SALL3 expression balance underlies lineage biases in human induced pluripotent stem cell differentiation

Abstract

Clinical applications of human induced pluripotent stem cells (hiPSCs) are expected, but hiPSC lines vary in their differentiation propensity. For efficient selection of hiPSC lines suitable for differentiation into desired cell lineages, here we identify SALL3 as a marker to predict differentiation propensity. SALL3 expression in hiPSCs correlates positively with ectoderm differentiation capacity and negatively with mesoderm/endoderm differentiation capacity. Without affecting self-renewal of hiPSCs, SALL3 knockdown inhibits ectoderm differentiation and conversely enhances mesodermal/endodermal differentiation. Similarly, loss- and gain-of-function studies reveal that SALL3 inversely regulates the differentiation of hiPSCs into cardiomyocytes and neural cells. Mechanistically, SALL3 modulates DNMT3B function and DNA methyltransferase activity, and influences gene body methylation of Wnt signaling-related genes in hiPSCs. These findings suggest that SALL3 switches the differentiation propensity of hiPSCs toward distinct cell lineages by changing the epigenetic profile and serves as a marker for evaluating the hiPSC differentiation propensity.

Fig. 1

Profiles of hiPSC lines showing differentiation propensities. a Outline of workflow for identification of biomarkers capable of predicting the differentiation propensity of hiPSCs. b Hierarchical clustering of gene expression in ten hiPSC lines. We identified 3362 probes with significantly different expression levels among ten hiPSC lines. c Expression profiles for lineage marker genes were summarized using PCA. The number indicates PC1 of each lineage among the ten hiPSC lines. d The line graph represents the rank of the first PC score of each lineage among the ten hiPSC lines. ^*P < 0.05, Spearman’s rank correlation coefficients

Fig. 2

SALL3 regulates differentiation into the three germ layers. a Venn diagrams illustrate overlaps among differentiation propensity marker candidate genes. SALL3 was the only gene showing an inverse correlation between ectoderm and mesoderm/endoderm differentiation. All candidate genes of differentiation propensity markers were listed in Supplementary Data 3. b Microarray data of SALL3 expression in ten hiPSC lines (n = 6, biological replicates). c SALL3 knockdown was confirmed by qRT-PCR analysis (n = 3, biological replicates). ^*P < 0.01, two-sided t test. d Western blot analysis of the total extracts obtained from control and SALL3 knockdown cells. LSD1 and β-actin were used as a nuclear protein control and loading control, respectively. Molecular weight is indicated as Mr (k). e qRT-PCR analysis of undifferentiated hPSC markers, OCT3/4 and NANOG. Total RNA was isolated from 253G1 SALL3 shRNA cells and 253G1 control shRNA cells in the undifferentiated state (n = 3, biological replicates). f–h qRT-PCR analysis of three germ layer-specific genes in EBs derived from 253G1 SALL3 shRNA cells and 253G1 control shRNA cells (n = 6, biological replicates). Ectoderm marker genes (f), endoderm marker genes (g), and mesoderm marker genes (h) are shown. ^*P < 0.05, ^**P < 0.01, two-sided t test. Error bars represent mean ± SD

Fig. 3

SALL3 knockdown accelerates cardiac differentiation. a Schematic diagram of culture procedures for cardiomyocyte. b qRT-PCR analysis of cardiomyocyte markers GATA4, NKX2.5, and TNNT2. Total RNA was isolated from 253G1 SALL3 shRNA cell derived and 253G1 control shRNA cell-derived cardiomyocytes (n = 3, biological replicates). ^*P < 0.01, two-sided t test. c Flow cytometry analysis of TNNT2 in 253G1 SALL3 shRNA cell-derived (right) and 253G1 control shRNA cell-derived (left) cardiomyocytes. d SALL3 mRNA levels in undifferentiated 253G1 SALL3 overexpressing cells (EF1α-SALL3). n = 3, biological replicates. ^*P < 0.01, two-sided t test. e qRT-PCR analysis of cardiomyocyte markers GATA4, NKX2.5, and TNNT2. Total RNA was isolated from 253G1 EF1α-SALL3 cell-derived and 253G1 control vector cell-derived cardiomyocytes (n = 3, biological replicates). ^*P < 0.01, two-sided t test. Error bars represent mean ± SD

Fig. 4

SALL3 expression regulates neural differentiation. a Schematic diagram of culture procedures for neural cells. b, c qRT-PCR analysis of neural cell markers PAX6, SOX1, NES, and TH. Total RNA from 253G1 SALL3 shRNA cell-derived and 253G1 control shRNA cell-derived neural cells (b) and 253G1 EF1α-SALL3 cell-derived and 253G1 control vector cell-derived neural cells (c) was isolated (n = 3, biological replicates). ^*P < 0.05, ^**P < 0.01, two-sided t test. d Immunofluorescence staining of PAX6 and OCT3/4 in 253G1 cell-derived (left), 253G1 SALL3 shRNA cell-derived (center), and 253G1 EF1α-SALL3 cell-derived neural cells (right). Scale bars, 200 µm. Error bars represent mean ± SD

Fig. 5

SALL3 interacts with DNMT3B and modulates DNMT function. a Lysate prepared from 253G1 cells was immunoprecipitated with anti-SALL3 or normal rabbit IgG. The immunoblot was analyzed with anti-DNMT3A, DNMT3B, and DNMT1 antibodies. Molecular weight is indicated as Mr (k). b Lysate prepared from 253G1 cells was immunoprecipitated with anti-DNMT3B or normal sheep IgG. The immunoblot was analyzed with anti-SALL3 antibody. c In vitro DNMT activity assay. DNMT activity of the nuclear fractions prepared from 253G1 (WT) cells and the 253G1 SALL3^−/− cells was measured (n = 3, biological replicates). ^*P < 0.01, two-sided t test. d qRT-PCR analysis of neural cell markers PAX6 and SOX1. Total RNA was isolated from 253G1-derived (WT), 253G1 SALL3^−/−-derived and 253G1 SALL3^−/− DNMT3B^−/mut-derived neural cells (n = 6, biological replicates). ^*P < 0.01, one-way ANOVA with post hoc Tukey–Kramer test. e qRT-PCR analysis of cardiomyocyte markers GATA4, NKX2.5, and TNNT2. Total RNA was isolated from 253G1-derived (WT), 253G1 SALL3^−/−-derived and 253G1 SALL3^−/− DNMT3B^-/mut-derived cardiomyocytes (n = 6, biological replicates). ^*P < 0.01, one-way ANOVA with post hoc Tukey–Kramer test. Error bars represent mean ± SD

Fig. 6

SALL3 regulates gene-body DNA methylation. a HumanMethylation450K BeadChip analysis of 253G1 SALL3 shRNA cells and 253G1 control shRNA cells. Diagram shows a comparison of the similarly methylated and differentially methylated probes (left) and hypermethylated and hypomethylated probes in 253G1 SALL3 shRNA cells, compared to the control cells (right). b Distribution of hypermethylated regions. Top: schematic showing the seven categories. Bottom: pie chart showing the percent distribution of hypermethylated probes in each category. c HumanMethylation450K BeadChip analysis of the WNT3A locus (left) and WNT5A locus (right). Top: schematic of a gene locus on the chromosome. Middle: schematic of a CpG island locus in two genes depicted by the UCSC Genome Browser. Bottom: methylation score. Gray vertical bars highlight CpG island and shore regions (n = 3, biological replicates). Asterisks highlight probes with a significant difference in the methylation score between the control and SALL3 knockdown (P < 0.05, two-sided t test). d The ChIP-seq data of the WNT3A locus (left) and WNT5A locus (right) are depicted by the UCSC Genome Browser. Top track: SALL3 protein binding to genomic regions in 253G1 control shRNA cells. Middle and bottom tracks: DNMT3B protein binding to genomic regions in 253G1 control shRNA cells and 253G1 SALL3 shRNA cells