\) and \(b\) should not significantly impact the amount of dimers formed between \(a\) and \(c\) , so the Loss function being defined as the sum of the component Badness functions is justified.  
+
+<|ref|>text<|/ref|><|det|>[[512, 264, 940, 293]]<|/det|>
+The Badness function in our implementation is defined as follows (Fig. 2b):  
+
+<|ref|>equation<|/ref|><|det|>[[530, 300, 940, 336]]<|/det|>
+\[\mathrm{Badness}(a,b) = \sum \frac{1}{(i + 1)(j + 1)}\cdot 2^{\mathrm{Length}}\cdot 2^{\mathrm{numGC}} \quad (2)\]  
+
+<|ref|>text<|/ref|><|det|>[[513, 344, 941, 450]]<|/det|>
+The sum in the Badness definition is over all subsequences between primer \(a\) and \(b\) with at least 4 nt of continuous complementarity. The lower limit of 4 nt is both based on our preliminary experimental studies in qPCR showing that up to 3 nt of complementarity at the \(3'\) ends of two primers will not result in significant primer dimers even in no- template control (NTC) reactions.  
+
+<|ref|>text<|/ref|><|det|>[[513, 450, 941, 644]]<|/det|>
+For each complementary subsequence, its Length and its number of GC nucleotides in complementary subsequence (numGC) contribute exponentially to Badness. Thus, the exponential components of Badness roughly reflect the partition function of the complementarity interaction, with G/C base pairs roughly twice as strong as A/T base pairs. We chose to use these simplistic parameters, rather literature base stacking thermodynamics parameters [11], because there is significant uncertainty in the effective salinity of the PCR reaction buffer, and because our previous studies on DNA thermodynamics suggests that previously reported \(\Delta H^{\circ}\) and \(\Delta S^{\circ}\) parameters do not extrapolate well to higher temperatures [12].  
+
+<|ref|>text<|/ref|><|det|>[[513, 646, 941, 840]]<|/det|>
+The distances of the complementary subsequence to the \(3'\) ends of primers \(a\) and \(b\) , denoted as \(i\) and \(j\) , are known to significantly affect the likelihood of primer dimer formation. In our preliminary qPCR experiments, we observed that a primer pair with 10 nt of complementarity at the \(5'\) end will not result in observable primer dimer formation, but a primer pair with 5 nt of complementarity at the \(3'\) would. Depending on whether the specific DNA polymerase used, different \(i\) - and \(j\) - based attenuation of Badness may be optimal for minimizing primer dimers. Because high- fidelity DNA polymerases with \(3' > 5'\) exonuclease activity can remove mismatched \(3'\) nucleotides, the optimal \(i\) - and \(j\) - based attenuation should be significantly weaker for high- fidelity DNA polymerases.  
+
+<|ref|>text<|/ref|><|det|>[[513, 841, 941, 917]]<|/det|>
+The evaluation of the Badness function is single largest component of software runtime cost, due to the large number of times the Badness function will be evaluated. For a primer that is 25 nt in length, there are 22 subsequences of length 4, 21 subsequences of length 5, etc. Evaluation of Badness for a
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[56, 50, 486, 216]]<|/det|>
+single primer pair would thus have time complexity of \(\mathrm{O}(P^{3})\) where \(P\) is the length of each primer. However, due to the additive nature of subsequence Badness components to the overall Badness function, we implement more rapid Badness evaluation using a hash table [13]. The time complexity to set up the hash table is \(\mathrm{O}(P^{2})\) for all subsequences of the first primer of a minimum length, and the time complexity to evaluate the Badness by stepping through all subsequences of the second primer is also \(\mathrm{O}(P^{2})\) . Consequently, the overall time complexity of evaluating the Loss function is \(\mathrm{O}(N^{2}\cdot P^{2})\) for all primers in \(\mathrm{S}_{0}\) .  
+
+<|ref|>text<|/ref|><|det|>[[56, 220, 486, 415]]<|/det|>
+4. Generate temporary primer set T based on \(\mathbf{S}_{g}\) . Step 4 begins the recursive optimization process. Based on the current primer set \(\mathbf{S}_{g}\) at generation \(g\) , we first randomly select one primer pair to "mutate." For that primer pair, we randomly select a different primer pair from the list of all candidate primer pairs generated in Step 1. Temporary primer set T is thus generated by combining this new primer pair with all remaining primers in set \(\mathbf{S}_{g}\) . Optionally, multiple primer pairs can be replaced simultaneously in this step to allow faster and more efficient exploration of the primer set space. In our preliminary in silico evaluations, we found that simultaneously mutating multiple primer pairs generally caused a slowdown of the optimization process.  
+
+<|ref|>text<|/ref|><|det|>[[56, 418, 486, 479]]<|/det|>
+5. Evaluate \(\mathbf{L}(\mathbf{T})\) and set \(\mathbf{S}_{g + 1}\) to be either T or \(\mathbf{S}_{g}\) . The Loss of temporary primer set T can be evaluated significantly faster than the initial evaluation of \(\mathrm{L}(\mathrm{S}_{0})\) , because many of the Badness components are the same as in \(\mathbf{S}_{g}\) :  
+
+<|ref|>equation<|/ref|><|det|>[[56, 496, 463, 529]]<|/det|>
+\[\mathrm{L}(\mathrm{T}) = \mathrm{L}(\mathrm{S}_{g}) - \sum_{b}(\mathrm{Badness}(a_{\mathrm{old}},b) - \mathrm{Badness}(a_{\mathrm{new}},b))\]  
+
+<|ref|>text<|/ref|><|det|>[[56, 549, 485, 579]]<|/det|>
+Thus, the time complexity of this step is only \(\mathrm{O}(N\cdot P^{2})\) , rather than \(\mathrm{O}(N^{2}\cdot P^{2})\) .  
+
+<|ref|>text<|/ref|><|det|>[[56, 582, 485, 808]]<|/det|>
+We next compare the value of \(\mathrm{L}(\mathrm{T})\) vs. \(\mathrm{L}(\mathrm{S}_{g})\) . If \(\mathrm{L}(\mathrm{T})\) is smaller than \(\mathrm{L}(\mathrm{S}_{g})\) , then the primer pair change was an improvement and accepted, so \(\mathrm{S}_{g + 1}\) is set to T. If \(\mathrm{L}(\mathrm{T})\) is larger than \(\mathrm{L}(\mathrm{S}_{g})\) , the change was detrimental, but we will still accept the change with a certain probability, as part of the simulated annealing algorithm [14]. To clarify, "simulated annealing" here refers to a specific computer science algorithm, and not a literal simulation of a physical DNA thermal annealing process. If we never accept any detrimental primer pair changes, then the approach degenerates to become a stochastic gradient descent approach. In preliminary in silico evaluations, we confirmed that stochastic gradient descent produces final primer sets with significantly worse Loss, because it becomes too easy to get stuck in a local Loss minima.  
+
+<|ref|>text<|/ref|><|det|>[[56, 812, 485, 917]]<|/det|>
+The probability of accepting a detrimental change depends on both the magnitude of the detriment \((\mathrm{L}(\mathrm{T}) - \mathrm{L}(\mathrm{S}_{g}))\) and the generation \(g\) of the optimization. Worse changes with higher \(\mathrm{L}(\mathrm{T})\) are accepted with lower probability, and later generations of the optimization (higher \(g\) ) are less tolerant of detrimental changes. In our implementations, the probability of setting \(\mathrm{S}_{g + 1}\) to be T when \(\mathrm{L}(\mathrm{T})\) is greater than \(\mathrm{L}(\mathrm{S}_{g})\) are  
+
+<|ref|>text<|/ref|><|det|>[[512, 52, 586, 65]]<|/det|>
+as follows:  
+
+<|ref|>equation<|/ref|><|det|>[[599, 74, 937, 115]]<|/det|>
+\[p = e^{(\mathrm{L}(\mathrm{S}_{g}) - \mathrm{L}(\mathrm{T})) / C(g)}\quad \mathrm{if~}g< g_{t} \quad (3)\]  
+
+<|ref|>text<|/ref|><|det|>[[512, 125, 940, 200]]<|/det|>
+\(C(g)\) is a function that is monotonically non- increasing in \(g\) , indicating decreasing tolerance to detrimental changes at later generations. The parameter \(g_{t}\) indicates the generation in which simulated annealing terminates, and we switch over to stochastic gradient descent.  
+
+<|ref|>text<|/ref|><|det|>[[512, 201, 940, 337]]<|/det|>
+6. Repeat steps 4 and 5. Steps 4 and 5 are repeated until either a pre-determined generation \(g\) , or until \(\mathrm{L}(\mathrm{S}_{g})\) is below a pre-determined threshold \(\mathrm{L}_{t}\) . In our implementation, we typically run the optimization to about \(1.5\cdot g_{t}\) to ensure we reach local minima. To further improve the overall quality of the generated primer set, we recommend running multiple SADDLE optimization processes with different starting conditions (initial primer sets) and selecting the best final primer set.  
+
+<|ref|>text<|/ref|><|det|>[[512, 343, 940, 614]]<|/det|>
+Design and Experimental Evaluation of a 96- plex Primer Set. We first used SADDLE to optimize the design of a 96- plex primer set, each amplicon target one arbitrarily selected exon of a different cancer- related gene [15] (Fig. 2c). Fig. 2d shows the calculated value of \(\mathrm{L}(\mathrm{S}_{g})\) at different generations \(g\) , and is representative of our typical optimization trajectory. We selected the designed primer sets at three different optimization generations for experimental testing: PS1 (initial unoptimized primer set), PS2 (primer set with intermediate Loss optimization), and PS3 (primer set with saturating Loss optimization). The primer set Loss decreased roughly 24- fold from PS1 to PS3; after 40,000 generations, only very marginal improvements were observed. We chose the primer set at 40,000 generations as PS3, rather than the one at 60,000 generations, because we know that our Loss function is an imperfect predictor of primer dimers. Overtraining on an imperfect Loss function can lead to worse experimental results.  
+
+<|ref|>text<|/ref|><|det|>[[512, 616, 940, 917]]<|/det|>
+We applied each of the three primer sets individually to human genomic DNA (10ng NA18562, sheared to a mean length of approximately 150 nt) and amplified for 17 cycles. We next constructed NGS libraries from the amplicons generated using PS1, PS2, and PS3, using a standard adaptor ligation protocol (Supplementary Section S1). After library preparation, capillary electrophoresis results show a clear increase of amplicons of the expected length from PS1 to PS2 to PS3 (Fig. 2e). In the NGS data analysis workflow, after the first step of adapter trimming, we separated NGS reads into three major species: on- target amplicons, dimers, and non- specific amplicons (Supplementary Section S2). On- target amplicons are the NGS reads that were successfully aligned to the intended amplicon sequences using Bowtie2[19]. The remaining NGS reads were aligned separately to each forward and reverse primer sequence. Reads with insert length shorter than the sum of the two aligned primers are classified as Dimers, and reads with insert length longer than the sum of the two aligned primers are classified as Non- specific amplicons (amplifying unintended regions of the human genome).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[58, 50, 490, 360]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[57, 368, 943, 529]]<|/det|>
+<center>FIG. 2: Implementation and experimental evaluation of a multiplex primer design algorithm based on the SADDLE framework. (a) Method for generating candidate primer sequences for a DNA target T. (b) Implementation of Badness function that can be rapidly evaluated using hash tables. (c) List of cancer genes selected as target sequences for a 96-plex primer set design. See Supplementary Excel spreadsheet for target selection details. (d) Loss function of primer sets \(S(g)\) across optimization generations \(g\) . The Loss function value decreases through the optimization and approaches a local minima after roughly 400 generations. We selected three different primer sets, constructed at generations 0, 200, and 400 for experimental evaluation; these are respectively called PS1, PS2, and PS3 for the remainder of this paper. (e) Capillary electrophoresis (Agilent Bioanalyzer 2100) analysis of amplicon products of PS1, PS2, and PS3. Here, 10 ng of the NA18562 human genomic DNA (Coriell) was used as input, and the median primer concentration was 45 nM in the PCR reaction. 17 cycles of PCR were performed using Vent (exo-) DNA polymerase (selected for its improved amplification ability for G/C-rich sequences). To facilitate more in-depth analysis by high-throughput sequencing (NGS), adapters/indexes were ligated to the amplicon products. No size selection was performed, in order to accurately reflect the fraction of primer dimer species following multiplex PCR. The on-target amplicons are expected to have an average length of roughly 250 nt. </center>  
+
+<|ref|>text<|/ref|><|det|>[[57, 550, 485, 745]]<|/det|>
+The amounts of these three species in the three primer set libraries are shown in Fig. 3a. Going from the PS1 to the PS3 library, the fraction of primer dimers dropped significantly, from \(90.7\%\) in the PS1 library to \(39.6\%\) in the PS2 library and then to \(4.9\%\) in the PS3 library. However, even with the decrease of dimers from the PS2 library to the PS3 library, the proportion of non- specific amplicons in these two libraries remained about the same. This is reasonable because the SADDLE Loss function was designed only minimizes primer Dimers, and does not consider likelihood of Non- specific amplicon formation. The distribution of amplicon length in NGS reads is consistent with the capillary electrophoresis results in 3 libraries (Fig. 3b, Supplementary section S3).  
+
+<|ref|>text<|/ref|><|det|>[[57, 748, 485, 884]]<|/det|>
+We next tested the PS3 primer set on 5 formalin- fixed, paraffin- embedded (FFPE) clinical tissue samples (1 breast cancer, 2 lung cancer, and 2 colorectal cancer samples, see also Supplementary Section S5). The beeswarm plot of the observed reads (Fig. 3d) show high consistency across the different samples, and are also consistent with our results from sheared genomic DNA. The identities and quantities of primer dimers formed, likewise, are similar between FFPE DNA samples and genomic DNA (Fig. 3e).  
+
+<|ref|>text<|/ref|><|det|>[[57, 887, 485, 916]]<|/det|>
+To demonstrate the scalability of SADDLE, we next designed and tested a 384 amplicon panel comprising 768  
+
+<|ref|>image<|/ref|><|det|>[[512, 50, 940, 360]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[512, 550, 940, 656]]<|/det|>
+primers. Due to the high cost of primer synthesis for this large panel, we only experimentally tested the final primer set design. Surprisingly, the observed Dimer fraction was only \(1\%\) for this library, using an input of 40 ng sheared NA18562 genomic DNA (Fig. 3f). Roughly \(56\%\) of the reads were Non- specific amplicons, resulting in a NGS library on- target rate of \(43\%\) (Supplementary section S6).  
+
+<|ref|>text<|/ref|><|det|>[[512, 670, 940, 835]]<|/det|>
+Accuracy of the Dimerization Prediction. We constructed the SADDLE Badness function based on our understanding of the mechanisms of primer dimer formation, but we know that this Badness function is imperfect both because our understanding of primer dimer formation is imperfect and because it is computationally too expensive to implement many classes of potentially more accurate Badness functions. Accurate assessment of how good or bad the current Badness function is at predicting Dimers, however, is critical to further incremental improvement in multiplex PCR primer design using SADDLE.  
+
+<|ref|>text<|/ref|><|det|>[[512, 841, 940, 916]]<|/det|>
+Through the course of SADDLE optimization, we expect that the Dimer prediction accuracy will get worse in later optimization generations, because we are selecting for primer sets with low expected Badness that will include false negatives. Experiments and analysis of PS1, PS2, and PS3 confirm this
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[84, 55, 914, 396]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[55, 406, 943, 594]]<|/det|>
+<center>FIG. 3: Experimental NGS results for SADDLE-designed primer sets. (a) Distribution of reads observed in NGS library constructed using PS1, PS2, and PS3. On Target reads are defined as those that aligned to the intended amplicons; Dimer reads are defined as those whose insert lengths are smaller than the sum of the two primer lengths; all other reads were classified as Non-specific. The vast majority of Non-specific reads align to other regions of the human genome, via a non-cognate pair of forward and reverse primers. The fraction of NGS reads mapped to Dimers dramatically decreases from PS1 to PS2 to PS3. (b) Distribution of NGS reads in the 3 primer set libraries. (c) Distribution of observed primer dimers, based on aligned reads. Because forward primers (FP) can also form primer dimers with other forward primers, we aligned the first and last 25 nucleotides of each NGS read to the merged set of FP and rPs, with primers 1 through 96 in the diagram showing fPs and primers 97 through 192 showing rPs. For clarity of visualization, the log number of reads of observed primer dimers are displayed via both coloration and circle size. (d) Performance of the PS3 primer set of formalin-fixed paraffin-embedded (FFPE) tissue samples from deidentified lung cancer patients. Because the NGS libraries for these 5 samples differed slightly in total reads, here we plotted the distribution of reads normalized to 1 million reads. (e) The observed primer dimer species and their corresponding NGS reads were relatively similar between cell line genomic DNA and FFPE samples. (f) Demonstration of a 384-plex primer set designed by SADDLE (768 primers). The main diagonal shows On Target reads. Only about \(1\%\) of all reads were primer dimers (Supplementary Section S6). </center>  
+
+<|ref|>text<|/ref|><|det|>[[57, 615, 485, 660]]<|/det|>
+understanding (Supplementary Section S7). The Dimer reads for each pair of primers from PS1 are plotted against the predicted Badness in Fig. 4a.  
+
+<|ref|>text<|/ref|><|det|>[[57, 675, 485, 916]]<|/det|>
+To facilitate discussions of Badness function accuracy in terms of sensitivity and specificity, we set two separate thresholds: the Reads Threshold (horizontal orange line) and the Badness Threshold (vertical dotted purple line). The plotted Reads Threshold in Fig. 4a corresponds to the mean on- target read depth, and the Badness Threshold plotted correspond to the value that maximizes prediction sensitivity plus specificity. For these Threshold values, we observe a sensitivity \(92.5\%\) ( \(\frac{62}{67}\) ) and a specificity of \(90.3\%\) ( \(\frac{33226}{36797}\) ). By adjusting the Badness Threshold value, we can change the tradeoff between sensitivity and specificity, resulting in a Receiver Operator Characteristic (ROC) curve (Fig. 4b). The area under the ROC curve (AUROC) is 0.9577, indicating very high Dimer prediction accuracy by the Badness function. When the Read Threshold is adjusted higher, the AUROC also increases (Fig. 4c), but the positive predictive value (PPV)  
+
+<|ref|>text<|/ref|><|det|>[[513, 616, 582, 629]]<|/det|>
+decreases.  
+
+<|ref|>text<|/ref|><|det|>[[513, 638, 941, 895]]<|/det|>
+We next examined the top 5 most dominant Dimer reads in the library (Fig. 4d) and compared them to the top 5 predicted dimer reads based on the Badness function (Fig. 4e). It is noteworthy that only 1 of the two different top 5 lists overlap. The other 4 predicted dimers did not contribute significantly experimentally, and the other 4 observed dimers were not predicted to have high risk for dimer formation. At a glance, it appears we over- weighted the possibility of forming primer dimers in which the \(3'\) - most nucleotide in unpaired, and we may need to adjust the Badness function to allow a stronger attenuation of Badness based on distance from the \(3'\) end. Additionally, it appears that the Badness function may be not scaled optimally, as the log10(Badness) ranges between 0 and 3.5, whereas the log10(Dimers) ranges between 0 and 5 (Fig. 4f). This may mean that the current algorithm over- weights weak potential dimers, at the expense of insufficiently avoiding strong predicted primer dimers.  
+
+<|ref|>text<|/ref|><|det|>[[528, 902, 940, 917]]<|/det|>
+Beyond the above observations, it is not clear why some
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[60, 50, 930, 555]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[55, 566, 943, 687]]<|/det|>
+<center>FIG. 4: Evaluation of prediction accuracy of the Badness function for individual primer dimer candidates. (a) Comparison of observed vs. predicted primer dimers for all possible pairs of primers in the PS1 library. The horizontal orange line shows the mean on-target reads for the 96 intended amplicons. By changing the Badness Threshold, different tradeoffs of dimer prediction sensitivity and specificity can be achieved. For the current Reads Threshold and Badness Threshold, we observe \(92.5\%\) sensitivity and \(90.3\%\) specificity. (b) Receiver Operator Characteristic (ROC) curve for dimer prediction sensitivity and specificity achieved by changing the Badness Threshold. (c) The Area Under the ROC (AUROC) value depends on the Reads Threshold, with a highest achievable AUROC of about 0.98. (d) The top 5 dimer species experimentally observed to form with highest number of aligned NGS reads. (e) The top 5 potential dimer species predicted to form based on the Badness function. Note that only the \(\# 4\) species are present are both list in panels (d) and (e). (f) Distribution of observed NGS reads and predicted Badness for all possible primer dimers. </center>  
+
+<|ref|>text<|/ref|><|det|>[[57, 708, 485, 828]]<|/det|>
+dimers are observed at much higher reads experimentally than others. For example, the top observed dimer only has a 5 nt overlap at the \(3'\) end, compared to a 7 nt overlap at the \(3'\) end for the rank 4 dimer. This is not consistent with our understanding of DNA hybridization and polymerase extension kinetics, and implies that we may not be able to generate a perfect Badness function even ignoring computational resource constraints.  
+
+<|ref|>text<|/ref|><|det|>[[57, 841, 485, 917]]<|/det|>
+Gene Fusion Detection with qPCR and Sanger Sequencing. Gene fusions are therapeutic targets and attractive diagnostic biomarkers to guide treatment [22- 25]. Currently, gene fusions are detected either in single- plex by qPCR for known high- frequency fusions (e.g. BCR- ABL1),  
+
+<|ref|>text<|/ref|><|det|>[[512, 708, 940, 753]]<|/det|>
+or by NGS. A highly multiplexed qPCR assay that can detect tens of potential gene fusions relevant to a particular disease could greatly increase the accessibility of gene fusion testing.  
+
+<|ref|>text<|/ref|><|det|>[[512, 766, 940, 917]]<|/det|>
+Here, we used SADDLE to design a set of 60 primers to detect 56 actionable gene fusions for non- small cell lung cancer (NSCLC) across 6 genes (ALK, ROS1, RET, NRTK1, NTRK2, and NRTK3). The number of primers are lower than 56:2 because the same exon can be fused with multiple partner genes or exons. We detect the fusions in complementary DNA (cDNA) reverse transcribed from RNA, in order to limit the complexity and length of the detection targets. For each fusion of interest, the primer set includes a forward primer (fP) targets the upstream partner gene and a reverse primer
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[105, 50, 888, 315]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[55, 323, 943, 485]]<|/det|>
+<center>FIG. 5: Highly multiplexed qPCR detection of gene fusions using SADDLE-designed primer sets. (a) Complementary DNA (cDNA) prepared through reverse transcription of RNA can have known target sequences at the exon breakpoints. Although it is trivial to design a single-plex qPCR assay to detect a single known fusion, such as BCR-ABL1 [20], we are not aware of any reports of highly multiplexed qPCR assays to simultaneously detect \(\geq 10\) different gene fusion cDNA species. For this assay, we designed a 60 primer set (46 forward, 14 reverse) that together can amplify 56 distinct gene fusion types commonly observed in non-small-cell lung cancer [21]. (b) Summary of observed qPCR cycle threshold (Ct) values for the 56 reactions, each with 1 of the 56 synthetic fusion DNA species across 6 genes (ALK, ROS1, RET, and NTRK1/NTRK2/NTRK3), each with 1700 copies. WT indicates wildtype commercial cDNA, and NTC indicates no template control. See Supplementary Section S8 for additional details and experimental results, including Sanger sequencing traces of each reaction product. (c) Example qPCR trace showing detection of the fusion DNA sequence joining NACC2 exon 4 to NTRK2 exon 13. (d) Clinical sample results on cDNA reverse transcribed from RNA from extracellular vesicles. Samples 3, 4, and 7 tested positive for a gene fusion, and sequence alignment of the Sanger sequencing results (right panels) show the exact identifies of the fusions. </center>  
+
+<|ref|>text<|/ref|><|det|>[[57, 503, 427, 518]]<|/det|>
+(rP) targets the downstream partner gene (Fig. 5a).  
+
+<|ref|>text<|/ref|><|det|>[[57, 520, 485, 671]]<|/det|>
+We first tested the multiplex PCR panel against synthetic samples bearing the gene fusions of interest (Fig. 5bc). In all cases, the positive samples were clearly distinguishable by cycle threshold (Ct) value against both commercial wildtype cDNA (WT) and the no- template control (NTC), with all \(\Delta \mathrm{Ct}\) values above 10. We also tested the panel on synthetic gene fusion samples with a variant allele frequency (VAF) of \(1\%\) (Supplementary Section S8). The \(1\%\) VAF samples were constructed by mixing synthetic gBlocks that contained a single fusion (the variant) with human cDNA (the wildtype).  
+
+<|ref|>text<|/ref|><|det|>[[57, 672, 485, 792]]<|/det|>
+Finally, we applied the gene fusion qPCR panel to clinical cDNA samples extracted from extracellular vesicles in blood plasma from NSCLC patients (Fig. 5d). Of the 10 clinical samples analyzed, 3 were called positive for gene fusions. To identify the exact gene fusion in these samples, we performed Sanger sequencing on the amplicons from the positive samples. Two samples were identified with EML4 exon20- ALK exon20, and one was identified with EML4 exon 15- ALK exon 20.  
+
+<|ref|>sub_title<|/ref|><|det|>[[57, 800, 157, 816]]<|/det|>
+## Discussion  
+
+<|ref|>text<|/ref|><|det|>[[57, 825, 485, 915]]<|/det|>
+In this study, we designed a multiplex PCR primer design algorithm SADDLE targeting numerous genomic regions in a single tube. We presented experimental validation of primer sets on a 96- plex cancer- related exons panel, demonstrating that the SADDLE was capable in selecting better primers by reducing dimerization in a multiplex PCR reac  
+
+<|ref|>text<|/ref|><|det|>[[512, 503, 940, 683]]<|/det|>
+tion. The dimer rate decreased going from the \(90.7\%\) in a naively designed PS1 to \(39.6\%\) in an intermediate PS2 and to \(4.9\%\) in an optimized PS3, resulting in an increased on- target rate as well as greater uniformity of on- target amplicons. In another 384- plex panel targeting random- selected SNPs in the human genome, the NGS library using the optimized primer set showed a dimer rate of \(1\%\) . SADDLE can reduce reagent costs and enable the amplification of hundreds of target templates simultaneously without wasting NGS reads. Importantly, library preparation using optimized primer sets generated by SADDLE does not depend on labor- intensive enzymatic cleavage or size selection steps to remove dimers.  
+
+<|ref|>text<|/ref|><|det|>[[512, 690, 940, 915]]<|/det|>
+The improvement of NGS library on- target rates through the reduction of primer dimers can allow significantly larger targeted panels to be possible using multiplex PCR library preparation. Because multiplex PCR generally requires less input DNA and are faster than ligation- based library preparation approaches, due to the low yields of end repair and ligation, we envision that SADDLE- designed primers can be useful for a variety of research and clinical applications where DNA sample quantities are limited and/or where rapid turnaround is needed. For example, in oncology tissue biopsies obtained through fine needle aspirates and core biopsies are frequently insufficient for standard NGS analysis, and cell- free DNA from peripheral blood plasma likewise are limited and impose sensitivity limitations to ligation- based approaches. Furthermore, in reproductive medicine, samples
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[56, 52, 485, 111]]<|/det|>
+from amniocentesis and preimplantation genetic screening (PGS) and preimplantation genetic diagnosis (PGD) are also very limited, and require rapid turnaround for molecular diagnostics due to the time- sensitivity of clinical decisions.  
+
+<|ref|>text<|/ref|><|det|>[[57, 113, 486, 414]]<|/det|>
+Through our analyses of predicted vs. observed dimers, we found that the parameters in the Loss function used in SADDLE could be adjusted to optimize dimer prediction performance, particularly in the \(3^{\prime}\) distance attenuation. However, with the current SADDLE algorithm, Non- specific amplicons now appear to dominate off- target rates, rather than Dimers. Thus, to further scale- up the panels that can be designed by SADDLE, it will be necessary to construct and optimize new Loss functions that penalize primer sets based on predicted off- target genomic amplification. Modification of the Loss function to minimize Non- specific amplicon formation would require significantly more work, as it requires consideration of the expected sample genome sequence. Whereas the current Loss function is "universal" in improving multiplex PCR primer set designs, a Loss function that considers Non- specific amplicons would inherently be suboptimal for primer dimer minimization. A Loss function predicting Non- specific amplification must also consider external factors, including the average length of the DNA molecules in the sample and nonpathogenic genomic polymorphisms.  
+
+<|ref|>text<|/ref|><|det|>[[58, 415, 486, 596]]<|/det|>
+In medical and research applications where the cost of NGS cannot be economically justified, qPCR assays will likely be the dominant tool for study of genomic variants. In qPCR, even single- plex primers can form significant dimers if poorly designed with Ct values below 30. Multiplex qPCR thus typically requires significant empirical optimization, even at around 4- plex [26]. SADDLE allowed us to successfully design a 60- primer qPCR panel targeting 56 gene fusions, and exact fusion identities can be determined through affordable Sanger sequencing. Thus, we envision that SADDLE can revolutionize the use of qPCR for highly multiplex molecular diagnostics.  
+
+<|ref|>text<|/ref|><|det|>[[58, 605, 485, 633]]<|/det|>
+Acknowledgements. This work was funded by NIH grant R01CA203964 to DYZ.  
+
+<|ref|>text<|/ref|><|det|>[[58, 643, 486, 747]]<|/det|>
+Author contributions. NGX and DYZ conceived the project. NGX performed the NGS experiments. NGX, MXW, and PS analyzed the NGS data. YW, YY, and JL designed and performed the qPCR- and Sanger- based gene fusion experiments. NGX and YW analyzed the qPCR and Sanger data. NGX and DYZ wrote the manuscript with input from all authors.  
+
+<|ref|>text<|/ref|><|det|>[[58, 757, 486, 894]]<|/det|>
+Additional information. Correspondence may be addressed to DYZ (dyz1@rice.edu). There is a patent pending on the Multiplex Primer Design Algorithm presented in this manuscript, and this patent has been exclusively licensed to Nuprobe Global. NGX, MXW, and PS declare a competing interest in the form of consulting for Nuprobe USA. DYZ declares a competing interest in the form of consulting for and significant equity ownership in Nuprobe Global, Torus Biosystems, and Pana Bio.  
+
+<|ref|>text<|/ref|><|det|>[[58, 903, 485, 917]]<|/det|>
+Data Availability. The reference and sample- specific  
+
+<|ref|>text<|/ref|><|det|>[[512, 52, 940, 127]]<|/det|>
+gDNA sequence data are available from the NCBI Nucleotide database, the Ensembl database, the COSMIC database, and the Foundation Medicine gene list. All other data supporting the findings of this study are available within the paper and its Supplementary Information files.  
+
+<|ref|>text<|/ref|><|det|>[[512, 136, 940, 197]]<|/det|>
+Code Availability. The MATLAB code used for multiplex PCR primer algorithm and the MATLAB code and Python code for NGS data processing are available at https://github.com/NinaGXie/SADDLE.  
+
+<|ref|>text<|/ref|><|det|>[[519, 235, 941, 917]]<|/det|>
+[1] Razavi, P. et al. High- intensity sequencing reveals the sources of plasma circulating cell- free DNA variants. Nat Med 25, 1928- 1937 (2019). [2] Cohen, J. D. et al. Detection and localization of surgically resectable cancers with a multi- analyte blood test. Science 359, 926- 930 (2018). [3] Claesson, M. J., Clooney, A. G., & O'toole, P. W. A clinician's guide to microbiome analysis. Nat Rev Gastro Hepat, 14(10), 585 (2017). [4] Mamanova, L. et al. Target- enrichment strategies for next- generation sequencing. Nat Methods 7, 111- 118 (2010). [5] Bailey, J. A., Gu, Z., Clark, R. A., Reinert, K., Samonte, R. V., Schwartz, S., ... & Eichler, E. E. Recent segmental duplications in the human genome. Science, 297(5583), 1003- 1007 (2002). [6] Goodwin, S., McPherson, J. D. & McCombie, W. R. Coming of age: ten years of next- generation sequencing technologies. Nat Rev Genet 17 333- 351 (2016). [7] Elnifro, E. M., Ashshi, A. M., Cooper, R. J., & Klapper, P. E. Multiplex PCR: optimization and application in diagnostic virology. Clin Microbiol Rev, 13(4), 559- 570 (2000). [8] Sun, J. M. et al. Small- cell lung cancer detection in never- smokers: clinical characteristics and multigene mutation profiling using targeted next- generation sequencing. Ann Oncol 26, 161- 166 (2015). [9] Leamon, J., Andersen, M., & Thornton, M. U.S. Patent No. 9,957,558. Washington, DC: U.S. Patent and Trademark Office (2018). [10] Khodakov, D., Wang, C., & Zhang, D. Y. Diagnostics based on nucleic acid sequence variant profiling: PCR, hybridization, and NGS approaches. Adv Drug Deliver Rev, 105, 3- 19 (2016). [11] SantaLucia J Jr, Hicks D. The thermodynamics of DNA structural motifs. Annu Rev Biophys Biomol Struct, 33:415- 440. doi:10.1146/annurev.biophys.32.110601.141800 (2004). [12] Bae, J. H., Fang, J. Z., & Zhang, D. Y. High- throughput methods for measuring DNA thermodynamics. Nucleic Acids Res, 48(15), e89- e89 (2020). [13] Maurer, W. D., & Lewis, T. G. Hash table methods. ACM Comput Surv, 7(1), 5- 19 (1975). [14] Kirkpatrick, S., Gelatt, C. D., & Vecchi, M. P. Optimization by simulated annealing. Science, 220(4598), 671- 680 (1983). [15] Foundation One® Current Gene List. https://www.foundationmedicineasia.com/content/dam/rfm/apac_v2- en/FOne_Current_Gene_List.pdf (2014) [16] Abyzov, A., Urban, A. E., Snyder, M. & Gerstein, M. CNVnator: an approach to discover, genotype, and characterize typical and atypical CNVs from family and population genome sequencing. Genome Res 21, 974- 984 (2011). [17] Corcoran, R. B., & Chabner, B. A. Application of cell- free DNA analysis to cancer treatment. New Engl J Med, 379(18), 1754- 1765 (2018). [18] Ye, J., Coulouris, G., Zaretskaya, I., Cutcutache, I., Rozen, S., & Madden, T. L. Primer- BLAST: a tool to design target- specific primers for polymerase chain reaction. BMC bioinformatics 13, 134 (2012). [19] Langmead, B. & Salzberg, S. L. Fast gapped- read alignment with Bowtie 2. Nat Methods 9 357- 359 (2012). [20] White, H., Deprez, L., Corbisier, P., Hall, V., Lin, F., Mazoua, S., ... & Emons, H. A certified plasmid reference material for the standardisation of BCR- ABL1 mRNA quantification by real- time quantitative PCR. Leukemia, 29(2), 369- 376 (2015). [21] Tang, Z., Zhang, J., Lu, X., Wang, W., Chen, H., Robinson, M. K., ... & Medeiros, L. J. Coexistent genetic alterations involving ALK, RET, ROS1 or MET in 15 cases of lung adenocarcinoma. Modern Pathol 31(2), 307- 312 (2018). [22] Mertens, F., Johansson, B., Fioretos, T. & Mitelman, F. The emerging complexity of gene fusions in cancer. Nat Rev Cancer 15, 371- 381 (2015). [23] Powers M. P. The ever- changing world of gene fusions in cancer: a secondary gene fusion and progression. Oncogene 38(47), 7197- 7199 (2019). [24] Latysheva, N. S. & Babu, M. M. Discovering and understanding oncogenic gene fusions through data intensive computational approaches. Nucleic Acids Res 44 4487- 4503 (2016).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[60, 52, 488, 117]]<|/det|>
+[25] Heyer, E. E., Deveson, I. W., Wooi, D., Selinger, C. I., Lyons, R. J., Hayes, V. M., ... & Blackburn, J. Diagnosis of fusion genes using targeted RNA sequencing. Nat Commun, 10(1), 1- 12 (2019).[26] Shen, Z., Qu, W., Wang, W., Lu, Y., Wu, Y., Li, Z., ... & Zhang, C. MPprimer: a program for reliable multiplex PCR primer design. BMC bioinformatics, 11(1), 1- 7 (2010).
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[57, 49, 220, 66]]<|/det|>
+## Methods Results  
+
+<|ref|>text<|/ref|><|det|>[[57, 78, 485, 259]]<|/det|>
+Oligonucleotides and repository samples. All the primers and synthetic DNA templates were purchased from Integrated DNA Technologies. Primers were purchased as standard desalted DNA oligonucleotides, and synthetic templates as desalted double strand fragments (gBlocks). DNA oligonucleotides solutions were stored at \(4^{\circ}\mathrm{C}\) . Human cell- line gDNA sample NA18562 (Correll Biorepository) was stored at \(- 20^{\circ}\mathrm{C}\) . The gDNA was mixed with synthetic DNA templates at various ratios to create samples containing different proportions of a specific variant sequence. Dilution of gDNA samples and synthetic DNA templates were made in 1x TE buffer with \(0.1\%\) Tween 20 (Sigma Aldrich).  
+
+<|ref|>text<|/ref|><|det|>[[57, 265, 485, 430]]<|/det|>
+Multiplex PCR protocol. Multiplex PCR was performed on a T100 Thermocycler or a C1000 Thermocycler (Bio- Rad). The total volume of each reaction was 50ul. DNA sample input ranged from 10 ng to 100 ng per tube. PCR reagents including vent (exo- ) polymerase, ThermoPol Reaction Buffer (10X), and dNTP (New England Biolabs) were used for enzymatic amplification. Thermal cycling started with a 3 min incubation step at \(95^{\circ}\mathrm{C}\) for polymerase activation, followed by 17 cycles of 30 s at \(95^{\circ}\mathrm{C}\) for DNA denaturing, 3 min at \(60^{\circ}\mathrm{C}\) for annealing, and 30 s at \(72^{\circ}\mathrm{C}\) for extension, followed by a final extension of 5 min at \(72^{\circ}\mathrm{C}\) .  
+
+<|ref|>text<|/ref|><|det|>[[57, 437, 485, 573]]<|/det|>
+End repair protocol. Multiplex PCR product was end- repaired using NEBNext \((\widehat{\mathbb{R}})\) UltraTM II End Repair/dA- Tailing Module (New England Biolabs). Each reaction was a mixture of \(3\mu \mathrm{l}\) NEBNext Ultra II End Prep Enzyme Mix, \(7\mu \mathrm{l}\) NEBNext Ultra II End Prep Reaction Buffer, \(20\mu \mathrm{l}\) multiplex PCR products, and \(30\mu \mathrm{l}\) H2O. End repair was performed on a Eppendorf Mastercycler. Thermal cycling started with the incubation at \(20^{\circ}\mathrm{C}\) for 30 min and \(65^{\circ}\mathrm{C}\) for 30 min, with the heated lid set to \(80^{\circ}\mathrm{C}\) .  
+
+<|ref|>text<|/ref|><|det|>[[57, 582, 485, 747]]<|/det|>
+Adapter ligation. End repair mixture was ligated with adapters using NEBNext \((\widehat{\mathbb{R}})\) UltraTM II Ligation Module (New England Biolabs). Each reaction was a mixture of \(30\mu \mathrm{l}\) NEBNext Ultra II Ligation Master Mix, \(1\mu \mathrm{l}\) NEBNext Ligation Enhancer, \(2.5\mu \mathrm{l}\) NEBNext Adaptor for Illumina, and \(60\mu \mathrm{l}\) previous End repair mixture. Ligation was performed on a Mastercycler from Eppendorf. Thermocycling started with the incubation at \(20^{\circ}\mathrm{C}\) for 15 min with the heated lid off; after adding \(3\mu \mathrm{l}\) USERTM enzyme to the ligation mixture, the reaction was incubated at \(37^{\circ}\mathrm{C}\) for 15 min with the heated lid set to \(55^{\circ}\mathrm{C}\) .  
+
+<|ref|>text<|/ref|><|det|>[[57, 756, 485, 800]]<|/det|>
+Index quantitative PCR. Following adapter ligation, Index qPCR was performed on CFX96 Touch Deep Well Real- Time PCR Detection system (Bio- Rad). Quantification of  
+
+<|ref|>text<|/ref|><|det|>[[512, 51, 940, 186]]<|/det|>
+Index quantitative PCR. Following adapter ligation, Index qPCR was performed on CFX96 Touch Deep Well Real- Time PCR Detection system (Bio- Rad). Quantification of different libraries was performed simultaneously in each well. Each reaction was a \(10\mu \mathrm{l}\) mixture, with \(1\mu \mathrm{l}\) i5 index, \(1\mu \mathrm{l}\) i7 index, \(1\mu \mathrm{l}\) ligation products, \(2\mu \mathrm{l}\) Milli- Q, and \(5\mu \mathrm{l}\) PowerUp SYBR Green Master Mix. Experiment was performed following a thermal cycling protocol with a 3 min incubation step at \(95^{\circ}\mathrm{C}\) for polymerase activation, followed by 40 cycles of 10 s at \(95^{\circ}\mathrm{C}\) for DNA denaturing and 30 s at \(60^{\circ}\mathrm{C}\) for annealing and extension. Ct values were obtained directly from the CFX96 system.  
+
+<|ref|>text<|/ref|><|det|>[[512, 196, 940, 377]]<|/det|>
+Index PCR. Index PCR was performed on a T100 Thermocycler or a C1000 Thermocycler (Bio- Rad). Index primers used were NEBNext \((\widehat{\mathbb{R}})\) Multiplex Oligos for Illumina \((\widehat{\mathbb{R}})\) (New England Biolabs). Each reaction was a mixture of \(2\mu \mathrm{l}\) each i5 and i7 index primers, \(5\mu \mathrm{l}\) ligation products, and PCR reagents including vent (exo- ) polymerase, ThermoPol Reaction Buffer (10X), and dNTP. The volume of each reaction was \(52\mu \mathrm{l}\) . Thermal cycling started with a 3 min incubation step at \(95^{\circ}\mathrm{C}\) for polymerase activation, followed by various cycles of 30 s at \(95^{\circ}\mathrm{C}\) for DNA denaturing and 30 s at \(60^{\circ}\mathrm{C}\) for annealing, and 30 s at \(72^{\circ}\mathrm{C}\) for extension, followed by a final extension of 5 min at \(72^{\circ}\) .  
+
+<|ref|>text<|/ref|><|det|>[[512, 387, 940, 477]]<|/det|>
+Column Purification. Multiplex PCR products and ligation products were all purified using DNA Clean & Concentrator Kits (ZYMO Research). The volume of DNA- binding buffer was \(250\mu \mathrm{l}\) for multiplex PCR products clean- up, and \(482.5\mu \mathrm{l}\) for ligation products clean- up; \(25\mu \mathrm{l}\) Milli- Q water was used as elution buffer for each reaction.  
+
+<|ref|>text<|/ref|><|det|>[[512, 487, 940, 547]]<|/det|>
+Beads Purification. Index PCR product was purified using AMPure XP beads (Beckman Coulter). For each \(50\mu \mathrm{l}\) reaction mixture, \(90\mu \mathrm{l}\) of beads was added; \(40\mu \mathrm{l}\) Milli- Q water was used as elution buffer.  
+
+<|ref|>text<|/ref|><|det|>[[512, 558, 940, 602]]<|/det|>
+Library quantitation. All the libraries were quantified using the QubitTM dsDNA HS Assay Kit (ThermoFisher Scientific).  
+
+<|ref|>text<|/ref|><|det|>[[512, 613, 940, 673]]<|/det|>
+Bioanalyzer. Sizes of PCR products and libraries were measured using Bioanalyzer High Sensitivity DNA Assay (Agilent), and DNA chips were run on the Agilent 2100 Bioanalyzer system.  
+
+<|ref|>text<|/ref|><|det|>[[512, 684, 940, 729]]<|/det|>
+Next- Generation Sequencing. All the libraries were loaded on a Miseq Reagent V2 for obtaining pair- end reads and were sequenced on a Miseq (Illumina).  
+
+<|ref|>text<|/ref|><|det|>[[512, 740, 940, 800]]<|/det|>
+Sanger sequencing. PCR products were purified and prepared using a BigDye Terminator v1.1 Cycle Sequencing Kit (Thermo Fisher Scientific) and were sequenced on a Thermo Fisher Scientific 3500 Series Genetic Analyzer.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[44, 44, 143, 70]]<|/det|>
+## Figures  
+
+<|ref|>image<|/ref|><|det|>[[60, 115, 920, 560]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[44, 592, 115, 611]]<|/det|>
+<center>Figure 1 </center>  
+
+<|ref|>text<|/ref|><|det|>[[40, 633, 950, 770]]<|/det|>
+Overview of Simulated Annealing Design using Dimer Likelihood Estimation (SADDLE). Given a set of DNA target sequences Ti, with i between 1 and N, the goal is to design a total of 2N PCR primers that can effectively amplify all DNA targets, while generating an acceptably low amount of primer dimer species. Steps 4 and 5 can be repeated a large number of times in order to improve (decrease) the Loss function value on the final primer set S. Multiple implementations, hyper- parameters, and parameters can be selected for each SADDLE step that can impact performance and speed.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[58, 66, 941, 377]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[44, 400, 117, 420]]<|/det|>
+<center>Figure 2 </center>  
+
+<|ref|>text<|/ref|><|det|>[[39, 443, 955, 805]]<|/det|>
+Implementation and experimental evaluation of a multiplex primer design algorithm based on the SADDLE framework. (a) Method for generating candidate primer sequences for a DNA target T. (b) Implementation of Badness function that can be rapidly evaluated using hash tables. (c) List of cancer genes selected as target sequences for a 96- plex primer set design. See Supplementary Excel spreadsheet for target selection details. (d) Loss function of primer sets S(g) across optimization generations g. The Loss function value decreases through the optimization and approaches a local minima after roughly 400 generations. We selected three different primer sets, constructed at generations 0, 200, and 400 for experimental evaluation; these are respectively called PS1, PS2, and PS3 for the remainder of this paper. (e) Capillary electrophoresis (Agilent Bioanalyzer 2100) analysis of amplicon products of PS1, PS2, and PS3. Here, 10 ng of the NA18562 human genomic DNA (Coriell) was used as input, and the median primer concentration was 45 nM in the PCR reaction. 17 cycles of PCR were performed using Vent (exo- ) DNA polymerase (selected for its improved amplification ability for G/C- rich sequences). To facilitate more in- depth analysis by high- throughput sequencing (NGS), adapters/indexes were ligated to the amplicon products. No size selection was performed, in order to accurately reflect the fraction of primer dimer species following multiplex PCR. The on- target amplicons are expected to have an average length of roughly 250 nt.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[55, 65, 940, 430]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[43, 460, 117, 479]]<|/det|>
+<center>Figure 3 </center>  
+
+<|ref|>text<|/ref|><|det|>[[39, 500, 951, 911]]<|/det|>
+Experimental NGS results for SADDLE- designed primer sets. (a) Distribution of reads observed in NGS library constructed using PS1, PS2, and PS3. On Target reads are defined as those that aligned to the intended amplicons; Dimer reads are defined as those whose insert lengths are smaller than the sum of the two primer lengths; all other reads were classified as Non- specific. The vast majority of Non- specific reads align to other regions of the human genome, via a non- cognate pair of forward and reverse primers. The fraction of NGS reads mapped to Dimers dramatically decreases from PS1 to PS2 to PS3. (b) Distribution of NGS reads in the 3 primer set libraries. (c) Distribution of observed primer dimers, based on aligned reads. Because forward primers (fP) can also form primer dimers with other forward primers, we aligned the first and last 25 nucleotides of each NGS read to the merged set of fP and rPs, with primers 1 through 96 in the diagram showing fPs and primers 97 through 192 showing rPs. For clarity of visualization, the log number of reads of observed primer dimers are displayed via both coloration and circle size. (d) Performance of the PS3 primer set of formalin- fixed paraffin- embedded (FFPE) tissue samples from deidentified lung cancer patients. Because the NGS libraries for these 5 samples differed slightly in total reads, here we plotted the distribution of reads normalized to 1 million reads. (e) The observed primer dimer species and their corresponding NGS reads were relatively similar between cell line genomic DNA and FFPE samples. (f) Demonstration of a 384-plex primer set designed by SADDLE (768 primers). The main diagonal shows On Target reads. Only about 1% of all reads were primer dimers (Supplementary Section S6).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[60, 60, 936, 570]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[44, 595, 118, 614]]<|/det|>
+<center>Figure 4 </center>  
+
+Evaluation of prediction accuracy of the Badness function for individual primer dimer candidates. (a) Comparison of observed vs. predicted primer dimers for all possible pairs of primers in the PS1 library. The horizontal orange line shows the mean on- target reads for the 96 intended amplicons. By changing the Badness Threshold, different tradeoffs of dimer prediction sensitivity and specificity can be achieved. For the current Reads Threshold and Badness Threshold, we observe \(92.5\%\) sensitivity and \(90.3\%\) specificity. (b) Receiver Operator Characteristic (ROC) curve for dimer prediction sensitivity and specificity achieved by changing the Badness Threshold. (c) The Area Under the ROC (AUROC) value depends on the Reads Threshold, with a highest achievable AUROC of about 0.98. (d) The top 5 dimer species experimentally observed to form with highest number of aligned NGS reads. (e) The top 5 potential dimer species predicted to form based on the Badness function. Note that only the #4 species are present are both list in panels (d) and (e). (f) Distribution of observed NGS reads and predicted Badness for all possible primer dimers.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[65, 50, 916, 345]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[44, 368, 118, 387]]<|/det|>
+<center>Figure 5 </center>  
+
+<|ref|>text<|/ref|><|det|>[[41, 409, 950, 749]]<|/det|>
+Highly multiplexed qPCR detection of gene fusions using SADDLE- designed primer sets. (a) Complementary DNA (cDNA) prepared through reverse transcription of RNA can have known target sequences at the exon breakpoints. Although it is trivial to design a single- plex qPCR assay to detect a single known fusion, such as BCR- ABL1 [20], we are not aware of any reports of highly multiplexed qPCR assays to simultaneously detect \(\boxed{10}\) different gene fusion cDNA species. For this assay, we designed a 60 primer set (46 forward, 14 reverse) that together can amplify 56 distinct gene fusion types commonly observed in non- small- cell lung cancer [21]. (b) Summary of observed qPCR cycle threshold (Ct) values for the 56 reactions, each with 1 of the 56 synthetic fusion DNA species across 6 genes (ALK, ROS1, RET, and NTRK1/NTRK2/NTRK3), each with 1700 copies. WT indicates wildtype commercial cDNA, and NTC indicates no template control. See Supplementary Section S8 for additional details and experimental results, including Sanger sequencing traces of each reaction product. (c) Example qPCR trace showing detection of the fusion DNA sequence joining NACC2 exon 4 to NTRK2 exon 13. (d) Clinical sample results on cDNA reverse transcribed from RNA from extracellular vesicles. Samples 3, 4, and 7 tested positive for a gene fusion, and sequence alignment of the Sanger sequencing results (right panels) show the exact identifies of the fusions.  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 771, 310, 798]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[44, 821, 765, 841]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[60, 859, 285, 959]]<|/det|>
+- SADDLESequence.xlsx- SADDLESequence.xlsx- SADDLESupv4.pdf- SADDLESupv4.pdf
+
+<--- Page Split --->

preprint/preprint__0fadf43a852d9e8dc0c5812bfff32104d323571f1b41834242e5098c378037ec/images_list.json

@@ -0,0 +1,70 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Figure 1 Interfaces of nano-bio interactions in the lungs.",
+    "footnote": [],
+    "bbox": [
+      [
+        160,
+        130,
+        835,
+        720
+      ]
+    ],
+    "page_idx": 25
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2 Effects of representative MeONPs in cells and mouse lungs.",
+    "footnote": [],
+    "bbox": [
+      [
+        140,
+        88,
+        870,
+        840
+      ]
+    ],
+    "page_idx": 26
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Figure 3 Performance of models constructed using eight machine learning algorithms.",
+    "footnote": [],
+    "bbox": [
+      [
+        184,
+        105,
+        802,
+        578
+      ]
+    ],
+    "page_idx": 28
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Figure 4 Experimental validation and diversity analysis of the established model.",
+    "footnote": [],
+    "bbox": [
+      [
+        205,
+        111,
+        777,
+        760
+      ]
+    ],
+    "page_idx": 29
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Figure 5 Identification of key descriptors for mechanism interpretation.",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 30
+  }
+]

preprint/preprint__0fadf43a852d9e8dc0c5812bfff32104d323571f1b41834242e5098c378037ec/preprint__0fadf43a852d9e8dc0c5812bfff32104d323571f1b41834242e5098c378037ec.mmd

@@ -0,0 +1,292 @@
+
+# Multimodal Feature Fusion Machine Learning for Predicting Chronic Injury Induced by Engineered Nanomaterials  
+
+Xuehua Li  
+
+lixuehua@dlut.edu.cn  
+
+Dalian University of Technology https://orcid.org/0000- 0003- 3924- 0987  
+
+## Article  
+
+Keywords: Nanotoxicity, lung fibrosis, predictive toxicology, structure- activity relationship, nano- bio interface.  
+
+Posted Date: November 3rd, 2023  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 3479434/v1  
+
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Additional Declarations: There is NO Competing Interest.  
+
+Version of Record: A version of this preprint was published at Nature Communications on March 20th, 2025. See the published version at https://doi.org/10.1038/s41467- 025- 58016- w.
+
+<--- Page Split --->
+
+Multimodal Feature Fusion Machine Learning for Predicting Chronic Injury Induced by Engineered Nanomaterials  
+
+Yang Huang \(^{1,4\#}\) , Jiayu Cao \(^{2\#}\) , Xuehua Li \(^{1*}\) , Qing Yang \(^{2}\) , Qianqian Xie \(^{3}\) , Xi Liu \(^{3}\) , Xiaoming Cai \(^{2*}\) , Jingwen Chen \(^{1}\) , Huixiao Hong \(^{5}\) , Ruibin Li \(^{3*}\) \(^{1}\) Key Laboratory of Industrial Ecology and Environmental Engineering (MOE), Dalian Key Laboratory on Chemicals Risk Control and Pollution Prevention Technology, School of Environmental Science and Technology, Dalian University of Technology, Dalian 116024, China \(^{2}\) School of Public Health, Soochow University, Suzhou, Jiangsu, 215123, China \(^{3}\) State Key Laboratory of Radiation Medicine and Protection, School of Radiation Medicine and Protection, Soochow University, Suzhou, Jiangsu, 215123, China \(^{4}\) School of Chemistry and Materials Science, Ludong University, Yantai, 264025, China \(^{5}\) National Center for Toxicological Research, U.S. Food and Drug Administration, Jefferson, AR 72079, USA #Equal contribution \(^{*}\) Address correspondence to Dr. Xuehua Li Dalian University of Technology, Linggong Road 2, Dalian 116024, P R China Email: lixuehua@dlut.edu.cn Dr. Xiaoming Cai 199 Ren'ai Road, 401 Building, Suzhou, 215123, Jiangsu, the People's Republic of China Email: xmcai@suda.edu.cn & Dr. Ruibin Li 199 Ren'ai Road, 401 Building, Suzhou, 215123, Jiangsu, the People's Republic of China Email: liruibin@suda.edu.cn
+
+<--- Page Split --->
+
+## Abstract  
+
+Concerns regarding chronic injuries (e.g., fibrosis and carcinogenesis) induced by nanoparticles raised public health concerns and needs to be rapidly assessed in hazard identification. Although in silico analysis is commonly used for risk assessment of chemicals, predicting chronic in vivo nanotoxicity remains challenging due to the complex interactions at multiple nano- bio interfaces. Herein, we developed a multimodal feature fusion analysis framework to predict the fibrogenic potential of metal oxide nanoparticles (MeONPs). Eighty- seven features derived from multiple MeONP- lung interfaces were used to develop a machine learning- based predictive framework. We identified viability damage and cytokine (IL- 1β and TGF- β1) production in macrophages and epithelial cells as key events that are closely associated with particle size, surface charge, and interactions with lysosomes. Experimental validations showed that the developed in silico model had 85% accuracy in predicting MeONP- induced lung fibrosis. The heterogeneity distribution of data points indicated good applicability of the predictive model. Our findings demonstrate the potential usefulness of this predictive model for risk assessment of nanomaterials and in assisting regulatory decision making.  
+
+Keywords: Nanotoxicity, lung fibrosis, predictive toxicology, structure- activity relationship, nano- bio interface.
+
+<--- Page Split --->
+
+## Main Text  
+
+More than 10,000 nanoproducts have been produced and used in various industries worldwide<sup>1</sup>. During the lifecycle of some nanoproducts, nanoparticles may be released to form aerosols and travel in the air<sup>2</sup>. Inhalation is the major exposure route for airborne nanoparticles and can cause respiratory injury in mammals<sup>3- 4</sup>. While certain engineered nanomaterials (ENMs) such as ZnO, CuO, Au, and Ag may produce acute lung inflammation via reactive oxygen species (ROS) generation<sup>5- 6</sup>, inflammasome activation<sup>7- 9</sup>, or proinflammatory cytokine release<sup>7,8, 10- 11</sup>, other ENMs such as carbon nanotubes and metal oxides may induce chronic respiratory toxicity such as lung fibrosis<sup>12- 16</sup> and carcinogenesis<sup>17- 19</sup>, which are permanent and irreversible lung injuries. The chronic adverse outcomes of ENMs have raised substantial product safety concerns and led to strict regulatory requirements on utilizing nanotechnology. For instance, carbon nanotubes have been added to the Substitute It Now (SIN) list due to their carcinogenic potential, which restricts their use in Europe<sup>20</sup>. Since the number of ENMs is rapidly growing and each nanomaterial may have various derivatives with differing size, shape, surface chemistry, crystallinity, and so on, a large number of animals would need to be sacrificed to adequately assess the risk of ENMs. It was estimated that the in vivo toxicity assessment of the first generation of ENMs in commercial use may take up to 50 years and cost billions of dollars<sup>21</sup>. To address this challenge, in silico predictive models have been developed to reduce experimental costs.
+
+<--- Page Split --->
+
+In recent years, there have been several attempts to predict the in vitro toxicity of nanoparticles<sup>11, 22- 30</sup>. ENM- induced cytotoxicity, including cell death and inflammatory effects, is often caused by a molecular initiating event (MIE) that is highly dependent on the physicochemical properties of ENMs<sup>11, 25- 26, 28</sup>. To establish in silico models, physicochemical properties have been extensively exploited to identify predictive features. For instance, Labouta et al. found that the cytotoxicity of ENMs could be primarily predicted based on their material chemistry, followed by nanoparticle size and concentration, cell type, and cytotoxicity biomarkers<sup>25</sup>. The surface chemistry, lipophilicity, zeta potential, shape, and size of ENMs also impact their uptake and toxicology behaviors<sup>31- 32</sup>. In addition, in silico analysis has been widely used to explore the quantitative structure- activity relationships (QSAR) of ENMs in various cell types, such as human lung cells<sup>27, 33</sup>, RAW 264.7 cells<sup>24, 34</sup>, human keratinous cells (HaCaT)<sup>35</sup>, THP- 1 cells<sup>11</sup>, and Escherichia coli<sup>36- 37</sup> for hazard ranking. However, there is no reliable model to predict the chronic respiratory toxicity of ENMs in vivo such as lung fibrosis. The prediction of chronic toxicity is a major challenge in nanotoxicity studies, as it involves multiple nano- bio interactions that cannot be entirely mirrored at one specific nano- bio interface. Taking lung fibrosis as an example, its pathogenic progress involves the interactions of nanoparticles with lung lining fluids and multiple subcellular organelles such as the plasma membrane, lysosome, mitochondria, and cytoplasmic components in immune/epithelial cells.  
+
+Herein, we aimed to establish a predictive framework for pulmonary fibrosis induced by metal oxide nanoparticles (MeONPs). Based on the biological fate of MeONPs in the lungs, we
+
+<--- Page Split --->
+
+prepared a library of 52 MeONPs and collected the potential predictive features at multiple interfaces between MeONPs and biological contexts such as membranes, lysosomes, mitochondria, and other cytoplasmic components (Figure 1A). We acquired a total of 87 multimodal features and two fibrogenic indexes, which were then subjected to machine learning modeling. We thoroughly evaluated performance of the developed predictive models using overall predictive accuracy (ACC), Matthews correlation coefficient (MCC), sensitivity (SE), specificity (SP), area under the receiver operating characteristic curve (AUC), and \(F1\) score. The random forest (RF) model performed the best and was further experimentally validated by five MeONPs. Our study presents the first in silico framework for decoding the in vitro to in vivo extrapolation (IVIVE) of ENM- induced lung fibrosis.  
+
+## Construction of a multimodal database  
+
+A high- quality database is essential for building reliable IVIVE models and identifying key events involved in the pathogenic process of lung fibrosis<sup>38</sup>. However, there is a lack of such a database to define the behaviors of MeONPs and their biological effects at multiple nano- bio interfaces. Therefore, we constructed a reliable database of 52 MeONPs by characterizing their physicochemical properties, examining their interactions with biological fluids and subcellular organelles, and assessing their fibrogenic effects in mouse lungs by detecting collagen deposition in lung tissues and TGF- \(\beta 1\) release in bronchoalveolar lavage fluid (BALF). The sources of the 52 MeONPs are listed in Table S1.
+
+<--- Page Split --->
+
+Input parameters. Nano- bio interactions occur mainly in the pulmonary alveoli which consist of lung lining fluid, macrophages, and epithelial cells. Initially, inhaled MeONPs may interact with the lining fluid, altering their surface chemistry and dispersion states and dissolving to release metal ions. We therefore examined the behavior of MeONPs in simulated lung fluid (SLF) (Table S1). Then, the MeONPs may interact with lung cells. To visualize the intracellular path of MeONPs, \(\mathrm{Bi}_2\mathrm{O}_3\) and \(\mathrm{Fe}_2\mathrm{O}_3\) were selected to conjugate with fluorescein isothiocyanate (FITC) for fluorescence microscopy imaging. As shown in Figure 1B, the labeled MeONPs showed a time- dependent cascaded distribution pattern from the extracellular matrix into the membrane, lysosome, and cytoplasm. At 8- 16 h, the MeONPs overlapped nicely with lysosomes (Figure 1C). Considering the acidic and enzymatic traits of lysosomes, we also examined the dissolution of MeONPs in phagolysosomal simulated fluid (PSF) (Table S1). Based on the distribution features of MeONPs, we examined the impacts of MeONPs on the cell membrane, lysosome, mitochondria activity, energy production, and redox homeostasis by detecting lactate dehydrogenase (LDH) leakage, lysosomal pH change, nicotinamide adenine dinucleotide hydride (NADH) content, adenosine triphosphate (ATP) level, and ROS generation, respectively. These interactions may lead to cytokine release and affect cell- cell communications such as the recruitment of immune cells and proliferation of profibrogenic cells. We measured the release of pro- inflammatory cytokines (TNF- \(\alpha\) , IL- 1 \(\beta\) , IL- 2, IL- 6, MCP- 1) in THP- 1 cells and growth factor (TGF- \(\beta 1\) ) in BEAS- 2B cells. For a snapshot view of the results, the measured levels of descriptors in cells incubated with 0- 200 \(\mu \mathrm{g / mL}\) MeONPs were ranked into three levels, leading to a visual display where high, moderate and negligible effects
+
+<--- Page Split --->
+
+are represented by red, yellow and blue colors, respectively. Figures 2A and S1 show the effects of ten representative MeONPs in THP- 1 and BEAS- 2B cells, respectively, while the others are listed in Table S2. Fifteen branch events were quantitatively measured for the 52 MeONPs at five different exposure concentrations (12.5, 25, 50, 100, 200 \(\mu \mathrm{g / mL}\) ), resulting in 75 input parameters (Table S2). Additionally, four descriptors were collected from the interactions between MeONPs and two biological media (SLF and PSF), and eight descriptors were acquired from the periodic table (Table S1). These data allowed us to establish a multimodal database consisting of 87 input parameters in total (Table 1).  
+
+In vivo toxicology endpoints. Mice were exposed to the 52 MeONPs via oropharyngeal aspiration to collect toxicological endpoints in vivo for assess their fibrogenic effects. Figure 2B illustrates the workflow of the animal experiments. The animals were subjected to three exposures to MeONP suspensions during the first three weeks and were sacrificed on day 90 to collect BALF and lung tissues for further examination. In the pathogenic process of lung fibrosis, active TGF- \(\beta 1\) plays a critical role in promoting fibroblast proliferation to secrete collagens. We, therefore, quantified the release of active TGF- \(\beta 1\) in BALF by ELISA and visualized collagen deposition by Masson’s trichrome staining of lung sections. Collagen staining images acquired from 203 whole slide images of lung sections were assessed by the Ashcroft score, a widely used index to rank pulmonary fibrosis (Figure 2C). Figure 2D shows the expression levels of TGF- \(\beta 1\) in BALF and the Ashcroft score of mouse lungs exposed to 10 representative MeONPs. The fold change of TGF- \(\beta 1\) (FCTGF- \(\beta 1\) ) and collagen staining images of the 52 MeONPs are shown in Table S3 and Figure S2, respectively. Among the 52 tested
+
+<--- Page Split --->
+
+MeONPs, 23 (Fe2O3- 4, TiO2- r- 1, TiO2- r- 2, TiO2- a- 1, TiO2- a- 2, TiO2- a- 3, \(\alpha\) - Al2O3, \(\gamma\) - Al2O3- 1, \(\gamma\) - Al2O3- 2, Dy2O3, NiO- 1, In2O3- 1, In2O3- 2, Tm2O3, Tb2O3, Bi2O3, Sm2O3, Y2O3, Yb2O3, Sb2O3, SnO2, CuO- 3 and ZnO- 1) consistently induced high levels of TGF- \(\beta 1\) with \(\mathrm{FC}_{\mathrm{TGF - }\beta 1}\geq 2\) and significant collagen deposition, 11 (CeO2- 2, \(\alpha\) - MnO2- 1, CuO- 2, CuO- 4, NiO- 2, MgO, MoO3, ZnO- 2, ZnO- 3, Fe2O3- 2 and Fe2O3- 3) had a moderate effect, and 18 (CeO2- 1, CeO2- 3, Co3O4- 1, Co3O4- 2, Co3O4- 3, Co3O4- 4, Co3O4- 5, \(\alpha\) - MnO2- 2, \(\alpha\) - MnO2- 3, Eu2O3- 1, Eu2O3- 2, Cr2O3, Gd2O3, Nd2O3, Er2O3, La2O3, CuO- 1 and Fe2O3- 1) showed similar levels of cytokine release and collagens to the vehicle control. Notably, four MeONPs had differing effects on TGF- \(\beta 1\) and collagen deposition: MgO, MoO3, ZnO- 2, and ZnO- 3 induced high TGF- \(\beta 1\) release, but had little effect on collagen deposition with an Ashcroft score \(< 2\) . Considering this difference, a MeONP was classified as a fibrogenic inducer if \(\mathrm{FC}_{\mathrm{TGF - }\beta 1}\geq 2\) or Ashcroft score \(\geq 2\) . Among all in vivo datapoints, 126 were identified as fibrogenic. For each in vivo datapoint, there were 87 associated input parameters for multimodal feature fusion (MFF) analysis.  
+
+## ML modeling in multimodal feature fusion predictive framework  
+
+Given the complexity of the 15 branched events involved in MeONP- induced lung fibrosis at different nano- bio interfaces, machine learning methods were selected to establish the predictive framework. An MFF predictive framework driven by machine learning was proposed to identify meaningful patterns between heterogeneous multidimensional events and lung fibrosis in animals. To identify a suitable machine learning algorithm for the database, models were developed and evaluated using eight algorithms, including random forest (RF), locally
+
+<--- Page Split --->
+
+weighted learning (LWL), C4.5 decision tree (C4.5), k- nearest neighbor (k- NN), support vector machine (SVM), Bayesnet, decision table (DT), and logistic regression (LGR). The dataset was randomly split into a training set containing 156 data points of 40 MeONPs and a test set consisting of 47 data points generated from 12 MeONPs. For each MeONP, all its four data points were included in either the training set or the test set. The six performance metrics values of the eight machine learning algorithms in the 10- fold cross- validation on the training set and in the external validation on the test set are plotted in Figure 3. Their detailed performances \((ACC, MCC, SE, SP, AUC,\) and \(F1\) values) of the eight algorithms are provided in Table S4. The formulae of the indexes \((ACC, MCC, SE, SP, AUC,\) and \(F1\) values) are listed in the supporting information. Among all models, the RF model had the best performance, with a strong robustness \((ACC = 89\%\) and \(AUC = 94\%\) in the 10- fold cross- validation) and a high predictive accuracy \((ACC = 84\%\) and \(AUC = 85\%\) in the external validation). The models developed with C4.5, SVM, and Bayesnet also exhibited a satisfactory performance, with \(ACC\) values 85 - 90% in the 10- fold cross- validations and 74 - 76% in the external validations. Their \(AUC\) values ranged from 84% to 90% in the cross- validations and 61% to 79% in the external validations. However, the remaining four classifiers showed poor performance in the external validations. The \(ACC\) values were below 70% for \(k\) - NN, DT, and LGR models, and the \(AUC\) for LWL model was lower less than 40%. Based on these results, RF was selected as the optimal algorithm for the MFF predictive framework. Thus, the RF model was further validated by animal tests.
+
+<--- Page Split --->
+
+## Experimental validation of the MFF predictive model  
+
+Five new MeONPs \(\mathrm{(Ho_2O_3}\) , \(\mathrm{Pr_6O_{11}}\) , \(\mathrm{Co_3O_4}\) , \(\mathrm{ZrO_2}\) , and \(\mathrm{CuO}\) ) were selected for validation based on two criteria: i) they were not included in the training or test sets, and ii) the positive/negative ratio in this validation set fell within the range of 1:1 to 2:1, which is consistent with the ratios observed in the training and test sets. These selected MeONPs were administered to 20 mice (n = 4 for each MeONP) for 90 days, and the extent of fibrosis was determined by measuring TGF- \(\beta 1\) in BALF (Figure S3A) and collagen deposition in lung tissues (Figure S3B). The resulting 20 data points from the animal tests were then compared to the predictions generated by the MFF models (Table S5). The models exhibited high predictive accuracy for this independent dataset of lung fibrosis, achieving an ACC of 85% and AUC of 98%. As shown in Figure 4A, 17 out of the 20 data points were accurately predicted, primarily for \(\mathrm{Ho_2O_3}\) , \(\mathrm{Pr_6O_{11}}\) , \(\mathrm{Co_3O_4}\) , and \(\mathrm{ZrO_2}\) . The remaining three data points generated from CuO were correctly predicted in terms of collagen staining but not TGF- \(\beta 1\) levels.  
+
+We further assessed the diversity and applicability domain of the validated MEF model. These two characteristics are crucial for ensuring the inclusion of representative data and defining the range within which the proposed framework can be reliably applied for new MeONPs. To characterize the applicability domain, we employed a descriptor standardization approach, which encompassed all 223 data points from the training set (156), test set (47), and experimental validation set (20) (Table S6). A similarity network was generated based on the descriptor spaces of the established models to visualize the distribution of the 223 data points (Figure 4B). The thickness of the lines between two data points represented the strength of their
+
+<--- Page Split --->
+
+correlation, and the sparse connections indicated high homogeneity within the descriptor spaces. Notably, the descriptor space of MeONPs exhibited significant heterogeneity, with data points scattered widely.  
+
+## Mechanism interpretation  
+
+We attempted to interpret the mechanism by analyzing the key biological events associated with MeONP- induced lung fibrosis. A Shapley Additive Explanation plot (SHAP) plot was used to identify the features that exert significant influence. Figure 5A illustrates the SHAP summary plot, where each dot represents an individual data point in the dataset. The horizontal position of the dots on the x- axis indicates the impact of the corresponding feature values on the model's predictions. The top seven features that exert the greatest influence on the model's predictions are IL- 1β, NADH in macrophages, TGF- β1, dissolution in PSF, zeta potential, hydrodynamic size, and NADH in epithelial cells. Notably, IL- 1β emerges as the most critical descriptor, contributing \(27.8\%\) to the overall feature importance, followed by NADH in THP- 1 at \(17.6\%\) . The remaining five descriptors accounted for \(54.5\%\) of the feature importance (Figure 5B). Based on the results of the SHAP analysis, we sought to elucidate the mechanisms underlying MeONP- induced lung fibrosis. As illustrated in Figure 5C, while the mucociliary escalator effectively removes large MeONP agglomerates from the airways, small particles, especially cationic particles, can enter lysosomes for decomposition. The acidic environment of the lysosomal fluids may accelerate the dissolution of MeONPs. Subsequently, the released metal ions or escaped MeONPs can interact with mitochondria, disrupting cellular metabolism
+
+<--- Page Split --->
+
+and eliciting ROS, leading to the activation of IL- 1β and TGF- β1. These cytokines, in turn, promote the recruitment of immune cells, proliferation of fibroblasts, and deposition of collagen, culminating in the development of pulmonary fibrogenesis. Notably, these key events have been well documented in the literature regarding fibrosis pathology and the pulmonary behavior of nanoparticles<sup>26, 39</sup>.  
+
+As the number of ENMs continues to expand, it is crucial to evaluate their potential profibrogenic risks before exposure to the environment. MFF has demonstrated high predictive accuracy in assessing the fibrogenic risk of MeONPs. To improve the practicality of the MFF model, it was transformed into a software named "Nano- induced Lung Fibrosis Prediction" (NILFP v 1.0.0) with a simplified user interface. It is available for noncommercial use at GitHub (https://github.com/huangyang2023/NILFPv1.0.0/releases/download/NILFPv1.0.0/NILFP.v1.0.0.zip). NILFP can be used for fibrogenic risk assessment of untested MeONPs, MeONP- based nanoproducts and beyond. It will greatly speed up the respiratory risk assessment of nano- enabled products.  
+
+## Conclusions  
+
+Our study presents a reliable in silico model for predicting the fibrogenic potential of ENMs through the construction of an MFF predictive framework. The MEF model achieved high accuracy (>85%) in predicting MeONP- induced lung fibrosis, making it a valuable tool for risk assessment. The developed model utilized diverse data points and was broadly applicable to different MeONPs. Furthermore, machine learning analysis identified seven key descriptors
+
+<--- Page Split --->
+
+and facilitated the interpretation of the underlying biological mechanism. Our research provides a cost- effective, time- efficient, and mechanism- driven alternative to the current practice of chronic nanotoxicity assessment in animals.  
+
+## Experimental section  
+
+## Materials  
+
+MTS and ATP assay kits were obtained from Promega (Madison, WI, USA). H2DCF- DA, Fetal bovine serum (FBS), penicillin, and streptomycin were obtained from Thermo Fisher Scientific (Grand Island, NY, USA). RPMI 1640 medium were purchased from Corning (Steuben, NY, USA). Bronchial epithelial cell medium were purchased from Sciencell (San Diego, CA, USA). ELISA kits for detection of IL- 1β, IL- 2, IL- 6, IL- 12, and TNF- α were came from BD biosciences (San Jose, CA, USA). ELISA kits for detection of TGF- β was came from RD biotechne (MN, USA). LDH assay kits were obtained from Leagene (Beijing, China). LysoSensor™ Yellow/Blue DND- 160 was purchased from Yeasen Ltd. (Shanghai, China). 4% paraformaldehyde were obtained from Biosharp (Anhui, China). Al₂O₃, TiO₂, ZnO, NiO, In₂O₃, Bi₂O₃, ZrO₂, MoO₃, SnO₂, MgO, Sb₂O₃, CeO₂- 1, CeO₂- 2, Y₂O₃, Dy₂O₃, and Yb₂O₃ were obtained from Aladin (Shanghai, China). The rest of the materials were made in the laboratory.  
+
+## Material characterization  
+
+Morphologies and primary sizes of MeONPs were examined by transmission electron microscopy (TEM) using a Tecnai G2 spirit BioTwin microscope (FEI, Oregon, USA) operated
+
+<--- Page Split --->
+
+at \(120\mathrm{kV}\) . MeONP suspensions ( \(50\mu \mathrm{g / mL}\) in deionized water) were placed on 200- mesh copper grids coated with carbon- coated formvar support film (Ted Pella, Inc., Redding, CA, USA) and air- dried at room temperature. The hydrodynamic diameters and surface charges of MeONP dispersions in water were determined by dynamic light scattering and zeta potential analysis using a Zetasizer Nano ZS90 instrument (Malvern Instruments Corp., UK), as previously described<sup>40</sup>.  
+
+## Examination of metal dissolution in simulated biological fluids  
+
+To investigate the dissolution behavior of MeONPs in phagolysosomal simulated fluid (PSF) and stimulated lung fluid (SLF), MeONPs were dissolved in \(10\mathrm{mL}\) of PSF buffer ( \(142\mathrm{mg / L}\) \(\mathrm{Na_2HPO_4}\) , \(6.65\mathrm{g / LNaCl}\) , \(62\mathrm{mg / LNa_2SO_4}\) , \(29\mathrm{mg / LCaCl_2\cdot H_2O}\) , \(250\mathrm{mg / L}\) glycine, \(8.09\mathrm{g / L}\) potassium phthalate, pH 4.5) or SLF buffer ( \(95\mathrm{mg / LMgCl_2}\) , \(6.019\mathrm{g / LNaCl}\) , \(298\mathrm{mg / LKCl}\) , \(126\mathrm{mg / LNa_2HPO_4}\) , \(63\mathrm{mg / LNa_2SO_4}\) , \(368\mathrm{mg / LCaCl_2\cdot 2H_2O}\) , \(574\mathrm{mg / LCH_3COONa}\) , \(2.604\mathrm{g / LNaHCO_3}\) , \(97\mathrm{mg / L}\) sodium citrate dihydrate, pH 7.4) at a concentration of \(50\mu \mathrm{g / mL}\) with probe sonication at \(32\mathrm{W}\) for \(10\mathrm{s}\) . The resulting MeONP suspensions were incubated for \(24\mathrm{h}\) at room temperature. The supernatants were collected by centrifugation at \(15,000\mathrm{RPM}\) for 10 min and analyzed for metal ion concentrations using inductively coupled plasma- atomic emission spectrometry (ICP- OES DUO 6500, Thermo Scientific, Massachusetts, USA). The percentage of MeONP dissolution was calculated using following equation:  
+
+\[\text{Dissolution \%} = \frac{c_{\mathrm{t}}\times\nu_{\mathrm{t}}}{c_{0}\times\nu_{0}\times R}\times 100\% \quad (1)\]
+
+<--- Page Split --->
+
+where \(\mathrm{Ct}\) and \(\mathrm{C_0}\) (50 \(\mu \mathrm{g / mL}\) ) are the concentrations of metal ions in the supernatant measured by ICP- OES and MeONPs before digestion, respectively; \(\mathrm{Vt}\) and \(\mathrm{V_0}\) are the volumes of digestive solution and MeONP suspension, respectively; R represents the mass ratio of metal elements in each specific MeONP.  
+
+## Assessment of in vitro toxicity  
+
+THP- 1 and BEAS- 2B cells were cultured in RPMI 1640 medium supplemented with \(10\%\) fetal bovine serum (Gemini) and BEpicM (Sciencell), respectively. To assess the effects of MeONPs, THP- 1 cells were primed with \(1 \mu \mathrm{g / mL}\) PMA and seeded in 96- well plates at a density of \(3 \times 10^{4}\) cells/well. BEAS- 2B cells were seeded in plates at a density of \(8 \times 10^{3}\) cells/well. After overnight culture, the cell media were removed and replaced with \(100 \mu \mathrm{L}\) aliquots of fresh medium containing MeONPs at concentrations of 0, 12.5, 25, 50, 100, and \(200 \mu \mathrm{g / mL}\) . The cells were incubated for \(24 \mathrm{~h}\) at \(37^{\circ} \mathrm{C}\) . The supernatants were then collected for detection of LDH release and cytokine production, including TGF- \(\beta 1\) , TNF- \(\alpha\) , IL- 1 \(\beta\) , IL- 2, IL- 6, and MCP- 1. The MeONPs treated cells were incubated with \(120 \mu \mathrm{L}\) of MTS working solution (5 mg/mL) in phenol red- free media for \(2 \mathrm{~h}\) at \(37^{\circ} \mathrm{C}\) to examine mitochondria activity, lysed in \(100 \mu \mathrm{L}\) working solution in ATP assay kit to assess energy metabolism by luminescence, or \(15 \mu \mathrm{g / mL}\) H2DCF- DA in the dark for \(30 \mathrm{~min}\) at \(37^{\circ} \mathrm{C}\) to detect ROS generation using a microplate reader at an excitation wavelength of \(488 \mathrm{~nm}\) and an emission wavelength of \(525 \mathrm{~nm}\) . Meanwhile, lysosomal pH was determined using the Lysosensor Yellow/Blue DND- 160 assay kit (40768ES50, Yishang) according to the manufacturer's instructions. Briefly, the culture media
+
+<--- Page Split --->
+
+in MeONP- treated cells was replaced with \(100~\mu \mathrm{L}\) PBS containing \(1\mu \mathrm{M}\) DND- 160 probes. After incubating at \(37^{\circ}\mathrm{C}\) for \(3\mathrm{min}\) , the cells were washed twice with PBS and detected using a microplate reader at \(384~\mathrm{nm}\) excitation and \(540~\mathrm{nm}\) emission. The relative fold changes of indexes were calculated by following equation:  
+
+\[\mathrm{FC} = \frac{I_{\mathrm{NP}} - I_{\mathrm{BL}}}{I_{\mathrm{Ctrl}} - I_{\mathrm{BL}}} \quad (2)\]  
+
+where \(I_{\mathrm{NP}}\) , \(I_{\mathrm{Ctrl}}\) and \(I_{\mathrm{BL}}\) represent the measured intensity of indexes in MeONP- treated cells, vehicle solution treated cells and blanks, respectively.  
+
+## Assessment of lung fibrosis in mice  
+
+A total of 208 animals were used to determine chronic lung fibrosis induced by 52 different MeONPs, with four animals in each group (n = 4). Female C57Bl/6 mice (8 weeks old) were purchased from Nanjing Peng Sheng Biological Technology (Nanjing, China). Animals were housed in groups of four under standard laboratory conditions (25°C; 60% relative humidity; 12 h light, 12 h dark cycle) and hygiene status (autoclaved food and acidified water) according to Soochow University guidelines for the care and treatment of laboratory animals. All animal experiments were approved by the Ethics Committee of Soochow University. Animals were exposed to MeONPs by an oropharyngeal instillation method. Briefly, MeONPs were suspended in PBS at \(1\mathrm{mg / mL}\) by a probe sonication (32 W) for 10 s. The animals were anesthetized by intraperitoneal injection of sodium pentobarbital (200 mg/kg) in a total volume of \(100~\mathrm{uL}\) . The anesthetized animals were held in a vertical position for the pulmonary aspiration of MeONP suspension (with a dose of \(2\mathrm{mg / kg}\) ) at the back of the tongue. Animals
+
+<--- Page Split --->
+
+included in vehicle and positive controls received 50 uL of PBS and 2 mg/kg quartz, respectively. The mice were exposed once a week for three weeks and sacrificed on day 90 by overdose of sodium pentobarbital (400 mg/kg). Bronchoalveolar lavage fluids (BALFs) and lung tissues were collected according to previous studies<sup>40</sup>. Briefly, the trachea was cannulated and then the lungs were gently lavaged 3 times with 1 mL of sterile PBS to obtain BALF. Aliquots of 50 uL BALF were used to measure TGF- \(\beta 1\) levels by the ELISA kits (BDLISA, China). Lung tissues were collected and stained Masson’s Trichrome staining according to a standard protocol<sup>41</sup>. The collagen deposition levels of the 203 mice were estimated according to the method reported by Ashcroft et al<sup>42</sup>.  
+
+## Construction of the MFF models  
+
+We developed classification models for lung fibrosis using eight different machine learning algorithms, namely C4.5 decision tree (C4.5), random forest (RF), logistic regression (LGR), k- nearest neighbor (kNN), locally weighted learning (LWL), support vector machine (SVM), Bayesnet, and decision table (DT). These classifiers represented various categories of supervised classifiers such as trees, Bayes, and rules. To validate the models, we randomly split the dataset into training and test sets, with 156 data points from 40 metal oxide nanoparticles (MeONPs) in the training set and 47 data points from 12 MeONPs in the test set. To avoid information leaking, we included all four data points of the same MeONP in either the training or test set. The criteria for classifying data points of MeONPs as fibrogenic potential were \(\mathrm{FC}_{\mathrm{TGF - }\beta 1} \geq 2\) or Ashcroft score \(\geq 2\) .
+
+<--- Page Split --->
+
+To optimize the model's performance, we exploited a 10- fold- cross- validation procedure based on a grid search to determine the best parameters in the machine learning methods. We performed all procedures using Weka software (Ver 3.8.5). We applied 10- fold cross- validation on the training set to assess the prediction accuracy of the models. External validation was performed on the test set. The performances of models were evaluated based on true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN). Model performance was evaluated using five metrics: sensitivity (SE) = TP/(TP + FN), specificity (SP) = TN/(TN + FP), overall predictive accuracy (ACC) = (TP + TN)/(TP + FP + TN + FN), F1 score = 2×SE×SP/(SE+SP), and Matthews' correlation coefficient (MCC) = (TP×TN - FP×FN)/√(TP+FP)(TP+FN)(TN+FP)(TN+FN). MCC ranges from -1 to +1, with extreme values of -1 and +1 in the case of perfect misclassification and perfect classification, respectively. We also calculated AUC by plotting the TP rate versus the FP rate at various threshold settings. We considered the performance of a model "excellent" if AUC ≥ 0.9, "very poor" if AUC < 0.6, "poor" if 0.7 > AUC ≥ 0.6, "fair" if 0.8 > AUC ≥ 0.7, and "good" if 0.9 > AUC ≥ 0.8.  
+
+We characterized the applicability domain of the prediction model using a descriptor standardization approach. Briefly, a data point was considered an outlier if all normalized descriptors for the data point were greater than 3; otherwise, it was a non- outlier. We generated a similarity network of the in vivo data points using Gephi software (V 0.9). We represented
+
+<--- Page Split --->
+
+the data points as nodes and colored them according to the chemical composition of MeONPs. We used the size of the circle to represent the fibrosis level (release of TGF- \(\beta 1\) ) in the lung. The thickness of the lines reflected the strength of the correlations between the data points. Tight and sparse connections indicated high and low homogeneity of the data points in descriptor spaces, respectively.  
+
+## Acknowledgements  
+
+This work was supported by the National Key Research and Development Program (2022YFC3902104, 2022YFE0124000) of China, National Natural Science Foundation (22176023, 21976126) of China, the Natural Science Foundation of Jiangsu Province (BK20211545), the Fundamental Research Funds for the Central Universities (DUT22QN216) and the Project of National Center for International Research on Intelligent Nano- Materials and Detection Technology in Environmental Protection, Soochow University (No. SDGH2202). We thank T. A. Patterson from National Center for Toxicological Research, U.S. Food and Drug Administration for his contribution to the improvement of the manuscript. This article reflects the views of the authors and does not necessarily reflect those of the U.S. Food and Drug Administration.  
+
+## Author contributions  
+
+X. Li, X. Cai and R. Li conceived the idea and designed the experiments. H. Yang established the predictive model. J. Cao performed most of the cell and animal experiments. Q. Xie performed ELISA assay. Q. Yang contributed to the histology assay of animal lungs. X. Liu performed the characterization of MeONPs. The writing of the manuscript was led by R. Li, X. Li, and X. Cai with participations from J. Chen, and H. Hong.  
+
+## Competing interests  
+
+The authors declare no competing interests.
+
+<--- Page Split --->
+
+## Additional information  
+
+The online version contains supplementary material available at https://github.com/huangyang  
+
+2023/NILFPv1.0.0/releases/download/NILFPv1.0.0/NILFP.v1.0.0.zip.
+
+<--- Page Split --->
+
+## References  
+
+1. Nanotechnology Products Database (NPD). [cited 2023 March 1st]Available from: https://product.statnano.com2. Kokot H, Kokot B, Sebastijanovic A, Voss C, Podlipec R, Zawilska P, et al. Prediction of chronic inflammation for inhaled particles: the impact of material cycling and quarantining in the lung epithelium. Advanced Materials 2020, 32(47): 2003913.3. Garcia-Mouton C, Hidalgo A, Cruz A, Perez-Gil J. The lord of the lungs: the essential role of pulmonary surfactant upon inhalation of nanoparticles. European Journal of Pharmaceutics and Biopharmaceutics 2019, 144: 230-243.4. Liu S, Xia T. Continued efforts on nanomaterial-environmental health and safety is critical to maintain sustainable growth of nanoindustry. Small 2020, 16(21): 2000603.5. Wu Z, Shi P, Lim HK, Ma Y, Setyawati MI, Bitounis D, et al. Inflammation increases susceptibility of human small airway epithelial cells to pneumonic nanotoxicity. Small 2020, 16(21): 2000963.6. Zhang H, Ji Z, Xia T, Meng H, Low-Kam C, Liu R, et al. Use of Metal Oxide Nanoparticle Band Gap To Develop a Predictive Paradigm for Oxidative Stress and Acute Pulmonary Inflammation. ACS Nano 2012, 6(5): 4349-4368.7. Cai X, Dong J, Liu J, Zheng H, Kaweeterawat C, Wang F, et al. Multi-hierarchical profiling the structure-activity relationships of engineered nanomaterials at nano-bio interfaces. Nature Communications 2018, 9: 4416.8. Li R, Ji Z, Chang CH, Dunphy DR, Cai X, Meng H, et al. Surface interactions with compartmentalized cellular phosphates explain rare earth oxide nanoparticle hazard and provide opportunities for safer design. ACS Nano 2014, 8(2): 1771-1783.9. Zhu M, Du L, Zhao R, Wang HY, Zhao Y, Nie G, et al. Cell-penetrating nanoparticles activate the inflammasome to enhance antibody production by targeting microtubule-associated protein 1-light chain 3 for degradation. ACS Nano 2020, 14(3): 3703-3717.10. Wu Z, Shi P, Lim HK, Ma Y, Setyawati MI, Bitounis D, et al. Inflammation Increases Susceptibility of Human Small Airway Epithelial Cells to Pneumonic Nanotoxicity. Small 2020, 16(21).11. Huang Y, Li X, Xu S, Zheng H, Zhang L, Chen J, et al. Quantitative structure-activity relationship models for predicting inflammatory potential of metal oxide nanoparticles. Environmental Health Perspectives 2020, 128(6): 67010.12. Barosova H, Maione AG, Septiadi D, Sharma M, Haeni L, Balog S, et al. Use of EpiAlveolar lung model to predict fibrotic potential of multiwalled carbon nanotubes. ACS Nano 2020, 14(4): 3941-3956.13. Wang X, Mansukhani ND, Guiney LM, Lee J-H, Li R, Sun B, et al. Toxicological profiling of highly purified metallic and semiconducting single-walled carbon nanotubes in the rodent lung and E. coli. ACS Nano 2016, 10(6): 6008-6019.14. Jiang W, Wang X, Osborne OJ, Du Y, Chang CH, Liao Y-P, et al. Pro-inflammatory and pro-fibrogenic effects of ionic and particulate arsenide and indium-containing semiconductor materials in the murine lung. ACS Nano 2017, 11(2): 1869-1883.15. Rahman L, Williams A, Gelda K, Nikota J, Wu D, Vogel U, et al. 21st century tools for nanotoxicology: transcriptomic biomarker panel and precision-cut lung slice organ mimic
+
+<--- Page Split --->
+
+system for the assessment of nanomaterial- induced lung fibrosis. Small 2020, 16(36): 2000272.  
+
+16. Zhu M, Feng W, Wang B, Wang T, Gu Y, Wang M, et al. Comparative study of pulmonary responses to nano- and submicron-sized ferric oxide in rats. Toxicology 2008, 247(2-3): 102-111.  
+
+17. Gao M, Wang Z, Zheng H, Wang L, Xu S, Liu X, et al. Two-dimensional tin selenide (SnSe) Nanosheets capable of mimicking key dehydrogenases in cellular metabolism. Angewandte Chemie-International Edition 2020, 59(9): 3618-3623.  
+
+18. Schraufnagel DE. The health effects of ultrafine particles. Experimental and Molecular Medicine 2020, 52(3): 311-317.  
+
+19. Oberdorster C, Maynard A, Donaldson K, Castranova V, Fitzpatrick J, Ausman K, et al. Principles for characterizing the potential human health effects from exposure to nanomaterials: elements of a screening strategy. Particle and Fibre Toxicology 2005, 2(1): 8.  
+
+20. Anonymous. The risks of nanomaterial risk assessment. Nature Nanotechnology 2020, 15(3): 163-163.  
+
+21. Choi JY, Ramachandran G, Kandlikar M. The impact of toxicity testing costs on nanomaterial regulation. Environmental Science & Technology 2009, 43(9): 3030-3034.  
+
+22. Yu F, Wei C, Deng P, Peng T, Hu X. Deep exploration of random forest model boosts the interpretability of machine learning studies of complicated immune responses and lung burden of nanoparticles. Science Advances 2021, 7(22).  
+
+23. Ji Z, Guo W, Wood EL, Liu J, Sakkiah S, Xu X, et al. Machine learning models for predicting cytotoxicity of nanomaterials. Chemical Research in Toxicology 2022, 35(2): 125-139.  
+
+24. Roy J, Roy K. Assessment of toxicity of metal oxide and hydroxide nanoparticles using the QSAR modeling approach. Environmental Science-Nano 2021, 8(11): 3395-3407.  
+
+25. Labouta HI, Asgarian N, Rinker K, Cramb DT. Meta-Analysis of Nanoparticle Cytotoxicity via Data-Mining the Literature. Acs Nano 2019, 13(2): 1583-1594.  
+
+26. Oh E, Liu R, Nel A, Gemill KB, Bilal M, Cohen Y, et al. Meta-analysis of cellular toxicity for cadmium-containing quantum dots. Nature Nanotechnology 2016, 11(5): 479.  
+
+27. Cao J, Pan Y, Jiang Y, Qi R, Yuan B, Jia Z, et al. Computer-aided nanotoxicology: risk assessment of metal oxide nanoparticles via nano-QSAR. Green Chemistry 2020, 22(11): 3512-3521.  
+
+28. Huang Y, Li X, Cao J, Wei X, Li Y, Wang Z, et al. Use of dissociation degree in lysosomes to predict metal oxide nanoparticle toxicity in immune cells: Machine learning boosts nano-safety assessment. Environment International 2022, 164: 107258.  
+
+29. Yan XL, Sedykh A, Wang WY, Yan B, Zhu H. Construction of a web-based nanomaterial database by big data curation and modeling friendly nanostructure annotations. Nature Communications 2020, 11(1): 2519.  
+
+30. Ban Z, Yuan P, Yu F, Peng T, Zhou Q, Hu X. Machine learning predicts the functional composition of the protein corona and the cellular recognition of nanoparticles. Proceedings of the National Academy of Sciences of the United States of America 2020, 117(19): 10492-10499.  
+
+31. Friedersdorf LE, Bjorkland R, Klaper RD, Sayes CM, Wiesner MR. Fifteen years of nanoEHS research advances science and fosters a vibrant community. Nature Nanotechnology 2019, 14(11): 996-998.
+
+<--- Page Split --->
+
+32. Wyrzykowska E, Mikolajczyk A, Lynch I, Jeliazkova N, Kochev N, Sarimveis H, et al. Representing and describing nanomaterials in predictive nanoinformatics. Nature Nanotechnology 2022, 17(9): 924-932.33. Tung Xuan T, Choi J-S, Jeon H, Byun H-G, Yoon T-H, Kim J. Quasi-SMILES-based nano-quantitative structure-activity relationship model to predict the cytotoxicity of multiwalled carbon nanotubes to human lung cells. Chemical Research in Toxicology 2018, 31(3): 183-190.34. Forest V, Hochepied J-F, Leclerc L, Trouve A, Abdelkebir K, Sarry G, et al. Towards an alternative to nano-QSAR for nanoparticle toxicity ranking in case of small datasets. Journal of Nanoparticle Research 2019, 21(5): 95.35. Gajewicz A, Schaeublin N, Rasulev B, Hussain S, Leszczynska D, Puzyn T, et al. Towards understanding mechanisms governing cytotoxicity of metal oxides nanoparticles: Hints from nano-QSAR studies. Nanotoxicology 2015, 9(3): 313-325.36. Puzyn T, Rasulev B, Gajewicz A, Hu X, Dasari TP, Michalkova A, et al. Using nano-QSAR to predict the cytotoxicity of metal oxide nanoparticles. Nature Nanotechnology 2011, 6(3): 175-178.37. Mu Y, Wu F, Zhao Q, Ji R, Qie Y, Zhou Y, et al. Predicting toxic potencies of metal oxide nanoparticles by means of nano-QSARs. Nanotoxicology 2016, 10(9): 1207-1214.38. Zhong S, Zhang K, Bagheri M, Burken JG, Gu A, Li B, et al. Machine learning: new ideas and tools in environmental science and engineering. Environmental Science & Technology 2021, 55(19): 12741-12754.39. Cai XM, Liu X, Jiang J, Gao M, Wang WL, Zheng HZ, et al. Molecular mechanisms, characterization methods, and utilities of nanoparticle biotransformation in nanosafety assessments. Small 2020, 16(36): 19.40. Huang Y, Li X, Xu S, Zheng H, Zhang L, Chen J, et al. Quantitative Structure-Activity Relationship Models for Predicting Inflammatory Potential of Metal Oxide Nanoparticles. Environ Health Perspect 2020, 128(6): 67010.41. Foot NC. The Masson Trichrome Staining Methods in Routine Laboratory Use. Stain technology 1933, 8(3): 101-110.42. Ashcroft T, Simpson JM, Timbrell V. Simple method of estimating severity of pulmonary fibrosis on a numerical scale. Journal of Clinical Pathology 1988, 41(4): 467-470.
+
+<--- Page Split --->
+
+
+Table 1. Multimodality input parameters   
+
+<table><tr><td>Type</td><td>Descriptors</td><td>Number</td></tr><tr><td rowspan="7">Periodic parameters</td><td>Molecular weight</td><td>1</td></tr><tr><td>Electronegativity of metal atom</td><td>1</td></tr><tr><td>Metal atom number</td><td>1</td></tr><tr><td>Oxygen atom number</td><td>1</td></tr><tr><td>Cation charge</td><td>1</td></tr><tr><td>Periodic number of metal atom</td><td>1</td></tr><tr><td>Group number of metal atom</td><td>1</td></tr><tr><td rowspan="4">Physicochemical properties in biological fluids</td><td>Atomic ratio of metal and oxygen</td><td>1</td></tr><tr><td>Hydrodynamic size</td><td>1</td></tr><tr><td>Zeta-potential</td><td>1</td></tr><tr><td>Dissolution in PSF</td><td>1</td></tr><tr><td rowspan="2">MeONP-membrane interaction</td><td>Dissolution in SLF</td><td>1</td></tr><tr><td>LDH release in THP-1 cells</td><td>5</td></tr><tr><td rowspan="2">MeONP-lysosome interaction</td><td>LDH release in BEAS-2B cells</td><td>5</td></tr><tr><td>pH changes in lysosomes</td><td>5</td></tr><tr><td rowspan="4">Impact on mitochondria activity</td><td>NADH level in THP-1 cells</td><td>5</td></tr><tr><td>NADH level in BEAS-2B cells</td><td>5</td></tr><tr><td>ATP level in THP-1 cells</td><td>5</td></tr><tr><td>ATP level in BEAS-2B cells</td><td>5</td></tr><tr><td rowspan="2">Impact on redox homeostasis</td><td>ROS generation in THP-1 cells</td><td>5</td></tr><tr><td>ROS generation in BEAS-2B cells</td><td>5</td></tr><tr><td rowspan="6">Impact on cell–cell communications</td><td>TNF-α release in THP-1 cells</td><td>5</td></tr><tr><td>IL-1β release in THP-1 cells</td><td>5</td></tr><tr><td>IL-2 release in THP-1 cells</td><td>5</td></tr><tr><td>IL-6 release in THP-1 cells</td><td>5</td></tr><tr><td>MCP-1 release in THP-1 cells</td><td>5</td></tr><tr><td>TGF-β1 release in BEAS-2B cells</td><td>5</td></tr><tr><td colspan="2">Total descriptors</td><td>87</td></tr></table>
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Figure 1 Interfaces of nano-bio interactions in the lungs. </center>  
+
+A) Schematic workflow of multimodal feature fusion (MFF) modeling. The figure depicts the travel path of nanoparticles in lungs where inhaled nanoparticles may deposit in alveoli and interact with macrophages and epithelial cells, leading to injury of various cellular components such as cell membranes, lysosomes, mitochondria, and other cellular components. Descriptive features at these nano-bio interfaces and fibrogenic indexes in vivo were collected to develop predictive models using machine learning methods. The established in silico model was further validated by animal tests. B) Confocal and C) TEM imaging of MeONPs in cells. THP-1 cells exposed to \(12.5 \mu \mathrm{g / mL}\) FITC-labeled or pristine MeONPs were collected. The fixed cells were stained with DAPI (blue) and Alexa Fluor™ 594 conjugated with WGA/anti-LAMP1 (red) for confocal imaging. Arrow and L indicate the MeONP and lysosome, respectively.
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Figure 2 Effects of representative MeONPs in cells and mouse lungs. </center>  
+
+A) Heatmap displaying the impacts of 10 representative MeONPs on THP-1 and BEAS-2B cells by detecting LDH leakage, ROS generation, NADH content, lysosomal pH change, ATP production, and cytokine release.
+
+<--- Page Split --->
+
+THP- 1 cells were exposed to 0, 12.5, 25, 50, 100, and \(200\mu \mathrm{g / mL}\) MeONPs for \(24\mathrm{h}\) , followed by LDH and cytokine measurement in supernatants and ROS, NADH, ATP and lysosomal pH detection in cells. BEAS- 2B cells were exposed to 0, 12.5, 25, 50, 100, and \(200\mu \mathrm{g / mL}\) MeONPs for \(24\mathrm{h}\) , followed by TGF- \(\beta 1\) detection in cells. The values of these descriptors in MeONP treatments were compared with control cells. The resulting ratios were expressed as fold changes (FCs) in the heatmap. B) Schematic illustration of MeONP instillation in mice. Mice were oropharyngeally administered with \(50\mu \mathrm{L}\) PBS (vehicle control), \(2\mathrm{mg / kg}\) MeONPs, and quarts (positive control) three times a week. The animals were sacrificed at day 90 to collect BALF and lung tissues for further examinations. C) TGF- \(\beta 1\) release in BALFs and Ashcroft score of stained lung sections. D) Masson’s trichrome staining of lung tissues exposed to representative MeONPs. BALFs were collected to measure TGF- \(\beta 1\) by ELISA \((n = 4)\) . \*p<0.05, \*\*p<0.01 and \*\*\*p<0.001 compared to the vehicle control by one- way ANOVA.
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Figure 3 Performance of models constructed using eight machine learning algorithms. </center>  
+
+The y- axis represents the values of the performance metrics. RF: random forest; C4.5: C4.5 decision tree; LWL: locally weighted learning; \(k\) - NN: k nearest neighbor; SVM: support vector machine; DT: decision table; LGR: logistic regression. \(ACC\) : overall predictive accuracy; \(MCC\) : Matthews correlation coefficient; \(SE\) : sensitivity; \(SP\) : specificity; \(AUC\) : the area under the receiver operating characteristic curve; \(F1\) : F1 score.
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Figure 4 Experimental validation and diversity analysis of the established model. </center>  
+
+A) Experimental and predicted results of lung fibrosis in the experimental validation set are depicted in the confusion matrix, with each node representing a data point. 
+B) The similarity network of the in vivo data points. Each node represents a data point which is colored according to the chemical composition of MeONPs. The size of a node indicates the fibrosis level in the lung, while the thickness of a line represents the strength of correlation between the two connected nodes. Tight and sparse connections denote high and low homogeneity of nodes for descriptor spaces, respectively.
+
+<--- Page Split --->
+![](images/Figure_5.jpg)
+  
+
+![PLACEHOLDER_30_1]
+
+<center>Figure 5 Identification of key descriptors for mechanism interpretation. </center>  
+
+A) SHAP summary plots displaying the effects of features and their values on the prediction. The y-axis of each plot contains the features included in the model sorted from the most (top) to least (bottom) important. The x-axis depicts the SHAP value, with each point referring to a SHAP value associated with a value of a certain feature. The color of the point displays whether the feature value is high (pink) or low (blue). B) Feature importance in the RF model. C) Schematic image of the proposed in chemico/in vitro-in vivo extrapolation of lung fibrosis. The key determinants of MeONP-induced lung fibrosis include IL-1β, TGF-β1, metabolic activity, hydrodynamic size, zeta potential, and metal ion release in PSF.
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+SupplementalExcelFile.xlsx SupportingInformation.pdf
+
+<--- Page Split --->

preprint/preprint__0fadf43a852d9e8dc0c5812bfff32104d323571f1b41834242e5098c378037ec/preprint__0fadf43a852d9e8dc0c5812bfff32104d323571f1b41834242e5098c378037ec_det.mmd

@@ -0,0 +1,367 @@
+<|ref|>title<|/ref|><|det|>[[44, 108, 890, 207]]<|/det|>
+# Multimodal Feature Fusion Machine Learning for Predicting Chronic Injury Induced by Engineered Nanomaterials  
+
+<|ref|>text<|/ref|><|det|>[[44, 230, 130, 248]]<|/det|>
+Xuehua Li  
+
+<|ref|>text<|/ref|><|det|>[[55, 258, 279, 275]]<|/det|>
+lixuehua@dlut.edu.cn  
+
+<|ref|>text<|/ref|><|det|>[[52, 303, 690, 323]]<|/det|>
+Dalian University of Technology https://orcid.org/0000- 0003- 3924- 0987  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 364, 103, 381]]<|/det|>
+## Article  
+
+<|ref|>text<|/ref|><|det|>[[44, 401, 900, 443]]<|/det|>
+Keywords: Nanotoxicity, lung fibrosis, predictive toxicology, structure- activity relationship, nano- bio interface.  
+
+<|ref|>text<|/ref|><|det|>[[44, 461, 339, 480]]<|/det|>
+Posted Date: November 3rd, 2023  
+
+<|ref|>text<|/ref|><|det|>[[44, 499, 475, 518]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 3479434/v1  
+
+<|ref|>text<|/ref|><|det|>[[44, 537, 916, 579]]<|/det|>
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+<|ref|>text<|/ref|><|det|>[[44, 598, 535, 617]]<|/det|>
+Additional Declarations: There is NO Competing Interest.  
+
+<|ref|>text<|/ref|><|det|>[[42, 653, 930, 696]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on March 20th, 2025. See the published version at https://doi.org/10.1038/s41467- 025- 58016- w.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[70, 90, 883, 154]]<|/det|>
+Multimodal Feature Fusion Machine Learning for Predicting Chronic Injury Induced by Engineered Nanomaterials  
+
+<|ref|>text<|/ref|><|det|>[[66, 167, 886, 900]]<|/det|>
+Yang Huang \(^{1,4\#}\) , Jiayu Cao \(^{2\#}\) , Xuehua Li \(^{1*}\) , Qing Yang \(^{2}\) , Qianqian Xie \(^{3}\) , Xi Liu \(^{3}\) , Xiaoming Cai \(^{2*}\) , Jingwen Chen \(^{1}\) , Huixiao Hong \(^{5}\) , Ruibin Li \(^{3*}\) \(^{1}\) Key Laboratory of Industrial Ecology and Environmental Engineering (MOE), Dalian Key Laboratory on Chemicals Risk Control and Pollution Prevention Technology, School of Environmental Science and Technology, Dalian University of Technology, Dalian 116024, China \(^{2}\) School of Public Health, Soochow University, Suzhou, Jiangsu, 215123, China \(^{3}\) State Key Laboratory of Radiation Medicine and Protection, School of Radiation Medicine and Protection, Soochow University, Suzhou, Jiangsu, 215123, China \(^{4}\) School of Chemistry and Materials Science, Ludong University, Yantai, 264025, China \(^{5}\) National Center for Toxicological Research, U.S. Food and Drug Administration, Jefferson, AR 72079, USA #Equal contribution \(^{*}\) Address correspondence to Dr. Xuehua Li Dalian University of Technology, Linggong Road 2, Dalian 116024, P R China Email: lixuehua@dlut.edu.cn Dr. Xiaoming Cai 199 Ren'ai Road, 401 Building, Suzhou, 215123, Jiangsu, the People's Republic of China Email: xmcai@suda.edu.cn & Dr. Ruibin Li 199 Ren'ai Road, 401 Building, Suzhou, 215123, Jiangsu, the People's Republic of China Email: liruibin@suda.edu.cn
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 94, 207, 113]]<|/det|>
+## Abstract  
+
+<|ref|>text<|/ref|><|det|>[[110, 128, 885, 660]]<|/det|>
+Concerns regarding chronic injuries (e.g., fibrosis and carcinogenesis) induced by nanoparticles raised public health concerns and needs to be rapidly assessed in hazard identification. Although in silico analysis is commonly used for risk assessment of chemicals, predicting chronic in vivo nanotoxicity remains challenging due to the complex interactions at multiple nano- bio interfaces. Herein, we developed a multimodal feature fusion analysis framework to predict the fibrogenic potential of metal oxide nanoparticles (MeONPs). Eighty- seven features derived from multiple MeONP- lung interfaces were used to develop a machine learning- based predictive framework. We identified viability damage and cytokine (IL- 1β and TGF- β1) production in macrophages and epithelial cells as key events that are closely associated with particle size, surface charge, and interactions with lysosomes. Experimental validations showed that the developed in silico model had 85% accuracy in predicting MeONP- induced lung fibrosis. The heterogeneity distribution of data points indicated good applicability of the predictive model. Our findings demonstrate the potential usefulness of this predictive model for risk assessment of nanomaterials and in assisting regulatory decision making.  
+
+<|ref|>text<|/ref|><|det|>[[115, 713, 881, 770]]<|/det|>
+Keywords: Nanotoxicity, lung fibrosis, predictive toxicology, structure- activity relationship, nano- bio interface.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 93, 225, 113]]<|/det|>
+## Main Text  
+
+<|ref|>text<|/ref|><|det|>[[110, 123, 886, 816]]<|/det|>
+More than 10,000 nanoproducts have been produced and used in various industries worldwide<sup>1</sup>. During the lifecycle of some nanoproducts, nanoparticles may be released to form aerosols and travel in the air<sup>2</sup>. Inhalation is the major exposure route for airborne nanoparticles and can cause respiratory injury in mammals<sup>3- 4</sup>. While certain engineered nanomaterials (ENMs) such as ZnO, CuO, Au, and Ag may produce acute lung inflammation via reactive oxygen species (ROS) generation<sup>5- 6</sup>, inflammasome activation<sup>7- 9</sup>, or proinflammatory cytokine release<sup>7,8, 10- 11</sup>, other ENMs such as carbon nanotubes and metal oxides may induce chronic respiratory toxicity such as lung fibrosis<sup>12- 16</sup> and carcinogenesis<sup>17- 19</sup>, which are permanent and irreversible lung injuries. The chronic adverse outcomes of ENMs have raised substantial product safety concerns and led to strict regulatory requirements on utilizing nanotechnology. For instance, carbon nanotubes have been added to the Substitute It Now (SIN) list due to their carcinogenic potential, which restricts their use in Europe<sup>20</sup>. Since the number of ENMs is rapidly growing and each nanomaterial may have various derivatives with differing size, shape, surface chemistry, crystallinity, and so on, a large number of animals would need to be sacrificed to adequately assess the risk of ENMs. It was estimated that the in vivo toxicity assessment of the first generation of ENMs in commercial use may take up to 50 years and cost billions of dollars<sup>21</sup>. To address this challenge, in silico predictive models have been developed to reduce experimental costs.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 90, 885, 775]]<|/det|>
+In recent years, there have been several attempts to predict the in vitro toxicity of nanoparticles<sup>11, 22- 30</sup>. ENM- induced cytotoxicity, including cell death and inflammatory effects, is often caused by a molecular initiating event (MIE) that is highly dependent on the physicochemical properties of ENMs<sup>11, 25- 26, 28</sup>. To establish in silico models, physicochemical properties have been extensively exploited to identify predictive features. For instance, Labouta et al. found that the cytotoxicity of ENMs could be primarily predicted based on their material chemistry, followed by nanoparticle size and concentration, cell type, and cytotoxicity biomarkers<sup>25</sup>. The surface chemistry, lipophilicity, zeta potential, shape, and size of ENMs also impact their uptake and toxicology behaviors<sup>31- 32</sup>. In addition, in silico analysis has been widely used to explore the quantitative structure- activity relationships (QSAR) of ENMs in various cell types, such as human lung cells<sup>27, 33</sup>, RAW 264.7 cells<sup>24, 34</sup>, human keratinous cells (HaCaT)<sup>35</sup>, THP- 1 cells<sup>11</sup>, and Escherichia coli<sup>36- 37</sup> for hazard ranking. However, there is no reliable model to predict the chronic respiratory toxicity of ENMs in vivo such as lung fibrosis. The prediction of chronic toxicity is a major challenge in nanotoxicity studies, as it involves multiple nano- bio interactions that cannot be entirely mirrored at one specific nano- bio interface. Taking lung fibrosis as an example, its pathogenic progress involves the interactions of nanoparticles with lung lining fluids and multiple subcellular organelles such as the plasma membrane, lysosome, mitochondria, and cytoplasmic components in immune/epithelial cells.  
+
+<|ref|>text<|/ref|><|det|>[[113, 829, 882, 885]]<|/det|>
+Herein, we aimed to establish a predictive framework for pulmonary fibrosis induced by metal oxide nanoparticles (MeONPs). Based on the biological fate of MeONPs in the lungs, we
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 92, 884, 462]]<|/det|>
+prepared a library of 52 MeONPs and collected the potential predictive features at multiple interfaces between MeONPs and biological contexts such as membranes, lysosomes, mitochondria, and other cytoplasmic components (Figure 1A). We acquired a total of 87 multimodal features and two fibrogenic indexes, which were then subjected to machine learning modeling. We thoroughly evaluated performance of the developed predictive models using overall predictive accuracy (ACC), Matthews correlation coefficient (MCC), sensitivity (SE), specificity (SP), area under the receiver operating characteristic curve (AUC), and \(F1\) score. The random forest (RF) model performed the best and was further experimentally validated by five MeONPs. Our study presents the first in silico framework for decoding the in vitro to in vivo extrapolation (IVIVE) of ENM- induced lung fibrosis.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 519, 511, 540]]<|/det|>
+## Construction of a multimodal database  
+
+<|ref|>text<|/ref|><|det|>[[111, 557, 884, 848]]<|/det|>
+A high- quality database is essential for building reliable IVIVE models and identifying key events involved in the pathogenic process of lung fibrosis<sup>38</sup>. However, there is a lack of such a database to define the behaviors of MeONPs and their biological effects at multiple nano- bio interfaces. Therefore, we constructed a reliable database of 52 MeONPs by characterizing their physicochemical properties, examining their interactions with biological fluids and subcellular organelles, and assessing their fibrogenic effects in mouse lungs by detecting collagen deposition in lung tissues and TGF- \(\beta 1\) release in bronchoalveolar lavage fluid (BALF). The sources of the 52 MeONPs are listed in Table S1.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[110, 90, 885, 892]]<|/det|>
+Input parameters. Nano- bio interactions occur mainly in the pulmonary alveoli which consist of lung lining fluid, macrophages, and epithelial cells. Initially, inhaled MeONPs may interact with the lining fluid, altering their surface chemistry and dispersion states and dissolving to release metal ions. We therefore examined the behavior of MeONPs in simulated lung fluid (SLF) (Table S1). Then, the MeONPs may interact with lung cells. To visualize the intracellular path of MeONPs, \(\mathrm{Bi}_2\mathrm{O}_3\) and \(\mathrm{Fe}_2\mathrm{O}_3\) were selected to conjugate with fluorescein isothiocyanate (FITC) for fluorescence microscopy imaging. As shown in Figure 1B, the labeled MeONPs showed a time- dependent cascaded distribution pattern from the extracellular matrix into the membrane, lysosome, and cytoplasm. At 8- 16 h, the MeONPs overlapped nicely with lysosomes (Figure 1C). Considering the acidic and enzymatic traits of lysosomes, we also examined the dissolution of MeONPs in phagolysosomal simulated fluid (PSF) (Table S1). Based on the distribution features of MeONPs, we examined the impacts of MeONPs on the cell membrane, lysosome, mitochondria activity, energy production, and redox homeostasis by detecting lactate dehydrogenase (LDH) leakage, lysosomal pH change, nicotinamide adenine dinucleotide hydride (NADH) content, adenosine triphosphate (ATP) level, and ROS generation, respectively. These interactions may lead to cytokine release and affect cell- cell communications such as the recruitment of immune cells and proliferation of profibrogenic cells. We measured the release of pro- inflammatory cytokines (TNF- \(\alpha\) , IL- 1 \(\beta\) , IL- 2, IL- 6, MCP- 1) in THP- 1 cells and growth factor (TGF- \(\beta 1\) ) in BEAS- 2B cells. For a snapshot view of the results, the measured levels of descriptors in cells incubated with 0- 200 \(\mu \mathrm{g / mL}\) MeONPs were ranked into three levels, leading to a visual display where high, moderate and negligible effects
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 92, 884, 388]]<|/det|>
+are represented by red, yellow and blue colors, respectively. Figures 2A and S1 show the effects of ten representative MeONPs in THP- 1 and BEAS- 2B cells, respectively, while the others are listed in Table S2. Fifteen branch events were quantitatively measured for the 52 MeONPs at five different exposure concentrations (12.5, 25, 50, 100, 200 \(\mu \mathrm{g / mL}\) ), resulting in 75 input parameters (Table S2). Additionally, four descriptors were collected from the interactions between MeONPs and two biological media (SLF and PSF), and eight descriptors were acquired from the periodic table (Table S1). These data allowed us to establish a multimodal database consisting of 87 input parameters in total (Table 1).  
+
+<|ref|>text<|/ref|><|det|>[[111, 420, 884, 909]]<|/det|>
+In vivo toxicology endpoints. Mice were exposed to the 52 MeONPs via oropharyngeal aspiration to collect toxicological endpoints in vivo for assess their fibrogenic effects. Figure 2B illustrates the workflow of the animal experiments. The animals were subjected to three exposures to MeONP suspensions during the first three weeks and were sacrificed on day 90 to collect BALF and lung tissues for further examination. In the pathogenic process of lung fibrosis, active TGF- \(\beta 1\) plays a critical role in promoting fibroblast proliferation to secrete collagens. We, therefore, quantified the release of active TGF- \(\beta 1\) in BALF by ELISA and visualized collagen deposition by Masson’s trichrome staining of lung sections. Collagen staining images acquired from 203 whole slide images of lung sections were assessed by the Ashcroft score, a widely used index to rank pulmonary fibrosis (Figure 2C). Figure 2D shows the expression levels of TGF- \(\beta 1\) in BALF and the Ashcroft score of mouse lungs exposed to 10 representative MeONPs. The fold change of TGF- \(\beta 1\) (FCTGF- \(\beta 1\) ) and collagen staining images of the 52 MeONPs are shown in Table S3 and Figure S2, respectively. Among the 52 tested
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[110, 92, 884, 580]]<|/det|>
+MeONPs, 23 (Fe2O3- 4, TiO2- r- 1, TiO2- r- 2, TiO2- a- 1, TiO2- a- 2, TiO2- a- 3, \(\alpha\) - Al2O3, \(\gamma\) - Al2O3- 1, \(\gamma\) - Al2O3- 2, Dy2O3, NiO- 1, In2O3- 1, In2O3- 2, Tm2O3, Tb2O3, Bi2O3, Sm2O3, Y2O3, Yb2O3, Sb2O3, SnO2, CuO- 3 and ZnO- 1) consistently induced high levels of TGF- \(\beta 1\) with \(\mathrm{FC}_{\mathrm{TGF - }\beta 1}\geq 2\) and significant collagen deposition, 11 (CeO2- 2, \(\alpha\) - MnO2- 1, CuO- 2, CuO- 4, NiO- 2, MgO, MoO3, ZnO- 2, ZnO- 3, Fe2O3- 2 and Fe2O3- 3) had a moderate effect, and 18 (CeO2- 1, CeO2- 3, Co3O4- 1, Co3O4- 2, Co3O4- 3, Co3O4- 4, Co3O4- 5, \(\alpha\) - MnO2- 2, \(\alpha\) - MnO2- 3, Eu2O3- 1, Eu2O3- 2, Cr2O3, Gd2O3, Nd2O3, Er2O3, La2O3, CuO- 1 and Fe2O3- 1) showed similar levels of cytokine release and collagens to the vehicle control. Notably, four MeONPs had differing effects on TGF- \(\beta 1\) and collagen deposition: MgO, MoO3, ZnO- 2, and ZnO- 3 induced high TGF- \(\beta 1\) release, but had little effect on collagen deposition with an Ashcroft score \(< 2\) . Considering this difference, a MeONP was classified as a fibrogenic inducer if \(\mathrm{FC}_{\mathrm{TGF - }\beta 1}\geq 2\) or Ashcroft score \(\geq 2\) . Among all in vivo datapoints, 126 were identified as fibrogenic. For each in vivo datapoint, there were 87 associated input parameters for multimodal feature fusion (MFF) analysis.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 635, 772, 657]]<|/det|>
+## ML modeling in multimodal feature fusion predictive framework  
+
+<|ref|>text<|/ref|><|det|>[[110, 673, 884, 888]]<|/det|>
+Given the complexity of the 15 branched events involved in MeONP- induced lung fibrosis at different nano- bio interfaces, machine learning methods were selected to establish the predictive framework. An MFF predictive framework driven by machine learning was proposed to identify meaningful patterns between heterogeneous multidimensional events and lung fibrosis in animals. To identify a suitable machine learning algorithm for the database, models were developed and evaluated using eight algorithms, including random forest (RF), locally
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[110, 90, 885, 852]]<|/det|>
+weighted learning (LWL), C4.5 decision tree (C4.5), k- nearest neighbor (k- NN), support vector machine (SVM), Bayesnet, decision table (DT), and logistic regression (LGR). The dataset was randomly split into a training set containing 156 data points of 40 MeONPs and a test set consisting of 47 data points generated from 12 MeONPs. For each MeONP, all its four data points were included in either the training set or the test set. The six performance metrics values of the eight machine learning algorithms in the 10- fold cross- validation on the training set and in the external validation on the test set are plotted in Figure 3. Their detailed performances \((ACC, MCC, SE, SP, AUC,\) and \(F1\) values) of the eight algorithms are provided in Table S4. The formulae of the indexes \((ACC, MCC, SE, SP, AUC,\) and \(F1\) values) are listed in the supporting information. Among all models, the RF model had the best performance, with a strong robustness \((ACC = 89\%\) and \(AUC = 94\%\) in the 10- fold cross- validation) and a high predictive accuracy \((ACC = 84\%\) and \(AUC = 85\%\) in the external validation). The models developed with C4.5, SVM, and Bayesnet also exhibited a satisfactory performance, with \(ACC\) values 85 - 90% in the 10- fold cross- validations and 74 - 76% in the external validations. Their \(AUC\) values ranged from 84% to 90% in the cross- validations and 61% to 79% in the external validations. However, the remaining four classifiers showed poor performance in the external validations. The \(ACC\) values were below 70% for \(k\) - NN, DT, and LGR models, and the \(AUC\) for LWL model was lower less than 40%. Based on these results, RF was selected as the optimal algorithm for the MFF predictive framework. Thus, the RF model was further validated by animal tests.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 93, 656, 115]]<|/det|>
+## Experimental validation of the MFF predictive model  
+
+<|ref|>text<|/ref|><|det|>[[111, 130, 884, 581]]<|/det|>
+Five new MeONPs \(\mathrm{(Ho_2O_3}\) , \(\mathrm{Pr_6O_{11}}\) , \(\mathrm{Co_3O_4}\) , \(\mathrm{ZrO_2}\) , and \(\mathrm{CuO}\) ) were selected for validation based on two criteria: i) they were not included in the training or test sets, and ii) the positive/negative ratio in this validation set fell within the range of 1:1 to 2:1, which is consistent with the ratios observed in the training and test sets. These selected MeONPs were administered to 20 mice (n = 4 for each MeONP) for 90 days, and the extent of fibrosis was determined by measuring TGF- \(\beta 1\) in BALF (Figure S3A) and collagen deposition in lung tissues (Figure S3B). The resulting 20 data points from the animal tests were then compared to the predictions generated by the MFF models (Table S5). The models exhibited high predictive accuracy for this independent dataset of lung fibrosis, achieving an ACC of 85% and AUC of 98%. As shown in Figure 4A, 17 out of the 20 data points were accurately predicted, primarily for \(\mathrm{Ho_2O_3}\) , \(\mathrm{Pr_6O_{11}}\) , \(\mathrm{Co_3O_4}\) , and \(\mathrm{ZrO_2}\) . The remaining three data points generated from CuO were correctly predicted in terms of collagen staining but not TGF- \(\beta 1\) levels.  
+
+<|ref|>text<|/ref|><|det|>[[111, 596, 884, 888]]<|/det|>
+We further assessed the diversity and applicability domain of the validated MEF model. These two characteristics are crucial for ensuring the inclusion of representative data and defining the range within which the proposed framework can be reliably applied for new MeONPs. To characterize the applicability domain, we employed a descriptor standardization approach, which encompassed all 223 data points from the training set (156), test set (47), and experimental validation set (20) (Table S6). A similarity network was generated based on the descriptor spaces of the established models to visualize the distribution of the 223 data points (Figure 4B). The thickness of the lines between two data points represented the strength of their
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 92, 885, 190]]<|/det|>
+correlation, and the sparse connections indicated high homogeneity within the descriptor spaces. Notably, the descriptor space of MeONPs exhibited significant heterogeneity, with data points scattered widely.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 247, 383, 268]]<|/det|>
+## Mechanism interpretation  
+
+<|ref|>text<|/ref|><|det|>[[112, 282, 884, 895]]<|/det|>
+We attempted to interpret the mechanism by analyzing the key biological events associated with MeONP- induced lung fibrosis. A Shapley Additive Explanation plot (SHAP) plot was used to identify the features that exert significant influence. Figure 5A illustrates the SHAP summary plot, where each dot represents an individual data point in the dataset. The horizontal position of the dots on the x- axis indicates the impact of the corresponding feature values on the model's predictions. The top seven features that exert the greatest influence on the model's predictions are IL- 1β, NADH in macrophages, TGF- β1, dissolution in PSF, zeta potential, hydrodynamic size, and NADH in epithelial cells. Notably, IL- 1β emerges as the most critical descriptor, contributing \(27.8\%\) to the overall feature importance, followed by NADH in THP- 1 at \(17.6\%\) . The remaining five descriptors accounted for \(54.5\%\) of the feature importance (Figure 5B). Based on the results of the SHAP analysis, we sought to elucidate the mechanisms underlying MeONP- induced lung fibrosis. As illustrated in Figure 5C, while the mucociliary escalator effectively removes large MeONP agglomerates from the airways, small particles, especially cationic particles, can enter lysosomes for decomposition. The acidic environment of the lysosomal fluids may accelerate the dissolution of MeONPs. Subsequently, the released metal ions or escaped MeONPs can interact with mitochondria, disrupting cellular metabolism
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 90, 886, 268]]<|/det|>
+and eliciting ROS, leading to the activation of IL- 1β and TGF- β1. These cytokines, in turn, promote the recruitment of immune cells, proliferation of fibroblasts, and deposition of collagen, culminating in the development of pulmonary fibrogenesis. Notably, these key events have been well documented in the literature regarding fibrosis pathology and the pulmonary behavior of nanoparticles<sup>26, 39</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[111, 285, 885, 616]]<|/det|>
+As the number of ENMs continues to expand, it is crucial to evaluate their potential profibrogenic risks before exposure to the environment. MFF has demonstrated high predictive accuracy in assessing the fibrogenic risk of MeONPs. To improve the practicality of the MFF model, it was transformed into a software named "Nano- induced Lung Fibrosis Prediction" (NILFP v 1.0.0) with a simplified user interface. It is available for noncommercial use at GitHub (https://github.com/huangyang2023/NILFPv1.0.0/releases/download/NILFPv1.0.0/NILFP.v1.0.0.zip). NILFP can be used for fibrogenic risk assessment of untested MeONPs, MeONP- based nanoproducts and beyond. It will greatly speed up the respiratory risk assessment of nano- enabled products.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 673, 241, 693]]<|/det|>
+## Conclusions  
+
+<|ref|>text<|/ref|><|det|>[[111, 710, 885, 887]]<|/det|>
+Our study presents a reliable in silico model for predicting the fibrogenic potential of ENMs through the construction of an MFF predictive framework. The MEF model achieved high accuracy (>85%) in predicting MeONP- induced lung fibrosis, making it a valuable tool for risk assessment. The developed model utilized diverse data points and was broadly applicable to different MeONPs. Furthermore, machine learning analysis identified seven key descriptors
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 92, 884, 189]]<|/det|>
+and facilitated the interpretation of the underlying biological mechanism. Our research provides a cost- effective, time- efficient, and mechanism- driven alternative to the current practice of chronic nanotoxicity assessment in animals.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 248, 331, 268]]<|/det|>
+## Experimental section  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 287, 203, 304]]<|/det|>
+## Materials  
+
+<|ref|>text<|/ref|><|det|>[[111, 321, 885, 736]]<|/det|>
+MTS and ATP assay kits were obtained from Promega (Madison, WI, USA). H2DCF- DA, Fetal bovine serum (FBS), penicillin, and streptomycin were obtained from Thermo Fisher Scientific (Grand Island, NY, USA). RPMI 1640 medium were purchased from Corning (Steuben, NY, USA). Bronchial epithelial cell medium were purchased from Sciencell (San Diego, CA, USA). ELISA kits for detection of IL- 1β, IL- 2, IL- 6, IL- 12, and TNF- α were came from BD biosciences (San Jose, CA, USA). ELISA kits for detection of TGF- β was came from RD biotechne (MN, USA). LDH assay kits were obtained from Leagene (Beijing, China). LysoSensor™ Yellow/Blue DND- 160 was purchased from Yeasen Ltd. (Shanghai, China). 4% paraformaldehyde were obtained from Biosharp (Anhui, China). Al₂O₃, TiO₂, ZnO, NiO, In₂O₃, Bi₂O₃, ZrO₂, MoO₃, SnO₂, MgO, Sb₂O₃, CeO₂- 1, CeO₂- 2, Y₂O₃, Dy₂O₃, and Yb₂O₃ were obtained from Aladin (Shanghai, China). The rest of the materials were made in the laboratory.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 792, 340, 809]]<|/det|>
+## Material characterization  
+
+<|ref|>text<|/ref|><|det|>[[113, 828, 884, 886]]<|/det|>
+Morphologies and primary sizes of MeONPs were examined by transmission electron microscopy (TEM) using a Tecnai G2 spirit BioTwin microscope (FEI, Oregon, USA) operated
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 92, 884, 308]]<|/det|>
+at \(120\mathrm{kV}\) . MeONP suspensions ( \(50\mu \mathrm{g / mL}\) in deionized water) were placed on 200- mesh copper grids coated with carbon- coated formvar support film (Ted Pella, Inc., Redding, CA, USA) and air- dried at room temperature. The hydrodynamic diameters and surface charges of MeONP dispersions in water were determined by dynamic light scattering and zeta potential analysis using a Zetasizer Nano ZS90 instrument (Malvern Instruments Corp., UK), as previously described<sup>40</sup>.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 365, 655, 384]]<|/det|>
+## Examination of metal dissolution in simulated biological fluids  
+
+<|ref|>text<|/ref|><|det|>[[112, 401, 884, 810]]<|/det|>
+To investigate the dissolution behavior of MeONPs in phagolysosomal simulated fluid (PSF) and stimulated lung fluid (SLF), MeONPs were dissolved in \(10\mathrm{mL}\) of PSF buffer ( \(142\mathrm{mg / L}\) \(\mathrm{Na_2HPO_4}\) , \(6.65\mathrm{g / LNaCl}\) , \(62\mathrm{mg / LNa_2SO_4}\) , \(29\mathrm{mg / LCaCl_2\cdot H_2O}\) , \(250\mathrm{mg / L}\) glycine, \(8.09\mathrm{g / L}\) potassium phthalate, pH 4.5) or SLF buffer ( \(95\mathrm{mg / LMgCl_2}\) , \(6.019\mathrm{g / LNaCl}\) , \(298\mathrm{mg / LKCl}\) , \(126\mathrm{mg / LNa_2HPO_4}\) , \(63\mathrm{mg / LNa_2SO_4}\) , \(368\mathrm{mg / LCaCl_2\cdot 2H_2O}\) , \(574\mathrm{mg / LCH_3COONa}\) , \(2.604\mathrm{g / LNaHCO_3}\) , \(97\mathrm{mg / L}\) sodium citrate dihydrate, pH 7.4) at a concentration of \(50\mu \mathrm{g / mL}\) with probe sonication at \(32\mathrm{W}\) for \(10\mathrm{s}\) . The resulting MeONP suspensions were incubated for \(24\mathrm{h}\) at room temperature. The supernatants were collected by centrifugation at \(15,000\mathrm{RPM}\) for 10 min and analyzed for metal ion concentrations using inductively coupled plasma- atomic emission spectrometry (ICP- OES DUO 6500, Thermo Scientific, Massachusetts, USA). The percentage of MeONP dissolution was calculated using following equation:  
+
+<|ref|>equation<|/ref|><|det|>[[348, 821, 880, 855]]<|/det|>
+\[\text{Dissolution \%} = \frac{c_{\mathrm{t}}\times\nu_{\mathrm{t}}}{c_{0}\times\nu_{0}\times R}\times 100\% \quad (1)\]
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 92, 884, 228]]<|/det|>
+where \(\mathrm{Ct}\) and \(\mathrm{C_0}\) (50 \(\mu \mathrm{g / mL}\) ) are the concentrations of metal ions in the supernatant measured by ICP- OES and MeONPs before digestion, respectively; \(\mathrm{Vt}\) and \(\mathrm{V_0}\) are the volumes of digestive solution and MeONP suspension, respectively; R represents the mass ratio of metal elements in each specific MeONP.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 287, 373, 304]]<|/det|>
+## Assessment of in vitro toxicity  
+
+<|ref|>text<|/ref|><|det|>[[112, 320, 885, 890]]<|/det|>
+THP- 1 and BEAS- 2B cells were cultured in RPMI 1640 medium supplemented with \(10\%\) fetal bovine serum (Gemini) and BEpicM (Sciencell), respectively. To assess the effects of MeONPs, THP- 1 cells were primed with \(1 \mu \mathrm{g / mL}\) PMA and seeded in 96- well plates at a density of \(3 \times 10^{4}\) cells/well. BEAS- 2B cells were seeded in plates at a density of \(8 \times 10^{3}\) cells/well. After overnight culture, the cell media were removed and replaced with \(100 \mu \mathrm{L}\) aliquots of fresh medium containing MeONPs at concentrations of 0, 12.5, 25, 50, 100, and \(200 \mu \mathrm{g / mL}\) . The cells were incubated for \(24 \mathrm{~h}\) at \(37^{\circ} \mathrm{C}\) . The supernatants were then collected for detection of LDH release and cytokine production, including TGF- \(\beta 1\) , TNF- \(\alpha\) , IL- 1 \(\beta\) , IL- 2, IL- 6, and MCP- 1. The MeONPs treated cells were incubated with \(120 \mu \mathrm{L}\) of MTS working solution (5 mg/mL) in phenol red- free media for \(2 \mathrm{~h}\) at \(37^{\circ} \mathrm{C}\) to examine mitochondria activity, lysed in \(100 \mu \mathrm{L}\) working solution in ATP assay kit to assess energy metabolism by luminescence, or \(15 \mu \mathrm{g / mL}\) H2DCF- DA in the dark for \(30 \mathrm{~min}\) at \(37^{\circ} \mathrm{C}\) to detect ROS generation using a microplate reader at an excitation wavelength of \(488 \mathrm{~nm}\) and an emission wavelength of \(525 \mathrm{~nm}\) . Meanwhile, lysosomal pH was determined using the Lysosensor Yellow/Blue DND- 160 assay kit (40768ES50, Yishang) according to the manufacturer's instructions. Briefly, the culture media
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 92, 883, 228]]<|/det|>
+in MeONP- treated cells was replaced with \(100~\mu \mathrm{L}\) PBS containing \(1\mu \mathrm{M}\) DND- 160 probes. After incubating at \(37^{\circ}\mathrm{C}\) for \(3\mathrm{min}\) , the cells were washed twice with PBS and detected using a microplate reader at \(384~\mathrm{nm}\) excitation and \(540~\mathrm{nm}\) emission. The relative fold changes of indexes were calculated by following equation:  
+
+<|ref|>equation<|/ref|><|det|>[[433, 239, 880, 273]]<|/det|>
+\[\mathrm{FC} = \frac{I_{\mathrm{NP}} - I_{\mathrm{BL}}}{I_{\mathrm{Ctrl}} - I_{\mathrm{BL}}} \quad (2)\]  
+
+<|ref|>text<|/ref|><|det|>[[115, 286, 881, 344]]<|/det|>
+where \(I_{\mathrm{NP}}\) , \(I_{\mathrm{Ctrl}}\) and \(I_{\mathrm{BL}}\) represent the measured intensity of indexes in MeONP- treated cells, vehicle solution treated cells and blanks, respectively.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 404, 420, 422]]<|/det|>
+## Assessment of lung fibrosis in mice  
+
+<|ref|>text<|/ref|><|det|>[[113, 440, 884, 888]]<|/det|>
+A total of 208 animals were used to determine chronic lung fibrosis induced by 52 different MeONPs, with four animals in each group (n = 4). Female C57Bl/6 mice (8 weeks old) were purchased from Nanjing Peng Sheng Biological Technology (Nanjing, China). Animals were housed in groups of four under standard laboratory conditions (25°C; 60% relative humidity; 12 h light, 12 h dark cycle) and hygiene status (autoclaved food and acidified water) according to Soochow University guidelines for the care and treatment of laboratory animals. All animal experiments were approved by the Ethics Committee of Soochow University. Animals were exposed to MeONPs by an oropharyngeal instillation method. Briefly, MeONPs were suspended in PBS at \(1\mathrm{mg / mL}\) by a probe sonication (32 W) for 10 s. The animals were anesthetized by intraperitoneal injection of sodium pentobarbital (200 mg/kg) in a total volume of \(100~\mathrm{uL}\) . The anesthetized animals were held in a vertical position for the pulmonary aspiration of MeONP suspension (with a dose of \(2\mathrm{mg / kg}\) ) at the back of the tongue. Animals
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 92, 884, 422]]<|/det|>
+included in vehicle and positive controls received 50 uL of PBS and 2 mg/kg quartz, respectively. The mice were exposed once a week for three weeks and sacrificed on day 90 by overdose of sodium pentobarbital (400 mg/kg). Bronchoalveolar lavage fluids (BALFs) and lung tissues were collected according to previous studies<sup>40</sup>. Briefly, the trachea was cannulated and then the lungs were gently lavaged 3 times with 1 mL of sterile PBS to obtain BALF. Aliquots of 50 uL BALF were used to measure TGF- \(\beta 1\) levels by the ELISA kits (BDLISA, China). Lung tissues were collected and stained Masson’s Trichrome staining according to a standard protocol<sup>41</sup>. The collagen deposition levels of the 203 mice were estimated according to the method reported by Ashcroft et al<sup>42</sup>.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 480, 400, 499]]<|/det|>
+## Construction of the MFF models  
+
+<|ref|>text<|/ref|><|det|>[[111, 518, 884, 888]]<|/det|>
+We developed classification models for lung fibrosis using eight different machine learning algorithms, namely C4.5 decision tree (C4.5), random forest (RF), logistic regression (LGR), k- nearest neighbor (kNN), locally weighted learning (LWL), support vector machine (SVM), Bayesnet, and decision table (DT). These classifiers represented various categories of supervised classifiers such as trees, Bayes, and rules. To validate the models, we randomly split the dataset into training and test sets, with 156 data points from 40 metal oxide nanoparticles (MeONPs) in the training set and 47 data points from 12 MeONPs in the test set. To avoid information leaking, we included all four data points of the same MeONP in either the training or test set. The criteria for classifying data points of MeONPs as fibrogenic potential were \(\mathrm{FC}_{\mathrm{TGF - }\beta 1} \geq 2\) or Ashcroft score \(\geq 2\) .
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 130, 884, 696]]<|/det|>
+To optimize the model's performance, we exploited a 10- fold- cross- validation procedure based on a grid search to determine the best parameters in the machine learning methods. We performed all procedures using Weka software (Ver 3.8.5). We applied 10- fold cross- validation on the training set to assess the prediction accuracy of the models. External validation was performed on the test set. The performances of models were evaluated based on true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN). Model performance was evaluated using five metrics: sensitivity (SE) = TP/(TP + FN), specificity (SP) = TN/(TN + FP), overall predictive accuracy (ACC) = (TP + TN)/(TP + FP + TN + FN), F1 score = 2×SE×SP/(SE+SP), and Matthews' correlation coefficient (MCC) = (TP×TN - FP×FN)/√(TP+FP)(TP+FN)(TN+FP)(TN+FN). MCC ranges from -1 to +1, with extreme values of -1 and +1 in the case of perfect misclassification and perfect classification, respectively. We also calculated AUC by plotting the TP rate versus the FP rate at various threshold settings. We considered the performance of a model "excellent" if AUC ≥ 0.9, "very poor" if AUC < 0.6, "poor" if 0.7 > AUC ≥ 0.6, "fair" if 0.8 > AUC ≥ 0.7, and "good" if 0.9 > AUC ≥ 0.8.  
+
+<|ref|>text<|/ref|><|det|>[[113, 752, 883, 888]]<|/det|>
+We characterized the applicability domain of the prediction model using a descriptor standardization approach. Briefly, a data point was considered an outlier if all normalized descriptors for the data point were greater than 3; otherwise, it was a non- outlier. We generated a similarity network of the in vivo data points using Gephi software (V 0.9). We represented
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 92, 882, 268]]<|/det|>
+the data points as nodes and colored them according to the chemical composition of MeONPs. We used the size of the circle to represent the fibrosis level (release of TGF- \(\beta 1\) ) in the lung. The thickness of the lines reflected the strength of the correlations between the data points. Tight and sparse connections indicated high and low homogeneity of the data points in descriptor spaces, respectively.  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 326, 314, 345]]<|/det|>
+## Acknowledgements  
+
+<|ref|>text<|/ref|><|det|>[[113, 358, 884, 640]]<|/det|>
+This work was supported by the National Key Research and Development Program (2022YFC3902104, 2022YFE0124000) of China, National Natural Science Foundation (22176023, 21976126) of China, the Natural Science Foundation of Jiangsu Province (BK20211545), the Fundamental Research Funds for the Central Universities (DUT22QN216) and the Project of National Center for International Research on Intelligent Nano- Materials and Detection Technology in Environmental Protection, Soochow University (No. SDGH2202). We thank T. A. Patterson from National Center for Toxicological Research, U.S. Food and Drug Administration for his contribution to the improvement of the manuscript. This article reflects the views of the authors and does not necessarily reflect those of the U.S. Food and Drug Administration.  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 655, 334, 675]]<|/det|>
+## Author contributions  
+
+<|ref|>text<|/ref|><|det|>[[114, 688, 884, 823]]<|/det|>
+X. Li, X. Cai and R. Li conceived the idea and designed the experiments. H. Yang established the predictive model. J. Cao performed most of the cell and animal experiments. Q. Xie performed ELISA assay. Q. Yang contributed to the histology assay of animal lungs. X. Liu performed the characterization of MeONPs. The writing of the manuscript was led by R. Li, X. Li, and X. Cai with participations from J. Chen, and H. Hong.  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 838, 321, 858]]<|/det|>
+## Competing interests  
+
+<|ref|>text<|/ref|><|det|>[[115, 877, 470, 895]]<|/det|>
+The authors declare no competing interests.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[117, 93, 353, 113]]<|/det|>
+## Additional information  
+
+<|ref|>text<|/ref|><|det|>[[115, 131, 884, 152]]<|/det|>
+The online version contains supplementary material available at https://github.com/huangyang  
+
+<|ref|>text<|/ref|><|det|>[[115, 170, 690, 189]]<|/det|>
+2023/NILFPv1.0.0/releases/download/NILFPv1.0.0/NILFP.v1.0.0.zip.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 84, 230, 101]]<|/det|>
+## References  
+
+<|ref|>text<|/ref|><|det|>[[113, 103, 884, 920]]<|/det|>
+1. Nanotechnology Products Database (NPD). [cited 2023 March 1st]Available from: https://product.statnano.com2. Kokot H, Kokot B, Sebastijanovic A, Voss C, Podlipec R, Zawilska P, et al. Prediction of chronic inflammation for inhaled particles: the impact of material cycling and quarantining in the lung epithelium. Advanced Materials 2020, 32(47): 2003913.3. Garcia-Mouton C, Hidalgo A, Cruz A, Perez-Gil J. The lord of the lungs: the essential role of pulmonary surfactant upon inhalation of nanoparticles. European Journal of Pharmaceutics and Biopharmaceutics 2019, 144: 230-243.4. Liu S, Xia T. Continued efforts on nanomaterial-environmental health and safety is critical to maintain sustainable growth of nanoindustry. Small 2020, 16(21): 2000603.5. Wu Z, Shi P, Lim HK, Ma Y, Setyawati MI, Bitounis D, et al. Inflammation increases susceptibility of human small airway epithelial cells to pneumonic nanotoxicity. Small 2020, 16(21): 2000963.6. Zhang H, Ji Z, Xia T, Meng H, Low-Kam C, Liu R, et al. Use of Metal Oxide Nanoparticle Band Gap To Develop a Predictive Paradigm for Oxidative Stress and Acute Pulmonary Inflammation. ACS Nano 2012, 6(5): 4349-4368.7. Cai X, Dong J, Liu J, Zheng H, Kaweeterawat C, Wang F, et al. Multi-hierarchical profiling the structure-activity relationships of engineered nanomaterials at nano-bio interfaces. Nature Communications 2018, 9: 4416.8. Li R, Ji Z, Chang CH, Dunphy DR, Cai X, Meng H, et al. Surface interactions with compartmentalized cellular phosphates explain rare earth oxide nanoparticle hazard and provide opportunities for safer design. ACS Nano 2014, 8(2): 1771-1783.9. Zhu M, Du L, Zhao R, Wang HY, Zhao Y, Nie G, et al. Cell-penetrating nanoparticles activate the inflammasome to enhance antibody production by targeting microtubule-associated protein 1-light chain 3 for degradation. ACS Nano 2020, 14(3): 3703-3717.10. Wu Z, Shi P, Lim HK, Ma Y, Setyawati MI, Bitounis D, et al. Inflammation Increases Susceptibility of Human Small Airway Epithelial Cells to Pneumonic Nanotoxicity. Small 2020, 16(21).11. Huang Y, Li X, Xu S, Zheng H, Zhang L, Chen J, et al. Quantitative structure-activity relationship models for predicting inflammatory potential of metal oxide nanoparticles. Environmental Health Perspectives 2020, 128(6): 67010.12. Barosova H, Maione AG, Septiadi D, Sharma M, Haeni L, Balog S, et al. Use of EpiAlveolar lung model to predict fibrotic potential of multiwalled carbon nanotubes. ACS Nano 2020, 14(4): 3941-3956.13. Wang X, Mansukhani ND, Guiney LM, Lee J-H, Li R, Sun B, et al. Toxicological profiling of highly purified metallic and semiconducting single-walled carbon nanotubes in the rodent lung and E. coli. ACS Nano 2016, 10(6): 6008-6019.14. Jiang W, Wang X, Osborne OJ, Du Y, Chang CH, Liao Y-P, et al. Pro-inflammatory and pro-fibrogenic effects of ionic and particulate arsenide and indium-containing semiconductor materials in the murine lung. ACS Nano 2017, 11(2): 1869-1883.15. Rahman L, Williams A, Gelda K, Nikota J, Wu D, Vogel U, et al. 21st century tools for nanotoxicology: transcriptomic biomarker panel and precision-cut lung slice organ mimic
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 85, 883, 103]]<|/det|>
+system for the assessment of nanomaterial- induced lung fibrosis. Small 2020, 16(36): 2000272.  
+
+<|ref|>text<|/ref|><|det|>[[113, 105, 883, 160]]<|/det|>
+16. Zhu M, Feng W, Wang B, Wang T, Gu Y, Wang M, et al. Comparative study of pulmonary responses to nano- and submicron-sized ferric oxide in rats. Toxicology 2008, 247(2-3): 102-111.  
+
+<|ref|>text<|/ref|><|det|>[[113, 162, 883, 218]]<|/det|>
+17. Gao M, Wang Z, Zheng H, Wang L, Xu S, Liu X, et al. Two-dimensional tin selenide (SnSe) Nanosheets capable of mimicking key dehydrogenases in cellular metabolism. Angewandte Chemie-International Edition 2020, 59(9): 3618-3623.  
+
+<|ref|>text<|/ref|><|det|>[[113, 220, 883, 256]]<|/det|>
+18. Schraufnagel DE. The health effects of ultrafine particles. Experimental and Molecular Medicine 2020, 52(3): 311-317.  
+
+<|ref|>text<|/ref|><|det|>[[113, 259, 883, 315]]<|/det|>
+19. Oberdorster C, Maynard A, Donaldson K, Castranova V, Fitzpatrick J, Ausman K, et al. Principles for characterizing the potential human health effects from exposure to nanomaterials: elements of a screening strategy. Particle and Fibre Toxicology 2005, 2(1): 8.  
+
+<|ref|>text<|/ref|><|det|>[[113, 317, 883, 353]]<|/det|>
+20. Anonymous. The risks of nanomaterial risk assessment. Nature Nanotechnology 2020, 15(3): 163-163.  
+
+<|ref|>text<|/ref|><|det|>[[113, 355, 883, 392]]<|/det|>
+21. Choi JY, Ramachandran G, Kandlikar M. The impact of toxicity testing costs on nanomaterial regulation. Environmental Science & Technology 2009, 43(9): 3030-3034.  
+
+<|ref|>text<|/ref|><|det|>[[113, 394, 883, 450]]<|/det|>
+22. Yu F, Wei C, Deng P, Peng T, Hu X. Deep exploration of random forest model boosts the interpretability of machine learning studies of complicated immune responses and lung burden of nanoparticles. Science Advances 2021, 7(22).  
+
+<|ref|>text<|/ref|><|det|>[[113, 452, 883, 508]]<|/det|>
+23. Ji Z, Guo W, Wood EL, Liu J, Sakkiah S, Xu X, et al. Machine learning models for predicting cytotoxicity of nanomaterials. Chemical Research in Toxicology 2022, 35(2): 125-139.  
+
+<|ref|>text<|/ref|><|det|>[[113, 511, 883, 547]]<|/det|>
+24. Roy J, Roy K. Assessment of toxicity of metal oxide and hydroxide nanoparticles using the QSAR modeling approach. Environmental Science-Nano 2021, 8(11): 3395-3407.  
+
+<|ref|>text<|/ref|><|det|>[[113, 549, 883, 585]]<|/det|>
+25. Labouta HI, Asgarian N, Rinker K, Cramb DT. Meta-Analysis of Nanoparticle Cytotoxicity via Data-Mining the Literature. Acs Nano 2019, 13(2): 1583-1594.  
+
+<|ref|>text<|/ref|><|det|>[[113, 588, 883, 624]]<|/det|>
+26. Oh E, Liu R, Nel A, Gemill KB, Bilal M, Cohen Y, et al. Meta-analysis of cellular toxicity for cadmium-containing quantum dots. Nature Nanotechnology 2016, 11(5): 479.  
+
+<|ref|>text<|/ref|><|det|>[[113, 626, 883, 682]]<|/det|>
+27. Cao J, Pan Y, Jiang Y, Qi R, Yuan B, Jia Z, et al. Computer-aided nanotoxicology: risk assessment of metal oxide nanoparticles via nano-QSAR. Green Chemistry 2020, 22(11): 3512-3521.  
+
+<|ref|>text<|/ref|><|det|>[[113, 685, 883, 740]]<|/det|>
+28. Huang Y, Li X, Cao J, Wei X, Li Y, Wang Z, et al. Use of dissociation degree in lysosomes to predict metal oxide nanoparticle toxicity in immune cells: Machine learning boosts nano-safety assessment. Environment International 2022, 164: 107258.  
+
+<|ref|>text<|/ref|><|det|>[[113, 743, 883, 798]]<|/det|>
+29. Yan XL, Sedykh A, Wang WY, Yan B, Zhu H. Construction of a web-based nanomaterial database by big data curation and modeling friendly nanostructure annotations. Nature Communications 2020, 11(1): 2519.  
+
+<|ref|>text<|/ref|><|det|>[[113, 801, 884, 856]]<|/det|>
+30. Ban Z, Yuan P, Yu F, Peng T, Zhou Q, Hu X. Machine learning predicts the functional composition of the protein corona and the cellular recognition of nanoparticles. Proceedings of the National Academy of Sciences of the United States of America 2020, 117(19): 10492-10499.  
+
+<|ref|>text<|/ref|><|det|>[[113, 859, 883, 914]]<|/det|>
+31. Friedersdorf LE, Bjorkland R, Klaper RD, Sayes CM, Wiesner MR. Fifteen years of nanoEHS research advances science and fosters a vibrant community. Nature Nanotechnology 2019, 14(11): 996-998.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 85, 884, 700]]<|/det|>
+32. Wyrzykowska E, Mikolajczyk A, Lynch I, Jeliazkova N, Kochev N, Sarimveis H, et al. Representing and describing nanomaterials in predictive nanoinformatics. Nature Nanotechnology 2022, 17(9): 924-932.33. Tung Xuan T, Choi J-S, Jeon H, Byun H-G, Yoon T-H, Kim J. Quasi-SMILES-based nano-quantitative structure-activity relationship model to predict the cytotoxicity of multiwalled carbon nanotubes to human lung cells. Chemical Research in Toxicology 2018, 31(3): 183-190.34. Forest V, Hochepied J-F, Leclerc L, Trouve A, Abdelkebir K, Sarry G, et al. Towards an alternative to nano-QSAR for nanoparticle toxicity ranking in case of small datasets. Journal of Nanoparticle Research 2019, 21(5): 95.35. Gajewicz A, Schaeublin N, Rasulev B, Hussain S, Leszczynska D, Puzyn T, et al. Towards understanding mechanisms governing cytotoxicity of metal oxides nanoparticles: Hints from nano-QSAR studies. Nanotoxicology 2015, 9(3): 313-325.36. Puzyn T, Rasulev B, Gajewicz A, Hu X, Dasari TP, Michalkova A, et al. Using nano-QSAR to predict the cytotoxicity of metal oxide nanoparticles. Nature Nanotechnology 2011, 6(3): 175-178.37. Mu Y, Wu F, Zhao Q, Ji R, Qie Y, Zhou Y, et al. Predicting toxic potencies of metal oxide nanoparticles by means of nano-QSARs. Nanotoxicology 2016, 10(9): 1207-1214.38. Zhong S, Zhang K, Bagheri M, Burken JG, Gu A, Li B, et al. Machine learning: new ideas and tools in environmental science and engineering. Environmental Science & Technology 2021, 55(19): 12741-12754.39. Cai XM, Liu X, Jiang J, Gao M, Wang WL, Zheng HZ, et al. Molecular mechanisms, characterization methods, and utilities of nanoparticle biotransformation in nanosafety assessments. Small 2020, 16(36): 19.40. Huang Y, Li X, Xu S, Zheng H, Zhang L, Chen J, et al. Quantitative Structure-Activity Relationship Models for Predicting Inflammatory Potential of Metal Oxide Nanoparticles. Environ Health Perspect 2020, 128(6): 67010.41. Foot NC. The Masson Trichrome Staining Methods in Routine Laboratory Use. Stain technology 1933, 8(3): 101-110.42. Ashcroft T, Simpson JM, Timbrell V. Simple method of estimating severity of pulmonary fibrosis on a numerical scale. Journal of Clinical Pathology 1988, 41(4): 467-470.
+
+<--- Page Split --->
+<|ref|>table<|/ref|><|det|>[[115, 120, 810, 710]]<|/det|>
+<|ref|>table_caption<|/ref|><|det|>[[115, 106, 450, 121]]<|/det|>
+Table 1. Multimodality input parameters   
+
+<table><tr><td>Type</td><td>Descriptors</td><td>Number</td></tr><tr><td rowspan="7">Periodic parameters</td><td>Molecular weight</td><td>1</td></tr><tr><td>Electronegativity of metal atom</td><td>1</td></tr><tr><td>Metal atom number</td><td>1</td></tr><tr><td>Oxygen atom number</td><td>1</td></tr><tr><td>Cation charge</td><td>1</td></tr><tr><td>Periodic number of metal atom</td><td>1</td></tr><tr><td>Group number of metal atom</td><td>1</td></tr><tr><td rowspan="4">Physicochemical properties in biological fluids</td><td>Atomic ratio of metal and oxygen</td><td>1</td></tr><tr><td>Hydrodynamic size</td><td>1</td></tr><tr><td>Zeta-potential</td><td>1</td></tr><tr><td>Dissolution in PSF</td><td>1</td></tr><tr><td rowspan="2">MeONP-membrane interaction</td><td>Dissolution in SLF</td><td>1</td></tr><tr><td>LDH release in THP-1 cells</td><td>5</td></tr><tr><td rowspan="2">MeONP-lysosome interaction</td><td>LDH release in BEAS-2B cells</td><td>5</td></tr><tr><td>pH changes in lysosomes</td><td>5</td></tr><tr><td rowspan="4">Impact on mitochondria activity</td><td>NADH level in THP-1 cells</td><td>5</td></tr><tr><td>NADH level in BEAS-2B cells</td><td>5</td></tr><tr><td>ATP level in THP-1 cells</td><td>5</td></tr><tr><td>ATP level in BEAS-2B cells</td><td>5</td></tr><tr><td rowspan="2">Impact on redox homeostasis</td><td>ROS generation in THP-1 cells</td><td>5</td></tr><tr><td>ROS generation in BEAS-2B cells</td><td>5</td></tr><tr><td rowspan="6">Impact on cell–cell communications</td><td>TNF-α release in THP-1 cells</td><td>5</td></tr><tr><td>IL-1β release in THP-1 cells</td><td>5</td></tr><tr><td>IL-2 release in THP-1 cells</td><td>5</td></tr><tr><td>IL-6 release in THP-1 cells</td><td>5</td></tr><tr><td>MCP-1 release in THP-1 cells</td><td>5</td></tr><tr><td>TGF-β1 release in BEAS-2B cells</td><td>5</td></tr><tr><td colspan="2">Total descriptors</td><td>87</td></tr></table>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[160, 130, 835, 720]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 725, 541, 741]]<|/det|>
+<center>Figure 1 Interfaces of nano-bio interactions in the lungs. </center>  
+
+<|ref|>text<|/ref|><|det|>[[115, 744, 883, 914]]<|/det|>
+A) Schematic workflow of multimodal feature fusion (MFF) modeling. The figure depicts the travel path of nanoparticles in lungs where inhaled nanoparticles may deposit in alveoli and interact with macrophages and epithelial cells, leading to injury of various cellular components such as cell membranes, lysosomes, mitochondria, and other cellular components. Descriptive features at these nano-bio interfaces and fibrogenic indexes in vivo were collected to develop predictive models using machine learning methods. The established in silico model was further validated by animal tests. B) Confocal and C) TEM imaging of MeONPs in cells. THP-1 cells exposed to \(12.5 \mu \mathrm{g / mL}\) FITC-labeled or pristine MeONPs were collected. The fixed cells were stained with DAPI (blue) and Alexa Fluor™ 594 conjugated with WGA/anti-LAMP1 (red) for confocal imaging. Arrow and L indicate the MeONP and lysosome, respectively.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[140, 88, 870, 840]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 860, 634, 876]]<|/det|>
+<center>Figure 2 Effects of representative MeONPs in cells and mouse lungs. </center>  
+
+<|ref|>text<|/ref|><|det|>[[113, 878, 883, 914]]<|/det|>
+A) Heatmap displaying the impacts of 10 representative MeONPs on THP-1 and BEAS-2B cells by detecting LDH leakage, ROS generation, NADH content, lysosomal pH change, ATP production, and cytokine release.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 84, 884, 295]]<|/det|>
+THP- 1 cells were exposed to 0, 12.5, 25, 50, 100, and \(200\mu \mathrm{g / mL}\) MeONPs for \(24\mathrm{h}\) , followed by LDH and cytokine measurement in supernatants and ROS, NADH, ATP and lysosomal pH detection in cells. BEAS- 2B cells were exposed to 0, 12.5, 25, 50, 100, and \(200\mu \mathrm{g / mL}\) MeONPs for \(24\mathrm{h}\) , followed by TGF- \(\beta 1\) detection in cells. The values of these descriptors in MeONP treatments were compared with control cells. The resulting ratios were expressed as fold changes (FCs) in the heatmap. B) Schematic illustration of MeONP instillation in mice. Mice were oropharyngeally administered with \(50\mu \mathrm{L}\) PBS (vehicle control), \(2\mathrm{mg / kg}\) MeONPs, and quarts (positive control) three times a week. The animals were sacrificed at day 90 to collect BALF and lung tissues for further examinations. C) TGF- \(\beta 1\) release in BALFs and Ashcroft score of stained lung sections. D) Masson’s trichrome staining of lung tissues exposed to representative MeONPs. BALFs were collected to measure TGF- \(\beta 1\) by ELISA \((n = 4)\) . \*p<0.05, \*\*p<0.01 and \*\*\*p<0.001 compared to the vehicle control by one- way ANOVA.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[184, 105, 802, 578]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 626, 750, 643]]<|/det|>
+<center>Figure 3 Performance of models constructed using eight machine learning algorithms. </center>  
+
+<|ref|>text<|/ref|><|det|>[[115, 645, 884, 720]]<|/det|>
+The y- axis represents the values of the performance metrics. RF: random forest; C4.5: C4.5 decision tree; LWL: locally weighted learning; \(k\) - NN: k nearest neighbor; SVM: support vector machine; DT: decision table; LGR: logistic regression. \(ACC\) : overall predictive accuracy; \(MCC\) : Matthews correlation coefficient; \(SE\) : sensitivity; \(SP\) : specificity; \(AUC\) : the area under the receiver operating characteristic curve; \(F1\) : F1 score.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[205, 111, 777, 760]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 784, 725, 799]]<|/det|>
+<center>Figure 4 Experimental validation and diversity analysis of the established model. </center>  
+
+<|ref|>text<|/ref|><|det|>[[115, 802, 884, 915]]<|/det|>
+A) Experimental and predicted results of lung fibrosis in the experimental validation set are depicted in the confusion matrix, with each node representing a data point. 
+B) The similarity network of the in vivo data points. Each node represents a data point which is colored according to the chemical composition of MeONPs. The size of a node indicates the fibrosis level in the lung, while the thickness of a line represents the strength of correlation between the two connected nodes. Tight and sparse connections denote high and low homogeneity of nodes for descriptor spaces, respectively.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[201, 85, 768, 250]]<|/det|>  
+
+<|ref|>image<|/ref|><|det|>[[201, 315, 825, 748]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 761, 658, 778]]<|/det|>
+<center>Figure 5 Identification of key descriptors for mechanism interpretation. </center>  
+
+<|ref|>text<|/ref|><|det|>[[115, 780, 884, 914]]<|/det|>
+A) SHAP summary plots displaying the effects of features and their values on the prediction. The y-axis of each plot contains the features included in the model sorted from the most (top) to least (bottom) important. The x-axis depicts the SHAP value, with each point referring to a SHAP value associated with a value of a certain feature. The color of the point displays whether the feature value is high (pink) or low (blue). B) Feature importance in the RF model. C) Schematic image of the proposed in chemico/in vitro-in vivo extrapolation of lung fibrosis. The key determinants of MeONP-induced lung fibrosis include IL-1β, TGF-β1, metabolic activity, hydrodynamic size, zeta potential, and metal ion release in PSF.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[42, 43, 312, 71]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[42, 93, 768, 113]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[60, 131, 323, 178]]<|/det|>
+SupplementalExcelFile.xlsx SupportingInformation.pdf
+
+<--- Page Split --->

preprint/preprint__0fc91687659d1cbd096c1f45e57db497948054dbc6edf1d233f7af3f1cd54319/images_list.json

@@ -0,0 +1,62 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_0.jpg",
+    "caption": "Figures and Captions:",
+    "footnote": [],
+    "bbox": [
+      [
+        130,
+        150,
+        870,
+        696
+      ]
+    ],
+    "page_idx": 12
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Fig. 2 | Bending test of PIN-PMN-PT. a, A load-depth curve obtained during bending a cantilever beam along a loading direction of [010]. The inset shows the enlarged curve of the rectangular area, where an abrupt decrease in mechanical load is evident. b, Snapshot captured from in-situ video corresponding to the maximum depth of the indenter. c, An SEM image showing cantilever beam after unloading. Irreversible deformation can be clearly revealed by comparison with Fig. S2g. d – e, A STEM-HAADF image and the corresponding GPA analysis of lattice rotation, with",
+    "footnote": [],
+    "bbox": [
+      [
+        120,
+        373,
+        880,
+        727
+      ]
+    ],
+    "page_idx": 13
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Fig. 3 | First principles atomistic investigation for the plasticity in PIN-PMN-PT. a, Three sub-unit cells for \\(\\mathrm{Pb(In_{1 / 2}Nb_{1 / 2})O_{3}}\\) (PIN), \\(\\mathrm{Pb(Mg_{1 / 3}Nb_{2 / 3})O_{3}}\\) (PMN) and \\(\\mathrm{PbTiO_{3}}\\) (PT). b, An example of relaxed atomic structure containing interfaces formed by one PIN, one PMN and one PT with one oxygen vacancy \\((V_{O}^{-})\\) at the interface of PIN and PMN. c, Calculated bulk modulus/shear modulus (B/G) ratios for various bulk, pristine interfaces, and interfaces with \\(V_{O}^{-}\\) . Higher B/G ratios (>1.75) suggest ductile behaviour in PIN-PMN-PT. d, Calculated valence charge density 2D contour plot (colours assigned recursively) on the (020) plane of the structure shown in b. The strength of covalent bonding is indicated by the colour bar. The presence of \\(V_{O}^{-}\\) eliminates the local covalent bonds.",
+    "footnote": [],
+    "bbox": [
+      [
+        123,
+        191,
+        861,
+        592
+      ]
+    ],
+    "page_idx": 14
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Fig. 4 | Analysis of the origin of extreme plasticity. a, O-EELS obtained from Mn- (blue), Sm- (green) doped and undoped (red) PIN-PMN-PT. The enlarged image (inset) shows the intensity difference of peak B for Mn-, Sm- doped and undoped PIN-PMN-PT, where a lower intensity suggests a higher \\(V_{O}^{*}\\) concentration. b, Engineering stress – strain curves obtained from in-situ compression tests of Mn- (blue sphere) and Sm- (green circle) doped PIN-PMN-PT. Images I and II show SEM images of the compressed Mn-doped and Sm-doped PIN-PMN-PT respectively.",
+    "footnote": [],
+    "bbox": [
+      [
+        123,
+        94,
+        875,
+        300
+      ]
+    ],
+    "page_idx": 15
+  }
+]

preprint/preprint__0fc91687659d1cbd096c1f45e57db497948054dbc6edf1d233f7af3f1cd54319/preprint__0fc91687659d1cbd096c1f45e57db497948054dbc6edf1d233f7af3f1cd54319.mmd

@@ -0,0 +1,326 @@
+
+# Extreme Room Temperature Compression and Bending in Ferroelectric Oxide Pillars  
+
+Ying Liu The University of Sydney Xiangyuan Cui The University of Sydney https://orcid.org/0000- 0002- 3946- 7324  
+
+Ranming Niu University of Sydney  
+
+Shujun Zhang University of Wollongong, Australia https://orcid.org/0000- 0001- 6139- 6887  
+
+Xiaozhou Liao University of Sydney https://orcid.org/0000- 0001- 8565- 1758  
+
+Scott Moss Defence Science and Technology Group  
+
+Peter Finkel US Naval Research Laboratory  
+
+Magnus Garbrecht The University of Sydney  
+
+Simon Ringer The University of Sydney https://orcid.org/0000- 0002- 1559- 330X  
+
+Julie Cairney ( julie.cairney@sydney.edu.au ) The University of Sydney https://orcid.org/0000- 0003- 4564- 2675  
+
+## Article  
+
+Keywords: Plastic Deformation, Ceramic Materials, Perovskite Oxide, Flexoelectric Polarization  
+
+Posted Date: June 1st, 2021  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 342103/v1  
+
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License
+
+<--- Page Split --->
+
+Version of Record: A version of this preprint was published at Nature Communications on January 17th, 2022. See the published version at https://doi.org/10.1038/s41467-022-27952-2.
+
+<--- Page Split --->
+
+## Extreme Room Temperature Compression and Bending in Ferroelectric Oxide Pillars  
+
+Y. Liu \(^{1,2}\) , 
+X.Y. Cui \(^{1,2}\) , 
+R.M. Niu \(^{1}\) , 
+S.J. Zhang \(^{3}\) , 
+X.Z. Liao \(^{1}\) , 
+S. Moss \(^{4}\) , 
+P. Finkel \(^{5}\) , 
+M. Garbrecht \(^{2}\) , 
+S.P. Ringer \(^{1,2}\) , 
+J.M. Cairney \(^{1,2*}\)  
+
+\(^{1}\) School of Aerospace, Mechanical & Mechatronic Engineering, The University of Sydney, NSW 2006, Australia; \(^{2}\) Australian Centre for Microscopy and Microanalysis, The University of Sydney, NSW 2006, Australia; \(^{3}\) ISEM, Australian Institute of Innovative Materials, University of Wollongong, NSW 2500, Australia; \(^{4}\) Aerospace Division, Defence Science and Technology Group, VIC 3207, Australia; \(^{5}\) US Naval Research Laboratory, Washington DC, 20375, USA  
+
+Plastic deformation in ceramic materials is normally only observed in nanometre- sized samples. However, we have observed unprecedented levels of plasticity ( \(>50\%\) plastic strain) and excellent elasticity ( \(6\%\) elastic strain) in perovskite oxide \(\mathrm{Pb(In_{1 / 2}Nb_{1 / 2})O_{3}}\) - Pb(Mg \(_{1 / 3}\) Nb \(_{2 / 3}\) )O \(_{3}\) - \(\mathrm{PbTiO_3}\) (PIN- PMN- PT), under compression along \(< 100>\) pc pillars up to 2.1 \(\mu \mathrm{m}\) in diameter. The extent of this deformation is much higher than has previously been reported for ceramic materials, and the sample size at which plasticity is observed is almost an order of magnitude larger. Bending tests also revealed over \(8\%\) flexural strain. Plastic deformation occurred by slip along \(\{110\} < 1\overline{1} 0>\) . Calculations indicate that the resulting strain gradients will give rise to extreme flexoelectric polarization. First principles models predict that a high concentration of oxygen vacancies \((V_{O}^{ - })\) weaken the covalent/ionic bonds, giving rise to the unexpected plasticity. Mechanical testing on \(V_{O}^{ - }\) - rich Mn- doped PIN- PMN- PT confirmed this prediction. These findings will facilitate the design of plastic ceramic materials and the development of flexoelectric- based nano- electromechanical systems.  
+
+Conventional wisdom dictates that most metals are ductile and almost all ceramics are brittle. The plasticity of metals is related to their atomic bonding. Valence electrons are not bound to a specific atom and there is little charge resistance during dislocation slip \(^{1}\) . For ceramics, the directional covalent or/and ionic bonds restrict slip due to electrostatic repulsion, resulting in brittle fracture with only limited strain (usually less than \(0.2\%\) ) \(^{1}\) . In many cases, the brittle nature of ceramics limits their application, and improvements to the brittle properties of ceramics materials have been sought for decades \(^{2}\) .
+
+<--- Page Split --->
+
+There are some rare exceptions to this rule. Crystals with the rock salt structure show limited plasticity due to their unique structure (slip occurs on \(\{110\}\) planes and along \(< 1\bar{1} 0>\) directions, where it does not bring similarly charged atoms together) \(^3\) . Among perovskite oxides, SrTiO \(_3\) (STO) has been reported to display around \(7\%\) plastic deformation under uniaxial compression at an extremely low strain rate \((10^{- 4})^4\) . More recently, good plasticity was reported in semiconductor \(\alpha\) - Ag \(_2\) S and InSe single crystals \(^5,^6\) . In \(\alpha\) - Ag \(_2\) S, excellent plasticity was attributed to planes with weak atomic interactions and irregularly distributed sulfur- silver and silver- silver bonds \(^5\) , while in InSe, the plasticity is thought to result from long- range In- Se Coulomb interactions across the van der Waals gap and soft intralayer In- Se bonding \(^6\) . Flash- sintered TiO \(_2\) has been compressed to \(\sim 10\%\) strain, attributed to a high- density of stacking faults, nanotwins, and dislocations \(^7\) . Plastic deformation observed in nano pillars, nanowires, etc., is mostly attributed to the low chance of smaller samples containing flaws, allowing the materials' intrinsic plasticity to be observed \(^8, 9, 10, 11, 12, 13, 14\) .  
+
+The existence of deformable ceramics has striking potential, but systems that display this characteristic must be identified and plasticity mechanisms need to be understood in order to guide the design of such materials. Because plastic deformation is not typical of ceramics, the applications have not yet been fully considered. It is expected such properties might enable applications such as sensors or even bendable and foldable electronics \(^{15}\) where flexible ceramic film capacitors are required \(^{16}\) .  
+
+Excellent elastic properties are especially desirable \(^{17}\) for functional oxides. A mechanical bending moment enables a dielectric material to polarize, giving rise to flexoelectricity. Flexoelectricity has a strong scaling effect and is therefore significant at micro/nano scales. For this reason, it has the potential to be used for electromechanical actuators and sensors that can be integrated into advanced nano- /micro- electromechanical systems (N/MEMS) \(^{18, 19}\) , meeting the requirement for the millions of micro- and nano- scale sensors to be employed during the expected rapid implementation of the Internet of Things.  
+
+Perovskite oxides are of great interest to both geophysics and materials science \(^{20}\) . In geophysics, a MgSiO \(_3\) - rich perovskite phase is thought to account for \(50 - 90\%\) of the volume of the region of
+
+<--- Page Split --->
+
+the earth that controls seismic activity \(^{21,22}\) (i.e. the \(670\mathrm{km}\) seismic discontinuity to the core- mantle boundary \(^{19}\) ). In the field of materials science, perovskites are of interest because they exhibit useful flexoelectric, dielectric, piezoelectric, ferroelectric, ferromagnetic, multiferroic, superconducting, and photovoltaic properties, as well as colossal magnetoresistance \(^{23}\) . Pb(In \(_{1 / 2}\) Nb \(_{1 / 2}\) )O \(_{3}\) - Pb(Mg \(_{1 / 3}\) Nb \(_{2 / 3}\) )O \(_{3}\) - PbTiO \(_{3}\) (PIN- PMN- PT) is a ternary relaxor ferroelectric perovskite. Single crystal PIN- PMN- PT exhibits outstanding flexoelectric, piezoelectric and electromechanical properties (flexoelectric coupling coefficient \(\mu_{12}\) of \(5 \times 10^{4}\mathrm{nC}\cdot \mathrm{m}^{- 1}\) , piezoelectric coefficient of \(\mathrm{d}_{33}\) \(\sim 2000\mathrm{pC / N}\) and electromechanical coupling factor of \(\mathrm{k}_{33} \sim 90\%\) ) compared to traditional Pb(Zr,Ti)O \(_{3}\) piezoelectric ceramics ( \(\mathrm{d}_{33} < 500\mathrm{pC / N}\) , \(\mathrm{k}_{33} < 75\%\) ) \(^{24,25,26}\) . These extraordinary electromechanical coupling functionalities mean that the mechanical properties are of great interest.  
+
+PIN- PMN- PT samples were first characterised by transmission electron microscopy (TEM). Even prior to deformation experiments, clues to the potential plastic behaviour of PIN- PMN- PT were already apparent. During preparation of thin specimens for TEM, it was noted that the edges of \(\sim 3\mu \mathrm{m}\) tripod- polished samples were not flat (Fig. S1a and b), and a high density of entangled dislocations (Fig. S1) was present in the resulting TEM samples. Compression, tensile and bending tests on PIN- PMN- PT were carried out by using a combination of TEM, scanning electron microscopy (SEM) and nanomechanical test systems. The experimental set up is shown in Fig. S2.  
+
+Because plasticity has previously been observed in nanoscale ceramics during compression, we first tested the properties of our PIN- PMN- PT by preparing round pillars with diameters from 130 nm to \(270\mathrm{nm}\) for compression experiments in a TEM. Results are shown in Figs. 1a – c and Figs. S3 and S4. Fig. 1a shows an engineering stress–strain curve from a \(140\mathrm{nm}\) diameter pillar. The slope of the curve starts to decrease from \(\sim 5\%\) strain. Two short stress plateaus appear when the strain reaches \(\sim 15\%\) and \(\sim 44\%\) respectively, typical of plastic deformation. The total compression strain of the pillar exceeds \(60\%\) , over \(50\%\) of which is plastic. This extreme strain far surpasses the expected deformability of ceramic materials \(^{27}\) and is much higher than has been previously reported in micro/nanopillars \(^{9,10,13}\) . Snapshots captured from a video of the compression are shown in Figs. 1b – d. Slip bands (indicated by yellow arrows) develop on the (011) crystallographic plane, along the \([01\bar{1}]\) direction. Similar phenomena were observed for the other eight pillars with diameters ranging from \(130 \sim 270\mathrm{nm}\) (Figs. S3 and S4).
+
+<--- Page Split --->
+
+As mentioned earlier, previous studies showed that STO single crystals displayed surprisingly high plasticity, with a plastic strain of \(\sim 7\%^{4,28}\) . Here, we compare the compression behaviour of PIN- PMN- PT and STO by also compressing single crystal STO pillars along the same orientation (Figs. S5 – S6). Five out of six STO pillars underwent brittle fracture. The smallest pillar, at \(150\mathrm{nm}\) , was the only one that did not fracture, suggesting that STO undergoes a brittle- to- plastic transition with a critical pillar diameter of around \(150\sim 180\mathrm{nm}\) , while PIN- PMN- PT has significantly better plasticity (all PIN- PMN- PT pillars show plastic deformation). The maximum observed plastic strain was \(17.8\%\) for an STO pillar with the diameter of \(180\mathrm{nm}\) , where PIN- PMN- PT pillars with a similar diameter typically displayed \(>40\%\) strain.  
+
+To determine the effect of pillar size on the deformation behaviour in PIN- PMN- PT, larger pillars were fabricated with diameters ranging from \(500\mathrm{nm}\) to \(2.1\mu \mathrm{m}\) . Most of them displayed plasticity and some were brittle. An engineering stress- strain curve of a \(1\mu \mathrm{m}\) diameter pillar is provided in Figs. 1d – f. Strain bursts were observed, characterized by serrated yielding in the stress- strain curve. Similar rapid bursts of deformation are typical of tests conducted on micrometer- scale metal pillars<sup>29</sup>. Video snapshots in Figs. 1e – f correspond to strain of \(14.3\%\) and \(39.3\%\) , respectively. Slip initiates along the (011) plane and \([01\bar{1}]\) direction, as indicated by the yellow arrow in Fig. 1e. With further deformation, another slip band (110) \([\bar{1} 10]\) is activated, indicated by the red arrow in Fig. 1f and deformation proceeds until the strain reaches \(39.3\%\) . Compression test results from fourteen more pillars with diameters ranging from \(500\mathrm{nm} - 2.1\mu \mathrm{m}\) are shown in Figs. S7 – S8 and another detailed example of extreme deformability for a \(2.1\mu \mathrm{m}\) diameter pillar can be found in Fig. S9 ( \([\bar{1} 10]\) [110] slip and \(39.1\%\) strain). About \(60\%\) , \(50\%\) and \(40\%\) strain were observed in pillars of \(500\mathrm{nm}\) , \(1\mu \mathrm{m}\) and \(2.1\mu \mathrm{m}\) diameters, respectively, as shown in Figs. S7a, S7b, S8c, and S9, which far surpasses the plasticity observed in STO. Figs. 1g – h summarises the results of all compression tests. All samples \(< 700\mathrm{nm}\) diameter underwent plastic deformation, while some larger samples were brittle. Both the strain and the yield strength were typically higher for smaller samples (Figs. \(1\mathrm{g} - \mathrm{h}\) ), consistent with the literature on size effects in metals and ceramic pillars<sup>29, 30, 31, 32</sup>. However, though both STO and PIN- PMN- PT are perovskite oxides, their deformability differs greatly, and the size effects kick in at a much larger size scale for PIN- PMN- PT. This suggests that the intrinsic plasticity of the PIN- PMN- PT is much greater. The elastic compression
+
+<--- Page Split --->
+
+strain is consistent for all samples, at an average of \(3.8\%\) and a maximum of \(6.2\%\) . The composition of PIN- PMN- PT used here is close to the morphotropic phase boundary at which an adaptive ferroelectric phase has been proposed, which can easily transform to other phases upon the mechanical strain. Our previous work shows that a reversible polydomain- rhombohedral to monodomain- orthorhombic phase transition happens under compression \(^{33}\) , which is thought to contribute to the large elastic strain observed here. To better understand the plastic deformation mechanism, deformed samples were further thinned by focused ion beam (FIB) into TEM foils. Scanning transmission electron microscopy – high- resolution high- angle annular dark- field (STEM- HAADF) images from a deformed area (Figs. 1i, j) show climb- dissociated pairs of partial dislocations with Burgers vector \(\frac{1}{2} a< 011>\) , separated by a stacking fault (see also Figs. S11 – S13).  
+
+Dog- bone pillars and cantilever beams were fabricated for tensile and bending tests. The experimental setup is described in Figs. S2d – g. Tensile tests (Fig. S14) of a dog bone sample of dimensions \(1.9 \mu \mathrm{m} \times 0.5 \mu \mathrm{m} \times 0.1 \mu \mathrm{m}\) revealed an elastic strain of \(4.0\%\) , but no plastic deformation. Fig. 2a shows a load- displacement curve obtained from an in- situ bending test. After deformation, the cantilever shows residual plastic deformation \((1.4\%)\) , consistent with an abrupt decrease in mechanical load, indicated by a red arrow in the inset curve in Fig. 2a. Figs 2b – c show video snapshots at maximum load and after unloading (an image prior to bending is shown in Fig. S2g). A maximum flexural strain of \(8.2\%\) , where \(6.8\%\) is elastic and \(1.4\%\) is plastic (details in SI) that occurs at the root of the cantilever beam. Fig. 2d is a low magnification high- resolution STEM- HAADF image taken from the area marked in green in Fig. 2c. Contrast is indicated by red arrows and numbers. Lattice rotation mapping derived from Geometric Phase Analysis (GPA, see methods) displays this contrast more clearly, Fig. 2e, highlighting dislocation cores \(^{34, 35}\) . Dislocations \(1 - 6\) are the same. Fig. 2f is a high- resolution STEM- HAADF image of dislocation #2, which consists of a pair of partial dislocations with Burgers vector of \(\frac{1}{2} a[01\bar{1}]\) and a stacking fault between them, consistent with the defects observed in compressed pillars. Dislocation #7 is different (Fig. S15b – c) and is assumed to be affected by the proximity of the surface of the cantilever beam.
+
+<--- Page Split --->
+
+For perovskite oxides, it is generally accepted that, at ambient temperature, the preferred slip system is \(\{110\} < 1\overline{1} 0>\) , with \(a< 1\overline{1} 0>\) dislocations \(^{19}\) . This type of dislocation is usually dissociated into two partials due to the high energy of two extra atomic planes. Previous studies on as- grown single/double crystals, polycrystals, or thin films show that \(a< 1\overline{1} 0>\) dislocations are dissociated either in a glide or a climb mode \(^{21, 34, 35, 36, 37}\) . Unexpectedly, we have observed climb- dissociated dislocation core structures, which would normally be expected to form at elevated temperatures because climb is a diffusion- assisted process \(^{38}\) . Instead, \(a< 110>\) dislocations formed during room temperature deformation might be expected to dissociate in a slip configuration \(^{38}\) , as was previously reported for compression- tested KNbO \(^{3}\) \(^{39}\) . We note here that a high density of point defects in the PIN- PMN- PT might enable the diffusion that is required to form climb- dissociated dislocations, leading to much better deformability compared to other perovskites such as STO or KNbO \(^{3}\) .  
+
+In perovskites oxides, vacancies are far more common than interstitials \(^{20}\) . \(V_{Pb}^{\prime \prime}\) and \(V_{O}^{- }\) are the most important vacancies in lead- based perovskites, where \(V_{Pb}^{\prime \prime}\) forms due to the volatility of lead at elevated temperature, or donor dopants, while \(V_{O}^{- }\) exists owing to the loss of oxygen at high temperature or acceptor dopants. The face- centred cubic lattice formed by \(\mathrm{Pb}^{2 + }\) and \(\mathrm{O}^{2 - }\) determines the dislocation and slip behavior \(^{20}\) . Consequently, the existence of \(V_{Pb}^{\prime \prime}\) and \(V_{O}^{- }\) in PIN- PMN- PT could considerably influence the observed plasticity. Studies of the effect of vacancies on the deformation behaviour of alloys or intermetallic compounds show varying results: vacancies may facilitate or deteriorate plasticity, depending on their type and distribution \(^{40, 41, 42}\) .  
+
+To trace the possible microscopic origin of the observed extreme plasticity, we conducted first principles atomistic simulations based on density functional theory (DFT). The results are given in Fig. 3. On the basis of a simplified model, PIN- PMN- PT is composed of three sets of subunits, PIN, PMN and PT (Fig. 3a). Atomic- scale Energy- Dispersive X- ray Spectroscopy (EDS) mapping (Fig. S16) indicates that the cations are uniformly distributed at the atomic level, suggesting a high density of mini- interfaces between the three subunits. Relaxed atomic structure and lattice constants of the bulk and interfaces are shown in Figs. S17 & S18, and Table S1 & S2. Calculated interface formation energies (shown in Fig. S19) suggest that the presence of interfaces promote
+
+<--- Page Split --->
+
+the concentration of \(V_{O}^{ - }\) but not \(V_{P b}^{\prime \prime}\) . Favourable \(V_{O}^{ - }\) sites in different side- by- side and top- down interface systems are shown in Fig. S20. Interestingly, these calculations show that it is energetically favourable to form oxygen vacancies (but not lead vacancies) at these interfaces to mitigate the large lattice mismatch (Fig. S21). That is, the three subunits that make up the PIN- PMN- PT naturally facilitate a uniformly- distributed high density of \(V_{O}^{ - }\) . As an example, the atomic structure of 1PIN- 1PMN- 1PT containing one oxygen vacancy is shown in Fig. 3b. To assess the corresponding ductility, we calculated the elastic constants and derived the bulk modulus (B) \(^{43}\) and the anisotropic shear modulus (G) on the (110) plane along \(< 1\bar{1} 0>\) direction for different single tetragonal crystalline species \(^{44}\) , as shown in Fig. 3c and Table S3. The Pugh’s B/G ratio is widely used to index ductility, with a critical value of 1.75 indicating a transition from brittle to ductile behaviour \(^{41,42}\) . For bulk PIN, PMN, and PT, and their pristine interfaces, the calculated B/G ratios are well below 1.75 (hence brittle). By contrast, the B/G ratios for interfaces containing \(V_{O}^{ - }\) are systematically enhanced, most well above 1.75 (hence ductile). Valence charge density analysis reveals that the presence of \(V_{O}^{ - }\) can dramatically weaken the covalent bonding (see Figs. 3d and S22). For comparison, \(V_{P b}^{\prime \prime}\) actually deteriorates the ductility. Thus, based on the DFT results, we attribute the extreme plasticity of PIN- PMN- PT to the high density of \(V_{O}^{ - }\) at the PIN/PMN/PT interfaces (see Fig. S20).  
+
+On the basis of DFT predictions, we investigated the \(V_{O}^{ - }\) levels and mechanical behaviour of PIN- PMN- PT crystals that are expected to be \(V_{P b}^{\prime \prime}\) - rich and \(V_{O}^{ - }\) - rich, (Sm- doped \(^{45}\) and Mn- doped \(^{26}\) crystals respectively), and compared them to the original un- doped PIN- PMN- PT crystal. Electron energy loss spectra (EELS) of O were collected to verify the existence of oxygen vacancies, shown in Fig. 4a. A lower intensity is observed for the O- k edge fine structure peak B compared to A for all three EELS curves. It is known that the peak at position B being lower than the peak at position A is an indication of oxygen deficiency in perovskite oxides \(^{46, 47, 48}\) , suggesting that \(V_{O}^{ - }\) with appreciable concentrations exist in all three samples. Furthermore, the inset image shows that peak B is larger for the Sm- doped sample than the un- doped crystal, indicating a lower \(V_{O}^{ - }\) concentration, and is smaller for the Mn- doped sample, indicating a higher \(V_{O}^{ - }\) fraction.
+
+<--- Page Split --->
+
+According to the DFT predictions, the \(V_{O}^{- }\) - rich (Mn- doped \(^{26}\) ) samples are more likely to be ductile and the \(V_{Pb}^{\prime \prime}\) - rich, (Sm- doped \(^{45}\) ) crystals are more likely to be brittle. Compression tests were performed on both samples. Example engineering stress- strain curves and SEM images of compressed pillars are shown in Fig. 4b (details in Figs. S23 – S24). Six \(\sim 600 \mathrm{nm}\) diameter pillars were fabricated for each sample type. All Mn- doped PIN- PMN- PT pillars showed plasticity, while half of the Sm- doped PIN- PMN- PT pillars underwent brittle fracture, indicating that the Mn- doped sample had superior plasticity. In the examples shown in Fig. 4b, the Sm- doped sample has fractured in a brittle way, while the Mn- doped sample has slip bands on the pillar and a stress plateau and strain burst on the stress- strain curve. The results of this comparison experiment are consistent with our hypothesis of \(V_{O}^{- }\) - induced plasticity.  
+
+It has been proposed by Zubko et al. that dislocations contribute significantly to flexoelectricity in \(\mathrm{STO}^{49}\) . Tang et al. and Gao et al. measured the strain gradient around dislocations by extracting Bi/Sr positions from STEM- HAADF images and calculating the flexoelectric polarization in multiferroic \(\mathrm{BiFeO_3}\) and paraelectric \(\mathrm{STO}^{50,51}\) , which was found to be several \(\mu \mathrm{C}\cdot \mathrm{cm}^{- 2}\) . The flexoelectric effect is expected to be extremely large, because relaxor ferroelectric PIN- PMN- PT shows outstanding flexoelectricity compared to other perovskite oxides. Take STO as an example, the flexoelectric coefficient \(\mu_{12}\) is about \(7 \mathrm{nC}\cdot \mathrm{m}^{- 1}24,25,49\) , while that of PIN- PMN- PT is about \(5.0 \times 10^{4} \mathrm{nC}\cdot \mathrm{m}^{- 1}24\) , a difference of 4 orders of magnitude.  
+
+Here, in order to measure the flexoelectric polarization around a pair of partial dislocations (introduced by plastic deformation), we extracted Pb atom positions from STEM- HAADF images firstly (details in SI) and calculated the maximum strain gradient \((\nabla S)\) to be about \(3.5 \times 10^{9} \mathrm{m}^{- 1}\) ([0 \(\overline{1} 1\) ] lattice strain gradient along the \([0 \overline{1} \overline{1} ]\) direction), which is 3 times that reported by Gao et al. around [010] dislocations in a STO bicrystal \(^{50}\) . Supposing the flexoelectric coefficient of PIN- PMN- PT [110] is comparable to that of \(\mu_{12}^{24}\) , the local flexoelectric polarization (1\~2 unit- cells) around dislocations is estimated to be about \(10^{7} \mu \mathrm{C}\cdot \mathrm{cm}^{- 2}\) according to \(P_{f} = u \times \nabla S\) , where \(P_{f}\) is flexoelectric polarization, \(u\) is flexoelectric coefficient, and \(\Delta S\) is gradient of the horizontal lattice constant along the vertical direction. However, this large calculated polarization is thought to be an over- estimate for two reasons. 1) In the case of such high strain gradients, higher- order coupling
+
+<--- Page Split --->
+
+terms of flexoelectric polarization and strain gradient, which is nonlinear, should not be neglected, and the magnitude of those terms is still unclear. 2) For smaller samples, permittivity \((\epsilon)\) is expected to decrease as a result of a size effect \(^{51}\) , and the flexoelectric coefficient \(\mu\) , which is a function of \(\epsilon\) in a manner of \(\mu = f \cdot \epsilon\) , should also be smaller than the corresponding bulk value (here \(f\) is flexo- coupling coefficient, about 10 V for PTO- based relaxor ferroelectrics). However, this extremely large polarization should give rise to a large number of bound charges. To screen these bound charges, free charges will accumulate. Transport properties or even magnetic properties around these dislocations can also be affected due to free charges. For slip bands, where a strain gradient also exists (as shown in Fig. S11e), the situation would be similar. As the strain gradient around a slip band is much smaller than it is around dislocations, the flexoelectric effect will be smaller. The movement of dislocations and the introduced slip bands make a functional region which is potentially applicable for flexoelectric based micro- and nano- scale electronic devices.  
+
+In addition to strain gradients around dislocations and slip bands, bending- induced elastic strain gradients are also of great interest for flexoelectricity because of their reversibility. The maximum elastic strain introduced by bending test is calculated to be \(6.8\%\) at the root of the cantilever beam, and the width \((b)\) of the cantilever beam is \(0.67 \mu \mathrm{m}\) , which gives rise to a strain gradient of about \(2 \times 10^{5} \mathrm{m}^{- 1} (\nabla S = \frac{6.8\%}{0.335 \mu \mathrm{m}} \approx 2 \times 10^{5} \mathrm{m}^{- 1})\) . Flexoelectric polarization from the elastic bending strain gradient is estimated to be about \(1 \times 10^{3} \mu \mathrm{C} \cdot \mathrm{cm}^{- 2}\) . Here strain gradient \(\nabla S\) is the horizontal ([001]) lattice strain gradient along the vertical direction ([010]). The calculated flexoelectric polarization is \(1 \sim 2\) orders of magnitude larger than the ferroelectric polarization of known ferroelectrics. For example, the ferroelectric polarization of \(\mathrm{PbTiO_3}\) is about \(75 \mu \mathrm{C} \cdot \mathrm{cm}^{- 2}\) , the ferroelectric polarization of \(\mathrm{BiFeO_3}\) is around \(90 \mu \mathrm{C} \cdot \mathrm{cm}^{- 2}\) , and the polarization of \(\mathrm{BaTiO_3}\) is about \(26 \mu \mathrm{C} \cdot \mathrm{cm}^{- 2}\) . \(^{52,53,54}\) The calculated flexoelectric polarization is also \(4 \sim 5\) times that of the recently- reported ferroelectric polarization of super- tetragonal \(\mathrm{PbTiO_3}^{55}\) . An even larger flexoelectric polarization would be expected if a lower strain rate is used, according to Deng's work \(^{56}\) . This large flexoelectric polarization is also likely to be an overestimate for reasons mentioned above. However, even if the real flexoelectric polarization is \(1 / 10\) of the calculated \(1 \times 10^{3} \mu \mathrm{C} \cdot \mathrm{cm}^{- 2}\) , it is still large enough
+
+<--- Page Split --->
+
+(100 μC·cm⁻²) to switch the local ferroelectric polarization, and to be used in flexoelectric based sensors.  
+
+The excellent deformability in \(V_{O}^{- }\) - rich PIN- PMN- PT is particularly promising for flexoelectric- based sensors, because it was reported that the effective flexoelectricity of oxygen- depleted perovskite oxide is two orders of magnitude larger than for a stoichiometric sample<sup>57</sup>. Combined with the scaling effects of flexoelectricity and super large flexoelectric coefficient of PIN- PMN- PT, these provide exciting opportunities for high performance flexoelectric based N/MEMS devices.  
+
+To summarize, we have revealed extreme deformability in relaxor ferroelectric PIN- PMN- PT micron/submicron single crystals pillars. A maximum elastic strain of \(>6\%\) and plastic strain \(>50\%\) were observed during compression tests, while a flexural strain of \(8.2\%\) was achieved for a bent cantilever beam. Pairs of \(\frac{1}{2} a< 011>\) climb- dissociated partial dislocations accommodate the plastic deformation. Based on first principles calculations, confirmed by experiments, we propose that the observed excellent plasticity is attributed to not only a decrease in the specimen size, but also a high \(V_{O}^{- }\) concentration. This suggests that it might be possible to alter the plasticity of ceramic materials by deliberate engineering of point defects, which paves the way towards the design of ductile ceramics, and implies that more attention should be paid to the previously ignored mechanical properties of functional oxides. The giant strain gradient generated by elastic bending and dislocations gives rise to considerable flexoelectric polarization, which can be used in sensors. These results will facilitate the development of flexoelectric- based flexible electronic devices and N/MEMS.
+
+<--- Page Split --->
+![](images/Figure_unknown_0.jpg)
+
+<center>Figures and Captions: </center>  
+
+Fig. 1 | Compression tests of sub- micro and micrometre scale pillars. a, An engineering stress – strain curve acquired during the compression of \(140 \mathrm{nm}\) diameter pillar, with a loading direction along [010]. b – c, Snapshots from a real time video recording of a compression test, at strains of \(18.7\%\) and \(60.1\%\) , respectively (labelled as yellow circles in a). Slip bands along (011) crystallographic plane and [011] direction is indicated by yellow arrows. Here, both slip plane and slip direction are determined from the change of contrast in TEM images. d, An engineering stress – strain curve from a compression test of a \(1 \mu \mathrm{m}\) pillar. e ��� f, Video snapshots corresponding to
+
+<--- Page Split --->
+
+strains of \(14.3\%\) and \(39.3\%\) (yellow circles in d). Two slip bands (oriented \((110)[01\bar{1}]\) and \((1\bar{1} 0)[110])\) are indicated by yellow and red arrows. \(\mathbf{g}\) , Strain as a function of pillar diameter, showing plastic strain (hollow circles) and total strain (spheres) for plastic- deformed pillars and fracture strain (red crosses) for brittle- fracture samples. \(\mathbf{h}\) , Yield strength as a function of pillar diameter, showing yield strength for plastic- deformed pillars (spheres) and brittle- fracture pillars (red crosses). Dashed black curve: fitted yield strength – diameter curve for plastic- deformed pillars, with a function of \(y = 56.9x^{-0.52}\) . Green arrows indicate the strain/stress value corresponding to the pillars shown in Fig. 1a – c and d – f. \(\mathbf{i} - \mathbf{j}\) , STEM- HAADF images showing pairs of partial dislocations with Burgers vectors of \(\frac{1}{2} a[011]\) (i) and \(\frac{1}{2} a[0\bar{1}\bar{1}]\) (j). The partial dislocations are separated by stacking faults.  
+
+![](images/Figure_2.jpg)
+
+<center>Fig. 2 | Bending test of PIN-PMN-PT. a, A load-depth curve obtained during bending a cantilever beam along a loading direction of [010]. The inset shows the enlarged curve of the rectangular area, where an abrupt decrease in mechanical load is evident. b, Snapshot captured from in-situ video corresponding to the maximum depth of the indenter. c, An SEM image showing cantilever beam after unloading. Irreversible deformation can be clearly revealed by comparison with Fig. S2g. d – e, A STEM-HAADF image and the corresponding GPA analysis of lattice rotation, with </center>
+
+<--- Page Split --->
+
+dislocations indicated by red arrows. Dislocations #1 and #7 are labelled. f, High resolution STEM- HAADF images showing dislocation #2, which includes a pair of partial dislocations with Burgers vector of \(\frac{1}{2} a[01\bar{1}]\) .  
+
+![](images/Figure_3.jpg)
+
+<center>Fig. 3 | First principles atomistic investigation for the plasticity in PIN-PMN-PT. a, Three sub-unit cells for \(\mathrm{Pb(In_{1 / 2}Nb_{1 / 2})O_{3}}\) (PIN), \(\mathrm{Pb(Mg_{1 / 3}Nb_{2 / 3})O_{3}}\) (PMN) and \(\mathrm{PbTiO_{3}}\) (PT). b, An example of relaxed atomic structure containing interfaces formed by one PIN, one PMN and one PT with one oxygen vacancy \((V_{O}^{-})\) at the interface of PIN and PMN. c, Calculated bulk modulus/shear modulus (B/G) ratios for various bulk, pristine interfaces, and interfaces with \(V_{O}^{-}\) . Higher B/G ratios (>1.75) suggest ductile behaviour in PIN-PMN-PT. d, Calculated valence charge density 2D contour plot (colours assigned recursively) on the (020) plane of the structure shown in b. The strength of covalent bonding is indicated by the colour bar. The presence of \(V_{O}^{-}\) eliminates the local covalent bonds. </center>
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Fig. 4 | Analysis of the origin of extreme plasticity. a, O-EELS obtained from Mn- (blue), Sm- (green) doped and undoped (red) PIN-PMN-PT. The enlarged image (inset) shows the intensity difference of peak B for Mn-, Sm- doped and undoped PIN-PMN-PT, where a lower intensity suggests a higher \(V_{O}^{*}\) concentration. b, Engineering stress – strain curves obtained from in-situ compression tests of Mn- (blue sphere) and Sm- (green circle) doped PIN-PMN-PT. Images I and II show SEM images of the compressed Mn-doped and Sm-doped PIN-PMN-PT respectively. </center>
+
+<--- Page Split --->
+
+## Acknowledgment:  
+
+Acknowledgment:The authors are grateful for the scientific and technical support from the Australian Centre for Microscopy and Microanalysis (ACMM) as well as the Microscopy Australia node at the University of Sydney. Thanks Dr Xianghai An from the University of Sydney for the fruitful discussion on the mechanical behaviour of materials. We are grateful for A/Prof. John Daniels, and PhD candidates Fan Ji and Tongzheng Xin from the University of New South Wales for helpful discussions regarding oxygen vacancies in perovskite oxides, and Dr. Jun Luo from TRS Technologies for providing single crystal samples. Thanks Prof. Gustau Catalan from Catalan Institute of Nanoscience and Nanotechnology (ICN2) for the discussion of flexoelectricity. This work was supported by the Australian Federal Government through the Next Generation Technologies Fund, and the DST Strategic Research Initiative in Advanced Materials and Sensors. We also acknowledge the assistance and high- performance computing (HPC) resources from the National Computational Infrastructure and the expert HPC facilitation from the team at the Sydney Informatics Hub at the University of Sydney. The authors would like to acknowledge the United States Office of Naval Research (ONR) and ONR Global for partially supporting this work.  
+
+## Author contributions:  
+
+S.M., J.C. and P.F. initiated studies into the nano and micro structural evolution under mechanical loading of PIN- PMN- PT nano plates. Y.L. and J.C. proposed the mechanical property experiments. Y.L., S.Z. and J.C. designed the experiment. Y.L. fabricated pillars, conducted in- situ experiment for compression and bending tests, and carried out aberration corrected (S)TEM investigation (TEM/STEM/EDS/EELS). R.N. prepared tensile test samples and conducted in- situ tensile tests. M.G. supported in the acquisition and analysis of aberration- corrected (S)TEM images and spectroscopic data. X.Y.C. and S.P.R. designed and conducted first- principles simulation. S.Z., S.M. and P.F. provided single crystal samples. J.C. and X.L. supervised the research. All authors contributed to the discussions and manuscript preparation.  
+
+Competing interests: The authors declare no competing interests.
+
+<--- Page Split --->
+
+## References:  
+
+1. Callister, W. D. & Rethwisch, D. G. Materials Science and Engineering: An Introduction (John Wiley & Sons, United States of America, 2009).  
+2. Pelleg, J. Mechanical Properties of Ceramic, Solid Mechanics and Its Applications, (Springer, Switzerland, 2014).  
+3. Wachtman, J. B., Cannon, W. R. & Matthewson, M. J. Mechanical properties of ceramics (John Wiley & Sons, United States of America, 2009).  
+4. Brunner, D., Taeri-Baghbadrani, S., Sigle, W. & Ruhle, M. Surprising results of a study on the plasticity in strontium titanate. J. Am. Ceram. Soc. 84, 1161–1163 (2001).  
+5. Shi, X., et al. Room-temperature ductile inorganic semiconductor. Nat. Mater. 17, 421–426 (2018).  
+6. Wei, T. R., et al. Exceptional plasticity in the bulk single-crystalline van der Waals semiconductor InSe. Science 369, 542–545 (2020).  
+7. Li, J., et al. Nanoscale stacking fault–assisted room temperature plasticity in flash-sintered TiO₂. Sci. Adv. 5, eaaw5519 (2019).  
+8. Chen, Y.J., An, X.H. & Liao, X.Z. Mechanical behaviors of nanowires. Appl. Phys. Rev. 4, 031104 (2017).  
+9. Michler, J., Wasmer, K., Meier, S., Ostlund, F. & Leifer, K. Plastic deformation of gallium arsenide micropillars under uniaxial compression at room temperature. Appl. Phys. Lett. 90, 043123 (2007).  
+10. Ostlund, F., et al. Brittle-to-Ductile Transition in Uniaxial Compression of Silicon Pillars at Room Temperature. Adv. Funct. Mater. 19, 2439–2444 (2009).  
+11. Han, X. D., et al. Low-temperature in situ large strain plasticity of ceramic SiC nanowires and its atomic-scale mechanism. Nano Lett. 7, 452–457 (2007).  
+12. Wang, Y. B., et al. Super Deformability and Young's Modulus of GaAs Nanowires. Adv. Mater. 23, 1356–1360 (2011).  
+13. Kiani, S., et al. Dislocation glide-controlled room-temperature plasticity in 6H-SiC single crystals. Acta Mater. 80, 400–406 (2014).  
+14. He, Y., Zhong, L., Fan, F., Wang, C., Zhu, T. & Mao, S. X., In situ observation of shear-driven amorphization in silicon crystals. Nat. Nanotechnol. 11, 866–871 (2016).
+
+<--- Page Split --->
+
+15. An, H. S., et al. Surface-agnostic highly stretchable and bendable conductive MXene multilayers. Sci. Adv. 4, eaaq0118 (2018).  
+
+16. Ma, B., Lee, T. H., Dorris, S. E., Korital, R. E. & Balachandran, U. Flexible ceramic film capacitors for high-temperature power electronics. Mater. Sci. Energy Technol. 2, 96-103 (2019).  
+
+17. Banerjee, A., et al. Ultralarge elastic deformation of nanoscale diamond. Science, 360, 300-302 (2018).  
+
+18. Wang, B., Gu, Y., Zhang, S. & Chen, L-Q. Flexoelectricity in Solids: Progress, Challenges, and Perspectives. Prog. Mater. Sci. 106, 100570 (2019).  
+
+19. Bhaskar, U. K., et al. A flexoelectric microelectromechanical system on silicon. Nature nanotechnology. 11, 263-266 (2016).  
+
+20. Crary, A. P., et al. Perovskite: A Structure of Great Interest to Geophysics and Materials Science (American Geophysical Union, Washington, D.C., 1989).  
+
+21. Ferre, D., Carrez, P. & Cordier, P. Modeling dislocation cores in SrTiO₃ using the Peierls-Nabarro model. Phys. Rev. B 77, 014106 (2008).  
+
+22. Miyagi, L., Kanitpanyacharoen, W., Kaercher, P., Lee, K. K. M. & Wenk, H.R. Slip Systems in MgSiO₃ Post-Perovskite: Implications for D" Anisotropy. Science 329, 1639-1641 (2010).  
+
+23. Tilley, R. J. D. Perovskites: Structure-Property Relationships (John Wiley & Sons, chichester, 2016).  
+
+24. Shu, L., Liang R., Rao, Z., Fei, L., Ke, S. & Wang, Y. Flexoelectric materials and their related applications: A focused review. J. Adv. Ceram. 8: 153-173 (2019).  
+
+25. Shu, L. et al. Flexoelectric behavior in PIN-PMN-PT single crystals over a wide temperature range. Appl. Phys. Lett. 111: 162901 (2017)  
+
+26. Zhang, S. J. & Li, F. High performance ferroelectric relaxor-PbTiO₃ single crystals: Status and perspective. J. Appl. Phys. 111, 031301 (2012).  
+
+27. Green, D. J. An Introduction to the Mechanical Properties of Ceramics. (Cambridge University Press, Cambridge, 1998).  
+
+28. Gumbsch, P., Taeri-Baghbadrani, S., Brunner, D., Sigle, W. & Ruhle, A. Plasticity and an inverse brittle-to-ductile transition in strontium titanate. Phys. Rev. Lett. 87, 085505 (2001).  
+
+29. Uchic, M. D., Dimiduk, D. M., Florando, J. N. & Nix, W. D. Sample dimensions influence strength and crystal plasticity. Science 305, 986-989 (2004).
+
+<--- Page Split --->
+
+30. Greer, J. R. & Hosson, J. T. M. D. Plasticity in small-sized metallic systems: Intrinsic versus extrinsic size effect. Prog. Mater. Sci. 56, 654–724 (2011).  
+
+31. Kiani, S., Yang, J. M. & Kodambaka, S. Nanomechanics of refractory transition-metal carbides: a path to discovering plasticity in hard ceramics. J. Am. Chem. Soc. 98, 2313–2323 (2015).  
+
+32. Zhu, T. & Li, J. Ultra-strength materials. Prog. Mater. Sci. 55, 710–757 (2010).  
+
+33. Liu, Y., Xia, J., Finkel, P., Moss, D. S., Liao, X. & Cairney, M. J. Real-time observation of stress-induced domain evolution in a [011] PIN-PMN-PT relaxor ferroelectric single crystal. Acta Mater. 175, 436–444 (2019).  
+
+34. Liu, Y., Tang, Y. L., Zhu, Y. L., Wang, W. Y. & Ma, X. L. Spatial Coupling of Ferroelectric Domain Walls and Crystallographic Defects in the PbTiO₃ Films. Adv. Mater. Interfaces 3, 1600342 (2016).  
+
+35. Tang, Y. L., Zhu, Y. L., Liu, Y., Wang, Y. J., Ma, X. L. Giant linear strain gradient with extremely low elastic energy in a perovskite nanostructure array. Nat. Commun., 8, 15994 (2017).  
+
+36. Zhang, Z. L., Sigle, W., Kurtz, W. & Ruhle, M. Electronic and atomic structure of a dissociated dislocation in SrTiO₃. Phys. Rev. B 66, 214112 (2002).  
+
+37. Mao, Z. & Knowles, K. M. Dissociation of lattice dislocations in SrTiO₃. Philos. Mag. A, 73, 699–708 (1996).  
+
+38. Matsunaga, T., Saka, H. Transmission electron microscopy of dislocations in SrTiO₃. Phil. Mag. Lett., 80, 597–604 (2000).  
+
+39. Mark, A. F., Castillo-Rodriguez, M. & Sigle, W. Unexpected plasticity of potassium niobate during compression between room temperature and 900 degrees C. J. Eur. Ceram. Soc. 36, 2781–2793 (2016).  
+
+40. Salehinia, I., Medyanik, S. N. Effects of Vacancies on the Onset of Plasticity in Metals-An Atomistic Simulation Study. Metall. Mater. Trans. A, 42a, 3868–3874. (2011).  
+
+41. Pan, Y., Lin, Y.H., Wang, H. & Zhang, C.M. Vacancy induced brittle-to-ductile transition of Nb<sub>5</sub>Si<sub>3</sub> alloy from first-principles. Mater. Design 86, 259–265 (2015).  
+
+42. Kindlund, H., et al. Vacancy-induced toughening in hard single-crystal V<sub>0.5</sub>Mo<sub>0.5</sub>N<sub>x</sub>/MgO (001) thin films. Acta Mater. 77, 394–400 (2014).  
+
+43. Liu, Y. et al. First-principles study of elastic properties in perovskite PbTiO<sub>3</sub>. Mater. Sci. Eng. A 472, 269–272 (2008).
+
+<--- Page Split --->
+
+44. He, Y., Schwarz, R. B., Migliori, A. & Whang, S. H. Elastic constants of single crystal \(\gamma\) – TiAl. J. Mater. Res. 10, 1187–1195 (1995).  
+
+45. Li, F., et al. Giant piezoelectricity of Sm-doped Pb(Mg<sub>1/3</sub>Nb<sub>2/3</sub>)O<sub>3</sub>-PbTiO<sub>3</sub> single crystals. Science, 364, 264–268 (2019).  
+
+46. Muller, D. A., Nakagawa, N., Ohtomo, A., Grazul, J. L. & Hwang, H. Y. Atomic-scale imaging of nanoengineered oxygen vacancy profiles in SrTiO<sub>3</sub>. Nature 430, 657–661 (2004).  
+
+47. Klie, R. F., Browning, N. D. Atomic scale characterization of oxygen vacancy segregation at SrTiO<sub>3</sub> grain boundaries. Appl. Phys. Lett. 77, 3737–3739 (2000).  
+
+48. Zhang, Z. L., Sigle, W., Ruhle, M. Atomic and electronic characterization of the \(a[100]\) dislocation core in SrTiO<sub>3</sub>. Phys. Rev. B 66, 094108 (2002).  
+
+49. Zubko, P., Catalan, G., Buckley, A., Welche, P. R. L. & Scott, J. F. Strain-Gradient-Induced Polarization in SrTiO<sub>3</sub> Single Crystals. Phys. Rev. lett. 99, 167601 (2007).  
+
+50. Gao, P., et al. Atomic-Scale Measurement of Flexoelectric Polarization at SrTiO<sub>3</sub> Dislocations. Phys. Rev. Lett. 120, 267601 (2018).  
+
+51. Saad, M. M., et al. Investigating the effects of reduced size on the properties of ferroelectrics. IEEE Trans. Ultrason. Ferroelectr. Freq. Control. 53, 2208–2225 (2016).  
+
+52. Tang, Y. L., et al. Observation of a periodic array of flux-closure quadrants in strained ferroelectric PbTiO<sub>3</sub> films. Science 348, 547–551 (2005).  
+
+53. Wang, J., et al. Epitaxial BiFeO<sub>3</sub> Multiferroic Thin Film Heterostructures. Science 299, 1719–1722 (2003).  
+
+54. Jaffe, B., Cook Jr., W. R. & Jaffe, H. Piezoelectric Ceramics. Academic Press, London and New York, p. 78 (1971).  
+
+55. Zhang, L. et al. Giant polarization in super-tetragonal thin films through interphase strain. Science 361: 494–497 (2018).  
+
+56. Deng, Y., et al. Hierarchically-structured large superelastic deformation in ferroelastic-ferroelectrics. Acta Mater. 181, 501–509 (2019).  
+
+57. Narvaez J, Vasquez-Sancho F, Catalan G. Enhanced flexoelectric-like response in oxide semiconductors. Nature 538, 219–221 (2016).
+
+<--- Page Split --->
+
+## Methods  
+
+## Materials:  
+
+The experimental work reported in this paper was performed using [011] poled PIN- PMN- PT single crystal plates (CTS Advanced Materials, with nominal composition 0.24PIN- 0.44PMN- 0.32PT, grown via the modified Bridgeman method) with MPB composition, a relative permittivity of 4000, dimensions of \(12 \times 4 \times 4 \mathrm{mm}^3\) , and surface polished to \(50 - 110 \mathrm{nm}\) . Sm- doped and Mn- doped PIN- PMN- PT (TRS Technologies) single crystals were grown by a modified Bridgeman method and STO is commercial single crystal.  
+
+## Sample preparation:  
+
+Micro- pillars preparation for compression, tensile and bending tests: The PIN- PMN- PT single crystal was first cut into slices of \(0.5 \mathrm{mm}\) in thickness, then further thinned using tripod polishing to \(\sim 500 \mathrm{nm}\) at the front edge. Pillars used for in- situ tests were fabricated at the thin edge by using FIB. Columnar pillars with an aspect ratio (height/diameter) of \(2:1 \sim 3:1\) were prepared for compression tests. The FIB was operated at \(30 \mathrm{kV}\) using a current of \(1 \mathrm{nA}\) for coarse milling and \(5 \mathrm{pA} \sim 300 \mathrm{pA}\) for final milling of pillars with diameters ranging from \(130 \mathrm{nm} \sim 2.1 \mu \mathrm{m}\) . The pillar taper angles are estimated to be around \(3^{\circ}\) . The diameter of the top surface was used for stress calculation, which is the first part of the sample to undergo plastic deformation. Cantilever beams for bending tests were prepared with FIB operating at \(30 \mathrm{kV}\) and using a current of \(50 \mathrm{pA}\) for final milling. The length, width and depth are \(6.5\) , \(0.67\) and \(0.8 \mu \mathrm{m}\) , respectively. Dog bone shaped pillars were prepared for tensile tests, and \(30 \mathrm{kV}\) , \(5 \mathrm{pA}\) were used for final milling.  
+
+TEM sample preparation: The deformed pillars were lifted- out using a tungsten manipulator onto a copper base, and then thinned to electron transparency ( \(\sim 50 \mathrm{nm}\) ) for TEM observation. \(10 \mathrm{kV}\) and \(10 \mathrm{pA}\) were used for FIB final milling. \(5 \mathrm{kV}\) , \(10 \mathrm{pA}\) and \(2 \mathrm{kV}\) , \(10 \mathrm{pA}\) were used for final cleaning of the surface. To protect the pillars from FIB damage, platinum was deposited around the pillars before thinning.  
+
+TEM sample preparation for O- K EELS: TEM samples for O- K EELS were prepared by grinding using tripod polisher and ion milling employing a Gatan precision ion polishing system II (PIPS II). \(4^{\circ}\) and \(0.5 \mathrm{kV}\) were used for final milling.  
+
+## In-situ mechanical tests:
+
+<--- Page Split --->
+
+In- situ compression experiments were carried out in both a TEM (JEOL JEM 2100) and an SEM (Zeiss Ultra), while in- situ tensile and bending tests were conducted in the SEM. The JEOL JEM 2100 uses a high brightness LaB6 electron source. It is equipped with Xarosa (4 k x 4 k) as well as Veleta Ultrascan (2 k x 2 k) cameras. In the TEM, in- situ compression tests of pillars with diameters around \(200\mathrm{nm}\) were carried out by using a Hysitron PI 95 Picoindenter with a flat diamond tip. As the load applied is limited to \(1.5\mathrm{mN}\) for the PI 95 Picoindenter, the requirement for thin sample in the TEM, we carried out the in- situ compression experiment of the larger pillars by using a Hysitron PI 85L picoindenter inside an SEM, with a specially designed system for applying loads up to \(10\mathrm{mN}\) . This system allows real- time observation of deformation process (i.e. slip band development, slip planes and slip directions). Load was applied to pillars by moving the indenter toward the pillars in the displacement control mode. The displacement rates were \(1\mathrm{nm}\cdot \mathrm{s}^{- 1}\) and \(2\mathrm{nm}\cdot \mathrm{s}^{- 1}\) for compression of pillars of around \(200\mathrm{nm}\) in diameter and from \(500\mathrm{nm}\sim 2.1\mu \mathrm{m}\) in diameter, respectively. For the tensile test, a displacement rate of \(1\mathrm{nm}\cdot \mathrm{s}^{- 1}\) was used. For the bending test, a higher displacement rate - \(4\mathrm{nm}\cdot \mathrm{s}^{- 1}\) was used.  
+
+## Microstructure investigation of the deformed pillars:  
+
+A JEOL JEM 2100 TEM and a FEI Themis- Z Double- corrected 60- 300 kV S/TEM were used to observe the compressed pillars. High- resolution STEM- HAADF images, EDS element mapping and O- K edge EELS were acquired using the FEI Themis- Z S/TEM. The convergence and collection angle under the STEM- HAADF mode are 17.9 mrad and \(50 - 200\mathrm{mrad}\) , respectively. Strain was analysed using free Geometric Phase Analysis script (by C.T. Koch) \(^{58}\) . EELS of the O- K edge was acquired under the TEM mode at a collection angle of \(100\mathrm{mrad}\) . Dual- EELS was used and zero peak was corrected for all three samples. The energy resolution is estimated to be \(1.0\mathrm{eV}\) , measured from full width at half maxima of zero loss peak, while an energy dispersion of \(0.025\mathrm{eV / ch}\) was employed. The point resolution of Themis- Z under the STEM mode is around \(0.6\mathrm{\AA}\) (operated at \(300\mathrm{kV}\) ). It is equipped with X- FEG high- brightness gun, Monochromator, ChemiSTEM (Super- X) EDS detectors as well as a Gatan Quantum ER/965 GIF (<0.14 eV (1s)) with Dual- EELS.  
+
+First- principles simulation: DFT calculations were performed using the plane- wave pseudopotential total energy method as implemented in the VASP code \(^{59,60}\) . Projector augmented wave potentials \(^{61}\) and the generalized gradient approximation \(^{62}\) were used for exchange- correlation. A plane- wave basis set was used with an energy cut off of \(500\mathrm{eV}\) . The summation
+
+<--- Page Split --->
+
+over the Brillouin zone for the bulk structures was performed on a \(\sim 0.06 \mathrm{\AA}^{- 1}\) spacing Monkhorst- Pack \(\mathbf{k}\) - point mesh for all calculations. For all systems, atomic relaxation was allowed until all the forces were less than \(0.01 \mathrm{eV / \AA}\) . For charge density calculations, \(\mathrm{Pb - 5d}\) , \(\mathrm{Nb - 4p}\) , \(\mathrm{Mg - 2p}\) , \(\mathrm{Ti - 3p}\) and \(\mathrm{In - 4d}\) semi- core states were treated as valence states to ensure high accuracy. Additional computational details can be found in the Supporting Information.  
+
+## References  
+
+58. Hytch, M. J., Snoeck, E., Kilaas, R. Quantitative measurement of displacement and strain fields from HREM micrographs. Ultramicroscopy 74, 131-146 (1998).  
+
+59. Kresse, G. & Furthmuller, J. Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set. Comp. Mater. Sci. 6, 15-50 (1996).  
+
+60. Kresse, G. & Furthmuller, J. Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Phys. Rev. B 54, 11169-11186 (1996).  
+
+61. Kresse, G. & Joubert, D. From ultrasoft pseudopotentials to the projector augmented-wave method. Phys. Rev. B 59, 1758-1775 (1999).  
+
+62. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation made simple. Phys. Rev. Lett. 77, 3865-3868 (1996)
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+SupportingInformation20210226. docx
+
+<--- Page Split --->

preprint/preprint__0fc91687659d1cbd096c1f45e57db497948054dbc6edf1d233f7af3f1cd54319/preprint__0fc91687659d1cbd096c1f45e57db497948054dbc6edf1d233f7af3f1cd54319_det.mmd

@@ -0,0 +1,423 @@
+<|ref|>title<|/ref|><|det|>[[45, 108, 852, 175]]<|/det|>
+# Extreme Room Temperature Compression and Bending in Ferroelectric Oxide Pillars  
+
+<|ref|>text<|/ref|><|det|>[[44, 196, 630, 280]]<|/det|>
+Ying Liu The University of Sydney Xiangyuan Cui The University of Sydney https://orcid.org/0000- 0002- 3946- 7324  
+
+<|ref|>text<|/ref|><|det|>[[44, 290, 235, 330]]<|/det|>
+Ranming Niu University of Sydney  
+
+<|ref|>text<|/ref|><|det|>[[44, 336, 720, 377]]<|/det|>
+Shujun Zhang University of Wollongong, Australia https://orcid.org/0000- 0001- 6139- 6887  
+
+<|ref|>text<|/ref|><|det|>[[44, 382, 592, 423]]<|/det|>
+Xiaozhou Liao University of Sydney https://orcid.org/0000- 0001- 8565- 1758  
+
+<|ref|>text<|/ref|><|det|>[[44, 429, 404, 470]]<|/det|>
+Scott Moss Defence Science and Technology Group  
+
+<|ref|>text<|/ref|><|det|>[[44, 475, 323, 515]]<|/det|>
+Peter Finkel US Naval Research Laboratory  
+
+<|ref|>text<|/ref|><|det|>[[44, 521, 273, 562]]<|/det|>
+Magnus Garbrecht The University of Sydney  
+
+<|ref|>text<|/ref|><|det|>[[44, 568, 630, 608]]<|/det|>
+Simon Ringer The University of Sydney https://orcid.org/0000- 0002- 1559- 330X  
+
+<|ref|>text<|/ref|><|det|>[[44, 613, 630, 654]]<|/det|>
+Julie Cairney ( julie.cairney@sydney.edu.au ) The University of Sydney https://orcid.org/0000- 0003- 4564- 2675  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 696, 102, 713]]<|/det|>
+## Article  
+
+<|ref|>text<|/ref|><|det|>[[44, 733, 857, 753]]<|/det|>
+Keywords: Plastic Deformation, Ceramic Materials, Perovskite Oxide, Flexoelectric Polarization  
+
+<|ref|>text<|/ref|><|det|>[[44, 772, 288, 790]]<|/det|>
+Posted Date: June 1st, 2021  
+
+<|ref|>text<|/ref|><|det|>[[44, 809, 463, 828]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 342103/v1  
+
+<|ref|>text<|/ref|><|det|>[[44, 847, 909, 890]]<|/det|>
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[42, 45, 940, 88]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on January 17th, 2022. See the published version at https://doi.org/10.1038/s41467-022-27952-2.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 89, 836, 109]]<|/det|>
+## Extreme Room Temperature Compression and Bending in Ferroelectric Oxide Pillars  
+
+<|ref|>text<|/ref|><|det|>[[115, 120, 881, 155]]<|/det|>
+Y. Liu \(^{1,2}\) , 
+X.Y. Cui \(^{1,2}\) , 
+R.M. Niu \(^{1}\) , 
+S.J. Zhang \(^{3}\) , 
+X.Z. Liao \(^{1}\) , 
+S. Moss \(^{4}\) , 
+P. Finkel \(^{5}\) , 
+M. Garbrecht \(^{2}\) , 
+S.P. Ringer \(^{1,2}\) , 
+J.M. Cairney \(^{1,2*}\)  
+
+<|ref|>text<|/ref|><|det|>[[115, 172, 865, 250]]<|/det|>
+\(^{1}\) School of Aerospace, Mechanical & Mechatronic Engineering, The University of Sydney, NSW 2006, Australia; \(^{2}\) Australian Centre for Microscopy and Microanalysis, The University of Sydney, NSW 2006, Australia; \(^{3}\) ISEM, Australian Institute of Innovative Materials, University of Wollongong, NSW 2500, Australia; \(^{4}\) Aerospace Division, Defence Science and Technology Group, VIC 3207, Australia; \(^{5}\) US Naval Research Laboratory, Washington DC, 20375, USA  
+
+<|ref|>text<|/ref|><|det|>[[113, 285, 876, 624]]<|/det|>
+Plastic deformation in ceramic materials is normally only observed in nanometre- sized samples. However, we have observed unprecedented levels of plasticity ( \(>50\%\) plastic strain) and excellent elasticity ( \(6\%\) elastic strain) in perovskite oxide \(\mathrm{Pb(In_{1 / 2}Nb_{1 / 2})O_{3}}\) - Pb(Mg \(_{1 / 3}\) Nb \(_{2 / 3}\) )O \(_{3}\) - \(\mathrm{PbTiO_3}\) (PIN- PMN- PT), under compression along \(< 100>\) pc pillars up to 2.1 \(\mu \mathrm{m}\) in diameter. The extent of this deformation is much higher than has previously been reported for ceramic materials, and the sample size at which plasticity is observed is almost an order of magnitude larger. Bending tests also revealed over \(8\%\) flexural strain. Plastic deformation occurred by slip along \(\{110\} < 1\overline{1} 0>\) . Calculations indicate that the resulting strain gradients will give rise to extreme flexoelectric polarization. First principles models predict that a high concentration of oxygen vacancies \((V_{O}^{ - })\) weaken the covalent/ionic bonds, giving rise to the unexpected plasticity. Mechanical testing on \(V_{O}^{ - }\) - rich Mn- doped PIN- PMN- PT confirmed this prediction. These findings will facilitate the design of plastic ceramic materials and the development of flexoelectric- based nano- electromechanical systems.  
+
+<|ref|>text<|/ref|><|det|>[[113, 654, 886, 831]]<|/det|>
+Conventional wisdom dictates that most metals are ductile and almost all ceramics are brittle. The plasticity of metals is related to their atomic bonding. Valence electrons are not bound to a specific atom and there is little charge resistance during dislocation slip \(^{1}\) . For ceramics, the directional covalent or/and ionic bonds restrict slip due to electrostatic repulsion, resulting in brittle fracture with only limited strain (usually less than \(0.2\%\) ) \(^{1}\) . In many cases, the brittle nature of ceramics limits their application, and improvements to the brittle properties of ceramics materials have been sought for decades \(^{2}\) .
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 87, 886, 418]]<|/det|>
+There are some rare exceptions to this rule. Crystals with the rock salt structure show limited plasticity due to their unique structure (slip occurs on \(\{110\}\) planes and along \(< 1\bar{1} 0>\) directions, where it does not bring similarly charged atoms together) \(^3\) . Among perovskite oxides, SrTiO \(_3\) (STO) has been reported to display around \(7\%\) plastic deformation under uniaxial compression at an extremely low strain rate \((10^{- 4})^4\) . More recently, good plasticity was reported in semiconductor \(\alpha\) - Ag \(_2\) S and InSe single crystals \(^5,^6\) . In \(\alpha\) - Ag \(_2\) S, excellent plasticity was attributed to planes with weak atomic interactions and irregularly distributed sulfur- silver and silver- silver bonds \(^5\) , while in InSe, the plasticity is thought to result from long- range In- Se Coulomb interactions across the van der Waals gap and soft intralayer In- Se bonding \(^6\) . Flash- sintered TiO \(_2\) has been compressed to \(\sim 10\%\) strain, attributed to a high- density of stacking faults, nanotwins, and dislocations \(^7\) . Plastic deformation observed in nano pillars, nanowires, etc., is mostly attributed to the low chance of smaller samples containing flaws, allowing the materials' intrinsic plasticity to be observed \(^8, 9, 10, 11, 12, 13, 14\) .  
+
+<|ref|>text<|/ref|><|det|>[[114, 455, 885, 606]]<|/det|>
+The existence of deformable ceramics has striking potential, but systems that display this characteristic must be identified and plasticity mechanisms need to be understood in order to guide the design of such materials. Because plastic deformation is not typical of ceramics, the applications have not yet been fully considered. It is expected such properties might enable applications such as sensors or even bendable and foldable electronics \(^{15}\) where flexible ceramic film capacitors are required \(^{16}\) .  
+
+<|ref|>text<|/ref|><|det|>[[114, 637, 885, 816]]<|/det|>
+Excellent elastic properties are especially desirable \(^{17}\) for functional oxides. A mechanical bending moment enables a dielectric material to polarize, giving rise to flexoelectricity. Flexoelectricity has a strong scaling effect and is therefore significant at micro/nano scales. For this reason, it has the potential to be used for electromechanical actuators and sensors that can be integrated into advanced nano- /micro- electromechanical systems (N/MEMS) \(^{18, 19}\) , meeting the requirement for the millions of micro- and nano- scale sensors to be employed during the expected rapid implementation of the Internet of Things.  
+
+<|ref|>text<|/ref|><|det|>[[114, 846, 884, 893]]<|/det|>
+Perovskite oxides are of great interest to both geophysics and materials science \(^{20}\) . In geophysics, a MgSiO \(_3\) - rich perovskite phase is thought to account for \(50 - 90\%\) of the volume of the region of
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 87, 885, 345]]<|/det|>
+the earth that controls seismic activity \(^{21,22}\) (i.e. the \(670\mathrm{km}\) seismic discontinuity to the core- mantle boundary \(^{19}\) ). In the field of materials science, perovskites are of interest because they exhibit useful flexoelectric, dielectric, piezoelectric, ferroelectric, ferromagnetic, multiferroic, superconducting, and photovoltaic properties, as well as colossal magnetoresistance \(^{23}\) . Pb(In \(_{1 / 2}\) Nb \(_{1 / 2}\) )O \(_{3}\) - Pb(Mg \(_{1 / 3}\) Nb \(_{2 / 3}\) )O \(_{3}\) - PbTiO \(_{3}\) (PIN- PMN- PT) is a ternary relaxor ferroelectric perovskite. Single crystal PIN- PMN- PT exhibits outstanding flexoelectric, piezoelectric and electromechanical properties (flexoelectric coupling coefficient \(\mu_{12}\) of \(5 \times 10^{4}\mathrm{nC}\cdot \mathrm{m}^{- 1}\) , piezoelectric coefficient of \(\mathrm{d}_{33}\) \(\sim 2000\mathrm{pC / N}\) and electromechanical coupling factor of \(\mathrm{k}_{33} \sim 90\%\) ) compared to traditional Pb(Zr,Ti)O \(_{3}\) piezoelectric ceramics ( \(\mathrm{d}_{33} < 500\mathrm{pC / N}\) , \(\mathrm{k}_{33} < 75\%\) ) \(^{24,25,26}\) . These extraordinary electromechanical coupling functionalities mean that the mechanical properties are of great interest.  
+
+<|ref|>text<|/ref|><|det|>[[113, 375, 885, 555]]<|/det|>
+PIN- PMN- PT samples were first characterised by transmission electron microscopy (TEM). Even prior to deformation experiments, clues to the potential plastic behaviour of PIN- PMN- PT were already apparent. During preparation of thin specimens for TEM, it was noted that the edges of \(\sim 3\mu \mathrm{m}\) tripod- polished samples were not flat (Fig. S1a and b), and a high density of entangled dislocations (Fig. S1) was present in the resulting TEM samples. Compression, tensile and bending tests on PIN- PMN- PT were carried out by using a combination of TEM, scanning electron microscopy (SEM) and nanomechanical test systems. The experimental set up is shown in Fig. S2.  
+
+<|ref|>text<|/ref|><|det|>[[113, 585, 885, 895]]<|/det|>
+Because plasticity has previously been observed in nanoscale ceramics during compression, we first tested the properties of our PIN- PMN- PT by preparing round pillars with diameters from 130 nm to \(270\mathrm{nm}\) for compression experiments in a TEM. Results are shown in Figs. 1a – c and Figs. S3 and S4. Fig. 1a shows an engineering stress–strain curve from a \(140\mathrm{nm}\) diameter pillar. The slope of the curve starts to decrease from \(\sim 5\%\) strain. Two short stress plateaus appear when the strain reaches \(\sim 15\%\) and \(\sim 44\%\) respectively, typical of plastic deformation. The total compression strain of the pillar exceeds \(60\%\) , over \(50\%\) of which is plastic. This extreme strain far surpasses the expected deformability of ceramic materials \(^{27}\) and is much higher than has been previously reported in micro/nanopillars \(^{9,10,13}\) . Snapshots captured from a video of the compression are shown in Figs. 1b – d. Slip bands (indicated by yellow arrows) develop on the (011) crystallographic plane, along the \([01\bar{1}]\) direction. Similar phenomena were observed for the other eight pillars with diameters ranging from \(130 \sim 270\mathrm{nm}\) (Figs. S3 and S4).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 113, 884, 345]]<|/det|>
+As mentioned earlier, previous studies showed that STO single crystals displayed surprisingly high plasticity, with a plastic strain of \(\sim 7\%^{4,28}\) . Here, we compare the compression behaviour of PIN- PMN- PT and STO by also compressing single crystal STO pillars along the same orientation (Figs. S5 – S6). Five out of six STO pillars underwent brittle fracture. The smallest pillar, at \(150\mathrm{nm}\) , was the only one that did not fracture, suggesting that STO undergoes a brittle- to- plastic transition with a critical pillar diameter of around \(150\sim 180\mathrm{nm}\) , while PIN- PMN- PT has significantly better plasticity (all PIN- PMN- PT pillars show plastic deformation). The maximum observed plastic strain was \(17.8\%\) for an STO pillar with the diameter of \(180\mathrm{nm}\) , where PIN- PMN- PT pillars with a similar diameter typically displayed \(>40\%\) strain.  
+
+<|ref|>text<|/ref|><|det|>[[112, 380, 885, 905]]<|/det|>
+To determine the effect of pillar size on the deformation behaviour in PIN- PMN- PT, larger pillars were fabricated with diameters ranging from \(500\mathrm{nm}\) to \(2.1\mu \mathrm{m}\) . Most of them displayed plasticity and some were brittle. An engineering stress- strain curve of a \(1\mu \mathrm{m}\) diameter pillar is provided in Figs. 1d – f. Strain bursts were observed, characterized by serrated yielding in the stress- strain curve. Similar rapid bursts of deformation are typical of tests conducted on micrometer- scale metal pillars<sup>29</sup>. Video snapshots in Figs. 1e – f correspond to strain of \(14.3\%\) and \(39.3\%\) , respectively. Slip initiates along the (011) plane and \([01\bar{1}]\) direction, as indicated by the yellow arrow in Fig. 1e. With further deformation, another slip band (110) \([\bar{1} 10]\) is activated, indicated by the red arrow in Fig. 1f and deformation proceeds until the strain reaches \(39.3\%\) . Compression test results from fourteen more pillars with diameters ranging from \(500\mathrm{nm} - 2.1\mu \mathrm{m}\) are shown in Figs. S7 – S8 and another detailed example of extreme deformability for a \(2.1\mu \mathrm{m}\) diameter pillar can be found in Fig. S9 ( \([\bar{1} 10]\) [110] slip and \(39.1\%\) strain). About \(60\%\) , \(50\%\) and \(40\%\) strain were observed in pillars of \(500\mathrm{nm}\) , \(1\mu \mathrm{m}\) and \(2.1\mu \mathrm{m}\) diameters, respectively, as shown in Figs. S7a, S7b, S8c, and S9, which far surpasses the plasticity observed in STO. Figs. 1g – h summarises the results of all compression tests. All samples \(< 700\mathrm{nm}\) diameter underwent plastic deformation, while some larger samples were brittle. Both the strain and the yield strength were typically higher for smaller samples (Figs. \(1\mathrm{g} - \mathrm{h}\) ), consistent with the literature on size effects in metals and ceramic pillars<sup>29, 30, 31, 32</sup>. However, though both STO and PIN- PMN- PT are perovskite oxides, their deformability differs greatly, and the size effects kick in at a much larger size scale for PIN- PMN- PT. This suggests that the intrinsic plasticity of the PIN- PMN- PT is much greater. The elastic compression
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 87, 886, 352]]<|/det|>
+strain is consistent for all samples, at an average of \(3.8\%\) and a maximum of \(6.2\%\) . The composition of PIN- PMN- PT used here is close to the morphotropic phase boundary at which an adaptive ferroelectric phase has been proposed, which can easily transform to other phases upon the mechanical strain. Our previous work shows that a reversible polydomain- rhombohedral to monodomain- orthorhombic phase transition happens under compression \(^{33}\) , which is thought to contribute to the large elastic strain observed here. To better understand the plastic deformation mechanism, deformed samples were further thinned by focused ion beam (FIB) into TEM foils. Scanning transmission electron microscopy – high- resolution high- angle annular dark- field (STEM- HAADF) images from a deformed area (Figs. 1i, j) show climb- dissociated pairs of partial dislocations with Burgers vector \(\frac{1}{2} a< 011>\) , separated by a stacking fault (see also Figs. S11 – S13).  
+
+<|ref|>text<|/ref|><|det|>[[112, 399, 886, 848]]<|/det|>
+Dog- bone pillars and cantilever beams were fabricated for tensile and bending tests. The experimental setup is described in Figs. S2d – g. Tensile tests (Fig. S14) of a dog bone sample of dimensions \(1.9 \mu \mathrm{m} \times 0.5 \mu \mathrm{m} \times 0.1 \mu \mathrm{m}\) revealed an elastic strain of \(4.0\%\) , but no plastic deformation. Fig. 2a shows a load- displacement curve obtained from an in- situ bending test. After deformation, the cantilever shows residual plastic deformation \((1.4\%)\) , consistent with an abrupt decrease in mechanical load, indicated by a red arrow in the inset curve in Fig. 2a. Figs 2b – c show video snapshots at maximum load and after unloading (an image prior to bending is shown in Fig. S2g). A maximum flexural strain of \(8.2\%\) , where \(6.8\%\) is elastic and \(1.4\%\) is plastic (details in SI) that occurs at the root of the cantilever beam. Fig. 2d is a low magnification high- resolution STEM- HAADF image taken from the area marked in green in Fig. 2c. Contrast is indicated by red arrows and numbers. Lattice rotation mapping derived from Geometric Phase Analysis (GPA, see methods) displays this contrast more clearly, Fig. 2e, highlighting dislocation cores \(^{34, 35}\) . Dislocations \(1 - 6\) are the same. Fig. 2f is a high- resolution STEM- HAADF image of dislocation #2, which consists of a pair of partial dislocations with Burgers vector of \(\frac{1}{2} a[01\bar{1}]\) and a stacking fault between them, consistent with the defects observed in compressed pillars. Dislocation #7 is different (Fig. S15b – c) and is assumed to be affected by the proximity of the surface of the cantilever beam.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 87, 885, 398]]<|/det|>
+For perovskite oxides, it is generally accepted that, at ambient temperature, the preferred slip system is \(\{110\} < 1\overline{1} 0>\) , with \(a< 1\overline{1} 0>\) dislocations \(^{19}\) . This type of dislocation is usually dissociated into two partials due to the high energy of two extra atomic planes. Previous studies on as- grown single/double crystals, polycrystals, or thin films show that \(a< 1\overline{1} 0>\) dislocations are dissociated either in a glide or a climb mode \(^{21, 34, 35, 36, 37}\) . Unexpectedly, we have observed climb- dissociated dislocation core structures, which would normally be expected to form at elevated temperatures because climb is a diffusion- assisted process \(^{38}\) . Instead, \(a< 110>\) dislocations formed during room temperature deformation might be expected to dissociate in a slip configuration \(^{38}\) , as was previously reported for compression- tested KNbO \(^{3}\) \(^{39}\) . We note here that a high density of point defects in the PIN- PMN- PT might enable the diffusion that is required to form climb- dissociated dislocations, leading to much better deformability compared to other perovskites such as STO or KNbO \(^{3}\) .  
+
+<|ref|>text<|/ref|><|det|>[[113, 444, 885, 652]]<|/det|>
+In perovskites oxides, vacancies are far more common than interstitials \(^{20}\) . \(V_{Pb}^{\prime \prime}\) and \(V_{O}^{- }\) are the most important vacancies in lead- based perovskites, where \(V_{Pb}^{\prime \prime}\) forms due to the volatility of lead at elevated temperature, or donor dopants, while \(V_{O}^{- }\) exists owing to the loss of oxygen at high temperature or acceptor dopants. The face- centred cubic lattice formed by \(\mathrm{Pb}^{2 + }\) and \(\mathrm{O}^{2 - }\) determines the dislocation and slip behavior \(^{20}\) . Consequently, the existence of \(V_{Pb}^{\prime \prime}\) and \(V_{O}^{- }\) in PIN- PMN- PT could considerably influence the observed plasticity. Studies of the effect of vacancies on the deformation behaviour of alloys or intermetallic compounds show varying results: vacancies may facilitate or deteriorate plasticity, depending on their type and distribution \(^{40, 41, 42}\) .  
+
+<|ref|>text<|/ref|><|det|>[[113, 699, 885, 901]]<|/det|>
+To trace the possible microscopic origin of the observed extreme plasticity, we conducted first principles atomistic simulations based on density functional theory (DFT). The results are given in Fig. 3. On the basis of a simplified model, PIN- PMN- PT is composed of three sets of subunits, PIN, PMN and PT (Fig. 3a). Atomic- scale Energy- Dispersive X- ray Spectroscopy (EDS) mapping (Fig. S16) indicates that the cations are uniformly distributed at the atomic level, suggesting a high density of mini- interfaces between the three subunits. Relaxed atomic structure and lattice constants of the bulk and interfaces are shown in Figs. S17 & S18, and Table S1 & S2. Calculated interface formation energies (shown in Fig. S19) suggest that the presence of interfaces promote
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 87, 885, 533]]<|/det|>
+the concentration of \(V_{O}^{ - }\) but not \(V_{P b}^{\prime \prime}\) . Favourable \(V_{O}^{ - }\) sites in different side- by- side and top- down interface systems are shown in Fig. S20. Interestingly, these calculations show that it is energetically favourable to form oxygen vacancies (but not lead vacancies) at these interfaces to mitigate the large lattice mismatch (Fig. S21). That is, the three subunits that make up the PIN- PMN- PT naturally facilitate a uniformly- distributed high density of \(V_{O}^{ - }\) . As an example, the atomic structure of 1PIN- 1PMN- 1PT containing one oxygen vacancy is shown in Fig. 3b. To assess the corresponding ductility, we calculated the elastic constants and derived the bulk modulus (B) \(^{43}\) and the anisotropic shear modulus (G) on the (110) plane along \(< 1\bar{1} 0>\) direction for different single tetragonal crystalline species \(^{44}\) , as shown in Fig. 3c and Table S3. The Pugh’s B/G ratio is widely used to index ductility, with a critical value of 1.75 indicating a transition from brittle to ductile behaviour \(^{41,42}\) . For bulk PIN, PMN, and PT, and their pristine interfaces, the calculated B/G ratios are well below 1.75 (hence brittle). By contrast, the B/G ratios for interfaces containing \(V_{O}^{ - }\) are systematically enhanced, most well above 1.75 (hence ductile). Valence charge density analysis reveals that the presence of \(V_{O}^{ - }\) can dramatically weaken the covalent bonding (see Figs. 3d and S22). For comparison, \(V_{P b}^{\prime \prime}\) actually deteriorates the ductility. Thus, based on the DFT results, we attribute the extreme plasticity of PIN- PMN- PT to the high density of \(V_{O}^{ - }\) at the PIN/PMN/PT interfaces (see Fig. S20).  
+
+<|ref|>text<|/ref|><|det|>[[113, 577, 886, 838]]<|/det|>
+On the basis of DFT predictions, we investigated the \(V_{O}^{ - }\) levels and mechanical behaviour of PIN- PMN- PT crystals that are expected to be \(V_{P b}^{\prime \prime}\) - rich and \(V_{O}^{ - }\) - rich, (Sm- doped \(^{45}\) and Mn- doped \(^{26}\) crystals respectively), and compared them to the original un- doped PIN- PMN- PT crystal. Electron energy loss spectra (EELS) of O were collected to verify the existence of oxygen vacancies, shown in Fig. 4a. A lower intensity is observed for the O- k edge fine structure peak B compared to A for all three EELS curves. It is known that the peak at position B being lower than the peak at position A is an indication of oxygen deficiency in perovskite oxides \(^{46, 47, 48}\) , suggesting that \(V_{O}^{ - }\) with appreciable concentrations exist in all three samples. Furthermore, the inset image shows that peak B is larger for the Sm- doped sample than the un- doped crystal, indicating a lower \(V_{O}^{ - }\) concentration, and is smaller for the Mn- doped sample, indicating a higher \(V_{O}^{ - }\) fraction.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 87, 885, 348]]<|/det|>
+According to the DFT predictions, the \(V_{O}^{- }\) - rich (Mn- doped \(^{26}\) ) samples are more likely to be ductile and the \(V_{Pb}^{\prime \prime}\) - rich, (Sm- doped \(^{45}\) ) crystals are more likely to be brittle. Compression tests were performed on both samples. Example engineering stress- strain curves and SEM images of compressed pillars are shown in Fig. 4b (details in Figs. S23 – S24). Six \(\sim 600 \mathrm{nm}\) diameter pillars were fabricated for each sample type. All Mn- doped PIN- PMN- PT pillars showed plasticity, while half of the Sm- doped PIN- PMN- PT pillars underwent brittle fracture, indicating that the Mn- doped sample had superior plasticity. In the examples shown in Fig. 4b, the Sm- doped sample has fractured in a brittle way, while the Mn- doped sample has slip bands on the pillar and a stress plateau and strain burst on the stress- strain curve. The results of this comparison experiment are consistent with our hypothesis of \(V_{O}^{- }\) - induced plasticity.  
+
+<|ref|>text<|/ref|><|det|>[[113, 392, 885, 596]]<|/det|>
+It has been proposed by Zubko et al. that dislocations contribute significantly to flexoelectricity in \(\mathrm{STO}^{49}\) . Tang et al. and Gao et al. measured the strain gradient around dislocations by extracting Bi/Sr positions from STEM- HAADF images and calculating the flexoelectric polarization in multiferroic \(\mathrm{BiFeO_3}\) and paraelectric \(\mathrm{STO}^{50,51}\) , which was found to be several \(\mu \mathrm{C}\cdot \mathrm{cm}^{- 2}\) . The flexoelectric effect is expected to be extremely large, because relaxor ferroelectric PIN- PMN- PT shows outstanding flexoelectricity compared to other perovskite oxides. Take STO as an example, the flexoelectric coefficient \(\mu_{12}\) is about \(7 \mathrm{nC}\cdot \mathrm{m}^{- 1}24,25,49\) , while that of PIN- PMN- PT is about \(5.0 \times 10^{4} \mathrm{nC}\cdot \mathrm{m}^{- 1}24\) , a difference of 4 orders of magnitude.  
+
+<|ref|>text<|/ref|><|det|>[[113, 641, 885, 901]]<|/det|>
+Here, in order to measure the flexoelectric polarization around a pair of partial dislocations (introduced by plastic deformation), we extracted Pb atom positions from STEM- HAADF images firstly (details in SI) and calculated the maximum strain gradient \((\nabla S)\) to be about \(3.5 \times 10^{9} \mathrm{m}^{- 1}\) ([0 \(\overline{1} 1\) ] lattice strain gradient along the \([0 \overline{1} \overline{1} ]\) direction), which is 3 times that reported by Gao et al. around [010] dislocations in a STO bicrystal \(^{50}\) . Supposing the flexoelectric coefficient of PIN- PMN- PT [110] is comparable to that of \(\mu_{12}^{24}\) , the local flexoelectric polarization (1\~2 unit- cells) around dislocations is estimated to be about \(10^{7} \mu \mathrm{C}\cdot \mathrm{cm}^{- 2}\) according to \(P_{f} = u \times \nabla S\) , where \(P_{f}\) is flexoelectric polarization, \(u\) is flexoelectric coefficient, and \(\Delta S\) is gradient of the horizontal lattice constant along the vertical direction. However, this large calculated polarization is thought to be an over- estimate for two reasons. 1) In the case of such high strain gradients, higher- order coupling
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 87, 885, 423]]<|/det|>
+terms of flexoelectric polarization and strain gradient, which is nonlinear, should not be neglected, and the magnitude of those terms is still unclear. 2) For smaller samples, permittivity \((\epsilon)\) is expected to decrease as a result of a size effect \(^{51}\) , and the flexoelectric coefficient \(\mu\) , which is a function of \(\epsilon\) in a manner of \(\mu = f \cdot \epsilon\) , should also be smaller than the corresponding bulk value (here \(f\) is flexo- coupling coefficient, about 10 V for PTO- based relaxor ferroelectrics). However, this extremely large polarization should give rise to a large number of bound charges. To screen these bound charges, free charges will accumulate. Transport properties or even magnetic properties around these dislocations can also be affected due to free charges. For slip bands, where a strain gradient also exists (as shown in Fig. S11e), the situation would be similar. As the strain gradient around a slip band is much smaller than it is around dislocations, the flexoelectric effect will be smaller. The movement of dislocations and the introduced slip bands make a functional region which is potentially applicable for flexoelectric based micro- and nano- scale electronic devices.  
+
+<|ref|>text<|/ref|><|det|>[[113, 468, 885, 867]]<|/det|>
+In addition to strain gradients around dislocations and slip bands, bending- induced elastic strain gradients are also of great interest for flexoelectricity because of their reversibility. The maximum elastic strain introduced by bending test is calculated to be \(6.8\%\) at the root of the cantilever beam, and the width \((b)\) of the cantilever beam is \(0.67 \mu \mathrm{m}\) , which gives rise to a strain gradient of about \(2 \times 10^{5} \mathrm{m}^{- 1} (\nabla S = \frac{6.8\%}{0.335 \mu \mathrm{m}} \approx 2 \times 10^{5} \mathrm{m}^{- 1})\) . Flexoelectric polarization from the elastic bending strain gradient is estimated to be about \(1 \times 10^{3} \mu \mathrm{C} \cdot \mathrm{cm}^{- 2}\) . Here strain gradient \(\nabla S\) is the horizontal ([001]) lattice strain gradient along the vertical direction ([010]). The calculated flexoelectric polarization is \(1 \sim 2\) orders of magnitude larger than the ferroelectric polarization of known ferroelectrics. For example, the ferroelectric polarization of \(\mathrm{PbTiO_3}\) is about \(75 \mu \mathrm{C} \cdot \mathrm{cm}^{- 2}\) , the ferroelectric polarization of \(\mathrm{BiFeO_3}\) is around \(90 \mu \mathrm{C} \cdot \mathrm{cm}^{- 2}\) , and the polarization of \(\mathrm{BaTiO_3}\) is about \(26 \mu \mathrm{C} \cdot \mathrm{cm}^{- 2}\) . \(^{52,53,54}\) The calculated flexoelectric polarization is also \(4 \sim 5\) times that of the recently- reported ferroelectric polarization of super- tetragonal \(\mathrm{PbTiO_3}^{55}\) . An even larger flexoelectric polarization would be expected if a lower strain rate is used, according to Deng's work \(^{56}\) . This large flexoelectric polarization is also likely to be an overestimate for reasons mentioned above. However, even if the real flexoelectric polarization is \(1 / 10\) of the calculated \(1 \times 10^{3} \mu \mathrm{C} \cdot \mathrm{cm}^{- 2}\) , it is still large enough
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 88, 883, 135]]<|/det|>
+(100 μC·cm⁻²) to switch the local ferroelectric polarization, and to be used in flexoelectric based sensors.  
+
+<|ref|>text<|/ref|><|det|>[[113, 181, 884, 333]]<|/det|>
+The excellent deformability in \(V_{O}^{- }\) - rich PIN- PMN- PT is particularly promising for flexoelectric- based sensors, because it was reported that the effective flexoelectricity of oxygen- depleted perovskite oxide is two orders of magnitude larger than for a stoichiometric sample<sup>57</sup>. Combined with the scaling effects of flexoelectricity and super large flexoelectric coefficient of PIN- PMN- PT, these provide exciting opportunities for high performance flexoelectric based N/MEMS devices.  
+
+<|ref|>text<|/ref|><|det|>[[112, 378, 885, 724]]<|/det|>
+To summarize, we have revealed extreme deformability in relaxor ferroelectric PIN- PMN- PT micron/submicron single crystals pillars. A maximum elastic strain of \(>6\%\) and plastic strain \(>50\%\) were observed during compression tests, while a flexural strain of \(8.2\%\) was achieved for a bent cantilever beam. Pairs of \(\frac{1}{2} a< 011>\) climb- dissociated partial dislocations accommodate the plastic deformation. Based on first principles calculations, confirmed by experiments, we propose that the observed excellent plasticity is attributed to not only a decrease in the specimen size, but also a high \(V_{O}^{- }\) concentration. This suggests that it might be possible to alter the plasticity of ceramic materials by deliberate engineering of point defects, which paves the way towards the design of ductile ceramics, and implies that more attention should be paid to the previously ignored mechanical properties of functional oxides. The giant strain gradient generated by elastic bending and dislocations gives rise to considerable flexoelectric polarization, which can be used in sensors. These results will facilitate the development of flexoelectric- based flexible electronic devices and N/MEMS.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[130, 150, 870, 696]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 95, 303, 112]]<|/det|>
+<center>Figures and Captions: </center>  
+
+<|ref|>text<|/ref|><|det|>[[112, 714, 886, 894]]<|/det|>
+Fig. 1 | Compression tests of sub- micro and micrometre scale pillars. a, An engineering stress – strain curve acquired during the compression of \(140 \mathrm{nm}\) diameter pillar, with a loading direction along [010]. b – c, Snapshots from a real time video recording of a compression test, at strains of \(18.7\%\) and \(60.1\%\) , respectively (labelled as yellow circles in a). Slip bands along (011) crystallographic plane and [011] direction is indicated by yellow arrows. Here, both slip plane and slip direction are determined from the change of contrast in TEM images. d, An engineering stress – strain curve from a compression test of a \(1 \mu \mathrm{m}\) pillar. e – f, Video snapshots corresponding to
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 87, 885, 355]]<|/det|>
+strains of \(14.3\%\) and \(39.3\%\) (yellow circles in d). Two slip bands (oriented \((110)[01\bar{1}]\) and \((1\bar{1} 0)[110])\) are indicated by yellow and red arrows. \(\mathbf{g}\) , Strain as a function of pillar diameter, showing plastic strain (hollow circles) and total strain (spheres) for plastic- deformed pillars and fracture strain (red crosses) for brittle- fracture samples. \(\mathbf{h}\) , Yield strength as a function of pillar diameter, showing yield strength for plastic- deformed pillars (spheres) and brittle- fracture pillars (red crosses). Dashed black curve: fitted yield strength – diameter curve for plastic- deformed pillars, with a function of \(y = 56.9x^{-0.52}\) . Green arrows indicate the strain/stress value corresponding to the pillars shown in Fig. 1a – c and d – f. \(\mathbf{i} - \mathbf{j}\) , STEM- HAADF images showing pairs of partial dislocations with Burgers vectors of \(\frac{1}{2} a[011]\) (i) and \(\frac{1}{2} a[0\bar{1}\bar{1}]\) (j). The partial dislocations are separated by stacking faults.  
+
+<|ref|>image<|/ref|><|det|>[[120, 373, 880, 727]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 745, 885, 898]]<|/det|>
+<center>Fig. 2 | Bending test of PIN-PMN-PT. a, A load-depth curve obtained during bending a cantilever beam along a loading direction of [010]. The inset shows the enlarged curve of the rectangular area, where an abrupt decrease in mechanical load is evident. b, Snapshot captured from in-situ video corresponding to the maximum depth of the indenter. c, An SEM image showing cantilever beam after unloading. Irreversible deformation can be clearly revealed by comparison with Fig. S2g. d – e, A STEM-HAADF image and the corresponding GPA analysis of lattice rotation, with </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 88, 883, 168]]<|/det|>
+dislocations indicated by red arrows. Dislocations #1 and #7 are labelled. f, High resolution STEM- HAADF images showing dislocation #2, which includes a pair of partial dislocations with Burgers vector of \(\frac{1}{2} a[01\bar{1}]\) .  
+
+<|ref|>image<|/ref|><|det|>[[123, 191, 861, 592]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 606, 885, 836]]<|/det|>
+<center>Fig. 3 | First principles atomistic investigation for the plasticity in PIN-PMN-PT. a, Three sub-unit cells for \(\mathrm{Pb(In_{1 / 2}Nb_{1 / 2})O_{3}}\) (PIN), \(\mathrm{Pb(Mg_{1 / 3}Nb_{2 / 3})O_{3}}\) (PMN) and \(\mathrm{PbTiO_{3}}\) (PT). b, An example of relaxed atomic structure containing interfaces formed by one PIN, one PMN and one PT with one oxygen vacancy \((V_{O}^{-})\) at the interface of PIN and PMN. c, Calculated bulk modulus/shear modulus (B/G) ratios for various bulk, pristine interfaces, and interfaces with \(V_{O}^{-}\) . Higher B/G ratios (>1.75) suggest ductile behaviour in PIN-PMN-PT. d, Calculated valence charge density 2D contour plot (colours assigned recursively) on the (020) plane of the structure shown in b. The strength of covalent bonding is indicated by the colour bar. The presence of \(V_{O}^{-}\) eliminates the local covalent bonds. </center>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[123, 94, 875, 300]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 315, 884, 468]]<|/det|>
+<center>Fig. 4 | Analysis of the origin of extreme plasticity. a, O-EELS obtained from Mn- (blue), Sm- (green) doped and undoped (red) PIN-PMN-PT. The enlarged image (inset) shows the intensity difference of peak B for Mn-, Sm- doped and undoped PIN-PMN-PT, where a lower intensity suggests a higher \(V_{O}^{*}\) concentration. b, Engineering stress – strain curves obtained from in-situ compression tests of Mn- (blue sphere) and Sm- (green circle) doped PIN-PMN-PT. Images I and II show SEM images of the compressed Mn-doped and Sm-doped PIN-PMN-PT respectively. </center>
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[116, 91, 269, 108]]<|/det|>
+## Acknowledgment:  
+
+<|ref|>text<|/ref|><|det|>[[113, 115, 885, 476]]<|/det|>
+Acknowledgment:The authors are grateful for the scientific and technical support from the Australian Centre for Microscopy and Microanalysis (ACMM) as well as the Microscopy Australia node at the University of Sydney. Thanks Dr Xianghai An from the University of Sydney for the fruitful discussion on the mechanical behaviour of materials. We are grateful for A/Prof. John Daniels, and PhD candidates Fan Ji and Tongzheng Xin from the University of New South Wales for helpful discussions regarding oxygen vacancies in perovskite oxides, and Dr. Jun Luo from TRS Technologies for providing single crystal samples. Thanks Prof. Gustau Catalan from Catalan Institute of Nanoscience and Nanotechnology (ICN2) for the discussion of flexoelectricity. This work was supported by the Australian Federal Government through the Next Generation Technologies Fund, and the DST Strategic Research Initiative in Advanced Materials and Sensors. We also acknowledge the assistance and high- performance computing (HPC) resources from the National Computational Infrastructure and the expert HPC facilitation from the team at the Sydney Informatics Hub at the University of Sydney. The authors would like to acknowledge the United States Office of Naval Research (ONR) and ONR Global for partially supporting this work.  
+
+<|ref|>sub_title<|/ref|><|det|>[[116, 508, 301, 525]]<|/det|>
+## Author contributions:  
+
+<|ref|>text<|/ref|><|det|>[[113, 532, 886, 761]]<|/det|>
+S.M., J.C. and P.F. initiated studies into the nano and micro structural evolution under mechanical loading of PIN- PMN- PT nano plates. Y.L. and J.C. proposed the mechanical property experiments. Y.L., S.Z. and J.C. designed the experiment. Y.L. fabricated pillars, conducted in- situ experiment for compression and bending tests, and carried out aberration corrected (S)TEM investigation (TEM/STEM/EDS/EELS). R.N. prepared tensile test samples and conducted in- situ tensile tests. M.G. supported in the acquisition and analysis of aberration- corrected (S)TEM images and spectroscopic data. X.Y.C. and S.P.R. designed and conducted first- principles simulation. S.Z., S.M. and P.F. provided single crystal samples. J.C. and X.L. supervised the research. All authors contributed to the discussions and manuscript preparation.  
+
+<|ref|>text<|/ref|><|det|>[[114, 794, 637, 813]]<|/det|>
+Competing interests: The authors declare no competing interests.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[114, 90, 214, 108]]<|/det|>
+## References:  
+
+<|ref|>text<|/ref|><|det|>[[110, 127, 869, 875]]<|/det|>
+1. Callister, W. D. & Rethwisch, D. G. Materials Science and Engineering: An Introduction (John Wiley & Sons, United States of America, 2009).  
+2. Pelleg, J. Mechanical Properties of Ceramic, Solid Mechanics and Its Applications, (Springer, Switzerland, 2014).  
+3. Wachtman, J. B., Cannon, W. R. & Matthewson, M. J. Mechanical properties of ceramics (John Wiley & Sons, United States of America, 2009).  
+4. Brunner, D., Taeri-Baghbadrani, S., Sigle, W. & Ruhle, M. Surprising results of a study on the plasticity in strontium titanate. J. Am. Ceram. Soc. 84, 1161–1163 (2001).  
+5. Shi, X., et al. Room-temperature ductile inorganic semiconductor. Nat. Mater. 17, 421–426 (2018).  
+6. Wei, T. R., et al. Exceptional plasticity in the bulk single-crystalline van der Waals semiconductor InSe. Science 369, 542–545 (2020).  
+7. Li, J., et al. Nanoscale stacking fault–assisted room temperature plasticity in flash-sintered TiO₂. Sci. Adv. 5, eaaw5519 (2019).  
+8. Chen, Y.J., An, X.H. & Liao, X.Z. Mechanical behaviors of nanowires. Appl. Phys. Rev. 4, 031104 (2017).  
+9. Michler, J., Wasmer, K., Meier, S., Ostlund, F. & Leifer, K. Plastic deformation of gallium arsenide micropillars under uniaxial compression at room temperature. Appl. Phys. Lett. 90, 043123 (2007).  
+10. Ostlund, F., et al. Brittle-to-Ductile Transition in Uniaxial Compression of Silicon Pillars at Room Temperature. Adv. Funct. Mater. 19, 2439–2444 (2009).  
+11. Han, X. D., et al. Low-temperature in situ large strain plasticity of ceramic SiC nanowires and its atomic-scale mechanism. Nano Lett. 7, 452–457 (2007).  
+12. Wang, Y. B., et al. Super Deformability and Young's Modulus of GaAs Nanowires. Adv. Mater. 23, 1356–1360 (2011).  
+13. Kiani, S., et al. Dislocation glide-controlled room-temperature plasticity in 6H-SiC single crystals. Acta Mater. 80, 400–406 (2014).  
+14. He, Y., Zhong, L., Fan, F., Wang, C., Zhu, T. & Mao, S. X., In situ observation of shear-driven amorphization in silicon crystals. Nat. Nanotechnol. 11, 866–871 (2016).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 880, 130]]<|/det|>
+15. An, H. S., et al. Surface-agnostic highly stretchable and bendable conductive MXene multilayers. Sci. Adv. 4, eaaq0118 (2018).  
+
+<|ref|>text<|/ref|><|det|>[[115, 143, 870, 186]]<|/det|>
+16. Ma, B., Lee, T. H., Dorris, S. E., Korital, R. E. & Balachandran, U. Flexible ceramic film capacitors for high-temperature power electronics. Mater. Sci. Energy Technol. 2, 96-103 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[115, 199, 839, 241]]<|/det|>
+17. Banerjee, A., et al. Ultralarge elastic deformation of nanoscale diamond. Science, 360, 300-302 (2018).  
+
+<|ref|>text<|/ref|><|det|>[[115, 254, 840, 297]]<|/det|>
+18. Wang, B., Gu, Y., Zhang, S. & Chen, L-Q. Flexoelectricity in Solids: Progress, Challenges, and Perspectives. Prog. Mater. Sci. 106, 100570 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[115, 310, 872, 352]]<|/det|>
+19. Bhaskar, U. K., et al. A flexoelectric microelectromechanical system on silicon. Nature nanotechnology. 11, 263-266 (2016).  
+
+<|ref|>text<|/ref|><|det|>[[115, 365, 848, 409]]<|/det|>
+20. Crary, A. P., et al. Perovskite: A Structure of Great Interest to Geophysics and Materials Science (American Geophysical Union, Washington, D.C., 1989).  
+
+<|ref|>text<|/ref|><|det|>[[115, 420, 857, 464]]<|/det|>
+21. Ferre, D., Carrez, P. & Cordier, P. Modeling dislocation cores in SrTiO₃ using the Peierls-Nabarro model. Phys. Rev. B 77, 014106 (2008).  
+
+<|ref|>text<|/ref|><|det|>[[115, 477, 840, 520]]<|/det|>
+22. Miyagi, L., Kanitpanyacharoen, W., Kaercher, P., Lee, K. K. M. & Wenk, H.R. Slip Systems in MgSiO₃ Post-Perovskite: Implications for D" Anisotropy. Science 329, 1639-1641 (2010).  
+
+<|ref|>text<|/ref|><|det|>[[115, 533, 876, 552]]<|/det|>
+23. Tilley, R. J. D. Perovskites: Structure-Property Relationships (John Wiley & Sons, chichester, 2016).  
+
+<|ref|>text<|/ref|><|det|>[[115, 564, 828, 607]]<|/det|>
+24. Shu, L., Liang R., Rao, Z., Fei, L., Ke, S. & Wang, Y. Flexoelectric materials and their related applications: A focused review. J. Adv. Ceram. 8: 153-173 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[115, 620, 863, 662]]<|/det|>
+25. Shu, L. et al. Flexoelectric behavior in PIN-PMN-PT single crystals over a wide temperature range. Appl. Phys. Lett. 111: 162901 (2017)  
+
+<|ref|>text<|/ref|><|det|>[[115, 675, 830, 718]]<|/det|>
+26. Zhang, S. J. & Li, F. High performance ferroelectric relaxor-PbTiO₃ single crystals: Status and perspective. J. Appl. Phys. 111, 031301 (2012).  
+
+<|ref|>text<|/ref|><|det|>[[115, 730, 880, 772]]<|/det|>
+27. Green, D. J. An Introduction to the Mechanical Properties of Ceramics. (Cambridge University Press, Cambridge, 1998).  
+
+<|ref|>text<|/ref|><|det|>[[115, 786, 852, 828]]<|/det|>
+28. Gumbsch, P., Taeri-Baghbadrani, S., Brunner, D., Sigle, W. & Ruhle, A. Plasticity and an inverse brittle-to-ductile transition in strontium titanate. Phys. Rev. Lett. 87, 085505 (2001).  
+
+<|ref|>text<|/ref|><|det|>[[115, 841, 864, 884]]<|/det|>
+29. Uchic, M. D., Dimiduk, D. M., Florando, J. N. & Nix, W. D. Sample dimensions influence strength and crystal plasticity. Science 305, 986-989 (2004).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[110, 88, 881, 131]]<|/det|>
+30. Greer, J. R. & Hosson, J. T. M. D. Plasticity in small-sized metallic systems: Intrinsic versus extrinsic size effect. Prog. Mater. Sci. 56, 654–724 (2011).  
+
+<|ref|>text<|/ref|><|det|>[[112, 144, 855, 187]]<|/det|>
+31. Kiani, S., Yang, J. M. & Kodambaka, S. Nanomechanics of refractory transition-metal carbides: a path to discovering plasticity in hard ceramics. J. Am. Chem. Soc. 98, 2313–2323 (2015).  
+
+<|ref|>text<|/ref|><|det|>[[112, 199, 720, 218]]<|/det|>
+32. Zhu, T. & Li, J. Ultra-strength materials. Prog. Mater. Sci. 55, 710–757 (2010).  
+
+<|ref|>text<|/ref|><|det|>[[112, 231, 856, 298]]<|/det|>
+33. Liu, Y., Xia, J., Finkel, P., Moss, D. S., Liao, X. & Cairney, M. J. Real-time observation of stress-induced domain evolution in a [011] PIN-PMN-PT relaxor ferroelectric single crystal. Acta Mater. 175, 436–444 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[112, 310, 877, 354]]<|/det|>
+34. Liu, Y., Tang, Y. L., Zhu, Y. L., Wang, W. Y. & Ma, X. L. Spatial Coupling of Ferroelectric Domain Walls and Crystallographic Defects in the PbTiO₃ Films. Adv. Mater. Interfaces 3, 1600342 (2016).  
+
+<|ref|>text<|/ref|><|det|>[[112, 366, 861, 409]]<|/det|>
+35. Tang, Y. L., Zhu, Y. L., Liu, Y., Wang, Y. J., Ma, X. L. Giant linear strain gradient with extremely low elastic energy in a perovskite nanostructure array. Nat. Commun., 8, 15994 (2017).  
+
+<|ref|>text<|/ref|><|det|>[[112, 421, 840, 465]]<|/det|>
+36. Zhang, Z. L., Sigle, W., Kurtz, W. & Ruhle, M. Electronic and atomic structure of a dissociated dislocation in SrTiO₃. Phys. Rev. B 66, 214112 (2002).  
+
+<|ref|>text<|/ref|><|det|>[[112, 477, 859, 520]]<|/det|>
+37. Mao, Z. & Knowles, K. M. Dissociation of lattice dislocations in SrTiO₃. Philos. Mag. A, 73, 699–708 (1996).  
+
+<|ref|>text<|/ref|><|det|>[[112, 533, 841, 576]]<|/det|>
+38. Matsunaga, T., Saka, H. Transmission electron microscopy of dislocations in SrTiO₃. Phil. Mag. Lett., 80, 597–604 (2000).  
+
+<|ref|>text<|/ref|><|det|>[[112, 588, 865, 655]]<|/det|>
+39. Mark, A. F., Castillo-Rodriguez, M. & Sigle, W. Unexpected plasticity of potassium niobate during compression between room temperature and 900 degrees C. J. Eur. Ceram. Soc. 36, 2781–2793 (2016).  
+
+<|ref|>text<|/ref|><|det|>[[112, 667, 875, 710]]<|/det|>
+40. Salehinia, I., Medyanik, S. N. Effects of Vacancies on the Onset of Plasticity in Metals-An Atomistic Simulation Study. Metall. Mater. Trans. A, 42a, 3868–3874. (2011).  
+
+<|ref|>text<|/ref|><|det|>[[112, 722, 867, 766]]<|/det|>
+41. Pan, Y., Lin, Y.H., Wang, H. & Zhang, C.M. Vacancy induced brittle-to-ductile transition of Nb<sub>5</sub>Si<sub>3</sub> alloy from first-principles. Mater. Design 86, 259–265 (2015).  
+
+<|ref|>text<|/ref|><|det|>[[112, 779, 867, 821]]<|/det|>
+42. Kindlund, H., et al. Vacancy-induced toughening in hard single-crystal V<sub>0.5</sub>Mo<sub>0.5</sub>N<sub>x</sub>/MgO (001) thin films. Acta Mater. 77, 394–400 (2014).  
+
+<|ref|>text<|/ref|><|det|>[[112, 834, 874, 876]]<|/det|>
+43. Liu, Y. et al. First-principles study of elastic properties in perovskite PbTiO<sub>3</sub>. Mater. Sci. Eng. A 472, 269–272 (2008).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[110, 88, 860, 130]]<|/det|>
+44. He, Y., Schwarz, R. B., Migliori, A. & Whang, S. H. Elastic constants of single crystal \(\gamma\) – TiAl. J. Mater. Res. 10, 1187–1195 (1995).  
+
+<|ref|>text<|/ref|><|det|>[[113, 143, 856, 186]]<|/det|>
+45. Li, F., et al. Giant piezoelectricity of Sm-doped Pb(Mg<sub>1/3</sub>Nb<sub>2/3</sub>)O<sub>3</sub>-PbTiO<sub>3</sub> single crystals. Science, 364, 264–268 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[113, 199, 860, 242]]<|/det|>
+46. Muller, D. A., Nakagawa, N., Ohtomo, A., Grazul, J. L. & Hwang, H. Y. Atomic-scale imaging of nanoengineered oxygen vacancy profiles in SrTiO<sub>3</sub>. Nature 430, 657–661 (2004).  
+
+<|ref|>text<|/ref|><|det|>[[113, 254, 870, 298]]<|/det|>
+47. Klie, R. F., Browning, N. D. Atomic scale characterization of oxygen vacancy segregation at SrTiO<sub>3</sub> grain boundaries. Appl. Phys. Lett. 77, 3737–3739 (2000).  
+
+<|ref|>text<|/ref|><|det|>[[113, 310, 863, 353]]<|/det|>
+48. Zhang, Z. L., Sigle, W., Ruhle, M. Atomic and electronic characterization of the \(a[100]\) dislocation core in SrTiO<sub>3</sub>. Phys. Rev. B 66, 094108 (2002).  
+
+<|ref|>text<|/ref|><|det|>[[113, 366, 813, 409]]<|/det|>
+49. Zubko, P., Catalan, G., Buckley, A., Welche, P. R. L. & Scott, J. F. Strain-Gradient-Induced Polarization in SrTiO<sub>3</sub> Single Crystals. Phys. Rev. lett. 99, 167601 (2007).  
+
+<|ref|>text<|/ref|><|det|>[[113, 421, 872, 465]]<|/det|>
+50. Gao, P., et al. Atomic-Scale Measurement of Flexoelectric Polarization at SrTiO<sub>3</sub> Dislocations. Phys. Rev. Lett. 120, 267601 (2018).  
+
+<|ref|>text<|/ref|><|det|>[[113, 477, 866, 521]]<|/det|>
+51. Saad, M. M., et al. Investigating the effects of reduced size on the properties of ferroelectrics. IEEE Trans. Ultrason. Ferroelectr. Freq. Control. 53, 2208–2225 (2016).  
+
+<|ref|>text<|/ref|><|det|>[[113, 533, 863, 576]]<|/det|>
+52. Tang, Y. L., et al. Observation of a periodic array of flux-closure quadrants in strained ferroelectric PbTiO<sub>3</sub> films. Science 348, 547–551 (2005).  
+
+<|ref|>text<|/ref|><|det|>[[113, 589, 861, 631]]<|/det|>
+53. Wang, J., et al. Epitaxial BiFeO<sub>3</sub> Multiferroic Thin Film Heterostructures. Science 299, 1719–1722 (2003).  
+
+<|ref|>text<|/ref|><|det|>[[113, 644, 843, 686]]<|/det|>
+54. Jaffe, B., Cook Jr., W. R. & Jaffe, H. Piezoelectric Ceramics. Academic Press, London and New York, p. 78 (1971).  
+
+<|ref|>text<|/ref|><|det|>[[113, 699, 856, 741]]<|/det|>
+55. Zhang, L. et al. Giant polarization in super-tetragonal thin films through interphase strain. Science 361: 494–497 (2018).  
+
+<|ref|>text<|/ref|><|det|>[[113, 754, 870, 797]]<|/det|>
+56. Deng, Y., et al. Hierarchically-structured large superelastic deformation in ferroelastic-ferroelectrics. Acta Mater. 181, 501–509 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[113, 810, 782, 852]]<|/det|>
+57. Narvaez J, Vasquez-Sancho F, Catalan G. Enhanced flexoelectric-like response in oxide semiconductors. Nature 538, 219–221 (2016).
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 90, 191, 107]]<|/det|>
+## Methods  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 125, 202, 142]]<|/det|>
+## Materials:  
+
+<|ref|>text<|/ref|><|det|>[[114, 149, 884, 300]]<|/det|>
+The experimental work reported in this paper was performed using [011] poled PIN- PMN- PT single crystal plates (CTS Advanced Materials, with nominal composition 0.24PIN- 0.44PMN- 0.32PT, grown via the modified Bridgeman method) with MPB composition, a relative permittivity of 4000, dimensions of \(12 \times 4 \times 4 \mathrm{mm}^3\) , and surface polished to \(50 - 110 \mathrm{nm}\) . Sm- doped and Mn- doped PIN- PMN- PT (TRS Technologies) single crystals were grown by a modified Bridgeman method and STO is commercial single crystal.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 315, 283, 333]]<|/det|>
+## Sample preparation:  
+
+<|ref|>text<|/ref|><|det|>[[113, 339, 885, 620]]<|/det|>
+Micro- pillars preparation for compression, tensile and bending tests: The PIN- PMN- PT single crystal was first cut into slices of \(0.5 \mathrm{mm}\) in thickness, then further thinned using tripod polishing to \(\sim 500 \mathrm{nm}\) at the front edge. Pillars used for in- situ tests were fabricated at the thin edge by using FIB. Columnar pillars with an aspect ratio (height/diameter) of \(2:1 \sim 3:1\) were prepared for compression tests. The FIB was operated at \(30 \mathrm{kV}\) using a current of \(1 \mathrm{nA}\) for coarse milling and \(5 \mathrm{pA} \sim 300 \mathrm{pA}\) for final milling of pillars with diameters ranging from \(130 \mathrm{nm} \sim 2.1 \mu \mathrm{m}\) . The pillar taper angles are estimated to be around \(3^{\circ}\) . The diameter of the top surface was used for stress calculation, which is the first part of the sample to undergo plastic deformation. Cantilever beams for bending tests were prepared with FIB operating at \(30 \mathrm{kV}\) and using a current of \(50 \mathrm{pA}\) for final milling. The length, width and depth are \(6.5\) , \(0.67\) and \(0.8 \mu \mathrm{m}\) , respectively. Dog bone shaped pillars were prepared for tensile tests, and \(30 \mathrm{kV}\) , \(5 \mathrm{pA}\) were used for final milling.  
+
+<|ref|>text<|/ref|><|det|>[[114, 625, 884, 750]]<|/det|>
+TEM sample preparation: The deformed pillars were lifted- out using a tungsten manipulator onto a copper base, and then thinned to electron transparency ( \(\sim 50 \mathrm{nm}\) ) for TEM observation. \(10 \mathrm{kV}\) and \(10 \mathrm{pA}\) were used for FIB final milling. \(5 \mathrm{kV}\) , \(10 \mathrm{pA}\) and \(2 \mathrm{kV}\) , \(10 \mathrm{pA}\) were used for final cleaning of the surface. To protect the pillars from FIB damage, platinum was deposited around the pillars before thinning.  
+
+<|ref|>text<|/ref|><|det|>[[114, 756, 883, 828]]<|/det|>
+TEM sample preparation for O- K EELS: TEM samples for O- K EELS were prepared by grinding using tripod polisher and ion milling employing a Gatan precision ion polishing system II (PIPS II). \(4^{\circ}\) and \(0.5 \mathrm{kV}\) were used for final milling.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 844, 320, 861]]<|/det|>
+## In-situ mechanical tests:
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 87, 885, 450]]<|/det|>
+In- situ compression experiments were carried out in both a TEM (JEOL JEM 2100) and an SEM (Zeiss Ultra), while in- situ tensile and bending tests were conducted in the SEM. The JEOL JEM 2100 uses a high brightness LaB6 electron source. It is equipped with Xarosa (4 k x 4 k) as well as Veleta Ultrascan (2 k x 2 k) cameras. In the TEM, in- situ compression tests of pillars with diameters around \(200\mathrm{nm}\) were carried out by using a Hysitron PI 95 Picoindenter with a flat diamond tip. As the load applied is limited to \(1.5\mathrm{mN}\) for the PI 95 Picoindenter, the requirement for thin sample in the TEM, we carried out the in- situ compression experiment of the larger pillars by using a Hysitron PI 85L picoindenter inside an SEM, with a specially designed system for applying loads up to \(10\mathrm{mN}\) . This system allows real- time observation of deformation process (i.e. slip band development, slip planes and slip directions). Load was applied to pillars by moving the indenter toward the pillars in the displacement control mode. The displacement rates were \(1\mathrm{nm}\cdot \mathrm{s}^{- 1}\) and \(2\mathrm{nm}\cdot \mathrm{s}^{- 1}\) for compression of pillars of around \(200\mathrm{nm}\) in diameter and from \(500\mathrm{nm}\sim 2.1\mu \mathrm{m}\) in diameter, respectively. For the tensile test, a displacement rate of \(1\mathrm{nm}\cdot \mathrm{s}^{- 1}\) was used. For the bending test, a higher displacement rate - \(4\mathrm{nm}\cdot \mathrm{s}^{- 1}\) was used.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 462, 560, 482]]<|/det|>
+## Microstructure investigation of the deformed pillars:  
+
+<|ref|>text<|/ref|><|det|>[[113, 487, 885, 795]]<|/det|>
+A JEOL JEM 2100 TEM and a FEI Themis- Z Double- corrected 60- 300 kV S/TEM were used to observe the compressed pillars. High- resolution STEM- HAADF images, EDS element mapping and O- K edge EELS were acquired using the FEI Themis- Z S/TEM. The convergence and collection angle under the STEM- HAADF mode are 17.9 mrad and \(50 - 200\mathrm{mrad}\) , respectively. Strain was analysed using free Geometric Phase Analysis script (by C.T. Koch) \(^{58}\) . EELS of the O- K edge was acquired under the TEM mode at a collection angle of \(100\mathrm{mrad}\) . Dual- EELS was used and zero peak was corrected for all three samples. The energy resolution is estimated to be \(1.0\mathrm{eV}\) , measured from full width at half maxima of zero loss peak, while an energy dispersion of \(0.025\mathrm{eV / ch}\) was employed. The point resolution of Themis- Z under the STEM mode is around \(0.6\mathrm{\AA}\) (operated at \(300\mathrm{kV}\) ). It is equipped with X- FEG high- brightness gun, Monochromator, ChemiSTEM (Super- X) EDS detectors as well as a Gatan Quantum ER/965 GIF (<0.14 eV (1s)) with Dual- EELS.  
+
+<|ref|>text<|/ref|><|det|>[[114, 809, 884, 907]]<|/det|>
+First- principles simulation: DFT calculations were performed using the plane- wave pseudopotential total energy method as implemented in the VASP code \(^{59,60}\) . Projector augmented wave potentials \(^{61}\) and the generalized gradient approximation \(^{62}\) were used for exchange- correlation. A plane- wave basis set was used with an energy cut off of \(500\mathrm{eV}\) . The summation
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 87, 884, 214]]<|/det|>
+over the Brillouin zone for the bulk structures was performed on a \(\sim 0.06 \mathrm{\AA}^{- 1}\) spacing Monkhorst- Pack \(\mathbf{k}\) - point mesh for all calculations. For all systems, atomic relaxation was allowed until all the forces were less than \(0.01 \mathrm{eV / \AA}\) . For charge density calculations, \(\mathrm{Pb - 5d}\) , \(\mathrm{Nb - 4p}\) , \(\mathrm{Mg - 2p}\) , \(\mathrm{Ti - 3p}\) and \(\mathrm{In - 4d}\) semi- core states were treated as valence states to ensure high accuracy. Additional computational details can be found in the Supporting Information.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 239, 209, 256]]<|/det|>
+## References  
+
+<|ref|>text<|/ref|><|det|>[[113, 281, 877, 322]]<|/det|>
+58. Hytch, M. J., Snoeck, E., Kilaas, R. Quantitative measurement of displacement and strain fields from HREM micrographs. Ultramicroscopy 74, 131-146 (1998).  
+
+<|ref|>text<|/ref|><|det|>[[114, 336, 806, 380]]<|/det|>
+59. Kresse, G. & Furthmuller, J. Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set. Comp. Mater. Sci. 6, 15-50 (1996).  
+
+<|ref|>text<|/ref|><|det|>[[114, 392, 875, 436]]<|/det|>
+60. Kresse, G. & Furthmuller, J. Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Phys. Rev. B 54, 11169-11186 (1996).  
+
+<|ref|>text<|/ref|><|det|>[[114, 448, 864, 491]]<|/det|>
+61. Kresse, G. & Joubert, D. From ultrasoft pseudopotentials to the projector augmented-wave method. Phys. Rev. B 59, 1758-1775 (1999).  
+
+<|ref|>text<|/ref|><|det|>[[113, 503, 852, 547]]<|/det|>
+62. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation made simple. Phys. Rev. Lett. 77, 3865-3868 (1996)
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[44, 42, 312, 70]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[44, 93, 765, 113]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[60, 130, 420, 150]]<|/det|>
+SupportingInformation20210226. docx
+
+<--- Page Split --->

preprint/preprint__1003c99ade06e9dbb20a5837a88e4801c0fa9dce9184fba16a6a52a8b20ccb57/images_list.json

@@ -0,0 +1,122 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Fig.1. Expression and Localization of Nat10 enriched in the nucleolus of oocytes in mice. (A) Western blot showing the relative expression levels of NAT10 protein among multiple organs in adult WT mice. GAPDH served as a loading control. (B) Dynamic mRNA expression levels of Nat10 from RNA-seq analyses in oocytes and preimplantation embryos in mice (GSE71434). ICM, Inner cell mass. (C) Quantitative RT-PCR results showing the relative expression levels of mouse Nat10 mRNA in oocytes, and preimplantation embryos. Data were presented as mean±SEM, n=3. GO, growing oocytes collected from postnatal 14-day-old (P14) female mice. (D) Immunofluorescence (IF) staining of NAT10 in growing (GO), GV, MI, and MII oocytes as indicated. Dashed circle indicates cellular membrane of oocytes. DNA was counterstained with 4',6-diamidino-2-phenylindole (DAPI). Scale bar, 20 μm. (E) IF images of 21-day-old WT ovarian cryosections stained with anti-NAT10 antibody (Red) and DAPI (Blue) for follicles at various stages (primordial, primary, secondary, early antral, and antral stages) as",
+    "footnote": [],
+    "bbox": [
+      [
+        120,
+        80,
+        874,
+        666
+      ]
+    ],
+    "page_idx": 27
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Fig.2. Pre-meiotic deletion of Nat10 caused follicular developmental arrest and premature ovarian failure (POF). (A) Schematic diagram showing the landmark timeline of oocyte development from embryonic meiotic cell-cycle progression to postnatal oocyte growth and maturation. Stra8-GFPcre is activated prior to Embryonic day 13.5 (E13.5); Zp3-cre is active starting from P5 in the primary follicles; Both Ubc-CreERT2 and Ddx4-CreERT2 lines are Cre-inducible in all tissues and specifically in the germline, respectively, upon tamoxifen injection. (B) A breeding scheme by crossing Nat10<sup>lox/lox</sup> with",
+    "footnote": [],
+    "bbox": [
+      [
+        130,
+        87,
+        870,
+        778
+      ]
+    ],
+    "page_idx": 29
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Fig.3. Embryonic Nat10 loss caused oocyte meiotic prophase I arrest at pachytene stage owing to deficient DSB repair. (A) Immunofluorescence staining of oocyte nuclear chromosome spreads by SYCP3 and SYCP1 markers in WT and Nat10-ScKO mouse ovaries at birth. Scale bar, \\(10 \\mu \\mathrm{m}\\) . Arrows point to the asynapsed structure of the lateral and central axes. (B) The statistic counts showing the percentage of oocytes at various stages as indicated. Data are presented as the mean±SEM, \\(n = 3\\) , \\(*\\) , \\(p< 0.05\\) by two-tailed Student's t-test. (C) IF staining by SYCP3 and yH2AX on surface-spread oocytes at pachytene and diplotene stages from WT and Nat10-ScKO mouse oocytes at birth. Scale bar, 10",
+    "footnote": [],
+    "bbox": [
+      [
+        120,
+        80,
+        875,
+        760
+      ]
+    ],
+    "page_idx": 31
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Fig.4. Postnatal Nat10 depletion caused ovarian developmental arrest at secondary follicles. (A) A breeding scheme for Nat10 KO in growing oocytes of primary follicles by crossing Nat10<sup>lox/lox</sup> with Zp3-Cre deleter to attain Nat10<sup>lox/-</sup>; Zp3-Cre (Nat10-ZcKO) female offspring. (B-C) Immunofluorescence staining by NAT10 (Red), NPM (Green) and Hoechst 33342(Blue) in the secondary follicles (B) and GV oocytes (C) from WT and Nat10-ZcKO ovaries. Scale bar, \\(20\\mu \\mathrm{m}\\) . (D) Fertility test showing the cumulative numbers of pups from breedings of WT and Nat10-ZcKO females with WT males during a half-year caging. Data are presented as the mean± SEM, \\(n = 5\\) ; \\(***\\) , \\(p< 0.0001\\) by two-tailed Student's t-test. (E) The gross morphology of ovaries derived from WT and Nat10-ZcKO mice at age of 1 month (M) (left) and 2 months (right). Scale bar, \\(200\\mu \\mathrm{m}\\) . (F) H&E staining showing ovarian histology from WT and Nat10-ZcKO mice at 1 M (Top) and 2 M (Bottom). Scale bar, \\(50\\mu \\mathrm{m}\\) . Follicles are indicated by arrows. (G) Comparison of the average numbers of follicles at indicated stages in the ovaries of WT and Nat10-ZcKO mice at 1M (Top) and 2M (Bottom). Follicles were counted on serial ovarian sections",
+    "footnote": [],
+    "bbox": [
+      [
+        123,
+        80,
+        870,
+        648
+      ]
+    ],
+    "page_idx": 33
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Fig.5. Postnatal Nat10 deficiency impedes oocyte chromatin NSN–SN configuration transition.",
+    "footnote": [],
+    "bbox": [
+      [
+        117,
+        88,
+        870,
+        606
+      ]
+    ],
+    "page_idx": 35
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_6.jpg",
+    "caption": "Fig.6. Postnatal Nat10 ablation led to defective oocyte meiotic maturation. (A) The gross morphology of oocytes collected at the time points as indicated for GV (0h), and cultured in vitro for MI",
+    "footnote": [],
+    "bbox": [
+      [
+        125,
+        80,
+        870,
+        856
+      ]
+    ],
+    "page_idx": 36
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_7.jpg",
+    "caption": "Fig.7. Mini-bulk SMART-seq2 analyses identified the dysregulated maternal transcriptome in Nat10-ZcKO oocytes. (A) A diagram showing mouse oocyte samples collected for mini-bulk SMART-seq2 analyses. (B) Bar graph showing the numbers of transcripts detected in WT and Nat10-ZcKO",
+    "footnote": [],
+    "bbox": [
+      [
+        119,
+        80,
+        880,
+        844
+      ]
+    ],
+    "page_idx": 38
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_8.jpg",
+    "caption": "Fig.8. Hairpin Adaptor-PolyA Tail length (HA-PAT) assay validated the deficient maternal mRNA decay in Nat10-ZcKO MII oocytes. (A) Venn diagram showing the overlapping of transcripts that were stabilized during GV-to-MII transition in Cnot6l-/- and Nat10-ZcKO MII oocytes (FC=[WT MII/Nat10-ZcKO MII] \\(\\geq 2\\) , \\(\\mathsf{p}< 0.05\\) ). (B) Fold change of relative expression levels of transcripts encoding ribosomal protein subunits in Nat10-ZcKO relative to WT oocytes at MII stage. The values of log2(FC[Nat10-ZcKO/WT]) are listed on the right column. (C) qPCR results showing the relative levels of indicated transcripts (Cnot6l, Cnot7 and Btg4) in WT and Nat10-ZcKO oocytes at MII stage. Data are presented as the mean±SEM, \\(n = 3\\) . \\*\\*\\*, \\(\\mathsf{p}< 0.0001\\) by two-tailed Student's t-test. (D) A schematic illustration depicting the design strategy and the key steps for Hairpin Adaptor-PolyA Tail length (HA-PAT) assay. The 1st strand of cDNA was synthesized with the hairpin adaptor (HA) primer in conjunction with a P5TSO primer containing three \"G\", via a mechanism of \"template-switching\". GSP, Gene-specific primer; A0, the PCR product resulting from the amplification with a gene-specific pair of GSPxF and",
+    "footnote": [],
+    "bbox": [
+      [
+        119,
+        80,
+        880,
+        658
+      ]
+    ],
+    "page_idx": 40
+  }
+]

preprint/preprint__1003c99ade06e9dbb20a5837a88e4801c0fa9dce9184fba16a6a52a8b20ccb57/preprint__1003c99ade06e9dbb20a5837a88e4801c0fa9dce9184fba16a6a52a8b20ccb57.mmd

@@ -0,0 +1,498 @@
+
+# Maternal N-acetyltransferase 10 (NAT10) orchestrates oocyte meiotic cell-cycle progression and maturation in mice  
+
+Jianqiang Bao ( jqbao@ustc.edu.cn )  
+
+The First Affiliated Hospital of University of Science and Technology of China https://orcid.org/0000- 0003- 1248- 2687  
+
+Xue Jiang  
+
+The First Affiliated Hospital of University of Science and Technology of China  
+
+Yu Cheng  
+
+University of Science and Technology of China  
+
+Yuzhang Zhu  
+
+University of Science and Technology of China  
+
+Caoling Xu  
+
+The First Affiliated Hospital of University of Science and Technology of China  
+
+Qiaodan Li  
+
+University of Science and Technology of China  
+
+Xuemei Xing  
+
+The First Affiliated Hospital of University of Science and Technology of China  
+
+Wenqing Li  
+
+The First Affiliated Hospital of University of Science and Technology of China  
+
+Jiaqi Zou  
+
+The First Affiliated Hospital of University of Science and Technology of China  
+
+Lan Meng  
+
+The First Affiliated Hospital of University of Science and Technology of China  
+
+Muhammad Azhar  
+
+The First Affiliated Hospital of University of Science and Technology of China  
+
+Yuzhu Cao  
+
+The First Affiliated Hospital of University of Science and Technology of China  
+
+Xianhong Tong  
+
+The First Affiliated Hospital of University of Science and Technology of China  
+
+Weibing Qin  
+
+Xiaoli Zhu  
+
+The First Affiliated Hospital of University of Science and Technology of China
+
+<--- Page Split --->
+
+## Article  
+
+Keywords: Nat10, meiosis, oocyte growth, oocyte maturation, PolyA Tail length assay (PAT)  
+
+Posted Date: September 15th, 2022  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 2033653/v1  
+
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License
+
+<--- Page Split --->
+
+# Maternal N-acetyltransferase 10 (NAT10) orchestrates oocyte meiotic cell-cycle progression and maturation in mice  
+
+Xue Jiang \(^{1,\dagger}\) , Yu Cheng \(^{2,\dagger}\) , Yuzhang Zhu \(^{3,\dagger}\) , Caoling Xu \(^{1,\dagger}\) , Qiaodan Li \(^{4,\dagger}\) , Xuemei Xing \(^{5}\) , Wenqing Li \(^{1}\) , Jiaqi Zou \(^{1}\) , Lan Meng \(^{1}\) , Muhammad Azhar \(^{1}\) , Yuzhu Cao \(^{5}\) , Xianhong Tong \(^{5}\) , Weibing Qin \(^{6*}\) , Xiaoli Zhu \(^{5*}\) , Jianqiang Bao \(^{1,7*}\)  
+
+\(^{1}\) The First Affiliated Hospital of USTC, School of Basic Medical Sciences, Division of Life Sciences and Medicine, University of Science and Technology of China (USTC), Hefei, Anhui 230001, China; \(^{2}\) School of Information Science and Technology, University of Science and Technology of China (USTC), Hefei, Anhui 230001, China; \(^{3}\) Division of Life Sciences and Medicine, University of Science and Technology of China (USTC), Hefei, Anhui 230001, China, \(^{4}\) Laboratory animal centre, University of Science and Technology of China (USTC), Hefei, Anhui 23,0001, China; \(^{5}\) Reproductive and Genetic Hospital, The First Affiliated Hospital of USTC, Division of Life Sciences and Medicine, University of Science and Technology of China (USTC), Hefei, Anhui 230001, China; \(^{6}\) NHC Key Laboratory of Male Reproduction and Genetics, Guangdong Provincial Reproductive Science Institute (Guangdong Provincial Fertility Hospital), Guangzhou 510600, China; \(^{7}\) Reproductive and Genetic Hospital, The First Affiliated Hospital of USTC, Hefei National Research Center for Physical Sciences at the Microscale, Biomedical Sciences and Health Laboratory of Anhui Province, School of Basic Medical Sciences, Division of Life Sciences and Medicine, University of Science and Technology of China (USTC), Hefei, Anhui 230001, China.  
+
+\(^{*}\) To whom correspondence should be addressed. Tel: +86 0551 63606389; Email: jqbao@ustc.edu.cn Correspondence may also be addressed to Xiaoli Zhu. Tel: +86 0551 63607625; Email: xiaolizh@ustc.edu.cn Correspondence may also be addressed to Weibing Qin. Tel: +86 020 87692825; Email: guardqin@163. com \(^{\dagger}\) The authors wish it to be known that, in their opinion, the first five authors should be regarded as Joint First Authors.  
+
+Running Title: Nat10 is essential for mouse oocyte development  
+
+Keywords: Nat10; meiosis; oocyte growth; oocyte maturation; PolyA Tail length assay (PAT)
+
+<--- Page Split --->
+
+## Abstract  
+
+In mammals, the production of mature oocytes necessitates rigorous regulation of the discontinuous meiotic cell- cycle progression at both the transcriptional and post- transcriptional levels; however, the factors underlying this sophisticated but explicit process during oocyte development remain largely unclear. Here we characterized the function of N- acetyltransferase 10 (Nat10), which was previously recognized as a "writer" for N4- acetylcytidine (ac4C) deposited on RNA molecules. We generated two germline- specific Nat10 knockout mouse models in the embryonic gonad and in postnatal growing oocytes, and another two tamoxifen- inducible Nat10 deletion models. We provided genetic evidence showing that Nat10 is essential for oocyte meiotic prophase I progression, oocyte growth and maturation in mice. Intriguingly, we discovered that Nat10 is required for sculpting the maternal transcriptome through timely degradation of polyA tail mRNAs during the Maternal- to- Zygotic transition (MZT). Importantly, we developed a novel method that outperformed the conventional methods for examining the polyA tail length (PAT), termed Hairpin Adaptor- polyA tail length (HA- PAT), in terms of the cost, sensitivity and efficiency. In summary, these findings altogether provide solid genetic evidence that unveils the indispensable role of maternal Nat10 in oocyte development, and lay a solid foundation for future mechanistic studies of varied domains of NAT10.
+
+<--- Page Split --->
+
+## Introduction  
+
+In mammals, the germ cells are distinct from somatic cells in that they entail two successive cell divisions following one round of DNA replication, producing haploid gametes with parentally exchanged, but equal amounts of genetic DNA transmitted to the offspring 1. This sexually dimorphic event is achieved through a highly conserved and tightly controlled process, namely, meiosis. In mice, the primordial germ cells (PGCs) migrate and colonize the genital ridge prior to embryonic day 10.5 (E10.5). Subsequently, stimulated by the microenvironmental signaling emanating from the surrounding soma, the PGCs are sexually determined to commit to either male (XY) or female (XX) germ cells on E11.5. In the female, the ovary is morphologically distinguishable from the male testis by the lack of cord- like structure on E12.5 2. Unlike the prolonged mitotic arrest of male germ cells in embryonic testis, female germ cells (herein referred to as oogonia) initiate meiosis quickly starting from E13.5, and sequentially undergo stages of meiotic prophase I (leptotene, zygotene and pachytene), but are fully arrested at diplotene stage until pubertal LH signaling 3. It is well- known that the topoisomerase SPO11- mediated double- strand breaks (DSBs) are a prerequisite for the timely pairing, synapsis, and recombination between homologous chromosomes. Nonetheless, little is known about how genome- wide DSBs are specifically generated and repaired and which factors are involved in this extended stage of meiotic prophase I 4.  
+
+After birth, individual oocytes arrested at the diplotene of prophase I are encircled by a flattened layer of granulosa cells, which together are called primordial follicles. A clutch of primordial follicles is periodically recruited to the growth phase, which develops sequentially through stages of primary, secondary, early antral, antral, and large antral follicles, ensued by the ovulation upon LH surge 2. This whole process is exquisitely coordinated, as evidenced by massive gene transcription in conjugation with selective degradation of specific transcripts, which together build up the maternal transcriptome. The growing oocytes acquire the meiotic competence, i.e., the capability of resuming meiosis to MII stage, during the transition from the secondary to early antral follicles, while the acquisition of developmental competence occurs in the late stage of antral follicles that gives rise to the matured MII oocytes with the ability to be fertilized and develop to term (referred to as oocyte maturation) 2, 5, 6, 7. Since the zygote is transcriptionally silent prior to zygotic gene activation, all the initial developmental events during early embryonic reprogramming are dependent on the maternal RNA transcriptome inherited from the oocyte. Remarkably, the maternal transcriptome is particularly rich in mRNA, which is stably stored and occupies up to \(20\%\) of the total RNA in a fully grown GV oocyte, in contrast to the average \(\sim 2\%\) mRNA in somatic cells 8, 9. The transcriptional activity peaks in the early growing oocytes, but gradually decreases and is considered silent in the fully grown oocyte. Interestingly, this process is accompanied by DNA configuration transition from the less condensed, non- surrounded nucleolus (NSN) to the surrounded nucleolus (SN) state, wherein fully condensed chromatin DNA encompasses the nucleolus 10. Of note, compared with SN oocytes, NSN oocytes exhibit higher activity of transcription and lower developmental competence, with most zygotes arresting at the 2- cell stage. Moreover, genome- wide RNA transcriptome analyses revealed that SN oocytes display different gene expression profiles and metabolic pathway enrichment compared with NSN oocytes 10, 11. These studies revealed the significance of maternal transcriptome integrity in coordinating the timely oocyte growth and
+
+<--- Page Split --->
+
+maturation, and implicated its pathogenic roles, when disrupted, underling female infertility. Nevertheless, which factors define the maternal transcriptome and how they coordinate with each other through continuous stages of folliculogenesis remain poorly understood.  
+
+Our interest in exploring the role of Nat10 in oocyte development was initially piqued for two reasons. First, during oocyte maturation, there is a profound polyA- shortening mechanism- mediated mRNA decay that dramatically reshapes the maternal transcriptome. How particular mRNAs are selected for destruction is not well- understood \(^{6,12}\) . Nonetheless, this event must occur at the post- transcriptional level since the DNA transcription is progressively shut down concurrent with the oocyte growth. Previous studies have shown that multiple RNA modifications, such as m6A, m5C and m1A, are involved in post- transcriptional RNA metabolism, including alternative splicing, mRNA decay, and mRNA translation \(^{13,14}\) . Nat10 is highly expressed in the mouse ovary and is thus far the only known "writer" for epi- transcriptomic modification- N4- acetylcytidine (ac4C). In HeLa cells, it has been shown to enhance the stability and protein translation for mRNAs \(^{15}\) . Second, as described above, oocyte development is under tight, spatiotemporally specific regulation through discontinuous meiotic cell- cycle progression – rapid progression at early meiotic prophase I in the embryonic gonad, lengthened late prophase I arrest at diplotene during prepubertal development, and quick cytoplasmic and nuclear maturation during the GV- MII transition. Compelling studies have shown evidence related to Nat10's function in cell- cycle control in somatic cells \(^{16,17,18}\) . Interestingly, specific deletion of Nat10 in the male germline elicited severe defects resulting in male infertility owing to corrupted meiotic cell- cycle progression in mice \(^{19}\) . This evidence altogether is reminiscent of the critical role of Nat10 in female oocyte development.  
+
+In this study, we generated two germline- specific Nat10 knockout (KO) mouse models by Stra8- and Zp3- driven Cre expression and two additional tamoxifen (TMX)- inducible Nat10 KO mouse models using Ddx4- CreERT2 and Ubc- CreERT2. We revealed that Nat10 plays a profound role and is indispensable for oocyte meiotic cell- cycle progression in embryonic gonads and in postnatal oocyte growth and maturation. Importantly, we designed and optimized a novel method, termed HA- PAT, that outperformed previous conventional methods adopted for polyA tail length examination in single oocytes. Taking advantage of this approach, we uncovered that Nat10- mediated polyA tail shortening is a critical mechanism that defines the oocyte maternal transcriptome. Together, we provided compelling genetic evidence that validated the essential roles of Nat10 in mouse oocyte development.
+
+<--- Page Split --->
+
+## Results  
+
+## NAT10 is highly expressed and localized to the nucleolus in mouse oocytes  
+
+ResultsNAT10 is highly expressed and localized to the nucleolus in mouse oocytesTo explore the potential function of Nat10 during oogenesis, we first evaluated the multiple- tissue expression pattern of Nat10 in an array of mouse organs. NAT10 protein was abundantly expressed in the ovary, with the highest levels detected in the thymus, spleen, and testis, which is consistent with a recent study (Fig. 1A) \(^{19}\) . We further re- analyzed the published bulk RNA- seq datasets in the growing oocytes and pre- implantation embryos (GSE71434) \(^{22}\) . It showed that Nat10 mRNA was dynamically regulated in the postnatal growing follicles and preimplantation embryos, with the highest mRNA levels detected in GV oocytes and the lowest levels in 2- cell embryos (Fig. 1B). In agreement with this finding, quantitative real- time PCR (qPCR) validated the similar expression trend of Nat10 mRNA levels in different stages of oocytes and embryos (Fig. 1C). Next, we performed the fluorescent immunostaining (IF) with a NAT10 antibody in isolated oocytes at various stages as well as in ovary cryosections. As shown in Fig. 1D, NAT10 protein is abundantly present in the nucleus of the GV oocytes. The intensity of NAT10 signal is reduced and dispersed in the nucleus of MI and MII oocytes owing to the breakdown of the nuclear membrane (Fig. 1D). On the basis of the morphology and the number of surrounding granulosa cells, the follicles can be categorized into various stages during postnatal folliculogenesis in mice, including primordial follicle (PrF), primary follicle (PF), secondary follicle (SF), early antral follicle (EAF), and antral follicle (AF) (Fig. 1E) \(^{23}\) . Consistent with its mRNA expression trend, IF revealed abundant NAT10 protein being detected in the central nucleus of the growing oocytes at various stages, with the highest intensity in the nucleus center encircled by a layer of highly condensed chromatin (Fig. 1E). This distinctive expression pattern raised the possibility that it might be localized in the nucleolus, as reported by a few previous studies \(^{18, 24, 25}\) . To verify this possibility, we co- immunostained NAT10 and a nucleolus- specific marker, nucleophosmin (NPM). As shown in Fig. 1F, NAT10 is well co- localized to the nucleolus with NPM in the oocytes through various stages of folliculogenesis. It is worth noting that NAT10 is also highly expressed in the nucleolus of granulosa cells, particularly in antral follicles, as revealed by its perfect co- localization with NPM (Fig. 1F). This evidence together implies that NAT10 might have important physiological roles in mouse oocyte development in vivo.  
+
+## Pre-meiotic deletion of Nat10 caused follicular developmental arrest at primary follicles and premature ovarian failure (POF)  
+
+Pre- meiotic deletion of Nat10 caused follicular developmental arrest at primary follicles and premature ovarian failure (POF)Next, to decipher the physiological function of Nat10 in vivo, we generated pre- meiotic stage- specific Nat10 knockout (KO) mice by crossing floxed Nat10 (Nat10<sup>lox/lox</sup>) alleles with Stra8- GFPCre knockin alleles to obtain oocyte conditional Nat10 KO females (Nat10<sup>lox/- ; Stra8- GFPCre, hereafter called Nat10- ScKO) (Fig. S1A- C). The floxed Exons #4 and #5 (E4/5) reside in the DUF1726 domain of NAT10 protein (Fig. S1A). Stra8- GFPCre is specifically activated around embryonic day 12 (E12) in the primordial germ cells (PGCs) prior to meiosis in the female embryonic gonad (Fig. 2A). When crossing these deleter mice with Nat10<sup>lox/lox</sup> mice, E4/5 of the Nat10 gene were removed, resulting in the frameshift translation and presumably nonsense- mediated mRNA decay (NMD) (Fig. 2A- D, Fig. S1B and C). Using isolated GV oocytes, both western blotting and qPCR showed that the protein and mRNA levels
+
+<--- Page Split --->
+
+of Nat10 were markedly reduced in Nat10- ScKO oocytes compared with WT oocytes (Fig. 2C and D). Fertility testing showed that female Nat10- ScKO mice were completely sterile when crossed with WT males during 6 months of breeding (Fig. 2E). At 1 month, the size of Nat10- ScKO ovaries was reduced by eightfold compared with that in WT ovaries (Fig. 2F). To trace at which stage the Nat10 abrogation impacted follicle development, we next carried out the histological examination of Hematoxylin&Eosin (HE)- stained sections of paraffin- embedded ovaries during the first wave of postnatal follicle development. This unveiled that the follicles managed to proceed but arrested at the stage of primary follicles before P21 in the Nat10- ScKO females, unlike the WT ovaries where antral follicles were frequently observed (Fig. 2G). After P21, the Nat10- ScKO ovaries progressively lost all the characteristic follicular structure but harbored tubule- like structures filled with homogeneous, immature granulosa cells, which resembled the seminiferous tubules in the male testis (Fig. 2G). All these tubules were devoid of oocytes in the center, presumably owing to the quick degeneration of the Nat10- null oocytes, reminiscent of premature ovarian failure (POF) (Fig. 2G). Indeed, this phenotype of transdifferentiation of ovarian cells to Sertoli- like cells has been previously reported, wherein deletion of Mtor in the primordial oocytes induced the conversion of granulosa cells to Sertoli- like cells, displaying a seminiferous tubule- like testicular structure present in the oocyte- specific Mtor KO ovary 26. To test this possibility, we next designed a panel of granulosa cell- and Sertoli/Leydig cell- specific primers. As shown in Fig. 2H, while the expression levels of granulosa cell- specific markers (Amh, Cyp19a1 and Esr2) were significantly reduced in Nat10- ScKO ovaries compared with WT ovaries, we only observed the mRNA expression levels of two markers, Sox9 in Sertoli cells and Cldn11 in Leydig cells, among a selection of testis- enriched markers (Cyp11b1, Hsd3b6, Gata1, data not shown) were markedly elevated. This evidence implicated the partial conversion of granulosa cells to Sertoli cells upon Nat10 KO. In summary, pre- meiotic ablation of Nat10 in the female gonad led to female infertility due to defective follicular developmental arrest at the primary follicle stage and premature ovarian failure.  
+
+## Premiotic loss of Nat10 led to oocyte meiotic arrest at pachytene stage owing to disturbed DSB repair  
+
+In contrast to the male germline, the female germ cells initiate meiotic prophase I division early following sex determination in the embryonic gonad, and sequentially undergo leptotene, zygotene, and pachytene, but are finally arrested at the diplotene stage perinatally until further hormonal stimulation (Fig. 2A). Nat10 has recently been shown to be essential for meiotic divisions during spermatogenesis in the testis 19, thus we next investigated whether Nat10 is required for female meiotic prophase I progression in vivo. Co- staining of the nuclear chromosome spreads by SYCP1 and SYCP3 markers showed the accumulation of aberrant pachytene- like cells with partially synapsed homologous chromosomes in the perinatal Nat10- ScKO ovaries (Fig. 3A). Statistical comparison validated the elevated percentage of pachytene/pachytene- like cells and, as a result, the decreased proportion of oocytes at the diplotene stage in Nat10- ScKO ovaries (Fig. 3A and B), suggesting the meiotic arrest of Nat10- ScKO oocytes at the pachytene stage.
+
+<--- Page Split --->
+
+Premature oocyte death during the first wave of folliculogenesis, as described above (Fig. 2G), is most often a consequence of a self- surveillance mechanism for the host to safeguard genome integrity against unsynapsed chromosomes or excessive double- strand DNA breaks (DSBs) \(^{27,28,29}\) . There was a much higher occurrence of aberrant chromosomes that were not fully synapsed, as described above, in the perinatal Nat10- ScKO ovaries (Fig. 3A and B), we thus next assessed the causative factors that account for the premature oocyte loss upon Nat10 KO. IF staining by \(\gamma\) H2AX revealed the elevated DSB signals in oocytes at the pachytene stage, but not at the diplotene stage, in embryonic Nat10- ScKO ovaries (Fig. 3C and D), suggesting defective DSB repair in Nat10- null pachytene oocytes \(^{30}\) . Further examination by staining with RPA2, a marker that exclusively labels unrepaired DSBs, unveiled that more RPA2 foci were present in the Nat10- ScKO oocytes at pachytene stage, rather than at diplotene stage (Fig. 3E and F), which presumably correspond to oocytes with elevated \(\gamma\) H2AX staining resulting from DSB repair deficiency in Nat10- ScKO ovaries \(^{30}\) . Together, this evidence suggests that Nat10 is essential for meiotic prophase I progression in the female embryonic gonad.  
+
+## Nat10 is essential for the chromatin configuration NSN-SN transition in growing oocytes during postnatal folliculogenesis  
+
+At birth, the oocytes arrest at the diplotene stage in meiotic prophase I, and each is surrounded by a single layer of flattened granulosa cells, which together constitute the primordial follicle pool (Fig. 2A). Upon pubertal stimulation by FSH and LH, the follicles sequentially enter the growth and maturation stages (Fig. 2A). We thus next evaluated whether Nat10 is required for oocyte growth through generation of a Nat10- specific deletion mouse model in growing oocytes by crossing the Nat10 \(^{lox/lox}\) alleles with female Zp3- Cre (henceforth termed Nat10- ZcKO), which is specifically activated in the oocytes of primary follicles (Fig. 2A, Fig. 4A) \(^{31}\) . IF staining using P21 ovaries and isolated GV oocytes showed that Nat10 protein is specifically eliminated from the oocytes, but not in the granulosa cells (Fig. 4B and C, Fig. S2A and B). During a half- year fertility testing, Nat10- ZcKO females were completely sterile, suggesting Nat10 is indispensable for oocyte growth during postnatal ovarian development (Fig. 4D- E). We next performed H&E staining and counted the average numbers of follicles at various stages in the postnatal ovary sections. At 1 month, there was a slight reduction in the size of the Nat10- ZcKO ovary (Fig. 4E), but the morphological features and the proportion of follicles at various stages were indistinguishable between the Nat10- ZcKO and WT ovaries (Fig. 4E- G). Nevertheless, after the first wave of folliculogenesis, Nat10- ZcKO oocytes appeared to quickly degenerate, resulting in developmental arrest at secondary follicles in Nat10- ZcKO ovaries, as compared with WT ovaries (Fig. 4E- G, Fig. S2C and D).  
+
+On the other hand, the permissive sterile phenotype in all the Nat10- ZcKO females implies that there exits functional deficiency in Nat10- ZcKO ovaries despite their morphological similarity to WT ovaries at the age of 1 month. Therefore, we next collected and stained oocytes from the PMSG- primed females at P21. The average numbers of GV oocytes retrieved were comparable between Nat10- ZcKO and WT ovaries (Fig. 5A and B). Notably, more bulged granules appeared to be observed in the cytoplasm of Nat10- ZcKO oocytes (Fig. 5A). In growing oocytes, cytoplasmic maturation is accompanied by the Non- Surrounded Nucleolus (NSN)- to- Surrounded Nucleolus (SN) DNA configuration transition \(^{10}\) . SN
+
+<--- Page Split --->
+
+oocytes mostly dominate the GV stage in late antral follicles and are considered meiotically competent. Compelling studies have previously shown that DNA is transcriptionally inert in SN oocytes, characterized by the enhanced modifications of H3K4me3 and H3K9me3, while NSN oocytes exhibit active DNA transcription with decreased H3K4me3 and H3K9me3 modifications 11. Indeed, further counting of NSN versus SN oocytes uncovered that the ratio of NSN to SN was distorted between Nat10- ZcKO and WT ovaries (Fig. 5C and D). IF staining by H3K4me3 showed markedly decreased H3K4me3 intensity in both NSN and SN oocytes in Nat10- ZcKO oocytes compared with WT oocytes (Fig. 5E and F). In contrast, H3K9me3 staining revealed dispersed and elevated chromatin signals in both NSN and SN oocytes in Nat10- ZcKO oocytes (Fig. 5G and H). This evidence suggests that the transcriptional machinery might be disrupted resulting in the compromised meiotic competence in Nat10- ZcKO oocytes. Together, these studies demonstrate that Nat10 is required for oocyte growth in developing follicles and hence for female fertility.  
+
+## Nat10 ablation impaired meiotic GV-MII progression owing to transcriptome disruption  
+
+The sterility and the premature oocyte death in the Nat10- ZcKO females suggested that Nat10 is pivotal for meiotic maturation (GV- MII progression). To test this possibility, we next sought to isolate and culture GV oocytes in vitro and determine whether Nat10- ZcKO oocytes are capable of resuming meiosis. At 3 h after IBMX release in M16 medium, most WT GV oocytes ( \(\sim 90\%\) ) resumed meiotic division and entered the pro- metaphase I, as evidenced by the nuclear membrane breakdown of the germinal vesicles (GVBD), while a lower percentage of Nat10- ZcKO GV oocytes ( \(\sim 58\%\) ) managed to complete GVBD (Fig. 6A and B, Fig. S4A- D). Consistently, the number of Nat10- ZcKO oocytes that progressed to MII stage significantly declined compared to that of WT oocytes ( \(23 \pm 2.5\) vs \(65 \pm 1.51\) ) (Fig. 6C). To determine the precise arrested stage and how meiotic divisions were impacted in Nat10- ZcKO oocytes, we collected the superovulated oocytes in vivo after PMSG/hCG injection and performed the co- staining with DAPI and a cytoskeleton marker, tubulin, in the formaldehyde- fixed oocytes. The average number of collected oocytes significantly declined in Nat10- ZcKO ovaries compared with WT ovaries ( \(3.8 \pm 1.15\) vs \(30.57 \pm 0.92\) ) (Fig. 6D). In agreement with previous findings, most Nat10- ZcKO oocytes were arrested at the MI stage, with a small fraction of oocytes exhibiting anaphase- to- telophase arrest in Prophase I (AI- TI) in Nat10- ZcKO ovaries compared to WT oocytes (Fig. 6E and F). Furthermore, we carried out an In vitro fertilization (IVF) assay using superovulated MII oocytes. Nat10- ZcKO oocytes appeared to be fertilized and developed to the 2- cell stage, similar to the WT oocytes (Fig. 6G and H). However, the proportion of 4- cell stage embryos markedly declined in the maternal Nat10 KO group. Together, these studies suggest that Nat10 is indispensable for oocyte meiotic maturation (Fig. 6).  
+
+Oocyte growth is under stringent transcriptional regulation, with peak transcription occurring in growing oocytes at the early stage of antral follicles 6, 11, 32. To address the factors underlying defective GV oocytes in Nat10- ZcKO ovaries, we carried out RNA- seq analyses with GV oocytes retrieved from WT and Nat10- ZcKO ovaries. In agreement with the sterile severity, a total of 1615 differentially expressed genes (DEGs) were identified (Cutoff: Fold change (FC) \(\geq 2\) , \(p< 0.05\) ), with similar numbers of genes up- regulated (839) and down- regulated (776) in the Nat10- ZcKO GV oocytes relative to the WT
+
+<--- Page Split --->
+
+oocytes (Fig. S3A and C, Supplementary Table S4). Interestingly, Gene ontology (GO) analyses revealed that most down- regulated genes were enriched in transcription- related pathways, whereas the up- regulated genes were related to tRNA processing and meiotic cell cycles (Fig. S3B and D). Moreover, there were a total of 583 genes related to cell- cycle progression showing an alternative splicing pattern in Nat10- ZcKO oocytes (Fig. S3E- G). This evidence altogether demonstrated that the maternal transcriptome was disrupted in Nat10- ZcKO oocytes at the GV stage resulting in the impaired oocyte maturation.  
+
+## Mini-bulk SMART-seq2 identified defective maternal mRNA decay in Nat10-null MII oocytes  
+
+Progression through oocyte development is accompanied by tightly regulated RNA transcription and degradation 6, 33, 34, 35, 36. Transcription is active in growing oocytes, but is shut down when oocytes progress to the SN- type GV stage. Once meiosis resumes, the GV maternal transcriptome undergoes a global but selective degradation of \(\sim 20\%\) polyA mRNAs, culminating in a characteristic maternal transcriptome in MII oocytes that differs from that in GV oocytes 7, 12, 37, 38. To interrogate the molecular mechanism underlying the disrupted oocyte maturation in Nat10- ZcKO oocytes, we performed RNA- seq analyses. Given that one Nat10- ZcKO female only superovulated \(\sim 4\) MII oocytes on average, we further optimized an in- house mini- bulk SMART- seq2 protocol that utilized 3- 5 oocytes for each biological replicate for RNA- seq (Fig. 7A and B). We first verified the validity of our mini- bulk SMART- seq2 method by comparing our data with published bulk RNA- seq result in WT oocytes. On average, our method detected \(\sim 13337\) genes in GV and \(\sim 12071\) genes in MII, which are comparable to the \(\sim 13629\) genes and \(\sim 12045\) genes detected in GV and MII (Cutoff: TPM \(\geq 1\) ), respectively, in the bulk oocyte RNA- seq datasets (Fig. 7B, Supplementary Table S4 and 5) 22. This result confirmed the validity and sensitivity of our mini- bulk SMART- seq2 protocol.  
+
+Next, we conducted the mini- bulk SMART- seq2 using 5 oocytes at MII stage from WT and Nat10- ZcKO females. This revealed a higher number of genes (1196) to be up- regulated than down- regulated (555) in Nat10- ZcKO MII oocytes (Fig. 7C, Fig. S4E and F, Supplementary Table S5). GO analyses showed that the up- regulated genes were mostly enriched in translation- and mRNA processing- related biological processes (Fig. 7D- F). By comparison, the down- regulated genes were enriched in transcription- related GO terms (Fig. S4G). A comparison of the relative expression levels of the transcripts showed that a higher number of transcripts was present in the Nat10- ZcKO MII oocytes than in the WT (TPM \(\geq 1\) ) (Fig. 7G and H), suggesting an aberrant accumulation of maternal transcripts. To distinguish what specific type of transcripts was affected, the expressed transcripts were divided into five bins according to their expression levels in the WT MII oocytes. This revealed that Nat10 KO caused global transcript up- regulation regardless of their expression abundance (Fig. 7I). Since there is global polyA mRNA degradation during GV- MII transition in WT oocytes, the expressed transcripts in WT MII oocytes were allocated into three types: Up- [FC(MII/GV) \(\geq 2\) , p<0.05], Down- [FC(MII/GV) \(\leq - 2\) , p<0.05], and Stable- type (remaining transcripts). A Sankey plot showed that the majority of up- regulated genes (965/1196) in the Nat10- ZcKO MII oocytes overlapped with the Down- type genes in the WT MII oocytes, suggesting that the 965 transcripts destined for decay during meiotic maturation failed to be eliminated in the Nat10- ZcKO MII oocytes (Fig. 7J). To determine which transcripts were susceptible to
+
+<--- Page Split --->
+
+degradation in the presence of Nat10, we defined a total of 2011 transcripts down- regulated in WT oocytes during the GV- MII transition and 1206 transcripts down- regulated in Nat10- ZcKO oocytes through GV- MII transition. Interestingly, a large fraction of 1416 transcripts (among 2011 transcripts in total) did not overlap with down- regulated transcripts in Nat10- ZcKO oocytes, suggesting that they were not timely degraded but aberrantly accumulated in Nat10- ZcKO oocytes (Fig. 7K). The degradation trend profiling verified that the 1416 transcripts indeed displayed a decreased degradation speed in the absence of Nat10 (Fig. 7L). Of note, a total of 442 genes displayed an aberrantly alternative splicing pattern (Fig. S4H and I). Taken together, these studies suggest that Nat10 is essential for the maintenance of normal maternal transcriptome by the timely degradation of selective maternal mRNAs.  
+
+## Hairpin Adaptor- PolyA Tail length (HA-PAT) assay, a simple, low-cost and sensitive method for validation of polyA mRNA degradation in single oocytes  
+
+CCR4- NOT deadenylase complex mediated polyA mRNA degradation has been well- documented to be responsible for maternal mRNA degradation during the MZT \(^{33,39,40}\) . Indeed, we discovered that \(\sim 65\%\) (782/1195) of transcripts up- regulated in Nat10- ZcKO MII oocytes were also accumulated in the Cnot6l (CCR4- NOT subunit) KO MII oocytes (Fig. 8A). Intriguingly, these up- regulated genes were especially enriched in ribosomal subunit components that were normally degraded through polyA tail- shortening mediated mRNA decay via the CCR4- NOT complex, a phenotype that was also observed in CCR4- NOT- deficient oocytes (Fig. 8B) \(^{33,41}\) . Further examination by qPCR analysis revealed that the mRNA levels of three members (Cnot6l, Cnot7 and Btg4) of CCR4- NOT pathway were significantly down- regulated in the Nat10- ZcKO MII oocytes compared with WT oocytes (Fig. 8C). This evidence led us to hypothesize that Nat10- driven maintenance of CCR4- NOT deadenylase activity is required for the polyA mRNA elimination and consequently the establishment of the maternal transcriptome.  
+
+We thus next tested whether the accumulated transcripts harbor long polyA tails in Nat10- ZcKO MII oocytes. Currently, two methods are commonly adopted for PolyA tail length (PAT) examination, including Ligase- mediated PAT (LM- PAT) and extension PAT (ePAT) \(^{42,43,44}\) . The original LM- PAT method relies on oligo(dT) \(_{12 - 18}\) hybridization and T4 ligation, followed by oligo- (dT) anchor primer- mediated PCR amplification \(^{43,44}\) ; ePAT exploits a hybridized oligo(dT) anchor primer as a DNA template for Klenow enzyme- mediated 3' extension ensued by PCR amplification \(^{42}\) . Given the limited availability of Nat10- null MII oocytes, we initially tried both methods using only 5\~10 oocytes as input materials; however, the PCR amplification either failed or produced very weak, unsatisfactory results, presumably owing to the low sensitivity of both methods when dealing with low- input samples (data not shown). Therefore, we next sought to develop a new PAT assay that can circumvent the drawbacks of previous approaches (Fig. 8D). After several rounds of optimization, we termed this novel method the Hairpin Adaptor- PolyA Tail length (HA- PAT), which is more sensitive, low- cost and time- saving than previous methods \(^{42,44}\) . HA- PAT utilizes an exquisitely designed hairpin adaptor that can self- hybridize plus an extended oligo (dT) \(_{8}\) with two additional degenerate W nucleotides at the 3' end (Fig. 8D). The loop sequence is connected through a C3 spacer, which provides a sufficiently flexible linker to enhance the self- hybridization of stem sequences. In practice, this strategy was optimized to facilitate specific anchoring at the 3' end of polyA tail and to reduce the random dT hybridization background. Moreover,
+
+<--- Page Split --->
+
+given that the 3' ends of polyA mRNA tails dominate with two U nucleotides in the oocytes 7, the two degenerate "WW" nucleotides will further capture and stabilize the binding between the hairpin adaptor and the "U"- terminated polyA tails. Therefore, this single hairpin adaptor not only provides an anchor primer for reverse transcription, but also harbors the full reverse primer sequence for subsequent cDNA PCR amplification. Thanks to its integral design strategy, the total reagent cost and hands- on operations were significantly reduced compared with the other two common methods (Fig. S5A). To test its efficiency, we selected two known genes (Rpl35a and Chchd2) that accumulated in Mll Nat10- null oocytes and the house- keeping Gapdh gene as a control. As shown in Fig. S5B, using 5 oocytes as input material without PCR preamplification, neither LM- PAT nor ePAT could detect Rpl35a and Chchd2 genes. In contrast, our HA- PAT assay yielded clear, differential smear bands between WT and Nat10- null oocytes, suggesting that the HA- PAT approach is more sensitive in detecting low- input mRNA samples (Fig. S5B). Under PCR preamplification for 16 cycles using five oocytes, LM- PAT showed strong signals for the smear polyA tails, whereas ePAT failed to produce satisfactory results (Fig. S8D). Nonetheless, our HA- PAT method outperformed the other two methods since it gave rise to stronger and clearer bands (Fig. S5B). This evidence validated the sensitivity and efficiency of the LM- PAT assay in detecting polyA tail length using a minute amount of input RNA material from single oocytes.  
+
+Taking advantage of the HA- PAT approach, we next selected a panel of representative genes showing up- regulation in each GO cluster identified by mini- bulk SMART- Seq2 for validation. We collected five hormone- primed oocytes at GV and Mll stages for each sample and executed 16 cycles of PCR amplification. As shown in Fig. 8E- F, all the polyA tails of the seven genes were lengthened in Mll Nat10- null oocytes, as judged by the elevated, smear PCR bands. Additionally, as a cross- validation, we further performed LM- PAT, which corroborated the similar findings (Fig. S6). Taken together, these results indicate that Nat10- mediated polyA- tail shortening and the resulting mRNA degradation sculpted the transcriptome of Mll oocytes.  
+
+## Evidence that maternal Nat10 is translationally required for pre-implantation embryo development  
+
+As an RNA acetyltransferase, Nat10 harbors a C- terminal tRNA- binding domain and was originally identified and experimentally validated to modify two well- known eukaryotic RNA substrates: tRNAsSer/Leu and 18S rRNA \(^{24,45}\) . Both types of RNA are canonical noncoding RNAs that are involved in mRNA translation to protein products. We next asked whether maternal Nat10 is physiologically important for protein translation. Oocytes are unique in their properties, as they amass abundant mRNAs, of which most are stored for translation during later maternal- zygotic transition (MZT), owing to the uncoupling of transcription and translation in maturing oocytes \(^{46}\) . As such, oocytes were often broadly exploited as a model to study cis- element mediated mRNA translation. To this end, we generated an in- house knockin, tamoxifen (TMX)- inducible Ddx4- cre mouse model (Ddx4- CreERT2), in which the Cre activity is specifically turned on in Ddx4- expressing germ cells upon TMX injection (data not shown, manuscript in preparation). Ddx4- CreERT2 females were crossed with Nat10 \(^{lox/lox}\) males to attain Nat10 \(^{lox/lox}\) ; Ddx4- CreERT2 offspring (henceforth termed Nat10- DcKO following TMX injection) (Fig. S7A). We performed the TMX injection for three consecutive days starting from P17,
+
+<--- Page Split --->
+
+when the oocytes had already proceeded to the GV stage in the antral follicles during the first wave of postnatal folliculogenesis. At P21, we injected PMSG and retrieved the oocytes 48 h later for culture in vitro. As shown in Fig. S7B, the mRNA levels of Nat10 were significantly reduced in Nat10- DcKO oocytes compared with WT oocytes, indicative of the successive deletion of Nat10 in oocytes. However, we did not observe any distinguishable difference in terms of the gross morphology, the average numbers, or the percentage of PMSG- primed oocytes at GV stage, and at MI or MII stages when cultured in vitro (Fig. S7C- I). The morphology or average numbers of superovulated MII oocytes following PMSG/hCG priming were also indistinguishable between WT and Nat10- DcKO females (Fig. S7J). Interestingly, we performed IVF using superovulated WT and Nat10- DcKO MII oocytes with WT sperm, which unveiled that the pre- implantation embryos were arrested at the 2- cell stage in Nat10- DcKO zygotes, suggesting that maternal Nat10 is essential for pre- implantation embryo development (Fig. S7K and L). Considering the high prevalence of Nat10 mRNA in oocytes at MI, MII and at 2- cell embryos (Fig. 1C), the absence of phenotypic outcome in Nat10- DcKO oocytes presumably hinted that the residual Nat10 mRNA or protein was still functional after TMX- induced Cre activation.  
+
+To further corroborate this finding, we utilized another well- established global Cre- inducible mouse model, namely Ubc- CreERT2, to perform similar crossings to obtain Nat10<sup>lox/lox</sup>; Ubc- CreERT2 (termed Nat10- UcKO) mice after TMX injection (Fig. S8A). Following TMX injection, Ubc- Cre was activated in all cell types, including somatic granulosa cells and oocytes in the ovary. Next, we carried out all subsequent experiments similar to those for Ddx4- CreERT2 as described above. In line with previous studies, all results were identical to the conclusions as drawn in Nat10- DcKO mice with only one exception – the preimplantation embryos were arrested at 1- cell zygote stage (Fig. S8). The discrepancy of earlier zygotic arrest between Nat10- DcKO- and Nat10- UcKO- derived embryos was likely due to the defective function of surrounding granulosa cells since Nat10 protein is also highly abundant in the nucleolus of the supporting granulosa cells (Fig. 1F). Given that no active transcription but exclusive translation occurs during late oocyte- to- zygote transition, these studies were reminiscent of an integral physiological role of Nat10 in coordination of mRNA translation during MZT.  
+
+Next, we sought out to examine how Nat10 impacted mRNA translation. Given the limited availability of oocytes, we decided to generate stable, Cre- inducible cell lines by exclusively utilizing our established Nat10<sup>lox/lox</sup>; Ubc- CreERT2 mice. We followed a “3T3” protocol and successfully generated two stable, 4- hydroxy tamoxifen (4'- OHT)- inducible, mouse embryonic fibroblast (MEF) cell lines from E13 embryos (Nat10<sup>lox/lox</sup>; Ubc- CreERT2) after 2- month in vitro culture (Fig. 9A) <sup>20</sup>. Immunoblotting and qPCR analyses demonstrated that Nat10 protein and mRNAs were vastly declined upon OHT induction, respectively (Fig. 9B and C), indicative of the successful establishment of two stable cell lines. Apparently, Nat10 depletion significantly reduced the cell proliferation without any effect on cell apoptosis, as evidenced by the Ki67 staining and the CCK8 assays (Fig. 9D- H). Reciprocally, rescue by Nat10 overexpression in Nat10- inducible KO MEF cells recovered and enhanced the cell division (Fig. 9F- H). Noteworthily, we consistently observed that longer induction of Nat10- UcKO MEF cells by 4'- OHT treatment led to complete cell cycle arrest and resulting cell death, suggesting that Nat10 is essential for cell survival. The permissive cell death in Nat10- UcKO MEF cells contrasted with the
+
+<--- Page Split --->
+
+viability observed in Nat10- null HeLa cells \(^{15}\) . Next, we pursued to explore how translation was impacted in Nat10 KO MEF cells. We collected similar numbers of MEF cells without and with 4'- OHT treatment for three days for polysome profiling. Consistently, we found that Nat10 KO cells displayed lower levels of mRNA- bound polysomes than non- treated MEF cells, indicating the translational efficiency was repressed in the Nat10- null MEFs (Fig. 9l). Intriguingly, both the levels of the 40S subunit and 80S ribosome were concurrently reduced in KO cells compared with WT cells (Fig. 9l and J). We reasoned that this presumably resulted from the assembly defect of the 40S ribosomal subunit protein binding to 18S rRNA, whose ac4C modification is essential for ribosome assembly as previously reported \(^{45}\) . In summary, these studies provided evidence that Nat10 is required for mouse pre- implantation embryo development, at least in part, through modulating mRNA translation.  
+
+## Discussion  
+
+## Oocyte development, RNA modification, and ac4C modification  
+
+Dysregulated expression of Nat10 has been recently linked to numerous diseases, such as Hutchinson- Gilford progeria syndrome (HGPS) \(^{47}\) , epithelial ovarian cancer \(^{48}\) , breast cancer \(^{16}\) , and spermatogenesis \(^{19}\) , and is most recently implicated in oocyte in vitro maturation (IVM) \(^{49}\) . Mature oocytes are specialized, transcriptionally inert cells that almost exclusively contribute to the cytoplasm of zygotes when fertilized by sperm. Therefore, all the early developmental events of preimplantation embryos occur at the post- transcriptional level and are dependent on the stored RNA content derived from oocytes, called the maternal transcriptome \(^{6, 32, 35, 36}\) . In mammals, a hallmark of gene expression for the maternal transcriptome is characterized by the uncoupling of transcription and translation – some accumulated mRNA species are immediately translated to support oocyte growth, while a large stock of other species is stabilized as ribonucleoprotein particles (RNPs) in a translationally inactive state, in growing follicles \(^{6, 38, 39}\) . When fully- grown GV oocytes resume meiosis, known as a process called meiotic maturation, however, the situation substantially changes – many previously active mRNAs become translationally silent, whereas some previously “dormant” mRNA species are translationally reactivated. A massive wave of RNA elimination is considered as a hallmark and a major driving force underlying oocyte- to- embryo transition (OET). It has been estimated that up to \(\sim 40\%\) polyA RNA is degraded during the GV- to- MII oocyte transition; however, the molecular mechanisms underlying this RNA remodeling event are not well understood \(^{6, 38, 39}\) .  
+
+Post- transcriptional RNA modifications constitute an exciting layer of so- called “epi- transcriptome” that modulates gene expression without altering RNA sequences. Recent evidence showing N6- methyladenosine (m6A) modification abundantly present in mRNA species is an attractive model that selectively marks any mRNA for degradation at the post- transcriptional level. For instance, m6A is specifically deposited by a large, heterogeneous multiprotein complex, known as the m6A “writer”, which comprises a core catalytic member of METTL3, along with METTL14, WTAP and KIAA1429 (VIRMA) cofactors \(^{50}\) . Genetic evidence in conjunction with m6A antibody- mediated RNA- immunoprecipitation sequencing (m6A- seq, or MeRIP- seq) showed that the m6A mark specifically decorates the 3'UTR close to the STOP codon region, and defective MZT transition is tightly associated with reduced m6A levels in m6A- deficient oocytes, suggestive of an essential role of m6A in active RNA
+
+<--- Page Split --->
+
+degradation through the MZT \(^{5, 14, 34, 51}\) . On the other hand, to be functional, the m6A mark on the substrate mRNA must first be recognized by the "reader" proteins, e.g., YTH domain- containing protein family (YTHDF1- 3, YTHDC1- 2), which interpret and relay substrates to the downstream signaling. However, unexpectedly, more recent compelling evidence validated that, either the "writer" METTL3/METTL14 or YTHDC2, exerts their critical biological functions independent of the m6A modification in somatic cells and the germline, respectively \(^{52, 53, 54}\) . For example, METTL3/METTL14 drives local chromatin remodeling through binding to target genomic loci, whereas YTHDC2 binds a class of mRNAs containing U- rich motif in their 3'UTRs and coding sequences. This evidence indicates that we might overinterpret the role of m6A in mediating its diverse biological functions observed in either the "writer" or the "reader" knockout mouse models. Indeed, this mechanistic discrepancy could be explained by methodological limitations - most previous studies that used high- throughput sequencing for mRNA m6A identification were highly dependent on the anti- m6A antibody, which is known to suffer from notable limitations, such as non- specificity, low resolution, and high input RNA materials \(^{55}\) . In addition, those conclusions were most often indirectly drawn based on the correlation between the phenotypic outcome and the declined overall levels of m6A, but lacked the direct evidence gauging the m6A stoichiometry to its functional readout.  
+
+In contrast to m6A modification, which is mostly prevalent ( \(\sim 0.5\%\) ) in mRNA, the abundance of ac4C modification in mammalian mRNAs is currently controversial \(^{56}\) . An initial study identified that ac4C tends to enrich in the 5'UTR region in thousands of coding mRNAs, on the basis of ac4C antibody- mediated immunoprecipitation and sequencing (acRIP- seq), similar to the m6A- seq strategy, in HeLa cells \(^{15, 57}\) . However, surprisingly, a later study unveiled no even a single ac4C site in mammalian mRNAs through a convincing chemistry- catalyzed strategy with quantitative and nucleotide resolution \(^{58, 59}\) . Importantly, as a positive control, they successfully identified the previously known ac4C sites in both rRNA and tRNAs, as well as low levels of a few hundred of ac4C acetylation sites in Nat10- overexpressing human cells, implying that their adopted method is sensitive and feasible for detecting ac4C modification \(^{58, 59}\) . Hence, the ac4C modification levels are, if present, extremely low in human mRNAs. The discrepancy between the two studies likely arises from the promiscuous binding of the ac4C antibody, as described above, which suffers from poor characterization and intrinsic cross- reactivity. We have found that, consistent with a recent study \(^{19}\) , the ac4C/C ratio approximates \(\sim 0.04\%\) in mRNAs from mouse testicular cells, but is roughly 1/10 of the m6A abundance in mouse testis (data not shown), based on mass- spectrometry (MS) measurement \(^{58, 59}\) . This prompted extremely low levels, or negligible levels of ac4C modification in mammalian mRNAs, and hence, we reasoned that Nat10 exerts its critical functions independent of ac4C acetylation. Nonetheless, owing to the large mRNA requirement, we were not able to perform the whole- transcriptome acRIP- seq to assess the effect of ac4C modification using isolated RNA from Nat10- null oocytes.  
+
+## The pleiotropic roles of Nat10 through oocyte meiotic progression and maturation  
+
+The NAT10 is a highly conserved protein from E. coli and yeast to mammalian species, and is the solely known enzyme responsible for ac4C modification on RNA substrates. It is a relatively large protein that comprises four distinct domains, including DUF1726, Helicase, GNAT and tRNA binding
+
+<--- Page Split --->
+
+domain. In this study, we employed four Cre deleter mouse models to cross with floxed Nat10 alleles to attain germline- specific or inducible Nat10 KO mice, yielding full Nat10- null mouse offspring without any of the four domains owing to premature frame- shift reading. Consistent with the high expression levels of Nat10 in the female germline, pre- meiotic ablation of Nat10 caused apparent meiotic arrest at the pachytene stage, resulting in premature ovarian failure and female infertility in adults (Fig. 2). We showed that this defect was likely caused by the deficient DSB repair as judged by the high persistent \(\gamma \mathrm{H}2\mathrm{AX}\) intensity and enhanced RPA2 remnant (Fig. 3). This conclusion agrees with a previous study showing that DNA- damaging agents induced Nat10 expression in a dosage- and time- dependent fashion \(^{60}\) . Nat10 depletion in growing oocytes of primary follicles disrupted the NSN- SN transition of GV oocytes and damaged meiotic maturation, as evidenced by the aberrant GV to MII transition \(^{10}\) . By optimized mini- bulk SMART- seq2 analyses, we discovered that a large number of genes enriched in cell cycles and DNA transcription were dysregulated in Nat10- null GV oocytes (Fig. S3). In particular, we revealed that many genes destined for degradation in the MII oocytes were aberrantly accumulated in Nat10- null MII oocytes, which was corroborated by a novel, in- house developed HA- PAT approach. Further evidence showed that Nat10 deletion caused the down- regulated expression of important members of the CCR4- NOT complex, at least in part, at the transcriptional level \(^{33,39}\) . These studies demonstrate that NAT10 is transcriptionally essential to maintain transcriptomic homeostasis to facilitate oocyte growth and maturation. Finally, we provided both genetic and in vitro evidences showing that NAT10 is translationally required for oocyte developmental competence, since Nat10 depletion caused a remarkable reduction in translational efficiency, as evidenced by the polysome profiling. This most likely resulted from the assembly defects of the ribosomes, a conclusion that is consistent with a previous study wherein down- regulation of Nat10 disrupted the biogenesis of 40S ribosomal subunit due to abolished ac4C modification on 18S rRNA \(^{24,45}\) .  
+
+In conclusion, we provide genetic and in vitro studies showing that Nat10 is transcriptionally required for the prophase I progression and meiotic resumption during oocyte growth and meiotic maturation. Nat10 is also translationally indispensable for the developmental competence of the oocytes. However, further experiments are urgently needed to elucidate the pathogenic roles of the distinct domains of NAT10 and to decipher whether there is a causative relationship, if present, between the ac4C RNA modification and the phenotypic outcome.  
+
+## Methods  
+
+## Mice  
+
+The floxed Nat10 (Nat10<sup>lox/lo</sup>) alleles were generated by GemPharmatech Co., Ltd. Conditional Nat10 knockout mice were achieved by crossing Nat10<sup>lox/lo</sup> mice with Stra8- GFPCre and with Zp3- Cre mice to attain the Nat10- ScKO and Nat10- ZcKO offspring, respectively. Induced Nat10 KO mice were generated by tamoxifen injection for three consecutive days in Ubc- CreERT2 or Ddx4- CreERT2 mice. The Stra8- GFPCre knock- in (KI) mouse line was generated in Ming- Han Tong's Lab at the Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences. Zp3- Cre and Ubc- CreERT2 KI mice were obtained from Jackson Laboratory. Ddx4- CreERT2 KI mouse line was generated in- house.
+
+<--- Page Split --->
+
+All mice were from the C57BL/6J background, and were bred in a specific pathogen- free (SPF) facility with a 12h light/dark cycle and with free access to food and water. All animal experiments were approved by the Animal Care and Research Committee of the University of Science and Technology of China (USTC).  
+
+## Inducible Nat10 KO cell lines and culture  
+
+We generated two tamoxifen- inducible, stable cell lines following a standard 3T3 protocol \(^{20}\) . In brief, the pregnant female mice were euthanized, and the embryos at embryonic day 13.5 (E13.5) were carefully dissected. After removing somatic organs, the embryonic body was chopped into small pieces and digested with DMEM medium containing \(0.25\%\) trypsin- EDTA in a \(37^{\circ}C\) water bath for 30 min. After filtering through a \(70 \mu \mathrm{m}\) cell strainer, the single cells were cultured in MEF medium (DMEM medium supplemented with \(10\%\) FBS (VivaCell, C04001) and \(1\%\) Penicillin- Streptomycin (Biosharp, BL505A)). The cells were subcultured and the medium was replenished every three days, with the passage number recorded. Cells were grown in a \(5\% \mathrm{CO}_2\) cell culture incubator (Heal Force) at \(37^{\circ}C\) . Two stable MEF lines were achieved after passage 23 following recovery of the MEF cells from the crisis around passage \(\sim 10 - 14\) (total time is \(\sim 2\) months).  
+
+## Lentiviral transduction  
+
+A mouse Nat10 cDNA plasmid clone (EX- Mm12636- M45) was purchased from GeneCopoeia, Inc. and cloned into pCDH- CMV vectors in- frame with a FLAG tag. The lentiviral vector was co- transduced into 293T cells alongside a packaging vector psPAX2 and a helper vector pCMV- VSVG using LIP2000 transfection reagent (Biosharp, BL623B) for lentivirus production. Viral particles were collected to infect MEF cells with \(8 \mathrm{mg / ml}\) Polybrene (Solarbio, H8761). The infected cells were positively selected with puromycin (2.5 mg/ml) (Solarbio, P8230) for 48 hours.  
+
+## Oocyte collection and in vitro culture  
+
+The 21- day- old female (P21) mice were injected with 5 IU of pregnant mare's serum gonadotropin (PMSG) (Ningbo Sansheng Pharmaceutical). After 48 h, the mice were euthanized and the oocytes at the GV stage were harvested in M2 medium (Nanjing Aibei Biotechnology Co., Ltd, M1250) and further cultured in M16 medium (Sigma, M7292) covered with mineral oil (Sigma, M5310) at \(37^{\circ}C\) in \(5\% \mathrm{CO}_2\) . Samples were imaged with a microscope (SZX7, Olympus).  
+
+## Superovulation and in vitro fertilization  
+
+For superovulation, 21- day- old female mice were injected with 5 IU of PMSG. After 46 h, the mice were injected with 5 IU of human chorionic gonadotropin (hCG) (Ningbo Sansheng Pharmaceutical). At post- hCG 16 h, the cumulus- oocyte complex (COC) was retrieved from the oviducts, and the oocytes were counted after digestion with \(0.3\%\) hyaluronidase (Sigma, H4272). For in vitro fertilization (IVF), superovulated female mice were euthanized 15 h after hCG injection, and the ampulla parts of the oviducts were collected in HTF medium covered by mineral oil. COCs were released and fertilized with WT sperm in HTF for 30 min at \(37^{\circ}C\) in a \(5\% \mathrm{CO}_2\) incubator, and further cultured in KSOM medium in vitro (Sigma, MR- 101).  
+
+## RNA isolation and real-time RT-PCR
+
+<--- Page Split --->
+
+Total RNA was isolated from mouse tissues with TRIzol Reagent following the manufacturer's instruction as described previously 21. In brief, freshly collected or frozen tissues were homogenized in TRIzol reagent. For MEF cells, total RNA was isolated with a SPARKeasy Cell RNA kit (Shandong Sparkjade Biotechnology Co., Ltd., AC0205). The quantity and quality of RNA samples were determined by measurement using a NanoPhotometer N50 (Implen, Germany). The RNA samples with OD values of \(260 / 280 \geq 1.9\) were selected for downstream analyses. Equal amounts of total RNA were loaded to synthesize cDNAs using the Hiscript III Reverse Transcriptase (Vazyme, R302- 1). Quantitative PCR (qPCR) was performed using Hieff® qPCR SYBR Green Master Mix (Yeasen, 11201ES03) on a Q2000B Real- Time PCR machine (LongGene). For oocytes, 5\~10 oocytes were lysed in 2 μl lysis buffer (0.2% Triton X- 100 and 2 IU/μl RNase inhibitor) followed by reverse transcription and PCR pre- amplification for 8\~16 cycles. The PCR products were diluted and used for qPCR. The primers are listed in Supplementary Table S1.  
+
+## Histological analysis  
+
+Hematoxylin & Eosin (HE) staining was performed following the standard procedure as described previously 21. In brief, ovary samples were freshly collected, and fixed in Bouin's solution at room temperature overnight. Paraffin- embedded samples were cut into slides with 5 μm thickness. Slides were de- paraffinized with xylene and re- hydrated, followed by staining with HE. The slides were then dehydrated and mounted with neutral resins. Images were taken on a light microscopy (MShot) with a MSX2 camera.  
+
+## Western blot analysis  
+
+Samples were freshly collected and lysed in RIPA solution [100mM Tris- HCl (PH7.4), 1% Triton X- 100, 1% Sodium deoxycholate, 0.1% SDS, 0.15M NaCl, supplemented with Protease inhibitor cocktail]. Protein concentrations were determined using a BCA protein assay kit. All protein samples were run in 8% of denatured sodium dodecyl sulfate polyacrylamide (SDS- PAGE) gel with Trelief® Prestained Protein Ladder (TSINGKE, TSP021), followed by wet- transfer to PVDF membranes. Subsequently, the membranes were blocked in 1XPBS with 5% non- fat milk and incubated with primary antibody followed by secondary antibody. Signals were visualized using an imaging system (SHST, Hangzhou, China). The following antibodies were used: rabbit anti- NAT10 (ZENBIO, 389412, 1:1000), mouse anti- PCNA (Proteintech, 60097- 1- lg, 1:2000), mouse anti- GAPDH (Proteintech, 60004- 1- lg, 1:10000), rabbit anti- Tubulin (Proteintech, 11224- 1- AP, 1:5000).  
+
+## Chromosome spreads analysis and immunofluorescent staining  
+
+For oocyte chromosome spreading analyses, newborn pups were sacrificed, and ovaries were dissected in hypotonic buffer [30mM Tris- HCL (pH=8.2), 50mM sucrose, 17mM trisodium citrate dihydrate, 5mM EDTA (pH=8.0), 1mM dithiothreitol (DTT), and 1mM phenylmethylsulfonyl fluoride (PMSF)] for 25 min. Next, ovaries were transferred to 100 mM sucrose, and single cells were released into sucrose solution using syringe needles. Single cells were spread and fixed in 1% PFA solution containing 0.15% Triton X- 100 on slides, followed by washing with 0.4% Photo- Flo. For immunofluorescent staining, cells were permeabilized with 0.3% Triton X- 100, blocked with 5% normal goat serum (Solarbio, SL038) in PBST, and incubated with primary antibodies diluted in blocking
+
+<--- Page Split --->
+
+solution at \(4^{\circ}C\) overnight. Antibodies used were as follows: mouse anti- SYCP3 (Abcam, ab97672, 1:1000), rabbit anti- SYCP1 (Abcam, ab15090, 1:1000), rabbit anti- SYCP3 (Proteintech, 23024- 1- AP, 1:200), mouse anti- \(\gamma \mathsf{H}_2\mathsf{A}\mathsf{X}\) (Millipore, 05- 636, 1:1000), rabbit anti- RPA2 (Proteintech, 10412- 1- AP, 1:400). After washing in 1XPBS (0.1% Tween) for three times, the samples were incubated with TRITC- conjugated Goat Anti- Rabbit IgG (Proteintech, gb2AF488) or 488- conjugated secondary antibodies (Proteintech, gb2AF488). For ovarian immunofluorescent staining, ovaries were fixed in \(4\%\) paraformaldehyde (PFA) overnight at \(4^{\circ}C\) on a rocker. Slides were dehydrated with \(10\%\) and \(20\%\) sucrose for 2 h each. Ovary samples were cut into \(8\mu \mathrm{m}\) slides. For immunofluorescence staining of MEF cells and oocytes, they were fixed in \(4\%\) PFA for 30min. The primary antibodies used were as follows: rabbit anti- NAT10 (Proteintech, 13365- 1- AP, 1:400), rabbit anti- NAT10 (ZENBIO, 389412, 1:500), rabbit anti- Nucleophosmin (Abcam, ab10530, 1:1000), rabbit anti- H3K4me3 (Abclonal, A2357, 1:200), rabbit anti- \(\alpha\) - Tubulin (Proteintech, 11224- 1- AP, 1:200), rabbit anti- KI67 (Servicebio, GB111141, 1:400), rabbit anti- FLAG (Proteintech, 80010- 1- RR, 1:400). Slides were imaged by a Leica THUNDER Imager Live Cell with a K5 camera driven by the Leica Application Suite Software. Image processing was performed by ImageJ software.  
+
+## PolyA-tail (PAT) length assay  
+
+For the Hairpin- Adaptor PAT (HA- PAT) assay, ten denuded oocytes were freshly lysed in \(2\mu \mathrm{l}\) lysis buffer (0.2% Triton X- 100 and 2 IU/ul RNase inhibitor). The hairpin adaptor was pre- annealed in 1Xoligo annealing buffer (50mM Tris, pH8.0, 50mM NaCl, 1mM EDTA). PolyA- tail mRNAs in the oocyte lysates were denatured at \(72^{\circ}C\) for 3 min and then hybridized with the annealed hairpin adaptor at \(25^{\circ}C\) for 10min. The reverse transcription mix contained 100 U of SuperScript IV, 10 U of RNase Inhibitor, 5 mM of DTT, 1 M of Betaine, 6 mM of MgCl2, 1 mM of dCTP (for template- switching) and 1.5 M of P5TSO. The \(1^{\mathrm{st}}\) strand cDNA was synthesized by reverse transcription at \(42^{\circ}C\) for 90 min. The full- length cDNAs were pre- amplified through semi- suppressive PCR for 8 or 16 cycles. The pre- amplified cDNA products were diluted and used for gene- specific PCR reactions using gene- specific primers (GSP) and polyA reverse primer 2. For the Ligation- Mediated PolyA Test (LM- PAT), oocyte samples were hybridized with oligo(dT) \(_{20}\) at \(42^{\circ}C\) for 30min, and then ligated with dT anchor primer by T4 DNA ligase (Sangon, B600511) at \(12^{\circ}C\) for 2h. Reverse transcription was performed with P5TSO and 1mM dCTP for 1h, ensued by PCR pre- amplification for 8 or 16 cycles using dT anchor and P5TSO. The diluted PCR products were used for gene- specific amplification. For extension PolyA Test (ePAT), oocyte polyA- tail RNAs were \(3^{\prime}\) - prime extended using ePAT anchor primer as a template with Klenow enzyme (New England Biolabs, K0210) at \(25^{\circ}C\) for 1h and \(80^{\circ}C\) for 10min. Reverse transcription was performed with P5TSO and 1mM dCTP for 1h, followed by pre- amplification for 8 or 16 cycles using ePAT anchor and P5TSO primers. PCR products were analyzed on a \(2\%\) agarose gel. All PCR primers are listed in Supplementary Table S2.  
+
+## Cell proliferation and apoptosis assays  
+
+For the cell proliferation assay, 2000 MEF and OHT- treated MEF cells were plated in 96- well plates with \(200\mu \mathrm{L}\) of fresh complete DMEM medium supplemented with \(10\%\) FBS. After incubation for 1, 3, 5 and 7 days, cell viabilities were measured using Cell Counting Kit- 8 (MedChemExpress, HY- K0301)
+
+<--- Page Split --->
+
+following the manufacturer's instructions. For apoptosis assays, an Annexin V- FITC/PI Apoptosis Detection Kit (YEASEN, 40302ES50) was used following the manufacturer's protocol. Samples were detected by the BD Accuri C6 flow cytometry, and the results were analyzed with FlowJo V10 software.  
+
+## Polysome profiling  
+
+For polysome profiling, cells were cultured in \(10 - cm\) dishes and treated with OHT for 3 days. Cells were washed with cold PBS supplemented with \(100\mu g / ml\) cycloheximide and collected by centrifugation. Cell pellets were lysed in lysis buffer [50 mM HEPES, \(2mM MgCl_2\) , \(100mM KCl\) , \(100\mu g / ml\) cycloheximide, \(1mM DTT\) , \(0.5\%\) Triton X- 100, \(10\%\) glycerol, and \(20U / ml\) EDTA- free protease inhibitor cocktail (Sigma, 11836170001)]. The lysate was cleared by centrifugation at \(12,000g\) for \(10\min\) at \(4^{\circ}C\) and the supernatant was loaded onto a \(20 - 50\%\) density gradient of sucrose cushion [30 mmol/l Tris- HCl (pH 7.5), \(100mmol / lNaCl\) , \(10mmol / lMgCl2\) , protease inhibitor cocktail (EDTA- free), and 100 units/ml RNase inhibitor (APExBIO, K1046)], ensued by ultracentrifugation in a rotor at \(38,000rpm\) for \(3h\) at \(4^{\circ}C\) . After centrifugation, the gradient was fractionated and the absorbance at \(254nm\) was continuously recorded using an ISCO fractionator (Teledyne ISCO).  
+
+## Mini-bulk SMART-seq2 for RNA-seq library preparation  
+
+The oocyte mini- bulk SMART- seq2 protocol was based on the well- established SMART- seq2 protocol with in- house optimized modifications as indicated below. In brief, after hormone challenge, five oocytes retrieved from each animal were washed at least five times in 1XPBS containing \(0.5\%\) BSA/PVP, and directly lysed in Lysis Buffer MasterMix ( \(0.3\%\) Triton, \(40U / \mu L\) RNase inhibitor, \(2.5\mu M\) oligodT30VN, and \(2.5\mu M\) dNTP mix). The oocyte Lysis mixture was allowed to undergo at least one freeze- thaw cycle at \(- 80^{\circ}C\) to facilitate complete cytosolic lysis. The \(1^{\text{st}}\) strand cDNA synthesis was performed at \(42^{\circ}C\) for \(90\min\) , followed by 14 cycles of PCR preamplification to attain the full- length cDNA products through ISPCR primer- mediated semi- suppressive PCR. \(1\mg\) of size- selected full- length cDNAs was used for Tn5- guided library preparation using Hieff NGS Fast Tagment DNA Library Prep Kit for Illumina (Yeasen, 12206ES96). Final dual- barcoded libraries were achieved through PCR amplification for 8 cycles with both index i5 and i7 primers prior to pooled library sequencing on the NovaSeq 6000 platform with PE150 mode (Novagene).  
+
+## RNA-Seq data analysis  
+
+Raw reads were processed to remove adaptor contaminants and low- quality bases. The clean reads were aligned to the mouse genome (mm10) using STAR, and uniquely mapped reads were counted with RSEM by default parameters (Supplementary Table S3). We quantified gene expression levels with TPM. For each sample, the expressed genes were defined with cutoff: TPM≥1. Differentially expressed genes (DEGs) were assessed with the DESeq2 package with a cutoff: Padj <0.05 and fold change (FC) ≥2. Gene Ontology (GO) enrichment was performed using DAVID (https://david.ncifcrf.gov/). rMATS was used to analyze the alternative splicing events. Statistical analyses were performed using R software (http://www.rproject.org). The RNA- seq data in this work have been submitted in the NCBI Gene Expression Omnibus (GEO) under accession number SRP392832.
+
+<--- Page Split --->
+
+## Statistical analysis  
+
+All experiments were performed at least in biological triplicates unless otherwise indicated. Statistical analysis was performed using Student's t- test unless otherwise stated. Values of \(p< 0.05\) were deemed statistically significant. Statistically significant values of \(p< 0.05\) , \(p< 0.01\) , \(p< 0.001\) and \(p< 0.0001\) are indicated by one, two, three and four asterisks, respectively. Statistical data were calculated by R or GraphPad Prism 6.  
+
+## Data availability  
+
+All the raw data and processed files have been deposited in the Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo) with the accession number: SRP392832
+
+<--- Page Split --->
+
+## References  
+
+1. Xu C, Cao Y, Bao J. Building RNA-protein germ granules: insights from the multifaceted functions of DEAD-box helicase Vasa/Ddx4 in germline development. Cell Mol Life Sci 79, 4 (2021).2. Jagarlamudi K, Rajkovic A. Oogenesis: transcriptional regulators and mouse models. Mol Cell Endocrinol 356, 31-39 (2012).3. Koubova J, Menke DB, Zhou Q, Capel B, Griswold MD, Page DC. Retinoic acid regulates sex-specific timing of meiotic initiation in mice. Proc Natl Acad Sci U S A 103, 2474-2479 (2006).4. Guan Y, et al. SCF ubiquitin E3 ligase regulates DNA double-strand breaks in early meiotic recombination. Nucleic Acids Res 50, 5129-5144 (2022).5. Hu Y, et al. Oocyte competence is maintained by m(6)A methyltransferase KIAA1429-mediated RNA metabolism during mouse follicular development. Cell Death Differ 27, 2468-2483 (2020).6. Svoboda P, Franke V, Schultz RM. Sculpting the Transcriptome During the Oocyte-to-Embryo Transition in Mouse. Curr Top Dev Biol 113, 305-349 (2015).7. Morgan M, et al. mRNA 3' uridylation and poly(A) tail length sculpt the mammalian maternal transcriptome. Nature 548, 347-351 (2017).8. Mole MA, Weberling A, Zernicka-Goetz M. Comparative analysis of human and mouse development: From zygote to pre-gastrulation. Curr Top Dev Biol 136, 113-138 (2020).9. Timmers HTM, Tora L. Transcript Buffering: A Balancing Act between mRNA Synthesis and mRNA Degradation. Mol Cell 72, 10-17 (2018).10. Ma JY, et al. Maternal factors required for oocyte developmental competence in mice: transcriptome analysis of non-surrounded nucleolus (NSN) and surrounded nucleolus (SN) oocytes. Cell Cycle 12, 1928-1938 (2013).11. Dumdie JN, et al. Chromatin Modification and Global Transcriptional Silencing in the Oocyte Mediated by the mRNA Decay Activator ZFP36L2. Dev Cell 44, 392-402 e397 (2018).12. Su YQ, et al. Selective degradation of transcripts during meiotic maturation of mouse oocytes. Dev Biol 302, 104-117 (2007).13. Hu L, et al. m(6)A RNA modifications are measured at single-base resolution across the mammalian transcriptome. Nat Biotechnol 40, 1210-1219 (2022).14. Zhao BS, et al. m(6)A-dependent maternal mRNA clearance facilitates zebrafish maternal-to-zygotic transition. Nature 542, 475-478 (2017).15. Arango D, et al. Acetylation of Cythidine in mRNA Promotes Translation Efficiency. Cell 175, 1872-1886 e1824 (2018).16. Liu HY, et al. Acetylation of MORC2 by NAT10 regulates cell-cycle checkpoint control and resistance to DNA-damaging chemotherapy and radiotherapy in breast cancer. Nucleic Acids Res 48, 3638-3656 (2020).17. Zheng J, et al. NAT10 regulates mitotic cell fate by acetylating Eg5 to control bipolar spindle assembly and chromosome segregation. Cell Death Differ 29, 846-860 (2022).18. Shen Q, et al. NAT10, a nucleolar protein, localizes to the midbody and regulates cytokinesis and acetylation of microtubules. Exp Cell Res 315, 1653-1667 (2009).
+
+<--- Page Split --->
+
+19. Chen L, et al. NAT10-mediated N4-acetylcytidine modification is required for meiosis entry and progression in male germ cells. Nucleic Acids Res, (2022).  
+
+20. Xu J. Preparation, culture, and immortalization of mouse embryonic fibroblasts. Curr Protoc Mol Biol Chapter 28, Unit 28 21 (2005).  
+
+21. Jiang X, et al. The Spin1 interactor, Spindoc, is dispensable for meiotic division, but essential for haploid spermatid development in mice. Reprod Biol Endocrinol 19, 144 (2021).  
+
+22. Zhang B, et al. Allelic reprogramming of the histone modification H3K4me3 in early mammalian development. Nature 537, 553-557 (2016).  
+
+23. Ivanova I, et al. The RNA m(6)A Reader YTHDF2 Is Essential for the Post-transcriptional Regulation of the Maternal Transcriptome and Oocyte Competence. Mol Cell 67, 1059-1067 e1054 (2017).  
+
+24. Sharma S, Langhendries JL, Watzinger P, Kotter P, Entian KD, Lafontaine DL. Yeast Kre33 and human NAT10 are conserved 18S rRNA cytosine acetyltransferases that modify tRNAs assisted by the adaptor Tan1/THUMPD1. Nucleic Acids Res 43, 2242-2258 (2015).  
+
+25. Liu X, et al. Deacetylation of NAT10 by Sirt1 promotes the transition from rRNA biogenesis to autophagy upon energy stress. Nucleic Acids Res 46, 9601-9616 (2018).  
+
+26. Guo J, et al. Oocyte stage-specific effects of MTOR determine granulosa cell fate and oocyte quality in mice. Proc Natl Acad Sci U S A 115, E5326-E5333 (2018).  
+
+27. Bolcun-Filas E, Rinaldi VD, White ME, Schimenti JC. Reversal of female infertility by Chk2 ablation reveals the oocyte DNA damage checkpoint pathway. Science 343, 533-536 (2014).  
+
+28. Rinaldi VD, Bolcun-Filas E, Kogo H, Kurahashi H, Schimenti JC. The DNA Damage Checkpoint Eliminates Mouse Oocytes with Chromosome Synapsis Failure. Mol Cell 67, 1026-1036 e1022 (2017).  
+
+29. MacQueen AJ, Hochwagen A. Checkpoint mechanisms: the puppet masters of meiotic prophase. Trends Cell Biol 21, 393-400 (2011).  
+
+30. Cloutier JM, Mahadevaiah SK, Ellnati E, Nussenzweig A, Toth A, Turner JM. Histone H2AFX Links Meiotic Chromosome Asynapsis to Prophase I Oocyte Loss in Mammals. PLoS Genet 11, e1005462 (2015).  
+
+31. Lan ZJ, Xu X, Cooney AJ. Differential oocyte-specific expression of Cre recombinase activity in GDF-9-iCre, Zp3cre, and Msx2Cre transgenic mice. Biol Reprod 71, 1469-1474 (2004).  
+
+32. Pan H, O'Brien M J, Wigglesworth K, Eppig JJ, Schultz RM. Transcript profiling during mouse oocyte development and the effect of gonadotropin priming and development in vitro. Dev Biol 286, 493-506 (2005).  
+
+33. Sha QQ, et al. CNOT6L couples the selective degradation of maternal transcripts to meiotic cell cycle progression in mouse oocyte. EMBO J 37, (2018).  
+
+34. Wu Y, et al. N(6)-methyladenosine regulates maternal RNA maintenance in oocytes and timely RNA decay during mouse maternal-to-zygotic transition. Nat Cell Biol 24, 917-927 (2022).  
+
+35. Ma J, Flemr M, Strnad H, Svoboda P, Schultz RM. Maternally recruited DCP1A and DCP2 contribute to messenger RNA degradation during oocyte maturation and genome activation in mouse. Biol Reprod 88, 11 (2013).  
+
+36. Li L, Lu X, Dean J. The maternal to zygotic transition in mammals. Mol Aspects Med 34, 919-938 (2013).
+
+<--- Page Split --->
+
+37. Reyes JM, Ross PJ. Cytoplasmic polyadenylation in mammalian oocyte maturation. Wiley Interdiscip Rev RNA 7, 71-89 (2016).  
+
+38. Tora L, Vincent SD. What defines the maternal transcriptome? Biochem Soc Trans 49, 2051-2062 (2021).  
+
+39. Yu C, et al. BTG4 is a meiotic cell cycle-coupled maternal-zygotic-transition licensing factor in oocytes. Nat Struct Mol Biol 23, 387-394 (2016).  
+
+40. Horvat F, et al. Role of Cnot6l in maternal mRNA turnover. Life Sci Alliance 1, e201800084 (2018).  
+
+41. Ma J, Fukuda Y, Schultz RM. Mobilization of Dormant Cnot7 mRNA Promotes Deadenylation of Maternal Transcripts During Mouse Oocyte Maturation. Biol Reprod 93, 48 (2015).  
+
+42. Janicke A, Vancuylenberg J, Boag PR, Traven A, Beilharz TH. ePAT: a simple method to tag adenylated RNA to measure poly(A)-tail length and other 3' RACE applications. RNA 18, 1289-1295 (2012).  
+
+43. Salles FJ, Strickland S. Rapid and sensitive analysis of mRNA polyadenylation states by PCR. PCR Methods Appl 4, 317-321 (1995).  
+
+44. Bilska A, Krawczyk PS, Dziembowski A, Mroczek S. Measuring the tail: Methods for poly(A) tail profiling. Wiley Interdiscip Rev RNA, e1737 (2022).  
+
+45. Ito S, et al. Human NAT10 is an ATP-dependent RNA acetyltransferase responsible for N4-acetylcytidine formation in 18 S ribosomal RNA (rRNA). J Biol Chem 289, 35724-35730 (2014).  
+
+46. Luong XG, Daldello EM, Rajkovic G, Yang CR, Conti M. Genome-wide analysis reveals a switch in the translational program upon oocyte meiotic resumption. Nucleic Acids Res 48, 3257-3276 (2020).  
+
+47. Larrieu D, Britton S, Demir M, Rodriguez R, Jackson SP. Chemical inhibition of NAT10 corrects defects of laminopathic cells. Science 344, 527-532 (2014).  
+
+48. Tan TZ, et al. Functional genomics identifies five distinct molecular subtypes with clinical relevance and pathways for growth control in epithelial ovarian cancer. EMBO Mol Med 5, 1051-1066 (2013).  
+
+49. Xiang Y, et al. NAT10-Mediated N4-Acetylcytidine of RNA Contributes to Post-transcriptional Regulation of Mouse Oocyte Maturation in vitro. Front Cell Dev Biol 9, 704341 (2021).  
+
+50. Poh HX, Mirza AH, Pickering BF, Jaffrey SR. Alternative splicing of METTL3 explains apparently METTL3-independent m6A modifications in mRNA. PLoS Biol 20, e3001683 (2022).  
+
+51. Sui X, et al. METTL3-mediated m(6)A is required for murine oocyte maturation and maternal-to-zygotic transition. Cell Cycle 19, 391-404 (2020).  
+
+52. Liu P, et al. m(6)A-independent genome-wide METTL3 and METTL14 redistribution drives the senescence-associated secretory phenotype. Nat Cell Biol 23, 355-365 (2021).  
+
+53. Saito Y, et al. YTHDC2 control of gametogenesis requires helicase activity but not m(6)A binding. Genes Dev 36, 180-194 (2022).  
+
+54. Su R, et al. METTL16 exerts an m(6)A-independent function to facilitate translation and tumorigenesis. Nat Cell Biol 24, 205-216 (2022).  
+
+55. Helm M, Motorin Y. Detecting RNA modifications in the epitranscriptome: predict and validate. Nat Rev Genet 18, 275-291 (2017).
+
+<--- Page Split --->
+
+56. Wiener D, Schwartz S. The eptitranscriptome beyond m(6)A. Nat Rev Genet 22, 119-131 (2021). 
+
+57. Arango D, Sturgill D, Oberdoerffer S. Immunoprecipitation and Sequencing of Acetylated RNA. Bio Protoc 9, e3278 (2019). 
+
+58. Sas-Chen A, et al. Dynamic RNA acetylation revealed by quantitative cross-evolutionary mapping. Nature 583, 638-643 (2020). 
+
+59. Thomas JM, Bryson KM, Meier JL. Nucleotide resolution sequencing of N4-acetylcytidine in RNA. Methods Enzymol 621, 31-51 (2019). 
+
+60. Liu H, Ling Y, Gong Y, Sun Y, Hou L, Zhang B. DNA damage induces N-acetyltransferase NAT10 gene expression through transcriptional activation. Mol Cell Biochem 300, 249-258 (2007).
+
+<--- Page Split --->
+
+## Acknowledgements  
+
+AcknowledgementsWe are grateful to Prof. Li He (School of Basic Medical Sciences, Division of Life Sciences and Medicine, University of Science and Technology of China (USTC), Hefei, Anhui) for generous assistance with oocyte imaging. We also thank all members of Bao laboratory for helpful discussion. This work was supported by grants from the Ministry of Science and Technology of China (2019YFA0802600); National Natural Science Foundation of China (31970793, 32170856); The open project of NHC Key Laboratory of Male Reproduction and Genetics (No. KF201901); the Fundamental Research Funds for the Central Universities" (WK9110000181, WK2070000156) and Startup funding (KY9100000001) from USTC.  
+
+## Author contributions  
+
+Author contributionsJ.Q.B., X.L.Z and W.B.Q. conceived, designed, and supervised the work. J.Q.B. wrote the manuscript. X.J. performed mouse crossing, chromosome spreads analysis and immunofluorescent staining. Y. C. collected equal amounts of mouse oocytes used in this study and western blot analysis. Y. Z. Z. did polyA- tail length assay. C.L.X analyzed the RNA- seq data and generated the figures. Q.D.L. collected mouse oocytes used for RNA- seq. X.M.X. did HE staining of mouse ovaries with the help of W.Q.L. J.Q.Z. and L.M. performed mouse genotyping. M.A. and Y.Z.C. helped with the immunofluorescence.  
+
+## Supplementary Data  
+
+Supplementary DataSupplementary Fig. S1\~8; Supplementary Table S1\~5  
+
+## Competing interests  
+
+The authors declare no competing interests.  
+
+## Correspondence  
+
+CorrespondenceCorrespondence and requests for materials should be addressed to Weibing Qin, Xiaoli Zhu or Jianqiang Bao.
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Fig.1. Expression and Localization of Nat10 enriched in the nucleolus of oocytes in mice. (A) Western blot showing the relative expression levels of NAT10 protein among multiple organs in adult WT mice. GAPDH served as a loading control. (B) Dynamic mRNA expression levels of Nat10 from RNA-seq analyses in oocytes and preimplantation embryos in mice (GSE71434). ICM, Inner cell mass. (C) Quantitative RT-PCR results showing the relative expression levels of mouse Nat10 mRNA in oocytes, and preimplantation embryos. Data were presented as mean±SEM, n=3. GO, growing oocytes collected from postnatal 14-day-old (P14) female mice. (D) Immunofluorescence (IF) staining of NAT10 in growing (GO), GV, MI, and MII oocytes as indicated. Dashed circle indicates cellular membrane of oocytes. DNA was counterstained with 4',6-diamidino-2-phenylindole (DAPI). Scale bar, 20 μm. (E) IF images of 21-day-old WT ovarian cryosections stained with anti-NAT10 antibody (Red) and DAPI (Blue) for follicles at various stages (primordial, primary, secondary, early antral, and antral stages) as </center>
+
+<--- Page Split --->
+
+indicated. Scale bar, \(20 \mu \mathrm{m}\) . Boxed insert area is a magnified view of the oocyte in the respective follicles. Arrows point to NAT10- positive nucleus of the oocyte in mouse developing follicles. (F) IF images of 21- day- old WT ovarian sections co- stained with NAT10 antibody (Red), Nucleophosmin (NPM, Green) and DAPI (Blue) in the follicles as indicated. Scale bar, \(20 \mu \mathrm{m}\) . Bottom panel is a magnified view of the oocyte in the respective follicles. Arrows point to the co- localization of NAT10 and NPM in the oocyte nucleolus at varied stages of follicles.
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Fig.2. Pre-meiotic deletion of Nat10 caused follicular developmental arrest and premature ovarian failure (POF). (A) Schematic diagram showing the landmark timeline of oocyte development from embryonic meiotic cell-cycle progression to postnatal oocyte growth and maturation. Stra8-GFPcre is activated prior to Embryonic day 13.5 (E13.5); Zp3-cre is active starting from P5 in the primary follicles; Both Ubc-CreERT2 and Ddx4-CreERT2 lines are Cre-inducible in all tissues and specifically in the germline, respectively, upon tamoxifen injection. (B) A breeding scheme by crossing Nat10<sup>lox/lox</sup> with </center>
+
+<--- Page Split --->
+
+Stra8- GFPCre to generate Nat10<sup>lox/- </sup>; Stra8- GFPCre (Nat10- ScKO) offspring. (C) Western blot analyses of the NAT10 protein levels in adult WT and Nat10- ScKO ovary. α- TUBULIN was used as a loading control. (D) Quantitative RT- PCR (qPCR) assay showing the relative expression levels of Nat10 mRNA in adult WT and Nat10- ScKO mouse ovary. Data are presented as mean± SEM, n=3. \*\*\*\*, p<0.0001 by two- tailed Student's t- test. (E) Fertility test showing the cumulative average numbers of pups from breeding of WT and Nat10- ScKO females with WT males. Data are presented as the mean± SEM, n=5, \*\*\*\*, p<0.0001 by two- tailed Student's t- test. (F) The gross morphology of ovaries derived from WT and Nat10- ScKO mice at 1M. Scale bar, 200 μm. (G) H&E staining of paraffin- embedded ovarian sections showing the histology of WT and Nat10- ScKO ovaries at postnatal days as indicated. Scale bar, 200 μm. High- resolution view of the boxed area is shown in parallel. Scale bar, 20 μm. Arrows point to follicles at stages as indicated. PrF, Primordial Follicle; PF, Primary Follicle; SF, Secondary Follicle; EAF, Early Antral Follicle; AF, Antral Follicle; LAF, Late Antral Follicle; (H) qPCR analyses of the relative expression levels for a cohort of genes showing specific or characteristic expression in ovarian granulosa cells (Left) or testicular Sertoli/Leydig cells (Right) in 1- month- old WT and Nat10- ScKO ovaries. Data are presented as the mean±SEM, n=3.
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Fig.3. Embryonic Nat10 loss caused oocyte meiotic prophase I arrest at pachytene stage owing to deficient DSB repair. (A) Immunofluorescence staining of oocyte nuclear chromosome spreads by SYCP3 and SYCP1 markers in WT and Nat10-ScKO mouse ovaries at birth. Scale bar, \(10 \mu \mathrm{m}\) . Arrows point to the asynapsed structure of the lateral and central axes. (B) The statistic counts showing the percentage of oocytes at various stages as indicated. Data are presented as the mean±SEM, \(n = 3\) , \(*\) , \(p< 0.05\) by two-tailed Student's t-test. (C) IF staining by SYCP3 and yH2AX on surface-spread oocytes at pachytene and diplotene stages from WT and Nat10-ScKO mouse oocytes at birth. Scale bar, 10 </center>
+
+<--- Page Split --->
+
+\(\mu \mathrm{m}\) . (D) The statistic counts showing the relative \(\gamma \mathrm{H}2\mathrm{AX}\) signal intensity calculated by ImageJ in pachytene and diplotene oocytes. Data are presented as mean± SEM, \(n = 3\) ; \*\*\*\*, \(p< 0.0001\) ; n.s., not significant by two- tailed Student's t- test. (E) IF staining on surface- spread oocytes by SYCP3 and RPA2 in WT and Nat10- ScKO mouse oocytes at birth. Scale bar, \(10 \mu \mathrm{m}\) . (F) Quantification of the numbers of RPA2 foci (representative of the unrepaired DSBs) in WT and Nat10- ScKO mouse oocytes at birth. Zyg, Zygotene; Pac, Pachytene; Pac- like, Pachytene- like; Dip, Diplotene. Data are presented as the mean± SEM, \(n = 3\) ; \* \(p< 0.05\) . n.s., not significant by two- tailed Student's t- test.
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Fig.4. Postnatal Nat10 depletion caused ovarian developmental arrest at secondary follicles. (A) A breeding scheme for Nat10 KO in growing oocytes of primary follicles by crossing Nat10<sup>lox/lox</sup> with Zp3-Cre deleter to attain Nat10<sup>lox/-</sup>; Zp3-Cre (Nat10-ZcKO) female offspring. (B-C) Immunofluorescence staining by NAT10 (Red), NPM (Green) and Hoechst 33342(Blue) in the secondary follicles (B) and GV oocytes (C) from WT and Nat10-ZcKO ovaries. Scale bar, \(20\mu \mathrm{m}\) . (D) Fertility test showing the cumulative numbers of pups from breedings of WT and Nat10-ZcKO females with WT males during a half-year caging. Data are presented as the mean± SEM, \(n = 5\) ; \(***\) , \(p< 0.0001\) by two-tailed Student's t-test. (E) The gross morphology of ovaries derived from WT and Nat10-ZcKO mice at age of 1 month (M) (left) and 2 months (right). Scale bar, \(200\mu \mathrm{m}\) . (F) H&E staining showing ovarian histology from WT and Nat10-ZcKO mice at 1 M (Top) and 2 M (Bottom). Scale bar, \(50\mu \mathrm{m}\) . Follicles are indicated by arrows. (G) Comparison of the average numbers of follicles at indicated stages in the ovaries of WT and Nat10-ZcKO mice at 1M (Top) and 2M (Bottom). Follicles were counted on serial ovarian sections </center>
+
+<--- Page Split --->
+
+after H&E staining. Data are presented as the mean± SEM, n=3; \*, p<0.05; \*\*, p<0.01; \*\*\*, p<0.001 by two-tailed Student's t-test. PO, Preovulatory Follicle.
+
+<--- Page Split --->
+![](images/Figure_5.jpg)
+
+<center>Fig.5. Postnatal Nat10 deficiency impedes oocyte chromatin NSN–SN configuration transition. </center>  
+
+(A) The gross morphology of oocytes at Germinal vesicle (GV) stage collected from PMSG-primed WT and Nat10-ZcKO females at P21. Scale bar, 100 μm. (B) Quantification of collected average numbers of GV oocytes. Data are presented as the mean±SEM, n=4. n.s., non-significant by two-tailed Student's t-test. (C) Hoechst 33342 (Blue) staining of the GV oocytes with non-surrounded nucleolus (NSN) and surrounded nucleolus (SN) chromatin configurations in WT and Nat10-ZcKO oocytes. Scale bar, 20 μm. (D) The percentage of NSN-type and SN-type oocytes isolated from WT and Nat10-ZcKO mice at P21. Data are presented as the mean±SEM, n=3. **, p<0.01 by two-tailed Student's t-test. (E-F) Immunofluorescence staining by H3K4me3 in NSN-type (Left) and SN-type (Right) oocytes from PMSG-primed WT and Nat10-ZcKO mice (E), and quantification of H3K4me3 intensity (F). Scale bar, 10 μm. Data are presented as the mean±SEM, n=3; ***, p<0.001 by two-tailed Student's t-test. (G-H) Immunofluorescence staining by H3K9me3 in NSN-type (Left) and SN-type (Right) oocytes from PMSG-primed WT and Nat10-ZcKO mice (G), and quantification of H3K9me3 intensity (H). Scale bar, 10 μm. Data are presented as the mean±SEM, n=3; *, p<0.05 by two-tailed Student's t-test.
+
+<--- Page Split --->
+![](images/Figure_6.jpg)
+
+<center>Fig.6. Postnatal Nat10 ablation led to defective oocyte meiotic maturation. (A) The gross morphology of oocytes collected at the time points as indicated for GV (0h), and cultured in vitro for MI </center>
+
+<--- Page Split --->
+
+(6h) and MII (16h) from PMSG- primed WT and Nat10- ZcKO females. Scale bar, \(100 \mu \mathrm{m}\) . (B- C) Percentage of oocytes at GVBD (B) and MII (C) after release of GV oocytes cultured in IBMX- containing medium from PMSG- primed WT and Nat10- ZcKO females. Data are presented as mean± SEM, \(n = 3\) . \(^{**}\) , \(p< 0.01\) , \(^{***}\) , \(p< 0.001\) by two- tailed Student's t- test. (D) Average numbers of superovulated oocytes at MII from WT (30.57±0.92) and Nat10- ZcKO (3.8±1.15) mice following PMSG and hCG injection in vivo. Data are presented as the mean± SEM, \(n = 5\) . \(^{***}\) , \(p< 0.001\) by two- tailed Student's t- test. (E- F) Immunofluorescence staining images of superovulated oocytes collected at 16h after hCG injection by \(\alpha\) - TUBULIN staining. Oocytes with MI arrest, Anaphase- to- telophase arrest in prophase I (AI- TI), and aberrant spindles were observed (E) and counted (F) in Nat10- ZcKO mice. Scale bar, \(40 \mu \mathrm{m}\) . Data are presented as the mean± SEM, \(n = 3\) . \(^{***}\) , \(p< 0.001\) by two- tailed Student's t- test. (G- H) Representative gross morphology of preimplantation embryos at various stages as indicated derived from superovulated WT and Nat10- ZcKO oocytes (after hCG priming) fertilized with WT sperm (G). Arrows point to the blastocysts; Quantitative comparison of the average numbers of preimplantation embryos at varied stages was shown (H). Scale bar, \(100 \mu \mathrm{m}\) . Data are presented as the mean± SEM, \(n = 5\) . \(^{***}\) , \(p< 0.0001\) by two- tailed Student's t- test.
+
+<--- Page Split --->
+![](images/Figure_7.jpg)
+
+<center>Fig.7. Mini-bulk SMART-seq2 analyses identified the dysregulated maternal transcriptome in Nat10-ZcKO oocytes. (A) A diagram showing mouse oocyte samples collected for mini-bulk SMART-seq2 analyses. (B) Bar graph showing the numbers of transcripts detected in WT and Nat10-ZcKO </center>
+
+<--- Page Split --->
+
+oocytes at GV and MII stages (TPM \(\geq 1\) ). Data are presented as the mean± SEM, n=3. n.s., nonsignificant by two- tailed Student's t- test. (C) Scatter plot of mini- bulk SMART- seq2 data showing differentially expressed genes (DEGs) in Nat10- ZcKO MII oocytes. Red color: Up- regulated; Blue color: Down- regulated; Cutoff: fold change (FC) \(\geq 2\) , adjusted p<0.05. The TPMs of WT and Nat10- ZcKO MII oocytes are listed in Supplementary Table S5. (D) Gene Ontology (GO) enrichment analysis of up- regulated genes in Nat10- ZcKO MII oocytes (Cutoff: FC \(\geq 2\) , adjusted p<0.05). (E) Heatmap of representative genes from four major functional GO categories showing up- regulated expression in Nat10- ScKO MII oocytes. The color intensity gradient from red to blue indicates the relative gene expression levels from high to low. (F) Bar plots showing the qPCR analyses of relative mRNA expression levels for a panel of up- regulated genes identified by mini- bulk SMART- seq2 in WT and Nat10- ZcKO MII oocytes. (G- H) Box plots showing the relative expression levels of the transcripts in WT and Nat10- ZcKO oocytes at the GV and MII stages as indicated (G); box plot showing the relative fold changes in mRNA levels in MII versus GV oocytes from WT and Nat10- ZcKO, respectively (H). The box indicates the upper and lower quartiles; the thick line in the box indicates the median, n = 3. P- values by a two- tailed Student's t- test are indicated. (I) Box plot showing gene expression levels in WT and Nat10- ZcKO oocytes at the GV and MII stages. Genes were divided into 5 bins according to their relative expression abundance in the WT MII oocytes. The box indicates upper and lower quartiles, n=3. (J) Sankey diagram showing the overlapping of the DEGs (1196 up- regulated vs 555 down- regulated) with genes exhibiting up- [FC(MII/GV) \(\geq 2\) , p<0.05], down- [FC(MII/GV) \(\leq -2\) , p<0.05], or stable expression patterns in WT MII relative to GV stage oocytes (TPM \(\geq 1\) ). (K) Venn diagram showing the overlapping of down- regulated transcripts between WT MII oocytes relative to GV oocytes (2011, Cutoff: TPM \(\geq 1\) , FC[GV/MII] \(\geq 5\) ), and Nat10- ZcKO MII oocytes relative to GV oocytes (1206, Cutoff: TPM \(\geq 1\) , FC[GV/MII] \(\geq 5\) ). The total 595 overlapping transcripts represent those that were concurrently down- regulated in both WT MII and Nat10- ZcKO MII oocytes. In other words, they were degraded regardless of Nat10 presence (Nat10- unrelated). (L) Degradation trend patterns of mouse maternal transcripts during the GV- MII transition in WT and Nat10- ZcKO oocytes. Each light- yellow line represents the expression levels of one gene, and the middle blue and red lines represent the median expression levels in WT and Nat10- ZcKO, respectively. Transcripts with TPM \(\geq 1\) at the GV stage were selected and analyzed.
+
+<--- Page Split --->
+![](images/Figure_8.jpg)
+
+<center>Fig.8. Hairpin Adaptor-PolyA Tail length (HA-PAT) assay validated the deficient maternal mRNA decay in Nat10-ZcKO MII oocytes. (A) Venn diagram showing the overlapping of transcripts that were stabilized during GV-to-MII transition in Cnot6l-/- and Nat10-ZcKO MII oocytes (FC=[WT MII/Nat10-ZcKO MII] \(\geq 2\) , \(\mathsf{p}< 0.05\) ). (B) Fold change of relative expression levels of transcripts encoding ribosomal protein subunits in Nat10-ZcKO relative to WT oocytes at MII stage. The values of log2(FC[Nat10-ZcKO/WT]) are listed on the right column. (C) qPCR results showing the relative levels of indicated transcripts (Cnot6l, Cnot7 and Btg4) in WT and Nat10-ZcKO oocytes at MII stage. Data are presented as the mean±SEM, \(n = 3\) . \*\*\*, \(\mathsf{p}< 0.0001\) by two-tailed Student's t-test. (D) A schematic illustration depicting the design strategy and the key steps for Hairpin Adaptor-PolyA Tail length (HA-PAT) assay. The 1st strand of cDNA was synthesized with the hairpin adaptor (HA) primer in conjunction with a P5TSO primer containing three "G", via a mechanism of "template-switching". GSP, Gene-specific primer; A0, the PCR product resulting from the amplification with a gene-specific pair of GSPxF and </center>
+
+<--- Page Split --->
+
+GSPxR primers; GSPxR primer was designed against an mRNA's 3' terminal preceding the polyA sequence; polyA- containing PCR products were amplified with GSPxF and a fixed HAPrimerR primers. The full sequence for hairpin adaptor (HA) is listed at the bottom. W indicates degenerate nucleotides (A or T); \* The asterisk indicates the phosphorothioate modification. (E- F) HA- PAT assay results showing changes in poly(A)- tail lengths of indicated transcripts in WT and Nat10- ZcKO oocytes at GV and MII stages. Experiments were performed in triplicates; a representative image is shown in the 2% agarose gel (E) and the length distribution shown in the densitometric curves (F).
+
+<--- Page Split --->
+
+0.7863. (F) CCK8 assay showing the cell proliferation rates among WT, OHT treatment and OHT plus Nat10 overexpression (OE) groups. (G-H) Comparison of cell proliferation as visualized by Ki67 labelling (G) and quantification (H) among WT (empty vector), OHT treatment and Nat10 overexpression groups. Scale bar, 50μm. Data are presented as mean±SEM, n=3. \*, p<0.05, \*\*, p<0.01 by two-tailed Student's t-test. (I-J) The polysome profiling displaying the translational efficiency and ribosome assembly in MEF cells analyzed by sucrose density gradient centrifugation. The graphic curves showed the polysome profiles of MEF cells treated by mock (Blue) or OHT (Red) (I). Comparison of the ratios of 60S to 40S (Left) and of 80S to 40S (Right) in MEF cells with mock treatment (Blue) or OHT (Red) (J). Data are presented as mean±SEM, n=3. \*\*\*, p<0.001 by two-tailed Student's t-test.
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+SupplementaryTableS4.DEGsbetweenWTandNat10ZcKOGVoocytes.csvSupplementaryTableS5.DEGsbetweenWTandNat10ZcKOMlloocytes.xlsxSupplementarymaterials.SupplementaryFigureS18andSupplementaryTablesS13.pdf
+
+<--- Page Split --->

preprint/preprint__1014f2957f8a5a80f5a91706c1899e6dcb17ad8f638e3ae989055850203d6b05/images_list.json

@@ -0,0 +1,77 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Figure 1 | Schematic diagram of a multilayer strategy and topology optimization for the design of cellular lattice metamaterial. a, illustration for the multilayer strategy. The nested multilayer cells can be constructed by scaling a surface, transforming a surface, or hybridizing other geometrical models. b, illustration for topology optimization. Through topology optimization, the multilayer model automatically converges into a beam-plate-shell-combined complex with reasonable material distribution. c, illustration for the technological route of the proposed design method.",
+    "footnote": [],
+    "bbox": [
+      [
+        110,
+        50,
+        890,
+        430
+      ]
+    ],
+    "page_idx": 3
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2 | Young's moduli and yield strengths comparisons for various shell-based cells. The metal iron is assigned as the constitute material, and Young's modulus, yield strength, and density are given as 210GPa, 400MPa, and \\(7800\\mathrm{Kg} / \\mathrm{m}^3\\) respectively. The geometrical cubic length for each cell is \\(15\\times 15\\times 15\\mathrm{mm}\\) in x-, y-, and z- directions. The Young's modulus and yield strength for the cellular material are calculated by extracting the slope and endpoint of the linear part of the stress-strain curves, respectively. Here, the normalized Young's modulus and yield strength are not the traditional concepts for solid constitute material, but effective concepts for describing mechanical performances for cellular material. The structural thickness was regularized to ensure that the relative density was limited to no more than \\(20\\%\\) otherwise, the thickness might become too large as the density increases thereby violating the presumption of shell. Since topology optimization digs holes on a surface, the mass losses were compensated in thickness for the optimized results by tuning the thickness and area fraction of the surface, leading to the same relative density as the original topologies. Subfigure (a) shows the compressive Young's moduli of Schwarz P, IWP, and Neovius and their optimized results in the material property space. Here 'Opt-' refers to optimized results under uniaxial loading. Subfigure (b) shows the compressive Young's moduli of the P set and its multilayer variants, and their optimized results in the material property space. Here, the script \"n\" in \"P.\" and \"Opt-P.\" indicates the number of layers. Subfigures (c) and (d) give the normalized Young's modulus and yield strength vs. the relative density of three different minimal surfaces (i.e., Schwarz P, IWP, Neovius) and their optimized results. Subfigures (e) and (f) give the normalized Young's modulus and yield strength vs. the relative density of the P set and their optimized results under uniaxial loading.",
+    "footnote": [],
+    "bbox": [
+      [
+        66,
+        58,
+        896,
+        558
+      ]
+    ],
+    "page_idx": 4
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Figure 3 | Physical experiment verification for constitutive relationships of unit cells of the P set and their optimized results with four different printing materials. a, comparison of fabricated models of the P set and Opt-P set with four different types of printing material, that is, thermoplastic polyurethane (TPU), nylon (PA12), stainless steel (SS316), and aluminum alloy (AlSi10Mg). b, comparison of numerical simulation results and practical physical experimental results of constitutive relationships of the P set and Opt-P set with four different types of printing material.",
+    "footnote": [],
+    "bbox": [
+      [
+        58,
+        40,
+        940,
+        520
+      ]
+    ],
+    "page_idx": 5
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Figure 4 | Physical experiment verification for constitutive relationships of 4X4X4 arrays of the P set and their optimized results with two different printing materials. a, illustration of different stress distributions of cells in different regions within a lattice. A cell along the boundary part of the lattice is akin to a cell with free boundary condition (FBC), while a cell inside the central part of the lattice is akin to a cell with periodic boundary condition (PBC). b, comparison of fabricated models of FBC-Opt-P set and PBC-Opt-P set with two different types of printing material, that is, nylon (PA12) and aluminum alloy (AlSi10Mg). Subfigures (c) and (d) give comparisons of numerical simulation results and practical physical experimental results of constitutive relationships of the FBC-Opt-P set and PBC-Opt-P set, respectively.",
+    "footnote": [],
+    "bbox": [
+      [
+        60,
+        45,
+        930,
+        472
+      ]
+    ],
+    "page_idx": 6
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Figure 5 | Elastoplastic property of Schwarz P set and their optimized results. a, energy comparison. b, elastoplastic deformation and collapse mode. c, force-displacement comparison during compression. d, normalized energy absorption rate comparison.",
+    "footnote": [],
+    "bbox": [
+      [
+        58,
+        42,
+        940,
+        533
+      ]
+    ],
+    "page_idx": 7
+  }
+]

preprint/preprint__1014f2957f8a5a80f5a91706c1899e6dcb17ad8f638e3ae989055850203d6b05/preprint__1014f2957f8a5a80f5a91706c1899e6dcb17ad8f638e3ae989055850203d6b05.mmd

@@ -0,0 +1,241 @@
+
+# Achieving extreme stiffness for beam-plate-shell- combined lattice metamaterial through a multilayer strategy and topology optimization  
+
+Jianbin Du dujb@tsinghua.edu.cn  
+
+Tsinghua University  Yang Liu  Tsinghua University  Yongzhen Wang  Tsinghua University  Hongyuan Ren  Tsinghua University  https://orcid.org/0009- 0001- 7078- 4694  Zhiqiang Meng  Tsinghua University  Xueqian Chen  Tsinghua University  Zuyu Li  University of Technology Sydney  Liwei Wang  Northwestern University  https://orcid.org/0000- 0002- 2777- 9718  Wei Chen  Northwestern University  Yifan Wang  Nanyang Technological University  https://orcid.org/0000- 0003- 2284- 520X  
+
+# Article  
+
+# Keywords:  
+
+Posted Date: September 21st, 2023  DOI: https://doi.org/10.21203/rs.3.rs- 3325404/v1
+
+<--- Page Split --->
+
+License: © ① This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Additional Declarations: There is NO Competing Interest.  
+
+Version of Record: A version of this preprint was published at Nature Communications on April 6th, 2024. See the published version at https://doi.org/10.1038/s41467-024-47089-8.
+
+<--- Page Split --->
+
+# Achieving extreme stiffness for beam-plate-shell-combined lattice metamaterial through a multilayer strategy and topology optimization  
+
+Yang Liu a, d, Yongzhen Wang a, Hongyuan Ren a, Zhiqiang Meng a, Xueqian Chen a, Zuyu Li b,1, Liwei Wang c,2  
+
+Wei Chen c,3, Yifan Wang d,4, Jianbin Du a,5  
+
+a School of Aerospace Engineering, Tsinghua University, Beijing, PR China  b School of Mechanical and Mechatronic Engineering, University of Technology Sydney, Ultimo, Australia  c Department of Mechanical Engineering, Northwestern University, Evanston, U.S.A  d School of Mechanical & Aerospace Engineering, Nanyang Technological University, Singapore  
+
+## Abstract  
+
+Metamaterials composed of different geometrical primitives have different properties. Corresponding to the fundamental geometrical forms of line, plane, and surface, beam-, plate-, and shell- based cellular lattice metamaterials enjoy many advantages in many aspects, respectively. To fully exploit the advantages of each structural archetype, we propose a multilayer strategy and topology optimization technique to design cellular lattice metamaterial in this study. Under the frame of the multilayer strategy, the design space is enlarged, and the design freedom is increased. Topology optimization is applied to explore better designs in the larger design space. Beam- plate- shell- combined metamaterials automatically emerge from the optimization to achieve extreme stiffness. Benefiting from high stiffness, energy absorption performances of optimized results also demonstrate substantial improvements under large geometrical deformation. The multilayer strategy and topology optimization can also bring a series of tunable dimensions for cellular lattice design, which helps achieve desired mechanical properties, such as isotropic elasticity and functionally grading material property, and superior performances in acoustic tuning, electrostatic shielding, and fluid field tuning. We envision that a broad array of novel synthetic and composite metamaterials with unprecedented performance can be designed with the multilayer strategy and topology optimization.  
+
+## Introduction  
+
+Cellular lattice metamaterials show many advantages in the design of synthetic and composite metamaterials due to sophisticated topologies and length scales. Nowadays, cellular lattices have been applied to many aspects of practical engineering, such as mechanical, acoustic, electromagnetic, and optical fields, and so on 1- 4.  
+
+According to the dominant structural elements, cellular lattice metamaterials can mainly be divided into three categories, that is, beam-, plate-, and shell- based lattices, corresponding to line, plane, and surface in structural geometry, respectively. Different types of lattices have different advantages as well as disadvantages. The beam- based lattices have been widely studied in theoretical research, numerical simulation, and experiment 5- 9, and particular functional properties, such as isotropy 10- 12, auxeticity 13, and chirality 14 can be achieved thanks to the flexibility of the line geometrical form. However, beam- based lattices are prone to stress concentration at the point joints between the strut elements where flaws or imperfections are more likely to occur 15. The plate- based lattices show superiority in mechanical performance such as stiffness and strength. Lattices with combinations of cubic and octet plate geometries can reach the Hashin- Shtrikman upper bound for the isotropic elastic moduli, including Young's modulus, bulk modulus, and shear modulus 16- 19. The experimental results showed that the stiffness and compressive strength of the plate lattice are always higher than those of the beam- based lattices with the same mass 17. Elastic isotropy can also be realized in plate- based lattices by combining certain lattice topologies, such as simple cubic, body- centered cubic, and face- centered cubic 20. Besides, plate- based lattices as metamaterials also demonstrated excellent sound and mechanical energy absorption performance 21. Nevertheless, imperfections can also be induced around connections and corners of plates, where risks of stress concentration and bulking potentially lie. Moreover, the tunable dimension for the plate element forming plate- based lattices is generally limited to the thickness due to the geometrical form of the plane. The lack of tunability restricts the applicability of lattice metamaterials for complex engineering scenarios. The shell- based lattices, whose cells are composed of continuous and smooth- curved surfaces, demonstrated outstanding strength, and stiffness at low density 22. Compared with plate- based lattices, shell- based lattices enjoy continuity and smoothness of the geometrical form of the surface. Theoretical and numerical analyses showed that the continuity and smoothness of the surface are very important in suppressing local buckling 23. The triply periodic minimal surfaces (TPMS) have attracted widespread attention in the materials and engineering fields due to their neoteric, symmetrical structures, and excellent mechanical properties 24- 26. TPMS naturally inherits the advantages of the surface. The mean curvature at each point of the TPMS is zero, which is a continuous and smooth surface and has no sharp edges 27- 29. The smooth transition between different crystal cells of the TPMS lattice will reduce the occurrence of stress concentration, thereby improving the overall mechanical properties 30- 32. Studies showed that the elasticity of cellular materials with particular topologies of TPMS can approach the Hashin- Shtrikman upper limit 33, 34. The advantageous geometrical form of the smooth surface of TPMS also attributed to superior performance in large geometrical deformation, leading to better energy absorption capacity 35- 40. Despite a series of excellent properties of shell- based lattices, their topological configurations usually have fewer variations and tunable dimensions are limited to the shape 41 and thickness 42.  
+
+To enrich the diversity and explore the better performance of cellular lattice metamaterial, we propose a multilayer strategy to enlarge the design space and associated topology optimization to effectively explore the vast design space for target- optimized performance. Specifically, the multilayer strategy can be realized through scaling, transforming, or hybridizing a series of single- layer elements to form nested models (Figure 1a). While the multilayer strategy leads to higher design freedom, it also imposes challenges on the design process to identify promising structures in an extremely large design space. To address this, we further develop a topology optimization technique to optimize topological configuration and improve the structural performance of lattices. Topology
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Figure 1 | Schematic diagram of a multilayer strategy and topology optimization for the design of cellular lattice metamaterial. a, illustration for the multilayer strategy. The nested multilayer cells can be constructed by scaling a surface, transforming a surface, or hybridizing other geometrical models. b, illustration for topology optimization. Through topology optimization, the multilayer model automatically converges into a beam-plate-shell-combined complex with reasonable material distribution. c, illustration for the technological route of the proposed design method. </center>  
+
+optimization targets on redistributing material to achieve reasonable configuration with prescribed objectives and constraints \(^{43,44}\) . To date, topology optimization has been applied to metamaterial design in many aspects \(^{45 - 51}\) . As can be seen in Figure 1b, optimized solutions in this study automatically converge into a comprehensive combination of beam, plate, and shell, which can fully utilize material with different structural archetypes. For example, the implementation scheme of the multilayer construction and topology optimization is illustrated in Figure 1c. A unit cell of Schwarz P forming the lattice is first identified as the candidate. Through scaling, transforming, and hybridizing, a nested multilayer model can be developed from the candidate cell. With prescribed objectives and constraints, the candidate cell and its multilayer variants are optimized. Finally, new lattices can be assembled with given unit cells. Through a series of analytical analyses, numerical simulations, as well as physical experiments, our findings show that the mechanical performances of optimized lattices demonstrate considerable improvement. On the other hand, the introduction of the multilayer strategy and topology optimization also brings a series of tunable dimensions, including shape, thickness, multimaterial, layer number, multilayer configuration, and area fraction (the ratio of the area of solid region and the area of the whole surface). With excellent tunability, particular structural mechanical properties, such as isotropic elasticity, and functionally grading stiffness, can be achieved conveniently for cellular lattice material. We also envision that the design method in this study will bring new opportunities in the application of acoustic absorption design, electrode design, fluid channel design, and so on.  
+
+## Results  
+
+The multilayer strategy is inspired to expand design space as the design freedom for a single layer model is limited. With a tailored multilayer configuration, the multilayer model is possible to achieve better results compared with a single- layer model at the same density. The topology optimization is targeted to redistribute material more efficiently with certain objectives, specifically, maximizing the total stiffness in this study.  
+
+## 1 Single-layer design  
+
+We first studied the design of single- layer cellular lattices. Three types of single- layer TPMS models are discussed, that is, Schwarz P, IWP, and Neovius. We compared the mechanical performances of the original designs and optimized results at the same relative densities to ensure fairness. As can be seen from the comparisons of normalized Young's moduli and yield strengths in Figure 2c and Figure 2d respectively (original data are referred to in in SI Fig. 4 and SI Table 1 in supplementary information (SI)), the "Opt- " results basically demonstrate improvements in stiffness and strength. The optimized geometries generally see uniaxial mechanical improvements as the optimizer concentrates more material in resisting loads from one unique direction (see strain energy distribution and uniaxial deformation comparisons in SI Fig. 5). In particular, the Opt- Neovius model achieves remarkable stiffness and strength increases (by around \(50\%\) ) at the same relative density, showing capability in reaching the theoretical Voigt bound (Figure 2a). Despite improvements through material redistribution, topology optimization is restricted to performing on the single surface. The optimization profit is destined to be limited due to limited design freedom.
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Figure 2 | Young's moduli and yield strengths comparisons for various shell-based cells. The metal iron is assigned as the constitute material, and Young's modulus, yield strength, and density are given as 210GPa, 400MPa, and \(7800\mathrm{Kg} / \mathrm{m}^3\) respectively. The geometrical cubic length for each cell is \(15\times 15\times 15\mathrm{mm}\) in x-, y-, and z- directions. The Young's modulus and yield strength for the cellular material are calculated by extracting the slope and endpoint of the linear part of the stress-strain curves, respectively. Here, the normalized Young's modulus and yield strength are not the traditional concepts for solid constitute material, but effective concepts for describing mechanical performances for cellular material. The structural thickness was regularized to ensure that the relative density was limited to no more than \(20\%\) otherwise, the thickness might become too large as the density increases thereby violating the presumption of shell. Since topology optimization digs holes on a surface, the mass losses were compensated in thickness for the optimized results by tuning the thickness and area fraction of the surface, leading to the same relative density as the original topologies. Subfigure (a) shows the compressive Young's moduli of Schwarz P, IWP, and Neovius and their optimized results in the material property space. Here 'Opt-' refers to optimized results under uniaxial loading. Subfigure (b) shows the compressive Young's moduli of the P set and its multilayer variants, and their optimized results in the material property space. Here, the script "n" in "P." and "Opt-P." indicates the number of layers. Subfigures (c) and (d) give the normalized Young's modulus and yield strength vs. the relative density of three different minimal surfaces (i.e., Schwarz P, IWP, Neovius) and their optimized results. Subfigures (e) and (f) give the normalized Young's modulus and yield strength vs. the relative density of the P set and their optimized results under uniaxial loading. </center>  
+
+Compared with the single- layer design, the use of multiple layers greatly enlarges the design space and enables more diverse geometries. Together with the powerful topology optimization technique, better results can be efficiently found in a vast design space. We elaborated on the implementation and advantages of the multilayer strategy and topology optimization, using the Schwarz P candidate as an example. Note that this scheme is also applicable to any other plate- and surface- based models. We defined a P set, including the P- 1 (original Schwarz P), - 2, - 4, and - 5, as well as their optimized results under uniaxial loading, i.e., the Opt- P set, including the Opt- P- 1, - 2, - 4, and - 5 (see their performances in the material property space in Figure 2b). All the optimized results were obtained with the same area fraction and tuned into the same mass through thickness adjustment. Mechanical performances of normalized Young's modulus and yield strength are compared in Figure 2e and Figure 2f, respectively (original data in SI Fig. 6 and SI Table 2). As shown, the model P- 2 shows slight reductions in both stiffness and strength, while P- 4 and P- 5 demonstrate considerable enhancements. Similar to the original Schwarz P, P- 2 also undergoes bending- dominant mechanical behavior. Since bending mode is much more sensitive to the structural thickness, it is sensible that P- 2 shows worse performance with thinner structural thickness at the same relative density. While the thicknesses of P- 4 and - 5 are smaller than P- 2, the tailored multilayer configurations transform the loads resisting mode. Specifically, P- 4 and - 5 undergo stretching- dominant mechanical behavior. In
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Figure 3 | Physical experiment verification for constitutive relationships of unit cells of the P set and their optimized results with four different printing materials. a, comparison of fabricated models of the P set and Opt-P set with four different types of printing material, that is, thermoplastic polyurethane (TPU), nylon (PA12), stainless steel (SS316), and aluminum alloy (AlSi10Mg). b, comparison of numerical simulation results and practical physical experimental results of constitutive relationships of the P set and Opt-P set with four different types of printing material. </center>  
+
+particular, the vertical plate elements provide sufficient stiffness under unilateral loading. Since the thickness of P- 4 is larger than P- 5, the vertical plate elements inside P- 4 are stronger, as a result, P- 4 outperforms P- 5 in both stiffness and strength. After optimization, Opt- P- 1 and - 2 become beam- based though maintaining the original shape outline. Opt- P- 4 and - 5 turn out to be beam- plate- shell- combined entities after material redistribution, and their stiffnesses are further brought into a higher level (increased by around \(50\%\) ), particularly Opt- P- 4 showing the capability of reaching the Voigt upper bound for anisotropic cellular materials. Generally, the larger the design space, the better the optimized outcome of topology optimization. However, it is noted that Opt- P- 5 shows worse static mechanical performances compared with Opt- P- 4. This is because the structural thickness of Opt- P- 5 is thinner than that of Opt- P- 4 with a larger structural area but the same mass. With the same thickness and mass, Opt- P- 5 should achieve better or at least the same performance as Opt- P- 4 (SI Fig. 7). Opt- P- 4 and - 5 share similar multilayer construction and optimized beam- plate- shell- combined configurations. It is noted that both the multilayer strategy and topology optimization contribute a lot to achieving extreme mechanical performance. On one hand, the tailored multiple layers develop more load- resisting paths, which helps in finding possible better deformation modes for given loading conditions. On the other hand, topology optimization further rationalizes the material distribution on the multiple layers. In the beam- plate- shell- combined configuration, the vertical plate elements mainly remain as they play an important role in resisting vertical loads. The horizontal plate elements degenerate into beam systems and show axial stretching deformation mode. The beam components pull the shell elements to prevent the curved shell from bending excessively under the compressing loads. In this sense, all three structural primitives are organically connected as a whole and exert their respective advantages efficiently (SI Fig. 8). Therefore, the beam- plate- shell- combined configuration substantiates the contribution of the multilayer strategy and topology optimization to remarkable improvements in mechanical performance.  
+
+The above- discussed comparison was based on data from numerical simulations for unit cells. We also studied the scale effect for the discussed cellular lattices. As presented in SI Fig. 9, lattices with three scales are compared, that is, unit cell, 4X4X4 array, and infinite array. As the lattice array becomes denser, the mechanical performance shows improvement, especially for the cases of P- 1 and - 2 as well as their optimized results. The benefits come from the changeover in loads carrying manner, that is, from bending- to stretching- dominated. As the Gibson- Ashby scaling power- law curve fits show (SI Fig. 9d), the fitting power values for P- 1 and - 2 are larger than 2, indicating that P- 1 and - 2 undergo bending- dominated deformation modes. This was alleviated through optimization and the fitting power values decreased from 2.272 and 2.414 to 1.747 (Opt- P- 1) and 2.159 (Opt- P- 2), respectively. As the lattice array increases, the fitting power values for P- 1 and - 2 as well as their optimized results further decrease, indicating their deformation modes become stretching- dominated. As for P- 4 and - 5 as well as their optimized results, their mechanical performances show
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Figure 4 | Physical experiment verification for constitutive relationships of 4X4X4 arrays of the P set and their optimized results with two different printing materials. a, illustration of different stress distributions of cells in different regions within a lattice. A cell along the boundary part of the lattice is akin to a cell with free boundary condition (FBC), while a cell inside the central part of the lattice is akin to a cell with periodic boundary condition (PBC). b, comparison of fabricated models of FBC-Opt-P set and PBC-Opt-P set with two different types of printing material, that is, nylon (PA12) and aluminum alloy (AlSi10Mg). Subfigures (c) and (d) give comparisons of numerical simulation results and practical physical experimental results of constitutive relationships of the FBC-Opt-P set and PBC-Opt-P set, respectively. </center>  
+
+independence on the lattice scale, and their corresponding fitting power values are close to 1 for different scales. As such, the shifting of loads carrying manner mainly comes from the multilayer strategy as it creates more load- resisting paths.  
+
+The scale effect is the result of boundary conditions. For a unit cell under uniaxial loading, the boundaries without connection to other cells are considered free boundary conditions (FBC). If all cells in a lattice are connected, then the periodic boundary conditions (PBC) are applied. Different boundary conditions may result in different deformation modes and different optimized solutions for the same geometrical model (SI Fig. 10). The models P- 1 and - 2 are sensitive to the boundary condition. One can see differences in their deformation modes as well as optimized topological configurations with FBC and PBC (SI Fig. 10). As a result, their mechanical performances rely much on lattice scale. P- 4 and - 5 are not sensitive to the boundary condition. Therefore, their deformation modes as well as optimized topological configurations are consistent. Generally, the calculated results with PBC are better than those with FBC. It is reasonable as the PBC plays a role of constraint in resisting external loads.  
+
+The physical compressing experiments in situ (SI Fig. 11) were conducted to validate the numerical results by subjecting the printed models to uniaxial compression using the universal testing machine. We studied two scales, that is, unit cell, and 4X4X4 array. For the scale of the unit cell, the P set and their optimized results are printed with four different kinds of material (Figure 3a), i.e., thermoplastic polyurethane (TPU), nylon (PA12), stainless steel (s316), and aluminum alloy (AlSi10Mg). Since the yielding strengths for diverse materials can be different and easily affected by local printing quality, here we only consider the comparison of uniaxial stiffnesses (compressing Young's moduli), i.e., the slopes of all stress- strain curves (original experimental data in SI Table 3). In comparison with the numerical simulations (Figure 3b), the basic trend of the physical experimental results shows much consistency with corresponding numerical results, yet still some minor differences remain for different kinds of printing material. The discrepancies are mainly caused by model weight (SI Fig. 12) and printing quality. Also, the anisotropy of additive manufacturing can affect the mechanical performance of the printed models.  
+
+For the 4X4X4 array scale, only the experiment for the original Schwarz P and the optimized P set were conducted as P- 2, - 4, and - 5 encountered fabricating difficulties. Particularly, P- 4 and - 5 are close- cell models, and the printing powder was unable to be removed in postprocess. We considered two array manners, that is, lattices arrayed by unit cells optimized with FBC and PBC, respectively. This is because the stress statuses for cells in different parts of a finite- arrayed lattice are inconsistent (Figure 4a). The 4X4X4 array P set and their optimized results with two materials, i.e., nylon (PA12), and aluminum alloy (AlSi10Mg), are presented in Figure 4b (original data in SI Fig. 13). As can be seen from the comparisons in Figure 4c and Figure 4d (original data in SI Table 4 and SI Table 5), the experimental results show consistency with corresponding numerical results, especially for the PA12 results. The inner ratio relationships normalized by the result of the original Schwarz P (100%) are just slightly different from that of corresponding numerical solutions. Only the Opt- P- 2 result shows a relatively larger discrepancy as a result of lesser material mass
+
+<--- Page Split --->
+![](images/Figure_5.jpg)
+
+<center>Figure 5 | Elastoplastic property of Schwarz P set and their optimized results. a, energy comparison. b, elastoplastic deformation and collapse mode. c, force-displacement comparison during compression. d, normalized energy absorption rate comparison. </center>  
+
+due to the post- printing process of powder cleaning (SI Fig. 13c). The practical stiffness performances for Opt- P- 4 and - 5 increased by around 3 and 2 times, respectively. The stiffnesses of AiSi10Mg results demonstrate larger gaps with corresponding numerical results, especially the Opt- P- 4 and - 5 outcomes. This is mainly because of the low printing quality, which leads to coarse surface and large granularity of manufactured models (SI Fig. 14). Consequently, the stiffnesses of printed models are impaired. In brief, the experimental results basically verified the correctness and effectiveness of the numerical simulations, and the physical experimental discrepancies from numerical results can be further narrowed by improving fabricating technology.  
+
+## 3 Energy absorption  
+
+In some practical engineering situations, many engineering crash scenarios, for instance, require light yet high energy absorption ratio materials under large geometrical deformation. High stiffness can have a positive influence on energy absorption. The impact direction usually is prescribed or predictable in some shock cases like aircraft landings and car pileups during traffic accidents. In such cases, the design of heterogeneous material with a high energy absorption ratio in one specific direction is meaningful. The P set and their optimized results arrayed with 4 unite cells in \(x\) , \(y\) , and \(z\) - directions were tested to examine their mechanical performance in energy absorption. The tests were carried out using uniaxial dynamic compression at a constant speed of 2mm/ms. The constitutive elastoplastic relationship is given in SI Table 6. The prescribed yield strength is 400MPa. Figure 5a displays the energy results of the P set and their optimized results during the whole impact period, including internal energy, plastic dissipation, strain energy, and kinetic energy. The strain energy and kinetic energy only account for small portions of the internal energy, while the plastic dissipation consists of the major part of the whole internal energy, and the tendency is nearly the same with the internal energy, indicating that the plastic behavior of a lattice is the key significance for performance in energy absorption. Apart from the plastic dissipation, strain energy, and kinetic energy, the artificial energy also makes up a small portion of the internal energy, but less than 10% for all examined cases, implying the dynamic simulations were effective. Figure 5b displays the arrayed models and their elastoplastic deformations and collapse modes. As shown, there are still many local regions under static deformation for the original Schwarz P model, making the lattice less efficient in energy absorption. P- 4 and - 5 as well as their optimized results, for example, are almost yielded all over the whole lattice, ensuring that all materials are fully utilized. In terms of the force- displacement curves during the whole impact period (Figure 5c), the forces of P- 4 and - 5 are much larger compared with other models in the P set. After optimization, the forces of Opt- P- 1 and - 2 show decreases, but remain stable during the whole impact. The forces of Opt- P- 4 and - 5 increase further but with observable fluctuations, indicating that clear buckling of components occurred during the impact. As we
+
+<--- Page Split --->
+
+integrate the force- displacement curves, we can obtain the energy absorption magnitudes. The comparison of normalized energy absorption by P- 1 (100%) is presented in Figure 5d. As can be seen, P- 2, Opt- P- 1, and - 2 show deterioration in energy absorption performance as they suffer more from the bending- dominant mechanical mode in large geometrical deformation. While benefiting from the high stiffness, P- 4 and - 5 as well as their optimized results are prominent in resisting dynamic loads. Especially for the Opt- P- 4 lattice, the energy absorption rate increased by 136%. Thus, it is useful to design and optimize multilayer beam- plate- shell- combined lattice metamaterial considering large geometrical deformation.  
+
+## Discussion  
+
+The main contribution of this work lies in providing a novel method combining a multilayer strategy and topology optimization to design cellular lattice metamaterial. Due to expanded design space and freedom, optimized multilayer designs at the same relative density can achieve better mechanical performance. Numerical simulations and physical experiments demonstrated that the obtained beam- plate- shell- combined lattice can achieve extreme stiffness.  
+
+In addition to the larger design space and better results, the multilayer strategy and topology optimization also bring excellent tunability to the design of cellular lattice metamaterial. Available tunable dimensions include shape, thickness, layer number, multilayer configuration, and area fraction, making the design of cellular lattice flexible and diverse. Benefiting from the high tunability, particular mechanical properties can be conveniently realized, such as isotropic elasticity (Extended Data Figure 1), and functionally grading stiffness (Extended Data Figure 2). Also, apart from mechanical design, the design method in this study shows exciting prospects when extending to the multiphysics- based design of metamaterials, such as acoustics, electrostatics, and flow- thermal coupling problems (Extended Data Figure 3). The multilayer strategy shows advantages in acoustic tuning, electrostatic shielding, and fluid field tuning, which can be useful in sound absorption for designing acoustic devices \(^{52}\) , electrodes for solid- state battery \(^{53}\) , and air- based actuators, such as pneumatic flexible tentacles of soft robots \(^{54}\) .  
+
+Despite advantageous properties accompanied by the multilayer strategy and topology optimization, side effects can also occur. For example, topology optimization naturally creates holes on the surface of a lattice, destroying the continuity of the surface so that more stress concentrations may be induced as the loads resisting mode changes. Also, some designs regarding specific problems require airtightness or watertightness, such as the design of pneumatic actuators. In our future work, we explore to design multilayer shell- based lattice whose thickness is variable across the whole surface with minimum limit constraint. As a result, the optimized results can achieve better performance meanwhile maintaining continuous surface. In this way, the multilayer- based design techniques can be further enriched and more choices are available for various designs purposes in practical application.  
+
+## Methods  
+
+## Multilayer construction  
+
+The multilayer lattice metamaterial is built upon TPMS, which can be generated through mathematical expressions. For example, the explicit term for the Schwarz P is stated as follows,  
+
+\[t = \cos X + \cos Y + \cos Z \quad (1)\]  
+
+where \(t\) is the mean curvature, \(X = 2\pi x / a\) , \(Y = 2\pi y / a\) , \(Z = 2\pi z / a\) , and \(a\) is the unit cell parameter. The mean curvature of TPMS vanishes at every point on the surface, that is, the Schwarz P is obtained when \(t = 0\) . The multilayer strategy can be realized through scaling, transforming, and hybridizing the candidate surface. In this way, the design space is expanded and the design freedom is increased. For example, the model P- 2 is constructed through a transformation from Schwarz P by changing the value of \(t\) . By further hybridizing three planes along \(x\) , \(y\) , and \(z\) - directions into the P- 2, P- 5 can be attained.  
+
+## Topology optimization  
+
+Topology optimization is a powerful technique to redistribute material to achieve optimal design with certain constraints. Other than optimizing the structural configuration, topology optimization can naturally introduce tunable dimensions, such as area fraction, making the design of multilayer shell- based metamaterial more flexible and diverse. In this study, we aim to minimize the compliance of the whole lattice. The optimization mathematical model can be formulated as follows,  
+
+\[\left\{ \begin{array}{l l}{\min_{\rho_{1},\cdots ,\rho_{N_{e}}}c(\rho_{e}) = \sum_{e = 1}^{N_{e}}\rho_{e}^{p}\mathbf{u}_{e}^{T}\mathbf{K}_{e}\mathbf{u}_{e}}\\ {\mathbf{K}\mathbf{U} = \mathbf{F}}\\ {\sum_{e = 1}^{N_{e}}\rho_{e}\nu_{e}\leq A_{f}}\\ {t\leq T_{h}}\\ {0\leq \rho_{e}\leq 1,} \end{array} \right. \quad (2)\]  
+
+where \(C\) is the objective function of static compliance, \(N_{e}\) the total number of finite elements in the admissible design domain, \(\mathbf{u}_{e}\) the elemental displacement field, \(\mathbf{K}_{e}\) the elemental stiffness matrix, \(\mathbf{K}\) the global stiffness matrix, \(\mathbf{U}\) global displacement field, \(\mathbf{F}\) the external loads, \(A_{f}\) the allowable area fraction, \(t\) the thickness, and \(T_{h}\) the prescribed maximum thickness. The design variable \(\rho_{e}\) is updated with lower and upper bounds of 0 and 1 respectively. The optimization problem can be efficiently solved using a novel ordinary differential equation (ODE) driven level- set density method \(^{55}\) .  
+
+Detailed information on the material characterization, optimization algorithm, practical implementation, as well as method verification for this study, is referred to in the SI text.  
+
+## Data availability  
+
+All data needed to evaluate the conclusions in this study are present in the paper and Supplementary Information.  
+
+## Code availability
+
+<--- Page Split --->
+
+All necessary information to generate the code used to evaluate the conclusions in this study is present in the paper and Supplementary Information. The original homemade code will be uploaded to Github. The reader can download and run it with guidance from the readme document.  
+
+## References  
+
+1. Chen X, Zhao J, Deng J, Jing Y, Pu H, Luo JJJoms. Low-frequency enhancement of acoustic black holes via negative stiffness supporting. 241, 107921 (2023).  
+2. Wu L, et al. A brief review of dynamic mechanical metamaterials for mechanical energy manipulation. 44, 168-193 (2021).  
+3. Luo C, Han CZ, Zhang XY, Zhang XG, Ren X, Xie YMJT-WS. Design, manufacturing and applications of auxetic tubular structures: A review. 163, 107682 (2021).  
+4. Kadic M, Milton GW, van Hecke M, Wegener M. 3D metamaterials. Nature Reviews Physics 1, 198-210 (2019).  
+5. Wang Z, et al. Additively-manufactured 3D truss-lattice materials for enhanced mechanical performance and tunable anisotropy: Simulations & experiments. Thin-Walled Structures 183, (2023).  
+6. Deng J, Li X, Liu Z, Wang Z, Li S. Mechanical Properties of Three-Dimensional Printed Combination-Design Truss Lattice Materials: Static and Dynamic Loading. Journal of Aerospace Engineering 35, (2022).  
+7. Zheng X, et al. Ultralight, ultrastiff mechanical metamaterials. Science 344, 1373-1377 (2014).  
+8. Meza LR, Das S, Greer JRJS. Strong, lightweight, and recoverable three-dimensional ceramic nanolattices. 345, 1322-1326 (2014).  
+9. Hyun S, Karlsson AM, Torquato S, Evans AG. Simulated properties of Kagome and tetragonal truss core panels. International Journal of Solids and Structures 40, 6989-6998 (2003).  
+10. Messner MCJ, JoltM, Solids Po. Optimal lattice-structured materials. 96, 162-183 (2016).  
+11. Li Z, Luo Z, Zhang L-C, Wang C-H. Topological design of pentamode lattice metamaterials using a ground structure method. Materials & Design 202, 109523 (2021).  
+12. Li Z, Gao W, Yu Wang M, Wang CH, Luo Z. Three-dimensional metamaterials exhibiting extreme isotropy and negative Poisson's ratio. International Journal of Mechanical Sciences 259, 108617 (2023).  
+13. Yang L, Harrysson O, West H, Cormier DJJJoS, Structures. Mechanical properties of 3D re-entrant honeycomb auxetic structures realized via additive manufacturing. 69, 475-490 (2015).  
+14. Frenzel T, Kadic M, Wegener MJS. Three-dimensional mechanical metamaterials with a twist. 358, 1072-1074 (2017).  
+15. Hashin Z, Shtrikman SJJotM, Solids Po. A variational approach to the theory of the elastic behaviour of multiphase materials. 11, 127-140 (1963).  
+16. Berger JB, Wadley HN, McMeeking RM. Mechanical metamaterials at the theoretical limit of isotropic elastic stiffness. Nature 543, 533-537 (2017).  
+17. Tancogne-Dejean T, Diamantopoulou M, Gorji MB, Bonatti C, Mohr D. 3D Plate-Lattices: An Emerging Class of Low-Density Metamaterial Exhibiting Optimal Isotropic Stiffness. Adv Mater 30, e1803334 (2018).  
+18. Duan S, Wen W, Fang DJAM. Additively-manufactured anisotropic and isotropic 3D plate-lattice materials for enhanced mechanical performance: Simulations & experiments. 199, 397-412 (2020).  
+19. Wang Y, Sigmund OJEML. Quasiperiodic mechanical metamaterials with extreme isotropic stiffness. 34, 100596 (2020).  
+20. Tancogne-Dejean T, Diamantopoulou M, Gorji MB, Bonatti C, Mohr DJAM. 3D plate-lattices: an emerging class of low-density metamaterial exhibiting optimal isotropic stiffness. 30, 1803334 (2018).  
+21. Li X, Yu X, Chua JW, Lee HP, Ding J, Zhai W. Microlattice Metamaterials with Simultaneous Superior Acoustic and Mechanical Energy Absorption. Small 17, e2100336 (2021).  
+22. Han SC, Lee JW, Kang K. A New Type of Low Density Material: Shellular. Adv Mater 27, 5506-5511 (2015).  
+23. Lee MG, Lee JW, Han SC, Kang K. Mechanical analyses of "Shellular", an ultralow-density material. Acta Materialia 103, 595-607 (2016).  
+24. Thomas N, Sreedhar N, Al-Ketan O, Rowshan R, Al-Rub RKA, Arafat HJD. 3D printed triply periodic minimal surfaces as spacers for enhanced heat and mass transfer in membrane distillation. 443, 256-271 (2018).  
+25. Afshar M, Anaraki AP, Montazeri HJM, C E. Compressive characteristics of radially graded porosity scaffolds architected with minimal surfaces. 92, 254-267 (2018).  
+26. Yoo D-JJJJOPE, Manufacturing. Computer-aided porous scaffold design for tissue engineering using triply periodic minimal surfaces. 12, 61-71 (2011).  
+27. Yoo D-JJJJOPE, Manufacturing. Heterogeneous porous scaffold design using the continuous transformations of triply periodic minimal surface models. 14, 1743-1753 (2013).  
+28. Gandy PJ, Bardhan S, Mackay AL, Klinowski JJCPL. Nodal surface approximations to the P, G, D and I-WP triply periodic minimal surfaces. 336, 187-195 (2001).  
+29. Schoen AH. Infinite periodic minimal surfaces without self-intersections. (1970).  
+30. Al-Ketan O, Al-Rub RKA, Rowshan R. Mechanical Properties of a New Type of Architected Interpenetrating Phase Composite Materials. Advanced Materials Technologies 2, (2017).  
+31. Al-Ketan O, Rezgui R, Rowshan R, Du H, Fang NX, Abu Al-Rub RK. Microarchitected Stretching-Dominated Mechanical Metamaterials with Minimal Surface Topologies. Advanced Engineering Materials 20, (2018).  
+32. Al-Ketan O, Al-Rub RKA, Rowshan RJJOMR. The effect of architecture on the mechanical properties of cellular structures based on the IWP minimal surface. 33, 343-359 (2018).  
+33. Chen Z, Xie YM, Wu X, Wang Z, Li Q, Zhou S. On hybrid cellular materials based on triply periodic minimal surfaces with extreme mechanical properties. Materials & Design 183, (2019).  
+34. Wang Y, Zhang X, Li Z, Gao H, Li XJPOtNAoS. Achieving the theoretical limit of strength in shell-based carbon nanolattices. 119, e2119536119 (2022).  
+35. Abueidda DW, Al-Rub RKA, Dalag AS, Lee D-W, Khan KA, Jasiuk IJMoM. Effective conductivities and elastic moduli of novel foams with triply periodic minimal surfaces. 95, 102-115 (2016).  
+36. Abueidda DW, et al. Mechanical properties of 3D printed polymeric cellular materials with triply periodic minimal surface architectures. 122, 255-267 (2017).  
+37. Maskery I, et al. Insights into the mechanical properties of several triply periodic minimal surface lattice structures made by polymer additive manufacturing. 152, 62-71 (2018).  
+38. Zhang L, et al. Energy absorption characteristics of metallic triply periodic minimal surface sheet structures under compressive loading. 23, 505-515 (2018).  
+39. Abueidda DW, Elhebeary M, Shiang C-S, Pang S, Abu Al-Rub RK, Jasiuk IM. Mechanical properties of 3D printed polymeric Gyroid cellular structures: Experimental and finite element study. Materials & Design 165, (2019).  
+40. Bonatti C, Mohr D. Mechanical performance of additively-manufactured anisotropic and isotropic smooth shell-lattice materials: Simulations & experiments. Journal of the Mechanics and Physics of Solids 122, 1-26 (2019).  
+41. Ma G, Zhang L, Wang MYJM, Design. Elastically isotropic open-cell uniform thickness shell lattices with optimized elastic moduli via shape optimization. 215, 110426 (2022).  
+42. Ma G, et al. Elastically-isotropic open-cell minimal surface shell lattices with superior stiffness via variable thickness design. 47, 102293 (2021).  
+43. Wang C, Zhao Z, Zhou M, Sigmund O, Zhang XS. A comprehensive review of educational articles on structural and multidisciplinary optimization. Structural and Multidisciplinary Optimization 64, 2827-2880 (2021).  
+44. Wang Y, Li X, Long K, Wei P. Open-Source Codes of Topology Optimization: A Summary for Beginners to Start Their Research. 137, 1-34 (2023).  
+45. Wu J, Sigmund O, Groen JP. Topology optimization of multi-scale structures: a review. Structural and Multidisciplinary Optimization 63, 1455-1480 (2021).  
+46. Alacoque L, Watkins RT, Tamijani AY. Stress-based and robust topology optimization for thermoelastic multi-material periodic microstructures. Computer Methods in Applied Mechanics and Engineering 379, 113749 (2021).  
+47. Lu Y, Wang Y. Structural optimization of metamaterials based on periodic surface modeling. Computer Methods in Applied Mechanics and Engineering 395, 115057 (2022).  
+48. Zeng Q, Duan S, Zhao Z, Wang P, Lei H. Inverse Design of Energy-Absorbing Metamaterials by Topology Optimization. Advanced Science 10, 2204977 (2023).  
+49. Xu W, Zhang P, Yu M, Yang L, Wang W, Liu L. Topology Optimization Via Spatially-Varying TPMS. IEEE Transactions on Visualization and Computer Graphics, 1-18 (2023).  
+50. Wang L, Boddapati J, Liu K, Zhu P, Daraio C, Chen W. Mechanical cloak via data-driven aperiodic metamaterial design. Proceedings of the National Academy of Sciences 119, e2122185119 (2022).  
+51. Liu Y, Li Z, Wei P, Wang W. Parameterized level-set based topology optimization method considering symmetry and pattern repetition constraints. Computer Methods in Applied Mechanics and Engineering 340, 1079-1101 (2018).  
+52. Li X, Yu C, Hua JW, Lee HP, Ding J, Zhai WJS. Microlattice metamaterials with simultaneous superior acoustic and mechanical energy absorption. 17, 2100336 (2021).  
+53. Werner J, Rodriguez-Calero G, Abruña H, Wiesner UJE, Science E. Block copolymer derived 3-D interpenetrating multifunctional gyroidal nanohybrids for electrical energy storage. 11, 1261-1270 (2018).  
+54. Chen F, Miao Y, Zhang L, Chen S, Zhu XJIR, Letters A. Triply Periodic Channels Enable Soft Pneumatic Linear Actuator with Single Material and Scalability. 7, 2668-2675 (2022).  
+55. Liu Y, Yang C, Wei P, Zhou P, Du J. An ODE-driven level-set density method for topology optimization. Computer Methods in Applied Mechanics and Engineering 387, 114159 (2021).  
+
+## Acknowledgements  
+
+The authors gratefully acknowledge the financial support provided by the National Natural Science Foundation of China (Grant Nos. 11772170) and the project of Beijing OptFuture Technology Co., Ltd (No. 20212002316).  
+
+## Author contributions  
+
+## Competing interests  
+
+The authors declare no competing interests.  
+
+## Additional information  
+
+Supplementary information is available for this manuscript.  
+
+## Extended Data
+
+<--- Page Split --->
+![PLACEHOLDER_10_0]
+  
+
+Extended Data Figure 1 | Anisotropic property of the P set. For a cubic symmetry unit cell, Zener's ratio can be used to quantify its anisotropy. When a lattice is homogeneous, its Zener's ratio equals 1. Subfigures a, b, c, and d give Young's moduli, shear moduli, Poisson's ratio, and Zener's ratio for the P set, respectively. As shown, among the P set, P- 2 basically has the same Young's and shear moduli and Poisson's ratio with P- 1, as a result, they show fewer differences in their Zener's ratio. P- 4 and - 5 show increases in Young's modulus and decreases in shear modulus as well as Poisson's ratio. Thus, their anisotropy properties are alleviated compared with P- 1. P- 4's Zener's ratio is close to 1. P- 5's Zener's ratio is almost equal to exact 1. Subfigures e, f, g, and h compare the uniaxial deformation, Young's modulus surface, shear deformation, and shear modulus surface for the P set, respectively. As can be seen, both Young's and shear modulus surfaces for P- 5 are close to a sphere, indicating the isotropic elasticity is reached. Note here each layer inside P- 5 has the same thickness. By further tuning the thickness of each layer, the anisotropy can be further tuned.
+
+<--- Page Split --->
+![PLACEHOLDER_11_0]
+  
+
+Extended Data Figure 2 | Stiffness- tunable lattice. The introduction of the multilayer strategy and topology optimization brings a series of tunable dimensions to facilitate the design of multiscale heterogeneous materials. As illustrated in subfigure a, a stiffness space can be developed through a series of tunable dimensions. As such, one can design a lattice with desired performance by simply assembling unit cells with specific properties as building blocks. For example, a candidate single- layer model with a tunable area fraction can be designated to construct stiffness- tunable lattices. Subfigure b displays several Schwarz P- based unit cells with different area fractions. Those unit cells were obtained through topology optimization. Subfigure c gives their optimization processes. As the area fraction decreases, the compliance increases. For optimized results, the relationship between stiffness and area fraction shows clear linearity in both situations of the same thickness but different masses and the same mass but different thicknesses (subfigure d). Exploiting this property, one can simply assemble grading microcells with grading area fractions to form functionally graded lattices (subfigure e). This can be useful in designing high- porosity artificial bone (subfigure f). Since the stiffness of the bone can be tunable, it stands a better chance to be adaptive to the realm of human body circumstance just like real human bones do. Another important tuning dimension to achieve variable stiffness is the thickness of the shell or plate elements inside a lattice. However, the optimized result is thickness- sensitive. Our findings show that the optimized topological configuration becomes more complex when the thickness of a shell or plate decreases. For example, as presented in subfigure h and subfigure i, optimization with two loading scenarios was studied to examine the thickness effect, i.e., uniaxial, and hydrostatic loadings. One can observe that more intricate components were shaped as the thickness became thinner. This is understandable because more components indicate more load- transferring paths. As the thickness of the shell decreases, the stiffness diminishes. Thus, the optimized results generate and distribute more routes to resist external loads, which is a trade of path number for stiffness. Interestingly, but not surprisingly, we found that the optimized solutions resembled the wing configuration of many insects, such as the dragonfly's wing (subfigure g). Due to the ultrathin thickness of the wing, it is necessary for a dragonfly to grow a sophisticated vein structure to support the wing in flight. Here we illustrate how mechanically tunable shell- based lattice can be realized with alterable area fraction and thickness. Note that more tunable dimensions can be activated with the introduction of the multilayer strategy and topology optimization.
+
+<--- Page Split --->
+![PLACEHOLDER_12_0]
+  
+
+Extended Data Figure 3 | Extension of multilayer lattices to multiphysics problems. The multilayer strategy also brings exciting prospects when extending to other physics- based designs of metamaterials. Subfigure a compares the performances of a monolayer and a bilayer Schwarz P- based lattice in dealing with acoustics, electrostatics, and flow- thermal coupling problems. The boundary conditions for the discussed three types of problems are illustrated in subfigures b, c, and d, respectively. As shown in subfigure e and subfigure f, the monolayer and bilayer models present different sound pressure distributions under the same sound source. The inner layer works as a wall blocking the transmission of sound. As a result, the total acoustic pressure on the external layer is near zero. In this sense, the multilayer strategy can help create a series of cavities, leading to more freedom for acoustic tuning. For the electrostatic problem, the comparison of the monolayer (subfigure g) and bilayer model (subfigure h) shows different distributions in electric field norm as well as electric potential. Since the floating potential is applied to model the metallic electrode, one can see that the electrostatic shielding is realized and the electric field norm outside the inner layer is zero. For the flow- thermal coupling problem, the temperature fields for the monolayer (subfigure i) and bilayer model (subfigure j) are basically the same as the conduction of heat converges into a steady state. However fluid velocity fields for the two models show different distributions, leading to different pressure levels within the lattice. Obviously, the multilayer configuration can change the flow manner.
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+Supplementaryinformation.pdf
+
+<--- Page Split --->

preprint/preprint__1014f2957f8a5a80f5a91706c1899e6dcb17ad8f638e3ae989055850203d6b05/preprint__1014f2957f8a5a80f5a91706c1899e6dcb17ad8f638e3ae989055850203d6b05_det.mmd

@@ -0,0 +1,301 @@
+<|ref|>title<|/ref|><|det|>[[44, 108, 936, 210]]<|/det|>
+# Achieving extreme stiffness for beam-plate-shell- combined lattice metamaterial through a multilayer strategy and topology optimization  
+
+<|ref|>text<|/ref|><|det|>[[44, 229, 280, 275]]<|/det|>
+Jianbin Du dujb@tsinghua.edu.cn  
+
+<|ref|>text<|/ref|><|det|>[[44, 303, 590, 744]]<|/det|>
+Tsinghua University  Yang Liu  Tsinghua University  Yongzhen Wang  Tsinghua University  Hongyuan Ren  Tsinghua University  https://orcid.org/0009- 0001- 7078- 4694  Zhiqiang Meng  Tsinghua University  Xueqian Chen  Tsinghua University  Zuyu Li  University of Technology Sydney  Liwei Wang  Northwestern University  https://orcid.org/0000- 0002- 2777- 9718  Wei Chen  Northwestern University  Yifan Wang  Nanyang Technological University  https://orcid.org/0000- 0003- 2284- 520X  
+
+<|ref|>title<|/ref|><|det|>[[44, 783, 101, 800]]<|/det|>
+# Article  
+
+<|ref|>title<|/ref|><|det|>[[44, 820, 135, 838]]<|/det|>
+# Keywords:  
+
+<|ref|>text<|/ref|><|det|>[[44, 857, 473, 914]]<|/det|>
+Posted Date: September 21st, 2023  DOI: https://doi.org/10.21203/rs.3.rs- 3325404/v1
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[42, 45, 910, 87]]<|/det|>
+License: © ① This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+<|ref|>text<|/ref|><|det|>[[42, 106, 530, 125]]<|/det|>
+Additional Declarations: There is NO Competing Interest.  
+
+<|ref|>text<|/ref|><|det|>[[42, 161, 950, 204]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on April 6th, 2024. See the published version at https://doi.org/10.1038/s41467-024-47089-8.
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[68, 78, 930, 115]]<|/det|>
+# Achieving extreme stiffness for beam-plate-shell-combined lattice metamaterial through a multilayer strategy and topology optimization  
+
+<|ref|>text<|/ref|><|det|>[[196, 146, 848, 161]]<|/det|>
+Yang Liu a, d, Yongzhen Wang a, Hongyuan Ren a, Zhiqiang Meng a, Xueqian Chen a, Zuyu Li b,1, Liwei Wang c,2  
+
+<|ref|>text<|/ref|><|det|>[[400, 166, 647, 179]]<|/det|>
+Wei Chen c,3, Yifan Wang d,4, Jianbin Du a,5  
+
+<|ref|>text<|/ref|><|det|>[[230, 184, 777, 225]]<|/det|>
+a School of Aerospace Engineering, Tsinghua University, Beijing, PR China  b School of Mechanical and Mechatronic Engineering, University of Technology Sydney, Ultimo, Australia  c Department of Mechanical Engineering, Northwestern University, Evanston, U.S.A  d School of Mechanical & Aerospace Engineering, Nanyang Technological University, Singapore  
+
+<|ref|>sub_title<|/ref|><|det|>[[58, 241, 123, 253]]<|/det|>
+## Abstract  
+
+<|ref|>text<|/ref|><|det|>[[57, 261, 940, 412]]<|/det|>
+Metamaterials composed of different geometrical primitives have different properties. Corresponding to the fundamental geometrical forms of line, plane, and surface, beam-, plate-, and shell- based cellular lattice metamaterials enjoy many advantages in many aspects, respectively. To fully exploit the advantages of each structural archetype, we propose a multilayer strategy and topology optimization technique to design cellular lattice metamaterial in this study. Under the frame of the multilayer strategy, the design space is enlarged, and the design freedom is increased. Topology optimization is applied to explore better designs in the larger design space. Beam- plate- shell- combined metamaterials automatically emerge from the optimization to achieve extreme stiffness. Benefiting from high stiffness, energy absorption performances of optimized results also demonstrate substantial improvements under large geometrical deformation. The multilayer strategy and topology optimization can also bring a series of tunable dimensions for cellular lattice design, which helps achieve desired mechanical properties, such as isotropic elasticity and functionally grading material property, and superior performances in acoustic tuning, electrostatic shielding, and fluid field tuning. We envision that a broad array of novel synthetic and composite metamaterials with unprecedented performance can be designed with the multilayer strategy and topology optimization.  
+
+<|ref|>sub_title<|/ref|><|det|>[[58, 420, 148, 433]]<|/det|>
+## Introduction  
+
+<|ref|>text<|/ref|><|det|>[[58, 440, 940, 478]]<|/det|>
+Cellular lattice metamaterials show many advantages in the design of synthetic and composite metamaterials due to sophisticated topologies and length scales. Nowadays, cellular lattices have been applied to many aspects of practical engineering, such as mechanical, acoustic, electromagnetic, and optical fields, and so on 1- 4.  
+
+<|ref|>text<|/ref|><|det|>[[57, 477, 940, 810]]<|/det|>
+According to the dominant structural elements, cellular lattice metamaterials can mainly be divided into three categories, that is, beam-, plate-, and shell- based lattices, corresponding to line, plane, and surface in structural geometry, respectively. Different types of lattices have different advantages as well as disadvantages. The beam- based lattices have been widely studied in theoretical research, numerical simulation, and experiment 5- 9, and particular functional properties, such as isotropy 10- 12, auxeticity 13, and chirality 14 can be achieved thanks to the flexibility of the line geometrical form. However, beam- based lattices are prone to stress concentration at the point joints between the strut elements where flaws or imperfections are more likely to occur 15. The plate- based lattices show superiority in mechanical performance such as stiffness and strength. Lattices with combinations of cubic and octet plate geometries can reach the Hashin- Shtrikman upper bound for the isotropic elastic moduli, including Young's modulus, bulk modulus, and shear modulus 16- 19. The experimental results showed that the stiffness and compressive strength of the plate lattice are always higher than those of the beam- based lattices with the same mass 17. Elastic isotropy can also be realized in plate- based lattices by combining certain lattice topologies, such as simple cubic, body- centered cubic, and face- centered cubic 20. Besides, plate- based lattices as metamaterials also demonstrated excellent sound and mechanical energy absorption performance 21. Nevertheless, imperfections can also be induced around connections and corners of plates, where risks of stress concentration and bulking potentially lie. Moreover, the tunable dimension for the plate element forming plate- based lattices is generally limited to the thickness due to the geometrical form of the plane. The lack of tunability restricts the applicability of lattice metamaterials for complex engineering scenarios. The shell- based lattices, whose cells are composed of continuous and smooth- curved surfaces, demonstrated outstanding strength, and stiffness at low density 22. Compared with plate- based lattices, shell- based lattices enjoy continuity and smoothness of the geometrical form of the surface. Theoretical and numerical analyses showed that the continuity and smoothness of the surface are very important in suppressing local buckling 23. The triply periodic minimal surfaces (TPMS) have attracted widespread attention in the materials and engineering fields due to their neoteric, symmetrical structures, and excellent mechanical properties 24- 26. TPMS naturally inherits the advantages of the surface. The mean curvature at each point of the TPMS is zero, which is a continuous and smooth surface and has no sharp edges 27- 29. The smooth transition between different crystal cells of the TPMS lattice will reduce the occurrence of stress concentration, thereby improving the overall mechanical properties 30- 32. Studies showed that the elasticity of cellular materials with particular topologies of TPMS can approach the Hashin- Shtrikman upper limit 33, 34. The advantageous geometrical form of the smooth surface of TPMS also attributed to superior performance in large geometrical deformation, leading to better energy absorption capacity 35- 40. Despite a series of excellent properties of shell- based lattices, their topological configurations usually have fewer variations and tunable dimensions are limited to the shape 41 and thickness 42.  
+
+<|ref|>text<|/ref|><|det|>[[57, 809, 940, 884]]<|/det|>
+To enrich the diversity and explore the better performance of cellular lattice metamaterial, we propose a multilayer strategy to enlarge the design space and associated topology optimization to effectively explore the vast design space for target- optimized performance. Specifically, the multilayer strategy can be realized through scaling, transforming, or hybridizing a series of single- layer elements to form nested models (Figure 1a). While the multilayer strategy leads to higher design freedom, it also imposes challenges on the design process to identify promising structures in an extremely large design space. To address this, we further develop a topology optimization technique to optimize topological configuration and improve the structural performance of lattices. Topology
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[110, 50, 890, 430]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[68, 444, 928, 508]]<|/det|>
+<center>Figure 1 | Schematic diagram of a multilayer strategy and topology optimization for the design of cellular lattice metamaterial. a, illustration for the multilayer strategy. The nested multilayer cells can be constructed by scaling a surface, transforming a surface, or hybridizing other geometrical models. b, illustration for topology optimization. Through topology optimization, the multilayer model automatically converges into a beam-plate-shell-combined complex with reasonable material distribution. c, illustration for the technological route of the proposed design method. </center>  
+
+<|ref|>text<|/ref|><|det|>[[57, 526, 940, 702]]<|/det|>
+optimization targets on redistributing material to achieve reasonable configuration with prescribed objectives and constraints \(^{43,44}\) . To date, topology optimization has been applied to metamaterial design in many aspects \(^{45 - 51}\) . As can be seen in Figure 1b, optimized solutions in this study automatically converge into a comprehensive combination of beam, plate, and shell, which can fully utilize material with different structural archetypes. For example, the implementation scheme of the multilayer construction and topology optimization is illustrated in Figure 1c. A unit cell of Schwarz P forming the lattice is first identified as the candidate. Through scaling, transforming, and hybridizing, a nested multilayer model can be developed from the candidate cell. With prescribed objectives and constraints, the candidate cell and its multilayer variants are optimized. Finally, new lattices can be assembled with given unit cells. Through a series of analytical analyses, numerical simulations, as well as physical experiments, our findings show that the mechanical performances of optimized lattices demonstrate considerable improvement. On the other hand, the introduction of the multilayer strategy and topology optimization also brings a series of tunable dimensions, including shape, thickness, multimaterial, layer number, multilayer configuration, and area fraction (the ratio of the area of solid region and the area of the whole surface). With excellent tunability, particular structural mechanical properties, such as isotropic elasticity, and functionally grading stiffness, can be achieved conveniently for cellular lattice material. We also envision that the design method in this study will bring new opportunities in the application of acoustic absorption design, electrode design, fluid channel design, and so on.  
+
+<|ref|>sub_title<|/ref|><|det|>[[58, 710, 115, 722]]<|/det|>
+## Results  
+
+<|ref|>text<|/ref|><|det|>[[58, 730, 940, 783]]<|/det|>
+The multilayer strategy is inspired to expand design space as the design freedom for a single layer model is limited. With a tailored multilayer configuration, the multilayer model is possible to achieve better results compared with a single- layer model at the same density. The topology optimization is targeted to redistribute material more efficiently with certain objectives, specifically, maximizing the total stiffness in this study.  
+
+<|ref|>sub_title<|/ref|><|det|>[[59, 791, 211, 805]]<|/det|>
+## 1 Single-layer design  
+
+<|ref|>text<|/ref|><|det|>[[58, 812, 940, 936]]<|/det|>
+We first studied the design of single- layer cellular lattices. Three types of single- layer TPMS models are discussed, that is, Schwarz P, IWP, and Neovius. We compared the mechanical performances of the original designs and optimized results at the same relative densities to ensure fairness. As can be seen from the comparisons of normalized Young's moduli and yield strengths in Figure 2c and Figure 2d respectively (original data are referred to in in SI Fig. 4 and SI Table 1 in supplementary information (SI)), the "Opt- " results basically demonstrate improvements in stiffness and strength. The optimized geometries generally see uniaxial mechanical improvements as the optimizer concentrates more material in resisting loads from one unique direction (see strain energy distribution and uniaxial deformation comparisons in SI Fig. 5). In particular, the Opt- Neovius model achieves remarkable stiffness and strength increases (by around \(50\%\) ) at the same relative density, showing capability in reaching the theoretical Voigt bound (Figure 2a). Despite improvements through material redistribution, topology optimization is restricted to performing on the single surface. The optimization profit is destined to be limited due to limited design freedom.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[66, 58, 896, 558]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[67, 570, 930, 767]]<|/det|>
+<center>Figure 2 | Young's moduli and yield strengths comparisons for various shell-based cells. The metal iron is assigned as the constitute material, and Young's modulus, yield strength, and density are given as 210GPa, 400MPa, and \(7800\mathrm{Kg} / \mathrm{m}^3\) respectively. The geometrical cubic length for each cell is \(15\times 15\times 15\mathrm{mm}\) in x-, y-, and z- directions. The Young's modulus and yield strength for the cellular material are calculated by extracting the slope and endpoint of the linear part of the stress-strain curves, respectively. Here, the normalized Young's modulus and yield strength are not the traditional concepts for solid constitute material, but effective concepts for describing mechanical performances for cellular material. The structural thickness was regularized to ensure that the relative density was limited to no more than \(20\%\) otherwise, the thickness might become too large as the density increases thereby violating the presumption of shell. Since topology optimization digs holes on a surface, the mass losses were compensated in thickness for the optimized results by tuning the thickness and area fraction of the surface, leading to the same relative density as the original topologies. Subfigure (a) shows the compressive Young's moduli of Schwarz P, IWP, and Neovius and their optimized results in the material property space. Here 'Opt-' refers to optimized results under uniaxial loading. Subfigure (b) shows the compressive Young's moduli of the P set and its multilayer variants, and their optimized results in the material property space. Here, the script "n" in "P." and "Opt-P." indicates the number of layers. Subfigures (c) and (d) give the normalized Young's modulus and yield strength vs. the relative density of three different minimal surfaces (i.e., Schwarz P, IWP, Neovius) and their optimized results. Subfigures (e) and (f) give the normalized Young's modulus and yield strength vs. the relative density of the P set and their optimized results under uniaxial loading. </center>  
+
+<|ref|>text<|/ref|><|det|>[[61, 789, 940, 950]]<|/det|>
+Compared with the single- layer design, the use of multiple layers greatly enlarges the design space and enables more diverse geometries. Together with the powerful topology optimization technique, better results can be efficiently found in a vast design space. We elaborated on the implementation and advantages of the multilayer strategy and topology optimization, using the Schwarz P candidate as an example. Note that this scheme is also applicable to any other plate- and surface- based models. We defined a P set, including the P- 1 (original Schwarz P), - 2, - 4, and - 5, as well as their optimized results under uniaxial loading, i.e., the Opt- P set, including the Opt- P- 1, - 2, - 4, and - 5 (see their performances in the material property space in Figure 2b). All the optimized results were obtained with the same area fraction and tuned into the same mass through thickness adjustment. Mechanical performances of normalized Young's modulus and yield strength are compared in Figure 2e and Figure 2f, respectively (original data in SI Fig. 6 and SI Table 2). As shown, the model P- 2 shows slight reductions in both stiffness and strength, while P- 4 and P- 5 demonstrate considerable enhancements. Similar to the original Schwarz P, P- 2 also undergoes bending- dominant mechanical behavior. Since bending mode is much more sensitive to the structural thickness, it is sensible that P- 2 shows worse performance with thinner structural thickness at the same relative density. While the thicknesses of P- 4 and - 5 are smaller than P- 2, the tailored multilayer configurations transform the loads resisting mode. Specifically, P- 4 and - 5 undergo stretching- dominant mechanical behavior. In
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[58, 40, 940, 520]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[70, 531, 930, 595]]<|/det|>
+<center>Figure 3 | Physical experiment verification for constitutive relationships of unit cells of the P set and their optimized results with four different printing materials. a, comparison of fabricated models of the P set and Opt-P set with four different types of printing material, that is, thermoplastic polyurethane (TPU), nylon (PA12), stainless steel (SS316), and aluminum alloy (AlSi10Mg). b, comparison of numerical simulation results and practical physical experimental results of constitutive relationships of the P set and Opt-P set with four different types of printing material. </center>  
+
+<|ref|>text<|/ref|><|det|>[[57, 616, 941, 838]]<|/det|>
+particular, the vertical plate elements provide sufficient stiffness under unilateral loading. Since the thickness of P- 4 is larger than P- 5, the vertical plate elements inside P- 4 are stronger, as a result, P- 4 outperforms P- 5 in both stiffness and strength. After optimization, Opt- P- 1 and - 2 become beam- based though maintaining the original shape outline. Opt- P- 4 and - 5 turn out to be beam- plate- shell- combined entities after material redistribution, and their stiffnesses are further brought into a higher level (increased by around \(50\%\) ), particularly Opt- P- 4 showing the capability of reaching the Voigt upper bound for anisotropic cellular materials. Generally, the larger the design space, the better the optimized outcome of topology optimization. However, it is noted that Opt- P- 5 shows worse static mechanical performances compared with Opt- P- 4. This is because the structural thickness of Opt- P- 5 is thinner than that of Opt- P- 4 with a larger structural area but the same mass. With the same thickness and mass, Opt- P- 5 should achieve better or at least the same performance as Opt- P- 4 (SI Fig. 7). Opt- P- 4 and - 5 share similar multilayer construction and optimized beam- plate- shell- combined configurations. It is noted that both the multilayer strategy and topology optimization contribute a lot to achieving extreme mechanical performance. On one hand, the tailored multiple layers develop more load- resisting paths, which helps in finding possible better deformation modes for given loading conditions. On the other hand, topology optimization further rationalizes the material distribution on the multiple layers. In the beam- plate- shell- combined configuration, the vertical plate elements mainly remain as they play an important role in resisting vertical loads. The horizontal plate elements degenerate into beam systems and show axial stretching deformation mode. The beam components pull the shell elements to prevent the curved shell from bending excessively under the compressing loads. In this sense, all three structural primitives are organically connected as a whole and exert their respective advantages efficiently (SI Fig. 8). Therefore, the beam- plate- shell- combined configuration substantiates the contribution of the multilayer strategy and topology optimization to remarkable improvements in mechanical performance.  
+
+<|ref|>text<|/ref|><|det|>[[57, 838, 940, 950]]<|/det|>
+The above- discussed comparison was based on data from numerical simulations for unit cells. We also studied the scale effect for the discussed cellular lattices. As presented in SI Fig. 9, lattices with three scales are compared, that is, unit cell, 4X4X4 array, and infinite array. As the lattice array becomes denser, the mechanical performance shows improvement, especially for the cases of P- 1 and - 2 as well as their optimized results. The benefits come from the changeover in loads carrying manner, that is, from bending- to stretching- dominated. As the Gibson- Ashby scaling power- law curve fits show (SI Fig. 9d), the fitting power values for P- 1 and - 2 are larger than 2, indicating that P- 1 and - 2 undergo bending- dominated deformation modes. This was alleviated through optimization and the fitting power values decreased from 2.272 and 2.414 to 1.747 (Opt- P- 1) and 2.159 (Opt- P- 2), respectively. As the lattice array increases, the fitting power values for P- 1 and - 2 as well as their optimized results further decrease, indicating their deformation modes become stretching- dominated. As for P- 4 and - 5 as well as their optimized results, their mechanical performances show
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[60, 45, 930, 472]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[66, 483, 928, 570]]<|/det|>
+<center>Figure 4 | Physical experiment verification for constitutive relationships of 4X4X4 arrays of the P set and their optimized results with two different printing materials. a, illustration of different stress distributions of cells in different regions within a lattice. A cell along the boundary part of the lattice is akin to a cell with free boundary condition (FBC), while a cell inside the central part of the lattice is akin to a cell with periodic boundary condition (PBC). b, comparison of fabricated models of FBC-Opt-P set and PBC-Opt-P set with two different types of printing material, that is, nylon (PA12) and aluminum alloy (AlSi10Mg). Subfigures (c) and (d) give comparisons of numerical simulation results and practical physical experimental results of constitutive relationships of the FBC-Opt-P set and PBC-Opt-P set, respectively. </center>  
+
+<|ref|>text<|/ref|><|det|>[[60, 592, 936, 618]]<|/det|>
+independence on the lattice scale, and their corresponding fitting power values are close to 1 for different scales. As such, the shifting of loads carrying manner mainly comes from the multilayer strategy as it creates more load- resisting paths.  
+
+<|ref|>text<|/ref|><|det|>[[57, 617, 940, 717]]<|/det|>
+The scale effect is the result of boundary conditions. For a unit cell under uniaxial loading, the boundaries without connection to other cells are considered free boundary conditions (FBC). If all cells in a lattice are connected, then the periodic boundary conditions (PBC) are applied. Different boundary conditions may result in different deformation modes and different optimized solutions for the same geometrical model (SI Fig. 10). The models P- 1 and - 2 are sensitive to the boundary condition. One can see differences in their deformation modes as well as optimized topological configurations with FBC and PBC (SI Fig. 10). As a result, their mechanical performances rely much on lattice scale. P- 4 and - 5 are not sensitive to the boundary condition. Therefore, their deformation modes as well as optimized topological configurations are consistent. Generally, the calculated results with PBC are better than those with FBC. It is reasonable as the PBC plays a role of constraint in resisting external loads.  
+
+<|ref|>text<|/ref|><|det|>[[57, 716, 940, 840]]<|/det|>
+The physical compressing experiments in situ (SI Fig. 11) were conducted to validate the numerical results by subjecting the printed models to uniaxial compression using the universal testing machine. We studied two scales, that is, unit cell, and 4X4X4 array. For the scale of the unit cell, the P set and their optimized results are printed with four different kinds of material (Figure 3a), i.e., thermoplastic polyurethane (TPU), nylon (PA12), stainless steel (s316), and aluminum alloy (AlSi10Mg). Since the yielding strengths for diverse materials can be different and easily affected by local printing quality, here we only consider the comparison of uniaxial stiffnesses (compressing Young's moduli), i.e., the slopes of all stress- strain curves (original experimental data in SI Table 3). In comparison with the numerical simulations (Figure 3b), the basic trend of the physical experimental results shows much consistency with corresponding numerical results, yet still some minor differences remain for different kinds of printing material. The discrepancies are mainly caused by model weight (SI Fig. 12) and printing quality. Also, the anisotropy of additive manufacturing can affect the mechanical performance of the printed models.  
+
+<|ref|>text<|/ref|><|det|>[[57, 839, 940, 950]]<|/det|>
+For the 4X4X4 array scale, only the experiment for the original Schwarz P and the optimized P set were conducted as P- 2, - 4, and - 5 encountered fabricating difficulties. Particularly, P- 4 and - 5 are close- cell models, and the printing powder was unable to be removed in postprocess. We considered two array manners, that is, lattices arrayed by unit cells optimized with FBC and PBC, respectively. This is because the stress statuses for cells in different parts of a finite- arrayed lattice are inconsistent (Figure 4a). The 4X4X4 array P set and their optimized results with two materials, i.e., nylon (PA12), and aluminum alloy (AlSi10Mg), are presented in Figure 4b (original data in SI Fig. 13). As can be seen from the comparisons in Figure 4c and Figure 4d (original data in SI Table 4 and SI Table 5), the experimental results show consistency with corresponding numerical results, especially for the PA12 results. The inner ratio relationships normalized by the result of the original Schwarz P (100%) are just slightly different from that of corresponding numerical solutions. Only the Opt- P- 2 result shows a relatively larger discrepancy as a result of lesser material mass
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[58, 42, 940, 533]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[68, 547, 928, 586]]<|/det|>
+<center>Figure 5 | Elastoplastic property of Schwarz P set and their optimized results. a, energy comparison. b, elastoplastic deformation and collapse mode. c, force-displacement comparison during compression. d, normalized energy absorption rate comparison. </center>  
+
+<|ref|>text<|/ref|><|det|>[[57, 607, 940, 683]]<|/det|>
+due to the post- printing process of powder cleaning (SI Fig. 13c). The practical stiffness performances for Opt- P- 4 and - 5 increased by around 3 and 2 times, respectively. The stiffnesses of AiSi10Mg results demonstrate larger gaps with corresponding numerical results, especially the Opt- P- 4 and - 5 outcomes. This is mainly because of the low printing quality, which leads to coarse surface and large granularity of manufactured models (SI Fig. 14). Consequently, the stiffnesses of printed models are impaired. In brief, the experimental results basically verified the correctness and effectiveness of the numerical simulations, and the physical experimental discrepancies from numerical results can be further narrowed by improving fabricating technology.  
+
+<|ref|>sub_title<|/ref|><|det|>[[58, 692, 206, 705]]<|/det|>
+## 3 Energy absorption  
+
+<|ref|>text<|/ref|><|det|>[[57, 713, 940, 949]]<|/det|>
+In some practical engineering situations, many engineering crash scenarios, for instance, require light yet high energy absorption ratio materials under large geometrical deformation. High stiffness can have a positive influence on energy absorption. The impact direction usually is prescribed or predictable in some shock cases like aircraft landings and car pileups during traffic accidents. In such cases, the design of heterogeneous material with a high energy absorption ratio in one specific direction is meaningful. The P set and their optimized results arrayed with 4 unite cells in \(x\) , \(y\) , and \(z\) - directions were tested to examine their mechanical performance in energy absorption. The tests were carried out using uniaxial dynamic compression at a constant speed of 2mm/ms. The constitutive elastoplastic relationship is given in SI Table 6. The prescribed yield strength is 400MPa. Figure 5a displays the energy results of the P set and their optimized results during the whole impact period, including internal energy, plastic dissipation, strain energy, and kinetic energy. The strain energy and kinetic energy only account for small portions of the internal energy, while the plastic dissipation consists of the major part of the whole internal energy, and the tendency is nearly the same with the internal energy, indicating that the plastic behavior of a lattice is the key significance for performance in energy absorption. Apart from the plastic dissipation, strain energy, and kinetic energy, the artificial energy also makes up a small portion of the internal energy, but less than 10% for all examined cases, implying the dynamic simulations were effective. Figure 5b displays the arrayed models and their elastoplastic deformations and collapse modes. As shown, there are still many local regions under static deformation for the original Schwarz P model, making the lattice less efficient in energy absorption. P- 4 and - 5 as well as their optimized results, for example, are almost yielded all over the whole lattice, ensuring that all materials are fully utilized. In terms of the force- displacement curves during the whole impact period (Figure 5c), the forces of P- 4 and - 5 are much larger compared with other models in the P set. After optimization, the forces of Opt- P- 1 and - 2 show decreases, but remain stable during the whole impact. The forces of Opt- P- 4 and - 5 increase further but with observable fluctuations, indicating that clear buckling of components occurred during the impact. As we
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[58, 40, 939, 116]]<|/det|>
+integrate the force- displacement curves, we can obtain the energy absorption magnitudes. The comparison of normalized energy absorption by P- 1 (100%) is presented in Figure 5d. As can be seen, P- 2, Opt- P- 1, and - 2 show deterioration in energy absorption performance as they suffer more from the bending- dominant mechanical mode in large geometrical deformation. While benefiting from the high stiffness, P- 4 and - 5 as well as their optimized results are prominent in resisting dynamic loads. Especially for the Opt- P- 4 lattice, the energy absorption rate increased by 136%. Thus, it is useful to design and optimize multilayer beam- plate- shell- combined lattice metamaterial considering large geometrical deformation.  
+
+<|ref|>sub_title<|/ref|><|det|>[[58, 125, 141, 137]]<|/det|>
+## Discussion  
+
+<|ref|>text<|/ref|><|det|>[[58, 145, 939, 196]]<|/det|>
+The main contribution of this work lies in providing a novel method combining a multilayer strategy and topology optimization to design cellular lattice metamaterial. Due to expanded design space and freedom, optimized multilayer designs at the same relative density can achieve better mechanical performance. Numerical simulations and physical experiments demonstrated that the obtained beam- plate- shell- combined lattice can achieve extreme stiffness.  
+
+<|ref|>text<|/ref|><|det|>[[58, 196, 940, 305]]<|/det|>
+In addition to the larger design space and better results, the multilayer strategy and topology optimization also bring excellent tunability to the design of cellular lattice metamaterial. Available tunable dimensions include shape, thickness, layer number, multilayer configuration, and area fraction, making the design of cellular lattice flexible and diverse. Benefiting from the high tunability, particular mechanical properties can be conveniently realized, such as isotropic elasticity (Extended Data Figure 1), and functionally grading stiffness (Extended Data Figure 2). Also, apart from mechanical design, the design method in this study shows exciting prospects when extending to the multiphysics- based design of metamaterials, such as acoustics, electrostatics, and flow- thermal coupling problems (Extended Data Figure 3). The multilayer strategy shows advantages in acoustic tuning, electrostatic shielding, and fluid field tuning, which can be useful in sound absorption for designing acoustic devices \(^{52}\) , electrodes for solid- state battery \(^{53}\) , and air- based actuators, such as pneumatic flexible tentacles of soft robots \(^{54}\) .  
+
+<|ref|>text<|/ref|><|det|>[[58, 305, 940, 393]]<|/det|>
+Despite advantageous properties accompanied by the multilayer strategy and topology optimization, side effects can also occur. For example, topology optimization naturally creates holes on the surface of a lattice, destroying the continuity of the surface so that more stress concentrations may be induced as the loads resisting mode changes. Also, some designs regarding specific problems require airtightness or watertightness, such as the design of pneumatic actuators. In our future work, we explore to design multilayer shell- based lattice whose thickness is variable across the whole surface with minimum limit constraint. As a result, the optimized results can achieve better performance meanwhile maintaining continuous surface. In this way, the multilayer- based design techniques can be further enriched and more choices are available for various designs purposes in practical application.  
+
+<|ref|>sub_title<|/ref|><|det|>[[58, 401, 125, 414]]<|/det|>
+## Methods  
+
+<|ref|>sub_title<|/ref|><|det|>[[58, 422, 226, 435]]<|/det|>
+## Multilayer construction  
+
+<|ref|>text<|/ref|><|det|>[[60, 442, 936, 469]]<|/det|>
+The multilayer lattice metamaterial is built upon TPMS, which can be generated through mathematical expressions. For example, the explicit term for the Schwarz P is stated as follows,  
+
+<|ref|>equation<|/ref|><|det|>[[374, 476, 928, 491]]<|/det|>
+\[t = \cos X + \cos Y + \cos Z \quad (1)\]  
+
+<|ref|>text<|/ref|><|det|>[[58, 498, 939, 563]]<|/det|>
+where \(t\) is the mean curvature, \(X = 2\pi x / a\) , \(Y = 2\pi y / a\) , \(Z = 2\pi z / a\) , and \(a\) is the unit cell parameter. The mean curvature of TPMS vanishes at every point on the surface, that is, the Schwarz P is obtained when \(t = 0\) . The multilayer strategy can be realized through scaling, transforming, and hybridizing the candidate surface. In this way, the design space is expanded and the design freedom is increased. For example, the model P- 2 is constructed through a transformation from Schwarz P by changing the value of \(t\) . By further hybridizing three planes along \(x\) , \(y\) , and \(z\) - directions into the P- 2, P- 5 can be attained.  
+
+<|ref|>sub_title<|/ref|><|det|>[[58, 572, 220, 585]]<|/det|>
+## Topology optimization  
+
+<|ref|>text<|/ref|><|det|>[[58, 592, 939, 643]]<|/det|>
+Topology optimization is a powerful technique to redistribute material to achieve optimal design with certain constraints. Other than optimizing the structural configuration, topology optimization can naturally introduce tunable dimensions, such as area fraction, making the design of multilayer shell- based metamaterial more flexible and diverse. In this study, we aim to minimize the compliance of the whole lattice. The optimization mathematical model can be formulated as follows,  
+
+<|ref|>equation<|/ref|><|det|>[[338, 650, 928, 771]]<|/det|>
+\[\left\{ \begin{array}{l l}{\min_{\rho_{1},\cdots ,\rho_{N_{e}}}c(\rho_{e}) = \sum_{e = 1}^{N_{e}}\rho_{e}^{p}\mathbf{u}_{e}^{T}\mathbf{K}_{e}\mathbf{u}_{e}}\\ {\mathbf{K}\mathbf{U} = \mathbf{F}}\\ {\sum_{e = 1}^{N_{e}}\rho_{e}\nu_{e}\leq A_{f}}\\ {t\leq T_{h}}\\ {0\leq \rho_{e}\leq 1,} \end{array} \right. \quad (2)\]  
+
+<|ref|>text<|/ref|><|det|>[[58, 783, 940, 849]]<|/det|>
+where \(C\) is the objective function of static compliance, \(N_{e}\) the total number of finite elements in the admissible design domain, \(\mathbf{u}_{e}\) the elemental displacement field, \(\mathbf{K}_{e}\) the elemental stiffness matrix, \(\mathbf{K}\) the global stiffness matrix, \(\mathbf{U}\) global displacement field, \(\mathbf{F}\) the external loads, \(A_{f}\) the allowable area fraction, \(t\) the thickness, and \(T_{h}\) the prescribed maximum thickness. The design variable \(\rho_{e}\) is updated with lower and upper bounds of 0 and 1 respectively. The optimization problem can be efficiently solved using a novel ordinary differential equation (ODE) driven level- set density method \(^{55}\) .  
+
+<|ref|>text<|/ref|><|det|>[[60, 848, 937, 873]]<|/det|>
+Detailed information on the material characterization, optimization algorithm, practical implementation, as well as method verification for this study, is referred to in the SI text.  
+
+<|ref|>sub_title<|/ref|><|det|>[[58, 882, 175, 895]]<|/det|>
+## Data availability  
+
+<|ref|>text<|/ref|><|det|>[[58, 903, 819, 917]]<|/det|>
+All data needed to evaluate the conclusions in this study are present in the paper and Supplementary Information.  
+
+<|ref|>sub_title<|/ref|><|det|>[[58, 925, 180, 938]]<|/det|>
+## Code availability
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[57, 42, 940, 80]]<|/det|>
+All necessary information to generate the code used to evaluate the conclusions in this study is present in the paper and Supplementary Information. The original homemade code will be uploaded to Github. The reader can download and run it with guidance from the readme document.  
+
+<|ref|>sub_title<|/ref|><|det|>[[57, 88, 141, 100]]<|/det|>
+## References  
+
+<|ref|>text<|/ref|><|det|>[[56, 102, 940, 696]]<|/det|>
+1. Chen X, Zhao J, Deng J, Jing Y, Pu H, Luo JJJoms. Low-frequency enhancement of acoustic black holes via negative stiffness supporting. 241, 107921 (2023).  
+2. Wu L, et al. A brief review of dynamic mechanical metamaterials for mechanical energy manipulation. 44, 168-193 (2021).  
+3. Luo C, Han CZ, Zhang XY, Zhang XG, Ren X, Xie YMJT-WS. Design, manufacturing and applications of auxetic tubular structures: A review. 163, 107682 (2021).  
+4. Kadic M, Milton GW, van Hecke M, Wegener M. 3D metamaterials. Nature Reviews Physics 1, 198-210 (2019).  
+5. Wang Z, et al. Additively-manufactured 3D truss-lattice materials for enhanced mechanical performance and tunable anisotropy: Simulations & experiments. Thin-Walled Structures 183, (2023).  
+6. Deng J, Li X, Liu Z, Wang Z, Li S. Mechanical Properties of Three-Dimensional Printed Combination-Design Truss Lattice Materials: Static and Dynamic Loading. Journal of Aerospace Engineering 35, (2022).  
+7. Zheng X, et al. Ultralight, ultrastiff mechanical metamaterials. Science 344, 1373-1377 (2014).  
+8. Meza LR, Das S, Greer JRJS. Strong, lightweight, and recoverable three-dimensional ceramic nanolattices. 345, 1322-1326 (2014).  
+9. Hyun S, Karlsson AM, Torquato S, Evans AG. Simulated properties of Kagome and tetragonal truss core panels. International Journal of Solids and Structures 40, 6989-6998 (2003).  
+10. Messner MCJ, JoltM, Solids Po. Optimal lattice-structured materials. 96, 162-183 (2016).  
+11. Li Z, Luo Z, Zhang L-C, Wang C-H. Topological design of pentamode lattice metamaterials using a ground structure method. Materials & Design 202, 109523 (2021).  
+12. Li Z, Gao W, Yu Wang M, Wang CH, Luo Z. Three-dimensional metamaterials exhibiting extreme isotropy and negative Poisson's ratio. International Journal of Mechanical Sciences 259, 108617 (2023).  
+13. Yang L, Harrysson O, West H, Cormier DJJJoS, Structures. Mechanical properties of 3D re-entrant honeycomb auxetic structures realized via additive manufacturing. 69, 475-490 (2015).  
+14. Frenzel T, Kadic M, Wegener MJS. Three-dimensional mechanical metamaterials with a twist. 358, 1072-1074 (2017).  
+15. Hashin Z, Shtrikman SJJotM, Solids Po. A variational approach to the theory of the elastic behaviour of multiphase materials. 11, 127-140 (1963).  
+16. Berger JB, Wadley HN, McMeeking RM. Mechanical metamaterials at the theoretical limit of isotropic elastic stiffness. Nature 543, 533-537 (2017).  
+17. Tancogne-Dejean T, Diamantopoulou M, Gorji MB, Bonatti C, Mohr D. 3D Plate-Lattices: An Emerging Class of Low-Density Metamaterial Exhibiting Optimal Isotropic Stiffness. Adv Mater 30, e1803334 (2018).  
+18. Duan S, Wen W, Fang DJAM. Additively-manufactured anisotropic and isotropic 3D plate-lattice materials for enhanced mechanical performance: Simulations & experiments. 199, 397-412 (2020).  
+19. Wang Y, Sigmund OJEML. Quasiperiodic mechanical metamaterials with extreme isotropic stiffness. 34, 100596 (2020).  
+20. Tancogne-Dejean T, Diamantopoulou M, Gorji MB, Bonatti C, Mohr DJAM. 3D plate-lattices: an emerging class of low-density metamaterial exhibiting optimal isotropic stiffness. 30, 1803334 (2018).  
+21. Li X, Yu X, Chua JW, Lee HP, Ding J, Zhai W. Microlattice Metamaterials with Simultaneous Superior Acoustic and Mechanical Energy Absorption. Small 17, e2100336 (2021).  
+22. Han SC, Lee JW, Kang K. A New Type of Low Density Material: Shellular. Adv Mater 27, 5506-5511 (2015).  
+23. Lee MG, Lee JW, Han SC, Kang K. Mechanical analyses of "Shellular", an ultralow-density material. Acta Materialia 103, 595-607 (2016).  
+24. Thomas N, Sreedhar N, Al-Ketan O, Rowshan R, Al-Rub RKA, Arafat HJD. 3D printed triply periodic minimal surfaces as spacers for enhanced heat and mass transfer in membrane distillation. 443, 256-271 (2018).  
+25. Afshar M, Anaraki AP, Montazeri HJM, C E. Compressive characteristics of radially graded porosity scaffolds architected with minimal surfaces. 92, 254-267 (2018).  
+26. Yoo D-JJJJOPE, Manufacturing. Computer-aided porous scaffold design for tissue engineering using triply periodic minimal surfaces. 12, 61-71 (2011).  
+27. Yoo D-JJJJOPE, Manufacturing. Heterogeneous porous scaffold design using the continuous transformations of triply periodic minimal surface models. 14, 1743-1753 (2013).  
+28. Gandy PJ, Bardhan S, Mackay AL, Klinowski JJCPL. Nodal surface approximations to the P, G, D and I-WP triply periodic minimal surfaces. 336, 187-195 (2001).  
+29. Schoen AH. Infinite periodic minimal surfaces without self-intersections. (1970).  
+30. Al-Ketan O, Al-Rub RKA, Rowshan R. Mechanical Properties of a New Type of Architected Interpenetrating Phase Composite Materials. Advanced Materials Technologies 2, (2017).  
+31. Al-Ketan O, Rezgui R, Rowshan R, Du H, Fang NX, Abu Al-Rub RK. Microarchitected Stretching-Dominated Mechanical Metamaterials with Minimal Surface Topologies. Advanced Engineering Materials 20, (2018).  
+32. Al-Ketan O, Al-Rub RKA, Rowshan RJJOMR. The effect of architecture on the mechanical properties of cellular structures based on the IWP minimal surface. 33, 343-359 (2018).  
+33. Chen Z, Xie YM, Wu X, Wang Z, Li Q, Zhou S. On hybrid cellular materials based on triply periodic minimal surfaces with extreme mechanical properties. Materials & Design 183, (2019).  
+34. Wang Y, Zhang X, Li Z, Gao H, Li XJPOtNAoS. Achieving the theoretical limit of strength in shell-based carbon nanolattices. 119, e2119536119 (2022).  
+35. Abueidda DW, Al-Rub RKA, Dalag AS, Lee D-W, Khan KA, Jasiuk IJMoM. Effective conductivities and elastic moduli of novel foams with triply periodic minimal surfaces. 95, 102-115 (2016).  
+36. Abueidda DW, et al. Mechanical properties of 3D printed polymeric cellular materials with triply periodic minimal surface architectures. 122, 255-267 (2017).  
+37. Maskery I, et al. Insights into the mechanical properties of several triply periodic minimal surface lattice structures made by polymer additive manufacturing. 152, 62-71 (2018).  
+38. Zhang L, et al. Energy absorption characteristics of metallic triply periodic minimal surface sheet structures under compressive loading. 23, 505-515 (2018).  
+39. Abueidda DW, Elhebeary M, Shiang C-S, Pang S, Abu Al-Rub RK, Jasiuk IM. Mechanical properties of 3D printed polymeric Gyroid cellular structures: Experimental and finite element study. Materials & Design 165, (2019).  
+40. Bonatti C, Mohr D. Mechanical performance of additively-manufactured anisotropic and isotropic smooth shell-lattice materials: Simulations & experiments. Journal of the Mechanics and Physics of Solids 122, 1-26 (2019).  
+41. Ma G, Zhang L, Wang MYJM, Design. Elastically isotropic open-cell uniform thickness shell lattices with optimized elastic moduli via shape optimization. 215, 110426 (2022).  
+42. Ma G, et al. Elastically-isotropic open-cell minimal surface shell lattices with superior stiffness via variable thickness design. 47, 102293 (2021).  
+43. Wang C, Zhao Z, Zhou M, Sigmund O, Zhang XS. A comprehensive review of educational articles on structural and multidisciplinary optimization. Structural and Multidisciplinary Optimization 64, 2827-2880 (2021).  
+44. Wang Y, Li X, Long K, Wei P. Open-Source Codes of Topology Optimization: A Summary for Beginners to Start Their Research. 137, 1-34 (2023).  
+45. Wu J, Sigmund O, Groen JP. Topology optimization of multi-scale structures: a review. Structural and Multidisciplinary Optimization 63, 1455-1480 (2021).  
+46. Alacoque L, Watkins RT, Tamijani AY. Stress-based and robust topology optimization for thermoelastic multi-material periodic microstructures. Computer Methods in Applied Mechanics and Engineering 379, 113749 (2021).  
+47. Lu Y, Wang Y. Structural optimization of metamaterials based on periodic surface modeling. Computer Methods in Applied Mechanics and Engineering 395, 115057 (2022).  
+48. Zeng Q, Duan S, Zhao Z, Wang P, Lei H. Inverse Design of Energy-Absorbing Metamaterials by Topology Optimization. Advanced Science 10, 2204977 (2023).  
+49. Xu W, Zhang P, Yu M, Yang L, Wang W, Liu L. Topology Optimization Via Spatially-Varying TPMS. IEEE Transactions on Visualization and Computer Graphics, 1-18 (2023).  
+50. Wang L, Boddapati J, Liu K, Zhu P, Daraio C, Chen W. Mechanical cloak via data-driven aperiodic metamaterial design. Proceedings of the National Academy of Sciences 119, e2122185119 (2022).  
+51. Liu Y, Li Z, Wei P, Wang W. Parameterized level-set based topology optimization method considering symmetry and pattern repetition constraints. Computer Methods in Applied Mechanics and Engineering 340, 1079-1101 (2018).  
+52. Li X, Yu C, Hua JW, Lee HP, Ding J, Zhai WJS. Microlattice metamaterials with simultaneous superior acoustic and mechanical energy absorption. 17, 2100336 (2021).  
+53. Werner J, Rodriguez-Calero G, Abruña H, Wiesner UJE, Science E. Block copolymer derived 3-D interpenetrating multifunctional gyroidal nanohybrids for electrical energy storage. 11, 1261-1270 (2018).  
+54. Chen F, Miao Y, Zhang L, Chen S, Zhu XJIR, Letters A. Triply Periodic Channels Enable Soft Pneumatic Linear Actuator with Single Material and Scalability. 7, 2668-2675 (2022).  
+55. Liu Y, Yang C, Wei P, Zhou P, Du J. An ODE-driven level-set density method for topology optimization. Computer Methods in Applied Mechanics and Engineering 387, 114159 (2021).  
+
+<|ref|>sub_title<|/ref|><|det|>[[58, 700, 202, 713]]<|/det|>
+## Acknowledgements  
+
+<|ref|>text<|/ref|><|det|>[[58, 720, 937, 746]]<|/det|>
+The authors gratefully acknowledge the financial support provided by the National Natural Science Foundation of China (Grant Nos. 11772170) and the project of Beijing OptFuture Technology Co., Ltd (No. 20212002316).  
+
+<|ref|>sub_title<|/ref|><|det|>[[58, 753, 207, 767]]<|/det|>
+## Author contributions  
+
+<|ref|>sub_title<|/ref|><|det|>[[58, 775, 205, 788]]<|/det|>
+## Competing interests  
+
+<|ref|>text<|/ref|><|det|>[[58, 795, 353, 809]]<|/det|>
+The authors declare no competing interests.  
+
+<|ref|>sub_title<|/ref|><|det|>[[58, 817, 222, 830]]<|/det|>
+## Additional information  
+
+<|ref|>text<|/ref|><|det|>[[58, 838, 451, 852]]<|/det|>
+Supplementary information is available for this manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[58, 859, 165, 872]]<|/det|>
+## Extended Data
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[68, 48, 911, 440]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[56, 445, 941, 562]]<|/det|>
+Extended Data Figure 1 | Anisotropic property of the P set. For a cubic symmetry unit cell, Zener's ratio can be used to quantify its anisotropy. When a lattice is homogeneous, its Zener's ratio equals 1. Subfigures a, b, c, and d give Young's moduli, shear moduli, Poisson's ratio, and Zener's ratio for the P set, respectively. As shown, among the P set, P- 2 basically has the same Young's and shear moduli and Poisson's ratio with P- 1, as a result, they show fewer differences in their Zener's ratio. P- 4 and - 5 show increases in Young's modulus and decreases in shear modulus as well as Poisson's ratio. Thus, their anisotropy properties are alleviated compared with P- 1. P- 4's Zener's ratio is close to 1. P- 5's Zener's ratio is almost equal to exact 1. Subfigures e, f, g, and h compare the uniaxial deformation, Young's modulus surface, shear deformation, and shear modulus surface for the P set, respectively. As can be seen, both Young's and shear modulus surfaces for P- 5 are close to a sphere, indicating the isotropic elasticity is reached. Note here each layer inside P- 5 has the same thickness. By further tuning the thickness of each layer, the anisotropy can be further tuned.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[66, 48, 920, 410]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[56, 426, 941, 701]]<|/det|>
+Extended Data Figure 2 | Stiffness- tunable lattice. The introduction of the multilayer strategy and topology optimization brings a series of tunable dimensions to facilitate the design of multiscale heterogeneous materials. As illustrated in subfigure a, a stiffness space can be developed through a series of tunable dimensions. As such, one can design a lattice with desired performance by simply assembling unit cells with specific properties as building blocks. For example, a candidate single- layer model with a tunable area fraction can be designated to construct stiffness- tunable lattices. Subfigure b displays several Schwarz P- based unit cells with different area fractions. Those unit cells were obtained through topology optimization. Subfigure c gives their optimization processes. As the area fraction decreases, the compliance increases. For optimized results, the relationship between stiffness and area fraction shows clear linearity in both situations of the same thickness but different masses and the same mass but different thicknesses (subfigure d). Exploiting this property, one can simply assemble grading microcells with grading area fractions to form functionally graded lattices (subfigure e). This can be useful in designing high- porosity artificial bone (subfigure f). Since the stiffness of the bone can be tunable, it stands a better chance to be adaptive to the realm of human body circumstance just like real human bones do. Another important tuning dimension to achieve variable stiffness is the thickness of the shell or plate elements inside a lattice. However, the optimized result is thickness- sensitive. Our findings show that the optimized topological configuration becomes more complex when the thickness of a shell or plate decreases. For example, as presented in subfigure h and subfigure i, optimization with two loading scenarios was studied to examine the thickness effect, i.e., uniaxial, and hydrostatic loadings. One can observe that more intricate components were shaped as the thickness became thinner. This is understandable because more components indicate more load- transferring paths. As the thickness of the shell decreases, the stiffness diminishes. Thus, the optimized results generate and distribute more routes to resist external loads, which is a trade of path number for stiffness. Interestingly, but not surprisingly, we found that the optimized solutions resembled the wing configuration of many insects, such as the dragonfly's wing (subfigure g). Due to the ultrathin thickness of the wing, it is necessary for a dragonfly to grow a sophisticated vein structure to support the wing in flight. Here we illustrate how mechanically tunable shell- based lattice can be realized with alterable area fraction and thickness. Note that more tunable dimensions can be activated with the introduction of the multilayer strategy and topology optimization.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[70, 50, 930, 576]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[56, 591, 941, 752]]<|/det|>
+Extended Data Figure 3 | Extension of multilayer lattices to multiphysics problems. The multilayer strategy also brings exciting prospects when extending to other physics- based designs of metamaterials. Subfigure a compares the performances of a monolayer and a bilayer Schwarz P- based lattice in dealing with acoustics, electrostatics, and flow- thermal coupling problems. The boundary conditions for the discussed three types of problems are illustrated in subfigures b, c, and d, respectively. As shown in subfigure e and subfigure f, the monolayer and bilayer models present different sound pressure distributions under the same sound source. The inner layer works as a wall blocking the transmission of sound. As a result, the total acoustic pressure on the external layer is near zero. In this sense, the multilayer strategy can help create a series of cavities, leading to more freedom for acoustic tuning. For the electrostatic problem, the comparison of the monolayer (subfigure g) and bilayer model (subfigure h) shows different distributions in electric field norm as well as electric potential. Since the floating potential is applied to model the metallic electrode, one can see that the electrostatic shielding is realized and the electric field norm outside the inner layer is zero. For the flow- thermal coupling problem, the temperature fields for the monolayer (subfigure i) and bilayer model (subfigure j) are basically the same as the conduction of heat converges into a steady state. However fluid velocity fields for the two models show different distributions, leading to different pressure levels within the lattice. Obviously, the multilayer configuration can change the flow manner.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[44, 42, 311, 70]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[44, 93, 765, 113]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[60, 130, 354, 150]]<|/det|>
+Supplementaryinformation.pdf
+
+<--- Page Split --->

preprint/preprint__1024b305173c41b8401ff5ab5cc9c8082f65a9cfffd5cbcad6232e5357b19148/images_list.json

@@ -0,0 +1,107 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_0.jpg",
+    "caption": "Figures",
+    "footnote": [],
+    "bbox": [
+      [
+        140,
+        120,
+        875,
+        620
+      ]
+    ],
+    "page_idx": 17
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2. Definitions of neuron classes and clustering methods. (A) In a single-class scenario,",
+    "footnote": [],
+    "bbox": [
+      [
+        125,
+        99,
+        880,
+        880
+      ]
+    ],
+    "page_idx": 18
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Figure 3. Primary motor cortex L6 (IT vs. CT). (A) Representation of the two clusters produced",
+    "footnote": [],
+    "bbox": [
+      [
+        112,
+        80,
+        884,
+        875
+      ]
+    ],
+    "page_idx": 20
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Figure 4. Classification of projection neuron types in the presubiculum. (A) Representation of 5 axonal clusters produced by Levene’s test and unsupervised hierarchical clustering of neurons from the presubiculum (n = 93). (B) Colormap of the axonal distributions of neurons (columns) across anatomical regions (rows), with darker shades representing more axonal projections. Parcel names highlighted in pink are hypothalamus related. Parcel names highlighted in yellow and light blue are thalamus related. (C) Neuron-to-target assignments for the identified axonal projection classes and corresponding anatomical regions (dotted line: contralateral). (D) Anterior view of the",
+    "footnote": [],
+    "bbox": [
+      [
+        110,
+        88,
+        888,
+        768
+      ]
+    ],
+    "page_idx": 22
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Figure 5. Spatial distributions of somata in the presubiculum across projection classes.",
+    "footnote": [],
+    "bbox": [
+      [
+        120,
+        100,
+        866,
+        880
+      ]
+    ],
+    "page_idx": 24
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_6.jpg",
+    "caption": "Figure 6. Divergent path distance comparison from one neuron class in the presubiculum to",
+    "footnote": [],
+    "bbox": [
+      [
+        113,
+        87,
+        883,
+        877
+      ]
+    ],
+    "page_idx": 26
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_7.jpg",
+    "caption": "Figure 7. Convergent path distance comparison from each presubiculum cluster to major targets. (A) Top left: box and whisker plot depicting the median, first and third quartiles, and full range of the path distances from neurons in ipsilateral (I) and contralateral (C) classes to the medial entorhinal cortex, dorsal zone. Bottom: the distance of an archetype neuron from class B27 (red), from its soma in the ipsilateral presubiculum (green) to the dMEC (purple), is significantly longer than the comparable distance of an archetype neuron from class D19 (brown). (B) Top left: box and whisker plot depicting the path distances from neurons in ipsilateral and contralateral classes to the parasubiculum. Bottom: the distance of an archetype neuron from class B27 (red), from its soma in the presubiculum (green) to the ipsilateral ParaS (purple), is significantly shorter than the",
+    "footnote": [],
+    "bbox": [
+      [
+        111,
+        90,
+        872,
+        731
+      ]
+    ],
+    "page_idx": 28
+  }
+]

preprint/preprint__1024b305173c41b8401ff5ab5cc9c8082f65a9cfffd5cbcad6232e5357b19148/preprint__1024b305173c41b8401ff5ab5cc9c8082f65a9cfffd5cbcad6232e5357b19148.mmd

@@ -0,0 +1,291 @@
+
+# Unsupervised classification of brain-wide axons reveals neuronal projection blueprint  
+
+Diek Wheeler  
+
+duhee1e5@gmu.edu  
+
+George Mason University https://orcid.org/0000- 0001- 8635- 0033  
+
+Shaina Banduri  
+
+George Mason University  
+
+Sruthi Sankararaman  
+
+George Mason University  
+
+Samhita Vinay  
+
+George Mason University  
+
+Giorgio Ascoli  
+
+George Mason University https://orcid.org/0000- 0002- 0964- 676X  
+
+Biological Sciences - Article  
+
+Keywords:  
+
+Posted Date: July 3rd, 2023  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 3044664/v1  
+
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Additional Declarations:  
+
+There is NO Competing Interest.  
+
+Supplemental Tables are not available with this version.  
+
+Version of Record: A version of this preprint was published at Nature Communications on February 20th, 2024. See the published version at https://doi.org/10.1038/s41467- 024- 45741- x.
+
+<--- Page Split --->
+
+# Unsupervised classification of brain-wide axons reveals neuronal projection blueprint  
+
+Wheeler, Diek W. \(^{1}\) , Banduri, Shaina \(^{1}\) , Sankararaman, Sruthi \(^{1}\) , Vinay, Samhita \(^{1}\) , Ascoli, Giorgio A. \(^{1*}\)
+
+<--- Page Split --->
+
+7 Long- range axonal projections are quintessential determinants of network connectivity, linking cellular organization and circuit architecture. Here we introduce a quantitative strategy to identify, from a given source region, all “projection neuron types” with statistically different patterns of anatomical targeting. We first validate the proposed technique with well- characterized data from layer 6 of the mouse primary motor cortex. The results yield two clusters, consistent with previously discovered cortico- thalamic and intra- telencephalic neuron classes. We next analyze neurons from the presubiculum, a less- explored region. Extending sparse knowledge from earlier retrograde tracing studies, we identify five classes of presubicular projecting neurons, revealing unique patterns of divergence, convergence, and specificity. We thus report several findings: (1) individual classes target multiple subregions along defined functions, such as spatial representation vs. sensory integration and visual vs. auditory input; (2) all hypothalamic regions are exclusively targeted by the same class also invading midbrain, a sharp subset of thalamic nuclei, and agranular retrosplenial cortex; (3) Cornu Ammonis, in contrast, receives input from the same presubicular axons projecting to granular retrosplenial cortex, also the purview of a single class; (4) path distances from the presubiculum to the same targets differ significantly between classes, as do the path distances to distinct targets within most classes, suggesting fine temporal coordination in activating distant areas; (5) the identified classes have highly non- uniform abundances, with substantially more neurons projecting to midbrain and hypothalamus than to medial and lateral entorhinal cortex; (6) lastly, presubicular soma locations are segregated among classes, indicating topographic organization of projections. This study thus demonstrates that classifying neurons based on statistically distinct axonal projection patterns sheds light on the functional organizational of their circuit.
+
+<--- Page Split --->
+
+The classification of neurons in the mammalian nervous system has long been a focus of intensive investigation. While local features from slice preparations in vitro may suffice to infer the circuit roles of GABAergic interneurons \(^{1 - 3}\) , long- range projecting axons are crucial architectural elements of neural organization \(^{4,5}\) constituting the conceptual and physical nexus between brain- wide circuits and synaptic communication \(^{6}\) . Thus, projection axons have long been digitally traced from serial sections after in vivo labeling and light microscopy imaging \(^{7 - 10}\) . At the same time, their macroscopic extent ( \(\sim 1\) cm span; \(\sim 1\) m cable length) and microscopic caliber ( \(\sim 100\) nm branch thickness) combine into a formidable technological challenge for large- scale collection \(^{11,12}\) . As a result, the number of completely reconstructed projection axons in any mammalian neural system has until recently remained into the low tens.  
+
+A source brain region projecting to N targets (where N typically ranges between 10 and 50 in the mouse cortex) could contain any combination of \(2^{N}\) - 1 distinct axonal projection types. Such a combinatorics challenge requires a large- scale data collection for proper classification. Projects based on fluorescent Micro- Optical Sectioning Tomography (fMOST) technology \(^{13 - 15}\) or the Janelia MouseLight platform \(^{16}\) , launched in recent years to address this need, produced nearly 10,000 mouse whole- brain single neuron reconstructions registered to a 3D Common Coordinate Framework (CCF) \(^{17}\) with consensus anatomical labeling \(^{18}\) . However, these newly available data do not themselves generate novel scientific insights, explain brain circuitry, or even disprove that axons might simply invade a random subset of the regional target areas \(^{19}\) . Rigorous methods are needed to test the hypothesis that specific projection types exist, to characterize their identities, and to quantify their population sizes \(^{20}\) .  
+
+This study introduces an original technique to objectively identify projection- based neuronal classes. To ascertain whether a collection of axonal projections might result from essentially random variation within the constraints of regional connectivity or likely reflects distinct neuron types, we begin from the foundational criterion for classification: if a set of items belongs to segregated classes, their pairwise inter- individual differences must be on average larger between than within classes. In other words, two items from the same class should tend to be more similar to each other than two items from separate classes. To implement this logic into a classification framework, we couple rigorous statistical testing with unsupervised hierarchical clustering. A unique strength of this approach is its entirely data- driven granularity: the continuous accumulation of new tracings will progressively refine the classification details with increasing statistical power. We can then characterize the identified projection classes by quantifying their population size, topographic soma distributions, and convergence and divergence patterns.  
+
+In the remainder of this article, we first propose a formal definition of and a quantitative solution for the classification problem. We validate our approach by applying it to layer 6 of the primary motor cortex, and then utilize it to study the presubiculum, a rather under- investigated region of the mouse brain. We next quantify the neuronal population sizes of the presubicular projection classes and characterize the spatial distribution of their somata. Finally, we analyze the patterns of divergence and convergence of presubicular projection classes. We conclude by discussing the biological interpretations of these results.
+
+<--- Page Split --->
+
+## Results.  
+
+Formal Definition and Quantitative Solution of the Classification Problem.  
+
+The axonal projections of each neuron in a source region can be represented as k- dimensional vectors, where k is the number of target regions invaded by the source region. Each of the k components of the vector quantifies the number of axonal points within the corresponding region (Figure 1). We explore the null hypothesis, \(\mathrm{H_0}\) , that all neurons from a source region belong to a single projection class (Figure 2A), as opposed to the alternative hypothesis, \(\mathrm{H_A}\) , that distinct projection classes exist from that source region (Figure 2B). If two hypothetical classes exist, the projections will be more similar between neurons within a class and more different across classes (Figure 2C). In such a two- class scenario, the combined within- and across- class distances would thus form a wider distribution than the distribution generated if all neurons belong to just a single class (Figure 2D). To formally test \(\mathrm{H_A}\) , we measure all pairwise differences between neurons (as arccosine vector distances, see Methods & Materials). We then generate the distribution of distances for \(\mathrm{H_0}\) by randomizing the projection patterns while preserving total axonal length both by neuron and target region. We achieve this single- class "continuum" by iterative stochastic swapping of axonal points between neurons across two target regions (see Figure 2E and Methods & Materials). We can then apply Levene's one- tail statistical test to ascertain whether the original distribution of pairwise distances has significantly larger variance than the randomized distribution. If the answer is positive, we must discard \(\mathrm{H_0}\) and accept \(\mathrm{H_A}\) . Starting from the top node in an unsupervised hierarchical clustering tree, we can thus repeat Levene's test on the neurons of each of the two subtrees, continuing the process until none of the variance differences are statistically significant (Figure 2F).  
+
+## Validation of the Approach.  
+
+To validate the above research design, we first analyzed 52 MouseLight layer 6 neurons from the primary motor cortex<sup>21</sup>. This anatomical area is known to contain two distinct projection classes with well- defined subdivisions: cortico- thalamic (CT) and cortico- cortical or intra- telencephalic (IT) neurons<sup>22</sup>. The variance of the distribution of pairwise axonal projection differences of these neurons was significantly larger than that of the randomized projections (p < 0.001; variance of real data = 373.4; variance of randomized data = 195.7), indicating the existence of distinct clusters. However, both subtrees after the first split of unsupervised hierarchical clustering returned a non- significant Levene's test (IT: \(\mathrm{p} = \mathrm{N/A}\) ; variance of real data = 219.9; variance of randomized data = 240.0; CT: \(\mathrm{p} > 0.05\) ; variance of real data = 295.1; variance of randomized data = 264.0), revealing exactly two clusters (Figure 3A). The first cluster, consisting of 21 neurons, projected almost exclusively to motor cortical targets; the second cluster of 31 neurons projected primarily to thalamic targets (Figures 3B- D and Table S1). These patterns were fully consistent with the axonal pathways of the IT and CT neuronal classes, respectively. This finding thus corroborates the validity of employing Levene's test of variance on pairwise difference distributions to identify statistically distinct classes in unsupervised hierarchical clustering.  
+
+Classification of Projection Neurons from the Mouse Presubiculum.
+
+<--- Page Split --->
+
+We then applied our analytic technique to a lesser- explored source region of the mouse brain: the presubiculum. Unsupervised clustering and the test of variance demonstrated that the 93 MouseLight neurons from the presubiculum form five distinct projection classes (Figures 4A- C). We designate each class by a letter (A- E) followed by the number of neurons in the class (Figure 4C). The first class, A38, primarily targets the lateral entorhinal cortex (LEC), accounting for 82% of axonal extent outside of the presubiculum. This class also invades the dorsoventral (granular) retrosplenial cortex as well as the hippocampal formation (dentate gyrus, CA3, CA2, CA1, and subiculum). The second class, B27, mainly targets the dorsal portion of the medial entorhinal cortex (dMEC), accounting for 92.5% of extra- presubicular axonal extent, as well as retrohippocampal zone and parasubiculum. Class C3 neurons mostly target the contralateral dMEC (42%) and LEC (40%), subiculum (14%), and parasubiculum (4%) through extensive callosal and commissural fibers. Class D19 has the most complex (and unreported) pattern of innervation: in addition to major projections to the subiculum (40.8%) and dentate gyrus (16.3%), it is the sole source of projections to the lateral (agranular) retrosplenial cortex, to the hypothalamus (including the lateral mammillary nucleus and 18 additional nuclei: Table S2), and to the superior and inferior colliculi in the midbrain. This neuronal class also projects to a subset of 8 thalamic nuclei, including the medial part of the anterior thalamic nucleus (ATN) and the lateral geniculate nucleus. Lastly, class E6 projects to a complementary set of 14 other thalamic nuclei, including the ventral, dorsal, anterior, and lateral parts of the ATN and the medial geniculate nucleus. Neurons from all five projection classes also have substantial collaterals within the presubiculum. Examples of projection neurons from each of the presubicular projection classes are depicted in Figures 4D- E.  
+
+Presubicular Projection Classes Have Non- Uniform Neuronal Population Sizes.  
+
+Next, we quantified the proportion of neurons in the mouse presubiculum that belong to each projection class. To this aim, we extracted the anterograde tract tracing density distributions from the Allen Institute regional connectivity atlas and matched the fractions of neurons in every class based on their axonal patterns by numerical optimization (see Non- Negative Least Squares in Methods & Materials). The results converged with very small residual error (<0.0006%) indicating a near- exact correspondence between single- neuron and regional projections. Class D19, reaching the midbrain, hypothalamus, lateral (agranular) retrosplenial, and the lateral geniculate (visual thalamus) accounted for a plurality (38.1%) of neurons. Class A38, targeting the hippocampus, subiculum, dorsoventral (granular) retrosplenial cortex, and lateral entorhinal cortex ("what" pathway), accounted for the second largest share (30.6%) of neurons. Class B27, projecting to the parasubiculum and medial entorhinal cortex ("where" pathway) consisted of 16.3% of projection neurons. Class E6, focused on other thalamic nuclei including medial geniculate (auditory), was responsible for 13.7% of presubicular neurons. The diffuse contralateral projections of class C3 comprised the remaining 1.3%.  
+
+When accounting for these relative proportions together with the MouseLight axonal projections (Table S2), we can estimate the contribution of each class to the presubicular projections in each collection of target regions. In particular, the dentate gyrus receives 21% of its presubicular afferents from class A38 and 79% from class D19. The subiculum receives 69% of its presubicular afferents from class D19, 30% from class A38, and 1% from class C3. The lateral entorhinal cortex receives 99% of presubicular afferents from class A38 and 1% from class C3. The dorsal medial entorhinal cortex and parasubiculum receive 99% of presubicular afferents from class B27 and 1%
+
+<--- Page Split --->
+
+from class C3. All other regions are targeted by individual classes: CA3, CA1, and the dorsoventral (granular) retrosplenial cortex by A38; the midbrain, hypothalamus, lateral (agranular) retrosplenial cortex, and part of the thalamic nuclei including mATN and LGN by D19; and the rest of the thalamic nuclei including dvATN and MGN by E6.  
+
+Spatial Distribution of Somata Reveals Topographic Organization of Neuronal Projections.  
+
+Computational geometry analysis of soma locations within the presubiculum demonstrated a clear spatial separation among the four main projection classes: A38, B27, D19, and E6 (the smallest class, C3, is largely contralateral projecting). Specifically, the convex hull volume of each neuron class overlapped only minimally ( \(\sim 5 - 20\%\) ) with that of other neuron classes (Figure 5A- C). In particular, class A38 was positioned more rostrally and dorsally relative to the caudal- ventral position of class B27, with approximately \(14\%\) of overlap (Figure 5A). The overlap of A38 was maximal with D19 \((21\%)\) ; however, while most A38 neurons had a selective somatic concentration in layer 2 ( \(34 / 38\) : \(89.5\%\) ), D19 had a somatic distribution across all 3 presubicular layers: \(21\%\) in layer 1 and \(26\%\) in layer 3 (Figure 5B). Class E6 had the most lateral positioning resulting in almost complete segregation from the other projection classes: there were so few overlapping somata that a proper convex hull volume of the overlap could not be calculated (Figures 5C- D).  
+
+Efferent Path Distances from the Same Neurons Significantly Vary by Target Region.  
+
+We tested whether the path distances from presubicular neurons of a given projection class differed across their divergent target regions (Figure 6). In these analyses of divergence, ipsilateral and contralateral targets were considered separately, as the latter are systematically farther than the former. For class A38 neurons, projection distances to the ipsilateral lateral entorhinal cortex, subiculum, and dentate gyrus are shorter than those to the ipsilateral hippocampus; moreover, projection distances to the ipsilateral lateral entorhinal cortex are longer than those to the ipsilateral subiculum and dentate gyrus. Similarly, projection distances to the contralateral subiculum and lateral entorhinal cortex are shorter than those to the contralateral hippocampus. Thus, presubicular efferent path distances differ less between ipsilateral and contralateral hippocampus than between other targets across brain hemispheres (Figure 6A). For class B27, projections to the ipsilateral parasubiculum have shorter paths than those to medial entorhinal cortex, dorsal zone, but the distances are comparable in the contralateral case (Figure 6B). Finally, for class D19, projections both to the ipsilateral medial anterior thalamic nucleus and lateral geniculate nucleus, and to the ipsilateral hypothalamus and lateral mammillary nucleus combined have longer paths than those to the ipsilateral midbrain (Figure 6C).  
+
+Afferent Path Distances to the Same Target Region Significantly Vary by Projection Class.  
+
+Next, we asked whether the axons from neurons of distinct projection classes converging onto their shared targets had different path distances. With the sole exception of the dentate gyrus, all target regions displayed a significant dependence of path distance on the presubicular neuron class (Figure 7A- C). For the ipsilateral medial entorhinal cortex, dorsal zone, projections from E6 and D19 have shorter distances than those from B27 and A38, and projections from B27 have shorter distances than those from A38. For the contralateral medial entorhinal cortex, in contrast, projections from B27 have longer distances than those from A38 (Figure 7A). For the ipsilateral
+
+<--- Page Split --->
+
+211 parasubiculum, path distances from D19 are longer than those from B27 (Figure 7B). Finally, for 212 both the contralateral subiculum and parasubiculum, path distances from B27 are longer than those 213 from A38 (Figure 7B- C).
+
+<--- Page Split --->
+
+## Discussion.  
+
+This study introduced an original method to objectively identify projection- based neuronal classes by pairing the Levene's test with unsupervised hierarchical clustering. We first conducted a confirmatory study on layer 6 of the primary motor cortex to verify that the proposed technique could reproduce known projection types in a previously explored area of the mammalian brain. The results yielded two clusters with axonal projections consistent with those of the corticothalamic and intratelencephalic neuron classes found in past studies, thereby confirming the validity of the technique<sup>23</sup>.  
+
+To test whether the technique could lead to novel insights, we then applied it to the presubiculum, a region with crucial cognitive function<sup>24</sup>, yet few studies on its circuitry<sup>25</sup>. The results yielded five clusters, indicating distinct neuron classes, which led us to reject the null hypothesis that projection neurons exhibit random variation within the constraints of regional connectivity from the presubiculum. In an earlier study<sup>26</sup>, retrograde tracing identified five classes of neurons projecting from the presubiculum, which target the retrosplenial cortex, contralateral subiculum, medial entorhinal cortex, anterior thalamic nucleus, and lateral mammillary nucleus. Our results confirm the existence of these five classes and add new information that reveals patterns of divergence (e.g., class A38 projects to the retrosplenial cortex, dentate gyrus, subiculum, and entorhinal cortex), convergence (e.g., the subiculum receives projections from classes A38, contralateral C3, and D19), and specificity (e.g., class E6 projects exclusively to the medial geniculate nucleus, and all hypothalamic regions receive projections solely from class D19).  
+
+The proposed clustering technique correctly distinguishes cortical (classes A38, B27, and C3) from subcortical (D19 and E6) pathways in the second binary split in the hierarchical classification. These results also add cellular level details to previously reported presubicular projections to retrosplenial cortex and thalamic reticular nuclei<sup>27</sup>, as well as a broader circuit context to the characterization of individual presubicular neurons targeting the medial entorhinal cortex<sup>28</sup>.  
+
+Furthermore, our findings reveal that several target regions are spatially subdivided according to the differing inputs between classes. These regions include the entorhinal cortex (lateral projections mainly from class A38 and medial projections primarily from class B27), retrosplenial cortex (dorsoventral granular projections almost exclusively from class A38 and lateral agranular projections solely from class D19), and thalamus (medial anterior thalamic nucleus and lateral geniculate nucleus projections principally from class D19 and dorsoventral anterior thalamic nucleus and medial geniculate nucleus projections predominantly from class E6). Some of these regional subdivisions also have known functional distinctions: for instance, the medial entorhinal cortex specializes in spatial representation while the lateral entorhinal cortex specializes in integrating sensory input<sup>29</sup>. Among the thalamic geniculate nuclei, the medial geniculate nucleus is part of the auditory pathway, whereas the lateral geniculate nucleus is part of the visual pathway<sup>4</sup>.  
+
+From a comparison of divergent path distances from one presubicular class to its major targets, along with a comparison of convergent path distances from each presubicular class to collectively major targets, we found that path distances to the same targets were significantly different between classes, as were the path distances to distinct targets within most classes. This might imply that
+
+<--- Page Split --->
+
+260 electrical impulses reach different targets with varying delays, both within the same class and 261 between classes.  
+
+Topographic analysis of presubicular classes revealed spatial separation between the somata of each class. This suggests the possibility of anatomically mapping the input and output of the circuitry specializing in head direction computations<sup>30</sup>. Our reported topography of presubicular projections classes is consistent with the recently observed local modularity of the head- direction microcircuit<sup>31</sup>, and may help clarify the relationship between the egocentric and allocentric spatial and episodic representations of the cortico- hippocampal system<sup>32</sup>.  
+
+As with many secondary data analyses, we have limited knowledge of, and control over, artifactual shortcomings in the utilized datasets due to possible idiosyncrasies in labeling, imaging, tracing, registration, and mapping. However, the technique introduced with this work is applicable to many disparate sources of data besides MouseLight, including fMOST<sup>13–15</sup> and even MapSeq/BarSeq<sup>33,34</sup>. These data sources follow separate experimental and computational protocols, allowing independent validation for the source regions in which these datasets overlap. Our results so far, in the cases of the mouse primary motor cortex and presubiculum, indicate that the executed analysis is robust to these possible confounding variables<sup>22</sup>.
+
+<--- Page Split --->
+
+## Conclusion.  
+
+Overall, this study revealed that neurons can be divided into distinct classes based on axonal projection patterns, as demonstrated in layer 6 of the primary motor cortex and the presubiculum. Our applied analyses can be used to similarly analyze neurons projecting from all other mouse brain regions with sufficient data. There are currently approximately 40 regions fitting this criterion in the existing datasets, but this number is expected to grow in the near future. Furthermore, we suggest the application of pairing Levene's test and unsupervised hierarchical clustering to other complementary datasets, such as single- cell transcriptomic datasets, to classify neurons across a molecular domain, in addition to an anatomical domain, as demonstrated here. Moreover, all these complementary datasets are broadly expected to continue to grow in sample size, brain coverage, and acquisition pace<sup>35,36</sup>, supporting a call to establish cloud- based, community accessible pipelines for robust, rigorous, and systematic neuronal characterization<sup>37,38</sup>.
+
+<--- Page Split --->
+
+## Material & Methods.  
+
+Data Extraction, Storage, and Conversion.  
+
+The location of each axonal data point for nearly 1100 neurons was extracted from the Janelia MouseLight public dataset<sup>21</sup> using the freeware JSONLab v1.5 (https://www.mathworks.com/matlabcentral/mc- downloads/downloads/submissions/33381/versions/22/download/zip). These data were contained in a JSON file for each neuron, where X- Y- Z coordinates and parcel information were provided for each axonal point of the neuron. The axonal points in each brain parcel were tabulated for all neurons and were stored in a matrix (Tables S1- 2), in which each row represents a neuron, each column represents a parcel, and the values in each cell represent the axonal counts of a particular neuron in a particular region (Figure 1).  
+
+Hypothesis Design.  
+
+To determine whether distinct projection classes of neurons exist from a particular parcel of the brain, hypothesis \(\mathrm{H_A}\) , we tested the pairwise differences between neurons from the experimental matrices described above. If only a single class of neurons exists, then only a single distribution of differences between neurons will be generated (Figure 2A). If two hypothetical classes exist, then the differences between neurons, evaluated two at a time, will be smaller within a given class than across the two classes (Figure 2B- C). In a multi- class scenario, a histogram of the differences between neurons should be wider than the distribution generated when all the neurons belong to just a single class (Figure 2D). To generate the distribution of differences for the null hypothesis, \(\mathrm{H_0}\) , a randomized control matrix was generated from the original experimental matrix through multiple iterations of the stochastic pairwise swapping of axonal counts from two neurons across two target regions (Figure 2E). This method randomized the projection patterns, yielding a "continuum" consistent with the regional connectivity of Figure 2A, while preserving axonal sizes (row sums) and regional targeting (column sums) of the original experimental matrix.  
+
+Levene's Test.  
+
+We assessed the hypothesis that the variance of experimental data was significantly larger than the variance of randomized data \((\alpha = 0.05)\) . For both the experimental and randomized matrices, we computed the arccosine between a pair of neuronal vectors, each composed of the axonal counts across all target regions (https://github.com/Projectomics/MATLAB). These angles measure the projection difference of two neurons across all brain parcels. We then performed a 1- tailed Levene's test<sup>39</sup> on the angle distributions of the experimental and randomized matrices to assess whether their variances differed significantly. To this aim, we used the MATLAB function vartestn with the TestType parameter set to LeveneAbsolute. If the experimental data had a greater variance than the randomized data, then the experimental data could be further divided into classes, consistent with the scenario presented in Figure 2B.  
+
+Unsupervised Hierarchical Clustering.
+
+<--- Page Split --->
+
+We used unsupervised hierarchical clustering to determine a biologically accurate division of neuron classes based on axonal projection patterns. Specifically, the MATLAB linkage function, with the “average” algorithm for computing distance between clusters, was utilized on the 93 MouseLight neurons originating in the presubiculum and the 52 MouseLight neurons originating in layer 6 of the primary motor cortex. The initial assumption (null hypothesis) was that all neurons were part of a single class. If Levene’s test yielded significant results, the number of class divisions was incremented, and the technique was again repeated on each class division. This iterative process continued until none of the subdivided classes yielded significant results, thereby yielding the final class divisions (Figure 2F).  
+
+## Non-Negative Least Squares.  
+
+To estimate the fractional counts of cells in each of k projection classes in each region, we matched their respective single- cell axonal patterns against the regional connectivity from anterograde tracing to the m known targets, as presented in the Allen Mouse Brain Connectivity Atlas (http://connectivity.brain-map.org/projection). The problem is equivalent to a set of constrained, weighted, linear equations that can be solved numerically by non- negative least- square (NNLS) optimization<sup>40</sup>. NNLS finds the values x that minimizes the Euclidean norm of (Ax - b) with the constraint \(x \geq 0^{41}\) , where x is the k- dimensional vector representing the fractions of neurons in each class; b is the m- dimensional vector representing the weights of the regional projections to each target; and A is a k- by- m matrix with rows representing the projections of each class (the sum of the summary vectors of the corresponding neurons) and columns representing target regions. NNLS was computed using the lsqnonneg function in MATLAB.  
+
+Matrix A and vector b were based on data from the MouseLight dataset (Table S2) and the Allen Mouse Brain Connectivity Atlas, respectively. Setting the target region to the whole brain in the Connectivity Atlas and the source region to the presubiculum resulted in 7 tracing experiments, which included projection volumes and projection densities for each target brain region. Cross referencing the targeted regions of the MouseLight axonal projections with target regions that appeared in all 7 anterograde tracing experiments resulted in a listing of 66 regions. Matrix A was created with rows representing these 66 brain regions and columns representing the 5 neuron classes found by pairing Levene’s test with unsupervised hierarchical clustering of the presubiculum data (Table S3). The average projection volume and density values for each of the 66 regions were calculated from the 7 experiments, and the averages were multiplied to populate the columns of vector b.  
+
+To obtain the highest confidence in the NNLS analysis, matrix A was sequentially “bi- normalized” first by axonal length and then by invaded region. Specifically, first each cell in matrix A was normalized so that each row summed to one. Next, each value was divided by the number of regions, 66, and multiplied by the number of clusters, 5, such that the sum of all values in matrix A equaled 5. Subsequently, each cell in matrix A was normalized so that each column summed to one. Vector b was normalized such that the sum of all values equaled to one. Finally, the squared Euclidean norm of the residual of the MATLAB function lsqnonneg was calculated as a proxy for the uncertainty of the analysis.  
+
+Soma Analysis.
+
+<--- Page Split --->
+
+To quantify the spatial separation among the somata among the neuron projection classes in the presubiculum, we performed a convex hull analysis for the location of the soma centers in each class using MATLAB. To create the convex hull, outliers were removed by iteratively going through all points in each class and calculating the volume of the convex hull without each point. If the volume differed by more than \(1 / \mathrm{n}\) of the volume of the original convex hull, which included all points, the point was considered an outlier and removed from the dataset. This established an algorithmic thresholding that corresponded well with the visual inspection of potential outliers. However, if removing the outliers resulted in fewer than four somata, the minimal number of points required to conduct a convex hull analysis, all points were considered. Between each pair of convex hulls, the proportion of the volume of overlap to the volume of the union of the convex hulls was used to assess the similarity between topographic locations.  
+
+Analysis of Divergence and Convergence.  
+
+Utilizing the original JSON data files, for every neuron in each presubiculum class, we measured the path distance from the soma to each axonal point in the target region. We then calculated the median path distance to each target region across all neurons in the class, and performed a Wilcoxon Signed Rank Test<sup>42</sup>, using the MATLAB function ranksum, to assess whether the path distances to each characteristic target of a particular class were significantly different. Using the same data files, we also performed a Wilcoxon Signed Rank Test to assess whether the path distances to each characteristic target between all clusters were significantly different. In both sets of comparisons, multiple testing was corrected for by False Discovery Rate to determine the significance of the resultant p- values.
+
+<--- Page Split --->
+
+## References.  
+
+1. Fishell, G. & Kepecs, A. Interneuron Types as Attractors and Controllers. Annu Rev Neurosci 43, 1–30 (2020).  
+2. Jiang, X. et al. Principles of connectivity among morphologically defined cell types in adult neocortex. Science 350, aac9462 (2015).  
+3. DeFelipe, J. et al. New insights into the classification and nomenclature of cortical GABAergic interneurons. Nat Rev Neurosci 14, 202–216 (2013).  
+4. Sherman, S. M. & Guillery, R. W. Distinct functions for direct and transthalamic corticocortical connections. J Neurophysiol 106, 1068–1077 (2011).  
+5. Winnubst, J., Spruston, N. & Harris, J. A. Linking axon morphology to gene expression: a strategy for neuronal cell-type classification. Curr Opin Neurobiol 65, 70–76 (2020).  
+6. Ascoli, G. A. Trees of the brain, roots of the mind. (The MIT Press, 2015).  
+7. François, C., Tande, D., Yelnik, J. & Hirsch, E. C. Distribution and morphology of nigral axons projecting to the thalamus in primates. J Comp Neurol 447, 249–260 (2002).  
+8. Rochefort, N. L. et al. Functional selectivity of interhemispheric connections in cat visual cortex. Cereb Cortex 19, 2451–2465 (2009).  
+9. Rojas-Piloni, G. et al. Relationships between structure, in vivo function and long-range axonal target of cortical pyramidal tract neurons. Nat Commun 8, 870 (2017).  
+10. Ropireddy, D., Scorcioni, R., Lasher, B., Buzsáki, G. & Ascoli, G. A. Axonal morphometry of hippocampal pyramidal neurons semi-automatically reconstructed after in vivo labeling in different CA3 locations. Brain Struct Funct 216, 1–15 (2011).  
+11. DeFelipe, J. From the connectome to the synaptome: an epic love story. Science 330, 1198–1201 (2010).  
+12. Helmstaedter, M. & Mitra, P. P. Computational methods and challenges for large-scale circuit mapping. Curr Opin Neurobiol 22, 162–169 (2012).  
+13. Gao, L. et al. Single-neuron projectome of mouse prefrontal cortex. Nat Neurosci 25, 515–529 (2022).  
+14. Gong, H. et al. High-throughput dual-colour precision imaging for brain-wide connectome with cytoarchitectonic landmarks at the cellular level. Nat Commun 7, 12142 (2016).  
+15. Peng, H. et al. Morphological diversity of single neurons in molecularly defined cell types. Nature 598, 174–181 (2021).  
+16. Economo, M. N. et al. A platform for brain-wide imaging and reconstruction of individual neurons. Elife 5, e10566 (2016).  
+17. Wang, Q. et al. The Allen Mouse Brain Common Coordinate Framework: A 3D Reference Atlas. Cell 181, 936-953.e20 (2020).  
+18. Chon, U., Vanselow, D. J., Cheng, K. C. & Kim, Y. Enhanced and unified anatomical labeling for a common mouse brain atlas. Nat Commun 10, 5067 (2019).  
+19. Reimann, M. W. et al. A null model of the mouse whole-neocortex micro-connectome. Nat Commun 10, 3903 (2019).  
+20. Armañanzas, R. & Ascoli, G. A. Towards the automatic classification of neurons. Trends in Neurosciences 38, 307–318 (2015).  
+21. Winnubst, J. et al. Reconstruction of 1,000 Projection Neurons Reveals New Cell Types and Organization of Long-Range Connectivity in the Mouse Brain. Cell 179, 268-281.e13 (2019).
+
+<--- Page Split --->
+
+451 22. Muñoz-Castañeda, R. et al. Cellular anatomy of the mouse primary motor cortex. Nature 452 598, 159- 166 (2021). 453 23. Shepherd, G. M. G. Corticostriatal connectivity and its role in disease. Nat Rev Neurosci 14, 454 278- 291 (2013). 455 24. Angelaki, D. E. & Laurens, J. The head direction cell network: attractor dynamics, 456 integration within the navigation system, and three-dimensional properties. Current Opinion 457 in Neurobiology 60, 136- 144 (2020). 458 25. Jacobs, H. I. L. et al. The presubiculum links incipient amyloid and tau pathology to memory 459 function in older persons. Neurology 94, e1916- e1928 (2020). 460 26. Preston- Ferrer, P., Coletta, S., Frey, M. & Burgalossi, A. Anatomical organization of 461 presubicular head- direction circuits. Elife 5, e14592 (2016). 462 27. Vantomme, G. et al. A Thalamic Reticular Circuit for Head Direction Cell Tuning and 463 Spatial Navigation. Cell Reports 31, 107747 (2020). 464 28. Honda, Y. & Furuta, T. Multiple Patterns of Axonal Collateralization of Single Layer III 465 Neurons of the Rat Presubiculum. Front. Neural Circuits 13, 45 (2019). 466 29. Knierim, J. J., Neunuebel, J. P. & Deshmukh, S. S. Functional correlates of the lateral and 467 medial entorhinal cortex: objects, path integration and local- global reference frames. Philos 468 Trans R Soc Lond B Biol Sci 369, 20130369 (2014). 469 30. Taube, J. S. The head direction signal: origins and sensory- motor integration. Annu Rev 470 Neurosci 30, 181- 207 (2007). 471 31. Balsamo, G. et al. Modular microcircuit organization of the presubicular head- direction map. 472 Cell Rep 39, 110684 (2022). 473 32. Wang, C., Chen, X. & Knierim, J. J. Egocentric and allocentric representations of space in 474 the rodent brain. Curr Opin Neurobiol 60, 12- 20 (2020). 475 33. Kebschull, J. M. & Zador, A. M. Cellular barcoding: lineage tracing, screening and beyond. 476 Nat Methods 15, 871- 879 (2018). 477 34. Sun, Q. et al. A whole- brain map of long- range inputs to GABAergic interneurons in the 478 mouse medial prefrontal cortex. Nat Neurosci 22, 1357- 1370 (2019). 479 35. David, K. K., Fang, H. Y., Peng, G. C. Y. & Gnadt, J. W. NIH BRAIN Circuits Programs: 480 An Experiment in Supporting Team Neuroscience. Neuron 108, 1020- 1024 (2020). 481 36. Ecker, J. R. et al. The BRAIN Initiative Cell Census Consortium: Lessons Learned toward 482 Generating a Comprehensive Brain Cell Atlas. Neuron 96, 542- 557 (2017). 483 37. Hsu, N. S. et al. The promise of the BRAIN initiative: NIH strategies for understanding 484 neural circuit function. Curr Opin Neurobiol 65, 162- 166 (2020). 485 38. Litvina, E. et al. BRAIN Initiative: Cutting- Edge Tools and Resources for the Community. J 486 Neurosci 39, 8275- 8284 (2019). 487 39. Levene, H. Robust Tests for Equality of Variances. in Contributions to Probability and 488 Statistics: Essays in Honor of Harold Hotelling 278- 292 (Stanford University Press, 1960). 489 40. Lawson, C. L. & Hanson, R. J. Solving Least Squares Problems. (Society for Industrial and 490 Applied Mathematics, 1995). 491 41. Lin, C.- J. Projected gradient methods for nonnegative matrix factorization. Neural Comput 492 19, 2756- 2779 (2007). 493 42. Wilcoxon, F. Individual comparisons by ranking methods. Biometrics Bulletin 1, 80- 83 494 (1945).
+
+<--- Page Split --->
+
+## Acknowledgements.  
+
+We thank Dr. Rodrigo Muñoz- Castañeda for help with validating the mapping of neuronal reconstructions to anatomical coordinates. This work was supported in part by NIH grants R01NS39600, U01MH114829, and RF1MH128693.  
+
+## Author contributions.  
+
+D.W.W., S.B., S.S., and S.V. contributed to the analysis and interpretation of data, to the writing of software, and to the writing of the manuscript. G.A.A. contributed to the conception of the project, to the analysis and interpretation of data, and to the writing of the manuscript.  
+
+## Competing interests.  
+
+All authors declare that they have no competing interests.  
+
+## Materials & Correspondence.  
+
+All correspondence and material requests should be addressed to G.A.A. (ascoli@gmu.edu).
+
+<--- Page Split --->
+![](images/Figure_unknown_0.jpg)
+
+<center>Figures </center>  
+
+<table><tr><td>Regions</td><td>AM</td><td>LM</td><td>fx</td><td>TH</td><td>MM</td><td>dhc</td><td>AV</td><td>LHA</td><td>CC</td></tr><tr><td>AA1090</td><td>159</td><td>129</td><td>84</td><td>83</td><td>78</td><td>60</td><td>42</td><td>39</td><td>35</td></tr><tr><td>AA1058</td><td>143</td><td>209</td><td>178</td><td>57</td><td>0</td><td>136</td><td>0</td><td>38</td><td>2</td></tr></table>  
+
+Figure 1. Brain- wide neuronal projections. CCF- registered reconstruction of two presubicular neurons (AA1090 in black and AA1058 in blue from the Janelia MouseLight project) invading 9 regions out of 40 potential targets along with the numbers of axonal points of the neurons in each highlighted region (posterior view of brain). CCF, common coordinate framework; AM, Anteromedial nucleus; AV, Anteroventral nucleus; cc, corpus callosum; dhc, dorsal hippocampal commissure; fx, fornix; LHA, Lateral hypothalamic area; LM, Lateral mammillary nucleus; MM, Medial mammillary nucleus; TH, other thalamic nuclei.
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Figure 2. Definitions of neuron classes and clustering methods. (A) In a single-class scenario, </center>
+
+<--- Page Split --->
+
+the distribution of differences between neurons can be calculated for all neuron pairs (pink double- arrows). (B) If two distinct classes exist, neurons (represented here as black dots) will tend to have more similar projections within their class (red double- arrows) and more different ones across classes (blue double arrow). (C) The differences within the classes (red distribution) will be smaller than those between classes (blue distribution). (D) The distribution of the combined frequency of differences, in a multi- class scenario (red- blue stacked areas; green half- height width), will be wider than that of a single- class distribution (pink curve; orange half- height width). (E) Diagram showing the randomization of projection patterns through the repeated pairwise swapping of axonal point counts between two neurons across two of their potential target regions, which preserves the column (for a given region) and row (for a given neuron) sums of the matrix. This swapping results in a projection pattern “continuum” that matches with the overall distribution that represents the 1- class null hypothesis. (F) Unsupervised hierarchical clustering groups a set of neurons into classes based on their relative pairwise differences or similarities, as modeled by a binary dendrogram. The top (root) of the dendrogram represents all neurons lumped into the same class, while the bottom (leaves) shows all neurons split into separate classes.
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Figure 3. Primary motor cortex L6 (IT vs. CT). (A) Representation of the two clusters produced </center>
+
+<--- Page Split --->
+
+by Levene's one- tailed test for the equality of variances and unsupervised hierarchical clustering, using MouseLight neurons from the primary motor cortex, layer 6 (n = 52). (B) Colormap of the axonal distributions of neurons (columns) across anatomical regions (rows), with darker shades representing more axonal projections. (C) Axonal pathways of representative IT (intratelencephalic, red) and CT (corticothalamic, blue) neurons with semitransparent surfaces of primary motor cortex layer 6 (green) and selected thalamic nuclei (pink). The two black dots indicate the cell body locations of the two representative cells from each class. (D) Axonal pathways of all IT and CT neurons in the MouseLight sample (same color coding).
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Figure 4. Classification of projection neuron types in the presubiculum. (A) Representation of 5 axonal clusters produced by Levene’s test and unsupervised hierarchical clustering of neurons from the presubiculum (n = 93). (B) Colormap of the axonal distributions of neurons (columns) across anatomical regions (rows), with darker shades representing more axonal projections. Parcel names highlighted in pink are hypothalamus related. Parcel names highlighted in yellow and light blue are thalamus related. (C) Neuron-to-target assignments for the identified axonal projection classes and corresponding anatomical regions (dotted line: contralateral). (D) Anterior view of the </center>
+
+<--- Page Split --->
+
+mouse brain with one neuron from each class. Color coding of neurons and semitransparent anatomical areas shown in A, B, and C. (E) Posterior view of the brain with all MouseLight presubicular neurons. \(CA3 + CA1\) : Cornu Ammonis areas 3 and 1; DG: dentate gyrus; Sub: subiculum; LEC: lateral entorhinal cortex; dMEC: dorsal portion of the medial entorhinal cortex; ParaS: parasubiculum; PostS: postsubiculum; Retrohippocampal region; DV(gr.)RtSpl: dorsal and ventral (granular) retrosplenial cortex; L(ag.)RtSpl: lateral (agranular) retrosplenial cortex; MidB: midbrain; Hyp: hypothalamus; PMdv+TU: dorsal and ventral premammillary nucleus and tuberal nucleus; MM+LZ: medial mammillary nucleus and hypothalamic lateral zone; MBO+LM: mammillary body and lateral mammillary nucleus; mATN+PT: medial anterior thalamic nucleus and parataenial nucleus; TH+LGN: thalamus and lateral geniculate nucleus; dvATN+MGN: dorsal and ventral anterior thalamic nucleus and medial geniculate nucleus; IAD+IAM: interanterodorsal and interanteromedial nucleus of the thalamus; LD+AD: lateral dorsal and anterodorsal nucleus of thalamus.
+
+<--- Page Split --->
+![](images/Figure_5.jpg)
+
+<center>Figure 5. Spatial distributions of somata in the presubiculum across projection classes. </center>
+
+<--- Page Split --->
+
+654 Convex hulls of neurons (spheres) from classes A38 (blue), B27 (red), D19 (brown), and E6 655 (green), and semitransparent presubiculum (green). (A) Left sagittal view of A38 and B27. (B) 656 Layer 1 (green), layer 2 (purple), and layer 3 (orange) of the presubiculum are highlighted in an 657 anterior coronal view, with somata from A38 in blue and D19 in brown. Most of the A38 somata 658 are concentrated in layer 2, while the D19 somata tend to be more concentrated in layers 1 and 3. 659 Somata that do not follow this pattern are indicated with a white dot inside of the circle. (C) Left 660 sagittal view of D19 and E6. (D) Posterior coronal view of B27 and E6. 661 662
+
+<--- Page Split --->
+![](images/Figure_6.jpg)
+
+<center>Figure 6. Divergent path distance comparison from one neuron class in the presubiculum to </center>
+
+<--- Page Split --->
+
+665 its targets. (A) Left: box and whisker plot depicting the median, first and third quartiles, and full 666 range of the path distances from class A38 to its major ipsilateral (I) and contralateral (C) targets. 667 Right: the path distance of an archetype neuron from class A38 (light blue), from its soma (black) 668 in the ipsilateral presubiculum (green) to the subiculum (purple), is significantly shorter than that 669 (dark blue) to the lateral entorhinal cortex (orange). (B) Left: box and whisker plot depicting the 670 distributions of path distances from class B27 to its major ipsilateral and contralateral targets. 671 Right: the path distance of an archetype neuron from class B27 (light red), from its soma (black) 672 in the ipsilateral presubiculum (green) to the parasubiculum (brown), is significantly shorter than 673 that (dark red) to the medial entorhinal cortex, dorsal zone (cyan). (C) Left: box and whisker plot 674 depicting the path distances from class D19 to its major ipsilateral targets. Right: the path distance 675 of an archetype neuron from class D19 (light brown), from its soma (black) in the ipsilateral 676 presubiculum (green) to the midbrain (magenta), is significantly shorter than that (dark brown) to 677 the hypothalamus and lateral mammillary nucleus (red). See Figure 4 for abbreviation definitions. 678 Significant differences in distances were calculated using a Wilcoxon Signed Rank Test performed 679 on neuronal path distances and multiple testing was corrected for by False Discovery Rate to 680 determine the significance of the resultant p- values.
+
+<--- Page Split --->
+![](images/Figure_7.jpg)
+
+<center>Figure 7. Convergent path distance comparison from each presubiculum cluster to major targets. (A) Top left: box and whisker plot depicting the median, first and third quartiles, and full range of the path distances from neurons in ipsilateral (I) and contralateral (C) classes to the medial entorhinal cortex, dorsal zone. Bottom: the distance of an archetype neuron from class B27 (red), from its soma in the ipsilateral presubiculum (green) to the dMEC (purple), is significantly longer than the comparable distance of an archetype neuron from class D19 (brown). (B) Top left: box and whisker plot depicting the path distances from neurons in ipsilateral and contralateral classes to the parasubiculum. Bottom: the distance of an archetype neuron from class B27 (red), from its soma in the presubiculum (green) to the ipsilateral ParaS (purple), is significantly shorter than the </center>
+
+<--- Page Split --->
+
+729 comparable distance of an archetype neuron from class D19 (brown). (C) Top left: box and whisker 730 plot of the path distances from neurons in contralateral classes to the subiculum. Bottom: the 731 distance of an archetype neuron from class A38 (blue), from its soma in the presubiculum (green) 732 to the contralateral Sub (purple), is significantly shorter than the comparable distance of an 733 archetype neuron from class B27 (red). See Figure 4 for abbreviation definitions. Significant 734 differences in distances were calculated using a Wilcoxon Signed Rank Test performed on 735 neuronal path distances and multiple testing was corrected for by False Discovery Rate to 736 determine the significance of the resultant p-values.
+
+<--- Page Split --->
+
+Supplemental TablesTable S1. Raw axonal counts for primary motor area layer 6. Table S2. Raw axonal counts for presubiculum. Table S3. Non- negative least- square normalizations.
+
+<--- Page Split --->

preprint/preprint__1024b305173c41b8401ff5ab5cc9c8082f65a9cfffd5cbcad6232e5357b19148/preprint__1024b305173c41b8401ff5ab5cc9c8082f65a9cfffd5cbcad6232e5357b19148_det.mmd

@@ -0,0 +1,362 @@
+<|ref|>title<|/ref|><|det|>[[44, 108, 880, 177]]<|/det|>
+# Unsupervised classification of brain-wide axons reveals neuronal projection blueprint  
+
+<|ref|>text<|/ref|><|det|>[[44, 196, 160, 214]]<|/det|>
+Diek Wheeler  
+
+<|ref|>text<|/ref|><|det|>[[55, 223, 235, 240]]<|/det|>
+duhee1e5@gmu.edu  
+
+<|ref|>text<|/ref|><|det|>[[44, 269, 631, 288]]<|/det|>
+George Mason University https://orcid.org/0000- 0001- 8635- 0033  
+
+<|ref|>text<|/ref|><|det|>[[44, 293, 183, 310]]<|/det|>
+Shaina Banduri  
+
+<|ref|>text<|/ref|><|det|>[[55, 315, 273, 333]]<|/det|>
+George Mason University  
+
+<|ref|>text<|/ref|><|det|>[[44, 339, 231, 357]]<|/det|>
+Sruthi Sankararaman  
+
+<|ref|>text<|/ref|><|det|>[[55, 362, 273, 380]]<|/det|>
+George Mason University  
+
+<|ref|>text<|/ref|><|det|>[[44, 386, 175, 404]]<|/det|>
+Samhita Vinay  
+
+<|ref|>text<|/ref|><|det|>[[55, 409, 273, 426]]<|/det|>
+George Mason University  
+
+<|ref|>text<|/ref|><|det|>[[44, 432, 168, 450]]<|/det|>
+Giorgio Ascoli  
+
+<|ref|>text<|/ref|><|det|>[[55, 455, 632, 473]]<|/det|>
+George Mason University https://orcid.org/0000- 0002- 0964- 676X  
+
+<|ref|>text<|/ref|><|det|>[[44, 515, 283, 534]]<|/det|>
+Biological Sciences - Article  
+
+<|ref|>text<|/ref|><|det|>[[44, 553, 135, 571]]<|/det|>
+Keywords:  
+
+<|ref|>text<|/ref|><|det|>[[44, 590, 283, 609]]<|/det|>
+Posted Date: July 3rd, 2023  
+
+<|ref|>text<|/ref|><|det|>[[44, 628, 473, 647]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 3044664/v1  
+
+<|ref|>text<|/ref|><|det|>[[44, 665, 909, 707]]<|/det|>
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+<|ref|>text<|/ref|><|det|>[[44, 727, 254, 746]]<|/det|>
+Additional Declarations:  
+
+<|ref|>text<|/ref|><|det|>[[44, 751, 318, 770]]<|/det|>
+There is NO Competing Interest.  
+
+<|ref|>text<|/ref|><|det|>[[44, 788, 528, 807]]<|/det|>
+Supplemental Tables are not available with this version.  
+
+<|ref|>text<|/ref|><|det|>[[42, 858, 944, 901]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on February 20th, 2024. See the published version at https://doi.org/10.1038/s41467- 024- 45741- x.
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[168, 91, 833, 150]]<|/det|>
+# Unsupervised classification of brain-wide axons reveals neuronal projection blueprint  
+
+<|ref|>text<|/ref|><|det|>[[137, 175, 861, 224]]<|/det|>
+Wheeler, Diek W. \(^{1}\) , Banduri, Shaina \(^{1}\) , Sankararaman, Sruthi \(^{1}\) , Vinay, Samhita \(^{1}\) , Ascoli, Giorgio A. \(^{1*}\)
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[66, 123, 884, 510]]<|/det|>
+7 Long- range axonal projections are quintessential determinants of network connectivity, linking cellular organization and circuit architecture. Here we introduce a quantitative strategy to identify, from a given source region, all “projection neuron types” with statistically different patterns of anatomical targeting. We first validate the proposed technique with well- characterized data from layer 6 of the mouse primary motor cortex. The results yield two clusters, consistent with previously discovered cortico- thalamic and intra- telencephalic neuron classes. We next analyze neurons from the presubiculum, a less- explored region. Extending sparse knowledge from earlier retrograde tracing studies, we identify five classes of presubicular projecting neurons, revealing unique patterns of divergence, convergence, and specificity. We thus report several findings: (1) individual classes target multiple subregions along defined functions, such as spatial representation vs. sensory integration and visual vs. auditory input; (2) all hypothalamic regions are exclusively targeted by the same class also invading midbrain, a sharp subset of thalamic nuclei, and agranular retrosplenial cortex; (3) Cornu Ammonis, in contrast, receives input from the same presubicular axons projecting to granular retrosplenial cortex, also the purview of a single class; (4) path distances from the presubiculum to the same targets differ significantly between classes, as do the path distances to distinct targets within most classes, suggesting fine temporal coordination in activating distant areas; (5) the identified classes have highly non- uniform abundances, with substantially more neurons projecting to midbrain and hypothalamus than to medial and lateral entorhinal cortex; (6) lastly, presubicular soma locations are segregated among classes, indicating topographic organization of projections. This study thus demonstrates that classifying neurons based on statistically distinct axonal projection patterns sheds light on the functional organizational of their circuit.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 125, 883, 300]]<|/det|>
+The classification of neurons in the mammalian nervous system has long been a focus of intensive investigation. While local features from slice preparations in vitro may suffice to infer the circuit roles of GABAergic interneurons \(^{1 - 3}\) , long- range projecting axons are crucial architectural elements of neural organization \(^{4,5}\) constituting the conceptual and physical nexus between brain- wide circuits and synaptic communication \(^{6}\) . Thus, projection axons have long been digitally traced from serial sections after in vivo labeling and light microscopy imaging \(^{7 - 10}\) . At the same time, their macroscopic extent ( \(\sim 1\) cm span; \(\sim 1\) m cable length) and microscopic caliber ( \(\sim 100\) nm branch thickness) combine into a formidable technological challenge for large- scale collection \(^{11,12}\) . As a result, the number of completely reconstructed projection axons in any mammalian neural system has until recently remained into the low tens.  
+
+<|ref|>text<|/ref|><|det|>[[113, 315, 883, 508]]<|/det|>
+A source brain region projecting to N targets (where N typically ranges between 10 and 50 in the mouse cortex) could contain any combination of \(2^{N}\) - 1 distinct axonal projection types. Such a combinatorics challenge requires a large- scale data collection for proper classification. Projects based on fluorescent Micro- Optical Sectioning Tomography (fMOST) technology \(^{13 - 15}\) or the Janelia MouseLight platform \(^{16}\) , launched in recent years to address this need, produced nearly 10,000 mouse whole- brain single neuron reconstructions registered to a 3D Common Coordinate Framework (CCF) \(^{17}\) with consensus anatomical labeling \(^{18}\) . However, these newly available data do not themselves generate novel scientific insights, explain brain circuitry, or even disprove that axons might simply invade a random subset of the regional target areas \(^{19}\) . Rigorous methods are needed to test the hypothesis that specific projection types exist, to characterize their identities, and to quantify their population sizes \(^{20}\) .  
+
+<|ref|>text<|/ref|><|det|>[[113, 525, 883, 735]]<|/det|>
+This study introduces an original technique to objectively identify projection- based neuronal classes. To ascertain whether a collection of axonal projections might result from essentially random variation within the constraints of regional connectivity or likely reflects distinct neuron types, we begin from the foundational criterion for classification: if a set of items belongs to segregated classes, their pairwise inter- individual differences must be on average larger between than within classes. In other words, two items from the same class should tend to be more similar to each other than two items from separate classes. To implement this logic into a classification framework, we couple rigorous statistical testing with unsupervised hierarchical clustering. A unique strength of this approach is its entirely data- driven granularity: the continuous accumulation of new tracings will progressively refine the classification details with increasing statistical power. We can then characterize the identified projection classes by quantifying their population size, topographic soma distributions, and convergence and divergence patterns.  
+
+<|ref|>text<|/ref|><|det|>[[113, 751, 883, 875]]<|/det|>
+In the remainder of this article, we first propose a formal definition of and a quantitative solution for the classification problem. We validate our approach by applying it to layer 6 of the primary motor cortex, and then utilize it to study the presubiculum, a rather under- investigated region of the mouse brain. We next quantify the neuronal population sizes of the presubicular projection classes and characterize the spatial distribution of their somata. Finally, we analyze the patterns of divergence and convergence of presubicular projection classes. We conclude by discussing the biological interpretations of these results.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 91, 183, 106]]<|/det|>
+## Results.  
+
+<|ref|>text<|/ref|><|det|>[[115, 125, 706, 144]]<|/det|>
+Formal Definition and Quantitative Solution of the Classification Problem.  
+
+<|ref|>text<|/ref|><|det|>[[115, 159, 883, 509]]<|/det|>
+The axonal projections of each neuron in a source region can be represented as k- dimensional vectors, where k is the number of target regions invaded by the source region. Each of the k components of the vector quantifies the number of axonal points within the corresponding region (Figure 1). We explore the null hypothesis, \(\mathrm{H_0}\) , that all neurons from a source region belong to a single projection class (Figure 2A), as opposed to the alternative hypothesis, \(\mathrm{H_A}\) , that distinct projection classes exist from that source region (Figure 2B). If two hypothetical classes exist, the projections will be more similar between neurons within a class and more different across classes (Figure 2C). In such a two- class scenario, the combined within- and across- class distances would thus form a wider distribution than the distribution generated if all neurons belong to just a single class (Figure 2D). To formally test \(\mathrm{H_A}\) , we measure all pairwise differences between neurons (as arccosine vector distances, see Methods & Materials). We then generate the distribution of distances for \(\mathrm{H_0}\) by randomizing the projection patterns while preserving total axonal length both by neuron and target region. We achieve this single- class "continuum" by iterative stochastic swapping of axonal points between neurons across two target regions (see Figure 2E and Methods & Materials). We can then apply Levene's one- tail statistical test to ascertain whether the original distribution of pairwise distances has significantly larger variance than the randomized distribution. If the answer is positive, we must discard \(\mathrm{H_0}\) and accept \(\mathrm{H_A}\) . Starting from the top node in an unsupervised hierarchical clustering tree, we can thus repeat Levene's test on the neurons of each of the two subtrees, continuing the process until none of the variance differences are statistically significant (Figure 2F).  
+
+<|ref|>sub_title<|/ref|><|det|>[[116, 525, 335, 544]]<|/det|>
+## Validation of the Approach.  
+
+<|ref|>text<|/ref|><|det|>[[113, 560, 883, 821]]<|/det|>
+To validate the above research design, we first analyzed 52 MouseLight layer 6 neurons from the primary motor cortex<sup>21</sup>. This anatomical area is known to contain two distinct projection classes with well- defined subdivisions: cortico- thalamic (CT) and cortico- cortical or intra- telencephalic (IT) neurons<sup>22</sup>. The variance of the distribution of pairwise axonal projection differences of these neurons was significantly larger than that of the randomized projections (p < 0.001; variance of real data = 373.4; variance of randomized data = 195.7), indicating the existence of distinct clusters. However, both subtrees after the first split of unsupervised hierarchical clustering returned a non- significant Levene's test (IT: \(\mathrm{p} = \mathrm{N/A}\) ; variance of real data = 219.9; variance of randomized data = 240.0; CT: \(\mathrm{p} > 0.05\) ; variance of real data = 295.1; variance of randomized data = 264.0), revealing exactly two clusters (Figure 3A). The first cluster, consisting of 21 neurons, projected almost exclusively to motor cortical targets; the second cluster of 31 neurons projected primarily to thalamic targets (Figures 3B- D and Table S1). These patterns were fully consistent with the axonal pathways of the IT and CT neuronal classes, respectively. This finding thus corroborates the validity of employing Levene's test of variance on pairwise difference distributions to identify statistically distinct classes in unsupervised hierarchical clustering.  
+
+<|ref|>text<|/ref|><|det|>[[115, 838, 650, 856]]<|/det|>
+Classification of Projection Neurons from the Mouse Presubiculum.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 90, 883, 459]]<|/det|>
+We then applied our analytic technique to a lesser- explored source region of the mouse brain: the presubiculum. Unsupervised clustering and the test of variance demonstrated that the 93 MouseLight neurons from the presubiculum form five distinct projection classes (Figures 4A- C). We designate each class by a letter (A- E) followed by the number of neurons in the class (Figure 4C). The first class, A38, primarily targets the lateral entorhinal cortex (LEC), accounting for 82% of axonal extent outside of the presubiculum. This class also invades the dorsoventral (granular) retrosplenial cortex as well as the hippocampal formation (dentate gyrus, CA3, CA2, CA1, and subiculum). The second class, B27, mainly targets the dorsal portion of the medial entorhinal cortex (dMEC), accounting for 92.5% of extra- presubicular axonal extent, as well as retrohippocampal zone and parasubiculum. Class C3 neurons mostly target the contralateral dMEC (42%) and LEC (40%), subiculum (14%), and parasubiculum (4%) through extensive callosal and commissural fibers. Class D19 has the most complex (and unreported) pattern of innervation: in addition to major projections to the subiculum (40.8%) and dentate gyrus (16.3%), it is the sole source of projections to the lateral (agranular) retrosplenial cortex, to the hypothalamus (including the lateral mammillary nucleus and 18 additional nuclei: Table S2), and to the superior and inferior colliculi in the midbrain. This neuronal class also projects to a subset of 8 thalamic nuclei, including the medial part of the anterior thalamic nucleus (ATN) and the lateral geniculate nucleus. Lastly, class E6 projects to a complementary set of 14 other thalamic nuclei, including the ventral, dorsal, anterior, and lateral parts of the ATN and the medial geniculate nucleus. Neurons from all five projection classes also have substantial collaterals within the presubiculum. Examples of projection neurons from each of the presubicular projection classes are depicted in Figures 4D- E.  
+
+<|ref|>text<|/ref|><|det|>[[115, 472, 747, 491]]<|/det|>
+Presubicular Projection Classes Have Non- Uniform Neuronal Population Sizes.  
+
+<|ref|>text<|/ref|><|det|>[[113, 507, 883, 752]]<|/det|>
+Next, we quantified the proportion of neurons in the mouse presubiculum that belong to each projection class. To this aim, we extracted the anterograde tract tracing density distributions from the Allen Institute regional connectivity atlas and matched the fractions of neurons in every class based on their axonal patterns by numerical optimization (see Non- Negative Least Squares in Methods & Materials). The results converged with very small residual error (<0.0006%) indicating a near- exact correspondence between single- neuron and regional projections. Class D19, reaching the midbrain, hypothalamus, lateral (agranular) retrosplenial, and the lateral geniculate (visual thalamus) accounted for a plurality (38.1%) of neurons. Class A38, targeting the hippocampus, subiculum, dorsoventral (granular) retrosplenial cortex, and lateral entorhinal cortex ("what" pathway), accounted for the second largest share (30.6%) of neurons. Class B27, projecting to the parasubiculum and medial entorhinal cortex ("where" pathway) consisted of 16.3% of projection neurons. Class E6, focused on other thalamic nuclei including medial geniculate (auditory), was responsible for 13.7% of presubicular neurons. The diffuse contralateral projections of class C3 comprised the remaining 1.3%.  
+
+<|ref|>text<|/ref|><|det|>[[115, 769, 883, 891]]<|/det|>
+When accounting for these relative proportions together with the MouseLight axonal projections (Table S2), we can estimate the contribution of each class to the presubicular projections in each collection of target regions. In particular, the dentate gyrus receives 21% of its presubicular afferents from class A38 and 79% from class D19. The subiculum receives 69% of its presubicular afferents from class D19, 30% from class A38, and 1% from class C3. The lateral entorhinal cortex receives 99% of presubicular afferents from class A38 and 1% from class C3. The dorsal medial entorhinal cortex and parasubiculum receive 99% of presubicular afferents from class B27 and 1%
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 90, 883, 160]]<|/det|>
+from class C3. All other regions are targeted by individual classes: CA3, CA1, and the dorsoventral (granular) retrosplenial cortex by A38; the midbrain, hypothalamus, lateral (agranular) retrosplenial cortex, and part of the thalamic nuclei including mATN and LGN by D19; and the rest of the thalamic nuclei including dvATN and MGN by E6.  
+
+<|ref|>text<|/ref|><|det|>[[115, 176, 836, 195]]<|/det|>
+Spatial Distribution of Somata Reveals Topographic Organization of Neuronal Projections.  
+
+<|ref|>text<|/ref|><|det|>[[115, 212, 883, 404]]<|/det|>
+Computational geometry analysis of soma locations within the presubiculum demonstrated a clear spatial separation among the four main projection classes: A38, B27, D19, and E6 (the smallest class, C3, is largely contralateral projecting). Specifically, the convex hull volume of each neuron class overlapped only minimally ( \(\sim 5 - 20\%\) ) with that of other neuron classes (Figure 5A- C). In particular, class A38 was positioned more rostrally and dorsally relative to the caudal- ventral position of class B27, with approximately \(14\%\) of overlap (Figure 5A). The overlap of A38 was maximal with D19 \((21\%)\) ; however, while most A38 neurons had a selective somatic concentration in layer 2 ( \(34 / 38\) : \(89.5\%\) ), D19 had a somatic distribution across all 3 presubicular layers: \(21\%\) in layer 1 and \(26\%\) in layer 3 (Figure 5B). Class E6 had the most lateral positioning resulting in almost complete segregation from the other projection classes: there were so few overlapping somata that a proper convex hull volume of the overlap could not be calculated (Figures 5C- D).  
+
+<|ref|>text<|/ref|><|det|>[[115, 420, 785, 439]]<|/det|>
+Efferent Path Distances from the Same Neurons Significantly Vary by Target Region.  
+
+<|ref|>text<|/ref|><|det|>[[115, 455, 883, 716]]<|/det|>
+We tested whether the path distances from presubicular neurons of a given projection class differed across their divergent target regions (Figure 6). In these analyses of divergence, ipsilateral and contralateral targets were considered separately, as the latter are systematically farther than the former. For class A38 neurons, projection distances to the ipsilateral lateral entorhinal cortex, subiculum, and dentate gyrus are shorter than those to the ipsilateral hippocampus; moreover, projection distances to the ipsilateral lateral entorhinal cortex are longer than those to the ipsilateral subiculum and dentate gyrus. Similarly, projection distances to the contralateral subiculum and lateral entorhinal cortex are shorter than those to the contralateral hippocampus. Thus, presubicular efferent path distances differ less between ipsilateral and contralateral hippocampus than between other targets across brain hemispheres (Figure 6A). For class B27, projections to the ipsilateral parasubiculum have shorter paths than those to medial entorhinal cortex, dorsal zone, but the distances are comparable in the contralateral case (Figure 6B). Finally, for class D19, projections both to the ipsilateral medial anterior thalamic nucleus and lateral geniculate nucleus, and to the ipsilateral hypothalamus and lateral mammillary nucleus combined have longer paths than those to the ipsilateral midbrain (Figure 6C).  
+
+<|ref|>text<|/ref|><|det|>[[115, 732, 828, 752]]<|/det|>
+Afferent Path Distances to the Same Target Region Significantly Vary by Projection Class.  
+
+<|ref|>text<|/ref|><|det|>[[115, 768, 883, 890]]<|/det|>
+Next, we asked whether the axons from neurons of distinct projection classes converging onto their shared targets had different path distances. With the sole exception of the dentate gyrus, all target regions displayed a significant dependence of path distance on the presubicular neuron class (Figure 7A- C). For the ipsilateral medial entorhinal cortex, dorsal zone, projections from E6 and D19 have shorter distances than those from B27 and A38, and projections from B27 have shorter distances than those from A38. For the contralateral medial entorhinal cortex, in contrast, projections from B27 have longer distances than those from A38 (Figure 7A). For the ipsilateral
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[52, 88, 884, 145]]<|/det|>
+211 parasubiculum, path distances from D19 are longer than those from B27 (Figure 7B). Finally, for 212 both the contralateral subiculum and parasubiculum, path distances from B27 are longer than those 213 from A38 (Figure 7B- C).
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 91, 209, 107]]<|/det|>
+## Discussion.  
+
+<|ref|>text<|/ref|><|det|>[[115, 124, 882, 249]]<|/det|>
+This study introduced an original method to objectively identify projection- based neuronal classes by pairing the Levene's test with unsupervised hierarchical clustering. We first conducted a confirmatory study on layer 6 of the primary motor cortex to verify that the proposed technique could reproduce known projection types in a previously explored area of the mammalian brain. The results yielded two clusters with axonal projections consistent with those of the corticothalamic and intratelencephalic neuron classes found in past studies, thereby confirming the validity of the technique<sup>23</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[115, 264, 882, 474]]<|/det|>
+To test whether the technique could lead to novel insights, we then applied it to the presubiculum, a region with crucial cognitive function<sup>24</sup>, yet few studies on its circuitry<sup>25</sup>. The results yielded five clusters, indicating distinct neuron classes, which led us to reject the null hypothesis that projection neurons exhibit random variation within the constraints of regional connectivity from the presubiculum. In an earlier study<sup>26</sup>, retrograde tracing identified five classes of neurons projecting from the presubiculum, which target the retrosplenial cortex, contralateral subiculum, medial entorhinal cortex, anterior thalamic nucleus, and lateral mammillary nucleus. Our results confirm the existence of these five classes and add new information that reveals patterns of divergence (e.g., class A38 projects to the retrosplenial cortex, dentate gyrus, subiculum, and entorhinal cortex), convergence (e.g., the subiculum receives projections from classes A38, contralateral C3, and D19), and specificity (e.g., class E6 projects exclusively to the medial geniculate nucleus, and all hypothalamic regions receive projections solely from class D19).  
+
+<|ref|>text<|/ref|><|det|>[[115, 490, 882, 577]]<|/det|>
+The proposed clustering technique correctly distinguishes cortical (classes A38, B27, and C3) from subcortical (D19 and E6) pathways in the second binary split in the hierarchical classification. These results also add cellular level details to previously reported presubicular projections to retrosplenial cortex and thalamic reticular nuclei<sup>27</sup>, as well as a broader circuit context to the characterization of individual presubicular neurons targeting the medial entorhinal cortex<sup>28</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[115, 594, 882, 805]]<|/det|>
+Furthermore, our findings reveal that several target regions are spatially subdivided according to the differing inputs between classes. These regions include the entorhinal cortex (lateral projections mainly from class A38 and medial projections primarily from class B27), retrosplenial cortex (dorsoventral granular projections almost exclusively from class A38 and lateral agranular projections solely from class D19), and thalamus (medial anterior thalamic nucleus and lateral geniculate nucleus projections principally from class D19 and dorsoventral anterior thalamic nucleus and medial geniculate nucleus projections predominantly from class E6). Some of these regional subdivisions also have known functional distinctions: for instance, the medial entorhinal cortex specializes in spatial representation while the lateral entorhinal cortex specializes in integrating sensory input<sup>29</sup>. Among the thalamic geniculate nuclei, the medial geniculate nucleus is part of the auditory pathway, whereas the lateral geniculate nucleus is part of the visual pathway<sup>4</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[115, 821, 882, 891]]<|/det|>
+From a comparison of divergent path distances from one presubicular class to its major targets, along with a comparison of convergent path distances from each presubicular class to collectively major targets, we found that path distances to the same targets were significantly different between classes, as were the path distances to distinct targets within most classes. This might imply that
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 90, 883, 125]]<|/det|>
+260 electrical impulses reach different targets with varying delays, both within the same class and 261 between classes.  
+
+<|ref|>text<|/ref|><|det|>[[115, 141, 883, 248]]<|/det|>
+Topographic analysis of presubicular classes revealed spatial separation between the somata of each class. This suggests the possibility of anatomically mapping the input and output of the circuitry specializing in head direction computations<sup>30</sup>. Our reported topography of presubicular projections classes is consistent with the recently observed local modularity of the head- direction microcircuit<sup>31</sup>, and may help clarify the relationship between the egocentric and allocentric spatial and episodic representations of the cortico- hippocampal system<sup>32</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[115, 263, 883, 404]]<|/det|>
+As with many secondary data analyses, we have limited knowledge of, and control over, artifactual shortcomings in the utilized datasets due to possible idiosyncrasies in labeling, imaging, tracing, registration, and mapping. However, the technique introduced with this work is applicable to many disparate sources of data besides MouseLight, including fMOST<sup>13–15</sup> and even MapSeq/BarSeq<sup>33,34</sup>. These data sources follow separate experimental and computational protocols, allowing independent validation for the source regions in which these datasets overlap. Our results so far, in the cases of the mouse primary motor cortex and presubiculum, indicate that the executed analysis is robust to these possible confounding variables<sup>22</sup>.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 91, 215, 108]]<|/det|>
+## Conclusion.  
+
+<|ref|>text<|/ref|><|det|>[[113, 123, 882, 318]]<|/det|>
+Overall, this study revealed that neurons can be divided into distinct classes based on axonal projection patterns, as demonstrated in layer 6 of the primary motor cortex and the presubiculum. Our applied analyses can be used to similarly analyze neurons projecting from all other mouse brain regions with sufficient data. There are currently approximately 40 regions fitting this criterion in the existing datasets, but this number is expected to grow in the near future. Furthermore, we suggest the application of pairing Levene's test and unsupervised hierarchical clustering to other complementary datasets, such as single- cell transcriptomic datasets, to classify neurons across a molecular domain, in addition to an anatomical domain, as demonstrated here. Moreover, all these complementary datasets are broadly expected to continue to grow in sample size, brain coverage, and acquisition pace<sup>35,36</sup>, supporting a call to establish cloud- based, community accessible pipelines for robust, rigorous, and systematic neuronal characterization<sup>37,38</sup>.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 91, 293, 108]]<|/det|>
+## Material & Methods.  
+
+<|ref|>text<|/ref|><|det|>[[115, 125, 453, 143]]<|/det|>
+Data Extraction, Storage, and Conversion.  
+
+<|ref|>text<|/ref|><|det|>[[115, 158, 883, 316]]<|/det|>
+The location of each axonal data point for nearly 1100 neurons was extracted from the Janelia MouseLight public dataset<sup>21</sup> using the freeware JSONLab v1.5 (https://www.mathworks.com/matlabcentral/mc- downloads/downloads/submissions/33381/versions/22/download/zip). These data were contained in a JSON file for each neuron, where X- Y- Z coordinates and parcel information were provided for each axonal point of the neuron. The axonal points in each brain parcel were tabulated for all neurons and were stored in a matrix (Tables S1- 2), in which each row represents a neuron, each column represents a parcel, and the values in each cell represent the axonal counts of a particular neuron in a particular region (Figure 1).  
+
+<|ref|>text<|/ref|><|det|>[[115, 333, 268, 351]]<|/det|>
+Hypothesis Design.  
+
+<|ref|>text<|/ref|><|det|>[[115, 367, 883, 595]]<|/det|>
+To determine whether distinct projection classes of neurons exist from a particular parcel of the brain, hypothesis \(\mathrm{H_A}\) , we tested the pairwise differences between neurons from the experimental matrices described above. If only a single class of neurons exists, then only a single distribution of differences between neurons will be generated (Figure 2A). If two hypothetical classes exist, then the differences between neurons, evaluated two at a time, will be smaller within a given class than across the two classes (Figure 2B- C). In a multi- class scenario, a histogram of the differences between neurons should be wider than the distribution generated when all the neurons belong to just a single class (Figure 2D). To generate the distribution of differences for the null hypothesis, \(\mathrm{H_0}\) , a randomized control matrix was generated from the original experimental matrix through multiple iterations of the stochastic pairwise swapping of axonal counts from two neurons across two target regions (Figure 2E). This method randomized the projection patterns, yielding a "continuum" consistent with the regional connectivity of Figure 2A, while preserving axonal sizes (row sums) and regional targeting (column sums) of the original experimental matrix.  
+
+<|ref|>text<|/ref|><|det|>[[115, 611, 228, 628]]<|/det|>
+Levene's Test.  
+
+<|ref|>text<|/ref|><|det|>[[115, 645, 883, 821]]<|/det|>
+We assessed the hypothesis that the variance of experimental data was significantly larger than the variance of randomized data \((\alpha = 0.05)\) . For both the experimental and randomized matrices, we computed the arccosine between a pair of neuronal vectors, each composed of the axonal counts across all target regions (https://github.com/Projectomics/MATLAB). These angles measure the projection difference of two neurons across all brain parcels. We then performed a 1- tailed Levene's test<sup>39</sup> on the angle distributions of the experimental and randomized matrices to assess whether their variances differed significantly. To this aim, we used the MATLAB function vartestn with the TestType parameter set to LeveneAbsolute. If the experimental data had a greater variance than the randomized data, then the experimental data could be further divided into classes, consistent with the scenario presented in Figure 2B.  
+
+<|ref|>text<|/ref|><|det|>[[115, 838, 423, 856]]<|/det|>
+Unsupervised Hierarchical Clustering.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 89, 883, 247]]<|/det|>
+We used unsupervised hierarchical clustering to determine a biologically accurate division of neuron classes based on axonal projection patterns. Specifically, the MATLAB linkage function, with the “average” algorithm for computing distance between clusters, was utilized on the 93 MouseLight neurons originating in the presubiculum and the 52 MouseLight neurons originating in layer 6 of the primary motor cortex. The initial assumption (null hypothesis) was that all neurons were part of a single class. If Levene’s test yielded significant results, the number of class divisions was incremented, and the technique was again repeated on each class division. This iterative process continued until none of the subdivided classes yielded significant results, thereby yielding the final class divisions (Figure 2F).  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 265, 344, 282]]<|/det|>
+## Non-Negative Least Squares.  
+
+<|ref|>text<|/ref|><|det|>[[115, 298, 883, 490]]<|/det|>
+To estimate the fractional counts of cells in each of k projection classes in each region, we matched their respective single- cell axonal patterns against the regional connectivity from anterograde tracing to the m known targets, as presented in the Allen Mouse Brain Connectivity Atlas (http://connectivity.brain-map.org/projection). The problem is equivalent to a set of constrained, weighted, linear equations that can be solved numerically by non- negative least- square (NNLS) optimization<sup>40</sup>. NNLS finds the values x that minimizes the Euclidean norm of (Ax - b) with the constraint \(x \geq 0^{41}\) , where x is the k- dimensional vector representing the fractions of neurons in each class; b is the m- dimensional vector representing the weights of the regional projections to each target; and A is a k- by- m matrix with rows representing the projections of each class (the sum of the summary vectors of the corresponding neurons) and columns representing target regions. NNLS was computed using the lsqnonneg function in MATLAB.  
+
+<|ref|>text<|/ref|><|det|>[[115, 507, 883, 699]]<|/det|>
+Matrix A and vector b were based on data from the MouseLight dataset (Table S2) and the Allen Mouse Brain Connectivity Atlas, respectively. Setting the target region to the whole brain in the Connectivity Atlas and the source region to the presubiculum resulted in 7 tracing experiments, which included projection volumes and projection densities for each target brain region. Cross referencing the targeted regions of the MouseLight axonal projections with target regions that appeared in all 7 anterograde tracing experiments resulted in a listing of 66 regions. Matrix A was created with rows representing these 66 brain regions and columns representing the 5 neuron classes found by pairing Levene’s test with unsupervised hierarchical clustering of the presubiculum data (Table S3). The average projection volume and density values for each of the 66 regions were calculated from the 7 experiments, and the averages were multiplied to populate the columns of vector b.  
+
+<|ref|>text<|/ref|><|det|>[[115, 716, 883, 855]]<|/det|>
+To obtain the highest confidence in the NNLS analysis, matrix A was sequentially “bi- normalized” first by axonal length and then by invaded region. Specifically, first each cell in matrix A was normalized so that each row summed to one. Next, each value was divided by the number of regions, 66, and multiplied by the number of clusters, 5, such that the sum of all values in matrix A equaled 5. Subsequently, each cell in matrix A was normalized so that each column summed to one. Vector b was normalized such that the sum of all values equaled to one. Finally, the squared Euclidean norm of the residual of the MATLAB function lsqnonneg was calculated as a proxy for the uncertainty of the analysis.  
+
+<|ref|>text<|/ref|><|det|>[[115, 873, 236, 889]]<|/det|>
+Soma Analysis.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 108, 883, 300]]<|/det|>
+To quantify the spatial separation among the somata among the neuron projection classes in the presubiculum, we performed a convex hull analysis for the location of the soma centers in each class using MATLAB. To create the convex hull, outliers were removed by iteratively going through all points in each class and calculating the volume of the convex hull without each point. If the volume differed by more than \(1 / \mathrm{n}\) of the volume of the original convex hull, which included all points, the point was considered an outlier and removed from the dataset. This established an algorithmic thresholding that corresponded well with the visual inspection of potential outliers. However, if removing the outliers resulted in fewer than four somata, the minimal number of points required to conduct a convex hull analysis, all points were considered. Between each pair of convex hulls, the proportion of the volume of overlap to the volume of the union of the convex hulls was used to assess the similarity between topographic locations.  
+
+<|ref|>text<|/ref|><|det|>[[115, 316, 445, 334]]<|/det|>
+Analysis of Divergence and Convergence.  
+
+<|ref|>text<|/ref|><|det|>[[115, 350, 883, 508]]<|/det|>
+Utilizing the original JSON data files, for every neuron in each presubiculum class, we measured the path distance from the soma to each axonal point in the target region. We then calculated the median path distance to each target region across all neurons in the class, and performed a Wilcoxon Signed Rank Test<sup>42</sup>, using the MATLAB function ranksum, to assess whether the path distances to each characteristic target of a particular class were significantly different. Using the same data files, we also performed a Wilcoxon Signed Rank Test to assess whether the path distances to each characteristic target between all clusters were significantly different. In both sets of comparisons, multiple testing was corrected for by False Discovery Rate to determine the significance of the resultant p- values.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[111, 90, 213, 107]]<|/det|>
+## References.  
+
+<|ref|>text<|/ref|><|det|>[[110, 120, 870, 875]]<|/det|>
+1. Fishell, G. & Kepecs, A. Interneuron Types as Attractors and Controllers. Annu Rev Neurosci 43, 1–30 (2020).  
+2. Jiang, X. et al. Principles of connectivity among morphologically defined cell types in adult neocortex. Science 350, aac9462 (2015).  
+3. DeFelipe, J. et al. New insights into the classification and nomenclature of cortical GABAergic interneurons. Nat Rev Neurosci 14, 202–216 (2013).  
+4. Sherman, S. M. & Guillery, R. W. Distinct functions for direct and transthalamic corticocortical connections. J Neurophysiol 106, 1068–1077 (2011).  
+5. Winnubst, J., Spruston, N. & Harris, J. A. Linking axon morphology to gene expression: a strategy for neuronal cell-type classification. Curr Opin Neurobiol 65, 70–76 (2020).  
+6. Ascoli, G. A. Trees of the brain, roots of the mind. (The MIT Press, 2015).  
+7. François, C., Tande, D., Yelnik, J. & Hirsch, E. C. Distribution and morphology of nigral axons projecting to the thalamus in primates. J Comp Neurol 447, 249–260 (2002).  
+8. Rochefort, N. L. et al. Functional selectivity of interhemispheric connections in cat visual cortex. Cereb Cortex 19, 2451–2465 (2009).  
+9. Rojas-Piloni, G. et al. Relationships between structure, in vivo function and long-range axonal target of cortical pyramidal tract neurons. Nat Commun 8, 870 (2017).  
+10. Ropireddy, D., Scorcioni, R., Lasher, B., Buzsáki, G. & Ascoli, G. A. Axonal morphometry of hippocampal pyramidal neurons semi-automatically reconstructed after in vivo labeling in different CA3 locations. Brain Struct Funct 216, 1–15 (2011).  
+11. DeFelipe, J. From the connectome to the synaptome: an epic love story. Science 330, 1198–1201 (2010).  
+12. Helmstaedter, M. & Mitra, P. P. Computational methods and challenges for large-scale circuit mapping. Curr Opin Neurobiol 22, 162–169 (2012).  
+13. Gao, L. et al. Single-neuron projectome of mouse prefrontal cortex. Nat Neurosci 25, 515–529 (2022).  
+14. Gong, H. et al. High-throughput dual-colour precision imaging for brain-wide connectome with cytoarchitectonic landmarks at the cellular level. Nat Commun 7, 12142 (2016).  
+15. Peng, H. et al. Morphological diversity of single neurons in molecularly defined cell types. Nature 598, 174–181 (2021).  
+16. Economo, M. N. et al. A platform for brain-wide imaging and reconstruction of individual neurons. Elife 5, e10566 (2016).  
+17. Wang, Q. et al. The Allen Mouse Brain Common Coordinate Framework: A 3D Reference Atlas. Cell 181, 936-953.e20 (2020).  
+18. Chon, U., Vanselow, D. J., Cheng, K. C. & Kim, Y. Enhanced and unified anatomical labeling for a common mouse brain atlas. Nat Commun 10, 5067 (2019).  
+19. Reimann, M. W. et al. A null model of the mouse whole-neocortex micro-connectome. Nat Commun 10, 3903 (2019).  
+20. Armañanzas, R. & Ascoli, G. A. Towards the automatic classification of neurons. Trends in Neurosciences 38, 307–318 (2015).  
+21. Winnubst, J. et al. Reconstruction of 1,000 Projection Neurons Reveals New Cell Types and Organization of Long-Range Connectivity in the Mouse Brain. Cell 179, 268-281.e13 (2019).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[60, 90, 880, 858]]<|/det|>
+451 22. Muñoz-Castañeda, R. et al. Cellular anatomy of the mouse primary motor cortex. Nature 452 598, 159- 166 (2021). 453 23. Shepherd, G. M. G. Corticostriatal connectivity and its role in disease. Nat Rev Neurosci 14, 454 278- 291 (2013). 455 24. Angelaki, D. E. & Laurens, J. The head direction cell network: attractor dynamics, 456 integration within the navigation system, and three-dimensional properties. Current Opinion 457 in Neurobiology 60, 136- 144 (2020). 458 25. Jacobs, H. I. L. et al. The presubiculum links incipient amyloid and tau pathology to memory 459 function in older persons. Neurology 94, e1916- e1928 (2020). 460 26. Preston- Ferrer, P., Coletta, S., Frey, M. & Burgalossi, A. Anatomical organization of 461 presubicular head- direction circuits. Elife 5, e14592 (2016). 462 27. Vantomme, G. et al. A Thalamic Reticular Circuit for Head Direction Cell Tuning and 463 Spatial Navigation. Cell Reports 31, 107747 (2020). 464 28. Honda, Y. & Furuta, T. Multiple Patterns of Axonal Collateralization of Single Layer III 465 Neurons of the Rat Presubiculum. Front. Neural Circuits 13, 45 (2019). 466 29. Knierim, J. J., Neunuebel, J. P. & Deshmukh, S. S. Functional correlates of the lateral and 467 medial entorhinal cortex: objects, path integration and local- global reference frames. Philos 468 Trans R Soc Lond B Biol Sci 369, 20130369 (2014). 469 30. Taube, J. S. The head direction signal: origins and sensory- motor integration. Annu Rev 470 Neurosci 30, 181- 207 (2007). 471 31. Balsamo, G. et al. Modular microcircuit organization of the presubicular head- direction map. 472 Cell Rep 39, 110684 (2022). 473 32. Wang, C., Chen, X. & Knierim, J. J. Egocentric and allocentric representations of space in 474 the rodent brain. Curr Opin Neurobiol 60, 12- 20 (2020). 475 33. Kebschull, J. M. & Zador, A. M. Cellular barcoding: lineage tracing, screening and beyond. 476 Nat Methods 15, 871- 879 (2018). 477 34. Sun, Q. et al. A whole- brain map of long- range inputs to GABAergic interneurons in the 478 mouse medial prefrontal cortex. Nat Neurosci 22, 1357- 1370 (2019). 479 35. David, K. K., Fang, H. Y., Peng, G. C. Y. & Gnadt, J. W. NIH BRAIN Circuits Programs: 480 An Experiment in Supporting Team Neuroscience. Neuron 108, 1020- 1024 (2020). 481 36. Ecker, J. R. et al. The BRAIN Initiative Cell Census Consortium: Lessons Learned toward 482 Generating a Comprehensive Brain Cell Atlas. Neuron 96, 542- 557 (2017). 483 37. Hsu, N. S. et al. The promise of the BRAIN initiative: NIH strategies for understanding 484 neural circuit function. Curr Opin Neurobiol 65, 162- 166 (2020). 485 38. Litvina, E. et al. BRAIN Initiative: Cutting- Edge Tools and Resources for the Community. J 486 Neurosci 39, 8275- 8284 (2019). 487 39. Levene, H. Robust Tests for Equality of Variances. in Contributions to Probability and 488 Statistics: Essays in Honor of Harold Hotelling 278- 292 (Stanford University Press, 1960). 489 40. Lawson, C. L. & Hanson, R. J. Solving Least Squares Problems. (Society for Industrial and 490 Applied Mathematics, 1995). 491 41. Lin, C.- J. Projected gradient methods for nonnegative matrix factorization. Neural Comput 492 19, 2756- 2779 (2007). 493 42. Wilcoxon, F. Individual comparisons by ranking methods. Biometrics Bulletin 1, 80- 83 494 (1945).
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 91, 285, 108]]<|/det|>
+## Acknowledgements.  
+
+<|ref|>text<|/ref|><|det|>[[115, 124, 828, 177]]<|/det|>
+We thank Dr. Rodrigo Muñoz- Castañeda for help with validating the mapping of neuronal reconstructions to anatomical coordinates. This work was supported in part by NIH grants R01NS39600, U01MH114829, and RF1MH128693.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 195, 300, 212]]<|/det|>
+## Author contributions.  
+
+<|ref|>text<|/ref|><|det|>[[115, 228, 877, 282]]<|/det|>
+D.W.W., S.B., S.S., and S.V. contributed to the analysis and interpretation of data, to the writing of software, and to the writing of the manuscript. G.A.A. contributed to the conception of the project, to the analysis and interpretation of data, and to the writing of the manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 299, 290, 316]]<|/det|>
+## Competing interests.  
+
+<|ref|>text<|/ref|><|det|>[[115, 333, 568, 351]]<|/det|>
+All authors declare that they have no competing interests.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 368, 365, 385]]<|/det|>
+## Materials & Correspondence.  
+
+<|ref|>text<|/ref|><|det|>[[115, 401, 841, 420]]<|/det|>
+All correspondence and material requests should be addressed to G.A.A. (ascoli@gmu.edu).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[140, 120, 875, 620]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 92, 183, 108]]<|/det|>
+<center>Figures </center>  
+
+<|ref|>table<|/ref|><|det|>[[177, 635, 820, 692]]<|/det|>
+
+<table><tr><td>Regions</td><td>AM</td><td>LM</td><td>fx</td><td>TH</td><td>MM</td><td>dhc</td><td>AV</td><td>LHA</td><td>CC</td></tr><tr><td>AA1090</td><td>159</td><td>129</td><td>84</td><td>83</td><td>78</td><td>60</td><td>42</td><td>39</td><td>35</td></tr><tr><td>AA1058</td><td>143</td><td>209</td><td>178</td><td>57</td><td>0</td><td>136</td><td>0</td><td>38</td><td>2</td></tr></table>  
+
+<|ref|>text<|/ref|><|det|>[[114, 697, 884, 823]]<|/det|>
+Figure 1. Brain- wide neuronal projections. CCF- registered reconstruction of two presubicular neurons (AA1090 in black and AA1058 in blue from the Janelia MouseLight project) invading 9 regions out of 40 potential targets along with the numbers of axonal points of the neurons in each highlighted region (posterior view of brain). CCF, common coordinate framework; AM, Anteromedial nucleus; AV, Anteroventral nucleus; cc, corpus callosum; dhc, dorsal hippocampal commissure; fx, fornix; LHA, Lateral hypothalamic area; LM, Lateral mammillary nucleus; MM, Medial mammillary nucleus; TH, other thalamic nuclei.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[125, 99, 880, 880]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[112, 877, 883, 898]]<|/det|>
+<center>Figure 2. Definitions of neuron classes and clustering methods. (A) In a single-class scenario, </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 90, 884, 352]]<|/det|>
+the distribution of differences between neurons can be calculated for all neuron pairs (pink double- arrows). (B) If two distinct classes exist, neurons (represented here as black dots) will tend to have more similar projections within their class (red double- arrows) and more different ones across classes (blue double arrow). (C) The differences within the classes (red distribution) will be smaller than those between classes (blue distribution). (D) The distribution of the combined frequency of differences, in a multi- class scenario (red- blue stacked areas; green half- height width), will be wider than that of a single- class distribution (pink curve; orange half- height width). (E) Diagram showing the randomization of projection patterns through the repeated pairwise swapping of axonal point counts between two neurons across two of their potential target regions, which preserves the column (for a given region) and row (for a given neuron) sums of the matrix. This swapping results in a projection pattern “continuum” that matches with the overall distribution that represents the 1- class null hypothesis. (F) Unsupervised hierarchical clustering groups a set of neurons into classes based on their relative pairwise differences or similarities, as modeled by a binary dendrogram. The top (root) of the dendrogram represents all neurons lumped into the same class, while the bottom (leaves) shows all neurons split into separate classes.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[112, 80, 884, 875]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[112, 876, 884, 896]]<|/det|>
+<center>Figure 3. Primary motor cortex L6 (IT vs. CT). (A) Representation of the two clusters produced </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 90, 883, 234]]<|/det|>
+by Levene's one- tailed test for the equality of variances and unsupervised hierarchical clustering, using MouseLight neurons from the primary motor cortex, layer 6 (n = 52). (B) Colormap of the axonal distributions of neurons (columns) across anatomical regions (rows), with darker shades representing more axonal projections. (C) Axonal pathways of representative IT (intratelencephalic, red) and CT (corticothalamic, blue) neurons with semitransparent surfaces of primary motor cortex layer 6 (green) and selected thalamic nuclei (pink). The two black dots indicate the cell body locations of the two representative cells from each class. (D) Axonal pathways of all IT and CT neurons in the MouseLight sample (same color coding).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[110, 88, 888, 768]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 770, 884, 893]]<|/det|>
+<center>Figure 4. Classification of projection neuron types in the presubiculum. (A) Representation of 5 axonal clusters produced by Levene’s test and unsupervised hierarchical clustering of neurons from the presubiculum (n = 93). (B) Colormap of the axonal distributions of neurons (columns) across anatomical regions (rows), with darker shades representing more axonal projections. Parcel names highlighted in pink are hypothalamus related. Parcel names highlighted in yellow and light blue are thalamus related. (C) Neuron-to-target assignments for the identified axonal projection classes and corresponding anatomical regions (dotted line: contralateral). (D) Anterior view of the </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[60, 90, 884, 325]]<|/det|>
+mouse brain with one neuron from each class. Color coding of neurons and semitransparent anatomical areas shown in A, B, and C. (E) Posterior view of the brain with all MouseLight presubicular neurons. \(CA3 + CA1\) : Cornu Ammonis areas 3 and 1; DG: dentate gyrus; Sub: subiculum; LEC: lateral entorhinal cortex; dMEC: dorsal portion of the medial entorhinal cortex; ParaS: parasubiculum; PostS: postsubiculum; Retrohippocampal region; DV(gr.)RtSpl: dorsal and ventral (granular) retrosplenial cortex; L(ag.)RtSpl: lateral (agranular) retrosplenial cortex; MidB: midbrain; Hyp: hypothalamus; PMdv+TU: dorsal and ventral premammillary nucleus and tuberal nucleus; MM+LZ: medial mammillary nucleus and hypothalamic lateral zone; MBO+LM: mammillary body and lateral mammillary nucleus; mATN+PT: medial anterior thalamic nucleus and parataenial nucleus; TH+LGN: thalamus and lateral geniculate nucleus; dvATN+MGN: dorsal and ventral anterior thalamic nucleus and medial geniculate nucleus; IAD+IAM: interanterodorsal and interanteromedial nucleus of the thalamus; LD+AD: lateral dorsal and anterodorsal nucleus of thalamus.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[120, 100, 866, 880]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[112, 877, 881, 899]]<|/det|>
+<center>Figure 5. Spatial distributions of somata in the presubiculum across projection classes. </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[61, 90, 884, 220]]<|/det|>
+654 Convex hulls of neurons (spheres) from classes A38 (blue), B27 (red), D19 (brown), and E6 655 (green), and semitransparent presubiculum (green). (A) Left sagittal view of A38 and B27. (B) 656 Layer 1 (green), layer 2 (purple), and layer 3 (orange) of the presubiculum are highlighted in an 657 anterior coronal view, with somata from A38 in blue and D19 in brown. Most of the A38 somata 658 are concentrated in layer 2, while the D19 somata tend to be more concentrated in layers 1 and 3. 659 Somata that do not follow this pattern are indicated with a white dot inside of the circle. (C) Left 660 sagittal view of D19 and E6. (D) Posterior coronal view of B27 and E6. 661 662
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[113, 87, 883, 877]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[112, 876, 883, 896]]<|/det|>
+<center>Figure 6. Divergent path distance comparison from one neuron class in the presubiculum to </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 90, 883, 370]]<|/det|>
+665 its targets. (A) Left: box and whisker plot depicting the median, first and third quartiles, and full 666 range of the path distances from class A38 to its major ipsilateral (I) and contralateral (C) targets. 667 Right: the path distance of an archetype neuron from class A38 (light blue), from its soma (black) 668 in the ipsilateral presubiculum (green) to the subiculum (purple), is significantly shorter than that 669 (dark blue) to the lateral entorhinal cortex (orange). (B) Left: box and whisker plot depicting the 670 distributions of path distances from class B27 to its major ipsilateral and contralateral targets. 671 Right: the path distance of an archetype neuron from class B27 (light red), from its soma (black) 672 in the ipsilateral presubiculum (green) to the parasubiculum (brown), is significantly shorter than 673 that (dark red) to the medial entorhinal cortex, dorsal zone (cyan). (C) Left: box and whisker plot 674 depicting the path distances from class D19 to its major ipsilateral targets. Right: the path distance 675 of an archetype neuron from class D19 (light brown), from its soma (black) in the ipsilateral 676 presubiculum (green) to the midbrain (magenta), is significantly shorter than that (dark brown) to 677 the hypothalamus and lateral mammillary nucleus (red). See Figure 4 for abbreviation definitions. 678 Significant differences in distances were calculated using a Wilcoxon Signed Rank Test performed 679 on neuronal path distances and multiple testing was corrected for by False Discovery Rate to 680 determine the significance of the resultant p- values.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[111, 90, 872, 731]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 734, 884, 893]]<|/det|>
+<center>Figure 7. Convergent path distance comparison from each presubiculum cluster to major targets. (A) Top left: box and whisker plot depicting the median, first and third quartiles, and full range of the path distances from neurons in ipsilateral (I) and contralateral (C) classes to the medial entorhinal cortex, dorsal zone. Bottom: the distance of an archetype neuron from class B27 (red), from its soma in the ipsilateral presubiculum (green) to the dMEC (purple), is significantly longer than the comparable distance of an archetype neuron from class D19 (brown). (B) Top left: box and whisker plot depicting the path distances from neurons in ipsilateral and contralateral classes to the parasubiculum. Bottom: the distance of an archetype neuron from class B27 (red), from its soma in the presubiculum (green) to the ipsilateral ParaS (purple), is significantly shorter than the </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[57, 90, 883, 230]]<|/det|>
+729 comparable distance of an archetype neuron from class D19 (brown). (C) Top left: box and whisker 730 plot of the path distances from neurons in contralateral classes to the subiculum. Bottom: the 731 distance of an archetype neuron from class A38 (blue), from its soma in the presubiculum (green) 732 to the contralateral Sub (purple), is significantly shorter than the comparable distance of an 733 archetype neuron from class B27 (red). See Figure 4 for abbreviation definitions. Significant 734 differences in distances were calculated using a Wilcoxon Signed Rank Test performed on 735 neuronal path distances and multiple testing was corrected for by False Discovery Rate to 736 determine the significance of the resultant p-values.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[57, 90, 630, 211]]<|/det|>
+Supplemental TablesTable S1. Raw axonal counts for primary motor area layer 6. Table S2. Raw axonal counts for presubiculum. Table S3. Non- negative least- square normalizations.
+
+<--- Page Split --->

preprint/preprint__10394e4b9f4446a1b17a9ba94de8cd766e54541a93fbbde182594cfbc2bc4d37/images_list.json

@@ -0,0 +1,62 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Fig. 1 | Coherent consolidation of nucleations and a mono-atom step-level flat surface. a, Series of schematic diagrams of the structural evolution from quasi-zero-dimensional \\(\\mathrm{Cu}(111)\\) nanodroplets to a quasi-two-dimensional single-crystal Cu film in three stages: (I) nucleation and lateral growth, (II) coherent coplanar merging, and (III) layer-by-layer growth. b, c, Cross-sectional high-resolution transmission electron microscopy (HRTEM) image of a perfectly flat surface (left) and strain field map by geometrical phase analysis (GPA) (right), observed in the \\([1\\bar{1} 0]\\) orientation. d, e, Cross-sectional HRTEM image of the interfacial region of the \\(\\mathrm{Cu / Al_2O_3}\\) heterostructure and its strain field map by GPA, with an orientation relationship of \\((111)_{\\mathrm{Cu}}[1\\bar{1} 0]_{\\mathrm{Cu}} / (001)_{\\mathrm{Al_2O_3}}[1\\bar{1} 0]_{\\mathrm{Al_2O_3}}\\) .",
+    "footnote": [],
+    "bbox": [
+      [
+        120,
+        100,
+        840,
+        525
+      ]
+    ],
+    "page_idx": 5
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Fig. 2 | Coherent consolidation process for a mono-atom step-level flat surface. a, Cross-sectional bright-field scanning transmission electron microscopy images. b, Corresponding illustrations. c, Topographic atomic force microscopy images acquired from samples of five different thicknesses, corresponding to initial deposition times of 15, 30, 45, 60 and 120 s. d, MSFS of the single-crystal Cu film with a root-mean-square (RMS) roughness of \\(\\sim 0.2 \\mathrm{nm}\\) . e, Schematic side view of d.",
+    "footnote": [],
+    "bbox": [
+      [
+        125,
+        87,
+        870,
+        455
+      ]
+    ],
+    "page_idx": 6
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Fig. 3 | Theoretical approach for coherent consolidation of nucleations into a coplanar layer. a, Scheme for the growth of a single nanodroplet via atomic diffusion and b, high-resolution scanning transmission electron microscopy image and fast Fourier transform pattern (inset) of a Cu nanodroplet oriented along the \\([1\\bar{1} 0]\\) zone axis, indicating single-crystal nanodroplet growth along the single-crystal [111] direction. c, Relative energy profiles for the diffusion of Cu atoms or atom clusters (up to four atoms) on an \\(\\mathrm{Al}_2\\mathrm{O}_3(0001)\\) substrate. Schemes for d, lateral nanodroplet growth and e, coherent merging to coplanar layers.",
+    "footnote": [],
+    "bbox": [
+      [
+        130,
+        88,
+        850,
+        565
+      ]
+    ],
+    "page_idx": 8
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Fig. 4 | Three phases of the initial stages of film growth: nucleation, coherent merging, and single-crystal thin film formation. a, Thickness-dependent resistivity \\((\\rho)\\) of SC films according to direct current electrical transport and optical Fourier transform-infrared spectroscopy measurements, as well as data from previous reports<sup>28,34</sup>. Stages I and II are distinguished by the divergence of resistivity (or zero conductivity) at a film thickness \\((d)\\) close to the nanodroplet size. Stages II and III are distinguished by the resistivity trend determined according to the resistivity slope change in \\(\\rho\\) as a function of \\(1 / d\\) (inset). b–e, Corresponding electron backscatter diffraction images of Cu thin films in (b) stage I, (c, d) stage II and (e) stage III.",
+    "footnote": [],
+    "bbox": [
+      [
+        137,
+        95,
+        833,
+        625
+      ]
+    ],
+    "page_idx": 11
+  }
+]

preprint/preprint__10394e4b9f4446a1b17a9ba94de8cd766e54541a93fbbde182594cfbc2bc4d37/preprint__10394e4b9f4446a1b17a9ba94de8cd766e54541a93fbbde182594cfbc2bc4d37.mmd

@@ -0,0 +1,260 @@
+
+# Coherent consolidation of trillions of nucleations for mono-atom step-level flat surfaces  
+
+Taewoo Ha  Institute for Basic Science  
+
+Yu- Seong Seo  Sungkyunkwan University  
+
+Teun- Teun Kim  University of Ulsan  
+
+Bipin Lamichhane  Mississippi State University  https://orcid.org/0000- 0002- 9503- 083X  
+
+Young- Hoon Kim  Department of Energy Science, Sungkyunkwan University  https://orcid.org/0000- 0001- 7343- 1512  
+
+Su Jae Kim  Pusan National University  https://orcid.org/0000- 0002- 6033- 8551  
+
+Yousil Lee  Pusan National University  
+
+Jong Kim  UNIST  https://orcid.org/0000- 0003- 4101- 0590  
+
+Sang Eon Park  Pusan National University  
+
+Kyung Ik Sim  Institute for Basic Science  https://orcid.org/0000- 0002- 9870- 3913  
+
+Jae Kim  Yonsei University  https://orcid.org/0000- 0002- 7840- 3630  
+
+Yong In Kim  Sungkyunkwan University  https://orcid.org/0000- 0002- 9540- 5996  
+
+Seon Kim  Sungkyunkwan University  
+
+Hu Young Jeong  Ulsan National Institute of Science and Technology  https://orcid.org/0000- 0002- 5550- 5298  
+
+Young Hee Lee  
+
+Center for Integrated Nanostructure Physics (CINAP), Institute for Basic Science (IBS), Sungkyunkwan University  https://orcid.org/0000- 0001- 7403- 8157  
+
+Seong- Gon Kim  
+
+Mississippi State University  https://orcid.org/0000- 0002- 1629- 0319
+
+<--- Page Split --->
+
+Young- Min Kim Sungkyunkwan University https://orcid.org/0000- 0003- 3220- 9004 Jungseek Hwang Sungkyunkwan University https://orcid.org/0000- 0002- 2555- 1218 Se- Young Jeong ( syjeong@pusan.ac.kr ) Pusan National University https://orcid.org/0000- 0003- 1019- 4403  
+
+## Article  
+
+Keywords: Coherent merging, coplanar layer, initial growth stages, mono- atom step- level flat surface, single- crystal thin films  
+
+Posted Date: September 16th, 2022  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 2018252/v1  
+
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License
+
+<--- Page Split --->
+
+# Coherent consolidation of trillions of nucleations for mono-atom step-level flat surfaces  
+
+Taewoo Ha \(^{1\dagger}\) , Yu-Seong Seo \(^{2\dagger}\) , Teun-Teun Kim \(^{3\dagger}\) , Bipin Lamichhane \(^{4}\) , Young-Hoon Kim \(^{5}\) , Su Jae Kim \(^{6}\) , Yousil Lee \(^{6}\) , Jong Chan Kim \(^{7}\) , Sang Eon Park \(^{6}\) , Kyung Ik Sim \(^{1,8}\) , Jae Hoon Kim \(^{8}\) , Yong In Kim \(^{5}\) , Seon Je Kim \(^{5}\) , Hu Young Jeong \(^{7,9}\) , Young Hee Lee \(^{1,5}\) , Seong-Gon Kim \(^{4}\) , Young-Min Kim \(^{1,5,*}\) , Jungseek Hwang \(^{2,*}\) & Se-Young Jeong \(^{10,11,*}\)  
+
+\(^{1}\) Center for Integrated Nanostructure Physics, Institute for Basic Science, Sungkyunkwan University, Suwon 16419, Republic of Korea. \(^{2}\) Department of Physics, Sungkyunkwan University, Suwon 16419, Republic of Korea. \(^{3}\) Department of Physics, University of Ulsan, Ulsan 44610, Republic of Korea. \(^{4}\) Department of Physics and Astronomy, Mississippi State University, Mississippi State, MS 39762, USA. \(^{5}\) Department of Energy Science, Sungkyunkwan University, Suwon 16419, Republic of Korea. \(^{6}\) Crystal Bank Research Institute, Pusan National University, Busan 46241, Republic of Korea. \(^{7}\) School of Materials Science and Engineering, Ulsan National Institute of Science and Engineering, Ulsan 44919, Republic of Korea. \(^{8}\) Department of Physics, Yonsei University, Seoul 03722, Republic of Korea. \(^{9}\) UNIST Central Research Facilities, Ulsan National Institute of Science and Technology, Ulsan 44919, Republic of Korea. \(^{10}\) Department of Cogno-Mechatronics Engineering, Pusan National University, Busan 46241, Republic of Korea. \(^{11}\) Department of Optics and Mechatronics Engineering, Pusan National University, Busan 46241, Republic of Korea. \(^{*}\) Corresponding authors. E- mail: youngmk@skku.edu (Y.- M. Kim); jungseek@skku.edu (J. Hwang); syjeong@pusan.ac.kr (S.- Y. Jeong) \(^{\dagger}\) These authors contributed equally to this work.  
+
+\(^{26}\) E- mail: youngmk@skku.edu (Y.- M. Kim); jungseek@skku.edu (J. Hwang); syjeong@pusan.ac.kr (S.- Y. Jeong) \(^{\dagger}\) These authors contributed equally to this work.
+
+<--- Page Split --->
+
+Constructing a mono- atom step- level ultra- flat material surface is challenging, especially for thin films, because it is prohibitively difficult for trillions of clusters to coherently merge. Even though a rough metal surface, as well as the scattering of carriers at grain boundaries, limits electron transport and obscures their intrinsic properties, the importance of the flat surface has not been emphasised sufficiently. In this study, we describe in detail the initial growth of copper thin films required for mono- atom step- level flat surfaces (MSFSs). Deposition using atomic sputtering epitaxy leads to the coherent merging of trillions of islands into a coplanar layer, eventually forming an MSFS, for which the key factor is suggested to be the individual deposition of single atoms. Theoretical calculations support that single sputtered atoms ensure the formation of highly aligned nanodroplets and help them to merge into a coplanar layer. The realisation of the ultra- flat surfaces is expected to greatly assist efforts to improve quantum behaviour by increasing the coherency of electrons.  
+
+Keywords: Coherent merging; coplanar layer; initial growth stages; mono- atom step- level flat surface; single- crystal thin films  
+
+Ultrathin metal films are indispensable in modern electronics and nanotechnology<sup>1- 3</sup>. During the past few decades, conventional metals have been studied extensively because the performance of metal- based devices is intimately related to their physical properties.<sup>4- 6</sup> Efforts to produce single- crystal (SC) copper (Cu) from Cu foil have been driven by competition<sup>7,8</sup> and interest in their nanocrystalline nature and potential applications in large two- dimensional (2D) components consisting of materials such as graphene and hexagonal boron nitride (h- BN).<sup>9- 11</sup> However, despite the importance of metal thin- film flatness, there have been few reports because it is challenging to control flatness. Since the contact between the metal electrode and the semiconductor material decisively affects the properties of electronic and optoelectronic devices,<sup>12,13</sup> a flat metal surface is proposed as a good solution to reduce contact resistance. The motion of electrons without scattering at surfaces and grain boundaries can also affect the carrier transport properties.<sup>14</sup> Single- crystal Cu thin films (SCCFs) on sapphire have recently been reported, and the formation of twin boundaries (TBs) have been investigated intensively.<sup>15- 17</sup> Two orientations (ORs) adjacent to a TB are rotated by a certain angle in- plane and satisfy the symmetry operation exactly, while two ORs adjacent to a grain boundary (GB) are tilted both in out- of- plane and in- plane directions. However, in view of electronic motion, a much clearer distinction between TB and GB must be followed by precise microscopic analysis. Twin
+
+<--- Page Split --->
+
+boundaries are not strong electron scatterers, \(^{18}\) while electrons are scattered at GBs if they have an in-plane mistilt even by \(1^{\circ}\) , even if ORs adjacent to GBs are aligned almost perfectly along the out-of-plane direction. The number of TBs produced near the interface gradually decreases and may disappear as the position approaches the surface. \(^{15}\) Nevertheless, for the grown thin film to lead to an ultra- smooth surface, it is necessary to discuss the initial growth process precisely. A study using scanning tunnelling microscopy demonstrated that nanocrystalline Cu films cannot be flat because valleys and ridges are created by out- of- plane grain rotation. \(^{19,20}\) A recent study revealed that thin Cu films fabricated on four- inch- diameter wafers using atomic sputtering epitaxy (ASE) have atomically flat surfaces overall, with occasional mono- atomic step edges, and the crystal quality is maintained even after a few years. \(^{21}\)  
+
+In this study, we describe in detail the initial growth phases of ultrathin Cu films prepared using ASE (Methods and Supplementary Fig. 1a). Through transmission electron microscopy (TEM) observations, we identified the structural evolution from quasi- zero- dimensional (0D) nanodroplets to quasi- 2D thin films in the initial growth stage. It was demonstrated experimentally and theoretically that the only way to consolidate the trillions of nucleations and to achieve a mono- atom step- level flat surface (MSFS) is by the individual deposition of single atoms, when considering long- range period lattice mismatch between Cu and substrate \(\mathrm{Al}_2\mathrm{O}_3\) and the formation of GB and TB.  
+
+## Growth of an ultrathin SCCF  
+
+Three approaches are typically used to grow thin films on a single- crystal substrate, i.e. the Volmer- Weber, Frank- van der Merwe and Stranski- Krastanov methods. \(^{22}\) In contrast, our ASE approach follows a new growth mode. The ideal initial growth of Cu on \(\mathrm{Al}_2\mathrm{O}_3\) substrate follows three stages (Fig. 1a): stage I, nanodroplet nucleation and lateral growth; stage II, coherent merging; and stage III, layer- by- layer growth of an SCCF. For successful completion of this process, each of trillions of nanodroplets must consist of a single- crystal, and all must be aligned in the same direction. A high- resolution TEM image of a perfectly flat surface of a 12- nm- thick film and its strain field map obtained by geometrical phase analysis, respectively, is shown in Fig. 1b and c. The Cu thin films thicker than 11.5 nm are aligned perfectly along the (111) plane, which is supported by X- ray diffraction (XRD), atomic force microscopy (AFM), electron backscatter diffraction (EBSD) mapping and scanning electron microscopy (SEM) and TEM images (Supplementary Fig. 1b- i). \(^{21}\) The interface between Cu and the \(\mathrm{Al}_2\mathrm{O}_3\) substrate was defect- and strain- free after growth of the 12- nm- thick film (Fig. 1d, e).
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Fig. 1 | Coherent consolidation of nucleations and a mono-atom step-level flat surface. a, Series of schematic diagrams of the structural evolution from quasi-zero-dimensional \(\mathrm{Cu}(111)\) nanodroplets to a quasi-two-dimensional single-crystal Cu film in three stages: (I) nucleation and lateral growth, (II) coherent coplanar merging, and (III) layer-by-layer growth. b, c, Cross-sectional high-resolution transmission electron microscopy (HRTEM) image of a perfectly flat surface (left) and strain field map by geometrical phase analysis (GPA) (right), observed in the \([1\bar{1} 0]\) orientation. d, e, Cross-sectional HRTEM image of the interfacial region of the \(\mathrm{Cu / Al_2O_3}\) heterostructure and its strain field map by GPA, with an orientation relationship of \((111)_{\mathrm{Cu}}[1\bar{1} 0]_{\mathrm{Cu}} / (001)_{\mathrm{Al_2O_3}}[1\bar{1} 0]_{\mathrm{Al_2O_3}}\) . </center>  
+
+## Coherent consolidation into a coplanar layer  
+
+The detailed states of early growth are shown in Fig. 2. Figure 2 shows a series of cross- sectional bright- field scanning TEM (BF- STEM) images<sup>23</sup> acquired from samples of various thicknesses (Fig. 2a) along with corresponding illustrations (Fig. 2b) and topographic images obtained by atomic force microscopy (AFM) (Fig. 2c).
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Fig. 2 | Coherent consolidation process for a mono-atom step-level flat surface. a, Cross-sectional bright-field scanning transmission electron microscopy images. b, Corresponding illustrations. c, Topographic atomic force microscopy images acquired from samples of five different thicknesses, corresponding to initial deposition times of 15, 30, 45, 60 and 120 s. d, MSFS of the single-crystal Cu film with a root-mean-square (RMS) roughness of \(\sim 0.2 \mathrm{nm}\) . e, Schematic side view of d. </center>  
+
+Two separate initial stages are apparent in Fig. 2, divided according to a thickness threshold of \(\sim 5 \mathrm{nm}\) . During stage I (Fig. 2a–c, first three panels), the Cu atoms form Cu(111) nanodroplets with single crystallinity; these nucleate at distances of \(20–30 \mathrm{nm}\) and consist of 18–31 layers (3–5 nm) on average, with a small height distribution (Supplementary Fig. 2). Because the area of an initial nanodroplet ranges from approximately 30 to \(100 \mathrm{nm}^2\) , the total number of nanodroplets on a 2-inch-diameter wafer is approximately \(\sim 10^{12}\) . During stage II, (Fig. 2a–c, fourth panels), Cu(111) nanodroplets begin to form conduction channels through Cu(111) lateral growth and coherent coplanar merging. The height of the nanodroplets increases more slowly (i.e. by a few layers) than the expected average deposition rate (i.e. \(\sim 7\) layers per 15 s of initial deposition time), indicating predominant lateral growth. AFM images of \(< 5 \mathrm{nm}\) - thick films (Fig. 2c, first three images) obtained at a \(1.0 \mathrm{nm}\) resolution (for an area of \(100 \times 100 \mathrm{nm}^2\) ) show the growth of islands separated by \(20–30 \mathrm{nm}\) , whereas thin films with a thickness of \(\geq 5 \mathrm{nm}\) (Fig. 2c, fourth image) obtained at a \(1.5 \mathrm{nm}\) resolution for a much larger area \((10 \times 10 \mu \mathrm{m}^2)\) exhibit a mainly flat surface. Remarkably, thin films thicker than \(10 \mathrm{nm}\) showed an atomically
+
+<--- Page Split --->
+
+flat surface with an exceptionally small root mean square (RMS) roughness of \(< 0.2 \mathrm{nm}\) . A thin film with RMS roughness of \(< 0.3 \mathrm{nm}\) has an atomically flat surface with occasional monoatomic step edges. A cross-sectional HRTEM image (Fig. 2d) and its graphical illustration (Fig. 2e) show the MSFS of the SCCF with an RMS roughness of \(\sim 0.2 \mathrm{nm}\) . The fourth and fifth AFM images in Fig. 2c provide experimental evidence for the illustration in Fig. 2e and show that the thin film grows to an atomically flat surface beyond a thickness of \(\sim 10 \mathrm{nm}\) . AFM images and RMS roughness as a function of thicknesses, EBSD and inverse pole figure (IPF), SEM images and XRD data for 12 Cu films exhibiting marked Pendellösung oscillations (thickness fringes) \(^{24}\) also supported the quality of these films (Supplementary Fig. 3). The estimated thicknesses of all 12 thin-film samples obtained via AFM are listed in Supplementary Table 1. Twin boundaries or GBs in stage II critically affect the final roughness of the surface. The regions in a polycrystal have random ORs, which are separated mostly by GBs and sometimes by TBs (Supplementary Fig. 4a), whereas regions in an SCCF have only two ORs and are separated by only TBs (Supplementary Fig. 4b). The two ORs are associated with two different stacking orders, i.e. ABCABC... and ACBACB..., and are separated by TBs with a closed path. Two different ORs in the SCCF must be rotated exactly by \(60^{\circ}\) to each other in-plane. The boundaries marked in blue (Supplementary Fig. 4b) resemble TBs but are GBs because they have rotational components that deviate slightly from \(60^{\circ}\) in-plane. \(^{15}\) Thus, it is not appropriate to identify TBs or GBs using an optical microscope or micrometre-scale EBSD map. Therefore, it is necessary to perform a misorientation line analysis at the nanoscale. Small ORs that are not observed at the micrometre scale are frequently observed at the nanometre scale. Two regions separated by a TB merge into a larger region of approximately \(5 - 6 \mathrm{nm}\) , and two enlarged regions double into a single OR at \(\sim 12 \mathrm{nm}\) . With increasing thin-film thickness, the region of a single OR doubles every \(5 - 6 \mathrm{nm}\) similarly, and eventually (when the thickness reaches \(\sim 80 \mathrm{nm}\) ) the number of TBs in the upper part of the thin film is significantly reduced (Supplementary Fig. 5). Ideally, a single-crystal thin film, such as that grown homoepitaxially, can be obtained via heteroepitaxy near the surface when the thickness exceeds \(\sim 80 \mathrm{nm}\) . Homoepitaxy-like heteroepitaxy of a Cu thin film on a hetero substrate is only possible when it is deposited by ASE, and the long-distance periodicity by the calculation of extended atomic distance mismatch \(^{21}\) is considered.
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Fig. 3 | Theoretical approach for coherent consolidation of nucleations into a coplanar layer. a, Scheme for the growth of a single nanodroplet via atomic diffusion and b, high-resolution scanning transmission electron microscopy image and fast Fourier transform pattern (inset) of a Cu nanodroplet oriented along the \([1\bar{1} 0]\) zone axis, indicating single-crystal nanodroplet growth along the single-crystal [111] direction. c, Relative energy profiles for the diffusion of Cu atoms or atom clusters (up to four atoms) on an \(\mathrm{Al}_2\mathrm{O}_3(0001)\) substrate. Schemes for d, lateral nanodroplet growth and e, coherent merging to coplanar layers. </center>  
+
+## Individual deposition of single atoms toward MSFS  
+
+Copper forms SC nanodroplets even during the early growth stage ( \(< 10 \mathrm{~s}\) ) (Figs. 2a and 3a, b). In this study, the droplet surfaces were composed of well- defined crystallographic facets (Fig. 3b). The nanodroplet FFT patterns indicated that the structure consists of a single phase of Cu (Fig. 3b, inset). These nanodroplets grow along the [111] direction of the \(\mathrm{Al}_2\mathrm{O}_3(0001)\) substrate
+
+<--- Page Split --->
+
+in one of two stacking orders (ABCABC... or ACBACB...). The height distribution of the nanodroplets was narrow; beyond a critical height, these droplets stop growing vertically and grow only laterally (Supplementary Fig. 2f). We used a simple model calculation to develop a new thin film growth mode that explains the conditions under which transition of the growth mechanism occurs (Methods). Our model showed that the evolution of the thin film growth mechanism depends on the relative strength of the surface tension and adhesion energy. Initially, deposition of Cu atoms occurs on the Al-terminated surface of the \(\mathrm{Al}_2\mathrm{O}_3(0001)\) substrate, which had a much lower surface tension \((1.59\mathrm{J / m^2})\) than the O-terminated surface \((4.26\mathrm{J / m^2})\) (Supplementary Fig. 6a and Supplementary Table S2). The adhesion energy of the Al-terminated surface was \(- 0.68\mathrm{J / m^2}\) , which was lower than the surface tension of \(\mathrm{Cu}(111)\) \((\tau_{\perp} = 1.34\mathrm{J / m^2})\) . Therefore, the requirement for a positive aspect ratio \((E_{a}< 2\tau_{\perp}\) , Methods) was met, and droplets began to grow spherically at a large aspect ratio \((R)\) of 0.63 (stage I). As the droplet grows, its bottom side flattens, and O atoms are incorporated into the interface between the Cu droplet and \(\mathrm{Al}_2\mathrm{O}_3(0001)\) substrate surface because the interfacial energy of the O-terminated surface is lower than that of the Al-terminated surface (Supplementary Fig. 5b and Supplementary Table S2). The transition of the interface from the Al- to O-terminated surface of the \(\mathrm{Al}_2\mathrm{O}_3(0001)\) substrate increased adhesion to the substrate because the adhesion energy of the fully developed interface of \(\mathrm{Cu}(111)\) on the O-terminated surface was \(4.81\mathrm{J / m^2}\) . When the adhesion became sufficiently strong \((E_{a} > 2\tau_{\perp}\) ; i.e. at the critical height), apparent vertical growth was halted (i.e. \(R\rightarrow 0\) ), and only lateral growth occurred (stage II, Fig. 3c). For SCCFs, the critical height was found to be \(5–6\mathrm{nm}\) (Supplementary Fig. 2f). Because all islands grew along the (111) direction in one of two possible ORs associated with stacking order, two nanodroplets with the same stacking sequence merged to a larger single-crystal droplet with a coplanar layer. However, when two nanodroplets with different ORs were merged, they became a larger droplet with a TB, which retains three-fold rotational symmetry.  
+
+The number of these separate nanodroplets depends on the diffusion rate of Cu atoms on the substrate. If the diffusion rate is high, then Cu atoms deposited on the substrate move over a large distance before coalescing with existing nanodroplets, forming fewer separate nanodroplet "islands". If the diffusion rate is low, then the Cu atoms have a shorter range of motion, leading to the formation of more independent islands. The diffusion rate of the Cu atoms depends strongly on whether they are deposited on the substrate as single atoms or clusters of multiple atoms. The relative energy profiles of Cu atom diffusion onto an \(\mathrm{Al}_2\mathrm{O}_3(0001)\) substrate are shown in Fig. 3c, as obtained from first-principles calculations. The activation energies of clusters of multiple atoms were higher than those of single atoms;
+
+<--- Page Split --->
+
+notably, the diffusion of clusters of two atoms required the most energy. This finding indicates that the deposition of single atoms during the early stage of nanodroplet nucleation is critical for the growth of an atomically flat SC thin film with uniform OR. Deposits of large clusters are likely to develop into a polycrystalline structure upon merging due to their random ORs. Although this requirement is challenging to fulfil in conventional sputtering systems, where Cu atoms are ejected and deposited as clusters of multiple atoms, the present ASE system meets this requirement.  
+
+To establish how much the sizes of the species falling off the target surfaces during the sputtering process differ between the conventional system and ASE system, we compared the surfaces of two targets of the general sputtering system and the ASE system after the sputtering process (Supplementary Fig. 7). While the target surface of the general sputtering system was very rough, with an average RMS roughness of 100 nm (Supplementary Fig. 7a), the target surface of the ASE system had a smooth surface, with an average RMS roughness of 4 nm (Supplementary Fig. 7b). Optical images (Supplementary Fig. 7c, d) and AFM surface images at different scales for both targets (Supplementary Fig. 7e- h) showed the critical differences. These results suggest that the difference in the size of the sputtered species during the sputtering process greatly influences thin film growth.  
+
+Once such a coplanar layer forms, the next layer is highly likely to grow in the same stacking order. After coherent merging, adhesion (with the layer of the adsorbate) became more dominant, and only layer growth occurred (stage III, Fig. 3e). Films thicker than 10 nm showed an ultra- flat, undistorted surface without multi- atomic step edges or grain boundaries (Fig. 1b).<sup>26</sup> In a highly stable thin film growth system, where adsorbate atoms are deposited individually as single atoms, growth strictly follows the energetics of the surface tension and adhesion energy. The growth of GB- free homoepitaxy- like thin films with MSFS is not only possible for Cu, but also for other metals such as Ag, Al, and Ni, if the principle of ASE is maintained.
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Fig. 4 | Three phases of the initial stages of film growth: nucleation, coherent merging, and single-crystal thin film formation. a, Thickness-dependent resistivity \((\rho)\) of SC films according to direct current electrical transport and optical Fourier transform-infrared spectroscopy measurements, as well as data from previous reports<sup>28,34</sup>. Stages I and II are distinguished by the divergence of resistivity (or zero conductivity) at a film thickness \((d)\) close to the nanodroplet size. Stages II and III are distinguished by the resistivity trend determined according to the resistivity slope change in \(\rho\) as a function of \(1 / d\) (inset). b–e, Corresponding electron backscatter diffraction images of Cu thin films in (b) stage I, (c, d) stage II and (e) stage III. </center>
+
+<--- Page Split --->
+
+## Initial growth stages  
+
+Three distinct stages of initial film growth are also revealed by thickness- dependent DC resistivity \((\rho)\) data (Fig. 4) obtained by DC electrical transport, \(^{27,28}\) Fourier transform- infrared spectroscopy (FTIR) and time- domain terahertz spectroscopy (Supplementary Fig. 8). The transport data showed excellent agreement among the three methods. The DC resistivity \((\rho)\) was close to the bulk value at thicknesses greater than \(\sim 11.5 \mathrm{nm}\) , which was confirmed by plotting \(\rho\) as a function of the inverse of the film thickness (Fig. 4e, inset). Abrupt divergence near a thickness of \(\sim 4.6 \mathrm{nm}\) indicated that the film was not conducting at lower thickness values. At thicknesses greater than \(\sim 4.6 \mathrm{nm}\) , the films had a finite \(\rho\) , indicating conduction channel formation. The results of analytical calculations using the Fuchs- Sondheimer (FS), \(^{29,30}\) Namba, \(^{31}\) and effective medium approximation (EMA) \(^{32}\) models are shown in Supplementary Fig. 9. The \(\rho\) values of films thicker than \(11.5 \mathrm{nm}\) were in good agreement with the FS model, \(^{33}\) whereas those of films thinner than \(11.5 \mathrm{nm}\) deviated from the FS model. The Namba model, which considers surface roughness, matched the data for a surface roughness of \(\sim 4.5 \mathrm{nm}\) , but could not explain \(\rho\) in stage I ( \(< 4.5 \mathrm{nm}\) ), in which Cu nanodroplets did not form conduction channels. The EMA method, which correctly predicts percolation for spherical grains, also described stages I and II well. The results indicate that the film does not occupy the full volume from the substrate surface to the film thickness during the initial growth stage (Supplementary Fig. 10). No reliable EBSDmaps were obtained from \(< 5 - \mathrm{nm}\) - thick films (i.e., stage I) (Fig. 4a), whereas thin films showed partial coverage at thicknesses between 5 and \(11.5 \mathrm{nm}\) (Fig. 4b) and complete coverage at \(>11.5 \mathrm{nm}\) (Fig. 4c). The EBSD map shown in Fig. 4c is indistinguishable from primary blue on the red- green- blue colour scale (0:0:255), indicating exact alignment along (111) with IPF and PF (Fig. 4d).  
+
+We anticipate that ultrathin SCCFs will be employed in a wide range of high- technology applications. For example, we introduce ultrathin transparent Cu honeycomb mesh electrodes (Supplementary Fig. 11). We fabricated microscale honeycomb meshes from both SCCFs and 15- nm- thick polycrystalline Cu films (PCCFs) using ultraviolet (UV) lithography and wet etching processes (Supplementary Fig. 11a- f). Structural analysis demonstrated that the SCCF mesh maintained the SC structure of Cu(111) even after UV lithography and wet etching, and had better transmittance and lower sheet resistance (Supplementary Fig. 11g). The sheet resistance of the SCCF mesh remained within a few percent of the initial value, even months after preparation (Supplementary Fig. 11h).  
+
+In summary, we implemented heteroepitaxy like homoepitaxy using the sputtering system
+
+<--- Page Split --->
+
+and grew thin films with MSFS. The critical requirement here is that each atom must be individually deposited during the sputtering process. Then, a myriad of nanodroplets merge through coherent lateral growth into coplanar layers at \(11.5 \mathrm{nm}\) , which is the critical thickness of a complete thin film. The thin film evolves to have an MSFS with an RMS roughness of \(< 0.3 \mathrm{nm}\) . In heteroepitaxy, the formation of TBs is inevitable but does not affect the formation of MSFS, whereas GBs have a strong effect. The individual deposition with single atoms enables the growth of thin films with MSFS not only for Cu, but also for other metals, such as Ag, Al and Ni.  
+
+## Methods  
+
+Preparation of thin SCCF using the ASE technique. The network of conducting wires, including cables, in the conventional sputtering system was replaced with SC Cu wires fabricated by cutting SC Cu wafers in a spiral fashion using a wire electrical discharge machining (wire- EDM). The Cu wafers were sliced from an SC ingot grown using the Czochralski method. In our setup, vibration caused by ambient noise was minimised as much as possible using a mechanical noise reduction system. Although minute vibrations appear not to cause significant degradation in conventional thin film growth, especially for PCCFs, such miniscule mechanical vibrations can cause irreversible stacking faults that could significantly disturb the initial nucleation and lateral growth processes, especially the coherent coplanar merging of nuclei. However, the present ASE system provides a stable environment for single- atom deposition with the objective of achieving atomically flat surfaces through the stacking of single atoms. The optimised sputtering conditions using the ASE system are as follows. A double- sided polished (001) \(\mathrm{Al}_2\mathrm{O}_3\) wafer with a thickness of \(430 \mu \mathrm{m}\) was used as the substrate. The deposition temperature and RF (13.56 MHz) power were approximately \(170^{\circ}\mathrm{C}\) and \(30 \mathrm{W}\) , respectively. The target- to- substrate distance was set at \(95 \mathrm{mm}\) . The base pressure was maintained at less than \(2 \times 10^{- 7}\) Torr and the working pressure at \(5.4 \times 10^{- 3}\) Torr with an Ar gas (99.9999% (6N)) flow of \(50 \mathrm{sccm}\) . The relationship between the deposition time and thickness of the thin film (or the average growth rate) was determined from the average deposition time of a \(200 \mathrm{nm}\) - thick film grown under the optimal conditions.  
+
+## Structural information  
+
+For high- resolution (scanning) transmission electron microscopy [HR(S)TEM] analysis, a series of Cu samples grown on \(\mathrm{Al}_2\mathrm{O}_3(0001)\) substrates at deposition times ranging from 15 to
+
+<--- Page Split --->
+
+120 s were cross-sectioned for HR(S)TEM imaging using dual-beam focused ion beam (FIB) slicing (Helios NanoLab 450; FEI Co., Hillsboro, OR, USA) and lift-out processes. Double caesium (Cs)- corrected TEM (JEM- ARM200F; JEOL Ltd., Tokyo, Japan) at \(200\mathrm{kV}\) was used to obtain BF- STEM and HRTEM images of the Cu samples. For high- resolution STEM imaging, the probe- forming semi- angle was 23 mrad. Statistically random background noise in the HR(S)TEM images was reduced using the 2D difference filtering method of a commercial software program (HREM- Filters Pro; HREM Research Inc., Tokyo, Japan). Chemical analysis of the Cu samples was performed using electron energy loss spectroscopy (EELS) with a post- column- type electron energy loss spectrometry system (GIF Quantum ER 965; Gatan, Pleasanton, CA, USA) equipped on the microscope.  
+
+For structural characterisation, XRD measurements were performed using a PANalytical Empyrean Series 2 diffractometer (Malvern PANalytical, Malvern, UK) with a Cu- Kα source (40 kV, 30 mA). Data were collected in the range of \(20^{\circ} < 20 < 90^{\circ}\) , with a step size of \(0.0167^{\circ}\) and dwell time of 0.5 s per point in all cases. EBSD measurements were performed to confirm the quality of the thin films. A SUPRA40 VP SEM (Zeiss, Oberkochen, Germany) was used to measure EBSD maps of the thin films. The EBSD maps, pole figures (PFs) and IPFs show the directions and distributions of the crystals within the films. Surface roughness and sample thickness were measured by AFM using an XE- 100 instrument (Park Systems, Suwon- si, South Korea). The basic scanning conditions included noncontact mode with an \(\sim 0.5\mathrm{Hz}\) scan rate and \(1,024 \times 1,024\) resolution. The sample thickness was determined from an AFM image of Cu film scraped off using a sharp material. The scan rate was reduced to compensate for the stepped shape. The height resolution of the AFM instrument was 1.8 (0.25) Å in high- (low- )- voltage mode.  
+
+## Transport and optical characteristics of SCCFs  
+
+For transport measurements, sheet resistance (Rs: \(\Omega /\mathrm{sq}\) ) was measured in the van der Pauw geometry with an HMS- 3000 Hall measurement system (Ecopia, Toronto, ON, Canada) under a 0.55- T magnetic field at room temperature. Resistivity data \((\rho :\Omega \cdot \mathrm{cm})\) were obtained by multiplying sheet resistance by film thickness measured via AFM.  
+
+We investigated the optical quality of the ultrathin SCCFs using both infrared/optical spectroscopy and time- domain THz techniques. We measured the transmittance spectra of 12 Cu films of thicknesses ranging from 3 to \(30\mathrm{nm}\) at room temperature. To conduct measurements over a wide spectral range from THz to UV wavelengths, we used a commercial THz spectrometer (TERA K15; Menlo, Planegg, Germany) for the spectral range below 100
+
+<--- Page Split --->
+
+\(\mathrm{cm}^{- 1}\) , a commercial FTIR- type spectrometer (Vertex 80v; Bruker, Karlsruhe, Germany) for the spectral range of \(400–25,000 \mathrm{cm}^{- 1}\) , and a commercial monochromatic spectrometer (Lambda 950; PerkinElmer, Waltham, MA, USA) for the spectral range from \(3,000 \mathrm{to} 50,000 \mathrm{cm}^{- 1}\) . Because the sapphire substrate has strong infrared- active phonons between \(150 \mathrm{and} 1,500 \mathrm{cm}^{- 1}\) , we could not obtain reliable spectra in this spectral range. However, no meaningful optical features of Cu films lie in this spectral range.  
+
+## Theoretical approach  
+
+All ab initio total energy calculations and geometry optimisations were performed using density functional theory (DFT) in the generalised gradient approximation (GGA) based on the Perdew- Burke- Ernzerhof functional \(^{36}\) and projected augmented plane- wave method, \(^{37}\) as implemented by Kresse et al. \(^{38}\) The \(\mathrm{Al}_{2}\mathrm{O}_{3}(0001)\) substrate was represented by a slab of 36 atomic layers of primitive unit cells containing 12 formula units, and the Cu thin film was represented by a slab of six layers of Cu atoms. The calculated lattice constants for bulk \(\mathrm{Al}_{2}\mathrm{O}_{3}\) are \(a = 4.785 \mathrm{\AA}\) and \(c = 13.06 \mathrm{\AA}\) , in good agreement with the experimental values. \(^{39}\) A vacuum length of \(15 \mathrm{\AA}\) was used, the bottom nine layers of the slab were fixed in their bulk positions, and the remaining atoms were fully relaxed until the Hellmann- Feynman force on each atom was \(< 0.001 \mathrm{eV / \AA}\) and the change in total energy was \(< 1 \times 10^{- 5} \mathrm{eV}\) . A supercell containing a slab of \(3 \times 2\) surface unit cells was used to simulate diffusion, and a slab of \(1 \times 1\) surface unit cells was used to calculate the adhesion energy of \(\mathrm{Cu}(111)\) on a \(\mathrm{Al}_{2}\mathrm{O}_{3}(0001)\) substrate. The electron wave functions were expanded in a plane- wave basis set with a cut- off energy of \(420 \mathrm{eV}\) , and Brillouin- zone integration for the slabs was performed using a \(5 \times 5 \times 1\) Monkhorst- Pack \(k\) - point grid. \(^{40}\) The nudged elastic band method \(^{41}\) was used to calculate the activation energy of diffusion with \(0.01 \mathrm{eV / \AA}\) of the force criterion for structure optimisation.  
+
+We modelled the growth of nanodroplets into a thin film on a substrate by considering the total energy \(E\) of the droplet on a substrate, as follows:  
+
+\[E = -E_{c}V + \int \tau dS - \int E_{a}dS_{a}, \quad (1)\]  
+
+where \(E_{c}\) and \(\tau\) are the cohesive energy per unit volume and surface tension, respectively, of a droplet with volume \(V\) and surface area \(S\) , and \(E_{a}\) is the adhesive energy per unit area of the interface between the droplet and a substrate with area \(S_{a}\) . To simplify the analysis, we assumed that the droplet was a cylinder of radius \(a\) and height \(h\) . The surface energy term can be written as the sum of two terms: \(\int \tau dS = 2\tau_{\perp}S_{\perp} + \tau_{\parallel}S_{\parallel}\) , where \(\tau_{\perp}(\tau_{\parallel})\) is the average surface tension of the top (side) face of the droplet and \(S_{\perp} = \pi a^{2}(S_{\parallel} = 2\pi ah)\) is the area of the top (side) face. To determine the shape of the droplet for a given amount (volume) of adsorbate atoms, we
+
+<--- Page Split --->
+
+expressed the total energy in terms of the droplet height \(h\) and volume \(V\) using \(a = \sqrt{V / \pi} h^{- 1 / 2}\) , as follows:  
+
+\[E = -E_{c}V + (2\tau_{\perp} - E_{a})\pi \big(\sqrt{V / \pi} h^{-1 / 2}\big)^{2} + \tau_{\parallel}2\pi \big(\sqrt{V / \pi} h^{-1 / 2}\big)h. \quad (2)\]  
+
+The sign conventions were selected such that all physical quantities were positive for typical substrates and adsorbates. For a given droplet of volume \(V\) ,  
+
+\[\frac{\partial E}{\partial h} = (2\tau_{\perp} - E_{a})V(-h^{-2}) + \tau_{\parallel}2\sqrt{\pi}\sqrt{V}\left(\frac{1}{2} h^{-1 / 2}\right).\]  
+
+\(E\) is at its minimum when \(\frac{\partial E}{\partial h} = 0\) :  
+
+\[\tau_{\parallel}\sqrt{\pi}\sqrt{V} h^{3 / 2} = (2\tau_{\perp} - E_{a})V. \quad (3)\]  
+
+When \(E_{a}< 2\tau_{\perp}\) , the ratio between the height and lateral size is:  
+
+\[\frac{h}{a} = 2\frac{\tau_{\perp} - E_{a} / 2}{\tau_{\parallel}}.\]  
+
+Generalising this result to droplets of different shapes, and noting that \(2a\) represents the lateral size, the evolution of the droplet shape can be expressed in terms of the aspect ratio \(R = h / 2a\) , as follows:  
+
+\[R = f\frac{\tau_{\perp} - E_{a} / 2}{\tau_{\parallel}} \qquad \text{for} E_{a}< 2\tau_{\perp} \quad (4a)\]  
+
+Where \(f\) is a form factor on the order of unity that depends on the specific shape of the droplet.  
+
+If \(E_{a}\geq 2\tau_{\perp}\) , the energy of the droplet in Eq. (2) is a monotonically increasing function of droplet height \(h\) , and minimising the energy causes \(h\) to approach zero (or \(a\) to increase limitlessly). Therefore,  
+
+\[R\rightarrow 0 \qquad \text{for} E_{a}\geq 2\tau_{\perp}, \quad (4b)\]  
+
+which is interpreted as a lack of droplet formation, with only layer growth occurring.  
+
+## Data availability  
+
+The authors declare that the main data supporting the findings of this study are available within the article and its Supplementary Information files.
+
+<--- Page Split --->
+
+## 414 References
+
+<--- Page Split --->
+
+448 Advances 3, 082105 (2013). 449 18. Lu, L., Shen, Y., Chen, X., Qian, L. & Lu, K. Ultrahigh Strength and High Electrical Conductivity in copper, Science 304, 422- 426 (2004). 450 19. Zhang, X. et al. Nanocrystalline copper films are never flat. Science 357, 397- 400 (2017). 451 20. Schiotz, J. & Jacobsen, K. W. Roughness in flatland. Nature Mater. 16, 1059- 1060 (2017). 452 21. Kim, S. J. et al. Flat- surface- assisted and self- regulated oxidation resistance of Cu (111). Nature 603, 434- 438 (2022). 453 22. Slavin, A. J. Growth modes of ultrathin metal films on dissimilar metal substrates. Prog. Surf. Sci. 50, 159- 172 (1995). 454 23. Williams, D. B. & Carter, C. B. Imaging. Transmission Electron Microscopy: A Textbook for Materials Science \(2^{\mathrm{nd}}\) ed. (Springer, 2009), pp. 369- 506. 455 24. Uragami, T. S. Pendellösung fringes in a finite crystal. J. Phys. Soc. Jpn. 13, 1141- 1161 (1971). 456 25. Tran, R. et al. Surface energies of elemental crystals. Sci. Data 3, 160080 (2016). 457 26. Li, L. et al. Surface- step- induced oscillatory oxide growth, Phys. Rev. Lett. 113, 136104 (2014). 458 27. Schmiedla, E., Wissmanna, P. & Finzelb, H.- U. The electrical resistivity of ultra- thin copper films. Z. Naturforsch. 63a, 739- 744 (2008). 459 28. Liu, H.- D., Zhao, Y.- P., Ramanath, G. S., Murarka, P. & Wang, G.- C. Thickness- dependent electrical resistivity of ultrathin (< 40 nm) Cu films. Thin Solid Films 384, 151- 156 (2001). 460 29. Fuchs, K. The conductivity of thin metallic films according to the electron theory of metals. Proc. Cambridge Philos. Soc. 34, 100- 108 (1938). 461 30. Sondheimer, E. H. The mean free path of electrons in metals. Adv. Phys. 1, 1- 42 (1952). 462 31. Namba, Y. Resistivity and temperature coefficient of thin metal films with rough surface. Jpn. J. Appl. Phys. 9, 1326- 1329 (1970). 463 32. Homes, C. C., Xu, Z. J., Wen, J. S. & Gu, G. D. Effective medium approximation and the complex optical properties of the inhomogeneous superconductor \(\mathrm{K}_{0.8}\mathrm{Fe}_{2 - y}\mathrm{Se}_2\) . Phys. Rev. B. 86, 144530 (2012). 464 33. Tanner, D. B. Classical theories for the dielectric function. Optical Effects in Solids (Cambridge Univ. Press, 2019), pp. 30- 42. 465 34. Schmiedla, E., Wissmanna, P. & Finzelb, H.- U. The electrical resistivity of ultra- thin
+
+<--- Page Split --->
+
+481 copper films. Z. Naturforsch. 63a, 739- 744 (2008). 482 35. Cho, Y. C. et al. Copper better than silver: Electrical resistivity of the grain- free single- crystal copper wire. Cryst. Growth Des. 10, 2780- 2784 (2010). 483 36. Perdew, J. P. et al. Generalized gradient approximation made simple. Phys. Rev. Lett. 78, 1396 (1997). 484 37. Blöchl, P. E. Projector augmented- wave method. Phys. Rev. B 50, 17953- 17979 (1994). 485 38. Kresse, G. & Joubert, D. From ultrasoft pseudopotentials to the projector augmented- wave method. Phys. Rev. B 59, 1758- 1775 (1999). 486 39. Izumi, F., Asano, H., Murata, H. & Watanabe, N. Rietveld analysis of powder patterns obtained by TOF neutron diffraction using cold neutron sources. J. App. Cryst. 20, 411- 418 (1987). 487 40. Monkhorst, H. J. & Pack, J. D. Special points for Brillouin- zone integrations. Phys. Rev. B 13, 5188- 5192 (1976). 488 41. Jonsson, H., Mills, G. & Jacobsen, K. W. Nudged elastic band method for finding minimum energy paths of transitions, Classical and Quantum Dynamics in Condensed Phase Simulations, ed. B. J. Berne, G. Ciccotti and D. F. Coker (World Scientific, 1998).  
+
+## Acknowledgements  
+
+This research was supported by the National Research Foundation of Korea (NRF) (nos., NRF- 2022R1A2B5B03001219, NRF- 2020R1A4A4078780, NRF- 2019R1A6A1A11053838, 2021R1C1C1006316, 2019R1I1A1A01058304, 2016M3D1A1919181, 2020R1A2C1006207, and 2021R1A2C101109811), Institute for Basic Science (IBS- R011- D1) and by the Commercialization Promotion Agency for R&D Outcomes(COMPA) funded by the Ministry of Science and ICT(MSIT) (2022RMD- S08). Use of the TEM instrument was supported by the Advanced Facility Center for Quantum Technology at SKKU.  
+
+## Author contributions  
+
+S.- Y.J., J.H., and Y.- M.K. conceived this study. S.J.K., S.E.P and Y.L. performed the Cu thin film growth and AFM, XRD, EBSD, SEM, and DC transport measurements. Y.H.K., Y.I.K., J.C.K., S.J.K., H.Y.J., and Y.- M.K. performed TEM measurements and analyses. T.H., T.T.K., K.I.S., and J.H.K. performed THz spectral measurements. Y.- S.S. and J.H. performed IR and optical measurements and EMA. S.- G.K. and B.L. carried out first- principles calculations. S.- Y.J. and Y.H.L. supervised the project. T.H., Y.- S.S., T.T.K., Y.- M.K., J.H., and S.- Y.J. wrote
+
+<--- Page Split --->
+
+515 the manuscript. All authors participated in the manuscript review.  
+
+## Competing interests  
+
+The authors have no competing interests.  
+
+## Additional information  
+
+Supplementary information The online version contains supplementary material available at https://doi.org./  
+
+Correspondence and requests for materials should be addressed to Seong- Gon Kim, Young- Min Kim, Jungseek Hwang or Se- Young Jeong.  
+
+Peer review information Nature Communications thanks  
+
+Reprints and permissions information is available at http://www.nature.com/reprints.  
+
+Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+- 2DCuSupplementaryInfoJSY.pdf
+
+<--- Page Split --->

preprint/preprint__10394e4b9f4446a1b17a9ba94de8cd766e54541a93fbbde182594cfbc2bc4d37/preprint__10394e4b9f4446a1b17a9ba94de8cd766e54541a93fbbde182594cfbc2bc4d37_det.mmd

@@ -0,0 +1,344 @@
+<|ref|>title<|/ref|><|det|>[[44, 106, 955, 175]]<|/det|>
+# Coherent consolidation of trillions of nucleations for mono-atom step-level flat surfaces  
+
+<|ref|>text<|/ref|><|det|>[[44, 196, 285, 238]]<|/det|>
+Taewoo Ha  Institute for Basic Science  
+
+<|ref|>text<|/ref|><|det|>[[44, 244, 283, 285]]<|/det|>
+Yu- Seong Seo  Sungkyunkwan University  
+
+<|ref|>text<|/ref|><|det|>[[44, 290, 220, 330]]<|/det|>
+Teun- Teun Kim  University of Ulsan  
+
+<|ref|>text<|/ref|><|det|>[[44, 336, 655, 377]]<|/det|>
+Bipin Lamichhane  Mississippi State University  https://orcid.org/0000- 0002- 9503- 083X  
+
+<|ref|>text<|/ref|><|det|>[[44, 381, 911, 424]]<|/det|>
+Young- Hoon Kim  Department of Energy Science, Sungkyunkwan University  https://orcid.org/0000- 0001- 7343- 1512  
+
+<|ref|>text<|/ref|><|det|>[[44, 428, 639, 470]]<|/det|>
+Su Jae Kim  Pusan National University  https://orcid.org/0000- 0002- 6033- 8551  
+
+<|ref|>text<|/ref|><|det|>[[44, 475, 283, 515]]<|/det|>
+Yousil Lee  Pusan National University  
+
+<|ref|>text<|/ref|><|det|>[[44, 521, 472, 562]]<|/det|>
+Jong Kim  UNIST  https://orcid.org/0000- 0003- 4101- 0590  
+
+<|ref|>text<|/ref|><|det|>[[44, 567, 283, 608]]<|/det|>
+Sang Eon Park  Pusan National University  
+
+<|ref|>text<|/ref|><|det|>[[44, 614, 640, 655]]<|/det|>
+Kyung Ik Sim  Institute for Basic Science  https://orcid.org/0000- 0002- 9870- 3913  
+
+<|ref|>text<|/ref|><|det|>[[44, 660, 562, 701]]<|/det|>
+Jae Kim  Yonsei University  https://orcid.org/0000- 0002- 7840- 3630  
+
+<|ref|>text<|/ref|><|det|>[[44, 706, 640, 748]]<|/det|>
+Yong In Kim  Sungkyunkwan University  https://orcid.org/0000- 0002- 9540- 5996  
+
+<|ref|>text<|/ref|><|det|>[[44, 753, 283, 794]]<|/det|>
+Seon Kim  Sungkyunkwan University  
+
+<|ref|>text<|/ref|><|det|>[[44, 800, 859, 841]]<|/det|>
+Hu Young Jeong  Ulsan National Institute of Science and Technology  https://orcid.org/0000- 0002- 5550- 5298  
+
+<|ref|>text<|/ref|><|det|>[[44, 846, 175, 864]]<|/det|>
+Young Hee Lee  
+
+<|ref|>text<|/ref|><|det|>[[44, 867, 928, 909]]<|/det|>
+Center for Integrated Nanostructure Physics (CINAP), Institute for Basic Science (IBS), Sungkyunkwan University  https://orcid.org/0000- 0001- 7403- 8157  
+
+<|ref|>text<|/ref|><|det|>[[44, 914, 175, 932]]<|/det|>
+Seong- Gon Kim  
+
+<|ref|>text<|/ref|><|det|>[[50, 936, 653, 955]]<|/det|>
+Mississippi State University  https://orcid.org/0000- 0002- 1629- 0319
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[42, 42, 640, 177]]<|/det|>
+Young- Min Kim Sungkyunkwan University https://orcid.org/0000- 0003- 3220- 9004 Jungseek Hwang Sungkyunkwan University https://orcid.org/0000- 0002- 2555- 1218 Se- Young Jeong ( syjeong@pusan.ac.kr ) Pusan National University https://orcid.org/0000- 0003- 1019- 4403  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 217, 102, 234]]<|/det|>
+## Article  
+
+<|ref|>text<|/ref|><|det|>[[42, 255, 912, 299]]<|/det|>
+Keywords: Coherent merging, coplanar layer, initial growth stages, mono- atom step- level flat surface, single- crystal thin films  
+
+<|ref|>text<|/ref|><|det|>[[44, 316, 352, 336]]<|/det|>
+Posted Date: September 16th, 2022  
+
+<|ref|>text<|/ref|><|det|>[[42, 354, 474, 373]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 2018252/v1  
+
+<|ref|>text<|/ref|><|det|>[[42, 391, 910, 434]]<|/det|>
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[70, 81, 870, 101]]<|/det|>
+# Coherent consolidation of trillions of nucleations for mono-atom step-level flat surfaces  
+
+<|ref|>text<|/ref|><|det|>[[70, 108, 885, 230]]<|/det|>
+Taewoo Ha \(^{1\dagger}\) , Yu-Seong Seo \(^{2\dagger}\) , Teun-Teun Kim \(^{3\dagger}\) , Bipin Lamichhane \(^{4}\) , Young-Hoon Kim \(^{5}\) , Su Jae Kim \(^{6}\) , Yousil Lee \(^{6}\) , Jong Chan Kim \(^{7}\) , Sang Eon Park \(^{6}\) , Kyung Ik Sim \(^{1,8}\) , Jae Hoon Kim \(^{8}\) , Yong In Kim \(^{5}\) , Seon Je Kim \(^{5}\) , Hu Young Jeong \(^{7,9}\) , Young Hee Lee \(^{1,5}\) , Seong-Gon Kim \(^{4}\) , Young-Min Kim \(^{1,5,*}\) , Jungseek Hwang \(^{2,*}\) & Se-Young Jeong \(^{10,11,*}\)  
+
+<|ref|>text<|/ref|><|det|>[[67, 258, 885, 700]]<|/det|>
+\(^{1}\) Center for Integrated Nanostructure Physics, Institute for Basic Science, Sungkyunkwan University, Suwon 16419, Republic of Korea. \(^{2}\) Department of Physics, Sungkyunkwan University, Suwon 16419, Republic of Korea. \(^{3}\) Department of Physics, University of Ulsan, Ulsan 44610, Republic of Korea. \(^{4}\) Department of Physics and Astronomy, Mississippi State University, Mississippi State, MS 39762, USA. \(^{5}\) Department of Energy Science, Sungkyunkwan University, Suwon 16419, Republic of Korea. \(^{6}\) Crystal Bank Research Institute, Pusan National University, Busan 46241, Republic of Korea. \(^{7}\) School of Materials Science and Engineering, Ulsan National Institute of Science and Engineering, Ulsan 44919, Republic of Korea. \(^{8}\) Department of Physics, Yonsei University, Seoul 03722, Republic of Korea. \(^{9}\) UNIST Central Research Facilities, Ulsan National Institute of Science and Technology, Ulsan 44919, Republic of Korea. \(^{10}\) Department of Cogno-Mechatronics Engineering, Pusan National University, Busan 46241, Republic of Korea. \(^{11}\) Department of Optics and Mechatronics Engineering, Pusan National University, Busan 46241, Republic of Korea. \(^{*}\) Corresponding authors. E- mail: youngmk@skku.edu (Y.- M. Kim); jungseek@skku.edu (J. Hwang); syjeong@pusan.ac.kr (S.- Y. Jeong) \(^{\dagger}\) These authors contributed equally to this work.  
+
+<|ref|>text<|/ref|><|det|>[[66, 747, 884, 840]]<|/det|>
+\(^{26}\) E- mail: youngmk@skku.edu (Y.- M. Kim); jungseek@skku.edu (J. Hwang); syjeong@pusan.ac.kr (S.- Y. Jeong) \(^{\dagger}\) These authors contributed equally to this work.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 82, 884, 397]]<|/det|>
+Constructing a mono- atom step- level ultra- flat material surface is challenging, especially for thin films, because it is prohibitively difficult for trillions of clusters to coherently merge. Even though a rough metal surface, as well as the scattering of carriers at grain boundaries, limits electron transport and obscures their intrinsic properties, the importance of the flat surface has not been emphasised sufficiently. In this study, we describe in detail the initial growth of copper thin films required for mono- atom step- level flat surfaces (MSFSs). Deposition using atomic sputtering epitaxy leads to the coherent merging of trillions of islands into a coplanar layer, eventually forming an MSFS, for which the key factor is suggested to be the individual deposition of single atoms. Theoretical calculations support that single sputtered atoms ensure the formation of highly aligned nanodroplets and help them to merge into a coplanar layer. The realisation of the ultra- flat surfaces is expected to greatly assist efforts to improve quantum behaviour by increasing the coherency of electrons.  
+
+<|ref|>text<|/ref|><|det|>[[115, 426, 880, 469]]<|/det|>
+Keywords: Coherent merging; coplanar layer; initial growth stages; mono- atom step- level flat surface; single- crystal thin films  
+
+<|ref|>text<|/ref|><|det|>[[111, 500, 884, 911]]<|/det|>
+Ultrathin metal films are indispensable in modern electronics and nanotechnology<sup>1- 3</sup>. During the past few decades, conventional metals have been studied extensively because the performance of metal- based devices is intimately related to their physical properties.<sup>4- 6</sup> Efforts to produce single- crystal (SC) copper (Cu) from Cu foil have been driven by competition<sup>7,8</sup> and interest in their nanocrystalline nature and potential applications in large two- dimensional (2D) components consisting of materials such as graphene and hexagonal boron nitride (h- BN).<sup>9- 11</sup> However, despite the importance of metal thin- film flatness, there have been few reports because it is challenging to control flatness. Since the contact between the metal electrode and the semiconductor material decisively affects the properties of electronic and optoelectronic devices,<sup>12,13</sup> a flat metal surface is proposed as a good solution to reduce contact resistance. The motion of electrons without scattering at surfaces and grain boundaries can also affect the carrier transport properties.<sup>14</sup> Single- crystal Cu thin films (SCCFs) on sapphire have recently been reported, and the formation of twin boundaries (TBs) have been investigated intensively.<sup>15- 17</sup> Two orientations (ORs) adjacent to a TB are rotated by a certain angle in- plane and satisfy the symmetry operation exactly, while two ORs adjacent to a grain boundary (GB) are tilted both in out- of- plane and in- plane directions. However, in view of electronic motion, a much clearer distinction between TB and GB must be followed by precise microscopic analysis. Twin
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 81, 884, 323]]<|/det|>
+boundaries are not strong electron scatterers, \(^{18}\) while electrons are scattered at GBs if they have an in-plane mistilt even by \(1^{\circ}\) , even if ORs adjacent to GBs are aligned almost perfectly along the out-of-plane direction. The number of TBs produced near the interface gradually decreases and may disappear as the position approaches the surface. \(^{15}\) Nevertheless, for the grown thin film to lead to an ultra- smooth surface, it is necessary to discuss the initial growth process precisely. A study using scanning tunnelling microscopy demonstrated that nanocrystalline Cu films cannot be flat because valleys and ridges are created by out- of- plane grain rotation. \(^{19,20}\) A recent study revealed that thin Cu films fabricated on four- inch- diameter wafers using atomic sputtering epitaxy (ASE) have atomically flat surfaces overall, with occasional mono- atomic step edges, and the crystal quality is maintained even after a few years. \(^{21}\)  
+
+<|ref|>text<|/ref|><|det|>[[113, 330, 884, 519]]<|/det|>
+In this study, we describe in detail the initial growth phases of ultrathin Cu films prepared using ASE (Methods and Supplementary Fig. 1a). Through transmission electron microscopy (TEM) observations, we identified the structural evolution from quasi- zero- dimensional (0D) nanodroplets to quasi- 2D thin films in the initial growth stage. It was demonstrated experimentally and theoretically that the only way to consolidate the trillions of nucleations and to achieve a mono- atom step- level flat surface (MSFS) is by the individual deposition of single atoms, when considering long- range period lattice mismatch between Cu and substrate \(\mathrm{Al}_2\mathrm{O}_3\) and the formation of GB and TB.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 551, 375, 568]]<|/det|>
+## Growth of an ultrathin SCCF  
+
+<|ref|>text<|/ref|><|det|>[[112, 576, 884, 889]]<|/det|>
+Three approaches are typically used to grow thin films on a single- crystal substrate, i.e. the Volmer- Weber, Frank- van der Merwe and Stranski- Krastanov methods. \(^{22}\) In contrast, our ASE approach follows a new growth mode. The ideal initial growth of Cu on \(\mathrm{Al}_2\mathrm{O}_3\) substrate follows three stages (Fig. 1a): stage I, nanodroplet nucleation and lateral growth; stage II, coherent merging; and stage III, layer- by- layer growth of an SCCF. For successful completion of this process, each of trillions of nanodroplets must consist of a single- crystal, and all must be aligned in the same direction. A high- resolution TEM image of a perfectly flat surface of a 12- nm- thick film and its strain field map obtained by geometrical phase analysis, respectively, is shown in Fig. 1b and c. The Cu thin films thicker than 11.5 nm are aligned perfectly along the (111) plane, which is supported by X- ray diffraction (XRD), atomic force microscopy (AFM), electron backscatter diffraction (EBSD) mapping and scanning electron microscopy (SEM) and TEM images (Supplementary Fig. 1b- i). \(^{21}\) The interface between Cu and the \(\mathrm{Al}_2\mathrm{O}_3\) substrate was defect- and strain- free after growth of the 12- nm- thick film (Fig. 1d, e).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[120, 100, 840, 525]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[114, 555, 884, 694]]<|/det|>
+<center>Fig. 1 | Coherent consolidation of nucleations and a mono-atom step-level flat surface. a, Series of schematic diagrams of the structural evolution from quasi-zero-dimensional \(\mathrm{Cu}(111)\) nanodroplets to a quasi-two-dimensional single-crystal Cu film in three stages: (I) nucleation and lateral growth, (II) coherent coplanar merging, and (III) layer-by-layer growth. b, c, Cross-sectional high-resolution transmission electron microscopy (HRTEM) image of a perfectly flat surface (left) and strain field map by geometrical phase analysis (GPA) (right), observed in the \([1\bar{1} 0]\) orientation. d, e, Cross-sectional HRTEM image of the interfacial region of the \(\mathrm{Cu / Al_2O_3}\) heterostructure and its strain field map by GPA, with an orientation relationship of \((111)_{\mathrm{Cu}}[1\bar{1} 0]_{\mathrm{Cu}} / (001)_{\mathrm{Al_2O_3}}[1\bar{1} 0]_{\mathrm{Al_2O_3}}\) . </center>  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 728, 502, 744]]<|/det|>
+## Coherent consolidation into a coplanar layer  
+
+<|ref|>text<|/ref|><|det|>[[115, 752, 883, 842]]<|/det|>
+The detailed states of early growth are shown in Fig. 2. Figure 2 shows a series of cross- sectional bright- field scanning TEM (BF- STEM) images<sup>23</sup> acquired from samples of various thicknesses (Fig. 2a) along with corresponding illustrations (Fig. 2b) and topographic images obtained by atomic force microscopy (AFM) (Fig. 2c).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[125, 87, 870, 455]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 462, 883, 550]]<|/det|>
+<center>Fig. 2 | Coherent consolidation process for a mono-atom step-level flat surface. a, Cross-sectional bright-field scanning transmission electron microscopy images. b, Corresponding illustrations. c, Topographic atomic force microscopy images acquired from samples of five different thicknesses, corresponding to initial deposition times of 15, 30, 45, 60 and 120 s. d, MSFS of the single-crystal Cu film with a root-mean-square (RMS) roughness of \(\sim 0.2 \mathrm{nm}\) . e, Schematic side view of d. </center>  
+
+<|ref|>text<|/ref|><|det|>[[113, 573, 884, 912]]<|/det|>
+Two separate initial stages are apparent in Fig. 2, divided according to a thickness threshold of \(\sim 5 \mathrm{nm}\) . During stage I (Fig. 2a–c, first three panels), the Cu atoms form Cu(111) nanodroplets with single crystallinity; these nucleate at distances of \(20–30 \mathrm{nm}\) and consist of 18–31 layers (3–5 nm) on average, with a small height distribution (Supplementary Fig. 2). Because the area of an initial nanodroplet ranges from approximately 30 to \(100 \mathrm{nm}^2\) , the total number of nanodroplets on a 2-inch-diameter wafer is approximately \(\sim 10^{12}\) . During stage II, (Fig. 2a–c, fourth panels), Cu(111) nanodroplets begin to form conduction channels through Cu(111) lateral growth and coherent coplanar merging. The height of the nanodroplets increases more slowly (i.e. by a few layers) than the expected average deposition rate (i.e. \(\sim 7\) layers per 15 s of initial deposition time), indicating predominant lateral growth. AFM images of \(< 5 \mathrm{nm}\) - thick films (Fig. 2c, first three images) obtained at a \(1.0 \mathrm{nm}\) resolution (for an area of \(100 \times 100 \mathrm{nm}^2\) ) show the growth of islands separated by \(20–30 \mathrm{nm}\) , whereas thin films with a thickness of \(\geq 5 \mathrm{nm}\) (Fig. 2c, fourth image) obtained at a \(1.5 \mathrm{nm}\) resolution for a much larger area \((10 \times 10 \mu \mathrm{m}^2)\) exhibit a mainly flat surface. Remarkably, thin films thicker than \(10 \mathrm{nm}\) showed an atomically
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[110, 78, 884, 820]]<|/det|>
+flat surface with an exceptionally small root mean square (RMS) roughness of \(< 0.2 \mathrm{nm}\) . A thin film with RMS roughness of \(< 0.3 \mathrm{nm}\) has an atomically flat surface with occasional monoatomic step edges. A cross-sectional HRTEM image (Fig. 2d) and its graphical illustration (Fig. 2e) show the MSFS of the SCCF with an RMS roughness of \(\sim 0.2 \mathrm{nm}\) . The fourth and fifth AFM images in Fig. 2c provide experimental evidence for the illustration in Fig. 2e and show that the thin film grows to an atomically flat surface beyond a thickness of \(\sim 10 \mathrm{nm}\) . AFM images and RMS roughness as a function of thicknesses, EBSD and inverse pole figure (IPF), SEM images and XRD data for 12 Cu films exhibiting marked Pendellösung oscillations (thickness fringes) \(^{24}\) also supported the quality of these films (Supplementary Fig. 3). The estimated thicknesses of all 12 thin-film samples obtained via AFM are listed in Supplementary Table 1. Twin boundaries or GBs in stage II critically affect the final roughness of the surface. The regions in a polycrystal have random ORs, which are separated mostly by GBs and sometimes by TBs (Supplementary Fig. 4a), whereas regions in an SCCF have only two ORs and are separated by only TBs (Supplementary Fig. 4b). The two ORs are associated with two different stacking orders, i.e. ABCABC... and ACBACB..., and are separated by TBs with a closed path. Two different ORs in the SCCF must be rotated exactly by \(60^{\circ}\) to each other in-plane. The boundaries marked in blue (Supplementary Fig. 4b) resemble TBs but are GBs because they have rotational components that deviate slightly from \(60^{\circ}\) in-plane. \(^{15}\) Thus, it is not appropriate to identify TBs or GBs using an optical microscope or micrometre-scale EBSD map. Therefore, it is necessary to perform a misorientation line analysis at the nanoscale. Small ORs that are not observed at the micrometre scale are frequently observed at the nanometre scale. Two regions separated by a TB merge into a larger region of approximately \(5 - 6 \mathrm{nm}\) , and two enlarged regions double into a single OR at \(\sim 12 \mathrm{nm}\) . With increasing thin-film thickness, the region of a single OR doubles every \(5 - 6 \mathrm{nm}\) similarly, and eventually (when the thickness reaches \(\sim 80 \mathrm{nm}\) ) the number of TBs in the upper part of the thin film is significantly reduced (Supplementary Fig. 5). Ideally, a single-crystal thin film, such as that grown homoepitaxially, can be obtained via heteroepitaxy near the surface when the thickness exceeds \(\sim 80 \mathrm{nm}\) . Homoepitaxy-like heteroepitaxy of a Cu thin film on a hetero substrate is only possible when it is deposited by ASE, and the long-distance periodicity by the calculation of extended atomic distance mismatch \(^{21}\) is considered.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[130, 88, 850, 565]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 600, 883, 768]]<|/det|>
+<center>Fig. 3 | Theoretical approach for coherent consolidation of nucleations into a coplanar layer. a, Scheme for the growth of a single nanodroplet via atomic diffusion and b, high-resolution scanning transmission electron microscopy image and fast Fourier transform pattern (inset) of a Cu nanodroplet oriented along the \([1\bar{1} 0]\) zone axis, indicating single-crystal nanodroplet growth along the single-crystal [111] direction. c, Relative energy profiles for the diffusion of Cu atoms or atom clusters (up to four atoms) on an \(\mathrm{Al}_2\mathrm{O}_3(0001)\) substrate. Schemes for d, lateral nanodroplet growth and e, coherent merging to coplanar layers. </center>  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 798, 561, 816]]<|/det|>
+## Individual deposition of single atoms toward MSFS  
+
+<|ref|>text<|/ref|><|det|>[[115, 823, 883, 915]]<|/det|>
+Copper forms SC nanodroplets even during the early growth stage ( \(< 10 \mathrm{~s}\) ) (Figs. 2a and 3a, b). In this study, the droplet surfaces were composed of well- defined crystallographic facets (Fig. 3b). The nanodroplet FFT patterns indicated that the structure consists of a single phase of Cu (Fig. 3b, inset). These nanodroplets grow along the [111] direction of the \(\mathrm{Al}_2\mathrm{O}_3(0001)\) substrate
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[110, 80, 884, 700]]<|/det|>
+in one of two stacking orders (ABCABC... or ACBACB...). The height distribution of the nanodroplets was narrow; beyond a critical height, these droplets stop growing vertically and grow only laterally (Supplementary Fig. 2f). We used a simple model calculation to develop a new thin film growth mode that explains the conditions under which transition of the growth mechanism occurs (Methods). Our model showed that the evolution of the thin film growth mechanism depends on the relative strength of the surface tension and adhesion energy. Initially, deposition of Cu atoms occurs on the Al-terminated surface of the \(\mathrm{Al}_2\mathrm{O}_3(0001)\) substrate, which had a much lower surface tension \((1.59\mathrm{J / m^2})\) than the O-terminated surface \((4.26\mathrm{J / m^2})\) (Supplementary Fig. 6a and Supplementary Table S2). The adhesion energy of the Al-terminated surface was \(- 0.68\mathrm{J / m^2}\) , which was lower than the surface tension of \(\mathrm{Cu}(111)\) \((\tau_{\perp} = 1.34\mathrm{J / m^2})\) . Therefore, the requirement for a positive aspect ratio \((E_{a}< 2\tau_{\perp}\) , Methods) was met, and droplets began to grow spherically at a large aspect ratio \((R)\) of 0.63 (stage I). As the droplet grows, its bottom side flattens, and O atoms are incorporated into the interface between the Cu droplet and \(\mathrm{Al}_2\mathrm{O}_3(0001)\) substrate surface because the interfacial energy of the O-terminated surface is lower than that of the Al-terminated surface (Supplementary Fig. 5b and Supplementary Table S2). The transition of the interface from the Al- to O-terminated surface of the \(\mathrm{Al}_2\mathrm{O}_3(0001)\) substrate increased adhesion to the substrate because the adhesion energy of the fully developed interface of \(\mathrm{Cu}(111)\) on the O-terminated surface was \(4.81\mathrm{J / m^2}\) . When the adhesion became sufficiently strong \((E_{a} > 2\tau_{\perp}\) ; i.e. at the critical height), apparent vertical growth was halted (i.e. \(R\rightarrow 0\) ), and only lateral growth occurred (stage II, Fig. 3c). For SCCFs, the critical height was found to be \(5–6\mathrm{nm}\) (Supplementary Fig. 2f). Because all islands grew along the (111) direction in one of two possible ORs associated with stacking order, two nanodroplets with the same stacking sequence merged to a larger single-crystal droplet with a coplanar layer. However, when two nanodroplets with different ORs were merged, they became a larger droplet with a TB, which retains three-fold rotational symmetry.  
+
+<|ref|>text<|/ref|><|det|>[[111, 699, 884, 914]]<|/det|>
+The number of these separate nanodroplets depends on the diffusion rate of Cu atoms on the substrate. If the diffusion rate is high, then Cu atoms deposited on the substrate move over a large distance before coalescing with existing nanodroplets, forming fewer separate nanodroplet "islands". If the diffusion rate is low, then the Cu atoms have a shorter range of motion, leading to the formation of more independent islands. The diffusion rate of the Cu atoms depends strongly on whether they are deposited on the substrate as single atoms or clusters of multiple atoms. The relative energy profiles of Cu atom diffusion onto an \(\mathrm{Al}_2\mathrm{O}_3(0001)\) substrate are shown in Fig. 3c, as obtained from first-principles calculations. The activation energies of clusters of multiple atoms were higher than those of single atoms;
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 82, 883, 248]]<|/det|>
+notably, the diffusion of clusters of two atoms required the most energy. This finding indicates that the deposition of single atoms during the early stage of nanodroplet nucleation is critical for the growth of an atomically flat SC thin film with uniform OR. Deposits of large clusters are likely to develop into a polycrystalline structure upon merging due to their random ORs. Although this requirement is challenging to fulfil in conventional sputtering systems, where Cu atoms are ejected and deposited as clusters of multiple atoms, the present ASE system meets this requirement.  
+
+<|ref|>text<|/ref|><|det|>[[113, 255, 884, 494]]<|/det|>
+To establish how much the sizes of the species falling off the target surfaces during the sputtering process differ between the conventional system and ASE system, we compared the surfaces of two targets of the general sputtering system and the ASE system after the sputtering process (Supplementary Fig. 7). While the target surface of the general sputtering system was very rough, with an average RMS roughness of 100 nm (Supplementary Fig. 7a), the target surface of the ASE system had a smooth surface, with an average RMS roughness of 4 nm (Supplementary Fig. 7b). Optical images (Supplementary Fig. 7c, d) and AFM surface images at different scales for both targets (Supplementary Fig. 7e- h) showed the critical differences. These results suggest that the difference in the size of the sputtered species during the sputtering process greatly influences thin film growth.  
+
+<|ref|>text<|/ref|><|det|>[[113, 500, 884, 714]]<|/det|>
+Once such a coplanar layer forms, the next layer is highly likely to grow in the same stacking order. After coherent merging, adhesion (with the layer of the adsorbate) became more dominant, and only layer growth occurred (stage III, Fig. 3e). Films thicker than 10 nm showed an ultra- flat, undistorted surface without multi- atomic step edges or grain boundaries (Fig. 1b).<sup>26</sup> In a highly stable thin film growth system, where adsorbate atoms are deposited individually as single atoms, growth strictly follows the energetics of the surface tension and adhesion energy. The growth of GB- free homoepitaxy- like thin films with MSFS is not only possible for Cu, but also for other metals such as Ag, Al, and Ni, if the principle of ASE is maintained.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[137, 95, 833, 625]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 644, 884, 860]]<|/det|>
+<center>Fig. 4 | Three phases of the initial stages of film growth: nucleation, coherent merging, and single-crystal thin film formation. a, Thickness-dependent resistivity \((\rho)\) of SC films according to direct current electrical transport and optical Fourier transform-infrared spectroscopy measurements, as well as data from previous reports<sup>28,34</sup>. Stages I and II are distinguished by the divergence of resistivity (or zero conductivity) at a film thickness \((d)\) close to the nanodroplet size. Stages II and III are distinguished by the resistivity trend determined according to the resistivity slope change in \(\rho\) as a function of \(1 / d\) (inset). b–e, Corresponding electron backscatter diffraction images of Cu thin films in (b) stage I, (c, d) stage II and (e) stage III. </center>
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 108, 295, 125]]<|/det|>
+## Initial growth stages  
+
+<|ref|>text<|/ref|><|det|>[[113, 130, 884, 667]]<|/det|>
+Three distinct stages of initial film growth are also revealed by thickness- dependent DC resistivity \((\rho)\) data (Fig. 4) obtained by DC electrical transport, \(^{27,28}\) Fourier transform- infrared spectroscopy (FTIR) and time- domain terahertz spectroscopy (Supplementary Fig. 8). The transport data showed excellent agreement among the three methods. The DC resistivity \((\rho)\) was close to the bulk value at thicknesses greater than \(\sim 11.5 \mathrm{nm}\) , which was confirmed by plotting \(\rho\) as a function of the inverse of the film thickness (Fig. 4e, inset). Abrupt divergence near a thickness of \(\sim 4.6 \mathrm{nm}\) indicated that the film was not conducting at lower thickness values. At thicknesses greater than \(\sim 4.6 \mathrm{nm}\) , the films had a finite \(\rho\) , indicating conduction channel formation. The results of analytical calculations using the Fuchs- Sondheimer (FS), \(^{29,30}\) Namba, \(^{31}\) and effective medium approximation (EMA) \(^{32}\) models are shown in Supplementary Fig. 9. The \(\rho\) values of films thicker than \(11.5 \mathrm{nm}\) were in good agreement with the FS model, \(^{33}\) whereas those of films thinner than \(11.5 \mathrm{nm}\) deviated from the FS model. The Namba model, which considers surface roughness, matched the data for a surface roughness of \(\sim 4.5 \mathrm{nm}\) , but could not explain \(\rho\) in stage I ( \(< 4.5 \mathrm{nm}\) ), in which Cu nanodroplets did not form conduction channels. The EMA method, which correctly predicts percolation for spherical grains, also described stages I and II well. The results indicate that the film does not occupy the full volume from the substrate surface to the film thickness during the initial growth stage (Supplementary Fig. 10). No reliable EBSDmaps were obtained from \(< 5 - \mathrm{nm}\) - thick films (i.e., stage I) (Fig. 4a), whereas thin films showed partial coverage at thicknesses between 5 and \(11.5 \mathrm{nm}\) (Fig. 4b) and complete coverage at \(>11.5 \mathrm{nm}\) (Fig. 4c). The EBSD map shown in Fig. 4c is indistinguishable from primary blue on the red- green- blue colour scale (0:0:255), indicating exact alignment along (111) with IPF and PF (Fig. 4d).  
+
+<|ref|>text<|/ref|><|det|>[[115, 672, 884, 889]]<|/det|>
+We anticipate that ultrathin SCCFs will be employed in a wide range of high- technology applications. For example, we introduce ultrathin transparent Cu honeycomb mesh electrodes (Supplementary Fig. 11). We fabricated microscale honeycomb meshes from both SCCFs and 15- nm- thick polycrystalline Cu films (PCCFs) using ultraviolet (UV) lithography and wet etching processes (Supplementary Fig. 11a- f). Structural analysis demonstrated that the SCCF mesh maintained the SC structure of Cu(111) even after UV lithography and wet etching, and had better transmittance and lower sheet resistance (Supplementary Fig. 11g). The sheet resistance of the SCCF mesh remained within a few percent of the initial value, even months after preparation (Supplementary Fig. 11h).  
+
+<|ref|>text<|/ref|><|det|>[[130, 894, 881, 912]]<|/det|>
+In summary, we implemented heteroepitaxy like homoepitaxy using the sputtering system
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 82, 883, 275]]<|/det|>
+and grew thin films with MSFS. The critical requirement here is that each atom must be individually deposited during the sputtering process. Then, a myriad of nanodroplets merge through coherent lateral growth into coplanar layers at \(11.5 \mathrm{nm}\) , which is the critical thickness of a complete thin film. The thin film evolves to have an MSFS with an RMS roughness of \(< 0.3 \mathrm{nm}\) . In heteroepitaxy, the formation of TBs is inevitable but does not affect the formation of MSFS, whereas GBs have a strong effect. The individual deposition with single atoms enables the growth of thin films with MSFS not only for Cu, but also for other metals, such as Ag, Al and Ni.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 305, 207, 323]]<|/det|>
+## Methods  
+
+<|ref|>text<|/ref|><|det|>[[111, 330, 884, 794]]<|/det|>
+Preparation of thin SCCF using the ASE technique. The network of conducting wires, including cables, in the conventional sputtering system was replaced with SC Cu wires fabricated by cutting SC Cu wafers in a spiral fashion using a wire electrical discharge machining (wire- EDM). The Cu wafers were sliced from an SC ingot grown using the Czochralski method. In our setup, vibration caused by ambient noise was minimised as much as possible using a mechanical noise reduction system. Although minute vibrations appear not to cause significant degradation in conventional thin film growth, especially for PCCFs, such miniscule mechanical vibrations can cause irreversible stacking faults that could significantly disturb the initial nucleation and lateral growth processes, especially the coherent coplanar merging of nuclei. However, the present ASE system provides a stable environment for single- atom deposition with the objective of achieving atomically flat surfaces through the stacking of single atoms. The optimised sputtering conditions using the ASE system are as follows. A double- sided polished (001) \(\mathrm{Al}_2\mathrm{O}_3\) wafer with a thickness of \(430 \mu \mathrm{m}\) was used as the substrate. The deposition temperature and RF (13.56 MHz) power were approximately \(170^{\circ}\mathrm{C}\) and \(30 \mathrm{W}\) , respectively. The target- to- substrate distance was set at \(95 \mathrm{mm}\) . The base pressure was maintained at less than \(2 \times 10^{- 7}\) Torr and the working pressure at \(5.4 \times 10^{- 3}\) Torr with an Ar gas (99.9999% (6N)) flow of \(50 \mathrm{sccm}\) . The relationship between the deposition time and thickness of the thin film (or the average growth rate) was determined from the average deposition time of a \(200 \mathrm{nm}\) - thick film grown under the optimal conditions.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 849, 316, 866]]<|/det|>
+## Structural information  
+
+<|ref|>text<|/ref|><|det|>[[115, 874, 883, 917]]<|/det|>
+For high- resolution (scanning) transmission electron microscopy [HR(S)TEM] analysis, a series of Cu samples grown on \(\mathrm{Al}_2\mathrm{O}_3(0001)\) substrates at deposition times ranging from 15 to
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 82, 883, 323]]<|/det|>
+120 s were cross-sectioned for HR(S)TEM imaging using dual-beam focused ion beam (FIB) slicing (Helios NanoLab 450; FEI Co., Hillsboro, OR, USA) and lift-out processes. Double caesium (Cs)- corrected TEM (JEM- ARM200F; JEOL Ltd., Tokyo, Japan) at \(200\mathrm{kV}\) was used to obtain BF- STEM and HRTEM images of the Cu samples. For high- resolution STEM imaging, the probe- forming semi- angle was 23 mrad. Statistically random background noise in the HR(S)TEM images was reduced using the 2D difference filtering method of a commercial software program (HREM- Filters Pro; HREM Research Inc., Tokyo, Japan). Chemical analysis of the Cu samples was performed using electron energy loss spectroscopy (EELS) with a post- column- type electron energy loss spectrometry system (GIF Quantum ER 965; Gatan, Pleasanton, CA, USA) equipped on the microscope.  
+
+<|ref|>text<|/ref|><|det|>[[113, 329, 884, 640]]<|/det|>
+For structural characterisation, XRD measurements were performed using a PANalytical Empyrean Series 2 diffractometer (Malvern PANalytical, Malvern, UK) with a Cu- Kα source (40 kV, 30 mA). Data were collected in the range of \(20^{\circ} < 20 < 90^{\circ}\) , with a step size of \(0.0167^{\circ}\) and dwell time of 0.5 s per point in all cases. EBSD measurements were performed to confirm the quality of the thin films. A SUPRA40 VP SEM (Zeiss, Oberkochen, Germany) was used to measure EBSD maps of the thin films. The EBSD maps, pole figures (PFs) and IPFs show the directions and distributions of the crystals within the films. Surface roughness and sample thickness were measured by AFM using an XE- 100 instrument (Park Systems, Suwon- si, South Korea). The basic scanning conditions included noncontact mode with an \(\sim 0.5\mathrm{Hz}\) scan rate and \(1,024 \times 1,024\) resolution. The sample thickness was determined from an AFM image of Cu film scraped off using a sharp material. The scan rate was reduced to compensate for the stepped shape. The height resolution of the AFM instrument was 1.8 (0.25) Å in high- (low- )- voltage mode.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 673, 528, 691]]<|/det|>
+## Transport and optical characteristics of SCCFs  
+
+<|ref|>text<|/ref|><|det|>[[115, 697, 883, 790]]<|/det|>
+For transport measurements, sheet resistance (Rs: \(\Omega /\mathrm{sq}\) ) was measured in the van der Pauw geometry with an HMS- 3000 Hall measurement system (Ecopia, Toronto, ON, Canada) under a 0.55- T magnetic field at room temperature. Resistivity data \((\rho :\Omega \cdot \mathrm{cm})\) were obtained by multiplying sheet resistance by film thickness measured via AFM.  
+
+<|ref|>text<|/ref|><|det|>[[115, 796, 883, 911]]<|/det|>
+We investigated the optical quality of the ultrathin SCCFs using both infrared/optical spectroscopy and time- domain THz techniques. We measured the transmittance spectra of 12 Cu films of thicknesses ranging from 3 to \(30\mathrm{nm}\) at room temperature. To conduct measurements over a wide spectral range from THz to UV wavelengths, we used a commercial THz spectrometer (TERA K15; Menlo, Planegg, Germany) for the spectral range below 100
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 82, 883, 224]]<|/det|>
+\(\mathrm{cm}^{- 1}\) , a commercial FTIR- type spectrometer (Vertex 80v; Bruker, Karlsruhe, Germany) for the spectral range of \(400–25,000 \mathrm{cm}^{- 1}\) , and a commercial monochromatic spectrometer (Lambda 950; PerkinElmer, Waltham, MA, USA) for the spectral range from \(3,000 \mathrm{to} 50,000 \mathrm{cm}^{- 1}\) . Because the sapphire substrate has strong infrared- active phonons between \(150 \mathrm{and} 1,500 \mathrm{cm}^{- 1}\) , we could not obtain reliable spectra in this spectral range. However, no meaningful optical features of Cu films lie in this spectral range.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 255, 304, 272]]<|/det|>
+## Theoretical approach  
+
+<|ref|>text<|/ref|><|det|>[[113, 279, 883, 670]]<|/det|>
+All ab initio total energy calculations and geometry optimisations were performed using density functional theory (DFT) in the generalised gradient approximation (GGA) based on the Perdew- Burke- Ernzerhof functional \(^{36}\) and projected augmented plane- wave method, \(^{37}\) as implemented by Kresse et al. \(^{38}\) The \(\mathrm{Al}_{2}\mathrm{O}_{3}(0001)\) substrate was represented by a slab of 36 atomic layers of primitive unit cells containing 12 formula units, and the Cu thin film was represented by a slab of six layers of Cu atoms. The calculated lattice constants for bulk \(\mathrm{Al}_{2}\mathrm{O}_{3}\) are \(a = 4.785 \mathrm{\AA}\) and \(c = 13.06 \mathrm{\AA}\) , in good agreement with the experimental values. \(^{39}\) A vacuum length of \(15 \mathrm{\AA}\) was used, the bottom nine layers of the slab were fixed in their bulk positions, and the remaining atoms were fully relaxed until the Hellmann- Feynman force on each atom was \(< 0.001 \mathrm{eV / \AA}\) and the change in total energy was \(< 1 \times 10^{- 5} \mathrm{eV}\) . A supercell containing a slab of \(3 \times 2\) surface unit cells was used to simulate diffusion, and a slab of \(1 \times 1\) surface unit cells was used to calculate the adhesion energy of \(\mathrm{Cu}(111)\) on a \(\mathrm{Al}_{2}\mathrm{O}_{3}(0001)\) substrate. The electron wave functions were expanded in a plane- wave basis set with a cut- off energy of \(420 \mathrm{eV}\) , and Brillouin- zone integration for the slabs was performed using a \(5 \times 5 \times 1\) Monkhorst- Pack \(k\) - point grid. \(^{40}\) The nudged elastic band method \(^{41}\) was used to calculate the activation energy of diffusion with \(0.01 \mathrm{eV / \AA}\) of the force criterion for structure optimisation.  
+
+<|ref|>text<|/ref|><|det|>[[115, 675, 881, 717]]<|/det|>
+We modelled the growth of nanodroplets into a thin film on a substrate by considering the total energy \(E\) of the droplet on a substrate, as follows:  
+
+<|ref|>equation<|/ref|><|det|>[[380, 722, 866, 743]]<|/det|>
+\[E = -E_{c}V + \int \tau dS - \int E_{a}dS_{a}, \quad (1)\]  
+
+<|ref|>text<|/ref|><|det|>[[113, 750, 883, 917]]<|/det|>
+where \(E_{c}\) and \(\tau\) are the cohesive energy per unit volume and surface tension, respectively, of a droplet with volume \(V\) and surface area \(S\) , and \(E_{a}\) is the adhesive energy per unit area of the interface between the droplet and a substrate with area \(S_{a}\) . To simplify the analysis, we assumed that the droplet was a cylinder of radius \(a\) and height \(h\) . The surface energy term can be written as the sum of two terms: \(\int \tau dS = 2\tau_{\perp}S_{\perp} + \tau_{\parallel}S_{\parallel}\) , where \(\tau_{\perp}(\tau_{\parallel})\) is the average surface tension of the top (side) face of the droplet and \(S_{\perp} = \pi a^{2}(S_{\parallel} = 2\pi ah)\) is the area of the top (side) face. To determine the shape of the droplet for a given amount (volume) of adsorbate atoms, we
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[60, 82, 880, 129]]<|/det|>
+expressed the total energy in terms of the droplet height \(h\) and volume \(V\) using \(a = \sqrt{V / \pi} h^{- 1 / 2}\) , as follows:  
+
+<|ref|>equation<|/ref|><|det|>[[245, 133, 868, 163]]<|/det|>
+\[E = -E_{c}V + (2\tau_{\perp} - E_{a})\pi \big(\sqrt{V / \pi} h^{-1 / 2}\big)^{2} + \tau_{\parallel}2\pi \big(\sqrt{V / \pi} h^{-1 / 2}\big)h. \quad (2)\]  
+
+<|ref|>text<|/ref|><|det|>[[60, 168, 880, 210]]<|/det|>
+The sign conventions were selected such that all physical quantities were positive for typical substrates and adsorbates. For a given droplet of volume \(V\) ,  
+
+<|ref|>equation<|/ref|><|det|>[[310, 214, 722, 245]]<|/det|>
+\[\frac{\partial E}{\partial h} = (2\tau_{\perp} - E_{a})V(-h^{-2}) + \tau_{\parallel}2\sqrt{\pi}\sqrt{V}\left(\frac{1}{2} h^{-1 / 2}\right).\]  
+
+<|ref|>text<|/ref|><|det|>[[60, 252, 393, 279]]<|/det|>
+\(E\) is at its minimum when \(\frac{\partial E}{\partial h} = 0\) :  
+
+<|ref|>equation<|/ref|><|det|>[[392, 284, 868, 307]]<|/det|>
+\[\tau_{\parallel}\sqrt{\pi}\sqrt{V} h^{3 / 2} = (2\tau_{\perp} - E_{a})V. \quad (3)\]  
+
+<|ref|>text<|/ref|><|det|>[[60, 312, 636, 331]]<|/det|>
+When \(E_{a}< 2\tau_{\perp}\) , the ratio between the height and lateral size is:  
+
+<|ref|>equation<|/ref|><|det|>[[456, 336, 576, 368]]<|/det|>
+\[\frac{h}{a} = 2\frac{\tau_{\perp} - E_{a} / 2}{\tau_{\parallel}}.\]  
+
+<|ref|>text<|/ref|><|det|>[[115, 372, 881, 440]]<|/det|>
+Generalising this result to droplets of different shapes, and noting that \(2a\) represents the lateral size, the evolution of the droplet shape can be expressed in terms of the aspect ratio \(R = h / 2a\) , as follows:  
+
+<|ref|>equation<|/ref|><|det|>[[437, 444, 878, 485]]<|/det|>
+\[R = f\frac{\tau_{\perp} - E_{a} / 2}{\tau_{\parallel}} \qquad \text{for} E_{a}< 2\tau_{\perp} \quad (4a)\]  
+
+<|ref|>text<|/ref|><|det|>[[115, 490, 880, 510]]<|/det|>
+Where \(f\) is a form factor on the order of unity that depends on the specific shape of the droplet.  
+
+<|ref|>text<|/ref|><|det|>[[115, 515, 881, 584]]<|/det|>
+If \(E_{a}\geq 2\tau_{\perp}\) , the energy of the droplet in Eq. (2) is a monotonically increasing function of droplet height \(h\) , and minimising the energy causes \(h\) to approach zero (or \(a\) to increase limitlessly). Therefore,  
+
+<|ref|>equation<|/ref|><|det|>[[486, 590, 878, 610]]<|/det|>
+\[R\rightarrow 0 \qquad \text{for} E_{a}\geq 2\tau_{\perp}, \quad (4b)\]  
+
+<|ref|>text<|/ref|><|det|>[[115, 616, 803, 636]]<|/det|>
+which is interpreted as a lack of droplet formation, with only layer growth occurring.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 690, 262, 707]]<|/det|>
+## Data availability  
+
+<|ref|>text<|/ref|><|det|>[[115, 714, 832, 757]]<|/det|>
+The authors declare that the main data supporting the findings of this study are available within the article and its Supplementary Information files.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[58, 82, 214, 100]]<|/det|>
+## 414 References
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[57, 75, 884, 903]]<|/det|>
+448 Advances 3, 082105 (2013). 449 18. Lu, L., Shen, Y., Chen, X., Qian, L. & Lu, K. Ultrahigh Strength and High Electrical Conductivity in copper, Science 304, 422- 426 (2004). 450 19. Zhang, X. et al. Nanocrystalline copper films are never flat. Science 357, 397- 400 (2017). 451 20. Schiotz, J. & Jacobsen, K. W. Roughness in flatland. Nature Mater. 16, 1059- 1060 (2017). 452 21. Kim, S. J. et al. Flat- surface- assisted and self- regulated oxidation resistance of Cu (111). Nature 603, 434- 438 (2022). 453 22. Slavin, A. J. Growth modes of ultrathin metal films on dissimilar metal substrates. Prog. Surf. Sci. 50, 159- 172 (1995). 454 23. Williams, D. B. & Carter, C. B. Imaging. Transmission Electron Microscopy: A Textbook for Materials Science \(2^{\mathrm{nd}}\) ed. (Springer, 2009), pp. 369- 506. 455 24. Uragami, T. S. Pendellösung fringes in a finite crystal. J. Phys. Soc. Jpn. 13, 1141- 1161 (1971). 456 25. Tran, R. et al. Surface energies of elemental crystals. Sci. Data 3, 160080 (2016). 457 26. Li, L. et al. Surface- step- induced oscillatory oxide growth, Phys. Rev. Lett. 113, 136104 (2014). 458 27. Schmiedla, E., Wissmanna, P. & Finzelb, H.- U. The electrical resistivity of ultra- thin copper films. Z. Naturforsch. 63a, 739- 744 (2008). 459 28. Liu, H.- D., Zhao, Y.- P., Ramanath, G. S., Murarka, P. & Wang, G.- C. Thickness- dependent electrical resistivity of ultrathin (< 40 nm) Cu films. Thin Solid Films 384, 151- 156 (2001). 460 29. Fuchs, K. The conductivity of thin metallic films according to the electron theory of metals. Proc. Cambridge Philos. Soc. 34, 100- 108 (1938). 461 30. Sondheimer, E. H. The mean free path of electrons in metals. Adv. Phys. 1, 1- 42 (1952). 462 31. Namba, Y. Resistivity and temperature coefficient of thin metal films with rough surface. Jpn. J. Appl. Phys. 9, 1326- 1329 (1970). 463 32. Homes, C. C., Xu, Z. J., Wen, J. S. & Gu, G. D. Effective medium approximation and the complex optical properties of the inhomogeneous superconductor \(\mathrm{K}_{0.8}\mathrm{Fe}_{2 - y}\mathrm{Se}_2\) . Phys. Rev. B. 86, 144530 (2012). 464 33. Tanner, D. B. Classical theories for the dielectric function. Optical Effects in Solids (Cambridge Univ. Press, 2019), pp. 30- 42. 465 34. Schmiedla, E., Wissmanna, P. & Finzelb, H.- U. The electrical resistivity of ultra- thin
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[55, 79, 884, 500]]<|/det|>
+481 copper films. Z. Naturforsch. 63a, 739- 744 (2008). 482 35. Cho, Y. C. et al. Copper better than silver: Electrical resistivity of the grain- free single- crystal copper wire. Cryst. Growth Des. 10, 2780- 2784 (2010). 483 36. Perdew, J. P. et al. Generalized gradient approximation made simple. Phys. Rev. Lett. 78, 1396 (1997). 484 37. Blöchl, P. E. Projector augmented- wave method. Phys. Rev. B 50, 17953- 17979 (1994). 485 38. Kresse, G. & Joubert, D. From ultrasoft pseudopotentials to the projector augmented- wave method. Phys. Rev. B 59, 1758- 1775 (1999). 486 39. Izumi, F., Asano, H., Murata, H. & Watanabe, N. Rietveld analysis of powder patterns obtained by TOF neutron diffraction using cold neutron sources. J. App. Cryst. 20, 411- 418 (1987). 487 40. Monkhorst, H. J. & Pack, J. D. Special points for Brillouin- zone integrations. Phys. Rev. B 13, 5188- 5192 (1976). 488 41. Jonsson, H., Mills, G. & Jacobsen, K. W. Nudged elastic band method for finding minimum energy paths of transitions, Classical and Quantum Dynamics in Condensed Phase Simulations, ed. B. J. Berne, G. Ciccotti and D. F. Coker (World Scientific, 1998).  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 501, 314, 520]]<|/det|>
+## Acknowledgements  
+
+<|ref|>text<|/ref|><|det|>[[115, 528, 884, 694]]<|/det|>
+This research was supported by the National Research Foundation of Korea (NRF) (nos., NRF- 2022R1A2B5B03001219, NRF- 2020R1A4A4078780, NRF- 2019R1A6A1A11053838, 2021R1C1C1006316, 2019R1I1A1A01058304, 2016M3D1A1919181, 2020R1A2C1006207, and 2021R1A2C101109811), Institute for Basic Science (IBS- R011- D1) and by the Commercialization Promotion Agency for R&D Outcomes(COMPA) funded by the Ministry of Science and ICT(MSIT) (2022RMD- S08). Use of the TEM instrument was supported by the Advanced Facility Center for Quantum Technology at SKKU.  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 752, 333, 771]]<|/det|>
+## Author contributions  
+
+<|ref|>text<|/ref|><|det|>[[115, 778, 884, 920]]<|/det|>
+S.- Y.J., J.H., and Y.- M.K. conceived this study. S.J.K., S.E.P and Y.L. performed the Cu thin film growth and AFM, XRD, EBSD, SEM, and DC transport measurements. Y.H.K., Y.I.K., J.C.K., S.J.K., H.Y.J., and Y.- M.K. performed TEM measurements and analyses. T.H., T.T.K., K.I.S., and J.H.K. performed THz spectral measurements. Y.- S.S. and J.H. performed IR and optical measurements and EMA. S.- G.K. and B.L. carried out first- principles calculations. S.- Y.J. and Y.H.L. supervised the project. T.H., Y.- S.S., T.T.K., Y.- M.K., J.H., and S.- Y.J. wrote
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[60, 84, 880, 103]]<|/det|>
+515 the manuscript. All authors participated in the manuscript review.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 158, 321, 178]]<|/det|>
+## Competing interests  
+
+<|ref|>text<|/ref|><|det|>[[115, 187, 450, 204]]<|/det|>
+The authors have no competing interests.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 234, 353, 254]]<|/det|>
+## Additional information  
+
+<|ref|>text<|/ref|><|det|>[[115, 263, 783, 305]]<|/det|>
+Supplementary information The online version contains supplementary material available at https://doi.org./  
+
+<|ref|>text<|/ref|><|det|>[[115, 337, 870, 380]]<|/det|>
+Correspondence and requests for materials should be addressed to Seong- Gon Kim, Young- Min Kim, Jungseek Hwang or Se- Young Jeong.  
+
+<|ref|>text<|/ref|><|det|>[[115, 411, 589, 429]]<|/det|>
+Peer review information Nature Communications thanks  
+
+<|ref|>text<|/ref|><|det|>[[115, 460, 825, 479]]<|/det|>
+Reprints and permissions information is available at http://www.nature.com/reprints.  
+
+<|ref|>text<|/ref|><|det|>[[115, 508, 831, 551]]<|/det|>
+Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[44, 42, 311, 70]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[44, 93, 765, 113]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[60, 130, 368, 150]]<|/det|>
+- 2DCuSupplementaryInfoJSY.pdf
+
+<--- Page Split --->

preprint/preprint__103a21253ab7e11fb48f962b0d3ee89f4952881f79d599aa792239b2927b753e/images_list.json

@@ -0,0 +1,107 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Fig.1",
+    "footnote": [],
+    "bbox": [
+      [
+        88,
+        138,
+        680,
+        455
+      ]
+    ],
+    "page_idx": 26
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Fig.2",
+    "footnote": [],
+    "bbox": [
+      [
+        92,
+        90,
+        655,
+        410
+      ]
+    ],
+    "page_idx": 28
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Fig.3",
+    "footnote": [],
+    "bbox": [
+      [
+        81,
+        92,
+        666,
+        590
+      ]
+    ],
+    "page_idx": 30
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Fig.4",
+    "footnote": [],
+    "bbox": [
+      [
+        63,
+        88,
+        655,
+        430
+      ]
+    ],
+    "page_idx": 32
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Fig.5",
+    "footnote": [],
+    "bbox": [
+      [
+        83,
+        88,
+        669,
+        500
+      ]
+    ],
+    "page_idx": 34
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_6.jpg",
+    "caption": "Fig.6",
+    "footnote": [],
+    "bbox": [
+      [
+        95,
+        90,
+        625,
+        560
+      ]
+    ],
+    "page_idx": 36
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_7.jpg",
+    "caption": "Fig.7",
+    "footnote": [],
+    "bbox": [
+      [
+        160,
+        94,
+        574,
+        301
+      ]
+    ],
+    "page_idx": 38
+  }
+]

preprint/preprint__103a21253ab7e11fb48f962b0d3ee89f4952881f79d599aa792239b2927b753e/preprint__103a21253ab7e11fb48f962b0d3ee89f4952881f79d599aa792239b2927b753e.mmd

@@ -0,0 +1,695 @@
+
+# Epigenetic modifications regulate cultivar-specific root development and metabolic adaptation to nitrogen availability in wheat  
+
+Jun Xiao ( jxiao@genetics.ac.cn )  
+
+Institute of Genetics and Developmental Biology, Chinese Academy of Sciences https://orcid.org/0000- 0002- 6077- 2155  
+
+Hao Zhang Institute of Genetics and Developmental Biology, Chinese Academy of Sciences  
+
+Zhiyuan Jin College of Life Sciences, Hebei Normal University  
+
+Fa Cui College of Agriculture, Ludong University  
+
+Long Zhao  
+
+Xiaoyu Zhang  
+
+Institute of Genetics and Developmental Biology, Chinese Academy of Sciences  
+
+Jinchao Chen  
+
+Institute of Genetics and Developmental Biology, Chinese Academy of Sciences  
+
+Jing Zhang  
+
+Institute of Genetics and Developmental Biology, Chinese Academy of Sciences  
+
+Yanyan Li  
+
+Institute of Genetics and Developmental Biology  
+
+Yongpeng Li  
+
+Center for Agricultural Resources Research, Institute of Genetics and Developmental Biology  
+
+Yanxiao Niu  
+
+College of Life Sciences, Hebei Normal University  
+
+Wenli Zhang  
+
+Nanjing University https://orcid.org/0000- 0003- 0710- 1966  
+
+Caixia Gao  
+
+Institute of Genetics and Developmental Biology https://orcid.org/0000- 0003- 3169- 8248  
+
+Xiangdong Fu  
+
+State Key Laboratory of Plant Cell and Chromosome Engineering, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences https://orcid.org/0000- 0001- 9285- 7543  
+
+Yiping Tong
+
+<--- Page Split --->
+
+The State Key Laboratory for Plant Cell and Chromosome Engineering, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences  
+
+## Lei Wang  
+
+State Key Laboratory of Plant Genomics, Center for Agricultural Resources Research, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences  
+
+## HongQing Ling  
+
+Institute of Genetics and Developmental Biology, Chinese Academy of Sciences  
+
+## Junming Li  
+
+Center for Agricultural Resources Research, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences  
+
+## Article  
+
+## Keywords:  
+
+Posted Date: April 20th, 2023  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 2801336/v1  
+
+License: © © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Additional Declarations: There is NO Competing Interest.  
+
+Version of Record: A version of this preprint was published at Nature Communications on December 12th, 2023. See the published version at https://doi.org/10.1038/s41467-023-44003-6.
+
+<--- Page Split --->
+
+# Epigenetic modifications regulate cultivar-specific root development and metabolic adaptation to nitrogen availability in wheat  
+
+Hao Zhang \(^{1,3}\) , Zhiyuan Jin \(^{4,5,6}\) , Fa Cui \(^{2}\) , Long Zhao \(^{1,3}\) , Xiaoyu Zhang \(^{1,3}\) , Jinchao Chen \(^{1,3}\) , Jing Zhang \(^{1,3}\) , Yanyan Li \(^{1,4}\) , Yongpeng Li \(^{4,5,6}\) , Yanxiao Niu \(^{5,6}\) , Wenli Zhang \(^{8}\) , Caixia Gao \(^{1,3}\) , Xiangdong Fu \(^{1,3}\) , Yiping Tong \(^{1,3}\) , Lei Wang \(^{1,4}\) , Hong- Qing Ling \(^{1,3,7*}\) , Junming Li \(^{4,5,6,*}\) , Jun Xiao \(^{1,3,6,9,*}\)  
+
+\(^{1}\) State Key Laboratory of Plant Cell and Chromosome Engineering, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Beijing 100101, China \(^{2}\) Key Laboratory of Molecular Module- Based Breeding of High Yield and Abiotic Resistant Plants in Universities of Shandong, College of Agriculture, Ludong University, Yantai 264025, China \(^{3}\) University of Chinese Academy of Sciences, Beijing 100049, China \(^{4}\) Center for Agricultural Resources Research, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Shijiazhuang 050022, Hebei, China \(^{5}\) Ministry of Education Key Laboratory of Molecular and Cellular Biology, Hebei Research Center of the Basic Discipline of Cell Biology, Hebei Key Laboratory of Molecular and Cellular Biology, College of Life Sciences, Hebei Normal University, Shijiazhuang 050024, China \(^{6}\) Hebei Collaboration Innovation Center for Cell Signaling, Shijiazhuang, 050024, China \(^{7}\) Hainan Yazhou Bay Seed Laboratory, Sanya, Hainan, China. \(^{8}\) State Key Laboratory for Crop Genetics and Germplasm Enhancement and utilization, CICMCP, Nanjing Agricultural University, Nanjing, Jiangsu 210095, China \(^{9}\) Centre of Excellence for Plant and Microbial Science (CEPAMS), JIC- CAS, Beijing, China \(^{*}\) Correspondence to: hqling@genetics.ac.cn; ljm@sjziam.ac.cn; jxiao@genetics.ac.cn  
+
+## Abstract  
+
+The breeding of crops with improved nitrogen use efficiency (NUE) is crucial for sustainable agriculture. Despite its importance, the way in which epigenetic modifications regulate cultivar- specific responses to low nitrogen (LN) constraints is not yet well understood. Here, we analyzed the chromatin landscapes in the roots, leaves, and seeds of two wheat cultivars (KN9204 and J411) that differ radically in NUE under varied nitrogen conditions. The chromatin regions responsible for regulating gene transcription exhibited clear cultivar- specificity between the two cultivars, and the regulation of nitrogen metabolism genes (NMGs) was closely linked to variation in histone modification levels instead of differences in DNA sequence. We also found that cultivar- specific histone modification regions contribute to the genetic regulation of NUE- related traits, such as the QTL locus of maximum root length of qMRL- 7B. Additionally, LN- induced H3K27ac and H3K27me3 dynamics enhanced root growth more significantly in KN9204, while strengthened the nitrogen uptake system remarkably in J411. Evidence from histone deacetylase inhibitor treatment and transgenic plants with loss function of the H3K27me3 methyltransferase further showed that changes in epigenetic modifications can alter the strategy for root development and nitrogen uptake in response to LN constraint. Taken together, our data highlight the importance of epigenetic regulation in mediating cultivar- specific adaptation to LN in wheat.  
+
+Key words: epigenetic regulation; root architecture; nitrogen uptake; wheat; low nitrogen
+
+<--- Page Split --->
+
+## Introduction  
+
+Improving nitrogen- use efficiency (NUE) has emerged as a pressing requirement for sustainable agriculture'. NUE is mainly composed of two components: N uptake efficiency (NUpE) and N utilize efficiency (NUte). NUpE pertains to the acquisition of nitrogen from the soil, while NUte signifies the yield generated per unit of nitrogen obtained. In situations where nitrogen availability is limited for crops such as winter wheat?, barley?, studies indicate that NUpE holds greater importance in determining NUE compared to NUte. To enhance both NUE and grain yield, specifically under low nitrogen conditions, it is imperative to boost nitrogen uptake, which is predominantly regulated by nitrate transporters. The concentration of nitrate in soil can experience substantial fluctuations due to external environmental factors; as a result, root system architecture plays a vital role in the efficient absorption of nitrate. Numerous efforts have been undertaken to assist plants in acquiring adequate nitrogen under such conditions, such as overexpressing NRT2 (a high affinity nitrogen transporter)4,5, and refining root architecture6,7. Nevertheless, further research is necessary to comprehensively understand the balance and coupling between root architecture response and nitrogen transporter expression, particularly when operating under low nitrogen conditions.  
+
+Kenong 9204 (KN9204) and Jing 411 (J411) are cultivars that exhibit diverse agricultural traits such as nitrogen use efficiency (NUE), root architecture, and productivity under low nitrogen conditions?. By utilizing these superior materials, we successfully identified several NUE- related quantitative trait loci (QTLs), encompassing factors such as root length, root tips, and grain protein content, derived from recombinant isogenic lines (RILs) generated through a cross between KN9204 and J411 in previous studies8- 10. We recently completed the reference genome sequence of KN9204 and identified 882 nitrogen metabolism genes (NMGs)11. Comparative transcriptome analysis between KN9204 and J411 revealed different responsive programs under low nitrogen constraint, especially in the yield- related spike tissue and in seeds during reproductive developmentl. The underground tissues (roots) are not as well investigated. Past research emphasizes the importance of the root architecture system (RSA) for NUE12 and the significant differences of RSA that exist between KN9204 and J4112,8. However, the regulatory mechanism underlying the diverse transcriptional programs linked to NUE in the roots of KN9204 and J411 still remains to be elucidated.  
+
+Recent studies have emphasized the crucial role that epigenetic factors play in regulating both nutrition uptake and metabolism, as well as the synergistic plant response to nutrient availability13. For example, the SET DOMAIN GROUP 8 (SDG8) gene is involved in regulating nitrogen (N) assimilation and lateral root response, by controlling the levels of H3K36me3 in response to changing nitrogen levels in Arabidopsis14. Similarly, the HISTONE DEACETYLASE 19 (HDA19) gene is responsible for controlling root cell elongation and modulating the expression of phosphorus (Pi)- homeostasis genes in Arabidopsis under phosphate starvation conditions15. In rice, the polycomb repressive complex 2 (PRC2), which is recruited by NITROGEN- MEDIATED TILLER GROWTH RESPONSE 5 (NGR5), regulates tillering by depositing H3K27me3 on branching- inhibitor genes16. Additionally, the H3K27me3 level at the AtNRT2.1 (nitrate transporter) locus influences root nitrate uptake, a process that is mediated by HIGH NITROGEN INSENSITIVE 9 (AtHNI9) and PRC217. Despite these advances in understanding the role of epigenetics in nutrient uptake and metabolism of plants like Arabidopsis and rice, the co- regulation of these processes and their response to nutrient availability in wheat remains largely unexplored.  
+
+In this study, we used the CUT&Tag18,19 to profile epigenomic modifications in two wheat cultivars, KN9204 and J411, for various histone modifications and histone variant in three different tissues
+
+<--- Page Split --->
+
+under varying nitrogen conditions. Our epigenomic maps offer a comprehensive understanding of the dynamic chromatin landscapes of these two wheat cultivars in response to low nitrogen conditions. By integrating our findings with previous QTL analysis, we also identified cultivar- specific transcription bias and epigenome divergence, particularly in the H3K27ac modification. Moreover, manipulating H3K27ac and H3K27me3 profiles through chemical inhibition or genetic mutation can significantly influence the LN adaptation strategy of different wheat cultivars.  
+
+## Results  
+
+## Profiling the tissue-specific chromatin landscape under different nitrogen conditions  
+
+To understand the epigenetic regulation of the transcriptomic dynamics (Fig. S1a), we did CUT&Tag of various histone modifications for the wheat cultivars KN9204 and J411 at normal nitrogen (NN) and LN conditions for roots (28 days), flag leaves (heading stage) and seeds (21 days after anthesis) (Fig. 1a), corresponding to transcriptomes generated previously<sup>11</sup>. The CUT&Tag assay showed reproducibility between two biological replicates for the various histone modifications (Fig.S1b). The majority of histone modification peaks were located in the distal region expected for H3K36me3 in wheat (Fig. S1c), as described in our recent report<sup>19</sup>. H3K27ac and H3K4me3 were associated with highly expressed genes, while H3K27me3 is enriched in low/non- expressed genes (Fig. S1d). H3K4me3 and H2A.Z did not show a preference for gene expression level (Fig. S1d). More than half of the H3K9me3 peaks were located in transposable elements (TEs) regions (Fig.S1e). Genes involved in biotic and abiotic stress responses had a higher H2A.Z level relative to that in developmental and hormone- related genes (Fig.S1f).  
+
+The chromatin state was systematically defined using ChromHMM<sup>20</sup>, which integrated combinatorial patterns of various histone marks (Fig. 1b). Five major categories were formed from the fifteen chromatin states (CS) identified, including Promoter (CS1- 4), Transcription (CS5- 8), Enhancer- like (CS9- 11), Repressive (CS12- 14), and No signal (CS15), each with different genome coverage, TE enrichment, and gene expression level (Fig. 1b). Both the Promoter and Enhancer- like states were associated with H3K27ac, H3K4me3, and H3K27me3, but located in the transcription start site (TSS) and intergenic region, respectively (Fig. 1b). Repressive states were mainly enriched with H3K27me3, covering approximately \(10\%\) of the genome, while the no signal state accounted for a major portion ( \(\sim 83\%\) ) of the genome. Therefore, a limited portion of the genome ( \(\sim 7\%\) ) is transcribed or involved in transcriptional regulation in the context of various histone modifications (Fig. 1b).  
+
+We wonder how chromatin state dynamics reflect the differences between wheat cultivars, tissue types, and nitrogen conditions. For example, genes related to nitrate uptake in the root (TaNRT2_3A), assimilation in the leaf (TaNIA_6D), and amino acid storage in the seed (TaPROT2_4D) showed varied CS under LN condition in KN9204 (Fig. 1c). Generally, Enhancer- like CS was the most variable chromatin region between the different cultivars, while Promoter CS was primarily influenced by nitrogen availability (Fig. 1d). CS10 and CS9 showed the highest frequency of change between wheat cultivars, while CS10 and CS2 exhibited greater variability under different nitrogen availabilities (Fig. S1g). Furthermore, we calculated variability scores for different histone marks and found that H3K27ac and H3K27me3 exhibited
+
+<--- Page Split --->
+
+the highest variability scores among the different nitrogen conditions and between the wheat cultivars (Fig. 1e). Further examination of the distribution patterns (Fig. S1c) allowed us to categorize these histone marks into distal and promoter peaks and evaluate their variability using Pearson correlation analysis (Fig.1f, Fig.S1h). We observed that distal H3K27ac showed cultivar- specificity, while promoter H3K27ac showed greater tissue- specificity (Fig. 1f). For H3K27me3, both distal and promoter peaks exhibited cultivar- specificities (Fig. S1h). Therefore, distinct genomic regions with variable chromatin states are shaped by differences in wheat cultivars, tissue types, and nitrogen conditions, particularly with regards to H3K27ac and H3K27me3 marks.  
+
+## Cultivar-bias expression of NMGs is mainly mediated by histone modification variations  
+
+NMGs are crucial for plants to uptake and utilize nitrogen, including genes that encode nitrate transporter (NPF, NRT2, NAR2, CLC, and SLAH), nitrate reductase (NIA, NIR), ammonium transporter (AMT), and amino acid transporter (APC) family members, as well as enzymes involved in ammonium assimilation (GS, GOGAT, GDH, ASN, AspAT) and transcription factors (TFs) related to N metabolism \(^{21,22}\) . Previously, we have identified a total of 882 NMGs in wheat through sequence similarity comparisons to six other plant species including Brachypodium distachyon, barley, rice, sorghum, maize, and Arabidopsis \(^{11}\) .  
+
+We compared the NMGs between KN9204 and J411 for variations in DNA variation and H3K27ac modifications in promoter regions, and changes in transcriptional level. The results revealed that about \(25\%\) of NMGs displayed different levels of expression between KN9204 and J411 (FDR \(< 0.05\) , fold- change \(\geq 1.5\) ), while only around \(5\%\) of NMGs had DNA sequence variations in the promoter regulatory regions (Fig. 2a). The variation in expression of NMGs between KN9204 and J411 existed for different gene families in various tissues (Fig. 2b, Supplemental Table S1). For instance, in roots, the expression levels of NRT2 and NIA in J411 were both higher under the LN condition relative to KN9204, while the NPF family genes was activated in KN9204 (Fig. S2a). Moreover, in flag leaves, NIA genes were up- regulated in KN9204, but not in J411. However, in seeds, the relative expression of genes that encode NPF and GS was much higher in KN9204 compared to J411 (Fig. S2a).  
+
+The bias expression of NMGs in different cultivars is unlikely to be affected by variations in the DNA sequence within their regulatory regions as only 16 genes overlapped (Fig. 2c, top). Rather, the majority of NMGs that showed cultivar- biased expression were marked by differential peaks of H3K27ac \((66.9\%)\) and H3K27me3 \((45.6\%)\) (Fig. 2c, Fig. S2b, Supplemental Table S1). In roots, the NRT2 family displayed similar changes in H3K27ac and H3K27me3 under LN/NN conditions in both cultivars but were more pronounced in J411 compared to KN9204 (Fig. 2d, top). For example, at the TaNRT2_6A (TraesCS6A02G031000) locus, H3K27ac gains and H3K27me3 losses in response to LN occurred in both cultivars but were more dramatic in J411 (Fig. 2e, top). This also aligns with the higher level of induced expression of TaNRT2_6A in J411 under LN conditions (Fig. 2e, top). TaNPF2.3, involved in nitrite transport from the root to the shoot \(^{23}\) , was activated in KN9204 with a decrease H3K27me3 under LN while it was accompanied by a significant decrease in H3K27ac in J411 (Fig. 2d, bottom). Consistently, TaNPF2.3_7B was increased in KN9204 but decreased in J411 under LN/NN conditions (Fig. 2e, bottom). Additionally, the nitrate contents in shoots relative
+
+<--- Page Split --->
+
+to roots were higher in KN9204 compared to J411 under LN conditions rather than NN conditions (Fig. 2f). Similarly, cultivar- specific H3K27ac dynamic in response to LN was associated with expression change of ammonium assimilation enzymes coding genes GS and GOGAT in seeds (Fig. S2c, d). Accordingly, KN9204 seeds have a higher protein content compared to J411 under LN conditions (Fig. S2e). Taken together, the varied levels of H3K27ac and H3K27me3 were associated with the expression bias of NMGs, leading to different nitrogen metabolism processes in KN9204 and J411.  
+
+## Cultivar-specific H3K27ac influences NUE-related traits by transcriptional regulation  
+
+Given the importance of H3K27ac in regulating the chromatin state and expression patterns of NMGs across different tissues and cultivars at LN/NN, we extended to identify cultivar- specific H3K27ac regions using K- means clustering (Fig. 3a). In general, the cultivar- specific H3K27ac peaks were primarily over- represented in distal regions while under- represented in promoter and genic regions when compared to all H3K27ac peaks (Fig. 3b). Moreover, cultivar- specific H3K27ac- marked promoters had close associations with cultivar- specific expressed genes in roots, leaves, and seeds (Fig. 3c, Fig. S3a, b). For instance, H3K27ac- marked genes in KN9204 were associated with cell wall biogenesis and nutrient reservoir activity (e.g., PROT1, LBD16, XTH19, CSLC5) (Fig. 3c, Fig. S3c), while H3K27ac specifically modulated genes involved in flavonol biosynthesis in J411 (Fig. S3c).  
+
+The H3K27ac peaks in the distal region, specific to each cultivar, were found to be in the same location as H3K4me3 and H2A.Z (Fig. S3d), suggesting that these epigenetically modified hotspots serve as 'enhancers' in regulating gene expression<sup>24- 26</sup>. A correlation between the H3K27ac and gene expression dynamics allowed us to assign these regions to potential targets within a 500 kb distance as reported previously<sup>27,28</sup> (Fig. S3e). A total of 58,493 unique pairs (22,003 distal H3K27ac regions and 6,357 target genes) between the distal cultivar- specific H3K27ac peaks and genes were identified (Fig. 3d), and the accuracy of the pairs was supported by a higher chromatin interaction ratio from published Hi- C data<sup>29</sup> (Fig. S3f). GO annotation of KN9204- specific distal H3K27ac- regulated genes showed enrichment in genes for cell wall synthesis, protexylem development, and nutrient reservoir activity, which is similar to the KN9204- specific promoter H3K27ac- marked genes. Among them, several genes are known to be involved in nutrient transport to the shoot, such as TaVND<sup>7</sup><sup>30</sup>, TaXTH<sup>25</sup><sup>31</sup>. No GO category was enriched in targets of J411- specific distal H3K27ac (Fig. 3d).  
+
+We examined the impact of cultivar- specific regulatory regions on DEGs located within QTLs identified via linkage analysis in a KN9204- J411 RIL population<sup>8- 10</sup>. In comparison to DNA sequence variations within promoter regions, cultivar- specific distal H3K27ac regions exhibit a stronger correlation with DEGs located within QTL regions such as nitrogen uptake (Nup), nitrogen concentration (Nct), maximum root length (MRL), and grain protein content (GPC) (Fig. S3g). We noticed a significant enrichment of target genes of KN9204- specific H3K27ac peaks in QTLs that consisted of NUE related- traits, wherein the elite genetic loci originated from KN9204 (Fig. 3e, Supplemental Table S2). Whereas, the J411- specific H3K27ac region associated genes were enriched in QTLs linked to leaf size- related traits (Fig. 3e). For example, while examining the previously discovered qMRL- 7B, we discovered 1,245 genes within the genetic region, of which 69 were DEGs, and only nine genes contained DNA sequence
+
+<--- Page Split --->
+
+variations within the promoter regions. In contrast, 34 genes had cultivar- specific distal H3K27ac peaks (Fig. 3f). We observed the higher expression levels of TraesCS7B02G317800 (TaHyPRP06_6B) and TraesCS7B02G326900 (TaXTH25_7B) in KN9204 compared to J411, which were correlated with KN9204- specific H3K27ac distal regulatory regions and were involved in regulating root development (Fig.3g, Fig. S3h). Furthermore, we confirmed the functional potential of these cultivar- specific regulatory regions through a luciferase reporter assay32 (Fig.3h, Fig. S3i). Thus, both promoter and distal cultivar- specific H3K27ac regions play a significant role in transcriptional regulation and contain genetic variations that are responsible for NUE- related traits.  
+
+## LN-induced H3K27ac dynamics affect divergent adaptive programs in KN9204 and J411  
+
+We further investigated the role of H3K27ac dynamics induced by LN in driving divergent adaptations between KN9204 and J411. H3K27ac shows varied dynamic patterns in different tissues under LN/NN conditions in KN9204 and J411 (Fig. 4a). KN9204 showed subtle changes in H3K27ac levels in roots, but more pronounced changes in flag leaves and seeds. Whereas J411 exhibited a significant loss and relatively lower gains in root but minor changes in above- ground tissues (Fig. 4a). Additionally, K- means clustering identified distinct H3K27ac dynamic patterns across different tissues in KN9204 and J411 (Fig. 4b, Fig. S4a).  
+
+Extensive loss of H3K27ac in roots of J411 (clusters C3, C5, and C6, \(\mathrm{n} = 105,261\) ) mainly located in distal regions (Fig. 4b, Fig. S4b). Whereas, gain of H3K27ac (clusters C4, C7, and C8, \(\mathrm{n} = 49,527\) ) were much less distributed in distal regions (Fig. 4b, Fig. S4b). Of note, many LN- induced H3K27ac- loss loci are specific to J411, but were maintained in KN9204 (C5 and C6 clusters). Such loss of H3K27ac in proximal regions (promoter and genic regions) caused down- regulation of genes that function in auxin homeostasis, cytokinin metabolism, and hormone signaling in J411 under LN/NN conditions (Fig. 4c). The gain of H3K27ac is linked to the activation of genes responsible for nitrate uptake, transport and assimilation, including TaNRT2 and TaNIA (Fig. 4c). In contrast, gain- of- H3K27ac in roots of KN9204 activated genes involved in cytokinin biosynthesis, auxin polarity transport, transporters for carbohydrate and organic cations, such as TaD27, TaPILS7, and TaLOG4 (Fig. 4c), which function in root growth under LN as reported33,34. Loss of H3K27ac in KN9204 led to down- regulation of genes involved in phosphate ion transport. Consistently, the root system of J411 has a lower response to LN as compared to KN9204, with more root tips and larger root diameters under LN conditions (Figs. 4d, e). This eventually translated into an increase in root surface area and a larger total root volume in KN9204. Thus, the dynamic changes in H3K27ac that occur in response to LN preferably enhanced root growth in KN9204 while they strengthened the nitrogen uptake system in root of J411. In KN9204 flag leaves, increased H3K27ac were found in genes related to sucrose metabolism and xylem development, while J411 had more of such genes related to reactive oxygen species response (Fig. S4c). In seeds, H3K27ac loss caused by LN occurred mainly in distal regions, thus have little influence on gene expression in both cultivars (Fig. S4d).  
+
+Whether the root morphological change in KN9204 and J411 under LN/NN is regulated by cultivar- biased LN- induced H3K27ac status? To determine this, we tested the effects of Trichostatin A (TSA), a chemical inhibitor of class I and II histone deacetylases (HDAC)35, on
+
+<--- Page Split --->
+
+the H3K27ac pattern in wheat seedlings using a hydroponic culture system (See method for details). The effects of TSA treatment were validated by western blotting (Fig. S4e). Root growth was induced by LN in KN9204 in both the TSA- treated and the untreated group, as indicated by increased root tip numbers under LN compared to NN conditions (Fig. 4f). In J411 under the mock condition, root growth was not induced by LN as expected, but root tip numbers increased substantially by LN with TSA treatment (Fig. 4f). Therefore, TSA treatment inhibited the extensive LN- induced loss of H3K27ac in J411, which restored the root growth response of J411 to LN conditions.  
+
+## H3K27me3 shapes distinct root developmental programs in KN9204 and J411 under LN  
+
+In addition to H3K27ac, H3K27me3 also showed varied dynamic patterns in roots, flag leaves, and seeds in response to LN in KN9204 and J411(Fig. 5a). In both KN9204 and J411, subtle changes in H3K27me3 occurred in flag leaves, while there were a large number of differential H3K27me3 regions in roots in J411 but in seeds in KN9204 (Fig. 5a). K- means clustering further identified different categories of dynamic H3K27me3 regions in roots and seeds under LN conditions for KN9204 and J411 (Fig. 5b, Fig. S5a).  
+
+The majority of the dynamic H3K27me3 regions were situated distally in both roots and seeds (Fig. S5b). In roots, the gain of H3K27me3 prevails compared to the loss in J411, for both proximal and distal regions (Fig. 5c), while for KN9204 there was more loss of H3K27me3 than gain in both regions, albeit with a smaller peak count than J411 (Fig. 5c). Additionally, genes marked with LN- induced gain- of- H3K27me3 (clusters C4, C5, and C7 in Fig. 5b) significantly overlapped with down- regulated genes in J411 (Fig. 5d, top). These genes are enriched in root growth- related processes, including auxin biosynthesis, regulation of cell differentiation, and root morphogenesis (Fig. 5e, top). For example, the gene for a Mob1- like transcription factor involved in root development<sup>36</sup> has increased H3K27me3 under LN conditions in J411 and the expression level is reduced, while no significant change in KN9204 (Fig. 5f, top). Conversely, the gain- of- H3K27me3 in KN9204 only overlapped with 24 genes that were down- regulated by LN (Fig. S5c), implying that LN- induced gain- of- H3K27me3 in roots might mainly reduce the root growth process in J411 but not so much in KN9204. Loss- of- H3K27me3 in KN9204 and J411 produced significant overlaps with genes where expression increased in response to LN (Fig. 5d, middle and bottom). Genes involved in the response to nitrite, nitrite transport, and nitrate assimilation were enriched in J411, while genes involved in nitrate transport and SL and GA biosynthetic processes, as well as primary cell wall biogenesis- related genes, were enriched in KN9204 (Fig. 5e, middle and bottom). For example, expression of TaNAR2_4A and TaKAO2_4A was activated separately in J411 and KN9204 with decreased H3K27me3 (Fig. 5f, middle and bottom). Thus, The LN- induced H3K27me3 loss in roots tends to activate root growth in KN9204. Conversely, in J411, the loss of H3K27me3 activates nitrite uptake and metabolism. In the seeds, gain or loss of H3K27me3 in KN9204 did not lead to much change in overall gene expression (Fig. S5d). The same is true for gain- of- H3K27me3 in J411 (Fig. S5d). However, loss- of- H3K27me3 induced a significant amount of up- regulation of gene expression in J411, though no specific GO term was enriched (Fig. S5d).  
+
+The deposition of H3K27me3 in plants depends on the recruitment of Polycomb repressive complex 2 (PRC2) by different DNA recognition factors<sup>37,38</sup>. To understand the drivers behind
+
+<--- Page Split --->
+
+the dynamic H3K27me3 landscape in two plant varieties (KN9204 and J411), we conducted a motif scanning analysis, which found that 'CGCCGCC' motif (50%) and 'GAGAGA' repeat (12.9%) were enriched in KN9204- and J411- specific dynamic H3K27me3 regions, respectively (Fig. 5g). Interestingly, these motifs were also present in Polycomb response element (PRE) fragments in Arabidopsis \(^{37,39}\) . Based on transcription factor (TF)- DNA recognition in other species \(^{40}\) , TaERF9_5B/TaERF9_5D and TaBPC_4A were selected for interaction tests with PRC2 components. Using yeast two- hybrid (Y2H) and bimolecular fluorescence complementation (BiFC) assays, we verified that TaERF9_9B/TaERF9_5D (AP2/ERF family) and TaBPC_4A could interact with EMF2 and SWN (subunits of PRC2) in vivo (Fig. 5h, i). Notably, BPC had been reported to interact with SWN and influence root development in Arabidopsis \(^{41}\) . Thus, different TFs likely mediate LN- induced divergent H3K27me3 changes between KN9204 and J411, resulting in different root development routes.  
+
+## Rewiring H3K27me3 modulates root development and nitrate uptake in response to LN  
+
+We investigated whether H3K27me3 plays a critical role in root development under LN conditions. There are nine genes encode H3K27me3 methyltransferases within three triads (Fig. S6a), showing varied expressions in roots under different N conditions (Fig. S6b). Considering the high expression of TaSWN in roots (Fig. S6b), we used the CRISPR- Cas9 system to create knock- out mutants of TaSWN (TraesCSA02G121300, TraesCS4B02G181400, and TraesCS4D02G184600) in the 'Bobwhite' (BW) background (Fig. S6c). Sequencing of transgenic wheat identified a Taswn- cr homozygous line with frameshift mutations in all three subgenome copies of TaSWN (Fig. S6c). We then used CUT&Tag to compare the genome- wide H3K27me3 patterns in the Taswn- cr and BW. The peak number and length of H3K27me3 in Taswn- cr was significantly decreased compared to BW (Fig. S6d). Similarly, the intensity of H3K27me3 on coding genes was decreased in Taswn- cr compared to BW (Fig. S6e). Therefore, TaSWN is indeed a H3K27me3 writer for some genomic areas in wheat.  
+
+Next, we profiled H3K27me3 patterns in BW and Taswn- cr to identify TaSWN- mediated H3K27me3 deposition in response to LN. K- means clustering identified different categories of dynamic H3K27me3 regions in BW and Taswn- cr under NN or LN conditions (Fig. 6a). Of note, the dynamic H3K27me3 peaks were predominantly located in distal regions, especially for clusters 1, 3, 6, and 7 (>80%) (Fig. S6f). Among them, clusters 1 and 6 showed reduced H3K27me3 in Taswn- cr compared to BW specifically under LN or both LN and NN conditions (Fig. 6a), which indicated a TaSWN- dependent manner. By overlapping with the LN- induced H3K27me3 peaks in KN9204 and J411 (Fig. 5b), we found that 10%- 16% (n = 1,748, 14,094 separately) of the LN- induced H3K27me3 peaks in KN9204 or J411 were mediated by TaSWN (Fig. 6b). There were also more genes influenced by SWN in J411 (n = 959) compared to KN9204 (n = 86) (Fig. 6c, Fig. S6g), suggesting that H3K27me3 regulation has more weight in J411 than in KN9204. Furthermore, we identified a set of genes which gained H3K27me3 in J411 under LN/NN conditions but lost H3K27me3 in Taswn- cr compared to BW (Fig. 6d). This group of genes is involved in the root development process; examples are MADS15, bHLH068, and MYB36 (Fig. 6d, e). The expression of these genes was induced in Taswn- cr under LN/NN, but no changes in BW (Fig. 6f). These results hint that TaSWN- mediated LN- induced gain- of- H3K27me3 in J411 partially reduces the root growth response to LN conditions. Therefore, we speculated that root growth in Taswn- cr plants would be more responsive to LN than BW.
+
+<--- Page Split --->
+
+Indeed, we found roots were more developed in Taswn- cr compared to BW under LN, as determined by total root length and the number of root tips (Fig. 6g, h). In addition, the relative level of induction of NRT2 expression in response to LN in Taswn- cr was lower than BW (Fig. 6i, left), which were similar to KN9204 cultivar with more developed roots (Fig. 6i, right). Consistently, the nitrate uptake rate was higher in BW compared to Taswn- cr as measured by \(^{15}\mathrm{N}\) uptake assay under LN condition (Fig. 6j, left), which similar to the trend of J411 and KN9204 (Fig. 6j, right).  
+
+The comparison of root morphological changes, nitrate uptake rate and transcriptional profiles in response to LN between BW and Taswn- cr highlighted that H3K27me3 plays important role in balancing root growth and nitrogen metabolism under LN constraint. Rewiring H3K27me3 could influence wheat cultivars for the decision- making between significantly enhancing root growth or remarkably strengthening the nitrogen uptake system to adapt to low nitrogen environments.  
+
+## Discussion  
+
+As the urgency to reduce nitrogen fertilizer application in crop production, considerable effort has been directed towards dissecting the genetic basis of NUE regulation in crops<sup>42- 44</sup>. However, epigenetic regulation, which functions in coordinating with transcription factors to manipulate gene expression, is not well studied in wheat. To fill this gap, we generated epigenomic datasets for three different tissues in two wheat cultivars that differ with respect to NUE (KN9204 and J411) under different nitrogen conditions (Fig. 1). Our analysis revealed that the epigenome, which varies more than DNA sequence variation, plays an important role in mediating the cultivar- specific low nitrogen response.  
+
+## Epigenetic variation contributes to cultivar-specific trait formation  
+
+Epigenetic modifications, especially H3K27me and H3K27ac, mediate transcriptional dynamics and contribute to different developmental programs or nitrogen metabolic processes between cultivars (Figs. 1, 2). Bias- expressed NMGs are associated with altered epigenetic regulation patterns rather than DNA sequence variations between KN9204 and J411 (Fig. 2). In addition, distal regulatory regions (H3K27ac and H3K27me3) clearly reflect cultivar specificity, with higher DNA sequence variations, and are associated with previously identified NUE- related QTLs (Fig. 3). In maize, many distal regulatory regions have been reported to regulate gene expression and are associated with agronomic trait variations<sup>45,46</sup>. Indeed, genes associated with cultivar- specific H3K27ac either in the promoter or in the distal region are enriched in QTL regions for mediating NUE- related agronomic traits, such as MRL, GPC, and Nup (Fig. 3), which could be good candidates for mediating the genetic difference between KN9204 and J411. Regarding the important role of epigenetic regulatory regions, more cultivars with defined NUE features could be used to profile the epigenome, especially H3K27ac and H3K27me3 in the future, which would enable Epi- GWAS analysis to uncover NUE regulatory mechanisms.  
+
+## Epigenetic regulation balances root growth and nitrogen metabolism under LN constraint  
+
+To absorb adequate nitrogen in nitrogen- limiting conditions, cultivars with different NUE features have various strategies, such as triggering root growth to have more root tips and
+
+<--- Page Split --->
+
+increase the total root system volume in KN9204, or powering up the nitrate uptake machinery via up- regulation of NRT2 transporters in J411 (Fig. 1). Interestingly, the different adaptive strategies in roots between KN9204 and J411 are correlated with changes in the dynamic epigenome, especially for H3K27ac and H3K27me3 (Figs. 4, 5). Gain- of- H3K27ac and loss- of- H3K27me3 coordinately enhance the expression of root development- related genes in KN9204 under LN conditions, whereas loss- of- H3K27ac and gain- of- H3K27me3 reduce root development in J411 under LN, but rather activate nitrate uptake transporters via gain- of- H3K27ac and loss- of- H3K27me3 (Fig. 7). Several transcription factors show the potential to establish such epigenetic modification specificity via the recognition of certain cis- acting motifs and recruiting histone modification writers or erasers, such as ERFs and BPCs (Fig. 5). Thus, precise epigenetic modification modulates root development and nutrient absorption, which is likely to be a general mechanism for balancing plant growth and environmental stimulus response or adaptation<sup>47</sup>. However, it is presently unknown how such epigenetic modification behaves in a cultivar- specific manner. An interesting finding is that the potential histone writer or eraser- guiding TFs are located within the QTL regions linked to the MRL and Nup traits (Supplemental Table 3). Further analysis would elucidate how LN can trigger different response patterns in those TF genes in cultivars that vary with respect to NUE.  
+
+## Manipulating epigenetic regulation to decouple root growth and nitrogen metabolism for NUE improvement in wheat  
+
+In addition to the correlation between cultivar- specific epigenetic dynamics and varied strategies in response to LN, we have shown that manipulating the epigenetic features could affect the strategy selection for LN adaptation. Chemical inhibition (TSA treatment) or genetic manipulation (Taswn- cr) that changes the epigenome landscape (H3K27ac and H3K27me3) could lead to altered root system development and coordinated NRT2 induction intensity under LN constraint (Figs. 4, 6). Similarly, several reports have shown that adjusting histone modification contributes to the response and/or tolerance to stress, such as HDA6- regulated salt stress<sup>48</sup>, and JMJ1- regulated dehydration stress<sup>49</sup>. Interestingly, the enhancement of root growth is likely coupled with the attenuated induction of the expression of nitrate transporter coding genes under LN constraint by histone modification, in particular H3K27me3 (Fig. 7). Theoretically speaking, precise epigenomic modification alterations at specific regions/genes could help to decouple such linkage, which could generate wheat with developed root architecture system and higher induction of nitrogen transporters simultaneously. To achieve this, instead of the genetic manipulation of the "writer" or "eraser" to histone modification, fine- tuning of the driver (for example ERFs) may serve as a better way of coordinating nitrogen metabolism and adaptive root growth to low nitrogen constraints in wheat cultivars.  
+
+## Materials and methods  
+
+## Plant materials and culture conditions  
+
+The root of KN9204 and J411 was harvested 4 weeks after transplanting in the nutrient solution, which corresponding to 28- day in previous study<sup>11</sup>, and immediately frozen in liquid nitrogen and stored at \(- 80^{\circ}\mathrm{C}\) . The nutrient solution for NN was as follows: \(1\mathrm{mM}\mathrm{Ca(NO_3)_2}\) , \(0.2\mathrm{mM}\) \(\mathrm{KH_2PO_4}\) , \(0.5\mathrm{mM}\mathrm{MgSO_4}\) , \(1.5\mathrm{mM}\mathrm{KCl}\) , \(1.5\mathrm{mM}\mathrm{CaCl_2}\) , \(1\times 10^{- 3}\mathrm{mM}\mathrm{H_3BO_3}\) , \(5\times 10^{- 5}\mathrm{mM}\) \((\mathrm{NH_4})_6\mathrm{Mo_7O_{24}}\) , \(5\times 10^{- 4}\mathrm{mM}\mathrm{CuSO_4}\) , \(1\times 10^{- 3}\mathrm{mM}\mathrm{ZnSO_4}\) , \(1\times 10^{- 3}\mathrm{mM}\mathrm{MnSO_4}\) , \(0.1\mathrm{mM}\)
+
+<--- Page Split --->
+
+Fe(III)- EDTA. For LN, \(0.02\mathrm{mM Ca(NO_3)_2}\) , \(2.48\mathrm{mM CaCl_2}\) (to compensate for the \(\mathrm{Ca^{2 + }}\) concentration in the nutrient solution), and other component was not changed. The flag leaf was harvested at heading stage \(^{11}\) in the field (Shijiazhuang, China), and seed was also harvested 21DAA \(^{11}\) in the field (shijiazhuang, China). In each NN plot, \(300\mathrm{kg / ha}\) of diamine phosphate and \(225\mathrm{kg / ha}\) of urea were applied before sowing, and \(150\mathrm{kg / ha}\) of urea was applied at the elongation stage every year. In the LN plots, no N fertilizer (N- deficient) was applied during the growing period.  
+
+## Generation of transgenic wheat plants  
+
+To obtain CRISPR transgenic wheat plants, the pU6- gRNA of TaSWN was annealed and inserted into pJIT163- Ubi- Cas9 vector. All constructed vectors were transformed into callus to generate the transgenic plants.  
+
+To identify mutations in TaSWN- 4A, TaSWN- 4B, or TaSWN- 4D, gene- specific primers were designed around the target site. Primers SWN- Check- F and SWN- Check- A- R were used to amplify TaSWN- 4A, SWN- Check- F and SWN- Check- B- R were used to amplify TaSWN- 4B, and SWN- Check- F and SWN- Check- D- R were used to amplify TaSWN- 4D. Primer sequences were listed in Supplemental Table S4. PCR products were checked on agarose gels and genotyped by Sanger sequencing.  
+
+## RNA-seq and CUT&Tag experiment  
+
+Total RNA was extracted using HiPure Plant RNA Mini Kit according to the manufacturer's instructions (Magen, R4111- 02), and libraries were sequenced using an Illumina Novaseq platform.  
+
+CUT&Tag experiment were done follow the previous described method \(^{19}\) . The nuclei were extracted by chooping fresh samples soaked in the HBM buffer ( \(25\mathrm{mM}\) Tris- HCl pH 7.6, 0.44 M sucrose, \(10\mathrm{mM}\) MgCl2, \(0.1\%\) Triton- X, \(10\mathrm{mM}\) Beta- mercaptoethanol, \(2\mathrm{mM}\) spermine, \(1\mathrm{mM}\) PMSF, EDTA- free protease inhibitor cocktail). After overnight incubation with corresponding antibody in \(4^{\circ}\mathrm{C}\) , the nuclei was incubated in \(50\mu \mathrm{l}\) wash buffer ( \(20\mathrm{mM}\) HEPES pH 7.5; \(150\mathrm{mM}\) NaCl; \(0.5\mathrm{mM}\) Spermidine; \(1\times\) Protease inhibitor cocktail) with secondary antibody (1:100; Guinea Pig anti- Rabbit IgG antibody) at \(4^{\circ}\mathrm{C}\) for around 1- 2 hour and then washed twice with wash buffer. \(\mathrm{pA - Tn5}\) complex ( \(\mathrm{pA - Tn5}\) 1:100 dilution in CT- 300 buffer:20 mM HEPES pH 7.5; \(300\mathrm{mM}\) NaCl; \(0.5\mathrm{mM}\) Spermidine; \(1\times\) Protease inhibitor cocktail) was incubated with nuclei in \(4^{\circ}\mathrm{C}\) for 2- 3h (Tn5, Vazyme, TD501- 01). After washing twice with CT- 300 buffer, the tagmentation of nuclei was done in \(300\mu \mathrm{l}\) Tagmentation buffer ( \(20\mathrm{mM}\) HEPES pH 7.5; \(300\mathrm{mM}\) NaCl; \(0.5\mathrm{mM}\) Spermidine; \(1\times\) Protease inhibitor cocktail; \(10\mathrm{mM}\) MgCl2) in \(37^{\circ}\mathrm{C}\) for 1h. \(10\mu \mathrm{l}\) 0.5M EDTA, \(3\mu \mathrm{l}\) \(10\%\) SDS and \(2.5\mu \mathrm{l}\) \(20\mathrm{mg / ml}\) Protease K were added to stop tagmentation reaction. The DNA was extracted with phenol:chloroform:isoamyl alcohol, precipitated with ethanol, and resuspended in ddH2O. The library was amplified 17 cycles by Q5 high fidelity polymerase (NEB, M0491L). Antibodies used for histone modifications are the same as previous reported \(^{19}\) . Libraries were purified with AMPure beads (Beckman, A63881) and sequenced using the Illumina Novaseq platform at Annoroad Gene Technology.
+
+<--- Page Split --->
+
+## \(^{15}\mathrm{N}\) -nitrate uptake activity assay  
+
+\(^{15}\mathrm{N}\) - nitrate uptake activity assay was performed as described before \(^{50}\) . Wheat seedlings were grown in LN nutrient solution (0.1 mM \(\mathrm{KNO_3}\) ) for 28 days, respectively. After that, the seedlings were subjected to a pre- treatment of 3 hours in LN nutrient solution (0.1 mM \(\mathrm{KNO_3}\) ). Subsequently, the seedlings were transferred to LN (0.1 mM \(\mathrm{KNO_3}\) ) was replaced by 0.1 mM \(^{15}\mathrm{N} - \mathrm{KNO_3}\) ) nutrient solution for \(^{15}\mathrm{N}\) - labelling for a 3 hours. Post \(^{15}\mathrm{N}\) - labelling, the roots were washed using 0.1 mM \(\mathrm{CaSO_4}\) solution and deionized water. Finally, the shoots and roots of the seedlings were separately collected and dried at \(70^{\circ}\mathrm{C}\) until they reached a constant weight. Then, the samples were ground to fine powder and \(^{15}\mathrm{N}\) - content was detected using an isotope ratio mass spectrometer (Isoprime 100).  
+
+## Nitrate \(\left(\mathrm{NO}_3\right)\) content assay  
+
+Nitrate assay was performed as described before \(^{51}\) . Standard curve was made based on different concentration of \(\mathrm{KNO_3}\) solution (deionized water as a control). Samples were boiled at \(100^{\circ}\mathrm{C}\) for 20 min. After Centrifuge of the boiled different samples, salicylic acid- sulphuric acid was added to supernatant. After incubation of 20 min, \(8\%\) (w/v) \(\mathrm{NaOH}\) solution was added. After cool down of the samples, measure the \(\mathrm{OD}_{410}\) value of each sample with the control for reference.  
+
+For samples collected of KN9204 and J411, grind each sample into powder in liquid nitrogen, detection procedure was done as described before. Finally, calculate the nitrate concentration using the following equation: \(\mathrm{Y} = \mathrm{CV} / \mathrm{W}\) .  
+
+\(\mathrm{Y}\) : nitrate content \((\mu \mathrm{g / g})\) , \(\mathrm{C}\) : nitrate concentration calculated with \(\mathrm{OD}_{410}\) into standard curve \((\mu \mathrm{g / ml})\) , \(\mathrm{V}\) : the total volume of extracted sample (ml), \(\mathrm{W}\) : weight of sample (g).  
+
+## Grain protein content assays  
+
+Grain protein content was measured by near- infrared reflectance spectroscopy (NIRS) with a Perten DA- 7200 instrument (Perten Instruments, Huddinge, Sweden) and expressed on a \(14\%\) moisture basis. The measurements were calibrated using calibration samples according to the manufacturer's instructions.  
+
+## Luciferase reporter assay  
+
+The genomic sequence of distal regulatory region was amplified and fused in- frame with the pMY155- mini35S vector \(^{32}\) to generate the reporter construct cultivar- specific- regions- mini35Spro:LUC. Primer sequences were listed in Supplemental Table S4. Then, mini35Spro:LUC (as control) and the reporter vector cultivar- specific- regions - mini35Spro:LUC were transformed into A.tumefaciens strain GV3101. The bacterial solution was injected to the back of the leaves of Nicotiana benthamiana (6- 8 leaf stage) using a syringe with the needle removed. The Nicotiana benthamiana were cultivated for 2- 3 days at a temperature of \(22^{\circ}\mathrm{C}\) and a light cycle of \(16\mathrm{h}\) light/8 h dark. Firefly luciferase (LUC) and Renilla luciferase (REN) activities were measured using a dual luciferase assay reagent (Promega, VPE1910). And the relative intensity was calculated by ratio between relative ratio (LUC: REN) of cultivar- specific regions and empty vector.
+
+<--- Page Split --->
+
+## Root system scanning  
+
+The root of different samples was scanned by ScanMaker i8000 plus, after analyzed by WinRHIZO software, five root traits were quantified, including total root length (Rl), root surface area (Rs), root volume (Rv), root diameter (Rd) and root tip number (Rt); and maximum root length is measured using a ruler.  
+
+LN response ratio (LRR) was calculated to reflect the root change under LN condition, which was \(\mathrm{(R_{LN}(root~trait~under~LN~condition) - R_{NN}(root~trait~under~NN~condition)) / R_{NN}}\) .  
+
+## TSA treatment  
+
+TSA (V900931- 5MG) was dissolved in DMSO, then directly add into nutrient solution (NN/LN) to final concentration of \(2\mu \mathrm{M}\) , with DMSO as mock. After treatment of 4 days, the root was harvested and immediately frozen in liquid nitrogen and stored at \(- 80^{\circ}\mathrm{C}\) .  
+
+## Protein interaction test (Y2H and BiFC)  
+
+For the Y2H between PRC2 and TaERF9_5B/ TaERF9_5D/TaBPC_4A, Full- length of TaERF9_5B, TaERF9_5D and TaBPC_4A were amplified using specific primers (listed in Supplemental Table S4) and fused with GAL4 AD in the pDEST22 vector. Full length of TaEMF2- 2A.2 and N- terminal of TaSWN were amplified using specific primers (listed in Supplemental Table S4) and fused with GAL4 BD in the pDEST32 vector. Interactions in yeast were tested on the SD/- Trp/- Leu/- His/- Ade medium.  
+
+For the BiFC analysis, the cDNA of TaERF9_5B/ TaERF9_5D/TaBPC_4Aand TaEMF2- 2A.2/TaSWN was amplified with primers (listed in Supplemental Table S4) and cloned into pSCYCE and pSCYNE vectors containing either C- or N- terminal portions of the enhanced cyan fluorescent protein. The resulting constructs were transformed into A. tumefaciens strain GV3101. Then these strains were injected into tobacco leaves in different combinations with p19. The CFP fluorescence was observed with a confocal laser- scanning microscope (FluoView 1000, Olympus).  
+
+## Western blot assays  
+
+Total histone proteins were extracted by using EpiQuik Total Histone Extraction Kit (OP- 0006- 100). The total histone proteins were then used for western blot using the antibodies listed below. Anti- H3 immunoblot was used as a loading control. Antibodies: anti- H3 (ab1791, Abcam), anti- H3K27ac (ab4729, Abcam). Immunoblotting was done by using the enhanced chemiluminescence (ECL) system.  
+
+## RNA-seq Data Processing  
+
+Adapter sequence and low- quality reads of RNA- seq library was removed by fastp (0.20.1)52, the cleaned reads was mapped to IWGSC Refseq v1.1 using hisat2 (2.1.0)53, and gene expression was quantified by featureCount (2.0.1)54. Differentially expressed genes were evaluated using the DESeq2 package (1.34.0)55 in R with an adjusted p value \(< 0.05\) and log2 fold- change \(>1\) . TPM (Transcripts Per Kilobase Million) values generated from the counts matrix were used to characterize gene expression and used for hierarchical clustering analysis.
+
+<--- Page Split --->
+
+For functional enrichment, GO annotation files were generated from IWGSC Annotation v1.1 and an R package clusterProfiler (4.2.2) \(^{56}\) was used for enrichment analysis.  
+
+## CUT&Tag Data Processing  
+
+Adapter sequence and low- quality reads of CUT&Tag library was removed by fastp (0.20.1) \(^{52}\) , the cleaned reads was mapped to IWGSC Refseq v1.1 using bwa mem algorithm (0.7.17) \(^{57}\) , We further filter the reads mapped with "samtools view - bS - F 1,804 - f 2 - q 30" to filter the low- quality mapped reads. Then the high- quality mapped reads were reduplicated using Picard- 2.20.5- 0. The de- duplicated bam files from two biological replicates were merged by samtools (1.5) \(^{58}\) , and merged bam file was converted into bigwig files using bamCoverage provided by deeptools (3.3.0) with parameters "- bs 10 - - effectiveGenomeSize 14,600,000,000 - - normalizeUsing RPKM - - smoothLength 50". The bigwig files were visualized using deeptools (3.3.0) \(^{59}\) and IGV (2.8.0.01) \(^{60}\) .  
+
+For peak calling, macs2 (2.1.4) \(^{61}\) was used. For narrow peaks (H3K27ac, H3K4me3, and H2A.Z) and broad peaks (H3K27me3, H3K36me3, and H3K9me3), parameters "- p 1e- 3 - - keep- dup all - g 14600000000" and "- - keep- dup all - g 14600000000 - - broad - - broad- cutoff 0.05" were used. Peak was annotated to the wheat genome using the R package ChIPseeker (v1.30.3) \(^{62}\) , as peaks annotated to three categories: promoter (- 3000bp of TSS), genic (TSS to TES) and distal (other). The MAnorm package \(^{63}\) was used for the quantitative comparison of CUT&Tag signals between samples with the following criteria: \(|\mathrm{M} \mathrm{value}| > 1\) and \(\mathrm{P} < 0.05\) .  
+
+## Chromatin state analysis  
+
+For chromatin state analysis, chromHMM (1.21) \(^{20}\) was used. "BinarizeBam" and "LearnModel" commands with default parameters were used for chromatin- state (CS) annotation. Multiple models were trained on these data, with CS numbers ranging from 2 to 20. The 15- state model was selected because it captured all the key information of CS. In previous studies, 15- state models were similarly trained on rice \(^{64}\) and Arabidopsis \(^{65}\) data. For chromatin states dynamic change analysis, bins (CS called, 200bp) were called dynamic if its state diverges between different samples. For variability score of histone modification, one minus jaccard index, which was calculated by bedtools (v2.29.2).  
+
+## Distal regulatory region-gene assignment  
+
+The distal regulatory regions annotation strategy was largely based on a previous study \(^{28}\) . Genes within 0.5M from a distal H3K27ac peak are considered candidate target genes. Then we generated null model as correlations between randomly selected peaks and randomly selected genes on different chromosomes, and enabling us to compute mean and standard deviation of this null distribution. For each potential link, after calculate correlation between gene expression (TPM) and distal H3K27ac signal (FPKM) in samples, we also compute p- values for the test correlations based on null model, then significantly pairs were selected as regulatory region- gene pairs.  
+
+## Detection of transcription factor-binding motifs  
+
+To detect recruiter of dynamic H3K27me3 changes, we downloaded the position weightmatrices of plant motifs from the JASPAR database \(^{40}\) , the motifs was scanned by
+
+<--- Page Split --->
+
+## Reference
+
+<--- Page Split --->
+
+single cells. Nat Commun 10, 1930 (2019). https://doi.org/10.1038/s41467- 019- 09982- 5Zhao, L. et al. Dynamic chromatin regulatory programs during embryogenesis of hexaploid wheat. Genome Biology 24, 7 (2023). https://doi.org/10.1186/s13059- 022- 02844- 2Ernst, J. & Kellis, M. Chromatin- state discovery and genome annotation with ChromHMM. Nat Protoc 12, 2478- 2492 (2017). https://doi.org/10.1038/nprot.2017.124Wang, Y. Y., Cheng, Y. H., Chen, K. E. & Tsay, Y. F. Nitrate Transport, Signaling, and Use Efficiency. Annu Rev Plant Biol 69, 85- 122 (2018). https://doi.org/10.1146/annurev- arplant- 042817- 040056Krapp, A. Plant nitrogen assimilation and its regulation: a complex puzzle with missing pieces. Curr Opin Plant Biol 25, 115- 122 (2015). https://doi.org/10.1016/j.pbi.2015.05.010Taochy, C. et al. The Arabidopsis root stele transporter NPF2.3 contributes to nitrate translocation to shoots under salt stress. Plant J 83, 466- 479 (2015). https://doi.org/10.1111/tpj.12901Giaimo, B. D., Ferrante, F., Herchenrother, A., Hake, S. B. & Borggrefe, T. The histone variant H2A.Z in gene regulation. Epigenetics Chromatin 12, 37 (2019). https://doi.org/10.1186/s13072- 019- 0274- 9Russ, B. E. et al. Regulation of H3K4me3 at Transcriptional Enhancers Characterizes Acquisition of Virus- Specific CD8(+) T Cell- Lineage- Specific Function. Cell Rep 21, 3624- 3636 (2017). https://doi.org/10.1016/j.celrep.2017.11.097Calo, E. & Wysocka, J. Modification of enhancer chromatin: what, how, and why? Mol Cell 49, 825- 837 (2013). https://doi.org/10.1016/j.molcel.2013.01.038Wang, M. et al. An atlas of wheat epigenetic regulatory elements reveals subgenome divergence in the regulation of development and stress responses. Plant Cell 33, 865- 881 (2021). https://doi.org/10.1093/plcell/koab028Trevino, A. E. et al. Chromatin accessibility dynamics in a model of human forebrain development. Science 367 (2020). https://doi.org/10.1126/science.aay1645Concia, L. et al. Wheat chromatin architecture is organized in genome territories and transcription factories. Genome Biol 21, 104 (2020). https://doi.org/10.1186/s13059- 020- 01998- 1Yamaguchi, M. et al. VASCULAR- RELATED NAC- DOMAIN7 directly regulates the expression of a broad range of genes for xylem vessel formation. Plant J 66, 579- 590 (2011). https://doi.org/10.1111/j.1365- 313X.2011.04514.xSasidharan, R. et al. Light quality- mediated petiole elongation in Arabidopsis during shade avoidance involves cell wall modification by xyloglucan endotransglucosylase/hydrolases. Plant Physiol 154, 978- 990 (2010). https://doi.org/10.1104/pp.110.162057Li, Z. et al. The bread wheat epigenetic map reveals distinct chromatin architectural and evolutionary features of functional genetic elements. Genome Biol 20, 139 (2019). https://doi.org/10.1186/s13059- 019- 1746- 8Tokunaga, H. et al. Arabidopsis lonely guy (LOG) multiple mutants reveal a central role of the LOG- dependent pathway in cytokinin activation. Plant J 69, 355- 365 (2012). https://doi.org/10.1111/j.1365- 313X.2011.04795.xBarbez, E. et al. A novel putative auxin carrier family regulates intracellular auxin homeostasis in plants. Nature 485, 119- 122 (2012). https://doi.org/10.1038/nature11001Yoshida, M., Kijima, M., Akita, M. & Beppu, T. Potent and specific inhibition of mammalian histone deacetylase both in vivo and in vitro by trichostatin A. Journal of Biological Chemistry 265, 17174- 17179 (1990). https://doi.org/10.1016/s0021- 9258(17)44885- xGonin, M. et al. CROWN ROOTLESS1 binds DNA with a relaxed specificity and activates OsROP and OsbHLH044 genes involved in crown root formation in rice. Plant J 111, 546- 566 (2022). https://doi.org/10.1111/tpj.15838
+
+<--- Page Split --->
+
+37 Xiao, J. et al. Cis and trans determinants of epigenetic silencing by Polycomb repressive complex 2 in Arabidopsis. Nat Genet 49, 1546- 1552 (2017). https://doi.org:10.1038/ng.3937  
+
+38 Bieluszewski, T., Xiao, J., Yang, Y. & Wagner, D. PRC2 activity, recruitment, and silencing: a comparative perspective. Trends in Plant Science 26, 1186- 1198 (2021). https://doi.org:10.1016/j.tplants.2021.06.006  
+
+39 Zhou, Y. et al. Telobox motifs recruit CLF/SWN- PRC2 for H3K27me3 deposition via TRB factors in Arabidopsis. Nature Genetics 50, 638- 644 (2018). https://doi.org:10.1038/s41588- 018- 0109- 9  
+
+40 Fornes, O. et al. JASPAR 2020: update of the open- access database of transcription factor binding profiles. Nucleic Acids Res 48, D87- D92 (2020). https://doi.org:10.1093/nar/gkz1001  
+
+41 Mu, Y. et al. BASIC PENTACYSTEINE Proteins Repress ABSCISIC ACID INSENSITIVE4 Expression via Direct Recruitment of the Polycomb- Repressive Complex 2 in Arabidopsis Root Development. Plant Cell Physiol 58, 607- 621 (2017). https://doi.org:10.1093/pcp/pcx006  
+
+42 Li, H., Hu, B. & Chu, C. Nitrogen use efficiency in crops: lessons from Arabidopsis and rice. J Exp Bot 68, 2477- 2488 (2017). https://doi.org:10.1093/jxb/erx101  
+
+43 Swarbreck, S. M. et al. A Roadmap for Lowering Crop Nitrogen Requirement. Trends Plant Sci 24, 892- 904 (2019). https://doi.org:10.1016/j.tplants.2019.06.006  
+
+44 Han, M., Okamoto, M., Beatty, P. H., Rothstein, S. J. & Good, A. G. The Genetics of Nitrogen Use Efficiency in Crop Plants. Annu Rev Genet 49, 269- 289 (2015). https://doi.org:10.1146/annurev- genet- 112414- 055037  
+
+45 Ricci, W. A. et al. Widespread long- range cis- regulatory elements in the maize genome. Nat Plants 5, 1237- 1249 (2019). https://doi.org:10.1038/s41477- 019- 0547- 0  
+
+46 Li, E. et al. Long- range interactions between proximal and distal regulatory regions in maize. Nat Commun 10, 2633 (2019). https://doi.org:10.1038/s41467- 019- 10603- 4  
+
+47 Xiao, J. et al. Wheat genomic study for genetic improvement of traits in China. Sci China Life Sci 65, 1718- 1775 (2022). https://doi.org:10.1007/s11427- 022- 2178- 7  
+
+48 Luo, M. et al. HD2C interacts with HDA6 and is involved in ABA and salt stress response in Arabidopsis. J Exp Bot 63, 3297- 3306 (2012). https://doi.org:10.1093/jxb/ers059  
+
+49 Huang, S. et al. Arabidopsis histone H3K4 demethylase JMJ17 functions in dehydration stress response. New Phytol 223, 1372- 1387 (2019). https://doi.org:10.1111/nph.15874  
+
+50 Liu, Y., Hu, B. & Chu, C. 15N- nitrate Uptake Activity and Root- to- shoot Transport Assay in Rice. Bio- Protocol 6 (2016). https://doi.org:10.21769/BioProtocol.1897  
+
+51 Zhao, L. & Wang, Y. Nitrate Assay for Plant Tissues. Bio Protoc 7, e2029 (2017). https://doi.org:10.21769/BioProtocol.2029  
+
+52 Chen, S., Zhou, Y., Chen, Y. & Gu, J. fastp: an ultra- fast all- in- one FASTQ preprocessor. Bioinformatics 34, i884- i890 (2018). https://doi.org:10.1093/bioinformatics/btv560  
+
+53 Kim, D., Paggi, J. M., Park, C., Bennett, C. & Salzberg, S. L. Graph- based genome alignment and genotyping with HISAT2 and HISAT- genotype. Nat Biotechnol 37, 907- 915 (2019). https://doi.org:10.1038/s41587- 019- 0201- 4  
+
+54 Liao, Y., Smyth, G. K. & Shi, W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics 30, 923- 930 (2014). https://doi.org:10.1093/bioinformatics/btf656  
+
+55 Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA- seq data with DESeq2. Genome Biol 15, 550 (2014). https://doi.org:10.1186/s13059- 014- 0550- 8  
+
+56 Wu, T. et al. clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. Innovation (Camb) 2, 100141 (2021). https://doi.org:10.1016/j.xinn.2021.100141  
+
+57 Li, H. Aligning sequence reads, clone sequences and assembly contigs with BWA- MEM.
+
+<--- Page Split --->
+
+arXiv (2013). https://doi.org/10.48550/arXiv.1303.3997Danecek, P. et al. Twelve years of SAMtools and BCFtools. GigaScience 10, giab008 (2021). https://doi.org/10.1093/gigascience/giab008Ramirez, F. et al. deepTools2: a next generation web server for deep- sequencing data analysis. Nucleic Acids Res 44, W160- 165 (2016). https://doi.org/10.1093/nar/gkw257Robinson, J. T. et al. Integrative genomics viewer. Nature Biotechnology 29, 24- 26 (2011). https://doi.org/10.1038/nbt.1754Zhang, Y. et al. Model- based Analysis of ChIP- Seq (MACS). Genome Biology 9, R137 (2008). https://doi.org/10.1186/gb- 2008- 9- 9- r137Yu, G., Wang, L.- G. & He, Q.- Y. ChIPseeker: an R/Bioconductor package for ChIP peak annotation, comparison and visualization. Bioinformatics 31, 2382- 2383 (2015). https://doi.org/10.1093/bioinformatics/btv145Shao, Z., Zhang, Y., Yuan, G.- C., Orkin, S. H. & Waxman, D. J. MAnorm: a robust model for quantitative comparison of ChIP- Seq data sets. Genome Biology 13, R16 (2012). https://doi.org/10.1186/gb- 2012- 13- 3- r16Zhao, L. et al. Integrative analysis of reference epigenomes in 20 rice varieties. Nat Commun 11, 2658 (2020). https://doi.org/10.1038/s41467- 020- 16457- 5Wu, L. Y. et al. Dynamic chromatin state profiling reveals regulatory roles of auxin and cytokinin in shoot regeneration. Dev Cell 57, 526- 542 e527 (2022). https://doi.org/10.1016/j.devcel.2021.12.019  
+
+## Figure legends  
+
+## Fig. 1. A 15-state model characterizes the dynamic chromatin landscape in KN9204 and J411  
+
+a. Experimental design for generating the epigenomic data sets in KN9204 and J411 under different nitrogen conditions.  
+b. Chromatin state definitions, genomic annotation enrichments, and expression levels of genes associated with each chromatin state. Chromatin states could be mainly organized into five categories: Pr (promoter), Tr (transcription), En (enhancer-like), Re (repressive), and Ns (No signal) states.  
+c. Schematic diagrams representing the chromatin state dynamics under different nitrogen conditions and tissues. State colors are as in (b).  
+d. The fractions of bases for the states of the five categories that vary between cultivar difference and nitrogen availability.  
+e. Variability scores of histone modifications between cultivar difference and nitrogen availability.  
+f. Cross-correlation heatmaps of all H3K27ac peaks which were located in the distal or promoter regions separately for the two cultivars and the two nitrogen conditions. Abbreviations were as follow: KN_NN_R: KN9204_NN_Root, KN_NN_L: KN9204_NN_Leaf, KN_NN_S: KN9204_NN_Seed, KN_LN_R: KN9204_LN_Root, KN_LN_L: KN9204_LN_Leaf, KN_LN_S: KN9204_LN_Seed, J4_NN_R: J411_NN_Root, J4_NN_L: J411_NN_Leaf, J4_NN_S: J411_NN_Seed, J4_LN_R:
+
+<--- Page Split --->
+
+J411_LN_Root, J4_LN_L: J411_LN_Leaf, J4_LN_S: J411_LN_Seed.  
+
+## Fig. 2. Epigenomic variations are vital to the expression bias of NMGs  
+
+a. Fraction of the regulatory regions and expression variation of nitrogen metabolism genes (NMGs) between wheat cultivars KN9204 and J411. Abbreviations were as follow: NPF: NRT1/PTR FAMILY, NRT2: Nitrate transporter 2, CLC: Chloride channel protein, SLAH: Slow anion channel-associated homologues, AMT: Ammonium transporter, APC: The amino acid-polyamine-choline transporter superfamily, NIA: Nitrate reductase, NIR: Nitrite reductase, GS: Glutamine synthetase, GOGAT: Glutamate synthase, ASN: Asparagine synthetase, AspAT: Aspartate aminotransferase, GDH: Glutamate dehydrogenase, TF: transcription factor.  
+
+b. Enrichment of LN-induced differentially-expressed NMGs based on their functional category in three tissues (Fisher's exact test was used to calculate the \(p\) -values for the overlaps). Abbreviations were same with (a).  
+
+c. Proportion of NMGs showing expression variation between KN9204 and J411 with regulatory region variation (top panel), H3K27ac variation (middle panel), and H3K27me3 variation (bottom panel).  
+
+d. Histone modification (H3K27ac and H3K27me3) levels of NRT2 and NPF2.3 in the roots of KN9204 and J411 between the two nitrogen availability levels (Wilcox test: ns: \(p > 0.05\) ; \(*: p \leq 0.05\) ; \(**: p \leq 0.01\) ; \(****: p \leq 0.0001\) ).  
+
+e. Representative tracks showing histone modifications and transcriptional changes of TaNRT2_6B and TaNPF2.3_7B in roots for the two wheat cultivars and two nitrogen levels.  
+
+f. Nitrate content of shoots and roots of KN9204 and J411 under two nitrogen conditions (Student's \(t\) -test; ns: \(p > 0.05\) ; \(*: p \leq 0.05\) ).  
+
+## Fig. 3. Distal regulatory region divergence is involved in NUE-trait variation between KN9204 and J411  
+
+a. K-means clustering of the variable H3K27ac peaks.  
+
+b. Peak distribution (distal, genic, or promoter) of cultivar-specific H3K27ac peaks in the wheat genome.  
+
+c. Mean gene expression in J411 (x-axis) versus mean gene expression in KN9204 (y-axis) for genes associated with proximal cultivar-biased H3K27ac peaks separately in the roots (top panel). Overlap between genes marked by cultivar-specific H3K27ac and DEGs between cultivars (blue circle) (bottom panel) (Fisher's exact test was used to calculate the \(p\) -values for the overlaps).  
+
+d. Assignment of distal cultivar-specific H3K27ac peaks to potential targets; representative genes and GO enrichment are shown on the right, and representative case is shown at the bottom. Abbreviations were as follow: NN_R: NN_Root, NN_L: NN_Leaf, NN_S: NN_Seed, LN_R: LN_Root, LN_L: LN_Leaf, LN_S: LN_Seed.  
+
+e. Enrichment of target genes with distal cultivar-specific H3K27ac peaks in QTLs between KN9204 and J411 (Fisher's exact test was used to calculate the \(p\) -values for the enrichment). Abbreviation: Rt: root tip number; Rd: root diameter; Nup: Nitrogen uptake content; Nct: Nitrogen concentration; MRL: Maximum root length; TKW: Thousand-kernel weight; GPC: Grain protein content; Rs: Root surface area; Rdw: Root dry weight; Ls: flag leaf size.  
+
+f. Schematic diagram illustrating the process used to narrow down potential key genes
+
+<--- Page Split --->
+
+regulated by distal cultivar- specific regulation in the QTL regions (QMLR- 7B in this case). g. Representative tracks of TaHyPRP06 regulated by distal cultivar- specific H3K27ac peaks in QMLR- 7B. Abbreviations were as follow: NN_R: NN_Root, LN_R: LN_Root. h. Luciferase reporter assay of cultivar- specific H3K27ac regions in track shown in (g) and Fig.S3. h (Student's \(t\) - test; ns: \(p > 0.05\) ; \(*:p\leq 0.05\) ).  
+
+## Fig. 4. Dynamics of H3K27ac reveals the divergent strategies of KN9204 and J411 to low nitrogen conditions  
+
+a. The number of differential H3K27ac peaks in response to LN of three tissues (seeds, roots, leaves) in KN9204 and J411. 
+b. Heatmaps showing the differential H3K27ac peaks in eight clusters in the roots of KN9204 and J411 under NN and LN conditions. 
+c. Dynamic proximal H3K27ac peaks and corresponding gene expression changes in the roots of KN9204 and J411. Representative genes and GO (gene ontology) enrichments are shown on the right. Abbreviations were as follow: KN_NN_R: KN9204_NN_Root, KN_LN_R: KN9204_LN_Root, J4_NN_R: J411_NN_Root, J4_LN_R: J411_LN_Root. 
+d. The LN-response-ratio (LRR) of root systems for KN9204 and J411 (Student's \(t\) -test; ns: \(p > 0.05\) ; \(*:p\leq 0.05\) ). 
+e. The number of root tips in KN9204 and J411 plants under different nitrogen conditions (Student's \(t\) -test; ns: \(p > 0.05\) ; \(*:p\leq 0.05\) ). 
+f. Phenotypes of the root systems of KN9204 and J411 seedlings in the control (mock) and TSA treatments (2 \(\mu \mathrm{M}\) ) under different nitrogen conditions. Scale bars=5cm. The right panel is the LRR of the root tips (Student's \(t\) -test; ns: \(p > 0.05\) ; \(*:p\leq 0.05\) ).  
+
+## Fig. 5. Different trends in H3K27me3 shapes adaptation bias between the cultivars KN9204 and J411  
+
+a. The number of differential (gain or loss) H3K27me3 peaks in response to LN of three tissues in KN9204 and J411. 
+b. Heatmaps showing the differential H3K27me3 peaks in the seven clusters in the roots of KN9204 and J411 under LN and NN conditions. 
+c. The number of differential H3K27me3 peaks (in proximal or distal regions) in the root of KN9204 and J411 plants in response to LN. 
+d. Overlap between genes with dynamic changes in H3K27me3 and DEGs in the roots (Fisher's exact test was used to calculate the \(p\) -values for the overlaps). 
+e. GO enrichment of the overlapping genes related to (d). 
+f. Representative tracks showing H3K27me3 and transcriptional changes in TaMob1L_6D (top panel), TaNAR2_4A (middle panel), and TaKAO2_4A (bottom panel) for the two cultivars under LN and NN. Abbreviations were as follow: KN_NN_R: KN9204_NN_Root, KN_LN_R: KN9204_LN_Root, J4_NN_R: J411_NN_Root, J4_LN_R: J411_LN_Root. 
+g. Sequence motif enrichment for the up-regulated H3K27me3 peaks under LN conditions in KN9204 and J411 (Fisher's exact test was used to calculate the \(p\) -values for the enrichment). 
+h. Yeast two-hybrid (Y2H) assays showing the interaction between TaERF9_5B/TaERF9_5D/TaBPC_4A and TaEMF2-2A.2/TaSWN (components of PRC2).
+
+<--- Page Split --->
+
+Transformed yeast cells were grown on synthetic media lacking Leu and Trp (- WL) and Leu, Trp, His, and Ade (- WLHA) or Leu, Trp, and His (- WLH) with 1mM 3AT.  
+
+i. BiFC analysis of the interaction between TaERF9_5B/TaERF9_5D/TaBPC_4A and TaSWN-N/TaEMF2_2A.2. Scale bars = 10 mm.  
+
+Fig. 6. Rewiring H3K27me3 strengthens root adaptation to low nitrogen fertilization level  
+
+a. Heatmaps showing the differential H3K27me3 peaks in 'BobWhite' (BW) and the Taswn- cr mutant plants in the eight clusters under NN and LN conditions.  
+
+b. The proportion of TaSWN-dependent peaks in the up-regulated H3K27me3 peaks under LN conditions in KN9204 and J411.  
+
+c. Overlap between genes marked by TaSWN-dependent H3K27me3 and marked by gain-of-H3K27me3 under LN conditions in J411 (Fisher's exact test was used to calculate the \(p\) -values for the overlap).  
+
+d. Dynamic changes in the TaSWN-dependent H3K27me3 peaks under different nitrogen conditions (NN and LN) in KN9204, J411, BW, and Taswn-cr plants. Representative gene names are shown on the right.  
+
+e. GO enrichment of the overlapping genes related to (d).  
+
+f. The expression changes and fold-changes of the genes in (d). (Wilcoxon test: ns: \(p > 0.05\) ; \(***: p \leq 0.0001\) ) FC: fold-change.  
+
+g. Scans of root systems of BW and Taswn-cr seedlings grown under different nitrogen conditions. Scale bars = 5 cm.  
+
+h. The number of root tips and total length of the root systems for BW and Taswn-cr seedlings grown under different nitrogen conditions. (Student's \(t\) -test; ns: \(p > 0.05\) ; \(*: p \leq 0.05\) ).  
+
+i. The fold-changes in the expression of NRT2 genes in BW, Taswn-cr, J411, and KN9204 seedlings grown under different nitrogen conditions.  
+
+j. \(^{15}\mathrm{N}\) -nitrate uptake activity of different seedlings (KN9204, J411, BW, and Taswn-cr) grown under LN conditions (Student's \(t\) -test; ns: \(p > 0.05\) ; \(*: p \leq 0.05\) ).  
+
+## Fig. 7. The balanced model of epigenetic regulation for divergent strategies to LN conditions in wheat  
+
+In the low- nitrogen condition (Input, left), different adaptation strategy of wheat is selected from the balance between root system development and nitrate uptake transporters (NRT2s). great gain- of- H3K27ac enhance the expression of root development- related genes in KN9204, Conversely, greater gain- of- H3K27me3 and minor gain- of- H3K27ac reduce root development in J411, but with an associated increase in nitrate uptake transporters (NRT2s) via gain- of- H3K27ac. After the knock- out of TaSWN (Taswn- cr), there is de- repression of root development, accompanied by the loss of H3K27me3 (Decision making, middle). Phenotypically, KN9204 has a developed root architecture (higher expression of root development genes) but lower nitrate uptake rate per unit weight (lower expression of NRT2s). In contrast, J411 has a diverse selection with regards to this balance (Output, right).  
+
+## Fig. S1. NUE epigenome dataset  
+
+a. The number of DEGs in the tissues in response to LN in KN9204 and J411.  
+b. PCA plots of H3K27ac, H3K27me3, H3K4me3, H3K9me3, H3K36me3, and H2A.Z in the NUE epigenome dataset. Each dot represents one sample. Two biological replicates were
+
+<--- Page Split --->
+
+sequenced for each stage. Abbreviations were as follow: KN_NN_R_1: KN9204_NN_Root rep1, KN_NN_R_2: KN9204_NN_Root rep2, KN_NN_L_1: KN9204_NN_Leaf rep1, KN_NN_L_2: KN9204_NN_Leaf rep2, KN_NN_S_1: KN9204_NN_Seed rep1, KN_NN_S_2: KN9204_NN_Seed rep2, KN_LN_R_1: KN9204_LN_Root rep1, KN_LN_R_2: KN9204_LN_Root rep2, KN_LN_L_1: KN9204_LN_Leaf rep1, KN_LN_L_2: KN9204_LN_Leaf rep2, KN_LN_S_1: KN9204_LN_Seed rep1, KN_LN_S_2: KN9204_LN_Seed rep2, J4_NN_R_1: J411_NN_Root rep1, J4_NN_R_2: J411_NN_Root rep2, J4_NN_L_1: J411_NN_Leaf rep1, J4_NN_L_2: J411_NN_Leaf rep2, J4_NN_S_1: J411_NN_Seed rep1, J4_NN_S_2: J411_NN_Seed rep2, J4_LN_R_1: J411_LN_Root rep1, J4_LN_R_2: J411_LN_Root rep2, J4_LN_L_1: J411_LN_Leaf rep1, J4_LN_L_2: J411_LN_Leaf rep2, J4_LN_S_1: J411_LN_Seed rep1, J4_LN_S_2: J411_LN_Seed rep2.  
+
+c. Peak distributions in gene regions of different histone marks in the wheat genome.  
+
+d. Heatmaps of epigenetic marks for all annotated wheat genes which were sorted according to their expression levels determined by RNA-Seq.  
+
+e. Proportion of peaks located in TE regions for the six different histone marks.  
+
+f. H2A.Z profiles of DEGs that respond to external stimuli (abiotic and biotic stress); RNA-seq data from \(^{27}\) .  
+
+g. Fraction of bases for 15 states that vary between the different cultivars (top) and nitrogen availability (bottom).  
+
+h. Cross-correlation heatmaps of all H3K27me3 peaks which are located in distal or promoter regions separately. Abbreviations were as follow: KN_NN_R: KN9204_NN_Root, KN_NN_L: KN9204_NN_Leaf, KN_NN_S: KN9204_NN_Seed, KN_LN_R: KN9204_LN_Root, KN_LN_L: KN9204_LN_Leaf, KN_LN_S: KN9204_LN_Seed, J4_NN_R: J411_NN_Root, J4_NN_L: J411_NN_Leaf, J4_NN_S: J411_NN_Seed, J4_LN_R: J411_LN_Root, J4_LN_L: J411_LN_Leaf, J4_LN_S: J411_LN_Seed.  
+
+## Fig. S2. Dynamic changes in the transcriptomes and epigenomes of NMGs  
+
+a. K-means clustering of LN-induced differentially-expressed NMGs in roots, leaves, and seeds. See also Supplemental Table 1. Abbreviations were as follow: NPF: NRT1/PTR FAMILY, NRT2: Nitrate transporter 2, CLC: Chloride channel protein, SLAH: Slow anion channel-associated homologues, AMT: Ammonium transporter, APC: The amino acid–polyamine–choline transporter superfamily, NIA: Nitrate reductase, NIR: Nitrite reductase, GS: Glutamine synthetase, GOGAT: Glutamate synthase, ASN: Asparagine synthetase, AspAT: Aspartate aminotransferase, GDH: Glutamate dehydrogenase, TF: transcription factor.  
+
+b. Percentage of NMGs in different categories marked by differential H3K27ac and H3K27me3. Abbreviations were same with (a).  
+
+c. H3K27ac levels of GS and GOGAT in the seeds of KN9204 and J411 between the two nitrogen availability levels (Wilcox test: ns: \(\mathrm{p} > 0.05\) ; \(\ast :p\leq 0.05\) ; \(\ast \ast :p\leq 0.01\) ).  
+
+d. Representative tracks showing histone modifications and transcriptional changes of TaGS_2D in seeds for the two wheat cultivars and two nitrogen levels.  
+
+e. Grain protein content (GPC) in seeds of KN9204 and J411 under different nitrogen conditions (Student's t-test; ns: \(\mathrm{p} > 0.05\) ; \(\ast :\mathrm{p}\leq 0.05\) ).
+
+<--- Page Split --->
+
+## Fig. S3. The functional influence of cultivar-biased H3K27ac peaks  
+
+Fig. S3. The functional influence of cultivar-biased H3K27ac peaksa. Mean gene expression in J411 (x-axis) versus mean gene expression in KN9204 (y-axis) for genes associated with proximal cultivar-specific H3K27ac peaks separately in the leaf and seed.b. Overlap between genes marked by cultivar-specific H3K27ac and DEGs between cultivars (blue circle) in the leaf (top panel) and seed (bottom panel) (Fisher's exact test was used to calculate the \(p\) -values for the overlaps).c. GO enrichment of genes that are marked by proximal cultivar-specific H3K27ac peaks in KN9204 and J411.d. Heatmaps showing the co-localization between cultivar-specific H3K27ac peaks and H3K4me3/H2A.Z.e. Schematic diagram showing the approach used to link distal cultivar-specific H3K27ac peaks to genes.f. Cross validation of distal cultivar-specific H3K27ac peaks assigned by Hi-C data.g. Fraction of DEGs with distal KN9204-specific/J411-specific regulation (left) or DNA variation within the promoters located in QTLs (right) between KN9204 and J411. Abbreviation: Rt: root tip number; Rd: root diameter; Nup: Nitrogen uptake content; Nct: Nitrogen concentration; MRL: Maximum root length; TKW: Thousand-kernel weight; GPC: Grain protein content; Rs: Root surface area; Rdw: Root dry weight; Ls: flag leaf size.h. Representative tracks of TaXTH25 regulated by distal cultivar-specific H3K27ac peaks in QMRL-7B. Abbreviations were as follow: NN_R: NN_Root, LN_R: LN_Root.i. Positive and negative control of luciferase reporter assay, region tested (positive: State5-8, negative: State5-10) got from data published before<sup>32</sup>.  
+
+## Fig. S4. Dynamic H3K27ac changes in KN9204 and J411  
+
+a. Heatmaps showing the differential H3K27ac peaks in the leaves and seeds of KN9204 and J411 under LN and NN conditions. 
+b. Peak distribution of differential H3K27ac peaks in roots, leaves, and seeds. 
+c. Dynamic H3K27ac and corresponding expression changes in the leaves of KN9204 and J411 seedlings. GO enrichments are shown on the right. Abbreviations were as follow: KN_NN_L: KN9204_NN_Leaf, KN_LN_L: KN9204_LN_Leaf, J4_NN_L: J411_NN_Leaf, J4_LN_L: J411_LN_Leaf. 
+d. Overlap between genes with up-regulated H3K27ac and the up-regulated DEGs for KN9204 and J411 in the seeds (Fisher's exact test was used to the calculate \(p\) -values for the overlaps). 
+e. Experiment design and western blotting of TSA (2 \(\mu \mathrm{M}\) ) treatment of KN9204 and J411 seedlings under the two nitrogen conditions. Abbreviations were as follow: KN_NN_R: KN9204_NN_Root, KN_LN_R: KN9204_LN_Root, J4_NN_R: J411_NN_Root, J4_LN_R: J411_LN_Root.  
+
+Fig. S5. Dynamic H3K27me3 changes in KN9204 and J411
+
+<--- Page Split --->
+
+a. Heatmaps showing the differential H3K27me3 peaks in seven clusters in the seeds of KN9204 and J411.  
+b. Distribution of differential H3K27me3 peaks in the roots and seeds.  
+c. Overlap between genes up-regulated for H3K27me3 and the down-regulated DEGs in the roots of KN9204 (Fisher's exact test was used to calculate the \(p\) -values for the overlaps).  
+d. Overlap between genes up-regulated for H3K27me3 and the down-regulated DEGs in the seeds of J411 (top panel) and KN9204 (bottom panel) (Fisher's exact test was used to calculate the \(p\) -values for the overlaps).  
+
+## Fig. S6. The generation and influence to global H3K27me3 level of Taswn-cr  
+
+a. A phylogenetic tree showing the evolutionary relationships between Ez proteins (one part of PRC2) from Arabidopsis and wheat.  
+b. Expression of TaSWN and TaCLF genes under LN and NN conditions in KN9204 and J411.  
+c. DNA sequence identification the mutated target sites in the three TaSWN genes in the Taswn-cr mutant.  
+d. Peak numbers and lengths of H3K27me3 peaks in Taswn-cr and BW plants under different nitrogen conditions (Wilcox test: ns: \(p > 0.05\) ; \*: \(p \leq 0.05\) ; \*\*: \(p \leq 0.01\) ; \*\*\*: \(p \leq 0.001\) ; \*\*\*\*: \(p \leq 0.0001\) ).  
+e. Profiles of H3K27me3 levels in Taswn-cr and BW under normal nitrogen conditions.  
+f. Peak distribution of the differential H3K27me3 peaks in Fig. 6a.  
+g. Overlap between genes marked by TaSWN-dependent H3K27me3 and genes marked by gain-H3K27me3 under LN condition in KN9204 (Fisher's exact test was used to calculate the \(p\) -values for the overlaps).
+
+<--- Page Split --->
+
+## Figures  
+
+![](images/Figure_1.jpg)
+
+<center>Fig.1 </center>  
+
+A 15- state model characterizes the dynamic chromatin landscape in KN9204 and J411
+
+<--- Page Split --->
+
+a. Experimental design for generating the epigenomic data sets in KN9204 and J411 under different nitrogen conditions.  
+
+b. Chromatin state definitions, genomic annotation enrichments, and expression levels of genes associated with each chromatin state. Chromatin states could be mainly organized into five categories: Pr (promoter), Tr (transcription), En (enhancer-like), Re (repressive), and Ns (No signal) states.  
+
+c. Schematic diagrams representing the chromatin state dynamics under different nitrogen conditions and tissues. State colors are as in (b).  
+
+d. The fractions of bases for the states of the five categories that vary between cultivar difference and nitrogen availability.  
+
+e. Variability scores of histone modifications between cultivar difference and nitrogen availability.  
+
+f. Cross-correlation heatmaps of all H3K27ac peaks which were located in the distal or promoter regions separately for the two cultivars and the two nitrogen conditions. Abbreviations were as follow: KN_NN_R: KN9204_NN_Root, KN_NN_L: KN9204_NN_Leaf, KN_NN_S: KN9204_NN_Seed, KN_LN_R: KN9204_LN_Root, KN_LN_L: KN9204_LN_Leaf, KN_LN_S: KN9204_LN_Seed, J4_NN_R: J411_NN_Root, J4_NN_L: J411_NN_Leaf, J4_NN_S: J411_NN_Seed, J4_LN_R:  
+
+J411_LN_Root, J4_LN_L: J411_LN_Leaf, J4_LN_S: J411_LN_Seed.
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Fig.2 </center>  
+
+## Figure 2  
+
+Epigenomic variations are vital to the expression bias of NMGs  
+
+a. Fraction of the regulatory regions and expression variation of nitrogen metabolism genes (NMGs) between wheat cultivars KN9204 and J411. Abbreviations were as follow: NPF: NRT1/PTR FAMILY, NRT2: Nitrate transporter 2, CLC: Chloride channel protein, SLAH: Slow anion channel-associated homologues,
+
+<--- Page Split --->
+
+AMT: Ammonium transporter, APC: The amino acid- polyamine- choline transporter superfamily, NIA: Nitrate reductase, NIR: Nitrite reductase, GS: Glutamine synthetase, GOGAT: Glutamate synthase, ASN: Asparagine synthetase, AspAT: Aspartate aminotransferase, GDH: Glutamate dehydrogenase, TF: transcription factor.  
+
+b. Enrichment of LN-induced differentially-expressed NMGs based on their functional category in three tissues (Fisher's exact test was used to calculate the p-values for the overlaps). Abbreviations were same with (a).  
+
+c. Proportion of NMGs showing expression variation between KN9204 and J411 with regulatory region variation (top panel), H3K27ac variation (middle panel), and H3K27me3 variation (bottom panel).  
+
+d. Histone modification (H3K27ac and H3K27me3) levels of NRT2 and NPF2.3 in the roots of KN9204 and J411 between the two nitrogen availability levels (Wilcox test: ns: p > 0.05; \*: p ≤ 0.05; \*\*: p ≤ 0.01; \*\*\*\*: p ≤ 0.0001).  
+
+e. Representative tracks showing histone modifications and transcriptional changes of TaNRT2_6B and TaNPF2.3_7B in roots for the two wheat cultivars and two nitrogen levels.  
+
+f. Nitrate content of shoots and roots of KN9204 and J411 under two nitrogen conditions (Student's t-test; ns: p > 0.05; \*: p ≤ 0.05).
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Fig.3 </center>  
+
+## Figure 3  
+
+Distal regulatory region divergence is involved in NUE- trait variation between KN9204 and J411  
+
+a. K-means clustering of the variable H3K27ac peaks.  
+
+b. Peak distribution (distal, genic, or promoter) of cultivar-specific H3K27ac peaks in the wheat genome.
+
+<--- Page Split --->
+
+c. Mean gene expression in J411 (x-axis) versus mean gene expression in KN9204 (y-axis) for genes associated with proximal cultivar-biased H3K27ac peaks separately in the roots (top panel). Overlap between genes marked by cultivar-specific H3K27ac and DEGs between cultivars (blue circle) (bottom panel) (Fisher's exact test was used to calculate the p-values for the overlaps).  
+
+d. Assignment of distal cultivar-specific H3K27ac peaks to potential targets; representative genes and GO enrichment are shown on the right, and representative case is shown at the bottom. Abbreviations were as follow: NN_R: NN_Root, NN_L: NN_Leaf, NN_S: NN_Seed, LN_R: LN_Root, LN_L: LN_Leaf, LN_S: LN_Seed.  
+
+e. Enrichment of target genes with distal cultivar-specific H3K27ac peaks in QTLs between KN9204 and J411 (Fisher's exact test was used to calculate the p-values for the enrichment). Abbreviation: Rt: root tip number; Rd: root diameter; Nup: Nitrogen uptake content; Nct: Nitrogen concentration; MRL: Maximum root length; TKW: Thousand-kernel weight; GPC: Grain protein content; Rs: Root surface area; Rdw: Root dry weight; Ls: flag leaf size.  
+
+f. Schematic diagram illustrating the process used to narrow down potential key genes regulated by distal cultivar-specific regulation in the QTL regions (QMLR-7B in this case).  
+
+g. Representative tracks of TaHyPRP06 regulated by distal cultivar-specific H3K27ac peaks in QMLR-7B. Abbreviations were as follow: NN_R: NN_Root, LN_R: LN_Root.  
+
+h. Luciferase reporter assay of cultivar-specific H3K27ac regions in track shown in (g) and Fig.S3.h (Student's t-test; ns: p > 0.05; \*: p 0.05).
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Fig.4 </center>  
+
+## Figure 4  
+
+Dynamics of H3K27ac reveals the divergent strategies of KN9204 and J411 to low nitrogen conditionsa. The number of differential H3K27ac peaks in response to LN of three tissues (seeds, roots, leaves) in KN9204 and J411.
+
+<--- Page Split --->
+
+b. Heatmaps showing the differential H3K27ac peaks in eight clusters in the roots of KN9204 and J411 under NN and LN conditions.  
+
+c. Dynamic proximal H3K27ac peaks and corresponding gene expression changes in the roots of KN9204 and J411. Representative genes and GO (gene ontology) enrichments are shown on the right. Abbreviations were as follow: KN_NN_R: KN9204_NN_Root, KN_LN_R: KN9204_LN_Root, J4_NN_R: J411_NN_Root, J4_LN_R: J411_LN_Root.  
+
+d. The LN-response-ratio (LRR) of root systems for KN9204 and J411 (Student's t-test; ns: \(\mathsf{p} > 0.05\) ; \*: \(\mathsf{p} \leq 0.05\) ).  
+
+e. The number of root tips in KN9204 and J411 plants under different nitrogen conditions (Student's t-test; ns: \(\mathsf{p} > 0.05\) ; \*: \(\mathsf{p} \leq 0.05\) ).  
+
+f. Phenotypes of the root systems of KN9204 and J411 seedlings in the control (mock) and TSA treatments (2 μM) under different nitrogen conditions. Scale bars=5cm. The right panel is the LRR of the root tips (Student's t-test; ns: \(\mathsf{p} > 0.05\) ; \*: \(\mathsf{p} \leq 0.05\) ).
+
+<--- Page Split --->
+![](images/Figure_5.jpg)
+
+<center>Fig.5 </center>  
+
+Different trends in H3K27me3 shapes adaptation bias between the cultivars KN9204 and J411a. The number of differential (gain or loss) H3K27me3 peaks in response to LN of three tissues in KN9204 and J411.
+
+<--- Page Split --->
+
+b. Heatmaps showing the differential H3K27me3 peaks in the seven clusters in the roots of KN9204 and J411 under LN and NN conditions.  
+
+c. The number of differential H3K27me3 peaks (in proximal or distal regions) in the root of KN9204 and J411 plants in response to LN.  
+
+d. Overlap between genes with dynamic changes in H3K27me3 and DEGs in the roots (Fisher's exact test was used to calculate the p-values for the overlaps).  
+
+e. GO enrichment of the overlapping genes related to (d).  
+
+f. Representative tracks showing H3K27me3 and transcriptional changes in TaMob1L_6D (top panel), TaNAR2_4A (middle panel), and TaKAO2_4A (bottom panel) for the two cultivars under LN and NN. Abbreviations were as follow: KN_NN_R: KN9204_NN_Root, KN_LN_R: KN9204_LN_Root, J4_NN_R: J411_NN_Root, J4_LN_R: J411_LN_Root.  
+
+g. Sequence motif enrichment for the up-regulated H3K27me3 peaks under LN conditions in KN9204 and J411 (Fisher's exact test was used to calculate the p-values for the enrichment).  
+
+h. Yeast two-hybrid (Y2H) assays showing the interaction between TaERF9_5B/TaERF9_5D/TaBPC_4A and TaEMF2-2A.2/TaSWN (components of PRC2).  
+
+Transformed yeast cells were grown on synthetic media lacking Leu and Trp (-WL) and Leu, Trp, His, and Ade (-WLHA) or Leu, Trp, and His (-WLH) with 1mM 3AT.  
+
+i. BiFC analysis of the interaction between TaERF9_5B/TaERF9_5D/TaBPC_4A and TaSWN-N/TaEMF2_2A.2. Scale bars = 10 mm.
+
+<--- Page Split --->
+![](images/Figure_6.jpg)
+
+<center>Fig.6 </center>  
+
+Rewiring H3K27me3 strengthens root adaptation to low nitrogen fertilization level  
+
+a. Heatmaps showing the differential H3K27me3 peaks in 'BobWhite' (BW) and the Taswn-cr mutant plants in the eight clusters under NN and LN conditions.
+
+<--- Page Split --->
+
+b. The proportion of TaSWN-dependent peaks in the up-regulated H3K27me3 peaks under LN conditions in KN9204 and J411.  
+
+c. Overlap between genes marked by TaSWN-dependent H3K27me3 and marked by gain-of-H3K27me3 under LN conditions in J411 (Fisher's exact test was used to calculate the p-values for the overlap).  
+
+d. Dynamic changes in the TaSWN-dependent H3K27me3 peaks under different nitrogen conditions (NN and LN) in KN9204, J411, BW, and Taswn-cr plants. Representative gene names are shown on the right.  
+
+e. GO enrichment of the overlapping genes related to (d).  
+
+f. The expression changes and fold-changes of the genes in (d). (Wilcoxon test: ns: p > 0.05; ****: p ≤ 0.0001) FC: fold-change.  
+
+g. Scans of root systems of BW and Taswn-cr seedlings grown under different nitrogen conditions. Scale bars = 5 cm.  
+
+h. The number of root tips and total length of the root systems for BW and Taswn-cr seedlings grown under different nitrogen conditions. (Student's t-test; ns: p > 0.05; *: p ≤ 0.05).  
+
+i. The fold-changes in the expression of NRT2 genes in BW, Taswn-cr, J411, and KN9204 seedlings grown under different nitrogen conditions.  
+
+j. 15N-nitrate uptake activity of different seedlings (KN9204, J411, BW, and Taswn-cr) grown under LN conditions (Student's t-test; ns: p > 0.05; *: p ≤ 0.05).
+
+<--- Page Split --->
+![](images/Figure_7.jpg)
+
+<center>Fig.7 </center>  
+
+## Figure 7  
+
+The balanced model of epigenetic regulation for divergent strategies to LN conditions in wheatIn the low- nitrogen condition (Input, left), different adaptation strategy of wheat is selected from the balance between root system development and nitrate uptake transporters (NRT2s). great gain- of- H3K27ac enhance the expression of root development- related genes in KN9204, Conversely, greater gain
+
+<--- Page Split --->
+
+of- H3K27me3 and minor gain- of- H3K27ac reduce root development in J411, but with an associated increase in nitrate uptake transporters (NRT2s) via gain- of- H3K27ac. After the knock- out of TaSWN (Taswn- cr), there is de- repression of root development, accompanied by the loss of H3K27me3 (Decision making, middle). Phenotypically, KN9204 has a developed root architecture (higher expression of root development genes) but lower nitrate uptake rate per unit weight (lower expression of NRT2s). In contrast, J411 has a diverse selection with regards to this balance (Output, right).  
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+- TableS1DifferentialexpressionandhistonemodificationsofNMGs.xlsx- TableS2Cultivarspecificregulation.xlsx- TableS3ERFinQTL.xlsx- TableS4Primerusedinthisstudy.xlsx- NCOMMS2315670RS.pdf
+
+<--- Page Split --->

preprint/preprint__1043f4f38f95212f0604c8a92c6978be37168898eb48b3de83a619b2b0f4a1dc/images_list.json

@@ -0,0 +1,108 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Fig. 1 Obesity increases Gal3 levels in islets and plasma. a, Gal3 levels in plasma of",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 4
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Fig. 2 Gal3 reduces islet and \\(\\beta\\) -cell GSIS. a-d, Glucose-stimulated insulin secretion (GSIS) in INS-1 cells (a), MIN6 cells (b), NC mice primary islets (c), and db/db mice primary islets (d) with Gal3 treatment (80 ng/ml for MIN6 cells, 250 ng/ml for INS-1 cells and islets) for 6 h. Gal3 existed all the time in the experimental process. a, c, d, n=4 biologically independent cell samples; b, n=6 biologically independent cell samples. e, Islet perfusion experiment in islets from 12-week-old NC mice with Gal3 (250 ng/ml) treatment for 6 h. f, KCl (30 mM) stimulated insulin secretion in mice islets with or without Gal3 (250 ng/ml, 1 h) treatment (n=4 biologically independent cell samples). g, Arginine (10 mM) stimulated insulin secretion in mouse islets with glucose (16.8 mM) and Gal3 (250 ng/ml, 1 h) treatment (n=5 biologically independent cell",
+    "footnote": [],
+    "bbox": [
+      [
+        145,
+        108,
+        880,
+        627
+      ]
+    ],
+    "page_idx": 35
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Fig. 3 Gal3 knockout improves diabetic phenotype in HFD-fed and db/db mice. a, Blood Gal3 levels in WT, Gal3+/- and Gal3-/- mice on HFD-fed (n=8 mice). b-e, Body weight (b), intravenous glucose tolerance test (IVGTT) (c), Area under curve (AUC) during IVGTT (d) and first-phase insulin secretion (e) in WT and Gal3+/- mice on HFD-fed after 6 h of fasting. n=7-9 mice. f-i, Plasma glucose (f), Insulin level (g), The AUC of the first phase of insulin secretion (from 0 to 10 min) and the second insulin secretion (from 10 to 120 min) (h), and GIR (i) in hyperglycemic clamp study in Gal3+/- mice on HFD-fed. n=6 mice. j, GSIS in primary islets from Gal3+/- mice on",
+    "footnote": [],
+    "bbox": [
+      [
+        144,
+        88,
+        853,
+        670
+      ]
+    ],
+    "page_idx": 36
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Fig. 4 Depletion macrophages improve GSIS in HFD mice and histologic and gene expression studies of WT and Gal3 KO islets. a, Experimental scheme of HFD mice by clodronate injection. b, Gal3 level in blood. c-f, IPGTT (c), AUC of IPGTT (d), first-phase insulin secretion (e) and GSIS in primary islet from clodronate- or control-treated",
+    "footnote": [],
+    "bbox": [
+      [
+        145,
+        106,
+        833,
+        796
+      ]
+    ],
+    "page_idx": 38
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Fig. 5 Dysfunction of cytosolic calcium in Gal3-treated MIN6 cells and mouse islets.",
+    "footnote": [],
+    "bbox": [
+      [
+        144,
+        83,
+        855,
+        543
+      ]
+    ],
+    "page_idx": 40
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_6.jpg",
+    "caption": "Fig. 6 Gal3 binds to CACNG1 and affects insulin secretion. a, Co- immunoprecipitation of Gal3-GFP and Flag-tagged CACNG1 in 293T cells and the",
+    "footnote": [],
+    "bbox": [
+      [
+        160,
+        85,
+        850,
+        860
+      ]
+    ],
+    "page_idx": 42
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_7.jpg",
+    "caption": "Fig. 7 Inhibition of Gal3 improves GSIS and glucose homeostasis in db/db mice. a-",
+    "footnote": [],
+    "bbox": [
+      [
+        144,
+        113,
+        848,
+        860
+      ]
+    ],
+    "page_idx": 44
+  },
+  {
+    "type": "image",
+    "img_path": "images/Extended_Data_Figure_4.jpg",
+    "caption": "1153 Extended Data Fig. 4 Gene expression studies about dedifferentiation and 1154 inflammation. a, immunofluorescence staining of F4/80 and Gal3 in pancreatic islets,",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 46
+  }
+]

preprint/preprint__105ebfaa1aa85e57a1bd1e06f9ff944b4e5c2031b921bdeebcccc8858f99a982/images_list.json

@@ -0,0 +1,62 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Fig. 1 | The structures and general \\(\\mathrm{d}I / \\mathrm{d}V\\) features of Graphene/WSe₂ heterostructure. a, A representative AFM image of the freshly mechanical exfoliated WSe₂ sheet. Inset: the AFM image of a typical monolayer WSe₂ island. b, A STM image of a typical graphene/WSe₂ heterostructure 5 QD. The height of the GQD is \\(\\sim 0.8 \\mathrm{nm}\\) and the width of edge area of the GQD is \\(\\sim 2.2 \\mathrm{nm}\\) . c, The zoom-in image of the area in black dashed squares from panel b. Inset: the FFT of graphene/WSe₂ heterostructure. The bright spots in the white dotted circles represent the reciprocal lattice of graphene, the bright spots in the blue dotted circles represent the reciprocal lattice of WSe₂, and",
+    "footnote": [],
+    "bbox": [
+      [
+        130,
+        92,
+        866,
+        626
+      ]
+    ],
+    "page_idx": 15
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Fig. 2 | Coexistence of WGMs confinement and ACSs in the GQDs. Top of a to c, The radially dI/dV spectroscopic maps of different GQDs. Bottom of a to c, The calculated space-energy maps of the LDOS of different GQDs with different value of \\(\\beta\\) and \\(r_0\\) . The red dotted lines indicate Dirac point energy. The black solid dots indicate the quasibound states via the WGM confinement, and the purple hollow dots indicate the ACSs.",
+    "footnote": [],
+    "bbox": [
+      [
+        130,
+        93,
+        866,
+        465
+      ]
+    ],
+    "page_idx": 17
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Fig. 3 | Lifting the orbital degeneracy of LLs in the GQD. Top panels of a to c, the radially dI/dV spectroscopic maps on the GQD ( \\(\\beta = 2.4\\) , \\(r_0 = 9 \\text{nm}\\) ) in the case of a series of magnetic fields. Bottom panels of a to c, the calculated space-energy maps of the LDOS of the GQD with different magnetic fields. The \\(m = -1\\) , \\(m = 0\\) , and \\(m = 1\\) indicate the split orbital states of the -1 LL. The red dotted lines indicate Dirac point energy.",
+    "footnote": [],
+    "bbox": [
+      [
+        130,
+        88,
+        868,
+        470
+      ]
+    ],
+    "page_idx": 18
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Fig. 4 | Evolution from the ACSs to unusual LLs in the GQD. The measured LLs at the center of the GQD ( \\(\\beta = 2.4\\) , \\(r_0 = 9 \\mathrm{nm}\\) ) as a function of the square root of the magnetic field \\(\\sqrt{B}\\) . The experimental results are superimposed onto the calculated map of LDOS in the GQD with \\(\\beta = 2.4\\) and \\(r_0 = 9 \\mathrm{nm}\\) . The ACS-R1 and ACS-R2 are two quasi-bound states due to atomic collapse resonance. The full width at half maximum of the peaks in the spectra was used to estimate the error bar in experiment (orange dots).",
+    "footnote": [],
+    "bbox": [
+      [
+        155,
+        90,
+        700,
+        440
+      ]
+    ],
+    "page_idx": 19
+  }
+]

preprint/preprint__105ebfaa1aa85e57a1bd1e06f9ff944b4e5c2031b921bdeebcccc8858f99a982/preprint__105ebfaa1aa85e57a1bd1e06f9ff944b4e5c2031b921bdeebcccc8858f99a982.mmd

@@ -0,0 +1,213 @@
+
+# Coexistence of electron whispering-gallery modes and atomic collapse states in graphene/WSe2 heterostructure quantum dots  
+
+Qi Zheng  Beijing Normal University  Yu-Chen Zhuang  Peking University  Qingfeng Sun  Peking University  Lin He ( \(\boxed{ \begin{array}{r l} \end{array} }\) helin@bnu.edu.cn )  Beijing Normal University  https://orcid.org/0000- 0001- 5251- 1687  
+
+## Article  
+
+Keywords: electron whispering- gallery modes, atomic collapse states, graphene/WSe2 heterostructure quantum dots  
+
+Posted Date: August 13th, 2021  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 763571/v1  
+
+License: © \(\circledast\) This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Version of Record: A version of this preprint was published at Nature Communications on March 24th, 2022. See the published version at https://doi.org/10.1038/s41467- 022- 29251- 2.
+
+<--- Page Split --->
+
+# Title: Coexistence of electron whispering-gallery modes and atomic collapse states in graphene/WSe₂ heterostructure quantum dots  
+
+Authors: Qi Zheng<sup>1,8</sup>, Yu- Chen Zhuang<sup>2,8</sup>, Qing- Feng Sun<sup>2,3,4,†</sup>, Lin He<sup>1,†</sup>  
+
+## Affiliations:  
+
+<sup>1</sup> Center for Advanced Quantum Studies, Department of Physics, Beijing Normal University, Beijing, 100875, People's Republic of China  <sup>2</sup> International Center for Quantum Materials, School of Physics, Peking University, Beijing, 100871, China  <sup>3</sup> Collaborative Innovation Center of Quantum Matter, Beijing 100871, China  <sup>4</sup> Beijing Academy of Quantum Information Sciences, West Bld. #3, No. 10 Xibeiwang East Road, Haidian District, Beijing 100193, China  
+
+<sup>8</sup>These authors contributed equally to this work.  
+
+<sup>†</sup>Correspondence and requests for materials should be addressed to Qing- Feng Sun (email: sunqf@pku.edu.cn) and Lin He (e- mail: helin@bnu.edu.cn).  
+
+The relativistic massless charge carriers with a Fermi velocity of about \(c / 300\) in graphene enable us to realize two distinct types of resonances ( \(c\) , the speed of light in vacuum). One is electron whispering- gallery mode in graphene quantum dots arising from the Klein tunneling of the massless Dirac fermions. The other is atomic collapse state, which has never been observed in experiment with real atoms due to the difficulty of producing heavy nuclei with charge \(Z > 170\) , however, can be realized near a Coulomb impurity in graphene with a charge \(Z \geq 1\) because of the "small" velocity of the Dirac excitations. Here, unexpectedly, we demonstrate that both the electron whispering- gallery modes and atomic collapse states coexist in graphene/WSe₂ heterostructure quantum dots due to the Coulomb- like potential near their edges. By applying a perpendicular magnetic field, evolution from the atomic collapse states to unusual Landau levels in the collapse regime are explored for the first time.
+
+<--- Page Split --->
+
+Many exotic electronic properties of graphene are rooted in its relativistic massless charge carriers (1- 4). For example, the massless Dirac fermions nature of the charge carriers in graphene enables us to demonstrate several oddball predictions by quantum electrodynamics (QED), among which the Klein tunneling (5) and atomic collapse (6- 8) are the two most famous effects that have attracted much attention. Very recently, it was demonstrated that the two effects lead to the formation of two types of quasibound states in graphene (9- 18). The Klein tunneling, i.e., the anisotropic transmission of the massless Dirac fermions across the potential barrier, in graphene leads to the formation of quasibound states in circular \(p\) - \(n\) junctions, i.e., graphene quantum dots (GQDs), via whispering- gallery modes (WGMs) (9- 15). Because of the “small” velocity of the Dirac fermions, a Coulomb impurity in graphene with a charge \(Z \geq 1\) can result in the formation of atomic collapse states (ACSs) around it (16,17). In previous experiments, pronounced resonances of the two types of the quasibound states were clearly observed (9- 18). Due to their distinct underlying origins, the two quasibound states are expected to be observed in the two different systems.  
+
+Here, we demonstrate the coexistence of the electron WGMs and ACSs in graphene/WS \(e_{2}\) heterostructure QDs. Because of the Coulomb- like potential near the edges of the QDs, we observe WGMs near the edge and ACSs in the center of the graphene/WS \(e_{2}\) heterostructure QDs. Moreover, the ACSs are further explored in the presence of high magnetic fields. The study about effect of magnetic fields on the ACSs has a long history (19). However, such a longstanding prediction remains highly controversial because that the theoretical results are contradictory (20- 25) and,
+
+<--- Page Split --->
+
+more importantly, an experimental verification of this fundamental prediction is still lacking up to now (25,26). Our experiments demonstrate that the ACSs still exist in magnetic fields and evolution from the ACSs to unusual Landau levels in the collapse regime are measured.  
+
+The graphene/WSe₂ heterostructure was obtained by using a wet transfer fabrication of a monolayer graphene on mechanical- exfoliated thick WSe₂ sheets (see methods for details of the sample preparation). In our experiment, nanoscale WSe₂ QDs were surprisingly observed on surface of freshly mechanical exfoliated WSe₂ sheets, as shown in atomic force microscopy (AFM) image of Fig. 1a. Usually, the thickness of the WSe₂ QDs is the same as a WSe₂ monolayer and the diameter is less than 10 nm. Such WSe₂ QDs also can be observed when mechanical- exfoliated thick WSe₂ is covered with graphene monolayer, i.e., in the graphene/WSe₂ heterostructure, as shown in scanning tunneling microscope (STM) image of Fig. 1b (also see Fig. S1 for AFM images of the graphene/WSe₂ heterostructure). At present, the exact origin for the emergence of the nanoscale WSe₂ QDs is unclear. In our experiment, nanoscale monolayer- thick WSe₂ anti- dots are usually observed around the WSe₂ QDs (See Fig. S2), which suggests that the WSe₂ QDs are generated from the anti- dots during the process of mechanical exfoliation. Figure 1c shows an atomic- resolved STM image around a graphene/WSe₂ heterostructure QD. No atomic defect and strain structure can be detected in graphene above the WSe₂ QD. Fast Fourier transform (FFT) images inside and outside the QD are identical (see inset of Fig. 1c and fig. S2b) and the relative rotation angle between graphene and WSe₂ is measured as about \(20.4^{\circ}\) (the rotation angle between the WSe₂ QD and the WSe₂ substrate is zero). A schematic side view of the graphene/WSe₂
+
+<--- Page Split --->
+
+heterostructure is shown in Fig. 1d.  
+
+It is interesting to find that the WSe \(_2\) QD strongly modifies electronic properties of the graphene above it. Figure 1e shows two representative scanning tunneling spectroscopy (STS) spectra of the graphene on and off the WSe \(_2\) QD. The STS, i.e., \(\mathrm{d}I / \mathrm{d}V\) , measurement of the graphene off the WSe \(_2\) QD shows a typical V- shaped spectrum profile of graphene with the Dirac point at \(E_D^{off} \approx - 0.2 \mathrm{eV}\) (p- doping). Whereas, spectrum of graphene on the WSe \(_2\) QD displays a series of resonance peaks with the Dirac point estimated as \(E_D^{on} \approx 0.3 \mathrm{eV}\) (n- doping). Obviously, the WSe \(_2\) QD generates a circular p- n junction, i.e., a GQD, on graphene. The almost equally spaced peaks in the spectrum are the quasibound states confined in the GQD via the WGMs (9- 15). Such a result is further confirmed by carrying out STS mapping at different resonance energies (Fig. 1f and Fig. S3). For the WGMs, the quasibound states, except the lowest one, display shell structures and are progressively closer to the GQD edge with increasing the energy (Fig. S7), as observed in our experiment. According to our experiment, the potential difference \((\Delta U)\) on and off the GQDs shows positive correlation to the ratio \((\eta)\) of the number of edge atoms to the number of inner atoms in the WSe \(_2\) QD (see Fig. S4). Therefore, the large potential difference on and off the GQD may arise from the edge of the WSe \(_2\) QD. The dangling bonds at the edge can significantly change the electronic property and work function of the WSe \(_2\) QD (27- 29), generating the large circular electrostatic potential on graphene covering it.  
+
+To further explore electronic properties of the GQDs, we performed the radially \(\mathrm{d}I / \mathrm{d}V\)
+
+<--- Page Split --->
+
+spectroscopic maps of several GQDs with different sizes and potentials, as shown in Fig. 2 (see insets of Fig. S5 for the STM images of the GQDs). By following the spatial dependence of global local density of states (LDOS) in the maps, it is interesting to note that the WSe₂ QDs generate a  
+
+Coulomb- like electrostatic potential: \(V_{\beta}(r) = \left\{ \begin{array}{ll} \hbar \nu_{F} \frac{\beta}{r_{0}}, & r \leq r_{0} \\ \hbar \nu_{F} \frac{\beta}{r}, & r > r_{0} \end{array} \right.\) , where \(\hbar\) is the reduced Planck constant, \(\nu_{F}\) is the Fermi velocity, \(r\) is the distance from the center of GQD, \(\beta = Z\alpha\) with \(\alpha \sim 2.5\) the fine structure constant of graphene (8), and \(r_{0}\) is the cutoff radius of Coulomb potential. We obtained different values of \(\beta\) and \(r_{0}\) for different GQDs. As shown in Fig. 2, the GQDs with different \(\beta\) exhibit quite different features of the quasibound states. For the \(\beta = 2\) GQD, there is only one resonance peak at the center of the GQD (Fig. 2a, Top). Whereas, for the \(\beta = 4.3\) GQD, besides several quasibound states confined via the WGMs at the edge of the GQD, there are three unequally spaced resonance peaks located at the center of the GQD (Fig. 2c, Top panels). To fully understand these unusual quasibound states, we numerically solved the problem for a Coulomb- like electrostatic potential with different values of \(\beta\) and \(r_{0}\) in the graphene monolayer (the values of \(\beta\) and \(r_{0}\) are extracted from our experimental results) (see supplementary information for the details). Bottom panels of Fig. 2 show the theoretical space- energy maps of the LDOS of the GQDs, which are in good agreement with the experimental results (see Fig. S5 for \(\mathrm{d}I / \mathrm{d}V\) spectra and corresponding simulated LDOS at different positions of the GQDs, see Fig. S7 for the spatial distribution of the quasibound states). According to our analysis, the resonance peaks located at the edge of GQDs arise from the quasibound states via the WGMs confinement. The energy levels  
+
+5 constant, \(\nu_{F}\) is the Fermi velocity, \(r\) is the distance from the center of GQD, \(\beta = Z\alpha\) with \(\alpha \sim 2.5\) the fine structure constant of graphene (8), and \(r_{0}\) is the cutoff radius of Coulomb potential. We obtained different values of \(\beta\) and \(r_{0}\) for different GQDs. As shown in Fig. 2, the GQDs with different \(\beta\) exhibit quite different features of the quasibound states. For the \(\beta = 2\) GQD, there is only one resonance peak at the center of the GQD (Fig. 2a, Top). Whereas, for the \(\beta = 4.3\) GQD, besides several quasibound states confined via the WGMs at the edge of the GQD, there are three unequally spaced resonance peaks located at the center of the GQD (Fig. 2c, Top panels). To fully understand these unusual quasibound states, we numerically solved the problem for a Coulomb- like electrostatic potential with different values of \(\beta\) and \(r_{0}\) in the graphene monolayer (the values of \(\beta\) and \(r_{0}\) are extracted from our experimental results) (see supplementary information for the details). Bottom panels of Fig. 2 show the theoretical space- energy maps of the LDOS of the GQDs, which are in good agreement with the experimental results (see Fig. S5 for \(\mathrm{d}I / \mathrm{d}V\) spectra and corresponding simulated LDOS at different positions of the GQDs, see Fig. S7 for the spatial distribution of the quasibound states). According to our analysis, the resonance peaks located at the edge of GQDs arise from the quasibound states via the WGMs confinement. The energy levels
+
+<--- Page Split --->
+
+of these quasibound states at the edge can be estimated as \(\hbar \nu_{F} / R_{e f f}\) (here \(R_{e f f}\) is the effective radius of the GQD), as observed in our experiment (see Fig. S4d) and reported in previous studies (10,12,13). Whereas the energy levels of the quasibound states at the center of the GQD follow an exponential function \(E_{n} = \frac{\hbar \nu_{F} \beta}{r_{0}} e^{-\frac{\pi}{r_{0}} n} + E_{D}\) , where \(\gamma = \sqrt{\beta^{2} - \left(m + 1 / 2\right)^{2}}\) , \(E_{D}\) is the energy of  
+
+5 Dirac point \([m\) denotes the orbital states \((m = 0,\pm 1,\pm 2,\ldots)]\) . This is a characteristic feature of the ACSs in the supercritical regime due to the Coulomb- like electrostatic potential (see Fig. S8 and the details of discuss) (6- 8). Therefore, our experimental results, supported by our theoretical calculation, strongly indicate the coexistence of the WGMs and ACSs in the graphene/WS \(e_{2}\) heterostructure GQDs. In previous studies of the Coulomb impurity in graphene with a supercritical charge, only the ACSs are observed because of the small \(r_{0}\sim 0.5 \mathrm{nm}\) (16,17). In this work, the Coulomb- like potential near the edges of the GQDs and the increase of about one order of magnitude of the \(r_{0}\) allow us to observe both the WGMs and the ACSs.  
+
+The Coulomb- like potential also strongly affects the electronic properties of the GQDs in the presence of magnetic fields. By applying a perpendicular magnetic field, we can observe well defined Landau levels (LLs) of massless Dirac fermions at positions away from the GQD (see Fig. S9a). When approaching the GQD, the Coulomb- like potential generates pronounced bending of the LLs (see Fig. S9a for the experimental result and theoretical simulation). Figure 3 shows radially spectroscopic maps around the \(\beta = 2.4\) GQD in three different magnetic fields. Near the edge of the GQD, the bending of the LLs follows the Coulomb- like electrostatic potential. Inside
+
+<--- Page Split --->
+
+the GQD, complex evolution of LDOS due to the transition from the confinement of the electrostatic potential to confinement of magnetic field is observed with increasing the magnetic field (see Fig. S10 for more experimental data). At \(B = 10 \mathrm{T}\) , we can observe LLs inside the GQD. However, the \(N = - 1\) LL is split into three peaks: two of them with higher energies are localized in the center of the GQD and the third one is mainly located at the edge of the GQD. The splitting does not occur in pairs and the energy spacing of the splitting is as large as \(\sim 40 \mathrm{meV}\) (Fig. S11), which removes valley and spin splitting as the origin of the observed phenomenon. The splitting LLs should be attributed to lifting the orbital degeneracy of LLs, which can be understood by considering the quantum-mechanical electron motion in the presence of a magnetic field and a Coulomb-like electrostatic potential. Considering the effect of the magnetic field and the electrostatic potential, the equation thus reads:  
+
+\[[\nu_{\vec{r}}\vec{\sigma}\cdot (-i\hbar \vec{\nabla} +e\vec{A}) + V_{\vec{\rho}}(\vec{r})]\psi (\vec{r}) = E\psi (\vec{r}), \quad (1)\]  
+
+where \(\vec{\sigma} = (\sigma_{x}, \sigma_{y})\) are the Pauli matrices, \(\vec{A} = (\vec{B} \times \vec{r}) / 2\) is the vector potential (21,25,26), \(e\) is the electron charge. Due to the axial symmetry of the electrostatic potential in the GQD, we can describe the eigenstates by the orbital quantum number \(m\) (here, we neglect spin). In the absence of the GQD, the eigen- energies \(E_{Nm}\) have infinite orbital degeneracy \([\psi_{Nm}(\vec{r})\) where \(m \geq - |N| ]\) independent of \(m\) because of translational invariance. The GQD lifts this orbital degeneracy \(m\) and the LLs are split into a series of sublevels, which exhibit similar behavior as that observed around charged impurities (17,26), due to the Coulomb- like electrostatic potential. However, previous experiments (17,26) in the presence of a magnetic field were limited to a charge impurity in the
+
+<--- Page Split --->
+
+subcritical regime. Further, the small cutoff radius of a charge impurity prohibits to explore the evolution from the ACSs to the LLs in experiment. Such difficulties can be naturally overcome in the studied GQDs.  
+
+The detailed comparison between experiment and theory can be made by numerically solving the problem for two dimensional massless Dirac fermion of graphene monolayer in the presence of Coulomb- like electrostatic potential \(V_{\beta}(\vec{r})\) and a magnetic field \(B\) (see supplementary information for the details). The calculated radially LDOS maps in the different magnetic fields display that the orbital degeneracy is lifted, which is well consistent with our experimental results (Fig. 3). Based on the calculated results, we can identify the orbital states of the split- 1 LL (Marked in Fig. 3). Thanks to the high- quality LLs in the GQD, \(m = - 1\) orbital state of - 1 LL can be clearly identified and exhibits some characteristics distinguished from that observed in the subcritical regime (17,26). The most important feature is that the \(m = - 1\) orbital state can be viewed as the evolution of the ACS with increasing magnetic field. At zero magnetic field, the broad ACS is located at the center of the GQD and, interestingly, the narrower \(m = - 1\) orbital state appears in the same energy region in the presence of high magnetic field. Such a result indicates directly connection of the ACS and the lowest orbital state ( \(m = - 1\) ) of the - 1 LL.  
+
+To better explore the evolution of the ACS in the presence of magnetic fields, we summarize the measured LLs at the center of the \(\beta = 2.4\) GQD as a function of the square root of the magnetic field \(\sqrt{B}\) (red dots in Fig. 4, see Fig. S12 for the corresponding STS spectra). The evolution of
+
+<--- Page Split --->
+
+LLs displayed a nonlinear dependence on the square root of the magnetic field, which is quite different from the feature of pristine graphene monolayer under magnetic field. The theoretical map of LDOS at the center of the \(\beta = 2.4\) GQD is also plotted as a function of \(\sqrt{B}\) (see supplementary information for the details), as shown in Fig. 4. With increasing magnetic field, the  
+
+5 perturbed LLs \((N = 0, N = - 1, N = - 2)\) display nonlinear dependence on \(\sqrt{B}\) . At a higher \(\sqrt{B}\) , the - 1 LL and - 2 LL are well distinctive, which split into low- energy orbital states \((m = - 1, m = 0)\) . However, we did not observe the splitting of the - 2 LL in the experiment, which is probably due to the large full width at half maximum (FWHM) of the LL peaks, prohibiting the observation of the splitting in the experiment. Furthermore, the ACS- R1 resonance is obvious at lower \(\sqrt{B}\) , and is well connected to the \(m = - 1\) orbital state of - 1 LL. Similarly, ACS- R2 resonance has the similar characteristic, connected to the \(m = - 1\) orbital state of - 2 LL. However, such a feature is harder to be recognized in the experiment due to the broadening peak of the - 2 LL. Our experiments, complemented by theoretical calculations, explicitly demonstrated the existence of ACSs in the presence of high magnetic fields and revealed the close connection between the ACS and the lowest orbital state \((m = - 1)\) of the LLs.  
+
+## References  
+
+1. Castro Neto, A. H., Peres, N. M. R., Novoselov, K. S. & Geim, A. K. The electronic properties of graphene. Rev. Mod. Phys. 81, 109-162 (2009).  
+2. Vozmediano, M. A. H., Katsnelson, M. I. & Guinea, F. Gauge fields in graphene. Physics Rep. 496, 109-148 (2010).
+
+<--- Page Split --->
+
+3. Das Sarma, S., Adam, S., Hwang, E. & Rossi, E. Electronic transport in two-dimensional graphene. Rev. Mod. Phys. 83, 407-470 (2011).  
+
+4. Goerbig, M. O. Electronic properties of graphene in a strong magnetic field. Rev. Mod. Phys. 83, 1193-1243 (2011).  
+
+5. Katsnelson, M. I., Novoselov, K. S. & Geim, A. K. Chiral tunnelling and the Klein paradox in graphene. Nat. Phys. 2, 620-625 (2006).  
+
+6. Pereira, V. M., Nilsson, J. & Castro Neto, A. H. Coulomb impurity problem in graphene. Phys. Rev. Lett. 99, 166802 (2007).  
+
+7. Shytov, A. V., Katsnelson, M. I. & Levitov, L. S. Vacuum Polarization and Screening of Supercritical Impurities in Graphene. Phys. Rev. Lett. 99, 236801 (2007).  
+
+8. Shytov, A. V., Katsnelson, M. I. & Levitov, L. S. Atomic collapse and quasi-rydberg states in graphene. Phys. Rev. Lett. 99, 246802 (2007).  
+
+9. Zhao, Y. et al. Creating and probing electron whispering-gallery modes in graphene. Science 348, 672-675 (2015).  
+
+15. Gutiérrez, C., Brown, L., Kim, C.-J., Park, J. & Pasupathy, A. N. Klein tunnelling and electron trapping in nanometre-scale graphene quantum dots. Nat. Phys. 12, 1069-1075 (2016).  
+
+11. Lee, J. et al. Imaging electrostatically confined Dirac fermions in graphene quantum dots. Nat. Phys. 12, 1032-1036 (2016).  
+
+12. Bai, K.-K. et al. Generating nanoscale and atomically-sharp p-n junctions in graphene via monolayer-vacancy-island engineering of Cu surface. Phys. Rev. B 97, 045413 (2018).
+
+<--- Page Split --->
+
+13. Fu, Z. Q., Bai, K. K., Ren, Y. N., Zhou, J. J. & He, L. Coulomb interaction in quasibound states of graphene quantum dots. Phys. Rev. B 101, 235310 (2020).  
+
+14. Fu, Z.-Q. et al. Relativistic Artificial Molecules Realized by Two Coupled Graphene Quantum Dots. Nano Lett. 20, 6738 (2020).  
+
+15. Ghahari, F. et al. An on/off Berry phase switch in circular graphene resonators. Science 356, 845 (2017).  
+
+16. Wang, Y. et al. Observing atomic collapse resonances in artificial nuclei on graphene. Science 340, 734-737 (2013).  
+
+17. Mao, J. et al. Realization of a tunable artificial atom at a supercritically charged vacancy in graphene. Nat. Phys. 12, 545-549 (2016).  
+
+18. Jiang, Y. et al. Tuning a circular p-n junction in graphene from quantum confinement to optical guiding. Nat. Nanotechnol. 12, 1045-1049 (2017).  
+
+19. Reinhardt, J. & Greiner, W. Quantum electrodynamics of strong fields. Rep. Prog. Phys. 40, 219-295 (1977).  
+
+20. Gamayun, O. V., Gorbar, E. V. & Gusynin, V. P. Magnetic field driven instability of a charged center in graphene. Phys. Rev. B 83, 235104 (2011).  
+
+21. Sobol, O. O., Pyatkovskiy, P. K., Gorbar, E. V. & Gusynin, V. P. Screening of a charged impurity in graphene in a magnetic field. Phys. Rev. B 94, 115409 (2016).  
+
+22. Zhang, Y., Barlas, Y. & Yang, K. Coulomb impurity under magnetic field in graphene: A semiclassical approach. Phys. Rev. B 85, 165423 (2012).
+
+<--- Page Split --->
+
+23. Maier, T. & Siedentop, H. Stability of impurities with Coulomb potential in graphene with homogeneous magnetic field. J. Math. Phys. 53, 095207 (2012).  
+
+24. Kim, S. C. & Eric Yang, S. R. Coulomb impurity problem of graphene in magnetic fields. Ann. Phys. 347, 21-31 (2014).  
+
+25. Moldovan, D., Masir, M. R. & Peeters, F. M. Magnetic field dependence of the atomic collapse state in graphene. 2D Mater. 5, 015017 (2018).  
+
+26. Luican-Mayer, A. et al. Screening charged impurities and lifting the orbital degeneracy in graphene by populating landau levels. Phys. Rev. Lett. 112, 036804 (2014).  
+
+27. Zhang, Y. et al. Electronic Structure, Surface Doping, and Optical Response in Epitaxial WSe₂ Thin Films. Nano Lett. 16, 2485-2491 (2016).  
+
+28. Addou, R. & Wallace, R. M. Surface Analysis of WSe₂ Crystals: Spatial and Electronic Variability. ACS Appl. Mater. Interfaces 8, 26400-26406 (2016).  
+
+29. Kahn, A. Fermi level, work function and vacuum level. Mater. Horiz. 3, 7-10 (2016).  
+
+## Acknowledgments:  
+
+15 This work was supported by the National Natural Science Foundation of China (Grant Nos. 11974050, 11674029, 11921005) and National Key R and D Program of China (Grant No. 2017YFA0303301). L.H. also acknowledges support from the National Program for Support of Top-notch Young Professionals, support from “the Fundamental Research Funds for the Central Universities”, and support from “Chang Jiang Scholars Program”.
+
+<--- Page Split --->
+
+## Author contributions  
+
+Q.Z. performed the sample synthesis, characterization and STM/STS measurements. Q.Z., Y.C.Z., and L.H. analyzed the data. Y.C.Z. carried out the theoretical calculations. L.H. conceived and provided advice on the experiment and analysis. Q.F.S. conceived and provided advice on the theoretical calculations. Q.Z. and L.H. wrote the paper with the input from others. All authors participated in the data discussion.  
+
+## Methods  
+
+CVD Growth of Graphene. The large area graphene monolayer films were grown on a \(20 \times 20 \mathrm{mm}^2\) polycrystalline copper (Cu) foil (Alfa Aesar, \(99.8\%\) purity, \(25 \mu \mathrm{m}\) thick) via a low pressure chemical vapor deposition (LPCVD) method. The cleaned Cu foil was loaded into one quartz boat in center of the tube furnace. Ar flow of 50 sccm (Standard Cubic Centimeter per Minutes) and \(\mathrm{H}_2\) flow of 50 sccm were maintained throughout the whole growth process. The Cu foil was heated from room temperature to \(1030^{\circ}\mathrm{C}\) in 30 min and annealed at \(1030^{\circ}\mathrm{C}\) for six hours. Then \(\mathrm{CH}_4\) flow of 5 sccm was introduced for 20 min to grow high- quality large area graphene monolayer. Finally, the furnace was cooled down naturally to room temperature.  
+
+Construction of graphene/WS \(\mathbf{e}_2\) heterostructure. We used conventional wet etching technique with polymethyl methacrylate (PMMA) to transfer graphene monolayer onto the substrate. PMMA was first uniformly coated on Cu foil with graphene monolayer. We transferred the Cu/graphene/PMMA film into ammonium persulfate solution, and then the underlying Cu foil was
+
+<--- Page Split --->
+
+etched away. The graphene/PMMA film was cleaned in deionized water for hours. The WSe \(_2\) crystal was separated into thick- layer WSe \(_2\) sheets by traditional mechanical exfoliation technology and then transferred to \(8 \times 8 \mathrm{mm}^2\) highly N- doped Si wafer [(100) oriented, 500 \(\mu \mathrm{m}\) thick]. We placed graphene/PMMA onto Si wafer which has been transferred with WSe \(_2\) sheets in advance. Finally, the PMMA was removed by acetone and then annealed in low pressure with Ar flow of 50 sccm and \(\mathrm{H}_2\) flow of 50 sccm at \(\sim 300^{\circ}\mathrm{C}\) for 1 hours.  
+
+AFM, STM and STS Measurements. The topographical images are measured by atomic force microscope (AFM, Bruker Multimode 8) with a tapping mode. We employed the n- doped Si tip coated with Platinum- Iridium (Bruker, SCM- PIT- V2, frequency 50- 100KHz, spring constant 1.5- 6 N/m) to characterize WSe \(_2\) and graphene/WSe \(_2\) heterostructure samples. STM/STS measurements were performed in low- temperature (77 K for Fig. S5a and c, 4.2 K for Fig. S5b) and ultrahigh- vacuum ( \(\sim 10^{- 10}\) Torr) scanning probe microscopes [USM- 1400 (77 K) and USM- 1300 (4.2 K)] from UNISOKU. The tips were obtained by chemical etching from a Pt/Ir (80:20%) alloy wire. The differential conductance (dI/dV) measurements were taken by a standard lock- in technique with an ac bias modulation of 5 mV and 793 Hz signal added to the tunneling bias.
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Fig. 1 | The structures and general \(\mathrm{d}I / \mathrm{d}V\) features of Graphene/WSe₂ heterostructure. a, A representative AFM image of the freshly mechanical exfoliated WSe₂ sheet. Inset: the AFM image of a typical monolayer WSe₂ island. b, A STM image of a typical graphene/WSe₂ heterostructure 5 QD. The height of the GQD is \(\sim 0.8 \mathrm{nm}\) and the width of edge area of the GQD is \(\sim 2.2 \mathrm{nm}\) . c, The zoom-in image of the area in black dashed squares from panel b. Inset: the FFT of graphene/WSe₂ heterostructure. The bright spots in the white dotted circles represent the reciprocal lattice of graphene, the bright spots in the blue dotted circles represent the reciprocal lattice of WSe₂, and </center>
+
+<--- Page Split --->
+
+the bright spots in the green dotted circles represent moiré structure of the graphene/ \(\mathrm{WSe_2}\) heterostructure. d, Schematic structure of the graphene/ \(\mathrm{WSe_2}\) heterostructure QD. e, The \(\mathrm{d}I / \mathrm{d}V\) spectra taken inside [marked by dark green pentagram in b] and outside [marked by blue pentagram in b] the GQD. f, The \(\mathrm{d}I / \mathrm{d}V\) maps with different energies [N1 and N2 marked in e] of the GQD.  
+
+5 the GQD.
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Fig. 2 | Coexistence of WGMs confinement and ACSs in the GQDs. Top of a to c, The radially dI/dV spectroscopic maps of different GQDs. Bottom of a to c, The calculated space-energy maps of the LDOS of different GQDs with different value of \(\beta\) and \(r_0\) . The red dotted lines indicate Dirac point energy. The black solid dots indicate the quasibound states via the WGM confinement, and the purple hollow dots indicate the ACSs. </center>
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Fig. 3 | Lifting the orbital degeneracy of LLs in the GQD. Top panels of a to c, the radially dI/dV spectroscopic maps on the GQD ( \(\beta = 2.4\) , \(r_0 = 9 \text{nm}\) ) in the case of a series of magnetic fields. Bottom panels of a to c, the calculated space-energy maps of the LDOS of the GQD with different magnetic fields. The \(m = -1\) , \(m = 0\) , and \(m = 1\) indicate the split orbital states of the -1 LL. The red dotted lines indicate Dirac point energy. </center>
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Fig. 4 | Evolution from the ACSs to unusual LLs in the GQD. The measured LLs at the center of the GQD ( \(\beta = 2.4\) , \(r_0 = 9 \mathrm{nm}\) ) as a function of the square root of the magnetic field \(\sqrt{B}\) . The experimental results are superimposed onto the calculated map of LDOS in the GQD with \(\beta = 2.4\) and \(r_0 = 9 \mathrm{nm}\) . The ACS-R1 and ACS-R2 are two quasi-bound states due to atomic collapse resonance. The full width at half maximum of the peaks in the spectra was used to estimate the error bar in experiment (orange dots). </center>
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+SupplementaryInformation.docx
+
+<--- Page Split --->

preprint/preprint__105ebfaa1aa85e57a1bd1e06f9ff944b4e5c2031b921bdeebcccc8858f99a982/preprint__105ebfaa1aa85e57a1bd1e06f9ff944b4e5c2031b921bdeebcccc8858f99a982_det.mmd

@@ -0,0 +1,275 @@
+<|ref|>title<|/ref|><|det|>[[44, 108, 920, 208]]<|/det|>
+# Coexistence of electron whispering-gallery modes and atomic collapse states in graphene/WSe2 heterostructure quantum dots  
+
+<|ref|>text<|/ref|><|det|>[[44, 230, 633, 410]]<|/det|>
+Qi Zheng  Beijing Normal University  Yu-Chen Zhuang  Peking University  Qingfeng Sun  Peking University  Lin He ( \(\boxed{ \begin{array}{r l} \end{array} }\) helin@bnu.edu.cn )  Beijing Normal University  https://orcid.org/0000- 0001- 5251- 1687  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 451, 102, 468]]<|/det|>
+## Article  
+
+<|ref|>text<|/ref|><|det|>[[44, 490, 925, 531]]<|/det|>
+Keywords: electron whispering- gallery modes, atomic collapse states, graphene/WSe2 heterostructure quantum dots  
+
+<|ref|>text<|/ref|><|det|>[[44, 550, 318, 569]]<|/det|>
+Posted Date: August 13th, 2021  
+
+<|ref|>text<|/ref|><|det|>[[44, 588, 463, 607]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 763571/v1  
+
+<|ref|>text<|/ref|><|det|>[[44, 625, 911, 667]]<|/det|>
+License: © \(\circledast\) This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+<|ref|>text<|/ref|><|det|>[[42, 704, 925, 747]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on March 24th, 2022. See the published version at https://doi.org/10.1038/s41467- 022- 29251- 2.
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[128, 91, 870, 154]]<|/det|>
+# Title: Coexistence of electron whispering-gallery modes and atomic collapse states in graphene/WSe₂ heterostructure quantum dots  
+
+<|ref|>text<|/ref|><|det|>[[115, 174, 696, 195]]<|/det|>
+Authors: Qi Zheng<sup>1,8</sup>, Yu- Chen Zhuang<sup>2,8</sup>, Qing- Feng Sun<sup>2,3,4,†</sup>, Lin He<sup>1,†</sup>  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 208, 217, 224]]<|/det|>
+## Affiliations:  
+
+<|ref|>text<|/ref|><|det|>[[70, 228, 886, 370]]<|/det|>
+<sup>1</sup> Center for Advanced Quantum Studies, Department of Physics, Beijing Normal University, Beijing, 100875, People's Republic of China  <sup>2</sup> International Center for Quantum Materials, School of Physics, Peking University, Beijing, 100871, China  <sup>3</sup> Collaborative Innovation Center of Quantum Matter, Beijing 100871, China  <sup>4</sup> Beijing Academy of Quantum Information Sciences, West Bld. #3, No. 10 Xibeiwang East Road, Haidian District, Beijing 100193, China  
+
+<|ref|>text<|/ref|><|det|>[[112, 391, 495, 409]]<|/det|>
+<sup>8</sup>These authors contributed equally to this work.  
+
+<|ref|>text<|/ref|><|det|>[[112, 412, 883, 452]]<|/det|>
+<sup>†</sup>Correspondence and requests for materials should be addressed to Qing- Feng Sun (email: sunqf@pku.edu.cn) and Lin He (e- mail: helin@bnu.edu.cn).  
+
+<|ref|>text<|/ref|><|det|>[[111, 476, 886, 911]]<|/det|>
+The relativistic massless charge carriers with a Fermi velocity of about \(c / 300\) in graphene enable us to realize two distinct types of resonances ( \(c\) , the speed of light in vacuum). One is electron whispering- gallery mode in graphene quantum dots arising from the Klein tunneling of the massless Dirac fermions. The other is atomic collapse state, which has never been observed in experiment with real atoms due to the difficulty of producing heavy nuclei with charge \(Z > 170\) , however, can be realized near a Coulomb impurity in graphene with a charge \(Z \geq 1\) because of the "small" velocity of the Dirac excitations. Here, unexpectedly, we demonstrate that both the electron whispering- gallery modes and atomic collapse states coexist in graphene/WSe₂ heterostructure quantum dots due to the Coulomb- like potential near their edges. By applying a perpendicular magnetic field, evolution from the atomic collapse states to unusual Landau levels in the collapse regime are explored for the first time.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[60, 90, 888, 650]]<|/det|>
+Many exotic electronic properties of graphene are rooted in its relativistic massless charge carriers (1- 4). For example, the massless Dirac fermions nature of the charge carriers in graphene enables us to demonstrate several oddball predictions by quantum electrodynamics (QED), among which the Klein tunneling (5) and atomic collapse (6- 8) are the two most famous effects that have attracted much attention. Very recently, it was demonstrated that the two effects lead to the formation of two types of quasibound states in graphene (9- 18). The Klein tunneling, i.e., the anisotropic transmission of the massless Dirac fermions across the potential barrier, in graphene leads to the formation of quasibound states in circular \(p\) - \(n\) junctions, i.e., graphene quantum dots (GQDs), via whispering- gallery modes (WGMs) (9- 15). Because of the “small” velocity of the Dirac fermions, a Coulomb impurity in graphene with a charge \(Z \geq 1\) can result in the formation of atomic collapse states (ACSs) around it (16,17). In previous experiments, pronounced resonances of the two types of the quasibound states were clearly observed (9- 18). Due to their distinct underlying origins, the two quasibound states are expected to be observed in the two different systems.  
+
+<|ref|>text<|/ref|><|det|>[[65, 682, 890, 910]]<|/det|>
+Here, we demonstrate the coexistence of the electron WGMs and ACSs in graphene/WS \(e_{2}\) heterostructure QDs. Because of the Coulomb- like potential near the edges of the QDs, we observe WGMs near the edge and ACSs in the center of the graphene/WS \(e_{2}\) heterostructure QDs. Moreover, the ACSs are further explored in the presence of high magnetic fields. The study about effect of magnetic fields on the ACSs has a long history (19). However, such a longstanding prediction remains highly controversial because that the theoretical results are contradictory (20- 25) and,
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 90, 885, 196]]<|/det|>
+more importantly, an experimental verification of this fundamental prediction is still lacking up to now (25,26). Our experiments demonstrate that the ACSs still exist in magnetic fields and evolution from the ACSs to unusual Landau levels in the collapse regime are measured.  
+
+<|ref|>text<|/ref|><|det|>[[67, 228, 886, 920]]<|/det|>
+The graphene/WSe₂ heterostructure was obtained by using a wet transfer fabrication of a monolayer graphene on mechanical- exfoliated thick WSe₂ sheets (see methods for details of the sample preparation). In our experiment, nanoscale WSe₂ QDs were surprisingly observed on surface of freshly mechanical exfoliated WSe₂ sheets, as shown in atomic force microscopy (AFM) image of Fig. 1a. Usually, the thickness of the WSe₂ QDs is the same as a WSe₂ monolayer and the diameter is less than 10 nm. Such WSe₂ QDs also can be observed when mechanical- exfoliated thick WSe₂ is covered with graphene monolayer, i.e., in the graphene/WSe₂ heterostructure, as shown in scanning tunneling microscope (STM) image of Fig. 1b (also see Fig. S1 for AFM images of the graphene/WSe₂ heterostructure). At present, the exact origin for the emergence of the nanoscale WSe₂ QDs is unclear. In our experiment, nanoscale monolayer- thick WSe₂ anti- dots are usually observed around the WSe₂ QDs (See Fig. S2), which suggests that the WSe₂ QDs are generated from the anti- dots during the process of mechanical exfoliation. Figure 1c shows an atomic- resolved STM image around a graphene/WSe₂ heterostructure QD. No atomic defect and strain structure can be detected in graphene above the WSe₂ QD. Fast Fourier transform (FFT) images inside and outside the QD are identical (see inset of Fig. 1c and fig. S2b) and the relative rotation angle between graphene and WSe₂ is measured as about \(20.4^{\circ}\) (the rotation angle between the WSe₂ QD and the WSe₂ substrate is zero). A schematic side view of the graphene/WSe₂
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 92, 395, 111]]<|/det|>
+heterostructure is shown in Fig. 1d.  
+
+<|ref|>text<|/ref|><|det|>[[111, 147, 886, 840]]<|/det|>
+It is interesting to find that the WSe \(_2\) QD strongly modifies electronic properties of the graphene above it. Figure 1e shows two representative scanning tunneling spectroscopy (STS) spectra of the graphene on and off the WSe \(_2\) QD. The STS, i.e., \(\mathrm{d}I / \mathrm{d}V\) , measurement of the graphene off the WSe \(_2\) QD shows a typical V- shaped spectrum profile of graphene with the Dirac point at \(E_D^{off} \approx - 0.2 \mathrm{eV}\) (p- doping). Whereas, spectrum of graphene on the WSe \(_2\) QD displays a series of resonance peaks with the Dirac point estimated as \(E_D^{on} \approx 0.3 \mathrm{eV}\) (n- doping). Obviously, the WSe \(_2\) QD generates a circular p- n junction, i.e., a GQD, on graphene. The almost equally spaced peaks in the spectrum are the quasibound states confined in the GQD via the WGMs (9- 15). Such a result is further confirmed by carrying out STS mapping at different resonance energies (Fig. 1f and Fig. S3). For the WGMs, the quasibound states, except the lowest one, display shell structures and are progressively closer to the GQD edge with increasing the energy (Fig. S7), as observed in our experiment. According to our experiment, the potential difference \((\Delta U)\) on and off the GQDs shows positive correlation to the ratio \((\eta)\) of the number of edge atoms to the number of inner atoms in the WSe \(_2\) QD (see Fig. S4). Therefore, the large potential difference on and off the GQD may arise from the edge of the WSe \(_2\) QD. The dangling bonds at the edge can significantly change the electronic property and work function of the WSe \(_2\) QD (27- 29), generating the large circular electrostatic potential on graphene covering it.  
+
+<|ref|>text<|/ref|><|det|>[[133, 863, 884, 884]]<|/det|>
+To further explore electronic properties of the GQDs, we performed the radially \(\mathrm{d}I / \mathrm{d}V\)
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 89, 886, 201]]<|/det|>
+spectroscopic maps of several GQDs with different sizes and potentials, as shown in Fig. 2 (see insets of Fig. S5 for the STM images of the GQDs). By following the spatial dependence of global local density of states (LDOS) in the maps, it is interesting to note that the WSe₂ QDs generate a  
+
+<|ref|>text<|/ref|><|det|>[[75, 216, 886, 280]]<|/det|>
+Coulomb- like electrostatic potential: \(V_{\beta}(r) = \left\{ \begin{array}{ll} \hbar \nu_{F} \frac{\beta}{r_{0}}, & r \leq r_{0} \\ \hbar \nu_{F} \frac{\beta}{r}, & r > r_{0} \end{array} \right.\) , where \(\hbar\) is the reduced Planck constant, \(\nu_{F}\) is the Fermi velocity, \(r\) is the distance from the center of GQD, \(\beta = Z\alpha\) with \(\alpha \sim 2.5\) the fine structure constant of graphene (8), and \(r_{0}\) is the cutoff radius of Coulomb potential. We obtained different values of \(\beta\) and \(r_{0}\) for different GQDs. As shown in Fig. 2, the GQDs with different \(\beta\) exhibit quite different features of the quasibound states. For the \(\beta = 2\) GQD, there is only one resonance peak at the center of the GQD (Fig. 2a, Top). Whereas, for the \(\beta = 4.3\) GQD, besides several quasibound states confined via the WGMs at the edge of the GQD, there are three unequally spaced resonance peaks located at the center of the GQD (Fig. 2c, Top panels). To fully understand these unusual quasibound states, we numerically solved the problem for a Coulomb- like electrostatic potential with different values of \(\beta\) and \(r_{0}\) in the graphene monolayer (the values of \(\beta\) and \(r_{0}\) are extracted from our experimental results) (see supplementary information for the details). Bottom panels of Fig. 2 show the theoretical space- energy maps of the LDOS of the GQDs, which are in good agreement with the experimental results (see Fig. S5 for \(\mathrm{d}I / \mathrm{d}V\) spectra and corresponding simulated LDOS at different positions of the GQDs, see Fig. S7 for the spatial distribution of the quasibound states). According to our analysis, the resonance peaks located at the edge of GQDs arise from the quasibound states via the WGMs confinement. The energy levels  
+
+<|ref|>text<|/ref|><|det|>[[70, 295, 886, 896]]<|/det|>
+5 constant, \(\nu_{F}\) is the Fermi velocity, \(r\) is the distance from the center of GQD, \(\beta = Z\alpha\) with \(\alpha \sim 2.5\) the fine structure constant of graphene (8), and \(r_{0}\) is the cutoff radius of Coulomb potential. We obtained different values of \(\beta\) and \(r_{0}\) for different GQDs. As shown in Fig. 2, the GQDs with different \(\beta\) exhibit quite different features of the quasibound states. For the \(\beta = 2\) GQD, there is only one resonance peak at the center of the GQD (Fig. 2a, Top). Whereas, for the \(\beta = 4.3\) GQD, besides several quasibound states confined via the WGMs at the edge of the GQD, there are three unequally spaced resonance peaks located at the center of the GQD (Fig. 2c, Top panels). To fully understand these unusual quasibound states, we numerically solved the problem for a Coulomb- like electrostatic potential with different values of \(\beta\) and \(r_{0}\) in the graphene monolayer (the values of \(\beta\) and \(r_{0}\) are extracted from our experimental results) (see supplementary information for the details). Bottom panels of Fig. 2 show the theoretical space- energy maps of the LDOS of the GQDs, which are in good agreement with the experimental results (see Fig. S5 for \(\mathrm{d}I / \mathrm{d}V\) spectra and corresponding simulated LDOS at different positions of the GQDs, see Fig. S7 for the spatial distribution of the quasibound states). According to our analysis, the resonance peaks located at the edge of GQDs arise from the quasibound states via the WGMs confinement. The energy levels
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 87, 886, 262]]<|/det|>
+of these quasibound states at the edge can be estimated as \(\hbar \nu_{F} / R_{e f f}\) (here \(R_{e f f}\) is the effective radius of the GQD), as observed in our experiment (see Fig. S4d) and reported in previous studies (10,12,13). Whereas the energy levels of the quasibound states at the center of the GQD follow an exponential function \(E_{n} = \frac{\hbar \nu_{F} \beta}{r_{0}} e^{-\frac{\pi}{r_{0}} n} + E_{D}\) , where \(\gamma = \sqrt{\beta^{2} - \left(m + 1 / 2\right)^{2}}\) , \(E_{D}\) is the energy of  
+
+<|ref|>text<|/ref|><|det|>[[72, 275, 886, 585]]<|/det|>
+5 Dirac point \([m\) denotes the orbital states \((m = 0,\pm 1,\pm 2,\ldots)]\) . This is a characteristic feature of the ACSs in the supercritical regime due to the Coulomb- like electrostatic potential (see Fig. S8 and the details of discuss) (6- 8). Therefore, our experimental results, supported by our theoretical calculation, strongly indicate the coexistence of the WGMs and ACSs in the graphene/WS \(e_{2}\) heterostructure GQDs. In previous studies of the Coulomb impurity in graphene with a supercritical charge, only the ACSs are observed because of the small \(r_{0}\sim 0.5 \mathrm{nm}\) (16,17). In this work, the Coulomb- like potential near the edges of the GQDs and the increase of about one order of magnitude of the \(r_{0}\) allow us to observe both the WGMs and the ACSs.  
+
+<|ref|>text<|/ref|><|det|>[[65, 619, 886, 890]]<|/det|>
+The Coulomb- like potential also strongly affects the electronic properties of the GQDs in the presence of magnetic fields. By applying a perpendicular magnetic field, we can observe well defined Landau levels (LLs) of massless Dirac fermions at positions away from the GQD (see Fig. S9a). When approaching the GQD, the Coulomb- like potential generates pronounced bending of the LLs (see Fig. S9a for the experimental result and theoretical simulation). Figure 3 shows radially spectroscopic maps around the \(\beta = 2.4\) GQD in three different magnetic fields. Near the edge of the GQD, the bending of the LLs follows the Coulomb- like electrostatic potential. Inside
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 90, 886, 528]]<|/det|>
+the GQD, complex evolution of LDOS due to the transition from the confinement of the electrostatic potential to confinement of magnetic field is observed with increasing the magnetic field (see Fig. S10 for more experimental data). At \(B = 10 \mathrm{T}\) , we can observe LLs inside the GQD. However, the \(N = - 1\) LL is split into three peaks: two of them with higher energies are localized in the center of the GQD and the third one is mainly located at the edge of the GQD. The splitting does not occur in pairs and the energy spacing of the splitting is as large as \(\sim 40 \mathrm{meV}\) (Fig. S11), which removes valley and spin splitting as the origin of the observed phenomenon. The splitting LLs should be attributed to lifting the orbital degeneracy of LLs, which can be understood by considering the quantum-mechanical electron motion in the presence of a magnetic field and a Coulomb-like electrostatic potential. Considering the effect of the magnetic field and the electrostatic potential, the equation thus reads:  
+
+<|ref|>equation<|/ref|><|det|>[[373, 541, 874, 568]]<|/det|>
+\[[\nu_{\vec{r}}\vec{\sigma}\cdot (-i\hbar \vec{\nabla} +e\vec{A}) + V_{\vec{\rho}}(\vec{r})]\psi (\vec{r}) = E\psi (\vec{r}), \quad (1)\]  
+
+<|ref|>text<|/ref|><|det|>[[111, 581, 886, 893]]<|/det|>
+where \(\vec{\sigma} = (\sigma_{x}, \sigma_{y})\) are the Pauli matrices, \(\vec{A} = (\vec{B} \times \vec{r}) / 2\) is the vector potential (21,25,26), \(e\) is the electron charge. Due to the axial symmetry of the electrostatic potential in the GQD, we can describe the eigenstates by the orbital quantum number \(m\) (here, we neglect spin). In the absence of the GQD, the eigen- energies \(E_{Nm}\) have infinite orbital degeneracy \([\psi_{Nm}(\vec{r})\) where \(m \geq - |N| ]\) independent of \(m\) because of translational invariance. The GQD lifts this orbital degeneracy \(m\) and the LLs are split into a series of sublevels, which exhibit similar behavior as that observed around charged impurities (17,26), due to the Coulomb- like electrostatic potential. However, previous experiments (17,26) in the presence of a magnetic field were limited to a charge impurity in the
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 90, 884, 195]]<|/det|>
+subcritical regime. Further, the small cutoff radius of a charge impurity prohibits to explore the evolution from the ACSs to the LLs in experiment. Such difficulties can be naturally overcome in the studied GQDs.  
+
+<|ref|>text<|/ref|><|det|>[[70, 230, 886, 750]]<|/det|>
+The detailed comparison between experiment and theory can be made by numerically solving the problem for two dimensional massless Dirac fermion of graphene monolayer in the presence of Coulomb- like electrostatic potential \(V_{\beta}(\vec{r})\) and a magnetic field \(B\) (see supplementary information for the details). The calculated radially LDOS maps in the different magnetic fields display that the orbital degeneracy is lifted, which is well consistent with our experimental results (Fig. 3). Based on the calculated results, we can identify the orbital states of the split- 1 LL (Marked in Fig. 3). Thanks to the high- quality LLs in the GQD, \(m = - 1\) orbital state of - 1 LL can be clearly identified and exhibits some characteristics distinguished from that observed in the subcritical regime (17,26). The most important feature is that the \(m = - 1\) orbital state can be viewed as the evolution of the ACS with increasing magnetic field. At zero magnetic field, the broad ACS is located at the center of the GQD and, interestingly, the narrower \(m = - 1\) orbital state appears in the same energy region in the presence of high magnetic field. Such a result indicates directly connection of the ACS and the lowest orbital state ( \(m = - 1\) ) of the - 1 LL.  
+
+<|ref|>text<|/ref|><|det|>[[113, 781, 886, 886]]<|/det|>
+To better explore the evolution of the ACS in the presence of magnetic fields, we summarize the measured LLs at the center of the \(\beta = 2.4\) GQD as a function of the square root of the magnetic field \(\sqrt{B}\) (red dots in Fig. 4, see Fig. S12 for the corresponding STS spectra). The evolution of
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 90, 886, 202]]<|/det|>
+LLs displayed a nonlinear dependence on the square root of the magnetic field, which is quite different from the feature of pristine graphene monolayer under magnetic field. The theoretical map of LDOS at the center of the \(\beta = 2.4\) GQD is also plotted as a function of \(\sqrt{B}\) (see supplementary information for the details), as shown in Fig. 4. With increasing magnetic field, the  
+
+<|ref|>text<|/ref|><|det|>[[75, 255, 886, 686]]<|/det|>
+5 perturbed LLs \((N = 0, N = - 1, N = - 2)\) display nonlinear dependence on \(\sqrt{B}\) . At a higher \(\sqrt{B}\) , the - 1 LL and - 2 LL are well distinctive, which split into low- energy orbital states \((m = - 1, m = 0)\) . However, we did not observe the splitting of the - 2 LL in the experiment, which is probably due to the large full width at half maximum (FWHM) of the LL peaks, prohibiting the observation of the splitting in the experiment. Furthermore, the ACS- R1 resonance is obvious at lower \(\sqrt{B}\) , and is well connected to the \(m = - 1\) orbital state of - 1 LL. Similarly, ACS- R2 resonance has the similar characteristic, connected to the \(m = - 1\) orbital state of - 2 LL. However, such a feature is harder to be recognized in the experiment due to the broadening peak of the - 2 LL. Our experiments, complemented by theoretical calculations, explicitly demonstrated the existence of ACSs in the presence of high magnetic fields and revealed the close connection between the ACS and the lowest orbital state \((m = - 1)\) of the LLs.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 725, 209, 743]]<|/det|>
+## References  
+
+<|ref|>text<|/ref|><|det|>[[112, 764, 886, 910]]<|/det|>
+1. Castro Neto, A. H., Peres, N. M. R., Novoselov, K. S. & Geim, A. K. The electronic properties of graphene. Rev. Mod. Phys. 81, 109-162 (2009).  
+2. Vozmediano, M. A. H., Katsnelson, M. I. & Guinea, F. Gauge fields in graphene. Physics Rep. 496, 109-148 (2010).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 90, 884, 155]]<|/det|>
+3. Das Sarma, S., Adam, S., Hwang, E. & Rossi, E. Electronic transport in two-dimensional graphene. Rev. Mod. Phys. 83, 407-470 (2011).  
+
+<|ref|>text<|/ref|><|det|>[[111, 174, 883, 237]]<|/det|>
+4. Goerbig, M. O. Electronic properties of graphene in a strong magnetic field. Rev. Mod. Phys. 83, 1193-1243 (2011).  
+
+<|ref|>text<|/ref|><|det|>[[75, 256, 884, 320]]<|/det|>
+5. Katsnelson, M. I., Novoselov, K. S. & Geim, A. K. Chiral tunnelling and the Klein paradox in graphene. Nat. Phys. 2, 620-625 (2006).  
+
+<|ref|>text<|/ref|><|det|>[[111, 338, 883, 402]]<|/det|>
+6. Pereira, V. M., Nilsson, J. & Castro Neto, A. H. Coulomb impurity problem in graphene. Phys. Rev. Lett. 99, 166802 (2007).  
+
+<|ref|>text<|/ref|><|det|>[[111, 420, 884, 484]]<|/det|>
+7. Shytov, A. V., Katsnelson, M. I. & Levitov, L. S. Vacuum Polarization and Screening of Supercritical Impurities in Graphene. Phys. Rev. Lett. 99, 236801 (2007).  
+
+<|ref|>text<|/ref|><|det|>[[111, 503, 884, 567]]<|/det|>
+8. Shytov, A. V., Katsnelson, M. I. & Levitov, L. S. Atomic collapse and quasi-rydberg states in graphene. Phys. Rev. Lett. 99, 246802 (2007).  
+
+<|ref|>text<|/ref|><|det|>[[111, 586, 884, 650]]<|/det|>
+9. Zhao, Y. et al. Creating and probing electron whispering-gallery modes in graphene. Science 348, 672-675 (2015).  
+
+<|ref|>text<|/ref|><|det|>[[66, 668, 884, 732]]<|/det|>
+15. Gutiérrez, C., Brown, L., Kim, C.-J., Park, J. & Pasupathy, A. N. Klein tunnelling and electron trapping in nanometre-scale graphene quantum dots. Nat. Phys. 12, 1069-1075 (2016).  
+
+<|ref|>text<|/ref|><|det|>[[111, 750, 883, 813]]<|/det|>
+11. Lee, J. et al. Imaging electrostatically confined Dirac fermions in graphene quantum dots. Nat. Phys. 12, 1032-1036 (2016).  
+
+<|ref|>text<|/ref|><|det|>[[111, 832, 884, 895]]<|/det|>
+12. Bai, K.-K. et al. Generating nanoscale and atomically-sharp p-n junctions in graphene via monolayer-vacancy-island engineering of Cu surface. Phys. Rev. B 97, 045413 (2018).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 90, 885, 155]]<|/det|>
+13. Fu, Z. Q., Bai, K. K., Ren, Y. N., Zhou, J. J. & He, L. Coulomb interaction in quasibound states of graphene quantum dots. Phys. Rev. B 101, 235310 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[115, 174, 884, 237]]<|/det|>
+14. Fu, Z.-Q. et al. Relativistic Artificial Molecules Realized by Two Coupled Graphene Quantum Dots. Nano Lett. 20, 6738 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[75, 256, 884, 319]]<|/det|>
+15. Ghahari, F. et al. An on/off Berry phase switch in circular graphene resonators. Science 356, 845 (2017).  
+
+<|ref|>text<|/ref|><|det|>[[115, 339, 884, 401]]<|/det|>
+16. Wang, Y. et al. Observing atomic collapse resonances in artificial nuclei on graphene. Science 340, 734-737 (2013).  
+
+<|ref|>text<|/ref|><|det|>[[115, 421, 884, 484]]<|/det|>
+17. Mao, J. et al. Realization of a tunable artificial atom at a supercritically charged vacancy in graphene. Nat. Phys. 12, 545-549 (2016).  
+
+<|ref|>text<|/ref|><|det|>[[115, 503, 884, 566]]<|/det|>
+18. Jiang, Y. et al. Tuning a circular p-n junction in graphene from quantum confinement to optical guiding. Nat. Nanotechnol. 12, 1045-1049 (2017).  
+
+<|ref|>text<|/ref|><|det|>[[115, 586, 884, 648]]<|/det|>
+19. Reinhardt, J. & Greiner, W. Quantum electrodynamics of strong fields. Rep. Prog. Phys. 40, 219-295 (1977).  
+
+<|ref|>text<|/ref|><|det|>[[66, 667, 884, 730]]<|/det|>
+20. Gamayun, O. V., Gorbar, E. V. & Gusynin, V. P. Magnetic field driven instability of a charged center in graphene. Phys. Rev. B 83, 235104 (2011).  
+
+<|ref|>text<|/ref|><|det|>[[113, 749, 884, 812]]<|/det|>
+21. Sobol, O. O., Pyatkovskiy, P. K., Gorbar, E. V. & Gusynin, V. P. Screening of a charged impurity in graphene in a magnetic field. Phys. Rev. B 94, 115409 (2016).  
+
+<|ref|>text<|/ref|><|det|>[[113, 832, 884, 894]]<|/det|>
+22. Zhang, Y., Barlas, Y. & Yang, K. Coulomb impurity under magnetic field in graphene: A semiclassical approach. Phys. Rev. B 85, 165423 (2012).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 90, 884, 154]]<|/det|>
+23. Maier, T. & Siedentop, H. Stability of impurities with Coulomb potential in graphene with homogeneous magnetic field. J. Math. Phys. 53, 095207 (2012).  
+
+<|ref|>text<|/ref|><|det|>[[112, 174, 882, 236]]<|/det|>
+24. Kim, S. C. & Eric Yang, S. R. Coulomb impurity problem of graphene in magnetic fields. Ann. Phys. 347, 21-31 (2014).  
+
+<|ref|>text<|/ref|><|det|>[[75, 256, 884, 319]]<|/det|>
+25. Moldovan, D., Masir, M. R. & Peeters, F. M. Magnetic field dependence of the atomic collapse state in graphene. 2D Mater. 5, 015017 (2018).  
+
+<|ref|>text<|/ref|><|det|>[[112, 339, 884, 401]]<|/det|>
+26. Luican-Mayer, A. et al. Screening charged impurities and lifting the orbital degeneracy in graphene by populating landau levels. Phys. Rev. Lett. 112, 036804 (2014).  
+
+<|ref|>text<|/ref|><|det|>[[112, 421, 884, 483]]<|/det|>
+27. Zhang, Y. et al. Electronic Structure, Surface Doping, and Optical Response in Epitaxial WSe₂ Thin Films. Nano Lett. 16, 2485-2491 (2016).  
+
+<|ref|>text<|/ref|><|det|>[[112, 503, 884, 565]]<|/det|>
+28. Addou, R. & Wallace, R. M. Surface Analysis of WSe₂ Crystals: Spatial and Electronic Variability. ACS Appl. Mater. Interfaces 8, 26400-26406 (2016).  
+
+<|ref|>text<|/ref|><|det|>[[112, 586, 819, 607]]<|/det|>
+29. Kahn, A. Fermi level, work function and vacuum level. Mater. Horiz. 3, 7-10 (2016).  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 644, 277, 662]]<|/det|>
+## Acknowledgments:  
+
+<|ref|>text<|/ref|><|det|>[[70, 697, 875, 885]]<|/det|>
+15 This work was supported by the National Natural Science Foundation of China (Grant Nos. 11974050, 11674029, 11921005) and National Key R and D Program of China (Grant No. 2017YFA0303301). L.H. also acknowledges support from the National Program for Support of Top-notch Young Professionals, support from “the Fundamental Research Funds for the Central Universities”, and support from “Chang Jiang Scholars Program”.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 119, 297, 137]]<|/det|>
+## Author contributions  
+
+<|ref|>text<|/ref|><|det|>[[75, 148, 886, 293]]<|/det|>
+Q.Z. performed the sample synthesis, characterization and STM/STS measurements. Q.Z., Y.C.Z., and L.H. analyzed the data. Y.C.Z. carried out the theoretical calculations. L.H. conceived and provided advice on the experiment and analysis. Q.F.S. conceived and provided advice on the theoretical calculations. Q.Z. and L.H. wrote the paper with the input from others. All authors participated in the data discussion.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 380, 192, 398]]<|/det|>
+## Methods  
+
+<|ref|>text<|/ref|><|det|>[[67, 420, 886, 730]]<|/det|>
+CVD Growth of Graphene. The large area graphene monolayer films were grown on a \(20 \times 20 \mathrm{mm}^2\) polycrystalline copper (Cu) foil (Alfa Aesar, \(99.8\%\) purity, \(25 \mu \mathrm{m}\) thick) via a low pressure chemical vapor deposition (LPCVD) method. The cleaned Cu foil was loaded into one quartz boat in center of the tube furnace. Ar flow of 50 sccm (Standard Cubic Centimeter per Minutes) and \(\mathrm{H}_2\) flow of 50 sccm were maintained throughout the whole growth process. The Cu foil was heated from room temperature to \(1030^{\circ}\mathrm{C}\) in 30 min and annealed at \(1030^{\circ}\mathrm{C}\) for six hours. Then \(\mathrm{CH}_4\) flow of 5 sccm was introduced for 20 min to grow high- quality large area graphene monolayer. Finally, the furnace was cooled down naturally to room temperature.  
+
+<|ref|>text<|/ref|><|det|>[[112, 750, 885, 893]]<|/det|>
+Construction of graphene/WS \(\mathbf{e}_2\) heterostructure. We used conventional wet etching technique with polymethyl methacrylate (PMMA) to transfer graphene monolayer onto the substrate. PMMA was first uniformly coated on Cu foil with graphene monolayer. We transferred the Cu/graphene/PMMA film into ammonium persulfate solution, and then the underlying Cu foil was
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 90, 886, 315]]<|/det|>
+etched away. The graphene/PMMA film was cleaned in deionized water for hours. The WSe \(_2\) crystal was separated into thick- layer WSe \(_2\) sheets by traditional mechanical exfoliation technology and then transferred to \(8 \times 8 \mathrm{mm}^2\) highly N- doped Si wafer [(100) oriented, 500 \(\mu \mathrm{m}\) thick]. We placed graphene/PMMA onto Si wafer which has been transferred with WSe \(_2\) sheets in advance. Finally, the PMMA was removed by acetone and then annealed in low pressure with Ar flow of 50 sccm and \(\mathrm{H}_2\) flow of 50 sccm at \(\sim 300^{\circ}\mathrm{C}\) for 1 hours.  
+
+<|ref|>text<|/ref|><|det|>[[111, 338, 886, 686]]<|/det|>
+AFM, STM and STS Measurements. The topographical images are measured by atomic force microscope (AFM, Bruker Multimode 8) with a tapping mode. We employed the n- doped Si tip coated with Platinum- Iridium (Bruker, SCM- PIT- V2, frequency 50- 100KHz, spring constant 1.5- 6 N/m) to characterize WSe \(_2\) and graphene/WSe \(_2\) heterostructure samples. STM/STS measurements were performed in low- temperature (77 K for Fig. S5a and c, 4.2 K for Fig. S5b) and ultrahigh- vacuum ( \(\sim 10^{- 10}\) Torr) scanning probe microscopes [USM- 1400 (77 K) and USM- 1300 (4.2 K)] from UNISOKU. The tips were obtained by chemical etching from a Pt/Ir (80:20%) alloy wire. The differential conductance (dI/dV) measurements were taken by a standard lock- in technique with an ac bias modulation of 5 mV and 793 Hz signal added to the tunneling bias.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[130, 92, 866, 626]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 647, 886, 911]]<|/det|>
+<center>Fig. 1 | The structures and general \(\mathrm{d}I / \mathrm{d}V\) features of Graphene/WSe₂ heterostructure. a, A representative AFM image of the freshly mechanical exfoliated WSe₂ sheet. Inset: the AFM image of a typical monolayer WSe₂ island. b, A STM image of a typical graphene/WSe₂ heterostructure 5 QD. The height of the GQD is \(\sim 0.8 \mathrm{nm}\) and the width of edge area of the GQD is \(\sim 2.2 \mathrm{nm}\) . c, The zoom-in image of the area in black dashed squares from panel b. Inset: the FFT of graphene/WSe₂ heterostructure. The bright spots in the white dotted circles represent the reciprocal lattice of graphene, the bright spots in the blue dotted circles represent the reciprocal lattice of WSe₂, and </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 90, 886, 262]]<|/det|>
+the bright spots in the green dotted circles represent moiré structure of the graphene/ \(\mathrm{WSe_2}\) heterostructure. d, Schematic structure of the graphene/ \(\mathrm{WSe_2}\) heterostructure QD. e, The \(\mathrm{d}I / \mathrm{d}V\) spectra taken inside [marked by dark green pentagram in b] and outside [marked by blue pentagram in b] the GQD. f, The \(\mathrm{d}I / \mathrm{d}V\) maps with different energies [N1 and N2 marked in e] of the GQD.  
+
+<|ref|>text<|/ref|><|det|>[[75, 259, 194, 277]]<|/det|>
+5 the GQD.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[130, 93, 866, 465]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 482, 884, 667]]<|/det|>
+<center>Fig. 2 | Coexistence of WGMs confinement and ACSs in the GQDs. Top of a to c, The radially dI/dV spectroscopic maps of different GQDs. Bottom of a to c, The calculated space-energy maps of the LDOS of different GQDs with different value of \(\beta\) and \(r_0\) . The red dotted lines indicate Dirac point energy. The black solid dots indicate the quasibound states via the WGM confinement, and the purple hollow dots indicate the ACSs. </center>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[130, 88, 868, 470]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 483, 884, 667]]<|/det|>
+<center>Fig. 3 | Lifting the orbital degeneracy of LLs in the GQD. Top panels of a to c, the radially dI/dV spectroscopic maps on the GQD ( \(\beta = 2.4\) , \(r_0 = 9 \text{nm}\) ) in the case of a series of magnetic fields. Bottom panels of a to c, the calculated space-energy maps of the LDOS of the GQD with different magnetic fields. The \(m = -1\) , \(m = 0\) , and \(m = 1\) indicate the split orbital states of the -1 LL. The red dotted lines indicate Dirac point energy. </center>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[155, 90, 700, 440]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[112, 461, 886, 690]]<|/det|>
+<center>Fig. 4 | Evolution from the ACSs to unusual LLs in the GQD. The measured LLs at the center of the GQD ( \(\beta = 2.4\) , \(r_0 = 9 \mathrm{nm}\) ) as a function of the square root of the magnetic field \(\sqrt{B}\) . The experimental results are superimposed onto the calculated map of LDOS in the GQD with \(\beta = 2.4\) and \(r_0 = 9 \mathrm{nm}\) . The ACS-R1 and ACS-R2 are two quasi-bound states due to atomic collapse resonance. The full width at half maximum of the peaks in the spectra was used to estimate the error bar in experiment (orange dots). </center>
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[44, 42, 312, 70]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[44, 93, 765, 113]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[61, 131, 366, 150]]<|/det|>
+SupplementaryInformation.docx
+
+<--- Page Split --->