Epithelial-specific genetic ablation of Itga6 affects intestinal hemidesmosomes

To assay the role of α6 integrin in intestinal homeostasis, we first induced a complete deletion of Itga6 in IECs using the Cre-lox approach18 (α6^ΔIEC; figure 1A and online supplementary figure S1). The resulting animals displayed abnormally loose and viscous stools, and frequently developed a rectal prolapse (figure 1B). To determine whether this phenotype was linked to hemidesmosome alterations, we examined the β4 integrin chain, the α6 hemidesmosome heterodimerising partner and found that both chains were removed from IECs at all stages examined (see figure 1C–E and online supplementary figure S2A, B). By contrast, epithelial expression of integrin β1, which can also heterodimerise with other α chains, did not vary (see online supplementary figure S2C), confirming that defects observed in these mice originated from a loss of the α6β4 integrin in the epithelium.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Efficient deletion of Itga6 in α6^ΔIEC mice results in compromised hemidesmosomes and epithelial fragility. (A) Strategy to generate an intestinal epithelium-specific Itga6 knockout (for details, see online supplementary figure S1A). The Itga6 floxed allele (α6 ^fl ) was obtained after insertion of two loxP cassettes (green triangles) at the Itga6 3′ end including the TM and the cytoplasmic A and B (α6A; α6B) exons. Crossing of the α6 ^fl/fl mice with the transgenic Villin-Cre line results in a truncated Itga6 copy, denoted α6^ΔIEC. (B) Morphology of the colorectal region in 15-week-old WT and α6^ΔIEC mice. White stars indicate stools. Scale bars, 5 mm. (C–H) Immunodetection of hemidesmosome markers in the colon of E16.5 embryos (C and D) and in intestinal segments of mice aged 9–16 weeks (E–H); (E) rectum; (F and G) jejunum; (H) colon; 4',6-diamidino-2-phenylindole (DAPI) marks nuclei (blue). (C and E) α6-integrin chain and (D) β4-integrin chain (green) with the mucin Muc2 (red). The remaining signal in (C and E) corresponds to α6 integrin in blood vessels, confirming the specificity of the deletion in the epithelium. (F) Plectin and (G) K8/K18 intermediate filaments (green), with collagen IV (red). (H) Laminin-γ2 chain (green). White arrows, epithelium/lamina propria interface; yellow arrowheads, hemidesmosome patches; stars, areas of epithelial detachment in mutants. Scale bars, 50 µm. (I) Histological analysis of the colon from 3-week-old mice; bracket, detached cells from the surface epithelium. Scale bar, 100 µm. (J) Scattered dot plots showing the protein concentration of epithelial cell lysates obtained from small intestinal tissue of pups aged 2 (P2) and 14 (P14) days submitted to a detachment assay; error bars, SD; *p<0.05, **p<0.01. (K) Scattered dot plots showing the plasma concentration of FITC-dextran (FD4) as a measurement of intestinal permeability in 6-week-old animals fed with FD4; error bars, SD; ****p<10⁻⁴. A, anus; C, colon; e, epithelium; lp, lamina propria; lu, lumen; m, muscle layer; mes, mesenchyme; P, prolapse; R, rectum; TM, transmembrane; w, weeks; WT, wildtype.

We next tested if Itga6 depletion in IECs impaired the assembly of the hemidesmosome component plectin and subsequently of keratins 8/18 (K8/18). Plectin, which interacts with the β4 integrin cytoplasmic domain, was detected in hemidesmosomal basal patches and at the apical membrane of control IECs (figure 1F). In α6^ΔIEC mice, plectin staining was strongly reduced basally but not apically; loss of basal plectin was already observed in late α6^ΔIEC embryos (see figure 1F and online supplementary figure S2D). Moreover, the K8/18 intermediate filament dimers, which normally are concentrated in the plectin patches basally, were diffusely distributed throughout the cytoplasm (figure 1G). As a result, the shape of epithelial cells appeared less columnar and more cuboidal than in controls (figure 1G). Despite these strong defects, the epithelium retained a characteristic polarised distribution of the apical marker villin, a structural component of microvilli and the tight junction marker cingulin/MAP115 (see online supplementary figure S3). Hence, like in the skin where it nucleates type I hemidesmosomes (containing BP180 and BP230),17 ,19–21 α6β4 integrin is essential for the assembly of type II hemidesmosomes (that are lacking BP180 and BP230) present in the intestine, but does not contribute to epithelial polarity.

Loss of intestinal hemidesmosomes impacts on gut barrier function

Loss of α6β4 integrin in the skin affects epithelial integrity.17 ,19–21 In the mutant intestine, collagen IV and plectin costaining revealed that the epithelial layer tends to detach from its BM and the subjacent lamina propria (figure 1F, G), without affecting the distribution of the hemidesmosomal laminin γ2 chain (figure 1H). From 3 weeks after birth, areas of epithelial detachment were observed throughout the mutant intestine (see figure 1I and online supplementary figure S4), suggesting that the disruption of hemidesmosomes results in epithelium weakening and loss of firm anchorage to the BM. To confirm these observations, we used a modified method of IEC dissociation22 in which control and mutant epithelium were subjected to ethylenediaminetetraacetic acid (EDTA) treatment and gentle agitation to assess their ability to resist a slight mechanical stress. Epithelial fragility was significantly increased in the mutant intestines compared with controls as early as 2 days postnatal (P2), and became more prominent at P14 (figure 1J). Alterations affecting the intestinal permeability are frequently observed in patients with IBD and reflect epithelial barrier dysfunction.23 To examine if such dysfunction occurred in mutant mice, we quantified intestinal permeability in vivo by feeding mice with fluorescein isothiocyanate (FITC)-dextran and measuring FITC levels in the blood plasma. Intestinal permeability was significantly increased in α6^ΔIEC mice (figure 1K). Thus, α6 integrin loss in IECs impairs hemidesmosome formation, epithelial morphology and gut barrier function.

α6^ΔIEC Mice develop spontaneous long-standing extensive colitis

As intestinal barrier defects often lead to inflammation,1 we examined its potential occurrence in α6^ΔIEC mice. We observed obvious inflammation in all mutants from 6 to 8 weeks of age as clinically illustrated by loose and/or bloody stools, rectal prolapse and mucus discharge when compared with their control littermates (see figures 1B and 2A). The intestinal wall appeared thickened throughout the colorectal region (figure 2B), with a histological score worsening from 3 weeks after birth (figure 2C). Mutant crypts in the distal colon were on average twice more elongated (figure 2D, F). As early as 3 weeks of age, the proliferation rate was also significantly increased in both colon and rectum of mutants but not in the jejunum as determined by staining for bromodeoxyuridin (BrdU) incorporation (see figure 2E, G and online supplementary figure S5A). These features (figure 2) suggested that α6^ΔIEC mice develop spontaneous long-standing extensive colitis from weaning onwards.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

α6^ΔIEC Mice develop colitis soon after weaning. (A) Representative images of WT and α6^ΔIEC adult colons observed by mini-endoscopy. Black arrow, lesion affecting the mutant mucosa; arrowheads, presence of abundant mucinous material in the lumen of the mutant colon. (B) Longitudinal H&E stained sections of the recto-anal region from animals aged 15 weeks. The bracket in mutants highlights the hyperplasia in mucosa and thickened muscle layers (arrowheads, ulcerated surface). Scale bars, 500 µm. (C) Histological scores displayed as scattered dot plots (error bars, SD) measured in the proximal and distal colons of WT and α6^ΔIEC mice from 2 to 9 weeks (w) of age. (D) Transverse H&E stained sections of the distal colons of WT and α6^ΔIEC mice at 4 weeks. Double arrows, crypt height. Scale bar, 50 µm. (E) Immunohistochemical BrdU detection (brown nuclei) on colon sections of 3-week-old animals; dashed lines and double arrows, area of expanded proliferation in the mutant colon. Scale bar, 50 µm. (F and G) Proliferation and crypt height measurement in colons of WT and α6^ΔIEC mice aged 2 and 3 weeks, and adult stages displayed as scattered dot plots (error bars, SD). (F) Quantification of crypt height. (G) Quantification of BrdU-positive cells. *p<0.05, **p<0.01. e, epithelium; lu, lumen; ns, not significant; w, weeks; WT, wildtype.

Colitis severity in α6^ΔIEC mice is independent of the adaptive immune response

Studies in mice and humans have strongly suggested that intestinal inflammation results from an impaired innate immune response to the gut microbiota, which results in dysregulation of the adaptive immune system.7 ,8 To examine if inflammation in α6^ΔIEC intestine follows this paradigm, we first defined the immunological profile by fluorescence activated cell sorting (FACS) of immune cells isolated from the colonic lamina propria (referred to as lamina propria mononuclear cells). Mutant mice displayed ∼five-fold increase in CD4⁺ T lymphocytes and of several myeloid cells including neutrophils, dendritic cells and monocytes (figure 3A, B). Notably, there was a >40-fold increased recruitment of Ly-6G⁺/CD11b⁺ neutrophils within the colonic mucosa of mutant mice when compared with controls (figure 3B). By contrast, the number of CD8⁺ T lymphocytes and inflammatory monocytes did not vary (figure 3A, B). Histological analysis confirmed the presence of extensive immune cell infiltrates in the mutant colon/rectum/prolapse area (see online supplementary figure S5B, C). Those were mostly identified as CD11b⁺ cells by immunostaining, and were recruited as soon as 4 weeks (figure 3C, C′). Of note, the recruitment of these cells was even stronger at 9 weeks (figure 3D, D′).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Inflammation onset in α6^ΔIEC mice is mediated by myeloid cells recruitment, independently of the adaptive immune system. (A and B) FACS analysis and quantification of immune cell subpopulations present in colonic lamina propria extracts from animals raised in conventional conditions. LPMCs originate from WT and α6^ΔIEC animals aged 9–15 weeks. Results are displayed as scattered dot plots (error bars, SD); each dot represents the number of positive cells present in the extract for each animal. (A) CD4⁺ and CD8⁺ T lymphocytes; (B) Cells of myeloid origin defined as: neutrophils, Ly-6G⁺ CD11b⁺; DC, Ly-6G⁻ CD11c⁺ MHC class II^intermediate; monocytes, Ly-6G⁻ Ly-6C⁺ MHC class II⁻; pro-inflammatory monocytes, Ly-6G⁻ Ly-6C⁺ CD11c^low MHC class II⁺ CD64⁺. (C–E) Immunodetection of CD11b⁺ cells (green) in the colon of WT and α6^ΔIEC mice aged 4 (C) and 9 (D) weeks, and of combined rag1^+/− α6^ΔIEC and rag1^−/− α6^ΔIEC animals (E). Scale bars, 100 µm. (C′–E′) Quantification of the number of CD11b⁺ infiltrating cells per mm² of colon from images illustrated in (C–E), displayed as scattered dot plots (error bars, SD). (C′) WT and α6^ΔIEC mice aged 3 and 4 weeks; (D′) 9-week-old WT and α6^ΔIEC mice; (E′) combined rag1^+/− α6^ΔIEC and rag1^−/− α6^ΔIEC; rag1 α6^ΔIEC mice were raised in SPF conditions. **p<0.01. DC, dendritic cells; e, epithelium; FACS, fluorescence activated cell sorting; Infl. Mono, pro-inflammatory monocytes; lp, lamina propria; LPMCs, lamina propria mononuclear cells; MHC, major histocompatibility complex; Mono., monocytes; Neutro., neutrophils; ns, not significant; SPF, specific pathogen-free; w, weeks.

To formally assess the potential involvement of T lymphocytes and B lymphocytes on disease onset, we crossed α6^ΔIEC mice with rag1 ^−/− mice, which lack mature B lymphocytes and T lymphocytes.24 Remarkably, histopathological analyses revealed that rag1 ^−/− α6^ΔIEC double mutant mice developed inflammation similar to that of single rag1^+/− α6^ΔIEC mutants, with the presence of large immune infiltrates throughout the colorectal mucosa (see online supplementary figure S5C) and high levels of CD11b⁺ myeloid cells (figure 3E, E′). Collectively, these results demonstrate that the onset of colitis is independent of T lymphocyte and B lymphocyte activation.

We next examined whether the inflammation observed in α6^ΔIEC mice was associated with the MyD88 pathway, a major player of the innate immune response. Colitis occurred similarly in both MyD88 ^−/− α6^ΔIEC and MyD88 ^+/− α6^ΔIEC mutant animals suggesting that MyD88 signalling was not primarily required to initiate inflammation in α6^ΔIEC intestine (see online supplementary figure S5D). Hence, inflammation onset did not involve classical toll-like receptor (TLR) pathways going through MyD88.

Early onset of colitis is linked to epithelial caspase-1 activation and IL-18 secretion in α6^ΔIEC mice

To further define the pathways mediating inflammation, we next examined the levels of well-characterised pro-inflammatory cytokines involved in IBD, released by colon explants cultured for 24 hours (see online supplementary materials and methods). Cytokine screening revealed that IL-1β and IL-18 levels were the most prominently increased in mutants compared with controls (see figure 4A and online supplementary figure S6A). Among eight additional cytokines that we assayed, IFNγ, IL-17 and IL-22 levels were also increased in α6^ΔIEC mice, though less prominently (see online supplementary figure S6A, B).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Colitis initiation in α6^ΔIEC mice is triggered by strong epithelial IL-18 secretion and colitis worsening by IL-1β hypersecretion. (A and B) Scattered dot plot quantification of IL-1β and IL-18 levels measured by ELISA starting from colon explants of WT and α6^ΔIEC mice at 9 weeks (A) and at 2, 3 or 4 weeks of age (B) (error bars, SD; ns, not significant; *p<0.05, **p<0.01, ***p<0.001). (C) Western blot analysis performed on protein extracts obtained from an enriched epithelial fraction (enriched IECs) or from whole colon segments (total colon lysates) of 2-week-old WT and α6^ΔIEC mice. Representative blots of three independent experiments performed on at least five animals per group are illustrated. Western blot with GAPDH and β-Tubulin are presented as loading controls. Casp1, caspase-1; enriched IECs, enriched intestinal epithelial cells; GAPDH, glyceraldehyde 3-phosphate dehydrogenase; w, weeks.

To better characterise the sequence of events occurring during disease onset, we specifically focused on the timing of IL-1β and IL-18 increase relative to the earliest abnormalities observed in mutant mice (see figure 2C–G). In particular, we examined animals aged 2–3 weeks when the epithelial barrier, the immune system and the microbiota are progressively established.25 We observed a clear difference between the initial secretion of both cytokines, since IL-18 was already significantly increased in the mutant colons at 2 weeks, whereas IL-1β was not detectable at this time point (figure 4B). From 3 weeks onwards, the secretion of both cytokines was significantly increased in the α6^ΔIEC colons (figure 4B).

To identify the most likely cellular source of IL-18 and IL-1β, we used western blot analysis, to screen their expression in whole protein extracts (total colon lysates) and in an enriched epithelial fraction (enriched IECs) obtained by scraping of the colorectal mucosa (see figure 4C and online supplementary figure S7). Since both cytokines are produced as inactive forms (pro-IL-18 and pro-IL-1β) that must be cleaved by caspase-1 to be active and secreted, we also checked the expression of active caspase-1 (resulting from pro-caspase-1 cleavage). Expression levels of pro-IL-18, pro-IL-1β and pro-caspase-1 inactive forms were similar in wildtype (WT) and α6^ΔIEC extracts, regardless of the stage (2 or 3 weeks) or the source of the extract (total colon vs enriched IECs) (see figure 4C and online supplementary figure S7). By contrast, a strong increase of cleaved IL-18 and cleaved caspase-1 was observed in α6^ΔIEC enriched IECs as early as 2 weeks (figure 4C) and was maintained at 3 weeks (see online supplementary figure S7), but was barely detectable in total mutant colon lysates. Cleaved-IL-1β was not or barely detected.

Altogether, our data establish that IL-18 is specifically activated in α6 deficient IECs prior to any clinical and histological signs of colitis, and that its epithelial release precedes the secretion of IL-1β. Furthermore, since the active form of IL-1β was not detected in the mutant IECs, it suggests that it was probably produced by immune cells infiltrating the mucosa.

α6^ΔIEC Colons display defective mucus barrier

One of the factors exacerbating inflammation in patients with IBD comes from defects in the mucus layer that protects from microbial translocation.3 ,26 Interestingly, very recent data have linked epithelial IL-18 signalling to mucus production by goblet cells (GCs) and to GC maturation.27

To examine the properties of the mucus layer in α6^ΔIEC mice, we used periodic acid Schiff/alcian blue (PAS/AB) staining. We found that the PAS⁺ GC which faced the lumen displayed a larger apical compartment or theca in 4-week-old mutant distal colon (see online supplementary figure S8A–D), as also observed in keratin K8-deficient mice.28 In addition, the number of PAS⁺ GC detected at the base of the mutant crypts was significantly increased, while the average number of PAS⁺ GC over crypt height was unchanged (see online supplementary figure S8A–D). From 6 weeks of age, Muc2 and PAS/AB staining revealed abundant mucus discharge presenting altered properties in the mutant colonic lumen (see online supplementary figure S8E, F), although the ratio GC/total epithelial cells was similar (59.0±4.6% for controls and 59.7±2.7% for α6^ΔIEC mice, n=3). At later stages, the normally bacteria-free inner mucus layer appeared irregular in α6^ΔIEC mice (see online supplementary figure S8G). To elucidate the potential cause of the mucus defects, we analysed the distribution of the anion transporter SLC26A3 and the Na⁺/H⁺ exchanger NHE3, described as being involved in the formation and maintenance of the mucus layer,29–31 and known to be affected in the K8^−/− mutant mice.28 We found that SLC26A3 and to a lesser extent NHE3 were less abundant at the apical surface of α6^ΔIEC colons (see online supplementary figure S8H, H′), arguing for their involvement in the observed mucus defects in association with the altered keratin distribution.

Antibiotic treatment improves disease severity

Alterations of the mucus layer can favour a physical interaction of bacteria with the epithelium.29 ,30 ,32 We thus examined the spatial segregation of the microbiota from the epithelium in the distal colon, using fluorescence in situ hybridisation (FISH) to detect bacterial 16S ribosomal RNA, and Muc2-immunostaining to delineate the mucin layer and the epithelial border. At 3 weeks of age, when histological defects and cytokine levels were already abnormal, bacteria were clearly physically separated from IECs (see online supplementary figure S9A). By contrast, bacteria were attached to the mutant colonic epithelial surface in older diseased mice as judged by the presence of bacterial 16S rRNA in direct contact with IECs (see online supplementary figure S9B).

Given the close contacts between bacteria and mutant IECs, we next assessed whether bacteria had a role in eliciting colitis in α6^ΔIEC mice. We treated 9-week-old animals with a cocktail of broad-spectrum antibiotics prior to collecting intestinal explants in order to deplete a large range of bacterial species. Overall, antibiotics reduced the levels of IL-1β to those observed in untreated controls and halved those observed for IL-18 (see figure 5A and online supplementary figure S9C), and had no effects on the levels of IFNγ, IL-17 and IL-22 (see online supplementary figure S9C). In addition, antibiotics improved disease severity by reducing the number of CD11b⁺ cells infiltrated into the mutant colonic mucosa (figure 5B), whereas they had no major effect on CD4⁺ T cells (figure 5B). Consequently, antibiotics reduced the colitis histological score in the mutant colon (figure 5C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Colitis in α6^ΔIEC mice is partially improved by antibiotic treatment. (A–C) Scattered dot plot representations showing the quantification of cytokine levels measured by ELISA starting from colon explants (A), of the CD11b⁺ and CD4⁺ cells infiltrating the colonic mucosa (B), and of histological scores (C), in 9-week-old WT and α6^ΔIEC mice treated with antibiotics (+ATB) or not. In each scattered dot plot, error bars represent SD; *p<0.05, **p<0.01, ***p<0.001. (A) ELISA quantification of IL-1β and IL-18 levels. (B) Quantification of CD11b⁺ and CD4⁺ infiltrating cells per surface of colon. (C) Colitis histological scores established in the proximal and distal colons. +ATB, with antibiotics; ns, not significant.

Overall, our results show that bacteria play a role in worsening colitis in α6^ΔIEC mice, but disease onset precedes bacterial attachment to the intestinal epithelial layer. In addition, the fact that antibiotic treatment only partially restored a normal healthy mucosa with normal IL-18 levels further supports the notion that IL-18 plays a key role in triggering inflammation.

α6 Integrin loss in adults recapitulates colitis occurrence observed in α6^ΔIEC animals

To discriminate between direct and indirect effects of α6 integrin ablation, we controlled the timing of Itga6 ablation in IECs by using the TAM-inducible Cre-ER^T2 system (α6^ΔIEC-TAM line).18 We treated adult mice with TAM at 8 weeks—stage at which animals have a fully mature intestine and immune system—and analysed them 2 weeks later (figure 6A). Immunodetection of α6β4 integrin in α6^ΔIEC-TAM colons revealed an efficient and homogeneous removal of the α6 and β4 signals in the mutant IECs (figure 6B). Interestingly, α6^ΔIEC-TAM mutants displayed colorectal swelling and loose stools (figure 6C), and an increased intestinal permeability (figure 6D) similar to those observed in the α6^ΔIEC line. To better define how early molecular signs of colitis arise, we examined animals at 6, 10 and 15 days post-TAM administration. At day 6 post-TAM, mutant colons displayed significant epithelial detachment and an increase in IL-18 secretion (see figure 6E and online supplementary figure S10A, B), as what also observed in α6^ΔIEC animals (see figures 1I, 4B and online supplementary figure S4B, C). α6^ΔIEC-TAM Mutant mice started to develop hyperplasia (assessed by the proliferative marker Ki-67), CD11b⁺ cell infiltration and increased IL-1β secretion at days 10 and 15 post-TAM (see figure 6E–F′ and online supplementary figure S10A–D), with no difference in T lymphocyte subsets (figure 6G, H). In line with our previous findings in the constitutive α6^ΔIEC model, antibiotic treatment applied prior to TAM administration reduced signs of inflammation (see figure 6E–F′ and online supplementary figure S10), without reducing epithelial detachment and restoring IL-18 secretion to control levels (see figure 6E and online supplementary figure S10). Thus, α6^ΔIEC-TAM mice phenocopied the α6^ΔIEC line.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6

Colorectal inflammation occurs quickly after α6 integrin ablation in adult IECs. (A) Experimental procedure of TAM treatment (red arrows indicate the days of TAM administration) and subsequent analysis (black arrow) of WT and α6^ΔIEC-TAM mice, illustrated for the time point 15 days post-TAM administration. Animal age is indicated in weeks. (A–D and F–H) Eight-week-old WT and α6^ΔIEC-TAM mice were treated with TAM and analysed 15 days after the first TAM gavage. (B) Immunodetection of α6-integrin and β4-integrin chains (green) on colon sections of WT and α6^ΔIEC-TAM. DAPI marks nuclei (blue). Scale bars, 50 µm. (C) Morphology of the colorectal region of WT and α6^ΔIEC-TAM mice. Scale bars, 5 mm. (D) Scattered dot plots showing the plasma concentration of FITC-dextran (FD4) in WT and α6^ΔIEC-TAM mice fed with FD4; error bars, SD. (E) ELISA quantification of the levels of IL-1β and IL-18 secreted by colon explants cultured for 24 hours. Samples were analysed 6, 10 and 15 days after the first TAM gavage in WT and α6^ΔIEC-TAM mice treated with antibiotics or not. Results are displayed as scattered dot plots; error bars, SD; ns, not significant. (F and F′) Immunodetection (F) and quantification (F′) of CD11b⁺ cells (green) in the colon of 10-week-old TAM treated WT and α6^ΔIEC-TAM mice, 15 days post TAM gavage. Scale bar, 100 µm. (G and H) Quantification by FACS analysis of the immune cell subpopulations present in LPMCs of 10-week-old TAM treated WT and α6^ΔIEC-TAM mice, 15 days post TAM gavage. (G) CD4⁺ and CD8⁺ T lymphocytes; (H) Cells of myeloid origin defined as Ly-6G⁺ CD11b⁺. Results are displayed as scattered dot plots (error bars, SD); each dot represents the number of positive cells present in the extract for each animal. *p<0.05; **p<0.01; ***p<0.001. A, anus; Asterisks (*), stools; +ATB, with antibiotics; C, colon; Ca, caecum; d, day; e, epithelium; FACS, fluorescence activated cell sorting; IECs, intestinal epithelial cells; LPMCs, lamina propria mononuclear cells; lp, lamina propria; lu, lumen; ns, not significant; R, rectum; TAM, tamoxifen; w, week.

We next aimed at further delineating the molecular signature of inflammation upon loss of the α6β4 integrin. To this end, we performed a transcriptome analysis on mRNAs extracted from the rectal compartment of WT and α6^ΔIEC-TAM mice 15 days post-TAM treatment. Microarray data (Affymetrix) were analysed using the ‘Ingenuity Pathway Analysis’ (IPA) which helped us to define molecular interaction networks. The most activated network in α6^ΔIEC-TAM mutants included the gene encoding IL-1β (see online supplementary figure S11A), confirming the leading role of the latter regulatory circuit in disease severity. Additional genes with ascribed function related to this ‘IL-1β-inflammatory network’ encoded pro-inflammatory molecules (IL1-RL1, CXCL5/CXCL6 and CXCR2) or extracellular matrix (ECM) remodelling factors (such as MMP10). Their differential expression was validated by RT-qPCR (see online supplementary figure S11B).

Altogether the phenocopy of colitis in adults having an already well-established immune system and microbiota highlights that the origin of colitis in the absence of α6 integrin is primarily associated with IEC detachment from the BM, secretion of epithelial IL-18 and exacerbation due to bacteria.

The spontaneous and fully penetrant development of adenocarcinomas in α6^ΔIEC mice requires activation of B lymphocytes and T lymphocytes

Some animal models of IBD can develop adenocarcinomas, although at a low prevalence unless treated with dextran sodium sulfate and azoxymethane.7 ,33 By contrast, we observed that all α6^ΔIEC mice spontaneously developed colorectal adenocarcinomas by 1 year of age, and without any further chemical treatment. Adenocarcinomas and/or high-grade dysplasia were mostly located in the rectal prolapse (figure 7A), and were found very rarely in the small intestine (two out of the eight cases analysed). Histopathological analyses revealed different degrees of mucosal infiltration (figure 7B) and numerous Ki-67 positive cells were observed in invading glands within the rectal prolapse (figure 7C). Among 14 mutant mice aged 48–80 weeks, all had adenocarcinomas (figure 7D). Interestingly, α6^ΔIEC-TAM mice treated with TAM at 8 weeks and analysed 1 year later or more developed also adenocarcinomas / high grade dysplasia (five out of the seven cases analysed) or at least low grade dysplasia (2/7) in the rectal prolapse, indicating that the depletion of α6 integrin even when induced in the mature intestine can trigger colitis-associated carcinogenesis (figure 7D). We next used positron emission tomography (PET scan) to test for the presence of metastasis. We detected a high metabolic level of fluorodeoxyglucose (FDG) uptake in the whole large intestine of 1-year-old α6^ΔIEC mice, but failed to observe distant metastasis (figure 7E).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7

All α6^ΔIEC mice spontaneously develop colitis-associated adenocarcinomas. (A–G) Tumour features in ≥1-year-old α6^ΔIEC mutant mice. (A) Wide-field view (top) and histological section (bottom) of a large prolapse with an infiltrating mucinous adenocarcinoma invading the submucosa and muscle layers (m). Dotted line, region comprising tumour lesions. Scale bars: 5 mm (top); 500 µm (bottom). (B) Views of prolapse-associated tumours: (left) intraepithelial adenocarcinoma and (right) highly infiltrating adenocarcinoma (dotted lines, arrows). Higher magnifications are shown in the lower panels. Scale bars: 100 µm. (C) Ki67 immunostaining (brown nuclei) on rectal sections from 1-year-old WT and α6^ΔIEC mice. Ki67⁺ cells are located at the base of the crypts (bar), and in the invading glands of the prolapse associated-adenocarcinoma (right panel, arrows). Scale bars: 100 µm. (D) Schematic representation showing the repartition of 1-year-old affected mice according to their genotype and the type of intestinal lesions they displayed; numbers mentioned in histogram columns and percentage (y-axis) of affected mice as well as the lesion types are indicated. The genotype of mice and the housing conditions (conventional (conv.) vs SPF) are indicated under the chart. (E) FDG-PET images of WT and α6^ΔIEC mice showing a significant FDG uptake in the mutant large intestine and the rectal prolapse area (arrows). (F) Section through the recto-anal region of a combined rag1^−/− α6^ΔIEC mutant showing a low grade dysplasia (dotted line, arrows). Lower panel: high magnification of the dysplasia. Scale bars: 100 µm. (G) Immunodetection of laminin γ2 chain (green) in WT rectum and α6^ΔIEC prolapse. DAPI (in blue) marks nuclei. Arrowheads, basement membrane underlying the surface epithelium. Arrows, increased deposits of laminin γ2 in the basement membrane and around tumour areas in α6^ΔIEC prolapse. Asterisks (*), invasive gland infiltrated into the sm. Scale bar: 100 µm. ADK, adenocarcinoma; Dys, low grade dysplasia; e, epithelium; FDG, fluorodeoxyglucose; Inf, inflammation; lp, lamina propria; lu, lumen; m, muscle layer; sm, submucosa; SPF, specific pathogen-free; PET, positron emission tomography.

As reported above, the adaptive immune system does not play an obvious role in the onset of colitis. To define whether or not activation of the adaptive immune system might also be dispensable for progression to adenocarcinomas, we compared single α6^ΔIEC and rag1^−/− α6^ΔIEC double mutant mice raised in specific pathogen-free (SPF) conditions. Out of eight α6^ΔIEC animals aged 45–76 weeks, three had developed an adenocarcinoma, one displayed a severe dysplasia, whereas the four remaining had inflammation (figure 7D). By contrast, none of the 50–64 weeks old rag1^−/− α6^ΔIEC double mutants (6/6) developed an adenocarcinoma; instead they showed very discrete low grade dysplasia lesions (4/6) and chronic inflammation (6/6) (figure 7D, F), suggesting a critical role for the T and B cell activation on disease progression.

To know if local alterations of the microenvironment, in particular of the laminins, might contribute to tumour development, WT and mutant rectal tissues were immunostained with laminin γ2, the hemidesmosome ligand of α6β4, that is known to be overexpressed and associated with poor prognosis in CRC.34 Large laminin γ2 deposits were detected throughout the whole tumoural area within the rectal prolapse in mutant mice while the signal was restricted to the BM at the upper part of the crypts in controls (figure 7G).

Finally, to better understand the mechanisms underlying tumour aggressiveness, we defined the specific transcriptomic signature associated with carcinogenesis in the α6^ΔIEC model by comparing gene expression in three distinct samples: (1) adenocarcinoma-comprising areas from the mutant prolapse, (2) inflamed but non-cancerous adjacent rectal mucosa and (3) rectal mucosa obtained from WT mice. A first IPA analysis allowed to identify significantly overexpressed or underexpressed genes in the tumour areas versus the inflamed mucosa (Student's t-test, p value ≤0.001; fold change ≥+2 or ≤−2) and to define the most activated gene network in the tumours (see online supplementary figure S12A). Many molecules involved in ECM composition and degradation/remodelling, as well as inflammatory components belong to this genes network. Interestingly, several of them have been described as being involved in colorectal carcinogenesis such as MMP9, MMP13 or TIMP1 (for review see ref. 35) or in inflammatory pathways (for review see ref. 36). Moreover, as previously described for the inflammation worsening, network analysis emphasised the prominent role of IL-1 cytokines and IL-1R signalling, suggesting that this inflammatory signature was required for tumour progression.

We further characterised the molecular signature of inflamed or cancerous lesions comparing them to the normal rectal mucosa by using a second IPA analysis which revealed the different categories of gene functions significantly affected. The results are summarised in online supplementary table S1 and figure S12B.

Altogether, the intestinal inflammation observed in α6^ΔIEC mice results from epithelial detachment and IL-18 secretion leading to defective mucus barrier, independently of T lymphocyte and B lymphocyte activation and of MyD88 signalling. Furthermore, our results suggest that intestinal tumourigenesis is likely caused by continuous IL-1 signalling together with BM remodelling and B cell and T cell activation.