Identification of a genetic locus controlling bacteria-driven colitis and associated cancer through effects on innate inflammation

As previously described (Erdman et al., 2003a,b), we found C57BL/6.RAG^−/− (B6.RAG) mice to be strongly resistant to H. hepaticus–induced typhlocolitis, showing only limited inflammation of the cecum, reduced splenomegaly, and no colitis compared with the severe inflammation observed in H. hepaticus–infected 129S6.RAG^−/− (129.RAG) mice (Fig. 1). The telomeric part of chromosome 3 has been identified in other models of colitis (using IL-10^−/− and Gnai2^−/− mice) as a major susceptibility locus (Beckwith et al., 2005; Borm et al., 2005). Consequently, we generated a novel congenic line on the 129.RAG background harboring a large interval of B6 origin on chromosome 3 (>40 Mb) delimited by the microsatellite markers D3Mit103 and D3Mit45 (Fig. 1 A). Upon H. hepaticus infection, phenotypic analysis of intercrossed mice from this 129S6.B6-(D3Mit103-D3Mit45).Rag2^−/− congenic line (abbreviated 129.C3B.RAG or C3B) revealed that the B6 chromosome 3 interval conferred significant protection from colonic inflammation and splenomegaly but not from cecal inflammation (Fig. 1, B–D). Moreover, the similar protection between heterozygote 129/B6 and homozygote B6/B6 congenic mice indicated a clear dominant effect of the B6 chromosome 3 locus (Fig. 1, B–D). In these mice, intestinal inflammation was strongly reduced in the proximal and distal parts of the colon compared with littermate or 129.RAG controls. Some degree of inflammation could still be found in the mid colon (unpublished data). The differential phenotype in response to chronic H. hepaticus infection was not linked to a change in bacterial burden, as cecal colonization by H. hepaticus was found to be equivalent in infected B6.RAG, 129.RAG, and intercrossed 129.C3B.RAG mice (unpublished data). Using a lymphocyte-replete model of H. hepaticus infection in IL-10^−/− mice, we found a marked reduction in colitis in B6.IL10^−/− compared with 129.IL10^−/− mice as described for the Cdcs 1 locus (Beckwith et al., 2005; Fig. 2). These results indicate that resistance to bacteria-induced colitis in the B6 compared with 129 strain extends to innate and lymphocyte replete models.

The 129.C3BR1.RAG (R1) recombinant was identified during the backcrossing process of the C3B congenic line. It retained the second half of the congenic interval, with a unique recombination event between the close microsatellite markers D3Mit348 and D3Mit319 (Table 1 and Fig. 3 A). Analysis of heterozygote (not depicted) and homozygote R1 mice after H. hepaticus infection showed an equivalent protected phenotype compared with C3B mice, with significantly reduced colitis and splenomegaly (Fig. 3 B). The susceptibility locus was therefore localized in the R1 interval spanning a region of 20 Mb, and the proximal boundary of this locus was delimited by the R1 recombination point.

Other subcongenic recombinants were generated by intercrossing 129.C3B.RAG and 129.RAG mice. In particular, several recombination events were found between the microsatellite markers D3Mit319 and D3Mit357, as exemplified by the R6, R9, and R12 recombinant lines (Table 1 and Fig. 3 A), indicating a probable recombination hotspot. Mice from the R12 line were found to be protected from colitis and splenomegaly upon H. hepaticus infection, whereas the R6 and R9 lines were found to be as susceptible as the 129.RAG strain (Fig. 3, B and C). The distal boundary of the susceptibility locus was therefore delimited by the recombination point between D3Mit319 and D3Mit357, excluding the large telomeric B6 region harbored by the R6 and R9 lines.

Taking advantage of the recombination hotspot between the markers D3Mit319 and D3Mit357, we were able to generate new recombinants from the R1 line, retaining only a small region from B6 origin. The reduced interval was centered on the marker D3Mit319, excluding D3Mit348 and D3Mit357, as defined in the R17 and R21 recombinant lines (Table 1 and Fig. 3 A). Upon H. hepaticus infection, mice from R17 and R21 lines were protected from colitis and splenomegaly (Fig. 3, B and C), confirming the previous definition of the susceptibility locus.

Precise mapping of the recombination events using SNP markers localized the proximal R1 point between the SNPs rs31622216 and rs31106293 in an interval spanning 147 Kb (Table 1 and Fig. 3 D). The distal hotspot of recombination was confirmed by mapping of R6, R9, R12, R17, and R21 recombination points, which revealed a common interval between the SNPs rs33862828 and rs30910985, spanning 202 Kb (Table 1 and Fig. 3 D). The susceptibility interval was then delimited by the SNPs rs31622216 and rs30910985, spanning 2.13 Mb (Fig. 3 D).

In the course of SNP genotyping for the R12 line, a new recombination event was identified and mapped between the microsatellite marker D3Mit319 and the SNP rs31282270 (Table 1 and Fig. 3 D). As mice from this new R12X line were protected upon chronic H. hepaticus infection (Fig. 3, B and C), the distal region containing the genes Pitx2 and Enpep was excluded from the susceptibility locus. The maximal interval was therefore delimited by the SNPs rs31622216 and rs31282270, spanning 1.71 Mb (Fig. 3 D), which define the Hiccs locus. Evaluation of cytokine production by lamina propria leukocytes (LPLs) isolated from colons of protected R17 and susceptible R9 mice 8 wk after H. hepaticus infection demonstrated that protection correlated with reduced secretion of proinflammatory cytokines including IL-17, IL-22, IFN-γ, and MIP-1α (Fig. 4).

Analysis of the 1.71 Mb susceptibility interval using the Ensembl and National Center for Biotechnology Information databases identified a cluster of known genes between the SNP rs31622216 and the microsatellite D3Mit319, as well as an empty region (“gene desert”) between D3Mit319 and the SNP rs31282270 (Fig. 3 D). The cluster included four genes of known function (Neurogenin2, Tifa, Larp7, and Alpk1), three genes identified by mRNA only (4930422G04Rik [abbreviated 493Rik], Ap1ar, and 5730508B09Rik [573Rik]), one hypothetical protein (LOC100043382), and a set of microRNAs (Mir302a-d and Mir367), as shown in Table 2.

We took a systematic approach for sequencing of the genes and determining their expression profile in the colon. Sequences for four bacterial artificial chromosome (BAC) clones of 129 origin covering the cluster of known genes in the susceptibility interval were determined using Illumina Solexa sequencing. The sequence of each BAC has been annotated and submitted to GenBank (accession nos. JQ664734, JQ664735, JQ664736, and JQ664737 for 416G21, 225D04, 164I03, and 32B05, respectively). The BAC assembly, covering 530 Kb, revealed >5,000 SNPs compared with the published B6 sequence, indicating that the 129 strain was quite polymorphic for this region compared with the B6 strain (unpublished data). More precisely, we were able to identify nonsynonymous polymorphisms in several of the candidate genes, as well as two SNPs in the microRNA sequences (Table 2). The genes Alpk1 and 493Rik were of particular interest as they showed 17 and 25 coding differences between B6 and 129 backgrounds, respectively (Table 2 and Table S1). We confirmed the exonic SNPs by sequencing PCR products from 129 and B6 cDNA, or genomic DNA in case of the microRNAs. We also found that these nonsynonymous SNPs were identical between the 129 and C3H/HeJ strains (Table S1), the latter being the susceptible background in the IL-10^−/− colitis model (Beckwith et al., 2005).

To complement the sequencing approach, extensive expression analysis was performed for all the known genes in the susceptibility interval. Levels of expression were determined by quantitative (Q) PCR on mRNA extracted from colon tissue homogenates, obtained from both uninfected and H. hepaticus–infected mice of resistant (R17) and susceptible (R9) recombinant lines (Fig. 5). We found significantly higher expression of Alpk1 and lower expression of Ap1ar and 573Rik in colon homogenates of uninfected and infected susceptible mice compared with resistant mice (Fig. 5). Of note, we could not detect Neurogenin2 or LOC100043382 mRNAs in the colon or cecum (unpublished data). Finally, no difference in expression of the microRNAs Mir302a-d and Mir367 was found in colon homogenates between resistant (R17) and susceptible (R9) mice, with or without H. hepaticus infection (unpublished data).

To gain insight into the mechanism of action of Hiccs and to identify the cells involved, reciprocal BM chimeras were generated using 129.RAG and 129.C3B.RAG mice (Fig. 6 A) or using protected (R17) and susceptible (R9) recombinant lines (Fig. 6 B). After reconstitution of the innate immune system by donor BM cells, chimeric mice were infected with H. hepaticus for 8–10 wk. Both approaches clearly illustrated the transfer of phenotype by the hematopoietic compartment. Indeed, susceptible recipients receiving BM from protected donors and infected with H. hepaticus exhibited significantly decreased colonic inflammation and splenomegaly, equivalent to the protected control groups (Fig. 6, A and B). Conversely, R17 or C3B recipients became susceptible to H. hepaticus–induced colitis when reconstituted with BM from R9 or 129.RAG donors, respectively (Fig. 6, A and B). As in previous experiments, inflammation was abrogated in the proximal and distal colon, whereas a reduced inflammatory infiltrate was observed in the mid colon (unpublished data). Severe inflammation of the cecum was observed in all infected chimeras, confirming that the Hiccs locus was not contributing to this phenotype (unpublished data).

As we identified cells of the BM compartment as the site of action of Hiccs, we investigated the expression of the candidate genes by leukocytes in the intestine. Thus, we sorted lamina propria cells isolated from the colon of H. hepaticus–infected recombinant mice using the pan-leukocyte marker CD45, isolated RNA, and assessed expression of each candidate gene by Q-PCR. Increased expression of Alpk1 and 573Rik mRNA was observed in CD45⁺ intestinal leukocytes compared with CD45⁻ nonhematopoietic cells (Fig. 6 C). Moreover, within the CD45⁺ leukocyte population there was increased expression of Alpk1 and 573Rik in cells isolated from susceptible (R9) compared with protected (R17) mice (Fig. 6 C).

The expression profile for the candidate genes was also evaluated in BM-derived macrophages (BMDMs) from both susceptible (R9) and protected (R17) recombinant lines. Alpk1, but not 573Rik, was expressed at higher levels in unstimulated BMDM derived from susceptible versus protected mice (Fig. 6 D). Furthermore, we observed increased expression of Alpk1 and 573Rik in LPS or H. hepaticus–stimulated BMDM compared with unstimulated regardless of their origin (R17 or R9 lines). Tifa and the other candidate genes showed no significant change of their mRNA expression in BMDM, with or without stimulation (Fig. 6 D and not depicted). Collectively these results indicate that susceptibility is determined by genes expressed in hematopoietic cells.

To gain further insight into how Hiccs controls the early inflammatory response, we performed a kinetic analysis of the innate response after H. hepaticus infection in protected and susceptible mice. As early as 3 to 8 d after H. hepaticus infection of susceptible 129.RAG mice, we observed a small increase in number of colonic LPL, as well as signs of systemic inflammation, identified by an increased spleen weight (Fig. 7 A). By day 14 after infection, clear differences in spleen weight and LPL numbers were apparent between susceptible 129.RAG and protected R17 mice (Fig. 7 A).

To monitor the composition of the colon lamina propria during the course of early inflammation, we performed FACS staining for different leukocyte populations: DCs, granulocytes, monocytes/macrophages, and ILC (Fig. 7 B). As early as 3 d after infection, a marked increase in both the frequency and absolute numbers of recruited granulocytes was found in LPL of susceptible 129.RAG mice compared with protected R17 mice (Fig. 7, B and 7C). In contrast, increased infiltration by DC and monocytes/macrophages was detected in susceptible mice only at later time points (Fig. 7 B).

These acute inflammatory changes were accompanied by an early peak of proinflammatory cytokines. This response was apparent by day 3 after infection, as LPL from 129.RAG mice showed increased production of TNF, IL-1β, and IFN-γ compared with LPL from R17 mice (Fig. 7 D). Somewhat unexpectedly, IL-17 levels were higher in R17 mice throughout the early inflammatory response. The reduction in IL-17A production observed in susceptible 129.RAG mice in the early inflammatory response was reversed at later time points (8 wk after infection) where higher levels of these cytokines correlated with disease susceptibility (Fig. 4). Lower levels of proinflammatory cytokines in R17 mice did not seem to reflect an overall hyporesponsiveness of the R17 line toward H. hepaticus, as BMDM of both protected (R17) and susceptible (R9) lines reacted similarly to LPS or H. hepaticus stimulation in terms of cytokine production and IκB phosphorylation (unpublished data).

The IL-23 axis is highly relevant for IBD development and has been shown to be a key pathway in mouse models of colitis (Hue et al., 2006; Kullberg et al., 2006). Early during H. hepaticus infection, we found that levels of IL-23p19 mRNA gradually increased and were also more elevated in LPL from 129.RAG compared with R17 mice (Fig. 7 D).

A novel population of ILCs that produce IFN-γ and IL-17 in response to IL-23 was recently identified to be crucially involved in the H. hepaticus innate colitis model (Buonocore et al., 2010). To examine whether ILC played a major role in the early inflammatory response dependent on Hiccs, 129.RAG and R17 mice were treated with ILC-depleting anti-Thy1 antibody and infected with H. hepaticus for 6 d (Buonocore et al., 2010). The depletion of ILC resulted in an abrogation of early granulocyte recruitment in the susceptible 129.RAG strain (Fig. 7 E). ILCs were found at a similar level in LPL of infected 129.RAG and R17 mice and were depleted equally well by the anti-Thy1 antibody (Fig. 7 B and not depicted). Furthermore, the production of TNF, IL-1β, and IFN-γ was not elevated in LPL from anti-Thy1–treated 129.RAG mice above levels observed in LPL from protected R17 mice (Fig. 7 F). Interestingly, MIP-1α, a chemokine involved in the recruitment of myeloid cells, was also increased in LPL from 129.RAG mice compared with R17 mice early after H. hepaticus infection and was abrogated by depletion of ILC (Fig. 7 F). Together, these results suggested a role for the Hiccs locus in early inflammatory events, such as production of inflammatory cytokines and granulocyte recruitment, that are driven by ILC.

To study if the Hiccs locus was also playing a role in the development of CAC, we tested the chromosome 3 congenic and recombinant lines in a H. hepaticus–driven innate tumor model. The development of epithelial dysplasia and invasive cancer in the colon of H. hepaticus–infected 129.RAG mice has been described to be evident 2 mo after infection (Erdman et al., 2003a,b). However we experienced a later development and lower frequency of neoplasia (unpublished data); therefore, we used a new tumor-promoting protocol that combined H. hepaticus infection and administration of the carcinogen azoxymethane (AOM; Nagamine et al., 2008). Even after 5 mo of H. hepaticus infection, the protected C3B group (pooled C3B and R17 mice) showed reduced susceptibility to colitis and systemic inflammation but not cecal inflammation, compared with the susceptible 129 group (pooled 129.RAG and R9 mice; Fig. 8 A). To assess colonic inflammation and tumor formation after H. hepaticus plus AOM treatment, dissected colons were stained with methylene blue. In mice from the susceptible 129 group, an increased thickening of the mid colon could be observed, as well as protruding lesions in the proximal part, indicative of aberrant crypt foci (Fig. 8 B). These aberrant crypt lesions were significantly decreased in mice from the C3B group and were absent in mice treated with AOM alone (Fig. 8 C), indicating the requirement for inflammation and H. hepaticus infection for the development of the lesions.

Further histological assessment of sequentially cut and hematoxylin/eosin-stained sections revealed the presence of H. hepaticus–induced adenocarcinomas that seem to follow the inflammation-dysplasia-carcinoma sequence as described for human CAC (Ullman and Itzkowitz, 2011). Carcinoma location correlated with the sites of highest inflammation, as illustrated in Fig. 8 D, which shows hyperplastic and heavily inflamed colonic mucosa with a focus of moderately differentiated adenocarcinoma (top left). This comprises several architecturally and cytologically abnormal crypts, which have invaded beyond the muscularis mucosa into the submucosa and have elicited a mild desmoplastic response (Fig. 8 D, bottom left). Carcinomas were found penetrating through the muscularis mucosa but not through the muscularis propria and would be staged in the TNM scoring scheme as pT1 (Sobin and Wittekind, 1997). Tumors were present throughout the colon with predominance in the rectum and correlated with foci of inflammation. Immunohistochemical staining of areas of CAC showed an aberrant expression of proliferation markers, such as Ki-67 and Cyclin D1, in the upper part of the crypts as well as an increased expression of COX2 (unpublished data).

Invasive colon carcinomas were found in 75% (12/16) of mice from the susceptible 129 group (pooled 129.RAG and R9 mice) treated with H. hepaticus plus AOM, with a mean of 1.2 carcinoma (range 0–4) per mouse (Fig. 8, E and F). Invading foci were small, with a diameter between 0.1 and 3 mm (mean size 0.81 ± 0.18 mm SEM). In contrast, in the protected C3B group (pooled 129.C3B.RAG and R17 mice) the incidence rate of colon tumors was reduced to 29% (4/14) and the mean tumor number was decreased to 0.3 carcinoma/mouse (Fig. 8, E and F). However, the mean tumor size (0.57 ± 0.21 mm SEM) in the protected C3B group was not significantly different from the 129 group and the severity of the residual tumors was similar, still proceeding to invasive adenocarcinoma (Fig. 8 F). Within the protected C3B group, the tumor incidence of the R17 recombinant mice was 14% (1/7) and therefore comparable or even lower than in the 129.C3B.RAG mice carrying the longer congenic interval at 42% (3/7). In summary, the B6 chromosome 3 congenic interval defining the Hiccs locus leads to a lower incidence rate and frequency of carcinoma after H. hepaticus infection and AOM administration.

Tumorigenesis induced by chronic inflammation has been associated with several human inflammatory diseases, including IBD, but the genetic factors that drive CAC are still largely unknown (Mantovani et al., 2008). Herein, we identify a 1.71-Mb interval on mouse chromosome 3 as a major susceptibility locus for H. hepaticus–induced innate colitis and associated colon cancer on the 129S6.RAG^−/− genetic background, designated Hiccs. This interval encompasses eight known genes and five microRNAs, of which several contain nonsynonymous SNPs and show expression variations in innate cells making them promising candidate genes. Hiccs acts in innate hematopoietic cells and controls ILC-driven initiation of inflammation after H. hepaticus infection. Importantly, the congenic interval is capable of reducing the frequency of colitis-associated invasive adenocarcinoma from 75 to 30%.

Host genetic factors that predispose to a dysregulated interplay between gut microbiota and the innate immune system are believed to play a key role in development of chronic intestinal inflammation and associated colon cancer (Saleh and Trinchieri, 2011). Although a defined bacterial etiology for CAC in humans has not yet been identified, recent GWASs have linked genes involved in bacterial sensing and handling, such as NOD2 or autophagy genes, with susceptibility to IBD (Khor et al., 2011). Similarly, in many mouse models of IBD, chronic inflammation and cancer development are dependent on the genetic background. A decrease in CAC after H. hepaticus infection was previously reported for the B6 background (Erdman et al., 2003a) and similar observations were made for innate immune colitis in the TRUC model between the BALB/c.TRUC and B6.TRUC strains (Ermann et al., 2011). However, to our knowledge, no systematic genome-wide analysis has been conducted in bacteria or chemically induced CAC models. Using a novel tumor induction protocol that combined AOM administration and H. hepaticus infection, we now identify Hiccs, a colitis susceptibility locus on chromosome 3 which contributes to colon tumorigenesis.

Hiccs is part of a larger region previously identified in two other models of colitis, using IL-10^−/− and Gnai2^−/− mice (Beckwith et al., 2005; Borm et al., 2005). The original Cdsc1 locus described in the IL-10^−/− model has been further subdivided into three distinct intervals covering a total of 26 Mb (Bleich et al., 2010). Recently, a study in TRUC mice also highlighted a role for Cdcs1 in an innate colitis model (Ermann et al., 2011). The complex genetic structure of the Cdcs1 locus is not seen in our H. hepaticus innate model, where we found that Hiccs, the reduced 1.7-Mb congenic interval, is sufficient to confer protection. However, Hiccs overlaps with the most telomeric sublocus Cdcs1.3, which covers 2.4 Mb (Bleich et al., 2010), suggesting that this region may represent a strong and common colitogenic determinant in several models of colitis and CAC. Similar to results in H. hepaticus–induced innate colitis, we found that H. hepaticus–induced colitis in lymphocyte-replete mice was less severe in B6.IL10^−/− mice than 129.IL10^−/− mice. Further experiments are required to determine whether the B6 Hiccs locus will be sufficient to confer resistance in this lymphocyte-replete setting. Interestingly, there is a clear synteny between the 1.7-Mb Hiccs interval on mouse chromosome 3 and a region of human chromosome 4 localized in 4q25, with clustering of the same genes in reverse orientation (as indicated by Ensembl database). Although an association of the 4q25 region with IBD has not been identified by recent GWAS (The Wellcome Trust Case Control Consortium, 2007; Anderson et al., 2011; Thompson and Lees, 2011), a disease-relevant role is still possible as GWAS based on SNPs analysis cannot detect all loci, particularly those involving rare variants (Manolio et al., 2009). Further studies are required to assess the potential contribution of the human syntenic interval in IBD and CAC.

Systematic analyses of sequence and expression of the eight known genes and five microRNAs localized in the 1.7-Mb interval identified several candidate genes. So far, the best candidate is Alpk1, as this gene shows 17 nonsynonymous nucleotide variations between 129 and B6 strains, 2 of which are located in the kinase domain, as well as a significantly higher expression level in susceptible mice. Alpk1 encodes an α kinase present in polarized epithelial cell lines that is essential for transport of lipid rafts carrying vesicles from the Golgi apparatus to the plasma membrane (Heine et al., 2005). Alpk1 was also recently identified as a gout-susceptibility factor and shown to play a role in the activation of proinflammatory cytokines in monocytes upon challenge with urate crystals (Wang et al., 2011). However, other genetic elements in the susceptibility interval also need to be considered, for example, 4930422G04Rik, which shows numerous amino acid differences but is not differentially expressed, as well as Ap1ar, which is differentially regulated. Tifa, through its function in TLR downstream signaling by TRAF6 binding would be an interesting candidate (Ea et al., 2004) but does not show strain-specific nonsynonymous polymorphisms or differential expression. As further reduction of the susceptibility interval using congenic recombinant lines is not feasible, a knockin or BAC transgenic approach will be required to conclusively identify the responsible gene. It is notable that upon H. hepaticus infection, B6.RAG mice showed a reduced inflammation of the cecum compared with 129.RAG mice. However, this phenotype was not found in infected 129.C3B.RAG mice or any of the subcongenic recombinants, which displayed a level of typhlitis equivalent to 129.RAG mice. This indicates that genetic loci outside the chromosome 3 interval control typhlitis susceptibility and act independently of Hiccs. Further investigation of these loci requires a genome-wide analysis focused on cecal inflammation.

Nevertheless, we were able to gain insight into the cellular mechanism underlying the susceptibility imparted by the Hiccs locus. In particular, BM chimera experiments demonstrated that colitis protection or susceptibility is transmissible with the genotype of the donor BM. This result parallels recent findings in the TRUC model of innate colitis, where the genetic background was also shown to control colitis severity through hematopoietic cells (Ermann et al., 2011). Innate inflammation in the absence of appropriate immune regulation seems to be a key driver of inflammation-induced cancer. Mediators of this process are proinflammatory cytokines and chemokines that shape a pro-neoplastic microenvironment (Mantovani et al., 2008; Balkwill and Mantovani, 2010; Saleh and Trinchieri, 2011). Recent studies exemplifying procarcinogenic effects of cytokines in the development of CAC showed that antiapoptotic effects of IL-6 on intestinal epithelial cells were mediated through STAT3 signaling (Bollrath et al., 2009; Grivennikov et al., 2009). Interestingly we see a clear effect of Hiccs on early innate immune responses, which could pave the way for later cancer development through protumorigenic inflammatory reactions. As early as 3 d after H. hepaticus infection, we observed a marked increase in neutrophil recruitment and in the production of the proinflammatory cytokines TNF, IL-1β, and IFN-γ in LPL from susceptible 129.RAG mice compared with protected R17 mice. Depletion of TNF was previously shown to abrogate colitis development in the H. hepaticus innate model (Maloy et al., 2003), and more recent studies demonstrated that the presence of Gr1⁺ neutrophils and elevated TNF and nitric oxide in the gut were prerequisites for tumor development (Erdman et al., 2009). TNF is a well known protumorigenic factor in the development of colon cancer in humans, and in mice and anti-TNF therapy has been successfully applied in IBD patients (Popivanova et al., 2008; de Silva et al., 2010). However, in the TRUC model, where dysregulated TNF expression by DC was suggested to be driving colitis and tumor development, late intervention with anti-TNF antibody could not cure colitis (Garrett et al., 2009). Together, these findings suggest an early role for TNF as a mediator that reshapes the microenvironment toward dysplasia and cancer.

Depletion of ILC by anti-Thy1 antibody treatment at the time of H. hepaticus infection blocked early neutrophil recruitment and expression of TNF, IL-1β, and IFN-γ by LPL from susceptible mice, and therefore, ILCs seem to be crucial in the chain of early inflammatory events. Our previous work has shown that both IL-17 and IFN-γ contribute to immune pathology in H. hepaticus–induced innate colitis (Buonocore et al., 2010). Strikingly, analysis of these ILC-derived cytokines very early in the response showed differential kinetics. Thus, IFN-γ was increased at day 3 in susceptible mice and remained high at 8 wk after infection. In contrast, IL-17 levels were low in susceptible mice in the first 14 d after infection but had risen by the chronic phase of disease. The situation was reversed in resistant mice that exhibited low IFN-γ throughout the time course but high early levels of IL-17 that had fallen by day 14. These results raise the interesting possibility that an acute IL-17 response is host protective and that later chronic production may be deleterious. Although ILCs clearly contribute to the amplification of the genetically controlled inflammatory response in susceptible mice, it remains to be determined whether the susceptibility locus acts directly on ILC or indirectly through other innate cells. Our earlier studies identified IL-23 as a pivotal player in chronic innate colitis (Hue et al., 2006). However, resistant and susceptible strains displayed similar early H. hepaticus–induced expression of IL-23 up to day 8, after which IL-23 levels continued to rise in susceptible mice. These results suggest IL-23 may drive the chronicity of the response and that differential production of other mediators, such as TNF or IL-1β, contribute to the Hiccs controlled early innate response.

Recently, several studies identified a role for the microbial flora and flora-derived virulence factors in the development of intestinal cancer in different mouse models (Garrett et al., 2009; Uronis et al., 2009; Wu et al., 2009). Interestingly, H. hepaticus–driven colitis has been shown to be strictly dependent on MyD88 expression by innate hematopoietic cells (Asquith et al., 2010). In contrast, colitis and CAC in the TRUC model seem to be MyD88 independent (Garrett et al., 2009). This illustrates the complexity of the interactions between the innate immune system and the gut microbiota in models where the same susceptibility locus may be involved. Preliminary studies failed to show differences in the frequency of major gut microbial phyla/families in 129.RAG versus congenic R17 mice (unpublished data), suggesting that it is the genetically controlled response to H. hepaticus infection itself that is the trigger for colitis and associated tumorigenesis in this model. In contrast, CAC in TRUC mice is associated with altered bacterial community structure and the presence of distinct colitogenic agents (Garrett et al., 2010). Why H. hepaticus, which is a commensal in many animal facilities, turns into a carcinogen in a particular host genetic background is not clear. An H. hepaticus virulence factor possibly involved in CAC development could be the toxin Cdt, which was shown to induce dysplasia in the liver and is thought to generate double-strand DNA breaks, thereby triggering the DNA damage response via Myc and ATM (Ge et al., 2007; Guerra et al., 2010). Therefore, it is tempting to speculate that besides the deleterious effects of chronic inflammation, H. hepaticus could actively be involved in transformation events leading to invasive adenocarcinoma.

Collectively, the Hiccs locus promotes through its actions in innate immune cells a procarcinogenic environment, where the accumulation of inflammatory cytokines and neutrophils could lead to increased DNA damage and eventually to cancer. Further characterization of the causal genetic element and the relevant molecular pathways that increase susceptibility to colitis and CAC in our mouse model could lead to the discovery of novel interplays between the innate immune system, the colonic epithelium, and the gut microbiota. This, in turn, could offer insights into human IBD and more generally into inflammation-driven cancer, as well as providing novel therapeutic targets for the treatment of these diseases.

All inbred mice used in this study were maintained in an accredited specific pathogen-free facility and experiments conducted in accordance with the UK Scientific Procedures Act (1986) under a Project License (PPL) authorized by the UK Home Office Animal Procedures Committee and approved by the Sir William Dunn School Ethical Review Committee. C57BL/6.IL10^−/− and 129S6.IL10^−/− mice were bred in our facility. Congenic 129S6.B6-(D3Mit103-D3Mit45).Rag2^−/− (abbreviated 129.C3B.RAG) mice were generated by iterative backcrossing to the 129S6.Rag2^−/− (129.RAG) background, using a marker-assisted speed congenic method (Markel et al., 1997). For the iterative backcrossing procedure, a total of 63 polymorphic microsatellite markers, evenly spaced on each chromosome, were used to screen the offspring at each generation. The final 129.C3B.RAG congenic line was established after the eighth generation of backcross. Primer sequences for the microsatellite markers were obtained from the Ensembl or MGI database. Mice intercrossed at seventh and eighth generation of backcross were infected with H. hepaticus and analyzed.

Subcongenic recombinant mice were generated by backcrossing the 129.C3B.RAG line with 129.RAG mice and by extensive genotyping using microsatellite markers and SNPs. Homozygote recombinant lines were established using the initial recombinant founder. To characterize precisely the location of the recombination point, SNPs were chosen in the various intervals based on their position on the chromosome 3 sequence (taken from Ensembl database). The identified SNPs were genotyped in recombinant lines by sequencing of PCR product (Source Bioscience Geneservice) using the following primer pairs: rs31622216, 5′-ATAGCCACCTCTCTAGAGTC-3′ and 5′-GGCAATCCTAAAAGCGTGAC-3′; rs31106293, 5′-GTTAATTGCAGAGCCACATAGC-3′ and 5′-TCTCAGTCCTTATCTCTCAGTGC-3′; rs31282270, 5′-TGGATGTGAGGAGATACTGTAGG-3′ and 5′-ACCTATGGTTACACTCATTCTGC-3′; rs33862828, 5′-TGCATATATTCAGCCAGAGACG-3′ and 5′-CAGATGCACAGCAGGAAGC-3′; and rs30910985, 5′-TAGAGTGGCTGCACCGTAGG-3′ and 5′-TGTGTGGAACCAGATGAAGC-3′.

H. hepaticus NCI-Frederick isolate 1A (strain 51449) was grown on blood agar plates containing trimethoprim, vancomycin, and polymyxin B (Oxoid) under microaerophilic conditions as described previously (Maloy et al., 2003). Mice were fed three times on alternate days with H. hepaticus (0.5–1.0 × 10⁸ CFU) by oral gavage. For the early time course experiments, mice were infected with a single dose of H. hepaticus. To minimize cage effects, susceptible mice (R9 or 129.RAG) were systematically co-housed with protected mice (R17). Recombinant congenic lines were also initially analyzed as heterozygotes with littermate controls. For quantitation of H. hepaticus colonization, cecal contents were collected upon sacrifice and DNA extracted using a DNA Stool kit (QIAGEN). Q-PCR was performed with primers specific for H. hepaticus cdtB gene as previously described (Ge et al., 2001), using the Chromo4 detection system.

Colonic and cecal inflammation in H. hepaticus–infected mice was assessed at 6–8 wk (if not otherwise stated). Samples of proximal, mid, and distal colon, as well as cecum, were fixed in 10% formalin, embedded in paraffin, and cut into 5-µm sections for H&E staining. Scoring of inflammation was performed using previously published criteria (Izcue et al., 2008). In brief, each sample was graded semiquantitatively from 0 to 3 for the four following criteria: degree of epithelial hyperplasia and goblet cell depletion, leukocyte infiltration in the lamina propria, area of tissue affected, and presence of markers of severe inflammation (such as crypt abscesses, submucosal inflammation, and ulcers). Scores for each criterion were added to give a final quantification for intestinal inflammation ranging from 0 to 12. For colitis scores, a mean value of the three sections (proximal, mid, and distal colon) was calculated. Photomicrographs of H&E-stained colon sections were taken with a Coolscope Slide Scanner (Nikon).

LPLs were isolated as previously described (Maloy et al., 2003). In brief, colons were longitudinally opened, cut into 1-cm pieces, and incubated (three times) in RPMI 1640 with 10% FCS and 5 mM EDTA at 37°C to remove epithelial cells. Tissue was then digested with 100 U/ml of type VIII collagenase in two cycles of 1 h at 37°C. The isolated cells were layered on a 30/40/75% Percoll gradient (GE Healthcare), which was centrifuged for 20 min at 600 g, and the 40/75% interface, containing mostly leukocytes, was recovered. For FACS sorting, LPL preparations were stained with an antibody against CD45.2 (BD). CD45-negative and -positive cell populations were sorted on a MoFlow cell sorter (Dako). To identify different cell populations during early colitis, we stained LPL by FACS using antibodies against F4/80 (BM8), CD11b (M1/70), Ly6G (Gr1, RB6-8C5), CD11c (N418), B220 (RA3-6B2), Thy1 (53–2.1), and Sca1 (D7; all from eBioscience).

BACs of 129S7 origin, from the SANGER library bMQ, were identified based on mapping of their extremities by Ensembl database. Four BAC clones, bMQ416G21, bMQ225D4, bMQ164I3, and bMQ32B5 (Source Bioscience Geneservice), were purified using a long-construct Maxiprep kit (QIAGEN). Gene content was confirmed by PCR amplification of the corresponding candidate gene. The BAC preparation was checked for integrity by NotI digestion and migration on agarose gel. The four BACs were pooled and send to the Gene Pool for Illumina Solexa high-throughput sequencing. Sequences of the four BACs were established by the Gene Pool using the published C57BL/6 sequence as framework and were integrated for analysis into Gbrowse by the Computational Biology Research Group. The mRNA sequence of each candidate gene was confirmed in C57BL/6, 129S6, and C3H/HeJ strains by de novo sequencing of PCR products amplified from the corresponding cDNA (or genomic DNA in the case of the candidate microRNAs).

For detection of cytokines, 10⁵ isolated LPL cells were cultured overnight in 100 µl RPMI 1640 with 10% FCS at 37°C. Cytokines in the supernatant were detected using the FlowCytomix Multiplex system for IL-17, IL-22, IFN-γ, IL-1β, IL-6, TNF, and MIP-1α (eBioscience). Samples were analyzed on a FACSCalibur (BD) according to the manufacturer’s instructions.

Isolated tissues (from colon or cecum) or cell preparations were snap-frozen in liquid nitrogen. Tissue material was homogenized in RLT buffer (QIAGEN) with β-mercaptoethanol using a homogenizer (MP Biomedicals). RNA isolation was performed using the RNeasy kit (QIAGEN) including a DNase I digest step. Content and purity of RNA was controlled with a NanoDrop spectrophotometer (Thermo Fisher Scientific). cDNA synthesis was performed using the Superscript III reverse transcription kit from Invitrogen. Quantitative real-time PCR for the candidate genes was done using SYBR green chemistry (SensiMix; Bioline) or the TaqMan system. cDNA samples were analyzed in triplicate using the Chromo4 detection system (Bio-Rad Laboratories) and values were normalized to Hprt. Analysis was performed according to the Δ-Ct method or, to show relative expression, the ΔΔCt method (Pfaffl, 2001).

The following primer pairs were used for the detection of candidate genes with SYBR green: QuantiTect Primer Assay for Hprt (QIAGEN): Larp7, 5′-GAAGATTGTGAGCGGAGAGC-3′ and 5′-CTCGAGGTTCCAACTGTGC-3′; 4930422G04Rik, 5′-TGTTAGGAAGATCGCCAAGC-3′ and 5′-AGCAGGACTCTGTTGGTTCC-3′; Alpk1, 5′-GCTATGCAGTGAAGCTGTGG-3′ and 5′-TGTCCATGCAAGATGAGTCG-3′; Mm_Alpk1_2_SG QuantiTect Primer Assay (QIAGEN): Tifa, 5′-GCCACTGGAAGACTCTCAGG-3′ and 5′-AACGTATACTGGCACATGTTGG-3′; Ap1ar, 5′-CGACAGCACCTCCTTAGACC-3′ and 5′-ACTCAGAGTTGCCGTTGTCG-3′; and 5730508B09Rik, 5′-GGCGAGCAGGATATTGAAGA-3 and 5′-ATCCTTGTGAAGCCCATGTT-3′. Primer pairs and probes for TaqMan were the following: Hprt, 5′-GACCGGTCCCGTCATGC-3′ and 5′-TCATAACCTGGTTCATCATCGC-3′, VIC/TAMRA probe 5′-ACCCGCAGTCCCAGCGTCGTC-3′; Neurog2, 5′-GCGTAGGATGTTCGTCAAATC-3′ and 5′-CAGCAGCATCAGTACCTCCTC-3′, FAM/TAMRA probe 5′-TTCCTCCTTCAACTCCAGAGTC-3; LOC100043382, 5′-TCTGAACATGGTTGTCCCAA-3′ and 5′-GTGAATCAGCCTGCCTTAGC-3′, FAM/TAMRA probe 5′-CCCTGCCTCCTGTCTTTCTT-3′; and IL-23p19, 5′-AGCGGGACATATGAATCTACTAAGAGA-3′ and 5′-GTCCTAGTAGGGAGGTGTGAAGTTG-3′, FAM/TAMRA probe 5′-GCCAGTTCTGCTTGCAAAGGATCCGC-3′.

For BM chimera experiments, recipient mice were lethally irradiated (2× 550 rads/5.5 Gys) before reconstitution with BM from the indicated donor line. BM was aseptically harvested from tibia and femur of the respective donor strain, and 5 × 10⁶ cells were injected into the tail vein of the irradiated recipients. Mice were allowed to reconstitute for 8 wk before being infected with H. hepaticus. Susceptible recipient mice (129.RAG or R9) were mixed with protected recipient (129.C3B.RAG or R17) before irradiation to minimize cage effects and maintained for the duration of the experiment.

For BMDMs, BM cells were isolated from femur and tibia of R9 or R17 recombinant mice and were cultured in L929 fibroblast-conditioned RPMI 1640 medium (containing 15% L929 supernatant and 10% FCS) for 7 d. The resulting macrophages were stimulated with 1 µg/ml LPS (Sigma-Aldrich) or 20 MOI of live H. hepaticus.

To deplete Thy1⁺ ILC, mice were treated with 1 mg anti-Thy1 antibody (YTS 154.7.7.10) or rat isotype control 6 d before and on the day of H. hepaticus infection. Depletion of ILC was controlled at day 6 after infection by FACS staining for lineage negative (CD11b, Gr1, and B220), Thy1^hi, Sca1^hi cells (Buonocore et al., 2010).

Mice were infected with H. hepaticus, and 6 wk after infection, AOM (Sigma-Aldrich) was administered i.p. once a week at a dose of 10 mg/kg for 5 wk according to a previously published protocol (Nagamine et al., 2008). 21 wk after H. hepaticus infection, mice were sacrificed and colons were removed. Whole colons were fixed in 10% formalin, stained in 2% methylene blue, and aberrant crypt area was analyzed using a dissection microscope and sliding caliper. Colons were then either cut in several pieces (covering the whole length) or swiss-rolled, and 5-µm paraffin sections were prepared. Randomly selected colon ribbons (200–300 sections per mouse) were prepared and every 10th section was H&E stained and graded by a pathologist. Low-grade dysplasia, high-grade dysplasia, and adenocarcinoma were scored blindly using criteria for grading dysplasia in human IBD, as well as the TNM classification of colorectal tumors. Criteria for dysplasia in human IBD: low grade dysplasia, requires intact to mild alteration of gland architecture and nuclear alterations that extend to mucosal surface without appreciable loss of nuclear polarity; high grade dysplasia, requires intact to severe alterations of gland architecture with clear evidence of a loss of nuclear polarity. TNM classification of colorectal tumors (Sobin and Wittekind, 1997): (pT = primary tumor) pTX, primary tumor cannot be assessed; pT0, no evidence of primary tumor; pT1, tumor invades submucosa; pT2, tumor invades muscularis propria; pT3, tumor invades through muscularis propria into subserosa or nonperitonealized pericolic or perirectal tissues; pT4, tumor directly invades other organs (pT4a) and/or involves the visceral peritoneum (pT4b).

Statistical analysis and graphical representations were done using Prism5 (GraphPad Software). For comparison of Q-PCR results and histological analysis, the nonparametric Mann-Whitney test was applied.

Table S1 gives details on the nonsynonymous SNPs found in the candidate genes (and microRNAs) of the susceptibility Hiccs locus, in particular their dbSNP ID number and genotype for C57BL/6, 129S6, and C3H/HeJ strains.

We would like to thank Carolina Arancibia and Julia Bollrath for critical reading of this manuscript, Richard Stillion for histology, Nigel Rust and Helen Ferry for cell sorting, B. Warren for advice on cancer pathology, N. Saunders for help with BAC sequencing, and S. Cobbold and H. Waldman for the Thy1 hybridoma. We wish to acknowledge the Computational Biology Research Group (Medical Sciences Division, Oxford UK) for use of their services in this project.

O. Boulard was supported by Cancer Research UK grants A7297 and A11663; S. Kirchberger by the European Commission research program Infla-Care (EC contract No. 223151) and an Austrian FWF Schroedinger Fellowship; D.J. Royston was supported by The Royal College of Pathologists and the Jean Shanks Foundation; and K.J. Maloy and F.M. Powrie were supported by the Wellcome Trust.

The authors declare no competing financial interest.