Improved maturity and yield of pluripotent stem cell-derived hepatocytes for 2D and 3D modeling of human liver pathophysiology

Robust and scalable differentiation of human pluripotent stem cells into high-quality hepatocyte-like cells (HLC) is achieved by acting on Wnt/b-catenin and TGFb pathways and taking advantage of latest knowledge on human liver development. Obtained HLC are comparable to primary human hepatocytes and superior to stem cell-derived hepatocytes generated with previously described protocols. As a proof of concept, such HLC were used to assess drug hepatotoxicity and study human ductal plate and bile duct morphogenesis in a complex liver organoid model. Comparison of liver-specific gene expression profile of our hepatocyte-like cells with previously described primary human hepatocytes (PHH: GSM2753380, GSM2753381 and GSM2753382 described in Gao et al. Stem Cell Rep. 2017;9:1813?1824) and stem cell-derived hepatocytes (HLC and hiHeps: GSM2411900, GSM2411902, GSM2411904 described in Warren et al. Cell Stem Cell 2017;20:547-557; GSM2753372, GSM2753373, GSM2753374 described in Gao et al. Stem Cell Rep. 2017;9:1813?1824). Prior to all analysis, the raw gene counts were normalized using the logCPM method by dividing each column by the total sum of its counts, multiply by a million and followed by the application of a log10-transform. Genes with the most variable expression were selected and transformed using the Z-score. Principal Component Analysis was generated using the Plotly's Python graphing library and sklearn Python module. Heatmap was created using Clustergrammer. We then compared HLC obtained with the new differentiation protocol with previously described HLC and PHH by performing an UMAP-based dimensionality reduction of the top 17 principal components obtained analyzing the 3000 most variable genes across datasets. For PHH, we used the following datasets: GSM2817112, GSM2815712, GSM2815717, GSM3146358, GSM3146359, GSM3067786, GSM3067787, GSM3067788, GSM3067804, GSM2753380, GSM2753381, GSM2753382 (described in Gao et al. Stem Cell Rep. 2017;9:1813?1824; Fu et al. Cell. Res. 2019;29:8?22; Xie et al. Cell. Res. 2019;29:696?710; Kim et al. J. Hepatol. 2019;70:97?107). For pluripotent stem cell-derived hepatocytes, we used the following datasets: GSM1817062, GSM2411900, GSM2411901, GSM2411902, GSM2411903, GSM2411904, GSM2411905, GSM2411906, GSM2411907, GSM2411908, GSM2411909, GSM2411910, GSM2411911, GSM2411912, GSM2411913, GSM2411914, GSM2411915, GSM2411916, GSM2411917, GSM2411918, GSM2411919, GSM2411920, GSM2411921, GSM2411922, GSM2411923, GSM2411924, GSM2411925, GSM2411926, GSM2411927, GSM2411928, GSM2411929, GSM2411930, GSM2411931, GSM2411932, GSM2411933, GSM2411934, GSM2411935, GSM2411936, GSM2411937, GSM2411938, GSM2411939, GSM2411940, GSM2411941, GSM2411942, GSM2411943, GSM2411944, GSM2411945, GSM2411946, GSM2411947, GSM2411948, GSM2411949, GSM2411950, GSM2411951, GSM2411952, GSM2411953, GSM2411954, GSM2411955, GSM2411956, GSM2411957, GSM2411958, GSM2411959, GSM2411960, GSM2411961, GSM2411962, GSM2411963, GSM2411964, GSM2411965, GSM2411966, GSM2411967, GSM2753372, GSM2753373, GSM2753374, GSM2753375, GSM2753376, GSM2753377, GSM2753378, GSM2753379, GSM3664948, GSM3664949, GSM3664950 (described in Warren et al. Cell Stem Cell 2017;20:547-557; Li et al. Nat. Commun. 2017;8:15166; Gao et al. Stem Cell Rep. 2017;9:1813?1824; Touboul et al. J. Hepatol. 2016;64:1315?1326). In brief, raw counts from different RNA-seq datasets were normalized (divided by the total count, multiplied by 1 million and natural-log transformed using log1p) and center-scaled. From the scaled data, the most variable genes (3,000) across datasets were used to conduct principal component analysis and the top principal components (17) were selected for dimensionality reduction and 2-dimention representation using the UMAP (Uniform Manifold Approximation and Projection) and DimPlot functions from the R Package Seurat.