ABSTRACT
Pancreatic and duodenal homeobox 1 (PDX1) is crucial for pancreas organogenesis, yet the dynamic changes in PDX1 binding in human or mouse developing pancreas have not been examined. To address this knowledge gap, we performed PDX1 ChIP-seq and single-cell RNA-seq using fetal human pancreata. We integrated our datasets with published datasets and revealed the dynamics of PDX1 binding and potential cell lineage-specific PDX1-bound genes in the pancreas from fetal to adult stages. We identified a core set of developmentally conserved PDX1-bound genes that reveal the broad multifaceted role of PDX1 in pancreas development. Despite the well-known dramatic changes in PDX1 function and expression, we found that PDX1-bound genes are largely conserved from embryonic to adult stages. This points towards a dual role of PDX1 in regulating the expression of its targets at different ages, dependent on other functionally congruent or directly interacting partners. We also showed that PDX1 binding is largely conserved in mouse pancreas. Together, our study reveals PDX1 targets in the developing pancreas in vivo and provides an essential resource for future studies on pancreas development.
INTRODUCTION
Environmental and genetic factors contribute to diabetes pathogenesis. Mutations of key transcription factors cause defects in endocrine pancreas development and impose susceptibility for diabetes. One of the most well-studied pancreatic transcription factors is the human diabetes gene pancreatic and duodenal homeobox 1 (PDX1), which plays essential roles in pancreas organogenesis, endocrine pancreas development, and the growth and function of insulin-secreting β-cells in both mouse and human (Pan and Wright, 2011; Jennings et al., 2020).
PDX1 is a hallmark transcription factor of pancreas development. One of the early signs of pancreas organogenesis is the induction of Pdx1/PDX1 in the foregut endoderm at embryonic day 8.5 (E8.5) in mouse (Guz et al., 1995) and at ∼29 days post conception (dpc) in human (Jennings et al., 2013). PDX1 expression is gradually enriched in both mouse and human pancreatic β-cells at later stages (reviewed by Pan and Wright, 2011). Rare homozygous loss of function of Pdx1/PDX1 causes pancreas agenesis in both mouse (Jonsson et al., 1994; Offield et al., 1996) and human (Stoffers et al., 1997b), whereas heterozygous mutations in PDX1 are linked to human type 2 diabetes (Stoffers et al., 1997a; Hani et al., 1999).
PDX1 genome-wide target genes have been identified in human embryonic stem cell-derived pancreatic progenitors (hPSC-PP) (Teo et al., 2015; Wang et al., 2015, 2018), mouse β-cell lines (Keller et al., 2007; Perelis et al., 2015), and adult mouse and human islets (Khoo et al., 2012). PDX1 directly regulates important transcription factors in pancreas development (Wang et al., 2018) and genes that function in endocrine system and metabolic disorders in adult islet cells (Khoo et al., 2012). However, the PDX1 cistrome of the developing pancreas has not been examined in mouse or human, mainly due to the limited availability of mouse embryonic pancreas tissue and to limited access to human fetal pancreas samples, respectively. As a result, a significant gap of knowledge remains with regard to PDX1 targets during pancreas development, particularly for human pancreas development. Single-cell studies have examined cell type-specific gene expression during the differentiation and maturation of mouse and human pancreatic cells (reviewed by Yu and Xu, 2020), including two studies of human fetal pancreas (Ramond et al., 2018; Yu et al., 2021). Integration of PDX1 cistrome with single-cell RNA-seq data could reveal PDX1 cell lineage-specific targets in the developing pancreas.
To address this critical gap, we performed PDX1 chromatin immunoprecipitation followed by high-throughput sequencing (ChIP-seq) using human fetal pancreata at 14 weeks gestation and mouse embryonic pancreata at E13.5 and E15.5, and single-cell RNA-seq using human fetal pancreata at 15-20 weeks gestation. By integrating our data with published datasets (Khoo et al., 2012; Bramswig et al., 2013; Krentz et al., 2018; Wang et al., 2018; Xin et al., 2018; Bastidas-Ponce et al., 2019), we characterized in parallel the transcriptomic signatures of human and mouse pancreas cells and integrated them with the PDX1 cistrome to identify the temporal pattern of PDX1 targets expressed in each lineage.
RESULTS
Temporal analysis of PDX1-bound genes in human developing pancreas and adult islets
To examine PDX1 genome-wide targets in human developing pancreas, we performed PDX1 ChIP-seq using human fetal pancreata at 14 weeks gestational age. All fetal pancreatic epithelial cells expressed PDX1 at this stage (Fig. S1A). HOMER de novo motif analysis showed that the consensus PDX1-binding motif was the most enriched (Fig. S1B) and histogram distribution of PDX1 peaks shows the majority of PDX1-binding sites were located within 2 kb of the transcriptional starting site (Fig. S1C). We compared these newly generated PDX1 cistromic data with published PDX1 ChIP-seq data from hPSC-PPs and adult islets (Tables S1-3) (Khoo et al., 2012; Wang et al., 2018). PDX1-binding events detected were a combination of different cell types due to the heterogeneity of cell types expressing PDX1 in the hPSC-PPs and fetal pancreata. There were 3934 genes bound by PDX1 in hPSC-PP cells only (Fig. 1A, Table S3). Canonical pathways with enrichment of these genes included DNA methylation and transcriptional repression signaling, which may help to establish proper epigenetic regulation (Wu and Sun, 2006), and retinoic acid-mediated apoptosis signaling, which may facilitate pancreatic lineage specification (Lorberbaum et al., 2020) (Fig. S2A, Table S4). There were 1092 genes bound by PDX1 in fetal pancreas only (Fig. 1A, Table S3). Canonical pathways with enrichment of these genes included the sirtuin signaling pathway and fatty acid β-oxidation I, which may help to maintain energy homeostasis for pancreas development in vivo (Yaney and Corkey, 2003; Wu et al., 2012) (Fig. S2A, Table S4). This set of PDX1-bound genes in fetal pancreas may provide clues to identify novel signals that maintain energy homeostasis to improve in vitro differentiation process using stem cells. A total of 3131 genes were bound by PDX1 in all three datasets, representing ∼50% of bound genes in the human fetal pancreas (Fig. 1A). Significantly, PDX1-bound genes in hPSC-PP, fetal and adult cells with important roles in specification and differentiation of pancreatic endocrine progenitors (e.g. NEUROG3, NKX6-1, NKX2-2, RFX3 and HNF1B) and/or in the function of adult β-cells (such as JUN, GCK and KCNJ11) (Fig. 1B). We analyzed changes in PDX1 binding in hPSC-PP, fetal and adult cells (Fig. 1B). We identified genes bound only in pancreas progenitors (NEUROD1, JUND and SIX3), during development (HNF4A, ONECUT1, CLDN4 and PTF1A) or only in adult islet (SYT4, NPY and CHGA) (Fig. 1B). In human adult islets, genes involved in β-cell maturation were bound by PDX1 (Table S3) (Khoo et al., 2012), including MAFA, a β-cell maturation driver, UCN3, a β-cell maturation marker, and GCK, a low-affinity glucokinase (reviewed by Liu and Hebrok, 2017), confirming the involvement of PDX1 in β-cell maturation. Moreover, a δ-cell differentiation regulator Hhex (Zhang et al., 2014) is bound by PDX1 in both developing pancreas and adult islet (Table S3). The δ-cell hormone gene SST is bound by PDX1 in adult human islet (Table S3). The data highlighted the role of PDX1 in δ-cell differentiation and function. Gene-ontology (GO) and upstream regulator analysis of each category for PDX1-bound genes is presented in Fig. S2A,B and Tables S4, S5. These findings expand the list of transcription factors and co-factors that are potentially either functionally congruent or directly interacting PDX1 partners.
Temporal analysis of PDX1 binding in human developing pancreas and adult islets. (A) Venn diagram of PDX1-bound genes in hPSC-PP cells, fetal pancreas and adult islet. (B) Temporal arrangement of representative stage-specific and conserved genes bound by PDX1 in hPSC-PP, fetal pancreas and adult islet.
Temporal analysis of PDX1 binding in human developing pancreas and adult islets. (A) Venn diagram of PDX1-bound genes in hPSC-PP cells, fetal pancreas and adult islet. (B) Temporal arrangement of representative stage-specific and conserved genes bound by PDX1 in hPSC-PP, fetal pancreas and adult islet.
Comparative single-cell transcriptome analysis of human fetal pancreas and adult islet cells reveals cell lineage-specific and temporally regulated genes
We used single-cell RNA-seq analysis to address the role of PDX1 in the various cell types found in the human fetal pancreas at 15-20 weeks gestational age, and we compared these newly generated fetal-cell transcriptomes with adult pancreas cell transcriptomes (Xin et al., 2018). Fetal and adult pancreatic cells largely overlapped in each cluster (Fig. 2A,B, Figs S3, S4, S6) by anchor-based method implemented in Seurat. The Seurat integration method identifies matching cell states of different datasets by considering mutual nearest neighbors as ‘anchors’. By using this anchor-based method, Seurat can exclude unique populations for the matching cell states. The differentially expressed genes among different cell lineages at fetal or adult stage were curated into transcriptomic signatures for each cell lineage (Fig. 2C-D, Fig. S5, Tables S6, S7).
Comparative single-cell transcriptome analysis of human fetal pancreas and adult islet cells reveals cell-lineage-specific and temporally regulated genes. (A,B) UMAP plots showing cell clusters (A) identified in single-cell RNA-seq from fetal and adult cells (B). (C) Heatmap showing differentially expressed genes in each cluster among human fetal pancreas cells. (D) Heatmap showing differentially expressed genes in each cluster among human adult pancreas cells. For C,D, the DEGs in each cluster were picked up by FDR <0.01 and log ratio >0.5. The DEGs commonly found in multiple clusters were excluded. If there were more than 10 DEGs in each cluster, only the top 10 DEGs by log ratio were included in the heatmap. The color codes of different clusters in A are related to different cell types in C and D. For example, the red cluster in A contains β-cells and the red bars in C,D show differentially expressed genes in β-cells. The acinar 1 cluster is in orange and the acinar 2 cluster is in yellow. Some clusters contain unidentified cell types, such as those in cyan.
Comparative single-cell transcriptome analysis of human fetal pancreas and adult islet cells reveals cell-lineage-specific and temporally regulated genes. (A,B) UMAP plots showing cell clusters (A) identified in single-cell RNA-seq from fetal and adult cells (B). (C) Heatmap showing differentially expressed genes in each cluster among human fetal pancreas cells. (D) Heatmap showing differentially expressed genes in each cluster among human adult pancreas cells. For C,D, the DEGs in each cluster were picked up by FDR <0.01 and log ratio >0.5. The DEGs commonly found in multiple clusters were excluded. If there were more than 10 DEGs in each cluster, only the top 10 DEGs by log ratio were included in the heatmap. The color codes of different clusters in A are related to different cell types in C and D. For example, the red cluster in A contains β-cells and the red bars in C,D show differentially expressed genes in β-cells. The acinar 1 cluster is in orange and the acinar 2 cluster is in yellow. Some clusters contain unidentified cell types, such as those in cyan.
Interestingly, we identified two clusters of putative acinar cells, acinar 1 and acinar 2, with acinar 1 containing almost exclusively fetal cells, while acinar 2 contained both fetal and adult cells. Despite separate sub-clustering, several transcripts had much higher expression in fetal compared with the adult cells (Fig. 3C,D). The expression of NKX6.1 and SOX9 in the acinar 1 cluster suggest that this cluster contains acinar cells in an early progenitor state. Overall, within the same cell lineage, a significant number of genes showed differential expression between fetal and adult cells (Fig. 3A-F). It is possible that contamination of highly expressed endocrine transcripts during single-cell encapsulation may affect differential expression analysis. Nevertheless, these gene profile variations over age constitute a resource for further investigation of maturational changes occurring during pancreas organogenesis and tissue homeostasis.
Heatmap showing differentially expressed genes between human fetal and adult pancreas, and potential PDX1 cell type-specific targets. (A) Genes differentially expressed between human fetal and adult β-cells. (B) Genes differentially expressed between human fetal and adult ductal cells. (C,D) Genes differentially expressed between human fetal and adult acinar cells (the acinar 1 cluster is mainly composed of fetal acinar cells; the acinar 2 cluster is composed of both fetal and adult acinar cells). (E) Genes differentially expressed between human fetal and adult δ-cells. (F) Genes differentially expressed between human fetal and adult α-cells. (G) Integrating PDX1 ChIP-seq with single-cell RNA-seq identifies stage-specific PDX1-bound genes differentially expressed in β, ductal and acinar cells.
Heatmap showing differentially expressed genes between human fetal and adult pancreas, and potential PDX1 cell type-specific targets. (A) Genes differentially expressed between human fetal and adult β-cells. (B) Genes differentially expressed between human fetal and adult ductal cells. (C,D) Genes differentially expressed between human fetal and adult acinar cells (the acinar 1 cluster is mainly composed of fetal acinar cells; the acinar 2 cluster is composed of both fetal and adult acinar cells). (E) Genes differentially expressed between human fetal and adult δ-cells. (F) Genes differentially expressed between human fetal and adult α-cells. (G) Integrating PDX1 ChIP-seq with single-cell RNA-seq identifies stage-specific PDX1-bound genes differentially expressed in β, ductal and acinar cells.
Integrating PDX1 ChIP-seq with single-cell RNA-seq identifies putative functional roles of PDX1 in target regulation in fetal pancreas and adult islets
We imputed plausible cell lineage-specific PDX1 targets in human pancreas by integrating single-cell transcriptomes with PDX1 cistrome data. By these criteria (binding and associated transcriptional expression), we called out ‘putative functional PDX1 targets’ upregulated in fetal or adult β-cells, ductal cells and acinar cells (Fig. 3G, Tables S8). PDX1 surprisingly bound important early acinar and ductal specific genes, not only during early development but also in the adult human islet. For example, fetal and adult-specific ductal genes, such as HHEX, ANXA2, KRT8 and KRT18 were bound by PDX1 from fetal pancreas to adult islet (Fig. 3G). Similarly, genes enriched in acinar cells, such as CPA1, RBPJ, SPINK1 and CELA2A, were expressed in fetal and/or adult cells, and were bound by PDX1 in fetal pancreas and in adult islets. By integrating PDX1-bound ductal/acinar genes in adult islets with histone mark ChIP-seq data of human adult β-cells (Bramswig et al., 2013), we found that the majority of PDX1 bound ductal/acinar genes in adult islets were marked by the active histone mark histone H3 lysine 4 trimethylation (H3K4me3) (Fig. S7), indicating potential active transcriptional status.
Temporal analysis of PDX1 binding and cell lineage-specific target expression from developing pancreas to adult islet in mouse
To examine species-conserved and -specific PDX1-binding in mouse pancreas, we performed ChIP-seq of whole mouse embryonic pancreata at E13.5 and E15.5 (Fig. S8A,B, Tables S9, S10). We compared these newly generated data with our previously published PDX1 cistrome of mouse adult islets (Khoo et al., 2012) (Table S11). Similar to the human cistromic analysis, the majority of PDX1 target genes (2040) were consistently bound by PDX1 in mouse pancreas from embryonic to adult stages (Fig. 4A). Some of the known genes required for pancreas development and adult β-cell function were continuously bound from E13.5-E15.5 into the adult islet, including Neurog3, Onecut1, Pax6, Foxa2, Mnx1, Nkx6-1 and Tshz1 (Pan and Wright, 2011; Raum et al., 2015). Pathway analysis and upstream regulator analysis revealed similar results to the analysis from human pancreas (Fig. S8C,D, Tables S12, S13). To characterize the functional roles of PDX1 in different cell types in mouse pancreas, we integrated the identified PDX1 targets with published single-cell RNA-seq data of mouse developing pancreas (Krentz et al., 2018; Bastidas-Ponce et al., 2019). We identified PDX1-bound genes that were highly expressed in multipotent progenitors (including Dlk1), endocrine progenitors (including Neurog3 and Neurod2), embryonic endocrine cells (including Pax6 and Mafb), and other cell types (Fig. 4B, Tables S14-S15).
Temporal analysis of PDX1 binding from developing pancreas to adult islet in mouse. (A) Venn diagram of PDX1-bound genes in E13.5, E15.5 and adult islets, and representative genes among the three datasets. (B) Temporal arrangement of representative stage-specific and conserved genes bound by PDX1 in mouse developing pancreas at E13.5 and E15.5, and in adult islets.
Temporal analysis of PDX1 binding from developing pancreas to adult islet in mouse. (A) Venn diagram of PDX1-bound genes in E13.5, E15.5 and adult islets, and representative genes among the three datasets. (B) Temporal arrangement of representative stage-specific and conserved genes bound by PDX1 in mouse developing pancreas at E13.5 and E15.5, and in adult islets.
PDX1 directs a developmentally and evolutionarily conserved ductal and endocrine program
We examined species-conserved and specific PDX1 targets in mouse and human pancreas at developmental and adult stages. We found that 762 and 1509 common mouse/human gene targets were bound by PDX1 in developing pancreas and adult islets, respectively (Fig. S9A-D, Tables S17-S19). Of these, 518 were conserved both developmentally and evolutionarily (Fig. 5A, Table S16). Integration of this data subset with our single-cell RNASeq-derived transcriptomic signatures suggested a stable occupancy by PDX1 of embryonic and adult ductal, as well as embryonic β cell, signature genes in adult islets (Fig. 5A). We next integrated the embryonic mouse and human PDX1-bound genes with the single-cell transcriptomic signatures of ductal, acinar and β-cells. None of the genes enriched in human fetal acinar cells or mouse embryonic acinar cells was part of the embryonic PDX1 cistrome. Ten genes expressed in embryonic ductal cells were PDX1 targets in pancreas development in both mouse and human (Fig. 5B). Six of these were also bound by PDX1 in adult islets in both mouse and human (bold in Fig. 5B). These ductal-specific genes are involved in mitochondrial function (ATP1B1 and ATP1A1), or related to pancreatic cancer progression (ACTN4, TPM1, CYR61 and CLDN3) (Westmoreland et al., 2012; Abou-Kheir et al., 2020). We found that some co-factors, potentially either functionally congruent or directly interacting PDX1 partners, showed stage-specific interactions with PDX1 (Table S19), suggesting that these co-factors contribute to temporally divergent regulatory functions of PDX1. In parallel, only four genes were expressed in embryonic β-cells and bound by PDX1 in embryonic mouse and fetal human samples: MEIS2, ISL1, PAX6, and SCGN – all of which play known and important roles in pancreas development (Swift et al., 1998; Ashery-Padan et al., 2004; Du et al., 2009; Malenczyk et al., 2018). MEIS2, ISL1 and PAX6 were also bound in the adult mouse and human islet.
PDX1 may direct a developmentally and evolutionarily conserved ductal and endocrine program. (A) Venn diagram and representative genes bound by PDX1 in developing pancreas and in adult islet in both mouse and human. (B) Genes bound by PDX1 in both mouse and human developmental pancreas, and expressed at high levels in embryonic ductal cells or β-cells. Genes in bold are also bound by PDX1 in both mouse and human adult islets.
PDX1 may direct a developmentally and evolutionarily conserved ductal and endocrine program. (A) Venn diagram and representative genes bound by PDX1 in developing pancreas and in adult islet in both mouse and human. (B) Genes bound by PDX1 in both mouse and human developmental pancreas, and expressed at high levels in embryonic ductal cells or β-cells. Genes in bold are also bound by PDX1 in both mouse and human adult islets.
DISCUSSION
Identification of genome-wide PDX1 targets in mouse and human developing pancreas in vivo
PDX1 has been studied extensively in the pancreas due to its roles in pancreas development and β-cell function, and its identification as a human diabetes gene. Previous studies examined PDX1 genome-wide targets in hPSC-PPs (Teo et al., 2015;Wang et al., 2015, 2018), which revealed the transcriptional networks regulated by PDX1 in the induced differentiation model system. By performing PDX1 ChIP-seq using human fetal pancreata at 14 weeks gestation and mouse embryonic pancreata at E13.5 and E15.5, we characterized the PDX1 cistrome of the developing pancreas. There are 11451 PDX1 targets identified in hPSC-PPs in vitro and 6293 PDX1 targets identified in fetal human pancreas in vivo. The much lower number of PDX1 targets in fetal pancreas than those in hPSC-PPs maybe due to a lower sensitivity of the ChIP-seq assay and/or to a lower ratio of endocrine lineage cells in human fetal pancreas than in hPSC-PPs. A higher degree of heterogeneity of cell types in fetal human pancreas may also contribute to this difference. PDX1 directly regulated key transcription factors and signaling pathways important in pancreatic lineage establishment and expansion. By integrating PDX1 ChIP-seq with single-cell RNA-seq data, we revealed potential cell lineage-specific PDX1 targets. When comparing the PDX1 cistrome of mouse and human pancreas, we identified an evolutionarily conserved gene program regulated by PDX1, which functions in key pathways in pancreas development and function, including Wnt/β-catenin signaling at the developmental stage (Sharon et al., 2019), Pi3K/AKT signaling at adult stage (Kaneko et al., 2010), and protein kinase A signaling at both developmental and adult stages (Kaihara et al., 2013) (Fig. S9C).
PDX1 may maintain cellular plasticity in adult β-cells by binding the same set of targets from developing pancreas to adult islet
Perhaps the most surprising and intriguing finding was the observation of binding of PDX1 on important acinar and ductal genes from early development to adult islet. During pancreas development, PDX1 is initially expressed in pancreatic progenitors to establish the pancreatic lineage and then gradually enriched in pancreatic β-cells to maintain β-cell function and identity (Pan and Wright, 2011; Jennings et al., 2020). The dramatic changes in PDX1 function could be caused by changes in PDX1 binding. However, although some PDX1 targets were stage specific, as also showed by Wang and colleagues (Wang et al., 2018), we found that PDX1 binding was largely stable from embryonic pancreas to adult islet in both mouse and human. These data indicate that a subset of PDX1-binding events in embryonic progenitors are maintained in the pancreatic islet, which may contribute to adult β-cell plasticity (Remedi and Emfinger, 2016), enabling β-cells to dedifferentiate and acquire a progenitor-like state under stress conditions.
PDX1 interacts with other transcription factors and cofactors to regulate pancreas development
PDX1 has been found to act as both an activator and a repressor in some cell types. In hESC-pancreatic progenitors, PDX1 activates pancreatic genes and represses hepatic genes to establish pancreatic lineages (Teo et al., 2015). For example, two canonical hepatic genes, FOXA3 (Kaestner et al., 1994) and HNF4A (Watt et al., 2003), were bound by PDX1 but not expressed in hESC-PPs and/or human fetal pancreas (Tables S3 and S8). In adult β-cells, PDX1 activates key β-cell functional genes and represses α-cell genes to maintain normal β-cell function and identity (Gao et al., 2014). PDX1 may interact with different transcription factors and/or recruit different co-factors to activate or repress the expression of its targets (Spaeth et al., 2016). Indeed, our upstream regulator analysis suggested that a large number of these PDX1 targets were co-regulated by other transcription factors, such as P53 and SOX2, and co-factors, such as LDB1 and EP300, at different stages (Figs S2B, S8D). We have previously identified PDX1 in complex with LDB1 in murine β-cells, providing validation for these bioinformatic predictions (Ediger et al., 2017). This pattern of interaction is remarkably well conserved between mouse and human.
In summary, our study comprehensively examines PDX1 genome-wide targets in mouse and human developing pancreas and reveals a core evolutionarily conserved gene program in the pancreatic islet. PDX1 surprisingly bound endocrine progenitor-, acinar- and duct-specific genes, from early development to adult islet in human pancreas, suggesting that PDX1 maintains a subset of its embryonic binding events in adult β-cells. Our data suggest a protective role of PDX1 in maintaining adult β-cell plasticity to allow adaptation and dedifferentiation under pathological conditions, and also set a valuable foundation for evaluating the efficiency and improving the protocols for human stem cell differentiation to the β-cell fate.
MATERIALS AND METHODS
Animals
Mouse experiments were performed at University of Pennsylvania, with IACUC approval. Timed pregnant CD1 mice were purchased from Charles River Laboratories. Dams were housed in AAALC approved rodent colony. Embryonic pancreata were dissected at E13.5 and E15.5.
Human fetal pancreas samples
A total of seven samples were used for these experiments (three pancreas samples at 14 weeks gestation, and one sample each at 15, 18, 19 and 20 weeks gestation). Gestational age was determined by ultrasound, usually using a crown-rump length. No other information on maternal history was available. Samples were received deidentified, in RPMI/5% FBS/antibiotics media on ice. All samples were processed for ChIP-seq or single cell RNA-seq on the day they were received, as detailed below. Fetal human pancreas tissues were obtained in the period of time between 2011-2017 from StemExpress (Los Angeles, CA, USA) and from Advance Bioscience Resources (Alameda, CA, USA) after elective termination of pregnancy. An informed consent was obtained at the time of the elective termination of pregnancy and is held by supplying company. A sample of the Advance Bioscience Resource consent was obtained and complies with NIH requirements (specifically, donation of human fetal tissues was obtained by someone other than the person who obtained the informed consent for abortion, occurred after the informed consent for abortion and does not affect the method of abortion; no enticements, benefits or financial incentives were used at any level of the process to incentivize abortion or the donation of human fetal tissues; and to be signed by both the woman and the person who obtains the informed consent). Experiments performed on these samples were exempt from IRB evaluation by the University of Pennsylvania and The Children's Hospital of Philadelphia.
ChIP-seq
We performed ChIP-seq for PDX1 in mouse embryonic pancreata (at E13.5 and E15.5) and in human fetal pancreata (at 14 weeks gestational age). We used the goat anti-PDX1 antibody (a gift from Chris Wright, Vanderbilt University, TN, USA; BCBC AB2027). Fifty to 60 individual mouse pancreata were pooled for one chromatin immunoprecipitation replicate. All analysis was performed in three replicates. Chromatin was prepared for individual replicates from three human fetal pancreata. ChIP-seq was performed as previously described (Khoo et al., 2012). All samples were sequenced in the Functional Genomic Core at University of Pennsylvania. The ChIP-seq reads were mapped to the mouse genome (mm10) using Bowtie.
PDX1 binding in hPSC-derived pancreas progenitor cells was obtained from previously published data by the Lickert laboratory (Wang et al., 2018). The PDX1 binding in adult human and mouse islets were previously published by our group (Khoo et al., 2012). After peak calling by HOMER, we continued the analysis with the ∼25% top peaks. For the mouse data, at E13.5 there were 4107 peaks (lowest HOMER score 5.28), at E15.5 were 5343 peaks (lowest HOMER score 5.03) and in mouse islet were 4559 peaks (lowest HOMER score 5.9) (Tables S9, S10). For the human data, in fetal pancreas there were 8183 peaks, representing the top 25% of the peaks (lowest HOMER score 11.1 – corresponding to 6294 genes); in hPSC-PP cells there were 21,292 peaks (lowest HOMER score – 18.9), corresponding to 11,451 genes; and for human adult islet there were 17,593 peaks (lowest HOMER score 18.2) – corresponding to 8307 genes. The majority of PDX1 peaks were within 2000 bp of the transcription start sites (TSS) (Fig. S1C). The data regarding the H3K4me3 and H3K27me3 histone marks in adult α, β and exocrine cells was previously published by Bramswig et al. (2013). Gene ontology analysis was performed with Ingenuity Pathway analysis software (Qiagen). BioVenn and DeepVenn were used to generate the Venn diagrams (Hulsen et al., 2008).
Single-cell RNA-seq and data analysis
Single islet cells were dissociated with TrypLe, and loaded on the 10x Genomics platform in the Center for Applied Genomics at The Children's Hospital of Philadelphia. Single-cell transcriptome libraries were sequenced on the HiSeq platform (Illumina). The cellranger (v.2.1,10× genomics) pipeline was used for barcode filtering, alignment (to GRCh38) and UMI counting. Secondary bioinformatic analysis was performed using the Seurat (v.2.3.4) packages in R. Briefly, sequenced 10× libraries were individually evaluated on multiple criteria to determine cells of interest, including low expression of Hemoglobin (average log2 UMI of HBA1, HBA2, HBB<8) and the number of UMI (0-65536). A total of 7421 fetal cells and 20,764 adult pancreatic cells were examined. Two single-cell RNA-seq samples of fetal and adult were integrated by anchor-based method implemented in Seurat. The top 30 principal components were used to perform clustering and visualization using a UMAP (uniform manifold approximation and projection) plot. We assessed cell type-specific gene expression signatures of acinar, ductal and endocrine cells (Fig. 2A,B). Using cell lineage-specific markers, we identified pancreatic cell clusters for α-cells (GCG), β-cells (INS), δ-cells (SST), PP cells (PPY), acinar cells (PRSS1) and ductal cells (CFTR) (Fig. 2A, Fig. S3). To identify differentially expressed genes between cell populations in fetal and adult cells, a non-parametric Wilcoxon rank sum test was performed, and P-values were adjusted using a Benjamini-Hochberg correction based on total number of genes in the dataset.
Human adult islet-cell transcriptomic data have been previously published (GSE114297; Xin et al., 2018). Mouse transcriptomic data of the developing pancreas were from Krentz et al. (2018) and Bastidas-Ponce et al. (2019).
Immunofluorescence
A small proportion of the human fetal pancreas samples at 14 weeks gestation were fixed with 4% PFA overnight, embedded in OCT and stored at −80C. Cryosections were then used for immunofluorescence for PDX1 with goat anti-Pdx1 (Santa Cruz Biotechnology, A-17, 1:500), rabbit anti-onecut1 (Santa Cruz Biotechnology, H-100, 1:1000) and guinea-pig anti-insulin (Abcam, ab7842, 1:500).
Acknowledgements
We thank Dr. Mitchell Lazar for his comments on our manuscript, and Dr Chris Wright for PDX1 antiserum and for his comments on our manuscript. We thank the University of Pennsylvania Diabetes Research Center (DRC) for the use of the Functional Genomics Core (P30-DK19525) and thank the Center for Applied Genomics at The Children's Hospital of Philadelphia.
Footnotes
Author contributions
Conceptualization: J.C.R., D.E.S., D.A.S.; Formal analysis: X.Y., J.C.R., J.K., R.Y., G.R., K.-J.W., D.E.S.; Investigation: J.C.R., J.Y., C.L., D.E.S.; Writing - original draft: X.Y., D.E.S., D.A.S.; Writing - review & editing: X.Y., K.-J.W., D.E.S., D.A.S.; Supervision: D.A.S.; Funding acquisition: D.A.S.
Funding
J.C.R. was supported by the National Institutes of Health (F32-DK-089747). D.E.S. was supported by an American Diabetes Association Junior Faculty Development Award (1-16-JDF086). D.A.S. was supported by the National Institutes of Health (R01DK105689 and R01DK121175). Deposited in PMC for release after 12 months.
Data availability
All mouse embryonic and human fetal PDX1 ChIP-seq data generated in this study are available in ArrayExpress (E-MTAB-3354). The human fetal pancreas single-cell RNA-seq data generated in this study have been deposited in GEO under accession number GSE201230.
References
Competing interests
The authors declare no competing or financial interests.