A, Venn diagram representation of binding sites identified by CDX2 ChIP-seq in proliferating and mature Caco-2 cells. B, Histograms of normalized CDX2 ChIP-seq tags at representative sites bound exclusively early in proliferating (left) or late in differentiated cells (right). Additional examples appear in Figure S3. C, Condition-specific CDX2 binding corresponds to condition-specific active chromatin. H3K4Me2-marked putative cis-elements were ranked from most differentially active chromatin signatures in proliferating (left) to differentiated cells (right) and plotted along the x-axis in bins of 750. CDX2 occupancy was then plotted relative to occupancy in areas where active chromatin did not vary (center). CDX2 binding sites specific to proliferating (blue triangles) or differentiated cells (red squares) correlated with proliferating and differentiated cell-specific active chromatin, respectively. D, Heat maps demonstrating correlation of stage-specific CDX2 occupancy with stage-specific gene expression. Top: Early (left, blue) versus late (right, red) genes were ranked and binned in groups of 100; genes with unaltered expression are represented in the center. The color map (blue - red) presents the scale of log₂ fold change (−1 to +1) in gene expression levels in the two stages. Bottom: The frequency of condition-specific CDX2 binding within 100 kb of each gene expression bin is indicated by the intensity of yellow shading. Numbers below the color map correspond to the minimum, mean, and maximum ratios (0.5, 1, and 2.4, respectively) between the average number of CDX2 binding sites within 100kb from the genes in the same bin over the average number of binding sites near all genes. E, Gene Ontology term analysis reveals that expressed genes with condition-specific CDX2 binding match functions classically attributed to differentiated (red) or proliferating (blue) cells. Complete analysis is found in Table S2.

A, Gene Ontology terms associated with mature intestinal functions are notably enriched among transcripts reduced in Cdx2^fV/fV;Villin-Cre^ER(T2) intestinal mucosa. B, Examples of terminal digestive enzyme transcripts reduced in Cdx2-null duodenum, jejunum, and ileum and C, a representative example of binding data from Caco-2 cells showing differentiated cell-specific CDX2 occupancy (arrow). D, Top: Transcripts reduced or unchanged in Cdx2^fV/fV;Villin-Cre^ER(T2) mice were detected in triplicate samples and grouped into bins of 100 by a ratio relative to controls (−0.5 to 0). Bottom: Differentiated cell-specific CDX2 binding frequency within 100 kb of the corresponding orthologs in Caco-2 cells, indicated by intensity of yellow shading (a range of 0.5–2.4, similar to Fig. 2d). Binding is observed more frequently near loci with reduced expression, suggesting direct regulation of many targets and inter-species conservation. E, To determine the consequence of Cdx2 deletion on chromatin structure, oligonucleotides were designed to query the left, center, and right nucleosomes by PCR in MNase protection assays at nine CDX2-bound putative enhancers in control and tamoxifen-treated Cdx2^fV/fV;Villin-Cre^ER(T2) (KO) adult mice. MNase-protected DNA is shown relative to the central nucleosome at each region. In villi depleted of Cdx2, MNase protection was relatively enhanced at center nucleosomes, consistent with reappearance of nucleosomes and loss of active enhancer chromatin. Five of the 9 tested regions gave this pattern; 4 examples are shown. Results are depicted schematically on the right, together with 1 example where nucleosome protection was not significantly affected in KO mice. Reductions in transcript levels for each gene in the Cdx2-depleted tissue are indicated. All error bars are SEM.

A, Individual examples of ChIP-seq in proliferating (blue) and differentiated (red) cells show condition-specific GATA6 and CDX2 occupancy at the same region in proliferating but not in differentiated cells; HNF4A and CDX2 co-occupy a region bound selectively in differentiated cells. Condition-specific H3K4me2 signals are evident at the regions selectively occupied by the TFs and demarcated here by the dotted lines. Additional examples appear in Figure S7. B, Frequency of co-occupancy among TFs in proliferating and differentiated cells as a function of the distance between peak factor-binding sites. Condition-specific CDX2 binding sites more frequently lie near GATA6-bound regions in proliferating cells and near HNF4A-occupied regions in mature cells. The frequency of overlap in binding peaks within a 300bp distance (dotted line) is indicated in pie charts (right). C, Heat maps of all condition-specific CDX2-bound sites, showing the extent of GATA6 and HNF4A co-occupancy at these regions and confirming the relationship between these factors in binding across the genome. Sites are ordered by robustness of CDX2 occupancy in each state of differentiation. The measure of each TF’s occupancy was defined as the relative ChIP signal in relation to the input signal at binding sites (see Experimental Procedures for mathematical details) and distributed between −4 and 12 (color map scale).

A, Caco-2 cells show progenitor properties while proliferating (left); after reaching confluence, the cells differentiate to resemble mature intestinal villus enterocytes (right). Histograms depict EdU incorporation, a marker of DNA replication, 1 h after exposure, as measured by fluorescence intensity. Diagrams in the upper right represent cell morphologies. B, Chromatin modifications associated with nascent (left) and active (right) enhancers are depicted, the latter characterized by relative increase in H3K4Me2 on flanking nucleosomes and depletion of signal in the nucleosome in between. C, H3K4Me2 ChIP-seq results in proliferating and differentiated Caco-2 cells at a representative region on chromosome 10. D, H3K4Me2 ChIP-seq tag counts across indicated 1-kb intervals in proliferating (left, blue) and differentiated (right, red) cells, including high- resolution image of the data shown in C. The top 2 panels show gain of active chromatin structure upon differentiation; the lower panel shows a region selectively active earlier in proliferating cells. E, Composite plots of H3K4Me2 ChIP-seq tag counts from the 1,000 regions with the greatest differential chromatin structure in proliferating cells (left, blue) and the same 1,000 regions in differentiated cells (right, red), aligned at center nucleosomes. TF motifs most significantly over-represented at the valleys of the composite plots are shown (details in Experimental Procedures). A complete list of identified motifs is found in Table S1. F, The same analysis as in E, but on the 1,000 regions with the greatest differential chromatin structure in differentiated cells.

A, Immunostaining verifies absence of intestinal Cdx2 expression in tamoxifen-treated Cdx2^fV/fV;Villin-Cre^ER(T2) mice (also see Supplemental Fig. S5). B, Reduced mRNA levels of Cdx2 and candidate target genes Sis and Isx in tamoxifen-treated Cdx2^fV/fV;Villin-Cre^ER(T2) intestines; Cdx1 transcript levels are unaffected. C, Cdx2^fV/fV;Cre^ER(T2) mice rapidly lose weight after tamoxifen treatment. D, Intestines of Cdx2^fV/fV;Villin-Cre^ER(T2) mice are dilated and the cecum is particularly engorged with fluid. E, H&E staining reveals intact tissue architecture in the Cdx2 mutant ileum. F, A profound deficiency is evident in Alkaline Phosphatase activity, a marker of enterocytes, especially in the distal small intestine. G–I, Periodic acid-Schiff (G), Lysozyme (H) and ChromograninA (I) were used as markers of goblet, Paneth, and enteroendocrine cells, respectively, in the ileum. J, In contrast to the respective single gene mutants, Cdx1^−/−;Cdx2^fV/fV;Villin-Cre^ER(T2) intestinal crypts show markedly reduced expression of the proliferative markers Ki67 and Pcna. Quantitation of the Ki67⁺ fraction of crypt cells showed significant reduction in the double mutant (P<0.0001; two tailed t-test). Box plot indicate median Ki67⁺ counts per crypt, with lower and upper quartiles and whiskers indicating minimum and maximum counts. Error bars are SEM for B and C. Scale bars, 33.3 µm except for F, 200uM.

A, DNA motif analysis at condition-specific CDX2-binding regions shows, in addition to CDX2 motifs, substantial enrichment of GATA motifs in proliferating (left) and of HNF4A motifs in differentiated cells (right). B, Over the course of Caco-2 differentiation, HNF4A transcript levels increase and GATA6 levels decrease; known markers of maturation (CDH17) or proliferation (MYC) increase or decrease, respectively, as expected. Error bars are SEM. C–D, ChIP-seq for GATA6 and HNF4A in proliferating (blue) and differentiated (red) cells reveals differences in preferred binding contexts. Venn diagrams indicate the number of sites bound uniquely in proliferating or differentiated cells (p<1e⁻¹⁰). HNF4A binding sites are more numerous later in differentiation, whereas GATA6 binds predominantly early in proliferating cells. For each occupancy dataset, the most enriched de novo binding motif is indicated (all within the statistical limit of p<1e⁻³⁰,) and each is similar to empirically defined TRANSFAC motifs (as shown in Fig. 3A). Identified binding regions were highly conserved and mainly found far from promoters (Figure S6). E, As in figure 2C, TF occupancy at regions of differentially active chromatin is compared to binding frequency at static chromatin regions. Top, GATA6 binding is better associated with the active H3K4Me2 pattern in proliferating than in differentiated cells, whereas HNF4A occupancy (bottom) correlates better with active H3K4Me2 in differentiated cells. F, GATA6 and HNF4A co-immunoprecipitate with CDX2, but not with IgG controls, indicating their presence in CDX2 complexes.

In proliferating cells, CDX2 binds with relative selectivity, frequently accompanied by GATA6, at H3K4Me2 regions with selectively open chromatin at loci active in progenitor cells. Genes associated with epithelial maturity are not expressed in these cells and their enhancers have a relatively closed structure with less H3K4Me2. As cells differentiate, CDX2 relocates from progenitor-specific target genes to those associated with maturity, now accompanied by HNF4A; the appearance of an active H3K4Me2 pattern at many loci depends on CDX2.