SOX2 regulates common and specific stem cell features in the CNS and endoderm derived organs

Abstract

Stem cells are defined by their capacities to self-renew and generate progeny of multiple lineages. The transcription factor SOX2 has key roles in the regulation of stem cell characteristics, but whether SOX2 achieves these functions through similar mechanisms in distinct stem cell populations is not known. To address this question, we performed RNA-seq and SOX2 ChIP-seq on embryonic mouse cortex, spinal cord, stomach and lung/esophagus. We demonstrate that, although SOX2 binds a similar motif in the different cell types, its target regions are primarily cell-type-specific and enriched for the distinct binding motifs of appropriately expressed interacting co-factors. Furthermore, cell-type-specific SOX2 binding in endodermal and neural cells is most often found around genes specifically expressed in the corresponding tissue. Consistent with this, we demonstrate that SOX2 target regions can act as cis-regulatory modules capable of directing reporter expression to appropriate tissues in a zebrafish reporter assay. In contrast, SOX2 binding sites found in both endodermal and neural tissues are associated with genes regulating general stem cell features, such as proliferation. Notably, we provide evidence that SOX2 regulates proliferation through conserved mechanisms and target genes in both germ layers examined. Together, these findings demonstrate how SOX2 simultaneously regulates cell-type-specific, as well as core transcriptional programs in neural and endodermal stem cells.

Fig 1. Cell-type-specific binding of SOX2 in CNS and endoderm progenitors.

(A) E11.5 Sox2-Gfp knock-in embryo shows expression in the cortex and spinal cord, while inset highlights expression in dissected stomach and lung/esophagus. (B) Overlapping expression of SOX2 and Ki67 in cycling progenitors of the E11.5 cortex, spinal cord, stomach and lung. Grey areas in diagrams represent SOX2⁺ stem cells, while tan surroundings in stomach and lung represent non-endodermal derived mesenchymal cells. (C) Central enrichment of SOX2 binding motifs in cortex, spinal cord, stomach and lung/esophagus SOX2 ChIP-seq peaks. The percentage of peaks that the given motifs are found centrally enriched, and their p-values are listed. (D) Venn diagram showing overlap between sites bound in the different SOX2 ChIP-seq experiments. 32 peaks overlapped in all four tissues. (E) Representative tracks showing germ-layer-specific and common SOX2 ChIP-seq reads aligned around Foxa2, Neurog2 and Axin2 gene loci, with magnification of selected peak sites. Tracks from cortex are in red, spinal cord in green, stomach in blue and lung/esophagus in yellow, with read scale maximum labelled on y-axis. Inset black lines are centred on peaks. (F) Toppcluster Fruchterman-Reingold Network of the top ten GO terms enriched for genes bound by SOX2 in the cortex, spinal cord, stomach and lung/esophagus. (G) Diffbind PCA of SOX2 ChIP-seq duplicates in cortex, spinal cord, stomach and lung/esophagus, as well as published SOX2 ChIP-seqs in ESCs, ESC derived NPCs and adult stomach.

Fig 2. SOX2 utilizes distinct partner factors to drive cell-type-specific gene expression.

(A) HOMER de novo transcription factor binding motifs enriched in specific and common SOX2 ChIP-seq peak sets. Mammalian transcription factors with consensus sites matching the motif, p-values for motif enrichment and the percentage of peaks the motifs are found in are inset next to each motif. (B) RNA-seq RPKMs in cortex, spinal cord, stomach and lung/esophagus of transcription factors matching the motifs enriched in HOMER analysis from Fig 2A. (C) Co-immunoprecipitation of Flag-tagged transcription factors, identified in Fig 2A as enriched in cortex specific (OTX1), lung/esophagus specific (FOXA1) or common (ZEB1) SOX2 peaks, and Myc-tagged full-length SOX2 with and without DNase I treatment. (D) Luciferase assays in P19 cells of common (Hoxa2 +296bp and Tm7sf2 -918bp), CNS specific (Dnajc19 -23kb and Pax3 +29kb) or endoderm specific (Foxp1 +98kb and Foxa2 +98kb) SOX2 bound regions, with depictions of motif arrangements within each enhancer above each assay. Each DNA-region was tested for the effect of co-transfected SOX2 and the specific transcription factor with enriched binding motifs as identified in Fig 2A (OTX1 for neural, FOXA1 for endodermal and ZEB1 for common). P-values are calculated with two-sided, unpaired t-tests. (E) Enrichment scores for overlap between genes specifically expressed by Deseq2 analysis between organ RNA-seqs, and genes specifically bound by SOX2 within 500kb in cortex, spinal cord, stomach or lung/esophagus. P-values are calculated by a Yates Chi-squared test. (F) Fold enrichment and p-value scores from Panther of selected GO terms for genes within 500kb of each specific and common ChIP-seq peak set. * = p<0.05, ** = p<0.01, *** = p<0.001.

Fig 3. SOX2 bound CRMs possess cell-type-specific enhancer activity.

(A) Schematic of the zebrafish reporter system used to analyze enhancer expression. Sections of 50 hour post-fertilization zebrafish embryos from three different anterior-posterior levels (a’: midbrain, b’: pharyngesophageal endoderm, c’: spinal cord) following injection of the Mrps28+58kb enhancer. Expression of SOX is shown in red and GFP in green. (B-D) Three examples each of SOX2 ChIP-seq tracks with read scale maximum values inset top left (cortex in red, spinal cord in green, stomach in blue and lung/esophagus in yellow) and reporter expression from enhancers commonly (B) or specifically bound by SOX2 in CNS (C) and endoderm (D). Arrows point to GFP expression in endoderm. Black lines under tracks indicate called SOX2 peaks.

Fig 4. SOX2 represses proliferation in the developing spinal cord and stomach.

(A) Percentage of cells expressing high or low levels of SOX2 labelled by a one hour pulse of BrdU in the E11.5 mouse spinal cord. Dotted lines in insets surround area of greatest BrdU incorporation. (B) Average background normalized SOX2 expression level and percentage of cells labelled by a one hour pulse of BrdU in the E11.5, E13.5 and E15.5 anterior and posterior stomach. (C) Percentage of electroporated cells in the chick spinal cord labelled by a 30 minute pulse of BrdU following misexpression of GFP, SOX2 or dnSOXB1. (D) Percentage of electroporated cells in E13.5 stomach explants labelled by a 30 minute pulse of BrdU following overexpression of GFP, SOX2 or dnSOXB1. All error bars represent standard deviations between experiments and p-values are calculated with two sided, unpaired t-tests (* = p<0.05, ** = p<0.01, *** = p<0.001).