Understanding human fetal pancreas development using subpopulation sorting, RNA sequencing and single-cell profiling

Abstract

To decipher the populations of cells present in the human fetal pancreas and their lineage relationships, we developed strategies to isolate pancreatic progenitors, endocrine progenitors and endocrine cells. Transcriptome analysis of the individual populations revealed a large degree of conservation among vertebrates in the drivers of gene expression changes that occur at different steps of differentiation, although notably, sometimes, different members of the same gene family are expressed. The transcriptome analysis establishes a resource to identify novel genes and pathways involved in human pancreas development. Single-cell profiling further captured intermediate stages of differentiation and enabled us to decipher the sequence of transcriptional events occurring during human endocrine differentiation. Furthermore, we evaluate how well individual pancreatic cells derived in vitro from human pluripotent stem cells mirror the natural process occurring in human fetuses. This comparison uncovers a few differences at the progenitor steps, a convergence at the steps of endocrine induction, and the current inability to fully resolve endocrine cell subtypes in vitro.

Keywords: Diabetes; Endocrine; Human; Islets; Pancreas; Stem cells.

Fig. 1.

Characterization of sorted cell populations from human fetal pancreata by RNA-seq at 9 WD. (A) Flow cytometry analysis of the expression of GP2 and ECAD on CD45⁻CD31⁻EPCAM⁺ cells at 9 WD. CD142 and SUSD2 expression was analyzed in GP2⁻ECAD⁺ and GP2⁻ECAD⁻ cells. FACS plots are representative of three independent pancreata at 9 WD. (B) Scheme of the experimental setup: GP2⁻ECAD⁺CD142⁺SUSD2⁻ (population A, yellow), GP2⁻ECAD⁺CD142⁻SUSD2⁻ (population B, orange), GP2⁻ECAD^lowCD142⁻SUSD2⁺ (population C, red) and GP2⁻ECAD^lowCD142⁻SUSD2⁻ (population D, blue) were cell sorted and analyzed by RNA-seq. All populations were CD45⁻CD31⁻EPCAM⁺ and derived from three independent pancreata at 9 WD. The population color code is re-used throughout the article. (C) Principal component analysis map on the RNA-seq from all four sorted populations in triplicate at 9 WD. (D) Heatmap displaying the expression of the 1007 differentially expressed genes in triplicate in all four populations in RPKM, along with the hierarchical clustering. Clusters are named on the right side and some gene examples are indicated in boxes. For more extensive examples, Table S1 highlights genes enriched in the four sorted cell populations.

Fig. 2.

Characterization of populations A, B, C and D at 9 WD. Heatmaps displaying the expression of multipotent genes (PDX1, NKX6-1, SOX9 and ONECUT1) (A); progenitor genes and early endocrine markers (PAX6, PAX4, NEUROD1, ACOT7, NEUROG3, ARX, FEV, ETV1 and NKX2-2) enriched in population C (B); Endocrine genes (IAPP, MAFA, FFAR1, G6PC2, ISL1, PCSK1, MAFB, GCK and PCSK2) enriched in population D (C); and pancreatic hormones (PPY, GHRL, INS, GCG and SST) in sorted populations A, B, C and D (D). Heatmaps are displayed using a log2 expression scale.

Fig. 3.

CD133 marks human fetal pancreatic ductal cells. (A) Flow cytometry analysis displaying the expression of CD133 in populations B, C and D at 11 WD. FACS plots are representative of three independent pancreata at 11 WD. (B) Cell frequencies of population B CD133⁺, population C CD133⁺ and population D CD133⁺ from three independent pancreata at 10 WD. (C) Expression of CFTR, CHGA, NEUROG3 and NKX2-2 (9-11 WD) by RT-qPCR in population B CD133⁺, population B CD133⁻, population C CD133⁻, population D CD133⁻ and population D CD133⁺. *P<0.05, **P<0.001 (t-test; data are mean±s.e.m. of three independent pancreata).

Fig. 4.

Endocrine cells in population D are the most granular. (A) Flow cytometry displaying the granulometry (SSC) in population C and population D_CD133⁻ at 10 WD. FACS plots are representative of three independent pancreata at 10 WD. (B) Frequency of SSC^hi and SSC^low in population D_CD133⁻ at 8, 9 and 12 WD (data are mean±s.e.m. of three independent pancreata at each stage). (C) Expression of CHGA, INS, NEUROD1 and PAX6 at (10-12 WD) by RT-qPCR in population D_CD133⁻, SSC^low or SSC^hi. *P<0.05 (t-test; data are mean±s.e.m. of three independent pancreata).

Fig. 5.

Single-cell profiling of sorted human fetal pancreatic cells. (A) Heatmap showing gene expression for a selected set of genes in individual cells sorted from human fetal pancreas based on cell-surface markers and granularity discriminating populations B_CD133⁻, C, D_HI and D_LO. The genes with greatest variance are shown. A heatmap with all genes is provided in Fig. S4A. Cells of populations B_CD133⁻ and C are derived from three individual pancreata at 9 WD; cells from populations D_HI and D_LO are derived from two individual pancreata at 9 WD. (B) t-SNE plot of single-cell qPCR data from human fetal pancreas colored by population (left panel) or according to expression level of selected genes (right panel). See also Fig. S5. (C) Developmental trajectory for endocrine differentiation in human fetal pancreas using pseudotemporal ordering with Monocle. Cells on the trajectory are colored according to population (left panel) or pseudotime (right panel). Specific genes characteristic for each branch are indicated on the left panel. See also Fig. S6.

Fig. 6.

Gene expression profile for human fetal endocrine progenitors and early endocrine cells. t-SNE plot of single-cell gene expression data for the endocrine-biased cluster (Cluster III-V) formed by the hierarchical clustering shown in Fig. 5A (left panel). Cells are colored according to population (left panel) or gene expression level (right panel) for selected genes. Colours reflect the Log2Ex scale.

Fig. 7.

Comparison of the single-cell expression profile of pancreatic cells generated in vitro with in vivo fetal and adult pancreata. (A) Heatmap showing the gene expression level in single cells of populations B and C isolated from human fetal pancreas (in vivo) or from hPSC-derived cultures (in vitro). Selected genes are indicated, colours reflect the Log2Ex scale. A heatmap with all gene annotations is provided in Fig. S9. (B) t-SNE plot of single-cell qPCR data for populations B and C isolated from human fetal pancreas (in vivo) or from hPSC-derived cultures (in vitro). Gene expression level in individual cells for RFX6 and CDX2 is indicated in the bottom plots. (C) t-SNE plot combining data from the present study on sorted hPSC-derived cells (populations C and D) with data from a previously published dataset comprising endocrine-biased hPSC-derived cells collected at different time points (early, stage 4 day 1 and 3; mid, stage 5 day 3+stage 6 day 2; late, stage 6 day 7+stage 7 day 7). Expression level of hormonal genes are mapped onto the t-SNE plot as indicated. (D) t-SNE plot combining the data from C with data on the sorted cell populations from human fetal pancreas (populations B, C and D_HI) from the present study and data on adult human islet cells (from the same previously published dataset as used in C). See also Fig. S10.