Wnt signaling plays an important role in embryonic development and tumorigenesis. These biological effects are exerted by activation of the β-catenin/TCF transcription complex and consequent regulation of a set of downstream genes. TCF-binding elements have been found in the promoter regions of many TCF target genes and characterized by a highly conserved consensus sequence. Utilizing this consensus sequence, we performed an in silico screening for new TCF target genes. Through computational screening and subsequent experimental analysis, we identified a novel TCF target gene, DKK1, which has been shown to be a potent inhibitor of Wnt signaling. Our finding suggests the existence of a novel feedback loop in Wnt signaling.
The canonical Wnt signaling transduction pathway is involved in many developmental processes as well as the development of tumors, such as colorectal cancer (Wodarz and Nusse, 1998; Polakis, 2000). These effects are achieved via activation of downstream target genes by the TCF/LEF family of transcription factors and their coactivator, β-catenin. In the absence of Wnt signaling, a multiprotein complex composed of Axin, glycogen synthase kinase-3β (GSK-3β), and tumor suppressor adenomatous polyposis coli (APC) phosphorylates β-catenin, leading to its ubiquitin/proteasome-mediated degradation. Wnt signaling negatively regulates the function of this complex and thereby stabilizes β-catenin, which is thereby able to translocate to the nucleus, form a complex with TCF, and induce expression of a variety of TCF target genes.
A number of TCF target genes have been identified to date, mainly by experimental approaches. These include a number of genes important for development or tumorigenesis; for example, Myc, CyclinD1, Axin2, Siamois (Brannon et al., 1997; He et al., 1998; Shtutman et al., 1999; Tetsu and McCormick, 1999; Yan et al., 2001; Jho et al., 2002; Lustig et al., 2002). Most of these genes have one or more TCF-binding elements near the transcription start site in their promoter region. TCF-binding elements are composed of a highly conserved consensus sequence 5′-IndexTermCTTTG[A/T][A/T]-3′ (van de Wetering et al., 1991; van de Wetering et al., 1997). To identify new potential TCF target genes, we developed a novel computational approach to search for gene promoters containing the TCF-binding consensus sequence. In the present study, we identify one such gene, DKK1, and experimentally demonstrated that it is indeed a novel TCF target. DKK1 is known to be a secreted protein that functions as a negative regulator of Wnt signaling and plays a crucial role in head formation in vertebrate development (Glinka et al., 1998; Fedi et al., 1999; Mao et al., 2001, 2002; Mukhopadhyay et al., 2001). Wnt ligands bind to the seven-transmembrane receptor, Frizzled, and the coreceptor lipoprotein-related protein 5 and 6 (LRP5/6). DKK1 forms a ternary complex with LRP5/6 and another receptor, Kremen, followed by endocytosis of this complex and removal of LRP5/6 from the cell surface. In addition to this previously known extracellular link, our data reveal a novel intranuclear link between Wnt signaling and its antagonist, DKK1.
Results and discussion
Taking advantage of the fact that the consensus sequence of TCF target genes is highly conserved, we performed in silico screening for genes containing the potential TCF-binding element, 5′-IndexTermCTTTG(T/A)(T/A)-3′, in their promoter region. However, we first had to overcome two difficulties.
First, in order to analyse promoter regions, we needed data on the transcription start sites of a large number of genes. However, most cDNA sequences stored in current databases lack precise information about their 5′-end termini, and a set of promoter sequences large enough for in silico screening is not available in conventional databases. Second, since the consensus sequences of transcription factor-binding elements are generally very short, there is a high likelihood of nonfunctional consensus sequences appearing frequently in genomic sequences. Hence, searching genomic DNA for transcription factor-binding elements based only on the consensus binding sequences results in numerous false positives.
To overcome the first difficulty, we prepared a data set consisting of human promoter sequences from the transcription start site data contained in DBTSS, a specialty database that includes information about transcription start sites based on the 5′-end sequences of full-length cDNAs (Suzuki et al., 2002). We overcame the second difficulty by applying a consensus sequence search in combination with phylogenetic footprinting, a technique that identifies transcriptional regulatory elements by finding evolutionally well-conserved regions between orthologous noncoding genomic sequences from different species (Loots et al., 2000; Pennacchio and Rubin, 2001). Thus, in addition to the human promoter sequence data set from DBTSS, we prepared a putative promoter sequence data set of mouse genes from genomic upstream sequences of mouse orthologous genes, then extracted highly conserved regions from each orthologous pair of promoter sequences, and surveyed these regions for conserved consensus sequences (Figure 1).
By this strategy, we were able to screen the sequences within 1 kbp upstream of the transcription start sites for 7477 genes. We looked for the presence of one or more conserved consensus TCF-binding elements within these conserved upstream regions with identity greater than or equal to 80% over at least 50 bp. We obtained 123 candidates for TCF target genes, including a known TCF target gene, cyclinD1. By contrast, when we performed similar searches using randomly permuted consensus sequences as a control, we obtained only 63.8 hits on an average (n=100). These observations suggest the biological significance of the conserved consensus TCF-binding sequences and the utility of our strategy.
On the other hand, we could not identify well-known target genes except cyclin D1 as candidate TCF target genes by our in silico screening. Since we screened the sequences within 1 kbp upstream of the transcription start sites, only a fraction of TCF target genes could be obtained as candidate genes. For example, genes that possess potential TBEs in introns are not expected to be obtained. Indeed, Axin2, which possesses several highly conserved potential TBEs in the first intron (Jho et al., 2002), was not included in our list of candidate target genes. Furthermore, to eliminate false positives, we used very strict criteria for phylogenetic footprinting. According to our criteria, the Myc promoter region containing two TBEs (He et al., 1998) was not identified as a candidate for TCF targets. Thus, our program needs to be improved in sensitivity while keeping accuracy.
Among the genes obtained, the DKK1 gene was striking in that it contained four TCF-binding elements, more than any other genes. As shown in Figure 2, the DKK1 promoter contains a TATA box just proximal to the transcription start site, and two regions evolutionally well-conserved between human and mouse. These regions presumably function as transcriptional regulatory elements. Each of the proximal and distal regions contains two consensus TCF-binding sequences. Consistent with the remarkable sequence conservation, our gel shift assays demonstrated that the four consensus TCF-binding sequences bind to the HMG domain of TCF4E in vitro (Figure 3). These facts prompted us to further examine whether DKK1 is indeed a TCF target gene.
We first examined whether the DKK1 promoter responds to the β-catenin/TCF complex. A reporter plasmid containing the DKK1 promoter upstream of a luciferase reporter gene was transfected into 293T cells along with β-catenin-S33Y, an oncogenic stable mutant form of β-catenin (Morin et al., 1997) (Figure 4a). The DKK1 promoter was activated up to threefold in a dose-dependent manner with increasing amounts of β-catenin-S33Y. When TCF4E was cotransfected with these plasmids, β-catenin-mediated transactivation increased in a dose-dependent manner to a maximum of more than fivefold. These results suggest that the DKK1 promoter contains functional TCF-binding elements and is indeed transactivated by the β-catenin/TCF complex in vivo.
We next attempted to identify functional TCF-binding elements in the DKK1 promoter. Plasmids containing various combinations of mutations within the potential TCF-binding elements and/or deletion of the distal promoter region were generated, and the responsiveness of each mutant construct to β-catenin-S33Y expression was measured by luciferase assays (Figure 4b). Mutations in the distal two consensus TCF-binding elements resulted in no remarkable change in responsiveness compared to the wild-type construct. Also, deletion of the region containing these two consensus elements did not affect β-catenin-mediated activation of the reporter gene. In contrast, mutation of either of the proximal two elements, particularly the second proximal element, attenuated the responsiveness to β-catenin. Furthermore, mutation of both elements almost completely eliminated the responsiveness to β-catenin. From these observations, we concluded that, out of the four potential TCF-binding elements, only the proximal two actually function as TCF-binding elements.
To demonstrate that DKK1 is a bona fide TCF target gene, we investigated whether expression level of DKK1 is affected by Wnt signaling in several cell lines (Figure 5a). Our RT–PCR analysis showed that DKK1 mRNA increased following activation of Wnt signaling in 293T cells by treatment with Wnt 3a or LiCl, an inhibitor of GSK3β (Hedgepeth et al., 1997). A similar increase in DKK1 expression was also induced by overexpression of β-catenin-S33Y. We also observed induction of DKK1 expression by Wnt signaling in mouse ES cells and human prostate cancer DU145 cells. Furthermore, we found that expression of the DKK1 mRNA was downregulated by expression of a dominant-negative mutant of TCF4E, TCF4E-ΔN, in human hepatoma HepG2 cells, a cell line in which Wnt signaling is constitutively activated due to a mutation in β-catenin (de La Coste et al., 1998). These results suggest that DKK1 expression is directly regulated by the β-catenin/TCF complex under various physiological circumstances.
To examine the time course of DKK1 induction, mRNA levels were assayed over time, following stimulation of 293T by LiCl (Figure 5b). Induction of DKK1 was detectable within 4 h after LiCl treatment and reached a maximum at 8 h after treatment. This rapid kinetics of DKK1 induction is consistent with the idea that DKK1 is a direct downstream target of Wnt signaling. Western blot analysis of conditioned medium from 293T cells confirmed that Wnt signaling upregulates DKK1 protein expression (Figure 5c): treatment with LiCl induced expression of DKK1 protein, which was secreted and accumulated in the culture medium. By contrast, we could not detect any endogenous DKK1 protein in the absence of stimulation.
The discovery that DKK1 expression can be upregulated via β-catenin/TCF, and the previously known capacity of DKK1 to inhibit Wnt signaling led us to the hypothesis that DKK1 participates in a novel negative feedback loop in Wnt signaling. To test this hypothesis, we examined whether the responsiveness of cells to Wnt signaling is altered by suppression of DKK1 induction. Using pTOP-tk-luciferase, a luciferase reporter construct that responds to Wnt signaling, we confirmed the response of 293T cells to Wnt1 and consequent DKK1 induction (Figure 6). Notably, this response was enhanced by cotransfection of an RNAi construct targeted against DKK1, which suppresses induction of DKK1 by Wnt1. Induction of the pTOP-tk-luciferase reporter by Wnt1 is normally high at 24 h and is attenuated by 48 h, presumably due to negative feedback mechanisms, but cotransfection of DKK1 RNAi partially abolished this attenuation. Taking these observations into consideration, it is reasonable to assume that DKK1 functions as a novel component of negative feedback loop that inhibits Wnt signaling.
Through computational screening and subsequent experimental analysis, we established that the DKK1 gene is a novel TCF target gene. Our results raise the possibility that DKK1 expression is upregulated in tumor cells in which β-catenin/TCF-mediated transcription is activated due to mutations in APC, β-catenin or Axin. Indeed, it has been reported that DKK1 is overexpressed in a large fraction of human hepatoblastoma, which frequently exhibit constitutively active Wnt signaling (Koch et al., 1999; Wirths et al., 2003). In addition, we showed that a dominant-negative mutant of TCF4E suppresses DKK1 expression in the HepG2 cell line, where Wnt signaling is constitutively activated. Furthermore, our findings suggest that DKK1 participates in a novel negative feedback loop in Wnt signaling, similar to Axin2, which is involved in the segmentation clock during somitogenesis (Yan et al., 2001; Jho et al., 2002; Lustig et al., 2002; Aulehla et al., 2003).
During early development of Xenopus and zebrafish, pre-MBT Wnt signaling induces DKK1 expression, which is essential for head formation (Glinka et al., 1998; Shinya et al., 2000). Notably, all four of the TCF-binding elements, which are evolutionally conserved between human and mouse, are also found in the zebrafish DKK1 promoter. In particular, the homology in the proximal two functional TCF-binding elements includes not only the consensus sequences but also adjacent regions, despite the long evolutionary distances separating these species (data not shown). We therefore speculate that DKK1 plays a crucial role as a TCF target gene in embryogenesis among different species.
We developed a novel computational approach for identifying TCF target genes from promoter sequence data. Our approach can be applied to the screening of target genes for transcription factors whose binding element sequence is relatively well conserved, like TCF. In addition to the promoter sequence data, utilization of the expanding body of expression profile data will help this method to be more useful for the elucidation of the transcriptional network.
Materials and methods
In silico screening
It was assumed that DBTSS oligocapped-cDNA clones would start most frequently at the transcription start site (Suzuki et al., 2002), therefore a set of human promoter sequences, which encompass 1 kbp upstream of the transcription start sites, was assembled from the human genome sequences. To obtain a set of putative mouse promoter sequences, each human cDNA sequence was subjected to blast analysis against the RTS database of FANTOM2 (ftp://fantom2.gsc.riken.go.jp/RTPS/rts-transcripts.fasta.gz) (The FANTOM Consortium and the RIKEN Genome Exploration Research Group Phase I & II Team, 2002), using the TBLASTX program with default parameters and a cutoff E <e-50. The best-fit genes, which were assumed to be the mouse orthologs, were mapped to the mouse genome, and 20 kbp of the upstream regions were obtained. For phylogenetic footprinting, each orthologous pair of promoter sequences was locally aligned using the bl2seq program in the NCBI BLAST suite. The human and mouse genome sequence data were downloaded from UCSC (http://genome.ucsc.edu/).
Construction of plasmids
For construction of the DKK1 promoter–luciferase reporter plasmids, a human BAC clone RPCl 11 HS, which contains the DKK1 locus, was purchased from the BACPAC Resources. The region containing the 1.3 kbp DKK1 promoter and 5′UTR was amplified by PCR from the BAC clone using primers, 5′-IndexTermTTCTCCTCTT AGTCTTTCTG-3′ and 5′-IndexTermCACTGCAT TTGGATAGCTGGT-3′, and cloned into pGEM-T Easy (Promega). Using this construct as a template, the following PCR was performed with restriction site-added primers, 5′-IndexTermGATGGATCCGTCAAGGTAAA T-3′ and 5′-IndexTermGATCCA-TGGCTCAGAAGGACTCAAGAGGGAGA-3′, and the product was cloned into BglII/NcoI-digested pGL3-Basic (Promega) following digestion with BamHI/NcoI for long type (−1037∼+163), or with BglII/NcoI for deleted short type (−537∼+163). Site-directed mutagenesis of TCF-binding consensus sequences was created by standard PCR techniques using Pyrobest DNA polymerase (TAKARA). DNA oligonucleotides encoding shRNAs were subcloned into the U6 promoter vector, pSHAG-1 (Paddison et al., 2002). The sequences of the region targeted for shRNAs in DKK1 cDNA was 5′-IndexTermCCTGAAAGAAGGTCAAGTGTGTACCAAGCAT-3′. pTOP-tk-luciferase and pFOP-tk-luciferase were obtained from V Korinek and H Clevers.
Cell culture, transfection, adenovirus, and conditioned medium
293T, DU145, and HepG2 cells were cultured as monolayers in DMEM (NISSUI) supplemented with 10% fetal bovine serum (JCS). The mouse ES cell line MG1.19 was cultured without feeders in LIF-supplemented medium as described previously (Niwa et al., 1998). All cells were maintained at 37°C in an atmosphere of humidified air with 5% CO2. Plasmids were transfected into these cells using LipofectAMINE 2000 (Life Technologies) or Lipofectamin plus reagent (Invitrogen). Ad-LacZ and Ad-β-catenin-S33Y were constructed as described previously (Sekiya et al., 2003). Ad-TCF-ΔN was provided by Y Nakamura. Cells were infected with adenoviruses at multiplicity of infection 40. Wnt3a conditioned medium was prepared from L Wnt-3A cells that was obtained from ATCC following the accompanying protocol.
Gel shift assays
Using a GST-fusion protein containing human hTCF4E amino acids 265–495, gel shift assays were performed as described previously (Tago et al., 2000). 32P-labeled double-stranded oligonucleotides were used as probes and sequences of their sense strand are as follows: TBE1, 5′-IndexTermGTCCCGGCCACTTTGATCTCACGCGTCTGCC T-3′; TBE2, 5′-IndexTermCCGCCATTGCCCTGATTCAAAGAACAACATTAAATG-3′; TBE3, 5′-IndexTermCCTCCCAGCGCTTTGAAATCCCATCCCGGCTT-3′; TBE4, 5′-IndexTermCCCATCCCGG CTTTGTTGTCTCCCT CCCAAGG-3′.
Cells were plated in 12-well dishes 18 h prior to transfection. Transfections were performed with Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. Luciferase assays were performed using the Dual-Luciferase Reporter Assay System following the manufacturer's protocol.
Total RNA was isolated from cell lines using NucleoSpin (MACHEREY-NAGEL). First-strand cDNA was synthesized using oligo-dT and Superscript III reverse transcriptase (Invitrogen). Amplification was carried out at 94°C for initial denaturation, followed by appropriate cycles at 94°C for 20 s, 60°C for 20 s, and 72°C for 30 s. The products were electrophoresed on a 2% agarose gel and detected by EtBr staining. Primer sequences are as follows: DKK1-FW, 5′-IndexTermTCCCCTGTGATTGCAGTAAA-3′; DKK1-REV, 5′-IndexTermTCCAAGAGATCCTTGCGTTC-3′; Axin2-FW, 5′-IndexTermAGTGTGAGGTCCACGGAAA C-3′; Axin2-REV, 5′-IndexTermCTTCACACTGCGATGCATTT-3′; GAPDH-FW, 5′-IndexTermACAGTCA GCCGCATCTTCTT-3′; GAPDH-REV, 5′-IndexTermGACAAGCTTCCCGTTCTCAG-3′; mouse DKK1-FW, 5′-IndexTermAGACACTTCTGGTCCAAGATC-3′; mouse DKK1-REV, 5′-IndexTermACAG GTAAGTGCCACACTGAG-3′; mouse GAPDH-FW, 5′-IndexTermTCCACACCCTGTT GCTA-3′; mouse GAPDH-REV, 5′-IndexTermACCACAGTC CATGCCATCAC-3′.
Antibody to DKK1 was prepared by immunizing rabbits with bacterially expressed GST fusion protein containing human DKK1 amino acids 32–178 and purified by affinity chromatography with a column to which the antigen had been linked. Serum-free conditioned medium from 293T cells was concentrated 10-fold by ultrafiltration with VIVASPIN (VIVASCIENCE) and the fraction absorbed to Wheat Germ Lectin-Sepaharose (Amersham) was analysed by Western blotting as described previously (Kawasaki et al., 2000).
Aulehla A, Wehrle C, Brand-Saberi B, Kemler R, Gossler A, Kanzler B and Herrmann BG . (2003). Dev. Cell, 4, 395–406.
Brannon M, Gomperts M, Sumoy L, Moon RT and Kimelman D . (1997). Genes Dev., 11, 2359–2370.
de La Coste A, Romagnolo B, Billuart P, Renard CA, Buendia MA, Soubrane O, Fabre M, Chelly J, Beldjord C, Kahn A and Perret C . (1998). Proc. Natl. Acad. Sci. USA, 95, 8847–8851.
Fedi P, Bafico A, Nieto Soria A, Burgess WH, Miki T, Bottaro DP, Kraus MH and Aaronson SA . (1999). J. Biol. Chem., 274, 19465–19472.
Glinka A, Wu W, Delius H, Monaghan AP, Blumenstock C and Niehrs C . (1998). Nature, 391, 357–362.
He TC, Sparks AB, Rago C, Hermeking H, Zawel L, da Costa LT, Morin PJ, Vogelstein B and Kinzler KW . (1998). Science, 278, 1606–1609.
Hedgepeth CM, Conrad LJ, Zhang J, Huang HC, Lee VM and Klein PS . (1997). Dev. Biol., 185, 82–91.
Jho EH, Zhang T, Domon C, Joo CK, Freund JN and Costantini F . (2002). Mol. Cell. Biol., 22, 1172–1183.
Kawasaki Y, Senda T, Ishidate T, Koyama R, Morishita T, Iwayama Y, Higuchi O and Akiyama T . (2000). Science, 289, 1194–1197.
Koch A, Denkhaus D, Albrecht S, Leuschner I, von Schweinitz D and Pietsch T . (1999). Cancer Res., 59, 269–273.
Loots GG, Locksley RM, Blankespoor CM, Wang ZE, Miller W, Rubin EM and Frazer KA . (2000). Science, 288, 136–140.
Lustig B, Jerchow B, Sachs M, Weiler S, Pietsch T, Karsten U, van de Wetering M, Clevers H, Schlag PM, Birchmeier W and Behrens J . (2002). Mol. Cell. Biol., 22, 1184–1193.
Mao B, Wu W, Davidson G., Marhold J, Li M, Mechler BM, Delius H, Hoppe D, Stannek P, Walter C, Glinka A and Niehrs C . (2002). Nature, 417, 664–667.
Mao B, Wu W, Li Y, Hoppe D, Stannek P, Glinka A and Niehrs C . (2001). Nature, 411, 321–325.
Morin PJ, Sparks AB, Korinek V, Barker N, Clevers H, Vogelstein B and Kinzler KW . (1997). Science, 275, 1787–1790.
Mukhopadhyay M, Shtrom S, Rodriguez-Esteban C, Chen L, Tsukui T, Gomer L, Dorward DW, Glinka A, Grinberg A, Huang SP, Niehrs C, Belmonte JC and Westphal H . (2001). Dev. Cell, 1, 423–434.
Niwa H, Burdon T, Chambers I and Smith A . (1998). Genes Dev., 12, 2048–2060.
Paddison PJ, Caudy AA, Bernstein E, Hannon GJ and Conklin DS . (2002). Genes Dev., 16, 948–958.
Pennacchio LA and Rubin EM . (2001). Nat. Rev. Genet., 2, 100–109.
Polakis P . (2000). Genes Dev., 14, 1837–1851.
Sekiya T, Adachi S, Kohu K, Yamada T, Higuchi O, Furukawa Y, Nakamura Y, Nakamura T, Tashiro K, Kuhara S, Ohwada S and Akiyama T . (2003). J. Biol. Chem., 279, 6840–6846.
Shinya M, Eschbach C, Clark M, Lehrach H and Furutani-Seiki M . (2000). Mech. Dev., 98, 3–17.
Shtutman M, Zhurinsky J, Simcha I, Albanese C, D'Amico M, Pestell R and Ben-Ze'ev A . (1999). Proc. Natl. Acad. Sci. USA, 96, 5522–5527.
Suzuki Y, Yamashita R, Nakai K and Sugano S . (2002). Nucleic Acids Res., 30, 328–331.
Tago K, Nakamura T, Nishita M, Hyodo J, Nagai S, Murata Y, Adachi S, Ohwada S, Morishita Y, Shibuya H and Akiyama T . (2000). Genes Dev., 14, 1741–1749.
Tetsu O and McCormick F . (1999). Nature, 398, 422–426.
The FANTOM Consortium and the RIKEN Genome Exploration Research Group Phase I & II Team (2002). Nature, 420, 563–573.
van de Wetering M, Cavallo R, Dooijes D, van Beest M, van Es J, Loureiro J, Ypma A, Hursh D, Jones T, Bejsovec A, Peifer M, Mortin M and Clevers H . (1997). Cell, 88, 789–799.
van de Wetering M, Oosterwegel M, Dooijes D and Clevers H . (1991). EMBO J., 10, 123–132.
Wirths O, Waha A, Weggen S, Schirmacher P, Kuhne T, Goodyer CG, Albrecht S, Von Schweinitz D and Pietsch T . (2003). Lab. Invest., 83, 429–434.
Wodarz A and Nusse R . (1998). Annu. Rev. Cell Dev. Biol., 14, 59–88.
Yan D, Wiesmann M, Rohan M, Chan V, Jefferson AB, Guo L, Sakamoto D, Caothien RH, Fuller JH, Reinhard C, Garcia PD, Randazzo FM, Escobedo J, Fantl WJ and Williams LT . (2001). Proc. Natl. Acad. Sci. USA, 98, 14973–14978.
We thank S Adachi, K Satoh, A Nishida and T Moriguchi for helpful discussion. This work was supported by Grants-in-Aid for Scientific Research on Priority Areas and the Organization for Pharmaceutical Safety and Research.
About this article
Cite this article
Niida, A., Hiroko, T., Kasai, M. et al. DKK1, a negative regulator of Wnt signaling, is a target of the β-catenin/TCF pathway. Oncogene 23, 8520–8526 (2004). https://doi.org/10.1038/sj.onc.1207892
- TCF-binding element
- negative feedback
Deciphering the role of Wnt signaling in acute myeloid leukemia prognosis: how alterations in DNA methylation come into play in patients’ prognosis
Journal of Cancer Research and Clinical Oncology (2020)
Gynecologic Oncology (2020)
Temozolomide induces activation of Wnt/β-catenin signaling in glioma cells via PI3K/Akt pathway: implications in glioma therapy
Cell Biology and Toxicology (2020)
Pathology - Research and Practice (2020)
Loss of mismatch repair signaling impairs the WNT–bone morphogenetic protein crosstalk and the colonic homeostasis
Journal of Molecular Cell Biology (2020)