Supplementary Figure 1: Human in vitro MPCs recapitulate key features of in vivo MPCs.

(a) Immunohistochemistry analysis of in vivo MPCs from Carnegie stage 16–18 human embryonic pancreas, and immunofluorescence analysis of in vitro MPCs show expression of stage-delimiting MPC TFs in both sources of MPCs. (b) Heatmap showing RNA-seq FPKM signal in MPCs and 23 control tissues for TFs that are enriched in pancreatic MPCs and for a similar number of known lineage-specific non-pancreatic TFs. (c) Expression correlation matrix showing Spearman coefficient values for transcript levels from in vivo and in vitro MPCs vs. 23 control tissues. (d) Z score correlation density plots. Comparisons of in vivo MPCs with an unrelated tissue (fetal heart, left panel), or between tissues from the same lineage, but different stages (adult and fetal heart, right panel), do not show high correlation, in contrast with data presented in Fig. 1c that shows highly correlated Z-scores for in vivo and in vitro MPCs. Spearman coefficient values are shown for each comparison. Color scale depicts number of transcripts. (e) Motif discovery in different FOXA2 ChIP-seq datasets, shows a similar binding motif for this TF in all samples. P values and percentages of bound versus background regions are indicated below each motif logo. (f) Regions enriched in FOXA2 and H3K4me1 in chromatin from in vivo MPCs also show H3K4me1-enrichment in in vitro MPCs, but not in control samples (mammary epithelial cells, myotubes, CD133 + umbilical cord blood and hESCs). The heatmap shows FOXA2 and H3K4me1 signal centered on these regions (see Methods for details). Note that even though the regions were pre-selected from in vivo MPC data, H3K4me1-enrichment is stronger in chromatin from in vitro MPCs, reflecting the larger number of cells used for ChIP-seq. (g) H3K4me1 and FOXA2 signals in the whole genome were binned in 5 Kb for H3K4me1 and 1 Kb for FOXA2. These signals were highly correlated in biological replicates (R > 0.8).