Introduction

The gastrointestinal (GI) epithelium, a high turnover organ, contains multiple cell types to fulfill complex functions including food digestion, nutrient absorption, pathogen insulation and clearance, as well as endocrine roles (Chin et al., 2017; Kurokawa et al., 2020; Noah et al., 2011). Enterocytes (ECs) constitute ∼90% of all differentiated cells and function as highly efficient cells for absorbing nutrients and water (Snoeck et al., 2005). Goblet cells (GCs) produce mucus to lubricate the mucosal surface (Gustafsson & Johansson, 2022). Enteroendocrine cells (EECs) sense luminal nutrients and secrete peptide hormones to regulate appetite, intestinal motility, and insulin release (Gribble & Reimann, 2019). Tuft cells (TCs) are chemosensory cells that respond to various stimuli including infections and hypoxia (Schneider et al., 2019). The crypts of GI epithelium harbor various types of stem cells that continually replenish lost epithelial cells (Ramadan et al., 2022; van der Flier & Clevers, 2009). While the prevailing belief is that Lgr5-expressing crypt base columnar (CBC) cells are actively dividing stem cells and Bmi-expressing " +4 " cells (positioned four cells above the base of the crypt) are slowly cycling stem cells, it’s important to note that crypt cells have the capacity for fate reprogramming and de-differentiation to sustain regenerative capability (Barker et al., 2007; Beumer & Clevers, 2016; Metcalfe et al., 2014; Tian et al., 2011; Yan et al., 2012). Intestinal stem cells generate transit-amplifying cells (TACs), which undergo multiple cell divisions and differentiate into all intestinal lineages while migrating towards the crypt orifice (Krausova & Korinek, 2014). In addition to stem and progenitor cells, the base of small intestinal crypts contains Paneth cells (PCs) that secret antimicrobial peptides, called defensins, thereby contributing to host defense against microbes (Barreto et al., 2022; Lueschow & McElroy, 2020). PCs also contribute to the stem cell niche function, e.g. secreting Wnt ligands, to maintain epithelial homeostasis of the small intestine (Sato et al., 2011).

The behavior of crypt cells is precisely regulated by multiple signals, which can act sequentially, additively, synergistically, or antagonistically (Nakamura et al., 2007). Among them, the Wnt/β-catenin signal pathway represents the principal force governing intestinal epithelium homeostasis, particularly in the preservation of stem cell proliferation and multipotency (Clevers et al., 2014; Mah et al., 2016; Pinto et al., 2003). Upon Wnt ligand binding, the β-catenin degradation complex is inactivated, leading to accumulation of β-catenin and its translocation into the nucleus, where it works with the co-activator TCF-4 to turn on the transcription of a plethora of target genes, including c-Myc, CCND1, AXIN2, and CD44 (Clevers, 2006; Clevers & Nusse, 2012; Nusse & Clevers, 2017). Inhibition of Wnt ligand secretion reduced number of functional stem cells and led to a faster crypt fixation rate (Huels et al., 2018). In the process of aging, reduced Wnt signaling in stem cells and the niche of intestinal crypts causes inhibition of proliferation and reduced numbers of intestinal stem cells (Nalapareddy et al., 2017; Pentinmikko et al., 2019). The Wnt signal is also essential for commitment, maturation, and location of PCs and other secretory cells (Andreu et al., 2008; van Es et al., 2005). On the other hand, Wnt, EGF and Delta-like 1/4 ligands presented by PCs sustained CBCs’ proliferative states and the production (Sato et al., 2011). Although the Wnt signal dosage, largely via post-translational regulations, controls early lineage specification, hematopoiesis, and tumorigenesis (Bakker et al., 2013; Luis et al., 2011), it remains elusive how Wnt signaling dosage, particularly through the transcriptional control of Ctnnb1, regulates composition and function of intestinal epithelia.

Aberrant activation of canonical Wnt signaling is associated with GI cancers, most notably colorectal cancer (CRC) (Zhao et al., 2022; Zhu & Li, 2023). Germinal mutations of the APC gene, which encodes the negative regulator of β-catenin, underlie the Familial Adenomatous Polyposis (FAP), a hereditary neoplastic syndrome and the pre-cancerous condition for CRC (Cheng et al., 2019; J. Li et al., 2021). Moreover, APC is mutated in up to 80% of human sporadic CRCs (J. Li et al., 2021). Inactivation of APC causes enhanced accumulation of β-catenin in nuclei and subsequent activation of proliferative and pro-survival signals in CRC cells (Cheng et al., 2019). However, largely due to lack of specificity, no molecular therapeutic strategy targeting the pathway has been incorporated into CRC treatment. Interestingly, the transcriptional activity of CTNNB1, the coding gene for β-catenin, is also enhanced in CRCs (Terrin et al., 2017); however, the underlying mechanism is largely unknown. Dissecting it would facilitate the development of new Wnt/β- catenin targeting means to treat CRC.

Enhancers are essential cis-regulatory DNA elements that control spatiotemporal gene expressions during cell fate specification and maintenance (Pachano et al., 2022). Active enhancers are often associated with cell fate specifiers, mostly transcription factors (TFs), and are enriched with specific histone modification such as H3K27ac (acetylated lysine 27 of histone H3) and H3K4me1 (mono-methylated lysine 4) of histone H3 (Jindal & Farley, 2021). Importantly, variations of enhancer sequences often lead to aberrant gene silence and activation, thereby causing developmental anomalies and cancers (Kvon et al., 2021; Sengupta & George, 2017; Zabidi & Stark, 2016). Although numerous putative enhancers have been identified by high- throughput studies, only a fraction of them were functionally annotated.

The proper expression of essential fate genes in different cell types is often controlled by specific enhancers(Kvon et al., 2021; Pachano et al., 2022). We previously identified a neocortical enhancer of Ctnnb1, the coding gene of β- catenin. The enhancer – neCtnnb1, is specifically active in the subventricular zone of developing neocortex and regulates the abundance of intermediate progenitors and production of superficial-layer pyramidal neurons (Wang et al., 2022). Here we have identified an intestinal enhancer for Ctnnb1 and name it as ieCtnnb1. Reporter mice showed that ieCtnnb1 is predominantly active in the base of small intestinal crypts and across the epithelium of large intestine. Knockout of ieCtnnb1 slows down turnovers of intestinal epithelia at physiology and cancer conditions in a Wnt/β-catenin dependent manner. The human ieCTNNB1 contains a SNP that is associated with CTNNB1 expression levels in human gastrointestinal epithelia. The enhancer activity of ieCTNNB1 in CRC tissues is higher than that in adjacent normal tissues and positively correlates with CTNNB1 expression levels. HNF4α and phosphorylated CREB1 at serine 133 (p-S133-CREB1) were found to associate with ieCtnnb1 and regulate Ctnnb1 transcription.

Results

ieCtnnb1 is a gastrointestinal enhancer of Ctnnb1

We previously identified an evolutionarily conserved upstream enhancer for Ctnnb1 – neCtnnb1, which regulates neocortical neurogenesis (Figure 1 - figure supplement 1A). We reasoned that there might be other potential enhancer regions that control Ctnnb1 expression in a cell-specific manner. Therefore, we scanned upstream regions of Ctnnb1 for active enhancer properties – open chromatin with enriched H3K27ac and H3K4me1 – in different tissues using data retrieved from the ENCODE database. Interestingly, we noticed that a region that is 29,068 base pair (bp) upstream of the transcription starting site (TSS) for Ctnnb1 displayed typical enhancer feature in intestinal tissues and resides in the same topologically associating domain (TAD) with Ctnnb1 (Figure 1A, Figure 1 - figure supplement 2A). The region, named as ieCtnnb1 (intestinal enhancer of Ctnnb1), is at open chromatin status (DNase I hypersensitive site) and enriched with H3K27ac and H3K4me1, two histone marks for active enhancers, in 8-week-old mouse small intestine tissue (Figure 1A). Importantly, the enrichment of H3K27ac at ieCtnnb1 is also prominent in embryonic intestine and stomach tissue, as well as in adult kidney and liver (Figure 1 - figure supplement 1B and D). In contrast, the enrichment was not detected in non-gastrointestinal tissues such as embryonic and adult brain (Figure 1 - figure supplement 1D). In line with signatures of active enhancers, the enrichment of H3K4me3 and H3K36me3 at ieCtnnb1 is negligible in all examined tissues (Figure 1A, Figure 1 - figure supplement 1B and C).

Figure 1.

We next performed in vitro and in vivo analyses to validate ieCtnnb1’s enhancer activity. First, the 2,153 bp core region (CR), representing the H3K27ac peak region of ieCtnnb1, could drive the expression of luciferase reporter in multiple cell types (Figure 1 - figure supplement 2B). Second, the ieCtnnb1 reporter mice (H11^i.enh) were constructed by inserting the reporter cassette – 5’-Shh’s promoter-ieCtnnb1-CR-LacZ-iCre-3’ – into the H11 locus (Figure. 1B). Of note, the Shh promoter alone at the H11 locus did not drive the expression of reporter gene in all examined tissue (Kvon et al., 2020). Wholemount β-galactosidase (β-gal) staining of 8-week-old tissues revealed prominent ieCtnnb1-LacZ signals along the gastrointestinal (GI) tract. The signal exhibited a proximal-low to distal-high gradient, with the ileum, cecum, and colon showing the strongest signals (Figure. 1B, Figure 1 - figure supplement 2C). The stomach and proximal duodenum also displayed reporter activity driven by ieCtnnb1 (Figure. 1B, Figure 1 - figure supplement 2C). β-gal staining of intestinal sections revealed that the ieCtnnb1-LacZ signal could be visualized at the bottom of small intestinal crypts and in the entire large intestinal epithelia, but is absent in sections of liver, kidney, and spleen (Figure 1 - figure supplement 2D and data not shown). Interestingly, the proximal- low to distal-high gradient of ieCtnnb1-LacZ signal along the small intestine corresponds to the expression pattern of β-catenin (Figure 1 - figure supplement 2E).

The transcription activity of ieCtnnb1 is strongest at the bottom of crypts, where Lgr5-expressing CBC cells and Paneth cells (PCs) reside. To more precisely elucidate cell types in which ieCtnnb1 is active, we crossed H11^i.enhmice with Lgr5-EGFP-IRES-CreERT2 (Lgr5-EGFP) mice to obtain 8-week-old H11^i.enh;Lgr5-EGFP mice. Small intestinal crypts were respectively harvested from Lgr5-EGFP and H11^i.enh;Lgr5-EGFP mice followed by enrichment of both eGFP+ and eGFP- cells (Figure 1 - figure supplement 2F). Quantitative RT- PCR (RT-qPCR) analyses indicated that LacZ transcripts could be detected in both eGFP+ and eGFP- crypt cells in H11^i.enh;Lgr5-EGFP mice (Figure 1 - figure supplement 2F). Next, sections of H11^i.enh crypts were subjected with β-gal staining followed by immunohistochemistry of OLFM4 (a CBC marker) or lysozyme (a PC marker). Data showed X-gal signals are in close proximity with both OLFM4 and lysozyme (Figure. 1C), suggesting that the activity of ieCtnnb1 is present in both CBCs and PCs.

We further examined the ieCtnnb1 activity in GI tracts of postnatal day 7 (P7) pups, when GI epithelia are rapidly expanding. The ieCtnnb1-LacZ signals could be visualized in epithelia of P7 small and large intestines, aligning well with the distribution of canonical Wnt signaling illustrated by the BAT-Gal mice (Figure 1 - figure supplement 2G and H). Together, ieCtnnb1 drives transcription in embryonic intestinal epithelia, the base of small intestinal crypts and large intestinal epithelia of adult mice.

ieCtnnb1 maintains the expression of Ctnnb1 in intestinal epithelia

We next examined whether ieCtnnb1 controls the transcription of Ctnnb1 in intestinal epithelia. For this purpose, we employed CRISPR/Cas9-mediated gene editing to delete the genomic region containing ieCtnnb1 (Figure 1 - figure supplement 2I). Homozygous ieCtnnb1 knockout (Ctnnb1^Δi.enh) mice were born in Mendelian ratios and thrived through adulthood. Interestingly, the body weight of 8-week-old male Ctnnb1^Δi.enh mice was slightly but significantly lighter than that of wildtype controls (Figure. 1D and F), but the length of small and large intestines of Ctnnb1^Δi.enh mice was comparable to that of control mice (Figure. 1E, G and H). We then collected crypts from small and large intestine, finding that the expression levels of Ctnnb1 were decreased by 26.5% and 22.3% respectively in Ctnnb1^Δi.enh crypts (Figure. 1I and J). In contrast, the expression level of Ctnnb1 in the Ctnnb1^Δi.enh liver remained unaltered (Figure 1 - figure supplement 2J). In addition, the expression of genes located in the same TAD as Ctnnb1 and ieCtnnb1, including Ulk4, Rpl14 and Entpd3, was not changed in Ctnnb1^Δi.enh small intestinal crypts (Figure 1 - figure supplement 2K). These data suggest that ieCtnnb1 specifically maintains Ctnnb1’s transcription in intestinal epithelia.

In epithelial tissues, most β-Catenin is localized at cell adhesion sites to maintain barrier integrity, while a smaller fraction translocates into the nucleus to active transcription upon Wnt ligand binding. We therefore carried out nuclear/cytosol fractionation assays, which revealed a more substantial decrease in the level of nuclear β-Catenin compared to cytosolic β-Catenin (49.5% versus 29.8%) in small intestinal crypts of Ctnnb1^Δi.enh mice (Figure. 1K and L). RNA-seq transcriptome and RT-qPCR analyses revealed a significant reduction in the expression levels of components in the Wnt signaling pathway, as well as Wnt target genes, in Ctnnb1^Δi.enhcrypts (Figure. 1M and N). Together, the loss of ieCtnnb1 undermined the Wnt signaling in intestinal epithelia by reducing Ctnnb1 transcription.

Loss of ieCtnnb1 altered cellular composition and expression profiles of small intestinal crypts

The canonical Wnt/β-catenin signaling is essential for the construction and homeostasis of intestinal epithelia (Mah et al., 2016). Because ieCtnnb1 maintains the expression level of Ctnnb1 of intestinal crypts, we investigated how the deletion of ieCtnnb1 affects cellular components and functions of intestinal crypts. We performed single-cell sequencing on EpCAM+ epithelial cells from small intestinal crypts of 10-week-old Ctnnb1^Δi.enh and control (H11^i.enh) mice (n = 2 female mice for each genotype) (Figure. 2A, Figure 2 - figure supplement 1A) (Ayyaz et al., 2019). Uniform Manifold Approximation and Projection (UMAP) was employed for dimensionality reduction to cluster cells, then well-established marker genes were mapped onto the UMAP representation to designate cell types (Haber et al., 2017) (Figure. 2B, Figure 2 - figure supplement 1B-D). A small population of crypt cells could not be clearly defined probably due to certain technical issues during crypt extraction, antibody staining, or cell sorting (Figure 2 - figure supplement 1D). First, in accordance with bulk RNA-seq and RT-qPCR results, the expression levels of Ctnnb1 were downregulated in almost all Ctnnb1^Δi.enh crypt cells, and the expression of Axin2 was decreased in CBCs and TACs (Figure. 2C and D). Second, the proportion of PCs was decreased by 49.2% in Ctnnb1^Δi.enh crypts, which was confirmed by the immunohistochemical (IHC) staining of Defensin 5α, a PC marker (Figure. 2E and F). The ieCtnnb1-driving LacZ could be detected in the PC cluster (n = 4 10-week-old mice) (Figure 2 - figure supplement 1E). Accordingly, the expression levels of examined PC markers – Lyz1, Pla2g2a, Wnt3 and Math1 - were all significantly compromised in Ctnnb1^Δi.enh crypts (Figure. 2G). The number of goblet cells (GCs) was slightly (not statistically significant) reduced upon loss of ieCtnnb1 (Figure 2 - figure supplement 1F). Third, the proliferative capability of CSCs and TACs were compromised in Ctnnb1^Δi.enh crypts, as the proportion of G2/M cells and the expression of Mki67 were decreased in the two cell types (Figure. 2H-J). Consistently, the numbers of Ki67-positive cells and EdU-labeled cells (4 hours pulse) were decreased by 16.1% and 21.9%, respectively, in Ctnnb1^Δi.enh small intestinal epithelia (Figure. 2K and L). Fourth, in secretory cells (PCs GCs and EECs), as well as enterocytes/absorptive cells (mature enterocytes (mECs) and immature enterocytes (imECs)), and stem cells of Ctnnb1^Δi.enh crypts, the expression of genes related to secretory functions including Lyz1, Ang4, Muc2, and Mmp7 was decreased. In contrast, the expression of markers for absorptive cells such as Fabp6, Apoa1, Fabp2, and Reg3b was elevated (Figure. 2M and N, Figure 2 - figure supplement 1G). Accordingly, gene ontology (GO) analyses showed in stem cells and secretory cells of Ctnnb1^Δi.enh crypts, upregulated genes were enriched with terms associated with absorptive functions including ribosome biogenesis and oxidative phosphorylation; whereas downregulated genes in stem cells and secretory cells were respectively enriched with terms related to cell division and secretory functions (Figure 2 - figure supplement 1H and I). In contrast, in mECs, downregulated genes were enriched with terms associated with absorptive functions (Figure 2 - figure supplement 1J).

Figure 2.

In summary, the loss of ieCtnnb1 results in a reduction in the number of PCs, compromises stem cell proliferation, and disrupts the expression of genes related to secretory and absorptive functions. These effects could be attributed to a combination of reduced Wnt/β-catenin signaling and dysregulated crosstalk among crypt cells.

ieCtnnb1 deletion inhibits tumorigenesis of intestinal cancers by suppressing Wnt/β-catenin signaling

Tumorigenesis of most colorectal cancer (CRC) is associated with aberrant activation of the Wnt/β-catenin signaling (Zhu & Li, 2023). To investigate whether knockout of ieCtnnb1 could suppress the carcinogenesis of Wnt-driven intestinal tumors, we crossbred Ctnnb1^Δi.enh mice with Apc^Min/+ mice. In Apc^Min/+mice, a nonsense mutation of the Apc gene results in the translation of truncated and dysfunction APC protein, leading to failed degradation and nuclear translocation of β-catenin. Apc^Min/+ mice spontaneously develop intestinal tumors due to enhanced canonical Wnt/β-catenin signaling in the intestinal epithelia, particularly in stem cells (Su et al., 1992). The survival period of Ctnnb1^Δi.enh;Apc^Min/+mice (median survival - 255 days) was significantly longer than the Apc^Min/+ mice (median survival - 165 days), with the former having a heavier body weight (Figure. 3A and C). Strikingly, the Ctnnb1^Δi.enh;Apc^Min/+ mice barely grow colorectal cancers and have much fewer intestinal tumors compared to Apc^Min/+mice, with significantly smaller tumor area (Figure. 3B, D-F, Figure 3 - figure supplement 1A-C). IHC analyses showed that tumors in Ctnnb1^Δi.enh;Apc^Min/+ large and small intestines had much fewer Ki67+ proliferating cells, expressed significant less β-catenin, but contained more Muc2-expressing Goblet cells, suggesting Ctnnb1^Δi.enh;Apc^Min/+tumors behave like normal epithelia (Figure. 3G-L, Figure 3 - figure supplement 1D-I). Consistently, RNA-seq transcriptome experiments showed that Ctnnb1^Δi.enh;Apc^Min/+colorectal cancers significantly downregulated genes associated with the Wnt signaling pathway, pluripotency of stem cells, and gastric cancer; whereas increased expression of genes related to normal epithelial functions including oxidative phosphorylation, adhesion, protein digestion and absorption, and Mucin biosynthesis (Figure. 3M-O). Thus, ablation of ieCtnnb1 deters tumorigenesis of intestinal cancers by suppressing the Wnt signaling.

Figure 3.

Human ieCTNNB1 drives the expression of reporter gene in the gastrointestinal tract

We next investigated whether the human genome also harbors an intestinal enhancer for CTNNB1. A region with open chromatin (DNase I hypersensitivity) and enrichment of H3K27ac/H3K4me1 was identified at 38,489 bp upstream of human CTNNB1 in adult small intestine and colon tissues (Figure. 4A). Moreover, the H3K27ac enrichment is also present in esophagus but not in non-GI tissues (Figure 4 - figure supplement 1A). Similar to ieCtnnb1, the 3 kb-long human genomic region, named as ieCTNNB1, resides in the same TAD as the promoter of CTNNB1 (Figure 4 - figure supplement 1B). We generated ieCTNNB1 reporter mice (H11^hi.enh) and observed that ieCTNNB1 could drive LacZ expression in the bottom of small intestinal crypts and the entire epithelia of large intestine, also exhibiting a proximal-low to distal-high pattern (Figure. 4B). Notably, the LacZ signal driven by human ieCTNNB1 was less intense than that driven by mouse ieCtnnb1, potentially attributed to species-specific TF-enhancer interactions.

Figure 4.

The H3K27ac-enriched region of ieCTNNB1 was divided into five equal parts for the luciferase reporter assay. Experiments demonstrated that each of the five ieCTNNB1 subregions individually increased luciferase activity in both HCT116 and HeLa cells (Figure. 4C, Figure 4 - figure supplement 1C). We then conducted CRISPR mediated activation and interference in HCT-15 colorectal cancer cells using three guide RNAs (gRNAs) targeting ieCTNNB1, which respectively enhanced and inhibited transcription of CTNNB1 (Figure. 4D). Together, ieCTNNB1 is the intestinal enhancer of CTNNB1 for human GI epithelia.

ieCTNNB1 is activated in colorectal cancer and its activity positively correlates with the expression of CTNNB1

CRC, one of the most prevalent cancers in human, is often associated with aberrant activation of the Wnt signaling (J. Li et al., 2021). We then studied whether the occurrence of CRC is associated with enhancer activation of ieCTNNB1. We performed paired analyses of chromatin immunoprecipitation sequencing (ChIP-seq) and RNA-seq data derived from 68 CRC samples and their neighboring normal (native) tissue (Figure. 5A, Figure 5 - figure supplement 1A). As anticipated, the expression levels of CTNNB1 in CRC samples are significantly higher than those in neighboring native tissues (Figure. 5B). Essentially, the enrichment of H3K27ac at ieCTNNB1 in CRC tissues were significantly higher than that in native tissues (Figure. 5C Figure 5 - figure supplement 1A). In contrast, the enrichment of H3K4me3 at CTNNB1’s promoter was comparable between cancer and native tissues (Figure. 5C). Moreover, the enrichment of H3K27ac at ieCTNNB1 positively correlated with the expression of CTNNB1; but the enrichment of H3K4me3 at CTNNB1’s promoter was not associated with CTNNB1’s expression (Figure. 5D). Therefore, it’s the activity of enhancer but not the promoter that reflects the occurrence of CRC and expression levels of CTNNB1.

Figure 5.

By examining Genotype Tissue Expression (GTEx) eQTL data browser, a single-nucleotide polymorphism (SNP) – rs15981379 (C>T) within ieCTNNB1 was discovered (Figure. 5A). Importantly, the C>T variation is inversely correlated with expression levels of CTNNB1 in the esophagus and transverse colon (Figure. 5E). Luciferase reporter assays in HCT116 and HeLa cells revealed that the C>T variation indeed compromised luciferase activity driven by the ieCTNNB1-region 1 (Figure. 5F).

HNF4α and p-S133-CREB1 bind to ieCTNNB1 and maintains the expression of CTNNB1

Characterizing associated trans-factors of ieCtnnb1 and ieCTNNB1 is essential for revealing related transcriptional mechanisms and holds clinical relevance. By examining Cistrome Data Browser for ChIP-seq data, eight trans- factors were found to associate with ieCTNNB1 in human colorectal cancer cells or ieCtnnb1 in mouse GI tissues (Figure. 6A, Figure 6 - figure supplement 1A). Moreover, they are relatively highly expressed and/or have positive correlation with CTNNB1’s expression in colon cancer samples. We then transduced shRNAs-expressing plasmids into HCT-15 cells to individually down-regulate their expressions, revealing that decreasing the expression of CREB1 (cAMP responsive element binding protein 1) or HNF4A (hepatocyte nuclear factor 4 alpha) could compromise CTNNB1’s transcription (Figure. 6B, Figure 6 - figure supplement 1B). Of note, HNF4α plays crucial roles in the normal development of the colon and the maintenance of epithelial barrier (Cattin et al., 2009; Garrison et al., 2006). The phosphorylation of CREB1 at serine 133 (p-S133-CREB1) is essential for its role in activation of gene transcription (Carlezon et al., 2005). ChIP-qPCR experiments showed that enrichment of p-S133-CREB1 was more prominent than CREB1 at ieCtnnb1 in small intestinal crypts and at ieCTNNB1 in HCT-15 cells (data not shown). Notably, HNF4α and p-S133-CREB1 bind to multiple subregions of ieCTNNB1 and ieCtnnb1, as well as the promoters of CTNNB1 and Ctnnb1 (pCTNNB1 and pCtnnb1) in HCT-15 cells and small intestinal crypts (Figure. 6C and D, Figure 6 - figure supplement 1C and D). By analyzing the GEIPA-COAD (colon adenocarcinoma) and READ (rectum adenocarcinoma) database, expression levels of CREB1 and HNF4A were found to be significantly higher in colon and rectum cancer samples than those of normal tissues (Figure. 6E). Furthermore, expression levels of CREB1 and HNF4A positively correlated with that of CTNNB1 (Figure. 6F). HNF4α and p-S133-CREB1 were significantly enriched at the promoter of Ctnnb1 in Apc mutation induced colon cancer tissues (Figure. 6G and H). Intriguingly, knockout of ieCtnnb1 greatly diminished the association of HNF4α and p-S133-CREB1 with Ctnnb1’s promoter (Figure. 6G and H). In summary, HNF4α and p-S133-CREB1 were identified as essential trans-factors that bind to ieCTNNB1 and maintain CTNNB1’s expression.

Figure 6.

Discussion

The intestinal epithelia replenish themselves rapidly throughout lifetime, which enables regeneration while predisposes to GI cancers (Clevers, 2013). The optimal strength and dynamics of canonical Wnt/β-catenin signaling is central to both homeostasis and tumorigenesis of GI tracts (Beumer & Clevers, 2016; Pinto et al., 2003; Silvia Maretto, 2003). While the stabilization and nuclear translocation of β-catenin is the key step for Wnt signaling activation, the transcriptional control of Ctnnb1 could be another layer of regulation (Rim et al., 2022; Zhang & Wang, 2020). Here we unveiled an enhancer-dependent transcriptional machinery that tunes expressions of Ctnnb1 in multiple cell types of intestinal crypts, thereby balancing homeostasis and tumorigenesis of intestinal epithelia (Figure supplement 8).

Our findings reinforce the notion that the canonical Wnt signaling is required for the establishment of the secretory cell lineage (Pinto & Clevers, 2005; Pinto et al., 2003; van Es et al., 2005). Expression levels of markers for secretory cells were greatly reduced whereas levels of markers for absorptive cells were elevated in all small intestinal crypt cells of Ctnnb1^Δi.enh mice. Fate determinations of CBC daughter cells occur upon leaving the stem cell niche, exposed to changing WNT, Notch and EGF levels. Progenitors losing contact with PC Notch ligands upregulate DLL1 and DLL4, commit to the secretory lineage, and induce Notch activity in neighboring cells, promoting enterocyte differentiation (Beumer & Clevers, 2020; van Es et al., 2012).

WNT signaling not only directs PC-fated progenitors toward the crypt bottom but also directly induces their differentiation (Andreu et al., 2008; Beumer & Clevers, 2020; van Es et al., 2005). Single-cell sequencing and reporter gene analyses indicated that the activity of ieCtnnb1 is present in PCs. Knockout of ieCtnnb1 decreases number of PCs by almost 50% but the pool size of CBCs was not significantly altered. A previous study showed that genetic removal of PCs leads to the concomitant loss of Lgr5+ stem cells (Sato et al., 2011), while other studies indicated that PCs are not mandatory niche cells for the intestinal epithelia (Durand et al., 2012; Garabedian et al., 1997; Kim et al., 2012). Although Wnt ligands provided by PCs may support Lgr5+ CBCs³², essential Wnt ligands are primarily supplied by subepithelial stromal populations (Degirmenci et al., 2018; Shoshkes-Carmel et al., 2018). The presence of redundancy at the stem cell niche enhances the accuracy in maintaining intestinal homeostasis.

ieCtnnb1 knockout compromises Wnt signaling dosage, leading to PC loss, dysregulation of genes controlling absorptive and secretory functions of intestinal epithelia, and reduced proliferation of stem cells. These combined effects could account for the lighter body weight observed in adult Ctnnb1^Δi.enhmice.

On the other hand, colon and small intestinal tumors were rarely observed in Ctnnb1^Δi.enh;Apc^Min/+ mice. Among the few Ctnnb1^Δi.enh tumors analyzed, they exhibited lower Ctnnb1 expression, reduced proliferation, and behaved more like normal intestinal epithelia. Different truncations or mutations of APC lead to specific tumor types throughout the body in a Wnt/β-catenin dosage- dependent manner. For example, although Apc^1638N;Ctnnb1^+/-mice were less prone to GI cancers, Apc^1638N;Ctnnb1^+/-female animals tended to grow breast cancers (Bakker et al., 2013). Therefore, the intestinal specificity of ieCtnnb1 makes it an ideal target to lower the strength of Wnt signaling in treating colorectal cancer without affecting other tissues (Hamdan & Johnsen, 2019; Orouji, 2022).

Both mouse ieCtnnb1 and human ieCTNNB1 were able to drive reporter gene expression at the base of small intestinal crypts and throughout the epithelium of the large intestine. Very intriguingly, unlike the highly conserved neCtnnb1 sequence among amniotes, mouse ieCtnnb1 and human ieCTNNB1 do not share significant homology in primary sequence (Wang et al., 2022). Nonetheless, both ieCtnnb1 and ieCTNNB1 localize at the intergenic region 30- 40 kb upstream of Ctnnb1’s promoter. Moreover, both ieCtnnb1 and ieCTNNB1 contain binding motifs for HNF4α and CREB1. HNF4α and p-S133-CREB1 were found to associate with ieCtnnb1 and Ctnnb1’s promoter to maintain Ctnnb1’s expression. These findings support the notion that conserved TF binding motifs stabilize and preserve enhancer functionality over evolution (Wong et al., 2020; Zabidi & Stark, 2016). Future studies could dissect how sequence variations of ieCtnnb1 are adapted to intestinal homeostasis of different species.

Methods

Mice and genotyping

All animal procedures were approved by the Animal Care and Ethical Committee of Medical Research Institute at Wuhan University. C57BL/6 back- ground mice were housed in a temperature- and humidity- controlled environment with 12hr light/dark cycle and food/water ad libitum. Ctnnb1^Δi.enhmice were generated in the Beijing Biocytogen. H11^i.enh reporter mice were generated in the Shanghai Model Organisms Center Inc. BAT-Gal mice were provided by Dr. Jun-Lei Chang (Jackson Lab, stock number 005317). Lgr5- EGFP-IRES-CreERT2 mice were gifts from Dr. Ye-Guang Chen (Jackson Lab, stock number 008875). APC^Min/+mice were gifts from Dr. Bo Zhong (Jackson Lab, stock number 002020). The primer set forward 5′-atcctctgcatggtcaggtc- 3′/reverse 5′-cgtggcctgattcattcc-3′ was used for genotyping of BAT-Gal mice, H11^i.enh mice and H11^hi.enh mice with a band size of 315 base pairs (bp). The primer set forward 5′-gtcctgtccgtcactattatcctggc-3′/reverse5′- ccactgccctgctaaagcattggt-3′/reverse 5′-tgattagttctccgggaagcccagt-3′ were used for genotyping of Ctnnb1^Δi.enh mice and band sizes for Ctnnb1^Δi.enhmice are 482 bp (wild-type allele) and 867 bp (ieCtnnb1 knockout allele). The primer set forward 5′-atctcatggcaaacagacct-3′/reverse5′-tcacaaatcatctcgcaga-3′ was used for genotyping of APC^Min/+ mice with a band size of 340 bp. The primer set forward 5′-ctgctctctgctcccagtct-3′/reverse5′-ataccccatcccttttgagc-3′/reverse 5′-gaacttcagggtcagcttgc-3′ were used for genotyping of Lgr5-EGFP-IRES- creERT2 mice and the band sizes for 386 bp (wild-type allele) and 119 bp (mutant allele).

Cell lines

HEK293T cells were gifts from Dr. Hong-Bing Shu (Wuhan University). HCT116, HCT-15 and HeLa cells were purchased from the China Center for Type Culture Collection. HEK293T, HeLa and HCT116 cells were grown in DMEM (Gibco) supplemented with 10% fetal bovine serum (FBS) (Life Technologies) and 1×penicillin/streptomycin (Gibco). HCT-15 cells were maintained in RPMI-1640 (Gibco) containing 10% FBS (Life Technologies) and 1×penicillin/streptomycin (Gibco). For routine culturing, cells were dissociated using 0.25% trypsin (Gibco).

Tissue fixation and section

Intestinal and tumor tissues were dissected and cleaned in pre-cooled phosphate buffered saline (PBS) and immersed in 4% paraformaldehyde (PFA) overnight at 4°C. For sectioning, intestines were embedded in Tissue-Tek® O.C.T. Compound (SAKURA) and cut to 8 μm thickness with a cryostat (Leica CM1950).

CRISPR/dCas9-mediated transcription activation (CRISPRa) and interference (CRISPRi) assay

CRISPRa and CRISPRi assay were performed as previously described (Wang et al., 2022). In brief, SgRNAs were designed using online tool (https://zlab.bio/guide-design-resources) and were listed in Table S2. SgRNAs were cloned into lenti sgRNA (MS2)_zeo backbone vector (Addgene, #61427). For CRISPRi assays, pHR-SFFV-dCas9-BFP-KRAB (Addgene, #46911) is used. For CRISPRa assays, lenti MS2-P65-HSF1_Hygro (Addgene, #61426) and dCAS9-VP64_GFP (Addgene, #61422) were used. To obtain lentiviral particles, HEK293T cells (5 × 10⁶ cells per 10-cm dish) were transiently transfected with 12 μg CRISPRa or CRISPRi constructs, 6 μg psPAX2 and 6 μg pMD2.G. The supernatant containing lentivirus particles was harvested at 72 hours after transfection and filtered through Millex-GP Filter Unit (0.45 μm pore size, Millipore). To obtain high virus titer, the supernatant was centrifuged at 10,000 × g for 100 minutes and concentrate to 100 μl. On the day before infection, HCT-15 cells (2 × 10⁵ cells per well) were inoculated in 12 well plates. 10 μl high titer viruses were added to each well. 72 hours after infection, cells were harvested for RT-qPCR analysis.

RNA isolation and cDNA synthesis

RNAiso Plus (TAKARA) is used for RNA isolation. 1 ml RNAiso Plus was added to DNase/RNase free EP tube to lyse tissues or cells, followed by adding one- fifth volume of chloroform to achieve phase separation. After vigorous shaking, centrifuge at 12,000 × g for 15 minutes at 4°C. The upper aqueous phase was transferred to a new DNase/RNase free EP tube and equal volume of isopropanol was added to precipitate RNA. Precipitation was dissolved in DNase/RNase-free water. Complementary DNAs (cDNAs) were synthesized by HiScript® II Q RT SuperMix for qPCR kit (Vazyme; R222-01).

Real-time quantitative reverse transcription PCR (RT-qPCR)

cDNA was subjected to RT-qPCR using the SYBR Green assay with 2× SYBR Green qPCR master mix (Bimake). PCR reactions were performed on a CFX Connect Real-Time PCR Detection System (Bio-rad) under the following condition: 5 minutes at 95°C, 40 cycles at 95°C for 15 seconds, 60°C for 20 seconds. The relative transcript level of each gene was normalized to Gapdh (mice) or ACTB (human) and calculated according to the 2^-ΔΔCt method. All primers used are shown in the Table S2.

Isolation of murine intestinal crypt cells

Murine small intestines were harvested and cut open longitudinally. After 3 times wash in cold PBS with Penicillin (200 U/ml) + Streptomycin (200 μg/ml) (Invitrogen), small intestines were cut into 5 cm segments and incubated in pre-cold PBS with 8 mM EDTA at 4°C for 2 hours. Then intestinal tissues were transferred into a centrifuge tube filled with new cold PBS. Shake vigorously for5 minutes and collect intestinal epithelial cells (villi plus crypts) from the supernatant. Centrifuge the supernatant to collect all epithelial cells or pass through 70 μm cell filter (Corning) to filter out villi. Crypts were spun down in a centrifuge at 300 × g for 5 minutes.

Enriching Lgr5-EGFP⁺ cells

Crypts were collected from Lgr5-EGFP and H11^i.enh;Lgr5-EGFP mice, followed by resuspension and single-cell digestion in DMEM containing 0.25% trypsin (Gibco) at 37°C for 5 minutes. Disaggregated cells were passed through 40 μm cell strainers (Corning). Lgr5-EGFP⁺ and Lgr5-EGFP^- cells were collected using a BD Arial III cell sorter.

X-Gal staining

X-Gal staining is performed as described (Wang et al., 2022). In short, frozen slides were placed in fixed buffer (0.2% PFA, 0.1 M PIPES buffer (pH 6.9), 2 mM MgCl2, 5 mM EGTA) for 10 minutes, followed by rinse for two times with PBS containing 2 mM MgCl2. Then slides were incubated with detergent solution (0.1 M PBS (pH 7.3), 2 mM MgCl2, 0.01% sodium deoxycholate, 0.02% Nonidet P-40) for 10 minutes. Slides were incubated with freshly prepared and filtered X-Gal staining solution (0.1 M PBS (pH 7.3), 2 mM MgCl2, 0.01% sodium-deoxycholate, 0.02% Nonidet P-40, 5 mM K3Fe(CN)6, 5 mM K4Fe(CN)6·3H2O and 1 mg/ml X-Gal) overnight in the dark at 37°C. For whole gastrointestinal tissues, tissues were fixed in freshly prepared 4% PFA for 10 minutes, followed by rinse with detergent solution for 5 minutes by three times. Tissues were incubated with freshly prepared and filtered X-Gal staining solution at 37°C for a few hours.

EdU pulse-labeling experiment

Mice were intraperitoneally injected with EdU (10 mg/kg of body weight). After 4 hours, harvest the intestines for frozen sectioning. Staining was performed using BeyoClick^TM EdU Cell Promotion Kit (C0075S, Beyotime, China) according to the manufacturer’s manual.

ChIP-qPCR assay

Cells were cross-linked with 1% formaldehyde for 10 minutes at room temperature and then quenched with 0.125 M glycine for 5 minutes and washed twice with cold PBS. Cells were resuspended with 500 μl lysis buffer (50 mM Tris-HCl, pH 8.0, 0.5% SDS, 5 mM EDTA) and subjected to sonication on ice with an output power of 100 W, 12 minutes, 0.5 seconds on, 0.5 seconds off. After centrifugation, 1% of the supernatant was taken out as input. The rest of the sonicated lysates was diluted into 0.1% SDS and divided into several equal parts. Immunoprecipitation was performed overnight at 4°C on a rotating wheel with sheared chromatin, protein G agarose beads and indicated antibodies: 2 μg HNF4α (Abcam, ab181604), 2 μg p-S133-CREB1 (Abcam, ab32096), 2 μg IgG antibody (ABclonal, AC005). The next day, beads were cleaned three times with Wash Buffer I (20 mM Tris-HCl, pH 8.0; 1% Triton X-100; 2 mM EDTA; 150 mM NaCl; 0.1% SDS), Wash Buffer II (20 mM Tris-HCl, pH 8.0; 1% Triton X- 100; 2 mM EDTA; 500 mM NaCl; 0.1% SDS), Wash Buffer III (10 mM Tris-HCl, pH 8.0; 1 mM EDTA; 0.25 M LiCl; 1% NP-40; 1% deoxycholate) and TE buffer. Protein-DNA complexes were de-crosslinked in 120 μl elution buffer (0.1 M NaHCO3, 1% SDS, 20 µg/ml proteinase K) and shaken at 65°C overnight. DNA purification kit (TIANGEN) was used to extract DNA fragments for qPCR.

Luciferase reporter assay

The luciferase reporter assay was performed as described previously (Li et al., 2017). Briefly, the core region of ieCtnnb1, 5 regions of ieCTNNB1 and the region containing the mutation site were respectively cloned into pGL3- promoter vector (Promega). HEK293T, HeLa and HCT116 cells were transfected with pGL3-promoter vector and the plasmids mentioned above. Renilla basic vector was co-transfected as a control for normalization of luciferase activity. After 48 hours of transfection, Promega Dual Glo test kit (Promega, E1910) was used to measure luciferase activity according to manufacturer’s instructions.

Immunohistochemical staining

Frozen sections were dried at 37°C, fixed with 4% PFA for 10 minutes, and then washed three times with PBS. Antigen retrieval was performed using citrate buffer (pH 6, epitope retrieval solution) at 95°C for 20 minutes. Slides were then placed in 3% H2O2 solution for a few hours to remove endogenous catalase. After washing with PBS three times, slides were blocked with 10% normal goat serum for 1 hour, then incubated with the following primary antibodies overnight at 4°C: rabbit anti-α-defensin5 (Abcam, ab80515, 1:500), rabbit anti-Ki67 (Abcam, ab15580, 1:500), rabbit anti-β-catenin (Cell Signaling Technology, 8480T, 1:100), rabbit anti-Muc2 (Thermo Fisher Scientific, PA5-21329, 1:500). The next day, slides were incubated with biotin labeled secondary antibodies at room temperature for 1 hour. Then, slides were incubated with the avidin- biotin-peroxidase complex (VECTASTAIN Elite ABC system, Vector Labs). Finally, colorization was performed in the reaction solution (Tris-HCl (pH 7.2), 5 mg/ml 3,3’-diaminobenzidine, 0.075% H2O2).

H&E staining

H&E staining kit (Beyotime Biotech, C0105S) was used for experiments. Frozen slides were dried at 65°C, then dripped with hematoxylin staining solution for 5 minutes. Excess staining solution was rinsed off with water. Slides were immersed in distilled water for 5 minutes, then rinsed with 95% ethanol, and stained with eosin solution for 30 seconds. Subsequently, slides were put into anhydrous ethanol for 2 minutes, and cleared with xylene for 5 minutes. After air dry, slides were sealed with neutral resin.

Immunoblotting analysis

Crypt cells were lysed by 1× SDS loading buffer at 95°C for 10 minutes. After short centrifugation, the supernatants were loaded onto 10% SDS- polyacrylamide gel electrophoresis gels. Then, proteins were transferred onto polyvinylidene fluoride membrane (Millipore). Membranes were blocked with 10% non-fat milk in tris-buffered saline containing 0.3% Tween 20 (TBST; pH 7.4) for 1 hour at room temperature. Immunoblotting was carried out used primary antibody: rabbit anti-β-catenin (Cell Signaling Technology, 8480T, 1:100); mouse anti-GAPDH (ABclonal, AC001, 1:1000); rabbit anti-H3 (ABclonal, A2348, 1:10,000). After three times of washing with TBST, the membrane was incubated with anti-rabbit immunoglobulin G (IgG)–conjugated horseradish peroxidase secondary for 1 hour at room temperature. Signals were detected using the ECL substrate (Thermo Fisher Scientific).

Alcian Blue PAS Staining

Alcian Blue PAS Stain Kit (Abcam, ab245876) was used for experiments. all materials and prepared reagents were equilibrated to room temperature and gently agitated prior to use. Tissue sections were applied with acetic acid solution (3%) for 2 minutes. Then alcian blue solution (pH 2.5) was applied for 15-20 minutes. Slides were rinsed with running tap water for 2 minutes followed by 2 changes of distilled water. Periodic acid solution was applied to tissue sections for 5 minutes. After slides were rinsed with two changes of distilled water, Schiff’s solution was applied to tissue sections for 10-20 minutes. Slides were rinsed with warm running tap water for 2 minutes followed by 2 changes of distilled water. Tissue sections were applied with hematoxylin for 2 minutes. Sections were rinsed in running tap water for 2 minutes followed by 2 changes of distilled water. Slides were dehydrated through graded alcohols and sealed with synthetic resin.

Cell Sorting

Crypt single cells were prepared as described above. Cells were resuspended using 1 ml PBS and stained with 5 μl anti-EpCAM-FITC (eBioScience, 11-5791- 82) at 4°C for 30 minutes. After washing with PBS three times, cells were resuspended using 1 M FACS buffer (PBS with 1%BSA) containing 0.2 μg/ml DAPI (BD Pharmingen, Cat. No. 564907). Cells were filtered with nylon membrane with a 40 μm pore size. EpCAM positive cells were sorted and collected based on fluorescence signals for subsequent single cell sequencing. Flow cytometry data were acquired on a BD Arial III flow cytometer and analyzed with FlowJo (version 10.6.2).

Bulk RNA sequencing

RNA-seq and data analyses were performed at Novogene Co., Ltd (Beijing, China). Crypt cells or tumor tissues were harvested and total RNAs were extracted. RNA integrity was assessed using the RNA Nano 6000 Assay Kit of the Bioanalyzer 2100 system (Agilent Technologies, CA, USA). RNA-seq library construction was performed as described previously (Xu et al., 2021). Briefly, first strand cDNA was synthesized using random hexamer primer and M-MuLV Reverse Transcriptase (RNase H-). Second strand cDNA synthesis was subsequently performed using DNA Polymerase I and RNase H. Then, the library fragments were purified with AMPure XP system (Beckman Coulter, Beverly, USA). Then PCR was performed with Phusion High-Fidelity DNA polymerase, Universal PCR primers and Index (X) Primer. At last, PCR products were purified (AMPure XP system). The clustering of the index-coded samples was performed on a cBot Cluster Generation System using TruSeq PE Cluster Kit v3-cBot-HS (Illumia) according to the manufacturer’s instructions. After cluster generation, the library preparations were sequenced on an Illumina Novaseq platform and 150 bp paired-end reads were generated.

Bulk RNA-seq data processing

For data analysis, raw data of fastq format were firstly processed through in- house perl scripts. In this step, clean data were obtained by removing reads containing adapter, reads containing ploy-N and low-quality reads from raw data. Index of the reference genome was built using Hisat2 v2.0.5 and paired- end clean reads were aligned to the reference genome using Hisat2 v2.0.5. featureCounts v1.5.0-p3 was used to count reads numbers mapped to each gene. Differential expression analysis of two groups was performed using the DESeq2 R package (1.20.0). The resulting P-values were adjusted using the Benjamini and Hochberg’s approach for controlling the false discovery rate.

Genes with an adjusted P-value <0.05 found by DESeq2 were assigned as differentially expressed. Gene Ontology (GO) enrichment analysis of differentially expressed genes was implemented by the clusterProfiler R package, in which gene length bias was corrected. clusterProfiler R package was also used to test the statistical enrichment of differential expression genes in KEGG pathways. Local version of the GSEA analysis tool (http://www.broadinstitute.org/gsea/index.jsp), GO, KEGG, Reactome, DO and DisGeNET data sets were used for GSEA independently.

Single-Cell RNA sequencing and data analysis

ScRNA-seq and data analysis were performed at Novogene Co., Ltd (Beijing, China). The protoplast suspension was loaded into Chromium microfluidic chips with 30 chemistry (v2 or v3) and barcoded with a 10× Chromium Controller (10X Genomics). RNA from the barcoded cells was subsequently reverse-transcribed and sequencing libraries constructed with reagents from a Chromium Single Cell 30 v2 reagent kit (10X Genomics) according to the manufacturer’s instructions. Sequencing was performed with Illumina HiSeq 2000 according to the manufacturer’s instructions. For data analysis, we use FastQC to perform basic statistics on the quality of the raw reads. Raw reads were demultiplexed and mapped to the reference genome by 10X Genomics Cell Ranger pipeline using default parameters. All downstream single-cell analyses were performed using Cell Ranger and Seurat unless mentioned specifically. In brief, for each gene and each cell barcode (filtered by CellRanger), unique molecule identifiers were counted to construct digital expression matrices. Secondary filtration by Seurat: a gene with expression in more than 3 cells was considered as expressed, and each cell was required to have at least 200 expressed genes.

Hi-C data analysis

Hi-C data analysis completed by Frasergen Medicine Co., Ltd (Wuhan, China). Hi-C data of BALB/c mice large intestinal epithelial cells were provided by Dr. Gen Zheng(Zheng et al., 2023). The Hi-C data of SW480 cells was obtained from deposited data(Orouji, 2022). According to the reported method, the dynamic changes of TAD structures can be explored by calculating the relative signal values of TAD boundaries. TAD boundaries that are common among individuals and are not within the 2 Mb range of chromosomal ends were included, followed by filtering out boundaries with interactions less than 10 in the 2 Mb upstream and downstream regions. The ratio of the interaction between the boundary and its 2 Mb upstream and downstream regions were calculated using log2 to obtain the signal value of the TAD boundary, then the minimum value was subtracted from it to finally obtain the distribution of the relative signal value of each sample TAD boundary.

ChIP-seq data analysis

The human colorectal cancer patient RNA-seq FPKM files and ChIP-seq data were from GEO dataset GSE156614 (Q. L. Li et al., 2021). All ChIP-seq raw fastq data were removed of adaptor sequence. Cutadapt (version 1.16) was used with the parameters -u 3 -u −10 -U 3 -U -10 -m 30. Cleaned reads were aligned to the human reference genome (hg19) using bowtie2 (version 2.4.4) with default settings. The CrossMap (0.6.5) package was used as a conversion between different versions of the human genome (hg19 and hg38). We calculated the normalized RPM (reads per million mapped reads) of ChIP-seq signal at ieCTNNB1 and CTNNB1 promoter. Briefly, reads aligned to the specified region were calculated with bedtools multicov command. These reads were normalized with RPM. RNA-seq data FPKM (Reads per kilobase per million mapped reads) were converted to TPM (Transcript per Kilobase per Million mapped reads) using R version 4.2.3.

Knockdown experiments

Target sequences for shRNAs were designed on the Merck Life Sciences website (https://www.sigmaaldrich.cn) and were listed in Table S2. Oligonucleotides for shRNAs were cloned into pLKO.1 vector (Addgene, #21297). The day before transfection, HEK293T cells (2 × 10⁵ cells per well) were seeded in 12-well plates. pLKO.1 constructs, psPAX2 and pMD2.G were co-transfected into HEK293T cells. After 72 hours, supernatant containing virus particles was harvested and passed through Millex-GP Filter Unit (0.45 μm pore size, Millipore). Viral particles were aliquoted and stored at −80°C. On the day before infection, HCT-15 cells (2 × 10⁵ cells per well) were seeded in 12-well plates. 20 μl virus suspension were added to each well. After 72 hours, cells were harvested for RT-qPCR to evalutate knockdown efficiency.

Reference database

ChIP-seq signals for H3K27ac, H3K4me3, H3K4me1, H3K36me3, and DNase I hypersensitivity data of stomach, intestines, liver, and other tissues at different stages were downloaded from the ENCODE (https://www.encodeproject.org/) and their identifiers are listed in Table S1. In addition, ChIP-seq data for listed transcription factors of cells or tissues were obtained from Cistrome Data Browser (http://cistrome.org). Expressions of HNF4A and CREB1 in human colon (COAD) and rectal (READ) cancers were obtained from GEPIA (gepia.cancer-pku.cn).

eQTL analysis

The GTEX database (https://www.gtexportal.org/home/) was used to search SNPs at CTNNB1 and the 1 MB upstream and downstream region, followed by eQTL analysis. A variant, rs115981379 (C>T), was identified inside CTNNB1. The expression levels of CTNNB1 in 428 C/C type and 37 C/T type esophagus samples, and 343 C/C type and 25 C/T type transverse colon samples were also obtained in the GTEX database.

Analyses of HNF4α and CREB1 binding motifs

The JASPAR database (https://jaspar.elixir.no/) was used to analyze binding motifs of HNF4α and CREB1 at ieCtnnb1/ieCTNNB1 and Ctnnb1/CTNNB1 promoter. Results are listed in Table S3.

Nuclear-cytoplasmic separation

Nuclear and cytoplasmic protein extraction kit (Beyotime Biotech, P0027) was used for experiments. Small intestinal crypts were collected from wild-type and Ctnnb1^Δi.enh mice as described above. According to manufacturer’s instructions, 20 μl cell precipitation was resuspended in 200 μl cytoplasmic protein extraction reagent A with 1 mM PMSF. After placing on ice for 15 minutes, 10 μl cytoplasmic protein extraction reagent B was added. After vigorous vortex, solutions were centrifuged at 12,000 × g for 5 minutes at 4°C. The supernatant was collected as the extracted cytoplasmic protein. The precipitate was resuspended with 50 μl nuclear protein extraction reagent with 1 mM PMSF. Solutions were placed on ice and vortex vigorously every 2 minutes for a total of 30 minutes. After centrifuging at 12,000 × g for 10 minutes at 4°C, the supernatant was collected as the extracted nuclear protein.

Statistics

All experiments reported in this study were repeated at least three independent times. In vivo analyses were conducted with at least three animals per condition. The number of animals in each experiment is depicted in Figure legends. The signal intensity of tumor stain region was generated using ImageJ software (version 6.0.0.260) as described previously (Wang et al., 2021). GraphPad Prism (version 8.0.2) was used to determine statistical significance. Unpaired two-tailed Student’s t-test were used for analysis between two groups of equal variances by F-tests. When the F-test of equal variance failed, the Welch t-tests were used. Data comparison of three or more groups with control groups were analyzed using one-way ANOVA followed by Dunnett’s multiple comparison test. Data comparison of three or more groups among each other were analyzed using one-way ANOVA followed by Tukey’s multiple comparison test. The exact P values were reported in each figure or indicated as ****, P < 0.0001; ***, P < 0.001; **, P < 0.01; *, P < 0.05, ns = not significant.

Data and code availability

The GEO accession number for the bulk RNA-seq and scRNA-seq data reported in this paper is GSE233979. RNA-seq data of mouse small intestinal crypt and colon tumor have been deposited at GEO: GSE233977. ScRNA-seq data of mouse small intestinal crypt cells in this study have been deposited at GEO: GSE233978. All data are publicly available as of the date of publication. Custom codes were described in detail at methods part. Any additional information required to analyze the data in this paper is available from authors upon reasonable request.

Acknowledgements

We thank Dr. Yeguang Chen for providing the Lgr5-EGFP-IRES-CreERT2 mice.

We thank the Core Facility and the Animal Facility of Medical Research Institute of Wuhan University for technical support. We thank all Zhou lab members for critical reading of the manuscript. Y. Zhou was supported by grants from National Key R&D Program of China (2022YFA0806603 and 2018YFA0800700), National Natural Science Foundation of China (32270876 and 31970770), and the Fundamental Research Funds for the Central Universities (2042022dx0003 and 2042023kf0234). Y.L. was supported by grants from National Key R&D Program of China (2018YFA0800700 and 2022YFA0806603), National Natural Science Foundation of China (31970676), and Hubei Natural Science Foundation (2022CFB128).

Author contributions

Y.Z. conceived and designed the study; X.H. and J.W. performed experiments and analyzed data; C.Z. assisted with processing ChIP-seq data; J.T. and X.P. retrieved and analyzed eQTL data. M.Y. and M.W. provided RNA-seq and ChIP-seq data of colorectal cancer samples; G.Z. provided Hi-C data; Y.Z. and X.H. wrote the paper. Y.Z. and Y.L. provided funding and supervised the study. All authors discussed results and commented to the manuscript.

Competing interests

The authors claimed that they have no competing interests.