Endocrinology Endocrine Society

Gene Signature of Proliferating Human Pancreatic ααα-Cells

Giselle Dominguez Gutierrez, Yurong Xin, Haruka Okamoto, Jinrang Kim, Ann-Hwee Lee, Min Ni, Christina Adler, George D. Yancopoulos, Andrew J. Murphy, and Jesper Gromada

Submitted: May 15, 2018 Accepted: July 04, 2018 First Online: July 11, 2018

Advance Articles are PDF versions of manuscripts that have been peer reviewed and accepted but

Endocrinology not yet copyedited. The manuscripts are published online as soon as possible after acceptance and before the copyedited, typeset articles are published. They are posted "as is" (i.e., as submitted by the authors at the modification stage), and do not reflect editorial changes. No corrections/changes to the PDF manuscripts are accepted. Accordingly, there likely will be differences between the Advance Article manuscripts and the final, typeset articles. The manuscripts remain listed on the Advance Article page until the final, typeset articles are posted. At that point, the manuscripts are removed from the Advance Article page.

DISCLAIMER: These manuscripts are provided "as is" without warranty of any kind, either express or particular purpose, or non-infringement. Changes will be made to these manuscripts before publication. Review and/or use or reliance on these materials is at the discretion and risk of the reader/user. In no event shall the Endocrine Society be liable for damages of any kind arising references to, products or publications do not imply endorsement of that product or publication. ADVANCE ARTICLE:

Downloaded from https://academic.oup.com/endo/advance-article-abstract/doi/10.1210/en.2018-00469/5051604 by [email protected] user on 17 July 2018 Endocrinology; Copyright 2018 DOI: 10.1210/en.2018-00469

Human α-cell proliferation gene signature

Gene Signature of Proliferating Human Pancreatic ααα-Cells

Giselle Dominguez Gutierrez*1, Yurong Xin*1, Haruka Okamoto1, Jinrang Kim1, Ann-Hwee Lee1, Min Ni1, Christina Adler1, George D. Yancopoulos1, Andrew J. Murphy1, and Jesper Gromada1 1Regeneron Pharmaceuticals, Inc., 777 Old Saw Mill River Road, Tarrytown, New York, 10591, USA Received 15 May 2018. Accepted 04 July 2018. *GDG and YX are co-first authors. Pancreatic α-cells proliferate with low rate and little is known about the control of this process. Here we report the characterization of human α-cells by large-scale single cell RNA sequencing coupled with pseudotime ordering. We identified two large subpopulations and a smaller cluster of proliferating α-cells with increased expression of genes involved in cell cycle regulation. The proliferating α-cells were differentiated, had normal levels of GCG expression and showed no signs of cellular stress. Proliferating α-cells were detected in both the G1S and G2M phases of the cell cycle. Human α-cells proliferate with 5-fold higher rate than human β-cells and express lower levels of the cell cycle inhibitors CDKN1A and CDKN1C. Collectively, this study provides the gene signatures of human α-cells and the genes involved in their cell division. The lower expression of two cell cycle inhibitors in human α-cells could account for their higher rate of proliferation compared to their insulin producing counterparts. RNA sequencing of single human α-cells and pseudotime ordering of their transcriptomes revealed proliferating cells in different stages of the cell cycle and the genes regulating this process. Endocrinology 1. Introduction The initiation and progression of the cell cycle requires a tightly regulated transcriptional program and results in cell division. Cyclins and cyclin dependent kinases (CDKs) form heterodimer complexes and are key regulators of this transcriptional network. Their activation and subsequent inactivation help drive the different phases of the cell cycle 1. Transcription factors also play a key role in the regulation of the cell cycle machinery 1. It is well established that the early postnatal stages in mice (days or weeks) and humans (few years) are associated with expansion of the endocrine islet mass 2,3. After this initial expansion, the proliferation rate decreases to very low levels 4. While it has been a focus to understand the mechanisms governing the α-cell expansion during develoment, insights into the factors and pathways orchestrating the proliferation of adult α-cells has only started to emerge. A recent study using transcriptomics analysis of islet cells from fetal and adult mice was able to capture cells undergoing proliferation 5. Not surprisingly, the majority of the proliferating cells pertained to the embryonicADVANCE and juvenile stages and they observed ARTICLEa higher ratio of proliferating α-cells compared to β-cells. While their findings shed light on the genes regulating division of α- and β- ADVANCE ARTICLE: cells in mice, comparable knowledge in humans is limited to detection of few proliferating α- cells with distinct gene signatures 6,7. In the present study, we obtained the transcriptomes of >6,000 α-cells by RNA sequencing of single islet cells isolated from 12 non-diabetic donors. The cells divided into three

subpopulations, two large and homogeneous subpopulations and a distinct cluster of proliferating α-cells. We used pseudotime analysis to investigate the progression of gene expression changes in the proliferating α-cells. This was obtained by projecting each cell onto a trajectory. The ordered sequence of the cells provides a higher resolution view of the genes important for the control of the various stages of the cell cycle. We also interrogated how these genes provide α- cells with greater proliferation capacity than their insulin producing counterparts obtained during the same sequencing effort.

2. Material and Methods Human islet processing Human cadaveric islets were procured from Prodo Labs. Information regarding the 12 non- diabetic donors analyzed in this study, along with their diabetes status and when available their hemoglobin A1c is provided in Supplemental Table 1. We received the islets 4-6 days after the isolation. Islets were plated in complete Prodo Islet Media (PIM-S) supplemented with glutamine/glutathione (PIM-G) and human AB serum (PIM-ABS) and incubated overnight in a tissue culture incubator at 37 ºC with a 5% CO2 in air atmosphere before dissociating into single cells. Handpicked islets were enzymatically digested at 37 ºC for 10 min using TrypLE Express (Life Technologies). Subsequently, the cells were filtered (30-µm strainer) and centrifuged. Afterwards, cells were re-suspended in 1X PBS containing 0.04% BSA. This process was immediately followed by loading and sequencing of the cells. Viability of the cells was measured using Trypan blue staining (91.2±3.3% cell viability; n=12). RNA fluorescence in situ hybridization of whole islets and dissociated islet cells Cytospin was used to place dissociated islet cells on microcope slides. Whole islets and dissociated cells were fixed in 10% neutral formalin for 35 min. Islets were embedded in paraffin

Endocrinology and cut in 6 µm sections. Disociated cells underwent a process of permeabilization followed by hybridization with mRNA probes for GCG and CDK1, MKI67, RRM2 and TOP2A. The hybridization process was performed as per the manufacture’s instructions (Advanced Cell Diagnostics). Fluorescein and Cy3 fluorescent signals were amplified using a fluorescent kit. Images were captured using a microscope slide scanner (Zeiss Axio Scan.Z1). To quantify the fluorescence intensity signal, the RNA FISH analysis module from HALO image software ( Indica Labs) was utilized. Single cell RNA sequencing and read mapping Single cells suspended in PBS with 0.04% BSA were loaded on a Chromium Single Cell Instrument (10X Genomics). RNAseq libraries were prepared using Chromium Single Cell 3’ Library, Gel Beads & Mutiplex Kit (10X Genomics). Paried-end sequencing was performed on Illumina NextSeq500 with Read 1 for 59-bp transcript read, and Read 2 for 14-bp I7 index for cell barcode, 8-bp I5 index for sample index, and 10-bp unique molecular identifier (UMI). The Cell Ranger Single-Cell Software Suite (10X Genomics, v1.1.0) was used to perform sample de- multiplexing,ADVANCE alignment, filtering, and UMI counting. TheARTICLE Human B37.3 Genome assembly and UCSC gene model were used for the alignment.

ADVANCE ARTICLE: Single cell data analysis As part of the quality control process, cells were removed if the number of detected genes was <500, total number of UMI was <3000, or viability score >0.2 8. Viability score was defined by the ratio of the sum of UMIs for MT-RNR2, MT-ND1, MT-CO1, MT-CO2, MT-ATP8, MT- ATP6, MT-CO3, MT-CYB expression to total UMI. Low viability was indicated by a higher

score. In order to identify islet endocrine cell types, densityMclust (mclust package) was used to define two expression groups (high and low) of GCG, INS, SST and PPY. Cells with more than one hormone in the high-expression group were excluded. Single-hormone cells were identified as α (GCG-high only), β (INS-high only), δ (SST-high only), PP (PPY-high only), and ε (GHRL- high only) cells. Clustering of retained endocrine and non-endocrine cells was identified by Seurat using 1,166 variable genes. Enriched genes among the α-cell subpopulations were found by FindMarkers in Seurat with p-value <0.05 and log-scale fold change >0.25. Enriched endocrine genes were obtained by comparing endocrine cells (α-, β-, δ-, PP-, and ε-cells) with the non- endocrine cells. Pseudotime trajectory reconstruction Monocle was used to reconstruct pseudotime ordering of α-cells with the default setting. The input genes for pseudotime ordering were α-cell subpopulation enriched genes. Cells from all twelve donors exhibited relatively uniform distribution along the trajectory, and were used for the pseudotime analysis (Supplemental Fig. 1B). The pseudotime trajectory had two states: Proliferating state and Other state. Branch-dependent significant genes were identified by BEAM function in monocle. Significant genes were defined with q-value <0.01. Differentially expressed genes along the proliferating branch were identified by removing cells in the Other state, and significant genes were those with q-value <1e-10. Pathway enrichment Enriched α-cell genes and branch-dependent genes were analyzed for pathway enrichment with GO, KEGG, and REACTOME by clusterProfiler (R package). Top enriched pathways were selected based on p-value. Biological process score calculation Endocrinology A composite score for each cell was calculated by the mean of gene expression (scaled UMI) in a predefined pathway or biological process. Three biological processes were used in the study: unfolded protein response (UPR), apoptosis and cell cycle. The gene sets were obtained from IPA Ingenuity, KEGG and 9 (Supplemental Table 2). In order to estimate empirical p-value of a composite score, the score distribution was generated by composite scores of randomly selected genes for each cell with 1000 iterations. Proliferating α- and β-cells were identified using cell cycle scores with empirical p-value <0.001. Accession number GSE114297

3. Results Human ααα-cells contain a subpopulation of highly proliferative cells Unbiased clustering of human islet cells revealed three subpopulations of α-cells; two large and homogenousADVANCE subpopulations with 26 and 6 uniquely en richedARTICLE genes and a smaller and more distinct subpopulation with 974 uniquely enriched genes (Fig. 1A and Supplemental Table 3).

ADVANCE ARTICLE: Pathway analysis of the third subpopulation revealed enrichment of genes associated with cell cycle and division (Fig. 1B). To explore the differentiation state of these subpopulations, and more specifically if the proliferative subgroup possesses progenitor features, we calculated a score for each cell that reflects the expression of endocrine marker genes. Fig. 1C shows no difference in the endocrine score between the three α-cell subpopulations. Furthermore, the three

α-cell subpopulations clustered together, but separated from non-endocrine cells in hierarchical clustering based on the expression of the highly abundant endocrine genes, demonstrating that both subpopulations represent fully differentiated α-cells (Fig. 1D). Thus, human α-cells comprise three subpopulations of fully differentiated cells with a smaller and distinctive proliferative subgroup among them. RNA-FISH confirms the existence of the ααα-cell proliferative subpopulation To validate the existence of proliferating α-cells, we performed RNA-FISH simultaneous to GCG and CDK1, MKI67, RRM2 and TOP2A using freshly isolated human islets. These genes are significantly enriched in the proliferating α-cells and important for cell cycle regulation 1,9,10. We identified α-cells co-expressing these proliferation markers, which were often undergoing mitosis (Fig. 2A). Quantification of the marker genes showed consistency in the fraction of cells with increased expression between the RNA sequencing and RNA-FISH analyses (Fig. 2B). Pseudotime analysis reveals proliferation trajectory To characterize the functional and temporal relationship between the α-cell subpopulations, we performed pseudotime analysis using their enriched marker genes (Supplemental Fig. 1A). The majority of the cells had similar transcriptome profiles and were found in the root (Root state). We detected one branch decision point leading to two states of relatively minor cell numbers (Fig. 3A). One of the states contained cells with high enrichment of genes associated with cell proliferation (Proliferating state). The “Other” state (Other state) was associated with 52 differentially expressed genes that did not translate into significant enrichment of any biological pathways (Fig. 3B and Supplemental Table 4). Interestingly, the Proliferating state was longer in comparison to the Other state in pseudotime. In fact, while the Root state and Other state measured up to 12 in pseudotime, the Proliferating state occupied from 12 to 23 in pseudotime (Supplemental Fig. 1A). We hypothesized that the long pseudotime of the Proliferating state Endocrinology might reflect different phases in the cell cycle. Using a cell cycle score, we found this to be the case, as there was a progressive increase in cell cycle score along the Proliferating state (Fig. 3C). Consistent with this, α-cells in the beginning of the Proliferating state showed higher G1S proliferation score, while cells at the middle-end of the Proliferating state showed higher G2M score (Fig. 3D and Supplemental Fig. 2). Proliferating state was composed of 123 cells. The original 33 proliferating α-cells in the “α-cell sub3 group” were located in the middle-end of the Proliferating state, indicating that “α-cell sub3 group” was primarily composed of cells in the G2M phase and with a strong proliferative gene signature (Supplemental Fig. 1C). Overall, pseudotime analysis allowed us to observe a trajectory of proliferation in human α-cells in the different phases of cell division. Cell cycle regulators of human ααα-cell proliferation To investigate the genes important in regulating α-cell proliferation, we focused on the pseudotime trajectory containing the Root state and Proliferating state branches. We found 942 genes to be differentially expressed along this trajectory with a q-value <1e-10 (Supplemental Table 5).ADVANCE As expected, pathway analysis revealed enrichment ARTICLE of cell cycle genes (Supplemental Table 6). From this category (KEGG hsa04110), we explored the expression of the 22 ADVANCE ARTICLE: significantly regulated cell cycle genes and plotted their expression in all cells along the pseudotime (Fig. 4 and Supplemental Fig. 3). Interestingly, we found 3 classes of cyclins to be significantly expressed along the pseudotime: CCNE2, a cyclin important during the G1/S phase; CCNA2 which plays a key role during the S phase and CCNB1-CCNB2 regulating events during the G2M transition and progression through mitosis 1. Of interest, these cyclin genes exhibited

different temporal expression profile along the pseudotime consistent with their function at different phases of the cell cycle (Fig. 4). As mentioned above, cyclins heterodimerize with CDKs, which serve as the catalytic subunits 1. Consistent with this, we found CDK1 to be enriched at the Proliferating state with a progressive increase in expression towards the middle- end of the trajectory (Fig. 4). Interestingly, previous research has shown that CDK1 can by itself complete progression of the mammalian cell cycle 11. Among other cell cycle regulators, we found differential expression of cell cycle inhibitors. Among the the Cip/Kip (CDK interacting protein/Kinase inhibitory protein) family, CDKN1A showed a modest increase at the beginning of the Proliferating state and was then mostly depleted for the remainder of the state (Supplemental Fig. 3). Members of this family of inhibitors have broad effects on cell cycle progression 1. Consistent with this, their low expression during most of the Proliferating state might be essential for the cell cycle to progress successfully. We also found members of the INK4 (inhibitors of CDK4) family to be differentially expressed: CDKN2A, CDKN2C and CDKN2D presented with different patterns of expression as well (Supplemental Fig. 3). Overall, we uncovered genes important for α-cell division. Their enrichment and expression patterns within the Proliferating state trajectory provide valuable information for the understanding of human α-cell proliferation. Transcriptional regulators of human ααα-cell proliferation To understand the regulation of cell cycle initiation and progression in human α-cells, we next focused on the 42 transcription factors and regulators that were differentially expressed along the pseudotime trajectory (Fig. 5 and Supplemental Fig. 4) (Supplemental Table 5). While some of them have not been well characterized, others have known functions in the cell cycle process. Fig. 5 shows TFDP1 to be enriched in the beginning of the Proliferating state. TFDP1 dimerizes with the E2F1 transcription factor and regulates expression of genes important during the G1/S 12

Endocrinology phase transition . Interestingly, we found that E2F1 was significantly enriched in the Proliferating state (q-value 9e-9). Since we use a stringent cutoff it was not included in the present list. We also found FOXM1 to be enriched in proliferating α-cells. Overexpression of FOXM1 in human islets induces the expression of cell cycle regulators (A-type cyclins, B-type cyclins and E-type cyclins), chromosome complex passenger genes (BIRC5, AURKB, PLK1 and CDCA8) and anaphase-promoting complex (APC) system factor (CDC20) among others 13. The majority of these FOXM1 target genes are significantly enriched in the α-cell Proliferating state (Supplemental Table 5). Although overexpression of FOXM1 in human islets induced proliferation and activation of many cell cycle genes, the expression of D-type cyclins was not affected 13. This is consistent with the lack of expression changes of the D-type cyclins in the current study. We observed HES1 expression to be increased in the Proliferating state (Figure 5). HES1 regulates proliferation of neural stem cells by blocking differentiation and the cell cycle inhibitor p21CIP1/WAF1 14. During pancreas development Hes1, through its regulation of p57Kip2 coordinates the balance between proliferation and cell cycle exit and thereby the number of progenitor cells 15. We alsoADVANCE found that the inhibitor of differentiation proteinsARTICLE ID1 and ID3 were enriched in the beginning of the Proliferating state (Figure 5). They are upregulated in several types of tumors ADVANCE ARTICLE: and form heterodimers with bHLH proteins to inhibit their DNA binding activity 16. Of interest, in neural stem cells these transcription factors release the negative auto loop regulation of Hes1 17. Lastly, the novel transcription factor TCF19 has been found to be important for proliferation and survival in the INS-1 β-cell line 18.

In addition to transcription factors, we investigated whether epigenetic processes could influence α-cell cycle regulation. We curated the database EpiFactors 19 and found 38 of the genes to be differentially expressed along the pseudotime trajectory (Figure 6 and Supplemental Fig. 5) (Supplemental Table 5). Among them, were factors with functions related to modification of histone methylation (EZH2), acetylation (ELP5, MORF4L2, NCOR1, PHF19, SUDS3, DNAJC1), deacetylation (DNAJC1), ubiquitination (UBE2B, UBE2D1, UBE2T) and phosphorylation (AURKA, AURKB, CDK1, PBK, PPP2CA, PPP4C). Interestingly, the PRC2 complex components EZH2, RBBP7 and PHF19 showed significant enriched expression in the Proliferating state trajectory (Figure 6). EZH2 is the catalytic component of the PRC2 complex. It is regulated by the E2F/pRB pathway and functions as a methyl transferase for lysine 9 (H3K9) and lysine 27 (H3K27) 20. In human breast cancer cells, EZH2 is a key contributor to the suppression of CDKN1C through its trimethylation of histone H3 lys 27 (H3K27me3) 21. In human insulinomas, EZH2 possesses a gain of copy number variation and its overexpression in combination with CDKN1C silencing was more efficient at inducing β-cell proliferation than modulation of expression of each factor alone 22. In human β-cells the levels of EZH2 mRNA and protein decrease with age 23. Thus, it is possible that epigenetic regulation by the PRC2 complex contributes to regulation of human α-cell proliferation. Overall, pseudotime analysis revealed 42 transcription factors or regulators and 38 epigenetic regulators with differential expression in human proliferating α-cells. While some of the factors are important for the control of the cell cycle in other cell types, little is known about their function in human α-cell proliferation. Hormone expression and cellular stress do not promote human ααα-cell proliferation We next sought to understand a potential trigger for human α-cell proliferation. In mouse β- cells, it has been reported that cellular stress and low expression of insulin promote cell 24,25 Endocrinology proliferation . We calculated a composite score representing the mRNA abundance of the genes involved in the unfolded protein response (UPR), and found that the proliferative α-cells are lower in UPR score compared with the cells in Root state (Fig. 7A). Apoptosis score was also low throughout the α-cell population, indicating the scaricity of apoptosing α-cells captured in our analysis (Fig. 7B). GCG expression did not correlate with α-cell proliferation (Supplemental Fig. 6). Taken together, these data suggest that the signal triggering α-cell proliferation is distinct from those for β-cells. CDKN1A and CDKN1C expression and human ααα- and βββ-cell proliferation During our sequencing effort, we also identified 6,241 β-cells 26. Consistent with previous findings 4,6, we detected more proliferating α-cells (0.94%; n=6,465) than β-cells (0.16%; n=6,241). These observations correlated well with the lower cell cycle composite scores of proliferating β-cells than those of α-cells (Fig. 8A). Then why α-cells are more likely to proliferate than β-cells? We hypothesized that cell cycle inhibitors might be more highly expressed in β-cells inhibiting cell cycle entry. Indeed, we found that cell cycle inhibitors CDKN1AADVANCE and CDKN1C were significantly higher in β -cellsARTICLE than in α-cells at the non- proliferating state (Fig. 8B). CDKN1C encodes the cyclin-dependent kinase inhibitor P57KIP2, ADVANCE ARTICLE: which inhibits the activity of several cyclin/CDK complexes 1. Importantly, their repression in human islets increased β-cell proliferation 22,27,28. Furthermore, loss of CDKN1C expression is detected in focal lesions of congenital hyperinsulinism, in pediatric and adult insulinomas as well as in Beckwith-Wiedemann Syndrome, which is associated with hypoglycemia 22,29-32. Thus, the

lower expression of these cell cycle inhibitors in α-cells could contribute to their higher proliferation rate.

4. Discussion RNA sequencing of >6,000 single human α-cells and pseudotime reconstruction of their transcriptomes revealed a trajectory composed of proliferating cells. We detected α-cells in the G1/S and G2/M phases of the cell cycle and uncovered temporal changes in expression of genes associated with cell cycle progression, as well as novel transcription factors and epigenetic regulatory components. As part of the analysis, we included proliferating β-cells identified in the same sequencing effort. We found that α-cells proliferate with 5-fold higher rate than β-cells, even when originating from the same donors. This is consistent with long-term labeling studies in mice, as well as recent single human islet cell studies demonstrating very slow turnover of β- cells in the adult stage 4,5,33. It is tempting to speculate that the higher expression of the cell cycle inhibitors CDKN1A and CDKN1C in β-cells could explain the lower rate of cell division. The proliferating β-cells were high in G1S score, but had an overall low expression of cell cycle related genes. Human α-cell mass remains relatively constant over time and is little affected by diabetes development 4,34,35. This would suggest equal rates of α-cell proliferation and death. It was therefore surprising that the apoptosis score was low in α-cells, and that a distinct α-cell subpopulation with high score did not emege in our pseudotime analysis. Additional work is therefore required to identify the mechanism(s) regulating α-cell death. Adult human α-cells remain fully differentiated while undergoing proliferation and no progenitor markers were detected. We were not able to detect the recently described α-related cell population characterized by the expression of the α-cell marker ARX, the progenitor marker Endocrinology SOX9 and having little to no expression of GCG 36. This is not surprising, since the α-related cell population was abundant in children and adolescent and our study was conducted on islet cells from adult donors. Collectively, these data suggest that the main source of new α-cells might change with age from a progenitor-like population in early life to replication of existing α-cells in adulthood. Apart from the highly proliferative subpopulation, the transcriptomes of the α-cells were fairly homogenous and no distinct functional states were detected in the pseudotime trajectory. Along these lines, we found no indication of stress or UPR activation. This could be explained by the observations that glucagon is less likely to misfold than insulin and that α-cells are more resistant to metabolic stress than β-cells 34. The lower propensity to develop cellular and metabolic stress might explain why α-cell mass, but not the β-cell mass remains unchanged in type 2 diabetes 35,37,38. In conclusion, we unveiled key genes important for control of human α-cell proliferation. We show how the expression of these genes change in a coordinated manner during cell division. For someADVANCE of these genes little is known about their function ARTICLE in the cell cycle, offering an opportunity for further exploration. Understanding the gene signatures of α-cells and the ADVANCE ARTICLE: mechanisms of proliferation could be of importance in ongoing research efforts to try to replenish the β-cell pool in type 2 diabetes via reprogramming of α-cells.

5. Acknowledgements The authors thank Samantha Intriligator for her help with preparing the manuscript.

Financial Support. The studies were funded by Regeneron Pharmaceuticals, Inc. Regeneron Pharmaceuticals http://dx.doi.org/10.13039/100009857, Jesper Gromada Author contributions: Y.X., G.D.G., H.O., J.K., and J.G. designed the studies, G.D.G., J.K., C.A., and M.N., conducted the studies, Y.X., G.D.G., H.O., J.K., A-H.L., and J.G. analyzed the data, Y.X., G.D.G., A-H.L., G.D.Y., A.J.M., and J.G. wrote the manuscript. Correspondence: Jesper Gromada, 777 Old Saw Mill River Rd, Tarrytown, NY 10591, Tel +1- 914-847-7971, Email: [email protected] Disclosure. The authors are employees and shareholders of Regeneron Pharmaceuticals, Inc

6. References 1. Yang VW. Chapter 15 - The Cell Cycle A2 - Johnson, Leonard R. In: Ghishan FK, Kaunitz JD, Merchant JL, Said HM, Wood JD, eds. Physiology of the Gastrointestinal Tract (Fifth Edition). Boston: Academic Press; 2012:451-471. 2. Ackermann AM, Gannon M. Molecular regulation of pancreatic beta-cell mass development, maintenance, and expansion. J Mol Endocrinol. 2007;38(1-2):193-206. 3. Meier JJ, Butler AE, Saisho Y, et al. Beta-cell replication is the primary mechanism subserving the postnatal expansion of beta-cell mass in humans. Diabetes. 2008;57(6):1584- 1594. 4. Wang YJ, Golson ML, Schug J, et al. Single-Cell Mass Cytometry Analysis of the Human Endocrine Pancreas. Cell Metab. 2016;24(4):616-626. Endocrinology 5. Qiu WL, Zhang YW, Feng Y, Li LC, Yang L, Xu CR. Deciphering Pancreatic Islet beta Cell and alpha Cell Maturation Pathways and Characteristic Features at the Single-Cell Level. Cell Metab. 2017;25(5):1194-1205 e1194. 6. Segerstolpe A, Palasantza A, Eliasson P, et al. Single-Cell Transcriptome Profiling of Human Pancreatic Islets in Health and Type 2 Diabetes. Cell Metab. 2016;24(4):593-607. 7. Wang YJ, Schug J, Won KJ, et al. Single-Cell Transcriptomics of the Human Endocrine Pancreas. Diabetes. 2016;65(10):3028-3038. 8. Xin Y, Kim J, Okamoto H, et al. RNA Sequencing of Single Human Islet Cells Reveals Type 2 Diabetes Genes. Cell Metab. 2016;24(4):608-615. 9. Dominguez D, Tsai YH, Gomez N, Jha DK, Davis I, Wang Z. A high-resolution transcriptome map of cell cycle reveals novel connections between periodic genes and cancer. Cell Res. 2016;26(8):946-962. 10. Grolmusz VK, Karaszi K, Micsik T, et al. Cell cycle dependent RRM2 may serve as proliferation marker and pharmaceutical target in adrenocortical cancer. Am J Cancer Res. 2016;6(9):2041-2053. 11. SantamariaADVANCE D, Barriere C, Cerqueira A, et al. Cdk1ARTICLE is sufficient to drive the mammalian cell cycle. Nature. 2007;448(7155):811-815. ADVANCE ARTICLE: 12. Martin K, Trouche D, Hagemeier C, Kouzarides T. Regulation of transcription by E2F1/DP1. J Cell Sci Suppl. 1995;19:91-94. 13. Davis DB, Lavine JA, Suhonen JI, et al. FoxM1 is up-regulated by obesity and stimulates beta-cell proliferation. Mol Endocrinol. 2010;24(9):1822-1834.

14. Kabos P, Kabosova A, Neuman T. Blocking HES1 expression initiates GABAergic differentiation and induces the expression of p21(CIP1/WAF1) in human neural stem cells. J Biol Chem. 2002;277(11):8763-8766. 15. Georgia S, Soliz R, Li M, Zhang P, Bhushan A. p57 and Hes1 coordinate cell cycle exit with self-renewal of pancreatic progenitors. Dev Biol. 2006;298(1):22-31. 16. Sikder HA, Devlin MK, Dunlap S, Ryu B, Alani RM. Id proteins in cell growth and tumorigenesis. Cancer Cell. 2003;3(6):525-530. 17. Bai G, Sheng N, Xie Z, et al. Id sustains Hes1 expression to inhibit precocious neurogenesis by releasing negative autoregulation of Hes1. Dev Cell. 2007;13(2):283-297. 18. Krautkramer KA, Linnemann AK, Fontaine DA, et al. Tcf19 is a novel islet factor necessary for proliferation and survival in the INS-1 beta-cell line. Am J Physiol Endocrinol Metab. 2013;305(5):E600-610. 19. Medvedeva YA, Lennartsson A, Ehsani R, et al. EpiFactors: a comprehensive database of human epigenetic factors and complexes. Database (Oxford). 2015;2015:bav067. 20. Bracken AP, Pasini D, Capra M, Prosperini E, Colli E, Helin K. EZH2 is downstream of the pRB-E2F pathway, essential for proliferation and amplified in cancer. EMBO J. 2003;22(20):5323-5335. 21. Yang X, Karuturi RK, Sun F, et al. CDKN1C (p57) is a direct target of EZH2 and suppressed by multiple epigenetic mechanisms in breast cancer cells. PLoS One. 2009;4(4):e5011. 22. Wang H, Bender A, Wang P, et al. Insights into beta cell regeneration for diabetes via integration of molecular landscapes in human insulinomas. Nat Commun. 2017;8(1):767. 23. Chen H, Gu X, Su IH, et al. Polycomb protein Ezh2 regulates pancreatic beta-cell Ink4a/Arf expression and regeneration in diabetes mellitus. Genes Dev. 2009;23(8):975-985. 24. Szabat M, Page MM, Panzhinskiy E, et al. Reduced Insulin Production Relieves

Endocrinology Endoplasmic Reticulum Stress and Induces beta Cell Proliferation. Cell Metab. 2016;23(1):179- 193. 25. Sharma RB, O'Donnell AC, Stamateris RE, et al. Insulin demand regulates beta cell number via the unfolded protein response. J Clin Invest. 2015;125(10):3831-3846. 26. Xin Y, Gutierrez GD, Okamoto H, et al. Pseudotime Ordering of Single Human β-Cells Reveals States of Insulin Production and Unfolded Protein Response. Diabetes 2018;10.2337/db18-0365. 27. Avrahami D, Li C, Yu M, et al. Targeting the cell cycle inhibitor p57Kip2 promotes adult human beta cell replication. J Clin Invest. 2014;124(2):670-674. 28. Robitaille K, Rourke JL, McBane JE, et al. High-throughput Functional Genomics Identifies Regulators of Primary Human Beta Cell Proliferation. J Biol Chem. 2016;291(9):4614- 4625. 29. Kassem SA, Ariel I, Thornton PS, Scheimberg I, Glaser B. Beta-cell proliferation and apoptosis in the developing normal human pancreas and in hyperinsulinism of infancy. Diabetes. 2000;49(8):1325-1333. 30. PercesepeADVANCE A, Bertucci E, Ferrari P, et al. Fami lialARTICLE Beckwith-Wiedemann syndrome due to CDKN1C mutation manifesting with recurring omphalocele. Prenat Diagn. 2008;28(5):447-449. ADVANCE ARTICLE: 31. Bhatti TR, Ganapathy K, Huppmann AR, et al. Histologic and Molecular Profile of Pediatric Insulinomas: Evidence of a Paternal Parent-of-Origin Effect. J Clin Endocrinol Metab. 2016;101(3):914-922.

32. Matsuoka S, Edwards MC, Bai C, et al. p57KIP2, a structurally distinct member of the p21CIP1 Cdk inhibitor family, is a candidate tumor suppressor gene. Genes Dev. 1995;9(6):650- 662. 33. Teta M, Long SY, Wartschow LM, Rankin MM, Kushner JA. Very slow turnover of beta-cells in aged adult mice. Diabetes. 2005;54(9):2557-2567. 34. Marroqui L, Masini M, Merino B, et al. Pancreatic alpha Cells are Resistant to Metabolic Stress-induced Apoptosis in Type 2 Diabetes. EBioMedicine. 2015;2(5):378-385. 35. Henquin JC, Rahier J. Pancreatic alpha cell mass in European subjects with type 2 diabetes. Diabetologia. 2011;54(7):1720-1725. 36. Lam CJ, Cox AR, Jacobson DR, Rankin MM, Kushner JA. Highly Proliferative Alpha- Cell Related Islet Endocrine Cells in Human Pancreata. Diabetes. 2018. 37. Butler AE, Janson J, Bonner-Weir S, Ritzel R, Rizza RA, Butler PC. Beta-cell deficit and increased beta-cell apoptosis in humans with type 2 diabetes. Diabetes. 2003;52(1):102-110. 38. Rahier J, Guiot Y, Goebbels RM, Sempoux C, Henquin JC. Pancreatic beta-cell mass in European subjects with type 2 diabetes. Diabetes Obes Metab. 2008;10 Suppl 4:32-42.

Figure 1. Heterogeneity in human α-cells. (A) Human α-cells segregate into three subpopulations (alpha sub1, alpha sub2 and alpha sub3). Other human islet cells are shown in light blue. (B) Top 5 GO Biological Process terms enriched for alpha sub2 enriched genes (p- value<1-10). (C) Boxplot of endocrine score in the α-cell subpopulations. Non-endocrine cells are presented for reference. Each dot represents an individual cell. (D) Expression heatmap of highly abundant endocrine genes shows no significant differences in expression between the three α- cell subpopulations. Non-endocrine cells are shown for reference.

Figure 2 – RNA-FISH confirms the existence of proliferating α-cells. (A) Representative images

Endocrinology of dissociated islet cells stained for enriched genes in alpha sub3 subpopulation (CDK1, MKI67, RRM2 and TOP2A in green), GCG (red) and DAPI (blue). Dividing cells with high expression of the marker genes are indicated with white arrows. (B) Proportion of gene positive (CDK1+, MKI67+, RRM2+, and TOP2A+) α-cells using single cell RNA sequencing and RNA-FISH. For single cell RNA sequencing, a gene was detected when normalized UMI was >0.4. The detection proportion was computed and averaged across 12 donors.

Figure 3 – Proliferative trajectory is revealed by pseudotime analysis. (A) Pseudotime states are highlighted by different colors. The cell number per state is as follows: “Root” (n=5959), “Proliferating” (n=123) and “Other” (n=383). (B) Heatmap showing branch dependent genes for Proliferating and Other branch (q-value <0.01). Directionality of the branch is shown above the heatmap. (C) Composite cell cycle score value was plotted in pseudotime ordering. Each circle represents an individual cell. The level of composite score is represented by the color and size of the individual circle. (D) Individual cell cycle score value for G1S and G2M was plotted against the pseudotime in the Root state and Proliferating state, to visually show each phase enrichment along theADVANCE trajectory. Cells from the Other state were removed ARTICLE from the analysis. The color of each trajectory represents the specific cell cycle phase. The vertical dotted line indicates the ADVANCE ARTICLE: division between the Root state and the Proliferating state.

Figure 4 – Differential expression of cell cycle factors along the pseudotime. Cells from the Other state were removed from the analysis. The expression (normalized UMI) of selected cell cycle associated genes is plotted against the pseudotime. The state of origin for each cell is

represented through different colors. Each bar represents an individual cell, and the percent of cells expressing the gene against the total number of cells in the state is shown underneath the gene name. Total number of cells in each state is shown the appropriate label. Cell cycle gene list was obtained from KEGG (Kyoto Encyclopedia of genes and genomes) database (hsa04110); and (q-value<1-10) was used to determine differentially expressed genes.

Figure 5 – Selected transcription factors and regulators differentially expressed along the pseudotime. Cells from the Other state were removed from the analysis. Differentially expressed transcription factor and regulator genes were determined using a (q-value<1-10). Gene expression (normalized UMI) was plotted against the pseudotime for selected transcription factors. The state of origin for each cell is represented through different colors. Each bar represents an individual cell, and the percent of cells expressing the gene against the total number of cells in the state is shown underneath the gene name. Total number of cells in each state is shown in the appropriate label.

Figure 6 – Epigenetic factors differentially expressed along the pseudotime. Cells from the Other state were removed from the analysis. Enriched epigenetic factors were obtained from the EpiFactors database 19. The expression (normalized UMI) of selected epigenetic factor genes is plotted against the pseudotime. The state of origin for each cell is represented through different colors. Each bar represents an individual cell, and the percent of cells expressing the gene against the total number of cells in the state is shown underneath the gene name. Total number of cells in each state is shown in the appropriate label. Epigenetic factors differentially expressed were determined using a (q-value<1-10).

Figure 7 – UPR and apoptosis signature expression in proliferative α-cells. A composite score

Endocrinology value was plotted in pseudotime ordering for (A) UPR and (B) Apoptosis. Each circle represents an individual cell. The level of composite score is represented by the color and size of the individual circle.

Figure 8 - Proliferation differs between α- and β-cells. (A) Proliferative α- and β-cells were aligned together against their cell cycle composite score. Proliferating cells were identified by p- value<0.001 for cell cycle composite score. (B) Boxplots of cyclin-dependent kinase inhibitor (CDKN1A and CDKN1C) genes showing their expression (normalized UMI) in non-proliferating α and β-cells.

ADVANCE ARTICLE ADVANCE ARTICLE:

Downloaded from https://academic.oup.com/endo/advance-article-abstract/doi/10.1210/en.2018-00469/5051604 by [email protected] user on 17 July 2018 A D

C Endocrinology Endocrinology

ADVANCE ARTICLE ADVANCE ARTICLE: ADVANCE ARTICLE:

Figure 1

Downloaded from https://academic.oup.com/endo/advance-article-abstract/doi/10.1210/en.2018-00469/5051604 by [email protected] user on 17 July 2018 A B

1.5 -cells) -cells α

α 1.0 (+) (+) 0.5 CDK1 CDK1 (% of total total of (% 0.0 RNA RNA- seq FISH

0.8 -cells -cells) 0.6 α α

(+) (+) 0.4

MKI67 0.2 MKI67 (% of total total of (% 0.0 RNA RNA- seq FISH

0.8 -cells)

-cells 0.6 α α

(+) (+) 0.4

RRM2 0.2 RRM2 (% of total total of (% 0.0 RNA RNA- seq FISH

1.0 0.8 -cells -cells) α α Endocrinology Endocrinology 0.6 (+) (+) 0.4

TOP2A 0.2 TOP2A (% of total total of (% 0.0 10 µm RNA RNA- seq FISH Dapi GCG /Marker GCG /Marker /Dapi

ADVANCE ARTICLE ADVANCE ARTICLE: ADVANCE ARTICLE:

Figure 2

Downloaded from https://academic.oup.com/endo/advance-article-abstract/doi/10.1210/en.2018-00469/5051604 by [email protected] user on 17 July 2018 A B Proliferating Root state Other

C D Endocrinology Endocrinology

ADVANCE ARTICLE ADVANCE ARTICLE: ADVANCE ARTICLE:

Figure 3

Downloaded from https://academic.oup.com/endo/advance-article-abstract/doi/10.1210/en.2018-00469/5051604 by [email protected] user on 17 July 2018 CCNA2 (4e−21) CCNB1 (1e−28) 0.05% 15% 0.89% 18% 10.0 30 7.5 20 5.0 10 2.5 0.0 0

CCNB2 (3e−40) CCNE2 (2e−14) 0.1% 21% 0.05% 12%

9 15 6 10 3 5 0 0 Expression(UMI)

CDC20 (1e−47) CDK1 (1e−82) 0.13% 18% 0.27% 36% 15 15 10 10 5 5 0 0 0 5 10 15 20 0 5 10 15 20 Pseudotime Endocrinology Endocrinology

ADVANCE ARTICLE ADVANCE ARTICLE: ADVANCE ARTICLE:

Figure 4

Downloaded from https://academic.oup.com/endo/advance-article-abstract/doi/10.1210/en.2018-00469/5051604 by [email protected] user on 17 July 2018 FOXM1 (1e−14) HES1 (3e−15) 0.034% 15% 6.9% 24% 5 15 4 3 10 2 5 1 0 0

ID1 (3e−87) ID3 (7e−13) 11% 37% 9.3% 25% 50 20 40 15 30 10 20 10 5 0 0 Expression(UMI)

TCF19 (9e−15) TFDP1 (2e−12) 0.27% 23% 15% 41% 5 4 7.5 3 5.0 2 1 2.5 0 0.0 0 5 10 15 20 0 5 10 15 20 Pseudotime Endocrinology Endocrinology

ADVANCE ARTICLE ADVANCE ARTICLE: ADVANCE ARTICLE:

Figure 5

Downloaded from https://academic.oup.com/endo/advance-article-abstract/doi/10.1210/en.2018-00469/5051604 by [email protected] user on 17 July 2018 EZH2 (5e−13) 0.18% 19%

4 3 2 1 0

PHF19 (2e−11) 1.8% 24% 6 4 2 0 Expression(UMI)

RBBP7 (3e−12) 23% 53%

0 0 5 10 15 20 Pseudotime Endocrinology Endocrinology

ADVANCE ARTICLE ADVANCE ARTICLE: ADVANCE ARTICLE:

Figure 6

Downloaded from https://academic.oup.com/endo/advance-article-abstract/doi/10.1210/en.2018-00469/5051604 by [email protected] user on 17 July 2018 A B Endocrinology Endocrinology

ADVANCE ARTICLE ADVANCE ARTICLE: ADVANCE ARTICLE:

Figure 7

Downloaded from https://academic.oup.com/endo/advance-article-abstract/doi/10.1210/en.2018-00469/5051604 by [email protected] user on 17 July 2018 A B

Cellcycle score 1

Proliferating alpha/beta cells

Alpha Beta Endocrinology Endocrinology

ADVANCE ARTICLE ADVANCE ARTICLE: ADVANCE ARTICLE:

Figure 8

Downloaded from https://academic.oup.com/endo/advance-article-abstract/doi/10.1210/en.2018-00469/5051604 by [email protected] user on 17 July 2018