SOX9+/PTF1A+ Cells Define the Tip Progenitor Cells of the Human Fetal Pancreas of the Second Trimester

Abstract Significant progress has been made in recent years in characterizing human multipotent progenitor cells (hMPCs) of the early pancreas; however, the identity and persistence of these cells during the second trimester, after the initiation of branching morphogenesis, remain elusive. Additionally, studies on hMPCs have been hindered by few isolation methods that allow for the recovery of live cells. Here, we investigated the tip progenitor domain in the branched epithelium of human fetal pancreas between 13.5 and 17.5 gestational weeks by immunohistological staining. We also used a novel RNA‐based technology to isolate live cells followed by gene expression analyses. We identified cells co‐expressing SOX9 and PTF1A, two transcription factors known to be important for pancreatic MPCs, within the tips of the epithelium and observed a decrease in their proportions over time. Pancreatic SOX9+/PTF1A+ cells were enriched for MPC markers, including MYC and GATA6. These cells were proliferative and appeared active in branching morphogenesis and matrix remodeling, as evidenced by gene set enrichment analysis. We identified a hub of genes pertaining to the expanding tip progenitor niche, such as FOXF1, GLI3, TBX3, FGFR1, TGFBR2, ITGAV, ITGA2, and ITGB3. YAP1 of the Hippo pathway emerged as a highly enriched component within the SOX9+/PTF1A+ cells. Single‐cell RNA‐sequencing further corroborated the findings by identifying a cluster of SOX9+/PTF1A+ cells with multipotent characteristics. Based on these results, we propose that the SOX9+/PTF1A+ cells in the human pancreas are uncommitted MPC‐like cells that reside at the tips of the expanding pancreatic epithelium, directing self‐renewal and inducing pancreatic organogenesis. stem cells translational medicine 2019;8:1249&1264


INTRODUCTION
The developing mammalian pancreas has a highly organized structure, in which subpopulations of progenitor cells and lineage-committed cells can be identified based on their molecular profile and location within the pancreatic epithelium, following the initiation of branching morphogenesis. Pancreatic multipotent progenitor cells (MPCs) in the early mouse embryo are known to express several key transcription factors such as Pdx1, Sox9, and Ptf1a, which are indispensable for MPC establishment, maintenance, and proliferation [1][2][3][4]. Pdx1, a master regulator of the pancreatic program, is responsible for initiation of pancreatogenesis and is strongly expressed by all MPCs. Sox9 is crucial to maintaining the MPC niche and its role was identified through the observation of pancreatic hypoplasia in Sox9 null mice [4] and lineage-tracing studies [5]. When branching morphogenesis initiates around e12.5 in the mouse [6], MPCs also express Ptf1a and other transcription factors and are localized at the tips of the branching epithelium. These cells remain multipotent with the potentials to give rise to the three major pancreatic lineages, that is, endocrine, acinar, and ductal cells, until they become restricted to the acinar lineage [7]. Thus, Pdx1, Sox9, and Ptf1a constitute core transcription factors of murine MPCs that are necessary to determine pancreatic development.
Importantly, studies on hMPCs have been hindered by scarce accessibility to fetal tissues and few isolation methods that allow for recovery of live cells. To isolate MPCs, most laboratories rely on the use of fluorescent reporter proteins in transgenic animal models or cell-surface markers obtained by screening of in vitroderived hMPC-like cells [26][27][28]. We recently reported the successful isolation of live renal progenitor cells from human fetal kidney using RNA-based fluorescent probes (Smartflare technology) combined with fluorescence-activated cell sorting (FACS) [29]. In this study, we investigated the utility of the Smartflare technology to isolate and profile SOX9+/PTF1A+ cells in the human fetal pancreas (hFP) during the second trimester (from 13.5 to 17.5WGA).
We report here the identification of a pool of cells residing at the tips of the branched epithelium co-expressing SOX9 and PTF1A. Genome-wide gene expression studies revealed that these cells express markers of uncommitted MPCs, are proliferative, and contribute to branching morphogenesis. Singlecell RNA profiling of a 15.4WGA pancreas further confirm an MPC-like signature of the SOX9+/PTF1A+ cells.
To the best of our knowledge, this is the first report that identifies the location and the genetic profile of hMPCs characterized by the expression of SOX9 and PTF1A in the second trimester hFP. Our work contributes to the understanding of human pancreatic development and pancreatic cell specification, and has potential implications in advancing in vitro methods to generate hMPCs on large scale and derive functional insulinsecreting cells suitable for clinical application.

MATERIALS AND METHODS hFP Tissue Collection and Digestion
After informed consent, deidentified hFP samples were collected under the Institutional Review Board's approval of the University of Southern California and the Children's Hospital Los Angeles. The only information collected was gestational age and whether there were any known genetic or structural abnormalities. A total of n = 22 samples between 13.5 and 17.5WGA were used. Specifically, for histological analysis, n = 5 samples were used; for flow cytometry analysis and probe testing, n = 8 samples were used. For FACS sorting and bulk RNA-seq, n = 8 samples were used. Single-cell (sc)RNA-seq was performed on n = 1 hFP at 15.4WGA. In addition, human fetal lung cells (approved under the same IRB protocol) were collected and used as control for FGFR2 antibody specificity.
To obtain single-cell suspensions, each hFP was digested in 1 mg/ml collagenase P (Roche, Basel, Switzerland) dissolved in Medium 199 (Life Technologies, Carlsbad, CA, USA) for 20 minutes at 37 C, then washed twice in PBS containing DNAse I (Thermo Fisher Scientific, Waltham, MA, USA), and incubated with trypsin for 5 minutes and either processed for flow cytometry analysis or plated in Medium 199 supplemented with 10% fetal bovine serum (ES cell grade, Life Technologies, Carlsbad, CA, USA), 1% Penicillin-Streptomycin (Life Technologies, Carlsbad, CA, USA), and 0.2% Primocin (InvivoGen, San Diego, CA, USA) for RNA probe selection. The same digestion protocol was used to obtain a single-cell suspension for scRNAseq, except cells were washed in PBS1X without DNAse I.

Histology and Immunohistochemistry
hFP tissues were fixed, dehydrated, paraffin-embedded, and H&E stained as previously [30]. Images were acquired with a Leica DM1000 microscope. For immunohistochemistry, heat-mediated antigen retrieval was performed in a citrate-based solution (Vector Laboratories, Burlingame, CA, USA), followed by blocking in PBS containing 2% BSA for 30 minutes at room temperature (RT). The primary antibodies and dilutions used are listed in Table 1.
Antibodies were incubated for 1 hour at RT or overnight at 4 C, following the manufacturer's instructions. Host-specific Alexa Fluor-conjugated secondary antibodies (anti-mouse Alexa Fluor 555, anti-rabbit Alexa Fluor 488, and anti-goat Alexa Fluor 647; Thermo Fisher Scientific, Waltham, MA, USA) were applied for 30 minutes at RT. Sections were mounted with Vectashield medium containing DAPI (Vector Laboratories, Burlingame, CA, USA) to stain nuclei. Antibodies against PDX1 and PTF1A were both raised in goat. To avoid cross-reaction in the triple staining of SOX9/PDX1/PTF1A, antibodies against PDX1 were directly conjugated with Zenon Alexa Fluor 647 using a Goat IgG Labeling Kit (Thermo Fisher Scientific, Waltham, MA, USA) (PDX1-AF647). To perform immunocytochemistry, following ON incubation with the RNA probe, cells were fixed in 4% paraformaldehyde (Santa Cruz Biotechnology, Dallas, TX, USA) for 10 minutes and permeabilized with PBS containing 0.1% Triton X-100 (Alfa Aesar, Haverhill, MA, USA). Staining was then performed using a SOX9 antibody. Images were acquired with a Leica fluorescent microscope (Model DFC360 FX). for 10 minutes, followed by incubation with anti-FGFR2 antibody (Novus Biologicals, Centennial, CO, USA, 2 μg/10 6 cells) for 1 hour. Cells were sorted using a FACSAria (Becton Dickinson [BD], Franklin Lakes, NJ, USA). Single fluorochrome controls and Smartflare uptake probes were used to set compensation. Upon sorting, cells were immediately pelleted and frozen for subsequent RNA extraction. A mouse IgG1 isotype control (2 μg/10 6 cells, R&D systems, Minneapolis, MN, USA) was used to validate the specificity of the FGFR2 antibody. Unstained and single-positive controls were used to perform area scaling, excluding autofluorescence, and, when needed, fluorochrome compensation.
For flow cytometry analysis, cells were fixed and permeabilized with the eBioscience Foxp3/Transcription Factor Staining Buffer Set (Thermo Fisher Scientific, Waltham, MA, USA). Cells were then blocked in human IgG solution for 10 minutes and incubated with antibodies against SOX9 and PTF1A for 30 minutes on ice. Analysis was performed on a FACSCanto (BD) using the FACSDiva software. The gating strategy was as follows: cells were first gated based on forward and side scatter (FSC/SSC) to exclude dead cells, then gated for FSC-Width (FSC-W) by FSC-Height (FSC-H) and SSC-Width (SSC-W) by SSC-Height (SSC-H) to exclude potential doublets. Final sorting gates were established based on the unstained control for each sample (Fig.S1).

Bulk RNA Sequencing
RNA extraction and RNA-seq were performed by Quickbiology (Quickbiology Inc., Pasadena, CA, USA). RNA integrity was checked with an Agilent Bioanalyzer 2100. Libraries were prepared according to the QIAseq FX Single Cell RNA Library Kit (Qiagen, Hilden, Germany) with 200-300 bp insert size using 25 ng total RNA as input. Final library quality and quantity was analyzed with an Agilent Bioanalyzer 2100 and a Life Technologies Qubit3.0 Fluorometer. Paired-end tags of 150 bp in length were sequenced and read on an Illumina HiSeq 4000 (Illumina Inc., San Diego, CA, USA).
Data processing was performed using the high-performance computing cluster at the University of Southern California (https:// hpcc.usc.edu/). Roughly, 50 million 150 bp paired-end sequences were aligned to the version 29 Gencode genome based on Genome Reference Consortium Human build number 38 using the STAR aligner with "GeneCounts" output [31]. The software Morpheus (https://software.broadinstitute.org/morpheus) was used to generate heatmaps of specific enrichment gene sets and perform hierarchical clustering. Stranded values of Reads Per Kilobase of transcript per Million mapped reads (RPKM) were used to produce the heatmaps. Differential gene expression was determined using the R/Bioconductor software "edgeR" (20). Genes are considered differentially expressed at false discovery rate (FDR)adjusted p value of <.05 and log2 fold change (log2FC) > 1.5 or less than −1.5.
Gene set enrichment analysis (GSEA) was performed on differentially expressed genes between SOX9+/PTF1A+ and SOX9−/PTF1A− cells (n = 3 donors). A total of 2,233 genes were entered into the Broad Institute javaGSEA Desktop application (http://software.broadinstitute.org/gsea/downloads.jsp) as an expression data set. The data set was filtered from an initial 21,876 differentially expressed molecules based on: protein coding genes, p < .05, FDR p < .05, and log2FC either < −1.5 or >1.5. The number of permutations was set to the default (1,000), and the permutation type was set to gene set due to having n = 3 samples per phenotype. The C5 gene ontology (GO) gene set (C5.all.v6.2.symbols.gmt) was obtained from the Molecular Signatures Database v6.2 to analyze the 2,233 genes. The gene set size filter was set to the default (min = 15, max = 500), resulting in 1,440 gene sets used in this analysis (run n = 2). All other fields remained at default settings. Leading Edge Analysis was run on 6 enriched gene sets of interest within the top 45 SOX9+/PTF1A+-enriched gene sets (p < .05, FDR p < .05, normalized enrichment score [NES] > 2.50). The NES is calculated by the software (http://software. broadinstitute.org/gsea/downloads.jsp).

Single Cell RNA Sequencing
After single-cell suspension of an hFP (15.4WGA), removal of red blood cells was performed using the red blood cell lysis solution (Miltenyi Biotec, Bergisch Gladbach, Germany) following manufacturer's recommendations. Approximately 9,000 cells were captured on a 10x Chromium device using a 10X V3 Single Cell 3 0 Solution kit (10xGenomics, Chromium Single Cell 3' Regent kit V3

In Situ Identification of SOX9+/PTF1A+ Cells in the Second Trimester
Differentiated structures such as endocrine islets, ducts, branched epithelium, and portions of stratified epithelium were identifiable by H&E staining (Fig. S2). Immunostaining of E-Cadherin was used to distinguish the pancreatic epithelium from the surrounding mesenchyme. Among E-Cadherin-expressing cells, PDX1 was ubiquitously expressed at both 13.5 and 17WGA (Fig. 1A). Subsequently, we co-stained cells with PDX1, SOX9, and PTF1A. Cells coexpressing the three markers (PDX1+/SOX9+/PTF1A+) were located at the tips of the branching epithelium at 13.5WGA (Fig. 1B) and 17WGA (Fig. 1C), as well as at 14 and 17.5WGA (Fig. S3A,B).

The Percentage of SOX9+/PTF1A+ Cells Decreases Overtime
Given that PDX1 expression was ubiquitous in the epithelium, we focused our attention to the SOX9+/PTF1A+ cells. The spatial and temporal identification of SOX9+/PTF1A+ cells within the human pancreatic epithelium is shown in a schematic diagram in Figure S3C. To determine the prevalence of SOX9+/PTF1A+ cells, we performed flow cytometry analysis on primary cell suspensions of hFP at 13.5W (n = 1), 14.5W (n = 2), 17W (n = 2), and 17.5W (n = 1). SOX9+/PTF1A+ cells were detected at all GA ( Fig. 1D). Plotting the percentage of each subpopulation (SOX9 +/PTF1A−, PTF1A+/SOX9−, SOX9+/PTF1A+) across GA showed that at earlier GA there was a higher prevalence (over 60%) of SOX9+/PTF1A+ cells, which decreased at later GA to less than 20% (Fig. 1E). In particular, we observed a large drop in the percentage of SOX9+/PTF1A+ cells between 13.5 and 14.5WGA. The opposite was true for PTF1A+/SOX9− cells, which increased during the same time frame. SOX9+/PTF1A− cells constituted a small population with a constant percentage of cells at all GA.

SOX9+/PTF1A+ Cells Do Not Express Acinar or Endocrine Proteins
SOX9+/PTF1A+ cells were further examined for the expression of endocrine (insulin and glucagon) or acinar (amylase) differentiation markers using immunohistological staining. Tip cells did not express insulin ( Fig. 2A), glucagon (Fig. 2B), or amylase ( Fig. 2C) at 14WGA, consistent with results published by others [32]. The effectiveness of the anti-amylase antibodies was verified by staining human adult pancreatic tissue (Fig. 2C). These results demonstrate that SOX9+/PTF1A+ cells do not express lineage-specific markers.
To further investigate the phenotype of these cells, we evaluated the expression of carboxypeptidase 1 (CPA1), which is expressed in early murine tip MPCs at e12.5 and later in committed acinar cells at e14 [7]. At 13.5WGA, CPA1 was expressed in a scattered pattern throughout the epithelium; all CPA1+ cells were also SOX9+ (Fig. 2D, arrowheads). In contrast, at 17WGA, some CPA1+ cells lost SOX9 expression (Fig. 2D, asterisks) while others maintained it (Fig. 2D, arrowheads). These results suggest that patterning of the tip progenitor cells versus committed acinar cells is determined around 17WGA. SOX9+/PTF1A+ tip cells also expressed PCNA, a proliferation marker, demonstrating that they were in a proliferative state (Fig. 2E, arrowheads).

Validation of the Smartflare Technique
Smartflare technology uses RNA probes to recognize RNA targets (including RNAs that code for intracellular proteins and nuclear factors) and can be used to sort live cells [33,34]. A 15WGA hFP was dissociated to a single-cell suspension and incubated with Cy5-conjugated SOX9 RNA probes (Fig. 3A). Cells were then fixed and stained with Alexa Fluor 488-conjugated SOX9 antibodies. The SOX9 RNA probe colocalized with SOX9 protein (Fig. 3B), demonstrating that the antibody and the RNA probe identified the same cells. To further validate the technology, we performed flow cytometry analysis using the SOX9 Smartflare probe and/or an antibody targeting SOX9. The SOX9 probe alone (Fig. 3C) or the anti-SOX9 antibody alone (Fig. 3D) stained~88% or~71% of the total cell population, respectively. When used in combination, the Smartflare probe and the antibody co-stained the same population, confirming specificity of the RNA probe (Fig. 3E). We calculated that Smartflare identified 97.5% of the SOX9-expressing cells based on antibody staining (see legend for Fig. 3). Two negative controls, a scramble probe sequence (Fig. 3F) and a probe for a kidney-specific transcription factor, CITED1 [29], did not stain these cells (Fig. 3G). Together, these results validate the specificity of the SOX9 RNA probe and corroborate the application of this technology for identifying and sorting live cells. Smartflare technology does not affect cell viability, as previously reported [29,[35][36][37][38].
To further confirm, anti-FGFR2 antibodies stained 5.3% positive in primary human fetal lung cells (Fig. S5E), which were used as positive control. Together, these results demonstrate that the majority of the SOX9+/PTF1A+ cells do not express FGFR2, at least at the cell-surface. We speculate that SOX9 +/PTF1A+ cells may use other FGFRs, such as FGFR1, 3, and 4, for signaling. Alternatively, human SOX9 may function differently compared to the murine gene in inducing Fgfr2 expression [5]. Given the very low expression of cell surface FGFR2, we focused only on SOX9+/PTF1A+/FGFR2− cells for subsequent analysis.  (Fig. 5A). This result was further corroborated by the hierarchical clustering analysis (one minus Pearson correlation) of overall gene expression (p < .05, FDR p < .05) (Fig. 5B), which showed that SOX9+/PTF1A+ cell populations at 13.5 and 14WGA clustered together, whereas a change with a slight increase in gene expression was appreciated at 17.5WGA. In contrast, SOX9 −/PTF1A− fractions displayed more divergence in expression patterns. This result is consistent with the drop in the percentage of SOX9+/PTF1A+ cells over time (Fig. 1E and Fig. S4), and suggests that during the second trimester more rapid changes were happening within the SOX9−/PTF1A− fractions, which include cells committed to not only endocrine, ductal, and acinar fate but also other cell types of the endothelial and mesenchymal lineages. We next analyzed the expression of markers that are characteristic of either MPC identity or lineage differentiation. SOX9+/PTF1A+ cells displayed enrichment   for genes expressed by multipotent progenitor cells compared to SOX9-/PTF1A-cells. Genes, such as the transcription factors MYC, YAP1, and GATA6, as well as CDH1 (E-Cadherin), were enriched in all three SOX9+/PTF1A+ populations, regardless of the GA analyzed (Fig. 5C, left panel). As expected, expression of endocrine and ductal differentiation markers showed an opposite trend (Fig. 5C, right panel) and was downregulated in SOX9+/PTF1A+ cells. However, the typical exocrine cell markers, CPA1 and GP2 for acinar and ALDH1A2 for centroacinar cells, were enriched only in the SOX9+/PTF1A+ cells isolated from the later but not earlier GAs (Fig. 5C, right panel). These results are consistent with the lack of committed acinar cells in the 14WGA hFP (Fig. 2C), and suggest that CPA1, GP2, and ALDH1A2 mark progenitor cells instead during the second trimester.
In mice, knockout of Hippo signaling in the whole pancreas causes defects in exocrine but not endocrine development [44,45], suggesting MPCs are not affected. In contrast, Wnt signaling has a major effect on MPC maintenance in murine fetal pancreas [46]. Hippo is known to regulate the expansion of pancreatic MPCs derived from embryonic stem cells [47]; however, whether Hippo plays a functional role in human fetal MPCs remained to be established. This result suggests that Hippo but not Wnt signaling may be the dominant pathway governing MPC functions at the second trimester.

Gene Set Enrichment Analysis (GSEA) Reveals Active Branching Morphogenesis of the SOX9+/PTF1A+ Cells
To corroborate our hypothesis and further characterize the SOX9+/PTF1A+ cells, we performed GSEA. Among the top-50 GO (Gene Ontology) sets, the extracellular matrix (ECM), tissue development and morphogenesis, and stem-cell proliferation emerged as major processes, as shown by their higher normalized enrichment scores (Fig. 6A). The rankings of six selected gene sets were further listed (Fig. 6B). Next, we performed leading-edge analysis to visualize the genes that were significantly enriched in the SOX9+/PTF1A+ cells, with regards to the ECM remodeling (Fig. 6C), epithelial branching morphogenesis (Fig. 6D), and cell proliferation (Fig. 6E). A few hubs of genes of interest were identified accordingly. In the GOs related to ECM, we noted genes for matrix remodeling (such as ADAM and MMP family genes) ( Fig. 6C; black boxes), matrix deposition (such as COL, GPC, FB, and LAM family genes) ( Fig. 6C; red boxes), and cell/ECM interactions (DDR and ITG family) ( Fig. 6C; blue boxes). In the GOs related to epithelial branching morphogenesis (Fig. 6D), various genes related to growth signals (Fig. 6D, asterisks) and cell/ECM interactions were also enriched. Finally, in the GO of cell proliferation, a matrix deposition gene (FBLN1) appeared again (Fig. 6E), as well as genes (FGFR1, FOXF1, GLI3, TBX3, TGFBR2, and YAP1) that were significantly represented in the GO of epithelial branching morphogenesis (Fig. 6D,E, asterisks). Taken together, these results suggest intimate relationships between branching morphogenesis, ECM remodeling and cell proliferation in the SOX9+/PTF1A+ cells.
Single-Cell (sc) RNA-Sequencing of a Human Pancreas of 15.4WGA To further dissect the developmentally-regulated processes of the hMPCs, we performed scRNA-seq (hFP, 15.4WGA). This GA comes after the reduction in the % SOX9+/PTF1A+ cells that was observed and therefore more representative of the MPCs of the second trimester (Fig. 1E). A total of 17 different clusters were identified and visualized in a t-Distributed Stochastic Neighbor Embedding (tSNE) plot, which reduces the highly complex gene expression patterns of the individual cells into a two-dimensional plot (Fig. 7A).
To determine whether the abovementioned markers were expressed beyond cluster #10, we removed the filter on SOX9 +/PTF1A+ cells and examined the gene expression patterns in all clusters (Fig. S6). Except for ITGB1, PCNA, and FGFR1, which were expressed in other clusters, SOX9, PTF1A, CDH1, FGFR2, and YAP1 were mostly exclusively represented in cluster #10. PDX1 was enriched in both cluster #10 and cluster #12, which identifies the endocrine cluster (Fig. S6), confirming its role in both MPCs and endocrine lineage. This result strengthens our findings from bulk RNA-seq.
Next, we investigated the expression of various lineage markers (Fig. S7). The mesenchyme marker VIM was expressed in the majority of the clusters at this developmental stage (Fig. S7A), whereas the marker for nascent endothelium (TIE1) was found only in cluster #6 (Fig. S7B). Cluster #12 was enriched for committed endocrine cells, as shown by GCG expression (Fig. S7C). Consistent with prior reports [22,24,48,49], PDX1, NKX6.1, and GATA6 were expressed in both MPC-like (#12) and endocrine (#10) cells (Fig. S7D-F). Compared to PDX1, NGN3 was expressed in fewer cells in the endocrine and the MPC clusters (Fig. S7G), supporting the idea that NGN3 is expressed transiently before commitment from progenitor cells into endocrine cells. Consistent with the lack of acinar cells (Fig. 2C), AMY2A was not detected (Fig. S7H) and AMY2B was minimally expressed (Fig. S7I). Interestingly, a pluripotent stem cell marker, NANOG, was expressed mostly in the MPC cluster (#10, Fig. S7J)-the significance of this finding is unknown, but  NANOG has been associated with the formation of pancreatic ductal adenocarcinoma [50].
To further identify potential functional pathways within the cluster containing SOX9+/PTF1A+ cells, we analyzed the genes that were differentially expressed between cluster #10 and the other clusters combined, using IPA analysis. The most significant processes are listed in Figure 7D, which included cell morphology, digestive system development and function, organ morphology, organismal development, cell death and survival, cellular assembly and organization, cellular development, cellular growth and proliferation, embryonic development, and organ development. Genes identified in each of these biological functions are listed in Figure S8. Together, these results confirm the bulk RNAseq data and demonstrate that SOX9+/PTF1A+ cells of the 15.4WGA are active in organ development, morphogenesis and proliferation.

DISCUSSION
Several mouse models have proved useful to elucidate the mechanisms involved in the maintenance of tip MPCs and their lineage commitment during pancreatic development [2,51]. Studies of the human pancreas revealed similarities between murine and human systems, including the expression of crucial transcription factors. However, important differences with regard to the timing and expression patterns of some of these factors were found [22,32]. Few studies reported on the development of the human pancreas in the second trimester and mostly focused on endocrine differentiation [52,53]. A recent work on the single-cell profiling of the hFP at 9WGA showed that distinct populations of progenitor cells and endocrine cells, each containing subpopulations of maturation intermediates, can be found at a single time point during development [54].
Information on human MPCs residing within the hFP has been limited. Although some studies have investigated the early embryonic stage of human pancreas development [24], others have mainly focused on the characterization of SOX9+ trunk progenitor cells. The spatial and temporal expression pattern of SOX9 during human pancreas development was dissected, with an emphasis on its role in islet cell differentiation [53]. However, the role of PTF1A in the hFP has been elusive. Our current studies build on the prior knowledge and show that SOX9+/PTF1A+ cells are found at the tips of the pancreatic branching epithelium, which is a different population than the previously characterized trunk progenitor cells.
A technological significance of our study is the application of RNA probes [33,34] to isolate live progenitor cells from human fetal pancreas. This method has proven valid in other cell types [35][36][37][38], and using this technology we have successfully isolated renal progenitor cells in a prior report [29]. Here, we isolated pancreatic cells expressing SOX9 and PTF1A transcripts, demonstrating a high correlation (97.5%) between the probe and the antibody detection of the same marker. However, it should be noted that Smartflare probes are not ideal for the quantitative analysis of subtle mRNA levels due to a narrow dynamic range of detection [55]. Nevertheless, our data validate the qualitative use of Smartflare probes to identify cells expressing SOX9 and PTF1A and study their genetic profile.
All of the three main pancreatic lineages, including endocrine, ductal and acinar, originate from an initial stratified epithelium containing the primordial MPCs. Once the process of branching morphogenesis begins, progenitor cells can be discriminated into trunk and tip progenitor cells, based on their location within the epithelium. The trunk progenitor cells eventually give rise to ductal and endocrine cells, whereas the MPC domain, located at the branching tips, maintains multipotency until becoming restricted to the acinar lineage. This switch of the tip multipotent progenitor cells to the acinar lineage coincides with the "second wave" of the endocrine pancreas development in mice [7,56]. The first wave is characterized by the presence of a few Ngn3+ endocrine progenitor cells and endocrine cells that co-express multiple hormones. In the second wave which starts around e12.5, an upregulation of the Ngn3+ endocrine progenitor cells is followed by the appearance of endocrine cells that express only single hormones. However, in the human pancreas a single-phase of NGN3 expression and endocrine commitment was detected starting at 8WGA and onward [24]. This further highlights the differences between the murine and the human pancreas development.
Information regarding the occurrence and timing of a developmental switch of the tip MPCs during human development is limited [57]. The proportion of SOX9+/PTF1A+ cells among the total cells in the hFP at 13.5WGA was higher than the proportion we observed at 14-17WGA. Concomitantly, the proportion of PTF1A+/SOX9− cells increased. Thus, the loss of SOX9 expression and the maintenance of PTF1A suggest a commitment of the SOX9+/PTF1A+ cells toward acinar lineage [58]. This was further corroborated by comparing SOX9 and CPA1 expression between earlier and later GA. Cpa1 was confirmed as MPC marker through lineage tracing studies [7]. Cpa1, Pdx1, Ptf1a, and c-Myc mark the murine tip MPC domain and these cells are multipotent before becoming restricted to the acinar lineage at mid-gestation around e14 [7]. Here, we observed that all CPA1+ cells were also SOX9+ at 13.5WGA, whereas at later GA, some of the CPA1+ cells were not SOX9+ (Fig. 2D), suggesting that patterning of the tip cells occurred in the hFP between 13.5 and 17WGA. However, contrary to a prior report [59], we did not observe expression of amylase at the protein level in any specimen (Fig. 2C). Similarly, mRNAs for AMY2A and AMY2B were largely absent in the scRNA-seq analysis. These results confirm amylase as a late differentiation marker for the acinar lineage.
Bulk RNA-sequencing profiling confirmed that SOX9+/PTF1A+ cells from n = 3 donor tissues clustered together, revealing similarities on overall gene expression; in contrast, the negative fractions displayed greater divergence. This result suggests that more rapid changes occur during developmental time within the non-MPC fractions of the hFP, which include committed or differentiating cells, whereas the SOX9+/PTF1A+ fraction preserves its identity/ gene expression profile (Fig. 5B). Interestingly, we found GATA6 enriched in the SOX9+/PTF1A+ cells (Figs. 5C and 7C). Contrary to what has been observed in the mouse [60], mutations of the human GATA6 gene have been associated with pancreatic agenesis [61][62][63], consistent with GATA6 as a MPC marker in human. A recent work also confirmed the role of GATA6 as a regulator of definitive endoderm (DE) and early pancreatic morphogenesis in GATA6-null human pluripotent stem cells, which failed to enter the DE program upon induction to differentiate [64].
Molecular signaling pathways such as Notch, FGF, and WNT/β-catenin are known to play critical roles in determining specification of nonhuman mammalian pancreas and regulating endocrine/exocrine differentiation [5,40,41,43]. Surprisingly, these pathways did not appear to play significant roles in the SOX9+/PTF1A+ cells, using various bioinformatics analyses. We, instead, found the Hippo pathway being relevant in the SOX9 +/PTF1A+ cells. The Hippo pathway is a universally conserved pathway involved in regulating organ size growth during development and controls the fine balance between pluripotent stem cells and lineage-restricted cell pools in mammals [65]. Evidence indicates that YAP, which is the main nuclear target of Hippo signaling and is negatively regulated by it, is highly expressed by the stem or progenitor cell niche of different adult epithelial tissues, including the intestinal crypts and the skin in both mice and humans [66,67]. The Hippo signaling pathway acts during the second wave in mice to determine proper architecture of the pancreas but not endocrine differentiation [44]. More recent studies demonstrate that the YAP/TEAD complex functions during early pancreas development to expand human MPCs, and is subsequently lost during lineage differentiation [47]. Inhibition of YAP results in enhanced differentiation of functional ß-cells from hPSCs, whereas a sustained in vitro activation of YAP results in impaired ß-cell differentiation [68]. Thus, our finding on the YAP/ TEAD complex being upregulated in the SOX9+/PTF1A+ cells isolated from the second trimester (Fig. 5E) is consistent with the idea that these cells are the uncommitted MPC-like cells with potential to direct organ size and architecture. expansion and organ growth [69]. From the GSEA analysis, we indeed found SOX9+/PTF1A+ cells enriched for expression of genes involved in matrix remodeling and cell proliferation, suggesting active branching and epithelium expansion. ECM remodeling proteins regulate matrix degradation and allow progenitor cells to expand [70]. We found ECM remodeling genes, including family proteins of metalloproteinases and a disintegrin and metalloproteinase with thrombospondin motif (ADAMTS), were upregulated in SOX9 +/PTF1A+ cells, compared to SOX9−/PTF1A− cells (Fig. 6B). Integrinmediated ECM-cell interaction has long been recognized for its importance in tissue development and maintenance of progenitor cells. We observed upregulation of ITGB1, consistent with previous findings in the hFP [71,72], and ITGA6, which may bind to ITGB1, was reported as a marker for the isolation of progenitor cells from the human fetal pancreas [73]. In addition, we detected enrichment of ITGAV [74], ITGA2 and ITGB3 at all GAs analyzed, which have not been previously reported as putative markers of human MPCs. FGF signaling is known to promote the proliferation of murine pancreatic progenitor cells via an Fgfr2/Sox9 regulatory axis. However, we detected very few SOX9+/PTF1A+ cells expressing FGFR2 protein at the cell surface (~1%, Fig. 4), despite the fact that approximately 40% of these cells express the FGFR2 transcripts, as detected in the scRAN-seq analysis (Fig. 7). Interestingly, FGFR1 transcript was enriched in the SOX+/PTF1A+ population (Fig. 7A) together with FGF7 and FGF2 (Fig. 6D). FGF2 is a direct ligand for FGFR1. The isoform IIIb of the FGFR1 was reported to be exclusively expressed in the rat pancreatic epithelium and as such proposed as a potential marker of pancreatic progenitor cells [75]. Inhibition of FGFR1 was shown to promote the terminal differentiation of human induced pluripotent stem cell (hiPSC)-derived endocrine progenitor cells into endocrine cells [76]. Therefore, FGFR1, rather than FGFR2, could be a more significant player of the FGF signaling in human pancreas development. Further investigation will be needed to validate this hypothesis.

CONCLUSION
We have demonstrated the presence of human MPC-like cells expressing SOX9 and PTF1A at the tips of the branching epithelium of the developing pancreas during the second trimester. Multiple gene-expression analyses of SOX9+/PTF1A+ cells indicate that these cells retain the profile of MPCs, including expression of known key transcription factors, proliferation, branching, and signaling pathways. Some newly-identified candidate genes, such as FGFR1, TGFBR2, FOXF1, GLI3, TBX3, as well as NANOG, ITGAV, ITGA2, and ITGB3, have not been previously described in either murine or human fetal pancreatic progenitor cells, making them suitable candidate for screening and enrichment of in vitro-derived human pancreatic progenitor cells. Mechanosignaling via integrin molecules might play a crucial role in maintaining the human pancreatic MPC niche. Further investigations into the functions of hMPCs are needed and are currently in progress in our laboratories.
In summary, the results we have described here provide a better understanding of the molecular programs that sustain pancreatic progenitor cells during the second trimester of human development. Although many aspects of the progenitor-cell molecular signature are conserved between mice and humans, we have identified some key differences that may provide insight in developing better protocols for differentiation of human pluripotent cells into the pancreatic lineages. Ultimately, these efforts may lead to therapeutically relevant cells such as beta cells that might be used for cell replacement therapy for type 1 diabetes.