Cdx2 regulates immune cell infiltration in the intestine

The intestinal epithelium is a unique tissue, serving both as a barrier against pathogens and to conduct the end digestion and adsorption of nutrients. As regards the former, the intestinal epithelium contains a diverse repertoire of immune cells, including a variety of resident lymphocytes, macrophages and dendritic cells. These cells serve a number of roles including mitigation of infection and to stimulate regeneration in response to damage. The transcription factor Cdx2, and to a lesser extent Cdx1, plays essential roles in intestinal homeostasis, and acts as a context-dependent tumour suppressor in colorectal cancer. Deletion of Cdx2 from the murine intestinal epithelium leads to macrophage infiltration resulting in a chronic inflammatory response. However the mechanisms by which Cdx2 loss evokes this response are poorly understood. To better understand this relationship, we used a conditional mouse model lacking all intestinal Cdx function to identify potential target genes which may contribute to this inflammatory phenotype. One such candidate encodes the histocompatability complex protein H2-T3, which functions to regulate intestinal iCD8α lymphocyte activity. We found that Cdx2 occupies the H3-T3 promoter in vivo and directly regulates its expression via a Cdx response element. Loss of Cdx function leads to a rapid and pronounced attenuation of H2-T3, followed by a decrease in iCD8α cell number, an increase in macrophage infiltration and activation of pro-inflammatory cascades. These findings suggest a previously unrecognized role for Cdx in intestinal homeostasis through H2-T3-dependent regulation of iCD8α cells.

www.nature.com/scientificreports/ that expresses CD8αα on the surface and CD3 intracellularly (iCD3) 22 . These cells are named innate CD8αα + (iCD8α) and they have the capacity to produce and secrete inflammatory cytokines and chemokines, are capable of engulfing and killing pathogenic bacteria and of processing and presenting antigens to MHC class II-restricted T cells [23][24][25] . H2-T3 functions to regulate various subtypes of intestinal CD8αα 23 . Activation of iCD8α cells has also been shown to contribute to the development of innate colitis induced by antibodies against CD40 24 .
We have found that that H2-T3 expression was rapidly lost following Cdx deletion in the intestine. H2-T3 is occupied by Cdx2 in vivo, and is regulated by a Cdx response element in heterologous expression assays, strongly suggesting it is a direct target gene. Attenuation of H2-T3 expression in Cdx mutants was associated with loss of TCR− CD8αα (iCD8α) lymphocytes specifically in the intestinal epithelium. In addition, iCD8α cells exhibited increased degranulation concomitant with macrophage infiltration and an increase in inflammatory markers in Cdx mutant intestine. Taken together, these findings suggest a novel mechanism by which Cdx impacts intestinal homeostasis. Cdx1 and Cdx2 play fundamental roles in the intestine presumably through regulation of expression of target genes. To identify such targets, we used RNA-seq, comparing global gene expression between control and Cdx mutants, combined with ChIP-seq to identify Cdx-dependent transcription units that were concomitantly occupied by Cdx2. Cdx1 −/− Cdx2 F/F VillinCreERT mice were treated with a single 5 mg dose of tamoxifen by oral gavage between 9 and 12 weeks of age to delete Cdx2, yielding animals lacking all Cdx function throughout the intestinal tract 26 ; these animals are termed Double Knock Out (DKO) hereafter for simplicity. To control for possible off-target drug effects, control littermates lacking the VillinCreERT transgene were treated with tamoxifen and processed in parallel. Such Cdx1 null mutants do not exhibit any intestinal phenotype and are therefore suitable controls 14 .

Identification of intestinal Cdx target genes.
We first determined the time course of Cre-mediated loss of Cdx2 in DKO mutant intestinal cells by western blot compared to attenuation of expression of SIS, a known Cdx intestinal target gene 18 . These analyses revealed significant loss of both Cdx2 protein (Fig. 1A) and SIS transcripts (Fig. 1B) 48 h post-deletion. This time point was therefore selected for RNA-seq analysis in order to mitigate secondary effects likely to occur upon protracted loss of Cdx function.
RNA-seq analysis revealed near-complete loss of Cdx2 exon 2, as anticipated (data not shown), consistent with western blot analysis. Partec FLOW analysis from biological triplicate samples revealed 1498 genes exhibiting a minimum of twofold change between Cdx mutant and control intestinal epithelium with a p-value of ≤ 0.05 ( Fig. 2A). Of these genes, 794 exhibited increased expression and 704 showed reduced levels relative to controls ( Fig. 2A). Gene ontology (GO) analysis revealed processes associated with immune cell regulation, including complement-related and antigen presentation genes, within the top pathways of transcripts affected by loss of Cdx function (Table 1).
To identify potential direct Cdx targets, we combined the RNA-seq data with the results from ChIP-seq analyses, using Cdx2 occupancy intervals which exhibited a minimum absolute peak summit of 3,000,000 and a p-value < 0.05 as determined by Partek-Flow. Peaks were annotated based on their proximity to known genes (Fig. 2B), yielding 246 candidates with occupancy within 2500 bp of transcription start sites (Fig. 2C). SIS was recovered in this group, and exhibited a 51 fold reduction in expression, validating the screen (Fig. 2D).   www.nature.com/scientificreports/ Among novel candidate targets was H2-T3, which encodes an MHC class Ib protein that regulates the development and function of intestinal intraepithelial iCD8α cells (Fig. 2D). RNA-seq revealed that H2-T3 expression was reduced at 48 h in DKO mutants compared to controls (Fig. 3A), while RT-qPCR analysis from independent biological samples demonstrated loss of H2-T3 expression as early as 24 h post-tamoxifen treatment, with an approximately 50 fold reduction at 72 h (Fig. 3B). Analysis of the sequences under the Cdx2 ChIP-seq peak at the H2-T3 locus revealed the sequence TTT ATT , a canonical Cdx response element 27 (CDRE), residing ~ 50 bp upstream of the transcriptional start site of the gene. ChIP-PCR analysis performed using primers flanking this interval confirmed Cdx2 occupancy (Fig. 3C).

Large IntesƟne
To assess the ability of Cdx to direct expression from the H2-T3 promoter, and the requirement for the CDRE in this process, a reporter construct containing 1.9 kb of sequences 5′ to the H2-T3 transcriptional start site was constructed harboring a wild type (TTT ATT ) or mutant (GGA TCC; Mut) Cdx binding motif (Fig. 4A). Co-transfection of the wild type reporter with a Cdx2 expression vector resulted in a threefold increase in luciferase expression in C2BBE1 cells, and this response was attenuated upon mutation of the CDRE (Fig. 4B). This demonstrates that the H2-T3 CDRE conveys transcriptional regulation by Cdx2 and, together with the ChIP-seq analysis, is consistent with H2-T3 being a direct Cdx target gene.
Cdx impacts iCD8α lymphocyte retention. The primary role of H2-T3 is to serve as a co-factor protein for retention of iCD8α lymphocytes by virtue of its high affinity for the αα homodimer 24 . This interaction also serves to regulate the activation of iCD8α cells, thereby controlling their secretion of various cytokines and granzymes which activate inflammatory responses and act as a chemokine for macrophages 23,28 . We therefore examined the impact of Cdx loss-of function on iCD8α cell retention, their degranulation, and infiltration of macrophage cells. To this end, intestinal cells were isolated from control and DKO mutants and interrogated by flow cytometry for CD45 (as a marker of total immune cells) as well TCRβ, CD8α, CD8β, CD4, F4/80, CD11c and CD206. iCD8α cells were inferred by the lack of expression of TCRβ, positive expression of CD8α and the absence of CD8β. F4/80 was used to identify total macrophage populations, while CD11c and CD206 were used to differentiate between pro-and anti-inflammatory macrophages, respectively, within the F4/80 population.
Cdx DKO mutants exhibited a reduction of iCD8α cells from a relative proportion of 45% total IEL cells to 15% by day four and approximately 1.5% by day five (Fig. 5). A concomitant increase in total macrophages was also seen as early as 4 days after Cdx2 deletion (Supplemental Fig. 1). Within the macrophage population, a threefold increase in the F4/80-CD11c positive cells was observed, consistent with a pro-inflammatory response in mutant intestine. No statistically significant changes were observed in the total CD4+ cell population suggesting that regulatory and helper T-cell populations were unaffected ( Supplementary Fig. 1).
Loss of H2-T3 in IECs has been shown to result in increased iCD8α degranulation leading to a localized inflammatory response including the recruitment of macrophages to the site of reaction 23,24 . Consistent with this, at 4 days post-Cdx2 deletion, we observed that the majority of CD8αα cells were granzyme B-positive. Conversely, we found a near-complete degranulation of CD8αα cells in DKO mutants compared to controls at day 5, while CD8αβ cells retained granzyme B (Fig. 6). While we do note a significant reduction in CD8αβ numbers, this occurs after the loss of iCD8α cells and is likely due to effects unrelated to the attenuation of H2-T3, as the αβ heterodimer is not known to interact with this ligand.

Discussion
Previous studies have identified Cdx1 and Cdx2 as key regulators of intestinal homeostasis and as contextdependent modifiers of colorectal cancer 10,26,29 , however, the underlying mechanisms by which these outcomes manifest are not fully understood. To address this, we used RNA-seq and ChIP-seq to identify novel Cdx target genes implicated in intestinal homeostasis, leading to our identification of H2-T3, which encodes a member of the major histocompatibility complex family, as a Cdx target. We found that Cdx2 occupied the H2-T3 locus in vivo in an interval which harbors a CDRE necessary to manifest Cdx-dependent transcription in heterologous expression assays. Loss of H2-T3 in Cdx mutant intestine correlated with reduced numbers of iCD8α cells as well as their degranulation, increased infiltration of macrophages and a pro-inflammatory response. We also  21 . Our analysis also showed that macrophage recruitment begins as early 4 days suggesting additional pro-inflammatory mechanisms are evoked by Cdx loss-of-function, the nature of which are presently unknown. Previous work has shown that Cdx2 deletion in the intestine leads to the recruitment of pro-inflammatory macrophages associated with Cdx-dependent cell non-autonomous signaling which potentiates transformation in a murine model of colorectal cancer (CRC) 30 . Our present observations suggest a possible mechanistic basis www.nature.com/scientificreports/ for this finding, and are consistent with the effects of disruption of H2-T3, which triggers a loss of the innate CD8αα + lymphocyte population (iCD8α) 21,23 and subsequent recruitment of pro-inflammatory immune cells 24 . iCD8α cells play functions similar to innate immune cells. They have the capacity to produce and secrete pro-inflammatory cytokines and chemokines, are capable of engulfing and killing pathogenic bacteria and of processing and presenting antigens to MHC class II-restricted cells 20 . iCD8α are characterized by expression of CD8αα homodimers and negative for expression of TCRα 23 . In addition to presenting H2-T3 to iCD8α cells, IECs also produce and present IL-15 to iCD8α cells 23 . This combination is required for proper development, maintenance and function of iCD8α cells 23 , the mis-regulation of which has been implicated in inflammatory bowel disease (IBD) 24 .
IBD can occur due to aberrant immune cell activity and may involve a T-cell autoimmune disorder 31,32 . This can manifest as several forms, including over-activity of innate T-cells or macrophage populations, either of which can result in a chronic inflammatory response [33][34][35] . In this regard, loss of expression of CDX2 has been associated with the development of ulcerative colitis [36][37][38] . While the basis underlying this observation is unresolved, increased TNFα secretion, triggered by activation of pro-inflammatory pathways, has been association with reduced expression of CDX2 in IBD patients 36,37 . Our present results suggest an additional pathway whereby attenuation of CDX2 leads to subsequent loss of H2-T3 and further promotes an inflammatory state through dysregulation and degranulation of iCD8α cells (Fig. 7). Also consistent with this model is the finding that activation of iCD8α cells contributes to the development of innate colitis induced by antibodies against CD40 24 , further supporting a link between Cdx, iCD8α cells and IBD. Although human iCD8α cells have been identified in the intestinal tract and may play a similar role as in the mouse, a human homologue of H2-T3 has not yet been reported.
Animals were treated with a single 5 mg dose of tamoxifen by oral gavage between 9 and 12 weeks of age to delete Cdx2, yielding animals lacking Cdx1 and Cdx2 throughout the intestinal tract and termed Double Knock Out (DKO) for simplicity. Control littermates lacking VillinCreERT were treated in parallel to control for off-target  www.nature.com/scientificreports/ sected in cold Phosphate Buffered Saline (PBS), opened longitudinally and rinsed with PBS. Intestines were then incubated in 5 ml of cold PBS 1.5 mM in EDTA for 15 min and tissue vortexed to dislodge epithelial cells. This process was repeated for a total of 6 cycles, with the 5th and 6th isolates, which were enriched in crypt cells, retained for analysis.

RNA-sequencing (RNA-seq).
For RNA-seq, intestinal cells were isolated as above 48 h post-treatment with 5 mg of tamoxifen from 3 mutant mice and 3 littermate controls. RNA was extracted using RNeasy as per the manufactures directions (Qiagen). RNA quality was assessed using an Agilent Bioanalyzer, poly-A + mRNA was enriched and fragmented, and libraries generated with cDNA which underwent rend tail repair and subsequent dA tailing followed by adaptor ligation. Samples with a minimum RNA integrity of 8, assessed by Bioanalyzer, were used to generate cDNA libraries which were sequenced on an Illumina HiSeq4000 system to a minimum of approximately 50 million reads per sample.
Chromatin immunoprecipitation-sequencing (ChIP-seq). ChIP-seq analysis was conducted as previously described 40 from two independent biological replicates. Briefly, nuclei were isolated by lysing intestinal cells in 4% Tween PBS and chromatin sheared by sonication to generate DNA fragments averaging 500 bp, verified by gel electrophoresis. Cdx2-associated chromatin was isolated using Dynabeads conjugated to an anti-Cdx2 antibody. Protein-DNA complexes were isolated, reverse crosslinked by heating at 65 °C for 18 h and DNA purified using Qiagen PCR purification kits as per the manufactures direction. Sequencing was conducted on an Illumina Nextgen sequencing column to 24-34 million reads per sample. RNA-seq and ChIP-seq in silico analyses were performed using Partek FLOW software. To assign potential direct Cdx2 targets, changes in transcript abundance between wild type and Cdx mutant intestinal cells were correlated with Cdx2 genome occupancy. Binding motif analysis was performed to identify putative Cdx2 binding elements (CDRE) under ChIP peaks, and targets further validated by ChIP-PCR using primers flanking such CDREs (Table 1)  Promoter analysis. Cell-based reporter assays were used to assess Cdx2-dependent regulation of the putative CDRE identified within the H2-T3 promoter. The reporter was generated from a 1.9 kb PCR amplicon of genomic DNA immediately 5′ to the H2-T3 transcriptional start site. This interval was then cloned into the pXP-2 luciferase plasmid and verified by sequencing. Site-directed mutagenesis was used to generate an identical reporter construct, mutating the H2-T3 CDRE (TTT ATT ) to a Bam HI recognition sequence (GGA TCC ).
C2BBE-1 and HEK293T cells (ATCC) were cultured in DMEM (Hyclone) supplemented with 5% FBS. Transferrin (0.01 mg/ml) was also included for C2BBE-1 cells. Cells were maintained at 37 °C in either 5% CO 2 (for HEK293T cells) or 10% CO 2 (for C2BBE-1 cells). Cultures were transfected with reporter constructs (750 ng) with a Cdx2 expression vector or empty expression vector control (250 ng) using Attractine transfection reagent (Qiagen) as per the manufacturer's directions. Cells were collected 48 h post-transfection for analysis of promoter activity using a Promega luciferase assay kit as per the manufacturer's direction. Luciferase activity was standardized to total protein content.
Flow cytometry. IEL cells were purified using a Percoll gradient as previously described 41 . An aliquot was subjected to heating at 65 °C for 20 min to induce cell death. Dead cells were pooled with live cells to serve as an internal control for live/dead staining, which was performed using zombie yellow (Biolegend) as per the manufacturers' directions. Cells were then washed twice in 2 ml of cold PBS for 5 min, fixed in 4% paraformaldehyde for 20 min at room temperature and pre-treated with anti-Fc antibody for 1 h and subsequently for 2 h with fluorophore-conjugated antibodies. Cells were then washed to remove excess antibody and analyzed by flow cytometry, gating to count live CD45+ cells and analyzed for the abundance of TCR+ , TCR− and sub population of CD8+ cells, F4/80 macrophages or M1/M2 macrophages (CD11c or CD206; see Supplementary Table 2).
Quantitative PCR (RT-qPCR). RNA was isolated from intestinal epithelial cells, cDNA generated by standard means and amplified by PCR using primers described in Supplementary Table 1. Thermo-cycler conditions were as follows: 95 °C for 5 min (Hot start), 95 °C for 30 s, 30 s for annealing (see Supplementary Table 1for annealing conditions), 72 °C for 30 s/kb for a total of 29 cycles, determined experimentally to fall with the linear phase of amplification. Products were resolved by agarose gel electrophoresis and quantified by densitometry using Alpha-ease software, standardized to β-actin expression.
For qPCR, 5 μl of cDNA was amplified for 40 cycles with SYBR Green (Go-Taq) in triplicate using a Bio-Rad CFX qPCR machine, followed by dissociation curve analysis. Cycling was 95 °C for 3 min, annealing at 58 °C for 30 s and elongation of the primers at 60 °C for 30 s. Data was analysed using the Bi-Rad CFX manager software and gene expression was expressed relative to GAPDH by (ΔCt) then standardized to control expression (ΔΔCt).
Western blot analysis. Cells from either intestinal epithelium or tissue culture were collected and lysed in RIPA buffer. Protein was quantified by Bradford assay as per the manufacturer's directions (Bio-Rad), lysates resolved by polyacrylamide gel electrophoresis and transferred to PVDF membranes. Membranes were blocked using 5% skim milk in Tris-buffered Saline + 0.1% Tween (TBST) and incubated with primary antibodies against Cdx2 16 or β-Actin (Santa Cruz) (diluted to 1:1,000 in 5% skim milk in TBST). Membranes were washed in TBST three times for 5 min each and incubated with HRP-conjugated secondary antibody (Santa Cruz Biotechnology, 1:10,000 in 5% skim milk in TBST). Following washing with TBST, reactivity was revealed by luminescence