|Home | About | Journals | Submit | Contact Us | Français|
The CDX1 and CDX2 homeoproteins are intestine‐specific transcription factors regulating homeostasis. We investigated their relevance in experimentally‐induced intestinal inflammation.
The response to intestinal inflammation induced by dextran sodium sulfate (DSS) was compared in wild type, Cdx1‐/‐ and Cdx2+/− mice. Intestinal permeability was determined in wild type and Cdx2+/− mice. Protein‐protein interactions were investigated by co‐immunoprecipitation and GST‐pulldown, and their functional consequences were assessed using Luciferase reporter systems.
Heterozygous Cdx2+/− mice, but not Cdx1‐/‐ mice, were hypersensitive to DSS‐induced acute inflammation as all these mice showed blood in the stools at day 1 of DSS treatment. Hypersensitivity was associated to a 50% higher intestinal permeability. In Cdx2+/‐ mice, the colonic epithelium was repaired during the week after the end of DSS treatment, whereas two weeks were required for wild type animals. Subsequently, no colonic tumour was observed in Cdx2+/− mice subjected to 5 repeated cycles of DSS, in contrast to the 2.7 tumours found per wild type mouse. Based on the fact that Smad3+/− mice, like Cdx2+/− mice, better repair the damaged intestinal epithelium, we found that the CDX2 protein interacts with SMAD3, independently of SMAD4, resulting in a 5‐fold stimulation of SMAD3 transcriptional activity. CDX1 also interacted with SMAD3 but it inhibited by 10‐fold the SMAD3/SMAD4‐dependent transcription.
The Cdx1 and Cdx2 homeobox genes have distinct effects on the outcome of a pro‐inflammatory challenge. This is mirrored by different functional interactions of the CDX1 and CDX2 proteins with SMAD3, a major element of the TGFβ signalling pathway.
Chronic bowel inflammation gives rise to high morbidity diseases characterised by successive periods of exacerbation and remission.1 It also increases the risk of cancer in relation with disease duration and early onset.2,3 Inflammatory bowel diseases depend on altered interactions between immunoregulatory factors in the gut mucosa and environmental factors.4 The intestinal epithelium, which is disrupted and repaired during ulceration and remission, participates in these interactions at the interface between the mucosa and the luminal content.
CDX1 and CDX2 are homeodomain transcription factors specific of the intestinal epithelium. During gut development, the Cdx2 gene dictates intestinal identity5 whereas Cdx1 has a less prominent role.6Cdx1 and Cdx2 have both similar and distinct effects on proliferation and differentiation in intestinal cell lines,7,8,9 and also in pathological conditions. For instance, the reduction of Cdx2 expression in Cdx2+/− mice facilitates the progression of colorectal cancer induced chemically or linked to a genetic predisposition, suggesting a tumour suppressor role.10,11 In contrast, the loss of Cdx1 reduces the hyperproliferative state induced by pRb/p300 deficiency.12
Since the intestinal epithelium is perturbed and repaired in inflammatory bowel diseases, we used Cdx1−/− and Cdx2+/− mice and the experimental model of colitis induced by dextran sodium sulfate (DSS) to investigate the role of the Cdx1 and Cdx2 genes on the outcome of the intestinal epithelium after pro‐inflammatory injury.
Cdx2+/−,13 Cdx1‐/‐6 and wild type mice in the 129sv/C57Bl6 background were housed under pathogen‐free conditions and given treatments according to the guidelines of the Ethic Committee of the University Louis Pasteur of Strasbourg. For acute treatments, two month‐old animals received 3% DSS (36–50 kDa, MP Biomedicals, Illkirch, France) in drinking water for five days and then tap water ad libitum. For chronic treatments, mice were subjected either to three cycles consisting of five days with 1.5% DSS in drinking water followed by 16 days with tap water, or to five cycles of five days with 3% DSS followed by 16 days with tap water. Blood in the stools was analysed using Haemoccult II® (SKD, Gagny, France). Mice were euthanised by cervical disruption. The gastro‐intestine was removed, flushed with PBS, mounted as Swiss Roll, fixed in 4% paraformaldehyde for 4 h and embedded in paraffin. Assessment of inflammation of the colon was performed in a blinded fashion by two pathologists with regard to stiffness, edema, ulcerations and thickness, as described.14
T84, HCT116 (ATCC) and HCT116 Smad4‐/‐ human colon cancer cells15 were cultured under standard conditions in Dulbecco's modified Eagle's medium with 10% fetal bovine serum at 37°C in humidified atmosphere under 5% CO2.
Intestinal permeability was determined in three‐month old wild type (n=12) and Cdx2+/− (n=12) mice by measuring the appearance of FITC‐dextran in the blood.16 Mice starved for 24 h received either tap water or 3% DSS in drinking water (6 mice per genotype) for 16 h. Then, FITC‐dextran (4.4 kDa; Sigma, St Quentin Fallavier, France) was administered by gavages (60 mg/100 g body weight). Blood was collected 4 h later by orbital retrobulbar puncture and centrifuged at 3000 g for 20 min at 4°C. Plasma (50 μL) was diluted volume to volume with PBS to determine fluorescence using a Molecular FX Imager (Biorad, Marnes‐la‐Coquette, France).
Immunohistochemistry was performed on 5 μm paraffin sections11 using primary antibodies raised against Ki67 (dilution 1:50, ab833, Abcam, Cambridge, UK), β‐catenin (dilution 1:150, clone 14, BD Transduction Laboratories, Lexington, Kentucky, USA), Cdx2 (dilution 1:500, CDX2‐88, Biogenex, San Ramon, California, USA) and Cdx1.11
The plasmids pCdx2‐S, pCdx1‐S, pCB6‐HA‐CDX2 and pCB6‐HA‐CDX1,7,17 those encoding 6xMyc‐Smad1, −2, −3, −4, −6 in pCDNA3.1 and Flag‐Smad3‐MH1‐L‐MH2, ‐MH1, ‐MH1‐L, ‐L‐MH2, ‐MH2 in pDEF318 and the reporter vectors CAGA‐Luc19 and pSI‐Luc20 have been described. To construct pGST‐CDX2 and pGST‐CDX1, a NheI site was introduced immediately downstream of the translation start site of pCdx1‐S and pCdx2‐S,17 the resulting plasmids were cut with NheI, filled in with T4 DNA polymerase, then cut with BamHI, and the fragments were introduced into the EcoRV/BamHI sites of pBC.21
T84 or HCT116 cells cultured at 70–80% confluence were transfected using Jet‐PEI (Polyplus Transfection, Illkirch, France) for 48 h, then scraped in ice‐cold NP40 lysis buffer (20 mM Tris at pH7.5, 150 mM NaCl, 1% Igepal) and a cocktail of protease inhibitors (Sigma), placed on ice for 10 min and centrifuged at 10.000 g for 10 min at 4°C. Protein content in the supernatant was assayed using Bradford reaction. For co‐immunoprecipitation or GST‐pulldown assays, protein extracts (1 mg) were incubated overnight at 4°C with agarose bead‐coupled anti‐HA antibody matrix (25 μl, 3F10, Roche, Indianapolis, IN) or for 3 h under gentle agitation at 4°C with glutathione‐Sepharose‐4B beads (25 μL, 50% vol/vol, GE Healthcare, Munich, Germany), then beads were washed three times for 15 min in lysis buffer and proteins were analysed by SDS‐PAGE and western blot. Western blots were performed as described22 using primary antibodies against HA (3F10, Roche, dilution 1:1000), Flag (M2, Sigma, dilution 1:1000), Myc (9E10, Santa Cruz, Heidelberg, Germany, dilution 1:1000), GST (G‐7781, Sigma, dilution 1:5000), CDX2 (CDX2‐88, Biogenex, dilution 1:5000) or SMAD3 (ab28379‐100, Abcam, dilution 1:1000), and secondary antibodies coupled to horseradish peroxydase (GE Healthcare): sheep anti‐mouse κ‐chain immunoglobulins (dilution 1:5000), donkey anti‐rabbit immunoglobulins (dilution 1:5000) or sheep anti‐rat immunoglobulins (dilution 1:5000). Co‐immunoprecipitation and GST‐pulldown assays were repeated at least three times.
Luciferase assays were performed in HCT116 or HCT116 Smad4‐/‐ cells co‐transfected with the indicated luciferase reporter plasmids and with pRL‐null (Promega, Charbonnières, France) to normalise transfection efficiency.22 Firefly luciferase activity was normalised with renilla luciferase activity and the data were expressed as fold of induction in the experimental condition as compared to the reporter control. Luciferase assays were repeated at least three times in triplicate.
Data were reported as mean (SEM) and analysed with the unpaired Student t test. P values <0.05 were considered statistically significant.
Cdx1‐/‐, Cdx2+/− and wild type mice were given 3% DSS in drinking water for five days and then allowed to recover with tap water. As generally observed with this experimental model, mortality by sepsis occurred in 10–15% of the animals. As shown in fig 1A1A,, no blood appeared in the stools of wild type and Cdx1‐/‐ mice before day two of treatment. At day two, blood was observed in less than 25% of wild type and Cdx1‐/‐ mice and the percentage progressively rose up to day five. In contrast, all Cdx2+/− heterozygotes exhibited blood in the stools as early as day one of treatment. At day one, structural alterations including edema and crypt atrophy already occurred in the colon of Cdx2+/− mice (fig 2A2Aa,b),a,b), in contrast to wild type and Cdx1‐/‐ mice (fig 2A2Ac).c). At day five, the colonic mucosa was severely impaired with gut wall thickening, massive round cell infiltration, edema and areas of epithelial denudation, irrespective of the genotype (not shown).
The five days treatment with DSS induced a progressive body weight loss. Consistent with the precocious appearance of blood in the stools in Cdx2+/− mice, body weight loss peaked two days after the end of DSS treatment in these animals (day seven) but the peak was delayed by four days in wild type and Cdx1‐/‐ mice (day 11) (fig 1B1B).). Consequently, body weight recovery was faster in Cdx2+/− than in wild type and Cdx1‐/‐ mice (fig 1B1B).). In wild type animals, severe mucosal alterations were observed at days six and eight (fig 2B2Ba,b)a,b) and regenerative epithelium appeared only in a third of the colonic surface at day 13 (fig 2B2Bc).c). In contrast, regenerative epithelium was already apparent at day six and day eight in Cdx2+/− heterozygotes (fig 2B2Bd,e)d,e) and the major part of the colonic surface was repopulated by structurally normal epithelium by day 13 (fig 2B2Bf).f). Differences between Cdx2+/− and wild type mice were confirmed by the inflammatory score (fig 33).
The hypersensitivity of Cdx2+/‐ mice to acute treatment with DSS, which is an abrasive agent, prompted us to compare intestinal permeability in wild type and Cdx2+/− mice by measuring the appearance of FITC‐dextran in blood after gavages. By this way, FITC‐dextran is absorbed at the level of the small intestine and colon.23 In basal condition, intestinal permeability was higher in Cdx2+/− than in wild type mice (fig 4A4A).). Permeability was also higher after a short exposure to 3% DSS for 16 h (fig 4B4B).). Unlike Cdx2+/− mice, intestinal permeability was twofold lower in Cdx1‐/‐ compared to wild type animals (not shown).
Despite their hypersensitivity to the inflammatory stress resulting from acute DSS treatment, Cdx2+/− mice exhibit a better recovery than wild type and Cdx1‐/‐ mice at the end of the pro‐inflammatory challenge. Therefore, we compared the behaviour of Cdx2+/− and wild type mice after chronic inflammation induced by repeated administration of DSS, to mimic sequences of exacerbation and remission. In a pilot experiment, 10 wild type and 10 Cdx2+/− mice were subjected to three cycles of DSS administration, each comprising five days with 1.5% DSS in drinking water followed by a 16‐day recovery period. Due to the mortality by sepsis at each cycle, five mice of each genotype survived the DSS treatments and were killed 10 weeks later for histological examination. Wild type mice, but not Cdx2+/− animals, exhibited edema in the colon, and there was an adenomatous polyp in the distal colon of one wild type mouse (not shown). Then, we repeated the cyclic treatment more drastically using 15 wild types and 15 Cdx2+/− mice subjected to five cycles of DSS at 3%. Three wild types and 4 Cdx2+/− mice survived and were sacrificed 12 weeks after the end of the DSS treatment. Cdx2+/− heterozygotes exhibited regular crypt architecture with a low rate of immune cells infiltration throughout the colon (fig 5A5Aa).a). In contrast, the colonic mucosa of wild type mice showed substantial infiltration of immune cells together with multiple areas characterised by a disorganised mucosal architecture. These areas appeared as flat epithelium above granular tissue, dilated mucinous glands and clusters of irregular crypts (fig 5A5Ab‐d).b‐d). Thus, unlike Cdx2+/− mice, wild type mice showed imperfect repair of the colonic epithelium upon chronic DSS‐induced colitis.
Next, we investigated the long‐term consequences of chronic inflammation using a new series of 40 mice, 20 of each genotype, subjected to five cycles of treatment with 3% DSS and sacrificed 50 weeks later. Six wild types and five Cdx2+/− animals survived the DSS treatment. Macroscopic examination at necropsy failed to detect any tumour in the small intestine and colon of Cdx2+/− animals. By contrast, five out of the six wild type mice showed tumours in the colon, giving a total of 16 tumours (fig 5B5Ba).a). Tumours corresponded to adenomatous polyps with low‐ to high‐grade dysplasia or adenocarcinoma that invaded the lamina propria into the submucosa (fig 5B5Bb).b). Tumour cells were highly proliferative as shown by strong Ki67 immunolabelling (fig 5B5Bc).c). The expression of several proteins was altered: β‐catenin distribution was heterogeneous with some cells exhibiting a predominant membranous localisation and others showing cytoplasmic/nuclear β‐catenin accumulation (fig 5B5Bd);d); CDX2 and CDX1 proteins expression was reduced and heterogeneous in the tumours compared to normal adjacent mucosa (fig 5B5Be,f).
Similarly to Cdx2+/− mice which better repair the damaged colon epithelium after DSS injury, accelerated re‐epithelialisation of the colon characterises Smad3+/− mice in a model of colitis.24 Therefore, we analysed the possibility for a connection between Cdx2 and Smad's in intestinal cells. For this purpose, HCT116 cells were transfected with the reporter plasmid CAGA‐Luc to monitor SMAD transcriptional activity (fig 6A6Aa).a). Co‐transfection of CAGA‐Luc with plasmids encoding either SMAD3 (without co‐transfection with SMAD4) or CDX2 led respectively to a modest two‐fold and 3.5‐fold increase of luciferase activity. However, luciferase activity was 10‐fold stimulated when both plasmids for SMAD3 and CDX2 were combined, and the effect was proportional to the amount of CDX2 (fig 6A6Ab).b). The synergistic effect of SMAD3 and CDX2 on CAGA‐Luc was independent of the co‐stimulator of SMAD3, SMAD4, since it was also effective in HCT116 Smad4‐/‐ cells which lack SMAD4 expression (fig 6A6Ab).b). Next, instead of the CAGA‐Luc reporter, the experiments were performed using the CDX2‐responding reporter plasmids SI‐Luc (fig 6A6Ac)c) and LPH‐Luc (not shown) containing the promoters of the intestinal genes for sucrase‐isomaltase and lactase‐phlorizin hydrolase. The transcriptional activity of both promoters was stimulated by CDX2, whereas SMAD3 alone or in combination with SMAD4 was without effect. Interestingly, the combination of CDX2 and SMAD3 did not further activate transcription over CDX2 alone. Therefore, SMAD3 and CDX2 functionally cooperate to stimulate the transcription of a SMAD‐responsive reporter, but not of CDX2‐responsive reporters.
In an attempt to address the molecular basis of the collaborative effect between CDX2 and SMAD3, T84 cells, which have an unaltered TGFβ signalling pathway, were transfected with the plasmid expressing HA‐tagged CDX2. Co‐immunoprecipitation experiments with anti‐HA antibody led to recover endogenous SMAD3 with HA‐tagged CDX2 (fig 6B6Ba).a). To confirm this result, HCT116 cells were co‐transfected with the plasmids encoding the GST‐CDX2 fusion protein and Myc‐SMAD3. GST‐CDX2 was able to pulldown Myc‐SMAD3 (fig 6B6Bb),b), indicating that CDX2 and SMAD3 belong to a same protein complex. Myc‐SMAD2 and Myc‐SMAD4 were also pulled‐down by GST‐CDX2, but not Myc‐SMAD1, suggesting that CDX2 differentiates between members of the SMAD family (data not shown). The SMAD3 protein is made of the MH1 domain responsible for nuclear import, DNA binding and transcription, a linker domain and the MH2 domain responsible for oligomerisation and transcription. By pull‐down assays in HCT116 cells with plasmids encoding GST‐CDX2 and Flag‐tagged versions of truncated forms of SMAD3, we identified MH2 as the domain of interaction with CDX2 (fig 6B6Bc).c). Reciprocally, we constructed plasmids encoding either the amino‐terminal transactivating domain of CDX2 or the DNA‐binding homeodomain linked to the carboxy‐terminal domain, but none of these truncated forms co‐purified with SMAD3, in contrast to the full‐length CDX2, suggesting that a complex conformation of CDX2 is required for interacting with SMAD3 (data not shown).
Having demonstrated the interaction between SMAD3 and CDX2, we tested the homologous homeodomain factor CDX1. Cotransfection of HCT116 cells with plasmids encoding GST‐CDX1 and Myc‐SMAD3 revealed an interaction between SMAD3 and CDX1 by GST‐pulldown (fig 8Aa). Like SMAD3, SMAD4 also interacted with GST‐CDX1 (not shown). Using Flag‐tagged truncated forms of SMAD3 we identified the domain MH2 of SMAD3 for interaction with CDX1, as previously shown for CDX2 (fig 7A7Ab).
Then we examined the functional effect of the SMAD3/CDX1 interaction. Upon co‐transfection in HCT116 cells, CDX1 did not change the low transcriptional activity of SMAD3 on the CAGA‐Luc reporter plasmid (not shown). Since SMAD3 cooperates with SMAD4 for optimal transcriptional activation, we analysed the effect of CDX1 on the SMAD3/SMAD4 heterodimer. Co‐transfection of CAGA‐Luc with the plasmids encoding SMAD3, SMAD4 and CDX1 showed that CDX1 inhibited SMAD3/SMAD4‐dependent transcription of CAGA‐Luc (fig 7Ba) in a dose‐dependent manner (fig 7B7Bb).b). The inhibitory effect of CDX1 was comparable to that exerted by SMAD6 (fig 7B7Ba).
Since CDX1 interacts with SMAD3/SMAD4 to inhibit the activity of this complex whereas CDX2 interacts with SMAD3 to stimulate its function, we asked if CDX2 can release the inhibition exerted by CDX1 on SMAD3/SMAD4. As illustrated in fig 7C7Ca,a, the addition of CDX2 counteracted the inhibitory effect of CDX1 on SMAD3/SMAD4. The opposing effect exerted by CDX2 on the inhibition by CDX1 was not caused by an interaction between CDX1 and CDX2 since GST‐CDX2 did not pulldown HA‐CDX1 and reciprocally, GST‐CDX1 did not pulldown HA‐CDX2 (fig 7C7Cb).
This study demonstrates that the intestine‐specific homeobox genes Cdx1 and Cdx2 have different impacts on the response of the intestinal epithelium to a pro‐inflammatory stress, since Cdx1‐/‐ mice behave like wild type animals, whereas Cdx2+/− mice exhibit both hypersensitivity to DSS‐induced colitis and improved mucosal repair.
The transcription factor CDX2 is at the head of the intestinal genetic program5 and controls the expression of a number of downstream genes, some of which being thought to play a role in inflammation. For instance, CDX2 is a positive regulator of the Muc2 and TFF3 genes,25,26 whose deficiency induces hypersensitivity to DSS‐induced colitis.26,27 Moreover, CDX2 controls cell‐cell interactions and the expression of cadherins7,28,29 In line with this, evidence is provided here that Cdx2+/− mice exhibit increased intestinal permeability, which itself could strengthen the abrasive effect of DSS and exacerbate the acute inflammatory response. Therefore, it is likely that a reduced expression of Cdx2 changes several functions and properties of the colon epithelium, which together converge towards the phenotype of hypersensitivity of Cdx2+/− mice to DSS‐induced colitis.
Epithelial repair after injury requires a complex interplay between cell migration and proliferation to resurface the mucosa and to restore crypt morphology. Cdx2 regulates the proliferation/differentiation balance.7,9 Moreover, we found that decreasing Cdx2 expression enhances the migration of intestinal cells,30 which may account for the improved intestinal repair observed in Cdx2+/− mice. Like Cdx2+/− mice, Smad3+/− animals also show a better intestinal recovery after experimental colitis,24 while the inhibition of regulatory SMADs appears as a general condition to accelerate healing in several epithelia.31,32 The improved recovery of the colonic epithelium after inflammatory injury in Cdx2+/− mice, as in Smad3+/− animals, is of significance since we report a protein/protein interaction between CDX2 and SMAD3 to synergistically stimulate SMAD3 transcriptional activity. Hence, the better healing of Cdx2+/− mice may result from the fact that reducing CDX2 subsequently diminishes the activity of SMAD3, therefore improving tissue repair. SMAD3 associates with SMAD4 to transduce TGFβ signals, but TGFβ receptors also signal via SMAD‐independent pathways. Of note, blocking TGFβ signalling by a dominant‐negative receptor slows down intestinal repair after inflammatory injury,33 in clear opposition with the phenotype of Smad3+/− mice.24 This suggests that signalling cascades downstream of TGFβ receptors must be correctly balanced between SMAD‐dependent and SMAD‐independent pathways during tissue repair. In this context, CDX2 may regulate this balance by binding and regulating the activity of SMAD3. In addition, we show here that CDX1, the homologue of CDX2, also binds SMAD3, yet inhibiting the SMAD3/SMAD4 complex, in contrast to the stimulatory effect exerted by CDX2 on SMAD3. The opposing effects of CDX1 and CDX2 are in line with the different response of Cdx1‐/‐ and Cdx2+/− mice to DSS‐induced colitis. CDX2 and CDX1 have overlapping and different effects on cell proliferation and differentiation,7,8,9 and we have previously reported that these two factors can differentially interact with components of the basic transcriptional machinery.34 Here, the binding of CDX2 and CDX1 to a common partner, SMAD3, while giving opposite effects, provides a new molecular mechanism to explain the divergent functions of these two homologous factors.
A tumour suppressor role has been attributed to Cdx2 since Cdx2+/− mice are hypersensitive to colon carcinogenesis induced chemically by AOM or linked to a genetic predisposition.10,11 Paradoxically, we found here that Cdx2+/− mice are resistant to colon cancers induced by chronic inflammation. AOM is a mutagenic compound, whereas DSS is a non‐mutagenic abrasive agent that compromises cell proliferation.35 Hence, CDX2 reduction has ultimately opposite consequences on colorectal cancer, depending on the mechanism of tumourigenesis. It is likely that the improved tissue repair in Cdx2+/− mice protects them from prolonged inflammation and subsequent tumour development. Reducing CDX2 could also attenuate prolonged inflammatory stress as it is a regulator of the peptide transporter PEPT1/SLC15A136 involved in the intake of bacterial peptides responsible for the production of accessory immune molecules.37 The sequence from chronic inflammation to cancerous evolution remains to be thoroughly explored. Chronic inflammation produces genotoxic reactive oxygen and nitrogen species for epithelial cells.38 It may also induce the recruitment of bone marrow‐derived stem cells that contribute to tumourigenesis, as recently shown in a murine model of gastric cancers.39
In conclusion, we report here a crucial role for Cdx2 in inflammation‐induced disruption and repair of the colonic epithelium and subsequently in colon cancer induced by chronic inflammation. Cdx1‐/‐ mice, unlike Cdx2+/− animals, are not hypersensitive to acute inflammation, revealing different functions of these two homeobox genes in the gut. The present study also uncovers a connection between the CDX1 and CDX2 proteins and members of the SMAD family of mediator of the BMP/TGFβ pathways. In line with this, the different effects of the Cdx1 and Cdx2 homeobox genes in the model of intestinal inflammation by DSS are mirrored by different effects of the CDX2 and CDX1 proteins in interaction with SMAD3.
This work was supported by INSERM and the Association François Aupetit. A.C. is a fellow of the INSERM/Région‐Alsace and of the Ligue contre le Cancer. I.G. is a fellow of the Association pour la Recherche sur le Cancer and of the Institut National contre le Cancer. We thank Dr J. Deschamps (Hubrecht Lab, Utrecht) for providing the Cdx1‐/‐ mice, Dr A. Atfi (Inserm U673, Paris) for the HCT116 Samd4‐/‐ cells and Smad‐expressing plasmids and Dr C. Foltzer‐Jourdainne (U682, Strasbourg) for statistical analyses.
DSS - dextran sodium sulfate
Competing interests: None.