The lineage-specific transcription factor CDX2 navigates dynamic chromatin to control distinct stages of intestine development

Abstract

Lineage-restricted transcription factors, such as the intestine-specifying factor CDX2, often have dual requirements across developmental time. Embryonic loss of CDX2 triggers homeotic transformation of intestinal fate, whereas adult-onset loss compromises crucial physiological functions but preserves intestinal identity. It is unclear how such diverse requirements are executed across the developmental continuum. Using primary and engineered human tissues, mouse genetics, and a multi-omics approach, we demonstrate that divergent CDX2 loss-of-function phenotypes in embryonic versus adult intestines correspond to divergent CDX2 chromatin-binding profiles in embryonic versus adult stages. CDX2 binds and activates distinct target genes in developing versus adult mouse and human intestinal cells. We find that temporal shifts in chromatin accessibility correspond to these context-specific CDX2 activities. Thus, CDX2 is not sufficient to activate a mature intestinal program; rather, CDX2 responds to its environment, targeting stage-specific genes to contribute to either intestinal patterning or mature intestinal function. This study provides insights into the mechanisms through which lineage-specific regulatory factors achieve divergent functions over developmental time.

Keywords: Chromatin; Development; Intestine; Lineage-specifying; Patterning; Transcription factor.

Fig. 1.

CDX2 is required for human intestinal development. (A) Differentiation schema used to differentiate hESCs into human intestinal organoids (HIOs). ACTA, activin A treatment, 100 ng/ml; FGF4+CHIR, 500 ng/ml FGF4 + 2 µM CHIR99021; ENR, 100 ng/ml Egf, 5% Noggin-conditioned media and 5% R-SPONDIN2-conditioned media. (B,C) Immunoblot (B) and RNA-seq (E-MTAB-4168; Tsai et al., 2017) (C) show that CDX2 expression is only induced robustly after differentiation of ESCs into endoderm with activin and then induction of midgut specification via treatment with FGF4 (500 ng/ml) and the WNT agonist CHIR99021 (2 µM; Chir). HIO-E, HIO early, pre-transplant; HIO-L, HIO late, after transplant into mouse kidney capsule for further intestinal maturation. DE, definitive endoderm. (D,E) HIOs result from treating ESCs with this differentiation protocol, but ESCs lacking CDX2 as a result of CRISPR targeting (CDX2-KO, Fig. S1) do not give rise to HIOs, and instead give rise to organoids expressing genes characteristic of foregut lineages, as seen by immunofluorescence staining (D, representative of at least three independent experiments) and qRT-PCR (E) on organoids collected after 30 days in ENR (EGF, Noggin, RSPO) culture conditions, with each independent biological sample comprising three to five organoids pooled together. *P<0.05, **P<0.01, ****P<0.0001 (unpaired t-test). Scale bars: 200 µm.

Fig. 2.

CDX2 has distinct transcriptional targets in the developing midgut versus the adult intestine. Human icon indicates data from human tissues. (A,B) CDX2 ChIP-seq performed in in vitro differentiated midgut (4 day FGF4+CHIR-treated endoderm, derived from human ESCs, as described in Fig. 1A) and in primary adult human duodenum epithelium (collected from duodenal tissue adjacent to the bile duct from patients undergoing Whipple procedures) show distinct binding profiles, as defined by k-means clustering of the identified genomic regions interacting with CDX2 (A), and compared using average signal traces at all indicated sites (B). Data from two merged biological replicates each, see Fig. S2A. (C) Distinct binding sites correspond to different sets of nearby target genes, as indicated by gene ontology (GO) analysis using DAVID (genes with their transcriptional start sites within 5 kb of CDX2-binding sites were used for analysis). (D) DNA-motif enrichment analysis finds that CDX2-binding motifs are present in both categories of binding regions, but other transcription factor-binding motifs are differentially enriched in the midgut-enriched (58,981 sites) or adult-enriched (19,752 sites) CDX2-binding regions, as defined using HOMER. (E) Examples of stage-specific CDX2-binding sites are shown using IGV to visualize merged replicates of the ChIP-seq experiments. (F) Ectopic expression of CDX2 in endoderm, using a doxycycline-inducible hESC line (Endoderm+4 days doxycycline treatment), is not sufficient to recapitulate CDX2 binding to the majority of its normal midgut target sites (Endoderm+4 days CHIR+FGF). Approximately 52% of CDX2 midgut sites (MACS P<1e⁻¹⁰) were not detected by MACS in the doxycycline-inducible condition (second cluster of sites in heatmap), even at a lower peak-calling stringency (P<1e⁻³), and 48% of the midgut regions were bound by expression of CDX2 in endoderm (first cluster of sites). Lack of binding cannot be attributed to differences in CDX2-binding motifs, which were robustly detected in each category of binding sites (HOMER). Also see Fig. S2 and Table S2.

Fig. 3.

Temporal-specific shifts in CDX2 binding is conserved between mice and humans, and corresponds to temporal shifts in intestinal gene expression. Mouse icon indicates experiments performed in mouse tissues. (A) CDX2 ChIP-seq performed at the indicated stages of intestinal development on isolated epithelium show distinct, stage-dependent binding profiles (two biological replicates each, see Fig. S3; for E13.5, 12 embryos were pooled per replicate; for E16.5, three embryos pooled per replicate; for E17.5, two embryos pooled per replicate). The heatmap depicts CDX2 ChIP-seq intensity at each of the regions defined as embryo- or adult-enriched via k-means clustering of all binding sites identified in adult or embryonic conditions (Fig. S3). (B) The average ChIP-seq signals at the binding regions shown in A is plotted. (C) Embryo-enriched and adult-enriched binding sites, as identified in A, correspond to different sets of nearby target genes, which show stage-specific gene expression. RNA-seq was conducted on E12.5 or adult intestinal epithelium and genes linked to embryo-enriched or adult-enriched CDX2 binding (within 5 kb) were analyzed for their distribution along the E12.5-to-adult expression continuum using GSEA analysis. The GSEA plots indicate that embryo-enriched sites are nearby genes that are expressed at higher levels in the embryo than in the adult (leftward shift), whereas the opposite trend is observed for expression of genes nearby CDX2 sites that are stronger in the adult. CDX2-bound genes are also dependent upon CDX2 for expression, as revealed by RNA-seq analysis in Cdx2 knockout tissues (Fig. S3B). ES, enrichment score. (D) Embryo- and adult-enriched CDX2-binding regions are nearby genes with distinct gene ontologies, corresponding to embryo-specific or adult-specific functions, as indicated in GO term analysis. (E,F) Representative examples of embryo-enriched (E) or adult-enriched (F) CDX2-binding sites are shown using IGV with merged biological replicates. Also see Fig. S3.

Fig. 4.

Developmental stage dictates the consequences of CDX2 loss. (A) Analysis of regional tissue identity on samples from control, Shh-Cre; CDX2^f/f or Villin-Cre^ERT2; CDX2^f/f embryos, treated with tamoxifen by oral gavage [5 mg of tamoxifen (Sigma, T5648) dissolved in 0.5 ml of corn oil] to induce CDX2 loss as indicated, and collected at E18.5 from the indicated intestinal region. PAS stain marks mucosubstances, including intestinal Goblet cells, which are identified by a punctate staining and are characteristic in the intestinal epithelium (Control or Villin-Cre^ERT2), whereas broad PAS staining at the apical cell surface is indicative of gastric foveolar cells, as observed in the jejunum of Shh-Cre; Cdx2^f/f embryos. TRP63 marks esophageal cells and is not typically found in the intestine, but appears ectopically in the ileum of Shh-Cre; Cdx2^f/f embryos, but not in control embryos, or embryos induced to delete Cdx2 in the later stages. The parietal cell marker ATP4B is typically expressed in the glandular stomach, but is ectopically expressed in the CDX2-negative jejunum of Shh-Cre; Cdx2^f/f embryos, but not in embryos deficient in intestinal CDX2 at the later stages tested. Each immunostain or histological stain is representative of at least three experiments on different embryos. (B) Summary of CDX2 loss-of-function phenotypes from the data shown in A and previous studies. Temporal-specific CDX2 loss at stages earlier than E12.5 induces homeotic transformations, as observed by Gao and colleagues using the Foxa3-Cre driver (∼E8.5) (Gao et al., 2009), in this study using the Shh-Cre driver (∼E9.5), or at ∼E13.5 using the VIllin-Cre driver (Grainger et al., 2010). However, acute loss of CDX2 at later developmental stages (>E13.5) and in the adult compromises intestinal digestive functions, but intestinal identity is preserved (Villin-Cre^ERT2 with tamoxifen treatment at E13.5, E15.5 or adult), as documented in these studies in the embryo (Gao et al., 2009; Grainger et al., 2010), and previously in the adult (Verzi et al., 2011).

Fig. 5.

CDX2 temporally enriched binding patterns follow a temporally dynamic chromatin landscape that requires CDX2 for maintenance. (A,B) CDX2 transcripts (from RNA-seq on isolated epithelium; A) and protein levels (via immunostaining; B) are consistent across mouse developmental time, yet CDX2 occupies distinct sets of genomic regions at different developmental stages (Fig. 3). (C) CDX2 chromatin binding shifts may be due to a shifting chromatin landscape across developmental time, as indicated by the ATAC-seq assay for accessible chromatin in isolated intestinal epithelial cells. Regions shown are the CDX2 condition-enriched binding regions (as defined in Fig. 3A). (D) Average ATAC-seq signal is robust at embryo-enriched CDX2-binding sites prior to E16.5, after which chromatin accessibility at these regions diminishes. The converse is true for adult-enriched CDX2-binding sites. (E,F) Specific examples of condition-enriched chromatin accessibility at CDX2-binding sites, as visualized using IGV. CDX2-binding regions are also dependent upon CDX2 binding to remain accessible to the ATAC-seq assay, as seen in CDX2 knockout epithelium (Fig. S5).