Intestinal DMBT1 Expression Is Modulated by Crohn’s Disease-Associated IL23R Variants and by a DMBT1 Variant Which Influences Binding of the Transcription Factors CREB1 and ATF-2

Objectives DMBT is an antibacterial pattern recognition and scavenger receptor. In this study, we analyzed the role of DMBT1 single nucleotide polymorphisms (SNPs) regarding inflammatory bowel disease (IBD) susceptibility and examined their functional impact on transcription factor binding and downstream gene expression. Methods Seven SNPs in the DMBT1 gene region were analyzed in 2073 individuals including 818 Crohn’s disease (CD) patients and 972 healthy controls in two independent case-control panels. Comprehensive epistasis analyses for the known CD susceptibility genes NOD2, IL23R and IL27 were performed. The influence of IL23R variants on DMBT1 expression was analyzed. Functional analysis included siRNA transfection, quantitative PCR, western blot, electrophoretic mobility shift and luciferase assays. Results IL-22 induces DMBT1 protein expression in intestinal epithelial cells dependent on STAT3, ATF-2 and CREB1. IL-22 expression-modulating, CD risk-associated IL23R variants influence DMBT1 expression in CD patients and DMBT1 levels are increased in the inflamed intestinal mucosa of CD patients. Several DMBT1 SNPs were associated with CD susceptibility. SNP rs2981804 was most strongly associated with CD in the combined panel (p = 3.0×10−7, OR 1.42; 95% CI 1.24–1.63). All haplotype groups tested showed highly significant associations with CD (including omnibus P-values as low as 6.1×10−18). The most strongly CD risk-associated, non-coding DMBT1 SNP rs2981804 modifies the DNA binding sites for the transcription factors CREB1 and ATF-2 and the respective genomic region comprising rs2981804 is able to act as a transcriptional regulator in vitro. Intestinal DMBT1 expression is decreased in CD patients carrying the rs2981804 CD risk allele. Conclusion We identified novel associations of DMBT1 variants with CD susceptibility and discovered a novel functional role of rs2981804 in regulating DMBT1 expression. Our data suggest an important role of DMBT1 in CD pathogenesis.


Introduction
The pathogenesis of the inflammatory bowel diseases (IBD) Crohn's disease (CD) and ulcerative colitis (UC) is not completely understood. Current models of disease pathogenesis suggest that IBD is caused by a complex interplay of environmental factors such as the microbiota of the host, the host immune system and genetic predisposition. The first IBD susceptibility gene NOD2/ CARD15 (which is particularly associated with small intestinal disease), was identified in 2001 and encodes an intracellular bacterial pattern recognition receptor (PRR) [1,2]. Since then, emerging evidence suggests that particularly autophagy genes such as ATG16L1 [3,4] and IRGM [5,6] as well as genes involved in the proinflammatory Th17 cell pathway such as IL23R, IL12B, JAK2, STAT3, CCR6, and IL2/IL21 [7,8,9,10] are associated with CD.
DMBT1, a glycoprotein of the scavenger receptor cystein-rich (SRCR) family, is a protein participating in antibacterial host defense on mucosal surfaces such as the intestine or the oral mucosa. It acts as a PRR that recognizes leucine-rich repeats of bacterial proteins [11]. It is known that DMBT1 inhibits NF-kB activation induced by the bacterial products lipopolysaccharide (LPS) and muramyldipeptide (MDP) [12], aggregates group A Streptococci by binding pili [13], and inhibits Salmonella enterica cytoinvasion [12], resulting in reduced bacterial adhesion to human epithelial cells [13,14,15,16]. Interestingly, DMBT1 is a target gene of NOD2 [12] and also a target of TLR4 for which we and others demonstrated associations with IBD susceptibility [17,18].
Studies with experimental dextran sulfate sodium (DSS) colitis models in DMBT1 knockout (KO) mice had heterogeneous outcomes. While in one study, no difference between KO and wild-type (WT) mice was observed [19], another study reported enhanced susceptibility to DSS colitis in KO mice [20] whereas a third study observed differences only at low doses of DSS [16]. However, to what extent these findings can be directly transferred to human IBD, remains to be determined. The contradictory results may reflect a general problem in using mouse models for human intestinal inflammation as it has been demonstrated recently that some mouse models of inflammation poorly mimic the more complex human conditions [21].
In humans, it has been demonstrated that DMBT expression is upregulated by proinflammatory stimuli such as TNF-a or LPS [12] and correlates with disease activity in IBD patients [20,22]. Recently, a copy number polymorphism in DMBT1, resulting in a substantially decreased number of scavenger receptor cysteine-rich (SRCR) domains, was identified as a risk factor for CD [20].
In this study, we aimed to further clarify the role of DMBT1 in intestinal epithelial cells (IEC) and in the pathogenesis of IBD. We investigated the IL-22-mediated transcriptional regulation of DMBT1 in the IEC lines HT-29 and DLD-1 and analyzed the influence of IL-22 expression-modulating, CD-associated IL23R variants on DMBT1 expression in biopsies from CD patients. Moreover, we analyzed seven DMBT1 gene variants and their haplotypes, including four so far not analyzed single nucleotide polymorphisms (SNPs) in a large panel of 2073 Caucasian individuals regarding their association with IBD risk and disease phenotype. We tested for gene-gene interaction (epistasis) of DMBT1 variants with known CD susceptibility variants in NOD2, IL23R and IL27 since previous studies demonstrated that these genes may also be involved in the modulation of DMBT1 expression [12,23,24]. In addition, we analyzed the functional impact of variants in DMBT1 on transcription factor binding to the respective DNA and the downstream gene expression and identified a CD-associated DMBT1 variant that is linked to the colonic DMBT1 gene expression in CD patients.

Ethics Statement
The study was approved by the Ethics committee of the Medical Faculty of the Ludwig-Maximilians-University Munich. Written, informed consent was obtained from all patients prior to the study. Study protocols were based on the ethical principles for medical research involving human subjects of the Helsinki Declaration (http://www.wma.net/e/policy/b3.htm).

Reagents and Antibodies
Human recombinant IL-22 was obtained from R&D Systems (Wiesbaden, Germany). Antibodies against CREB1 and ATF-2 were from Santa Cruz Biotechnology (Heidelberg, Germany) and p84 antibody was from Abcam (Cambridge, UK). The secondary anti-rabbit antibody was from GE Healthcare (Freiburg, Germany).

RNA Isolation, Reverse Transcription and Quantitative PCR
Total RNA from IEC was isolated using the RNeasy Mini Kit from Qiagen (Hilden, Germany) and 500 ng were reverse transcribed using the Transcriptor First Strand cDNA Synthesis Kit from Roche (Mannheim, Germany). Total RNA from intestinal biopsies was isolated using a Branson sonifier to disrupt the tissue followed by RNA extraction with Trizol reagent and chloroform. Quantitative real-time PCR was performed as previously described [23] using a LightCycler 480 instrument (Roche Diagnostics, Mannheim, Germany) and SYBR green detection format. Gene expression was normalized to the b-actin expression in the respective samples.

Western Blot
Western Blot analysis was performed according to standard procedures [25] with nuclear extracts isolated from IEC lines as described [23]. Briefly, 20 mg of nuclear extract or 50 mg of total protein were separated on an 8-16% gradient polyacrylamide gel and were transferred to a PVDF membrane. Membranes were blocked with 5% milk in TBS-T and incubated with the primary antibody overnight at 4uC. Following incubation with the secondary HRP-coupled antibody, luminescent detection was performed with the ECL system (Pierce) and a CCD camera (Peqlab, Erlangen, Germany).
siRNA Transfection DLD-1 cells were reverse transfected in 24 well plates (for RNA isolation) or 10 cm plates (for nuclear extract isolation) with siRNA (Life Technologies, Darmstadt, Germany) using Lipofectamine RNAiMAX (Life Technologies) following the manufacturer's guidelines. Total mRNA and nuclear protein was isolated and specific knockdown was assessed by quantitative PCR and Western Blot, respectively.

In silico Analysis of Transcription Factor Binding Sites
Genomic sequences including SNPs rs2981745 and rs2981804 were analyzed for potential transcription factor binding sites of human transcription factors with the TFSEARCH program (http://www.cbrc.jp/research/db/TFSEARCH.html) which is based on the TRANSFAC database [26]. The threshold score for binding site prediction was set to 75.0 (score = 100.0*('weighted sum'2min)/(max2min); max. score = 100). For each SNP, both alleles including the flanking sequences 10 bp upstream and downstream were compared.

Nuclear Extracts and Electrophoretic Mobility Shift Assay (EMSA)
Isolation of nuclear extracts from IEC lines HT-29, DLD-1, HCT116 and SW480 was performed according to standard procedures [27] with minor modifications [23]. EMSA analysis was performed essentially as described [23]. Briefly, 100 fmol of biotinylated, double-stranded probes were incubated with 5 mg of nuclear extract from cell lines as indicated in the presence of 1 mg of poly d(I)-d(C) for 30 min. Where indicated, 1 mg of anti-CREB1 or anti-ATF-2 antibody or a non-specific negative control antibody was added. For competition experiments, 50-fold molar excess of unlabelled probe was included in the reaction before adding the labelled probe. Samples were separated on a 6% polyacrylamide gel (2.5 hours at 100 V for the DMBT1 probe, 1.5 hours for the CREB1 probe) and were transferred to a positively charged nylon membrane. Detection of biotinylated probes was performed with streptavidin-HRP using the LightShift Chemiluminescent EMSA kit from Pierce (Thermo Fisher Scientific, Bonn, Germany). The sequences of the two different DMBT1 probes used were 59-CCTGCTAACGTAACCAAATTGCTA-39 and 59-CCTGCTAACGTAGCCAAATTGCTA-39. The sense orientation for each probe is given and the polymorphic nucleotide comprising DMBT1 SNP rs2981804 is underlined. The sequence for the CREB1 probe was 59-AGAGATTGCCTGACGTCAGA-GAGCTAG-39 and the sequence for the ATF-2 probe was 59-CTTAGTTACGTAATAATTGT-39 (consensus sequence underlined). The unspecific competitor probe had the sequence 59-GATCCTTCTGGGCCGTCCTAGATC-39.

Plasmid Cloning, Transient Transfection and Luciferase Assay
The genomic region harboring SNP rs2981804 was amplified by PCR from HT-29 cells (genotype AA) or DLD1 cells (genotype GG). The resulting fragments were cloned either into the luciferase reporter plasmid pGL4.23 (Promega, Mannheim, Germany) which contains a minimal promoter (minP) in front of the luciferase gene resulting in low basal expression or into the pGL4.13 vector with a strong SV40 promoter/enhancer with high basal expression. For both vector backbones, three constructs were cloned: 1) DMBT1 insert upstream, i.e. 59 of the promoter into the multiple cloning site, 2) DMBT1 insert 2 kb downstream, i.e. 39 of the promoter, and 3) DMBT1 insert 39 and 59 of the promoter. All constructs were verified by sequencing. DLD1 or HT-29 cells were transfected in 96-well plates with 100 or 200 ng of plasmid, respectively, using Lipofectamine LTX (Life Technologies), together with 10 or 20 ng of Renilla luciferase plasmid. After 24 hours, luciferase activity was detected using the Dual-Glo Luciferase Assay System (Promega). Relative light units (RLU) were normalized to the levels of Renilla luciferase. A detailed description of all experimental cloning procedures and the primers used can be found in the methods S1, figures S1 and S2 and table S1.

Study Population
The study population (n = 2073) consisted of 818 Crohn's disease (CD) patients and 972 healthy, unrelated controls [(623 CD patients and 762 controls in the discovery panel (recruited from the IBD center of the University Hospital Munich-Grosshadern) and 195 CD patients and 210 controls in the replication panel (from the LMU Munich Innenstadt Campus)]. In addition, we analyzed 283 UC patients recruited from the IBD center of the University Hospital Munich-Grosshadern. All patients were of Caucasian origin.
Phenotypic data were collected blind to the results of the genotypic data and consisted of demographic and clinical parameters (behaviour and anatomic location of IBD, diseaserelated complications, surgical or immunosuppressive therapy) which were recorded by two senior gastroenterologists using patient charts analysis and a detailed questionnaire including an interview at time of enrolment. The diagnosis of CD and UC was based on established guidelines based on endoscopic, radiological, and histopathological parameters. The phenotypic classification of CD patients was based on the Montreal classification [28], including age at diagnosis (A), location (L), and behaviour (B) of disease. In patients with UC, anatomic location was also assessed in accordance to the Montreal classification based on the criteria ulcerative proctitis (E1), left-sided UC (distal UC; E2), and extensive UC (pancolitis; E3). Patients with indeterminate colitis were excluded from the study. The demographic characteristics of the study population are presented in table 1.

Sampling of Intestinal Biopsies from Patients with Crohn's Disease
In a subgroup of 27 CD patients, 75 intestinal biopsies from inflamed and non-inflamed intestinal tissue were collected during routine endoscopy. Written informed consent was obtained from all patients prior to biopsy sampling. The study was approved by the Ethics Committee of the Medical Faculty of the Ludwig-Maximilians-University Munich. Detailed patient characteristics are summarized in table 2.

DNA Extraction and Genotyping
For genotyping, genomic DNA was isolated from peripheral blood leukocytes from all study participants using the DNA blood mini kit (Qiagen, Hilden, Germany) following the manufacturer's instructions. The study participants were genotyped for seven SNPs in the DMBT1 gene region.
The DMBT1 SNPs rs2981745, rs3013236 (corresponds to p.Leu54Ser) and rs1052715 (p.Pro1707Pro) were investigated in the study of Renner and co-workers [20]. Additionally, the tagging SNPs rs2981778, rs11523871 (p.Pro42Thr), rs2981804 and rs2277244 (p.His585Tyr) were selected from the data of the International HapMap project covering the DMBT1 gene plus 10 kb flanking the centromeric and telomeric end of the gene, respectively, and using a setting of r 2 of 0.8 (Fig. 1A). Genotyping was performed by PCR and melting curve analyses using a pair of fluorescence resonance energy transfer (FRET) probes in a LightCycler480 instrument. Detailed genotyping methodology and the primers/probes used for genotyping are summarized in the methods S1 and in tables S2 and S3. In addition, detailed haplotype and subphenotype analyses were performed. Gene-gene interactions (epistasis) of DMBT1 variants with variants in NOD2, IL23R and IL27 were analyzed. Those genotype data were available from previous studies [8,29,30,31,32,33]. IL-22 serum protein levels were analyzed in previous study [34].

Statistical Analysis
Single-marker allelic tests were performed with Pearson's x 2 test. Student's t-test was applied for quantitative variables. All tests were two-tailed and P-values ,0.05 were considered nominally significant. Correction for multiple testing was performed using Bonferroni's method. We tested 7 DMBT SNPs for association with two phenotypes (CD and UC status) resulting in a corrected P-value of 0.0036 (0.05/(2*7)). Empirical P-values were derived using 10,000,000 point-wise permutations. Odds ratios (OR) were calculated for the risk allele at each SNP. Data were evaluated by using the SPSS 13.0 software (SPSS Inc., Chicago, IL, U.S.A.) and R-2.4.1.
(http://cran.r-project.org). Permutation-based P-values, interaction P-values as well as haplotype and LD analysis were conducted using PLINK v 1.07 (http://pngu.mgh.harvard.edu/purcell/plink/). We ran a sliding window approach, varying the window size from 2 to 7 included markers and using the option ''hap-logistic''. Haplotype omnibus P-values are based on jointly testing all H haplotype effects at the specific position resulting in a H-1 degrees of freedom test. For haplotype specific tests, individuals presenting with the specific haplotype are compared to all other individuals.

Results
The Th17 Cytokine IL-22 Up-regulates STAT3-dependent DMBT1 Expression in Intestinal Epithelial Cells We and others have shown that IL-22 is an inducer of antimicrobial peptides like defensins in IEC and other tissues [35,36,37]. DMBT1 is an antibacterial scavenger receptor (9) and we recently demonstrated that the STAT3-activating cytokine IL-27 induces DMBT1 expression in IEC [23]. Given that IL-22 is We have recently demonstrated that CD-associated variants in IL23R modulate IL-22 expression. Higher IL-22 serum levels were found in carriers of CD risk-associated IL23R SNPs and lower IL-22 serum levels in carriers of CD-protective IL23R SNPs [34] (summarized in table 3). Having shown that IL-22 induces DMBT1 expression, we further analyzed these IL23R variants regarding their association with DMBT1 expression. DMBT1 mRNA expression was determined by quantitative PCR in a total of n = 75 inflamed and not inflamed intestinal biopsies collected from 27 CD patients (for patient characteristics, see table 2). DMBT1 expression in carriers of the minor allele (homozygous and heterozygous) of the respective IL23R SNPs was divided by the expression in homozygous carriers of the wild-type (WT) allele (table 3, Fig. 3A). The odds ratios (ORs) for the analyzed IL23R SNPs were available from a previous study [8]. In 5 out of 10 IL23R SNPs, there were significant differences (p,0.05) in DMBT1 expression between carriers of the minor allele and WT carriers (table 3). Similar to our previous results obtained for    Fig. 3A). The correlation of OR and DMBT1 expression was independent of the inflammation status of the analyzed biopsies ( figure S3).

Intestinal DMBT1 Expression is Increased in Active Crohn's Disease
Having demonstrated that DMBT1 is up-regulated by IL-22 and IL-27 [23] and given the increased expression of IL-22 and IL-27 in active IBD as shown by us and others [34,35,38], we next compared DMBT1 mRNA expression levels in all inflamed (n = 30 biopsies) and not inflamed (n = 45 biopsies) intestinal biopsies from the above described 27 CD patients. DMBT1 expression was significantly higher (16.6-fold) in inflamed colonic biopsies compared to not inflamed regions (Fig. 3B, p = 7.6610 24 ).

Associations of DMBT1 Gene Variants and Haplotypes with Crohn's Disease in the German Population
Having shown that DMBT1 expression is associated with disease activity in CD patients, we next aimed to determine whether SNPs in the DMBT1 gene region might influence IBD susceptibility. We genotyped 7  case-control panels, we demonstrated significant disease associations with certain DMBT1 SNPs ( To elucidate if the two most strongly IBD-associated SNPs represent independent disease association, we conditioned association analysis on the strongest signal, i.e. rs2981804 for CD and rs2981745 for UC. Even after conditioning, effects for rs2981745 in CD (p = 1.0610 25 ) and rs2981804 in UC (p = 4.5610 24 )  were investigated using the Montreal classification of CD and UC [28]. However, all statistical significant phenotypic associations were rather weak and were not robust to multiple testing (table S8-S10).  (3) and represents the fold increase or decrease in DMBT1 expression in carriers of the IL23R minor allele (homozygous and heterozygous) compared to DMBT1 levels in IL23R wild-type carriers for the respective IL23R SNP. The IL-22 serum levels of column 6 summarizes the results of a previous study [34] while the odds ratios of column 7 represent the results of a previous detailed genotype analysis [8], in which all patients analyzed here participated. The correlation coefficient between the DMBT1 quotient of column (5) and the ORs of column (7)   In order to investigate for potential disease associations with certain DMBT1 haplotypes, we performed a detailed haplotype analysis considering all possible haplotypes with a frequency of at least 1% in the whole sample (table S11 and S12). In CD patients, all haplotype groups analyzed were significantly associated with the disease after correction for multiple testing including corrected omnibus P-values of ,10 210 in 60% of all analyzed haplotype groups (table S11). The strongest association with CD was found for a haplotype consisting of all seven analyzed SNPs (omnibus Pvalue: 6.14610 218 , table 5 and table S11). This association was mainly attributable to two haplotypes: CGATGCA and CGA-TACA, which differ only in the rs2981804 allele (underlined). The rare CGATGCA haplotype was significantly more common in controls (haplotype frequency (HF) 0. 13 Analysis for Epistasis between DMBT1 and NOD2, IL23R and IL27 Gene Variants Regarding IBD Susceptibility We then aimed to uncover potential epistasis of the DMBT1 variants rs2981745 and rs2981804 with other replicated IBD susceptibility gene variants. We focused on NOD2, IL23R and IL27, given that DMBT1 is a target gene of NOD2 [12] and IL-23R [24]. Moreover, we recently demonstrated that IL23R modulates the expression of the Th17 cytokine IL-22 [34]. IL-22 is an epithelial barrier-protective cytokine [35] that is a transcriptional activator of DMBT1 expression as we demonstrated above. As IL-27 is a Th17 cell inhibiting cytokine [39], we also included variants in the IL27 gene region for which we and others recently demonstrated an association with IBD susceptibility [32,40]. Moreover, we have recently shown that DMBT1 is also a transcriptional target of IL-27 [23]. Data on NOD2, IL23R and IL27 gene variants in our study population were available from previous studies [8,32,33]. No evidence for epistasis between DMBT1 variants and SNPs in NOD2 and the IL27 gene region was found (table S13 and S14). For the IL23R SNP rs1004819, the most strongly IBD-associated IL23R SNP in our cohort [8], there was significant epistasis with the DMBT1 SNPs rs2981745 (p = 0.042) and rs2981804 (p = 0.031) in UC but not CD patients (table S15). However, the identified statistical interaction was not robust to multiple testing.

Allelic Variants of DMBT1 SNP rs2981804 Alter the DNA Binding of the Transcription Factors CREB1 and ATF-2
As both most strongly IBD-associated DMBT1 SNPs rs2981745 and rs2981804 are located in non-coding regions of DMBT1 (rs2981745:59-untranslated region (UTR); rs2981804:6th intron, Fig. 1A), we hypothesized that these SNPs might be located within DNA binding sequences of nuclear proteins. We therefore screened the respective genomic regions including the SNP and 10 bp upstream and downstream with the online tool TFsearch [26]. For rs2981745, no significant differences in DNA binding scores above the defined threshold were obtained for the two alleles (data not shown). For rs2981804, the program predicted differential binding of several transcription factors (table 6, Fig. 1B).
In EMSA experiments with probes containing either the IBD risk allele A or the protective allele G of rs2981804 and the surrounding genomic sequences, nuclear extracts from IEC lines HT-29, DLD-1, HCT116 and SW480 bound strongly to the probe containing the risk allele A. The probe with the IBDprotective allele G showed clearly weaker protein binding (Fig. 4A). To analyze which proteins were bound to the probes, antibodies against phosphorylated cAMP responsive element binding protein 1 (CREB1) or activating transcription factor 2 (ATF-2) were incubated together with the DNA binding reaction with nuclear extracts from DLD-1 cells. For both transcription factors, higher binding scores were predicted for the A allele in comparison to the G allele (table 6). The addition of an ATF-2 or a CREB1 antibody diminished but did not abolish DNA binding of protein to the DMBT1 probe (Fig. 4B, lanes 3, 4) as well as did unlabelled competitor probes with binding sequences for ATF-2 or CREB1 (Fig. 4B, lanes 5, 8). Addition of 50-fold excess of an unlabelled DMBT1 probe with the A allele completely inhibited binding to the labelled probe while a DMBT1 probe with the G allele had a weaker effect (Fig. 4B, lanes 6, 7). Control reactions (Fig. 4B, lanes 9, 10) with a non-specific antibody or a non-specific DNA probe which did not inhibit specific DNA-protein complex formation demonstrated that two specific protein complexes with two distinct sizes are bound to the DMBT1 probe (Fig. 4B, see arrows). Next, we performed EMSA experiments using a labelled CREB1 consensus probe. CREB1 protein binding was proven by a CREB1-specific antibody (Fig. 4C, lane 4). Excess of unlabelled DMBT1 probe was able to repress protein binding to the CREB1 probe (Fig. 4C, lane 7). Moreover, on the same gel, binding reactions with a labelled DMBT1 probe were included (Fig. 4C, lanes 8-10). A more slowly migrating (and therefore larger) protein complex bound to the DMBT1 probe that had much stronger signal intensity (in comparison to the CREB1 probe) could be observed.
Then we transfected DLD-1 cells with a non-specific control siRNA or siRNA against CREB1 and ATF-2 48 hours prior to protein isolation. SiRNA-mediated down-regulation of CREB1 resulted in a nearly complete loss of protein binding to the DMBT1 probe with the risk allele (Fig. 4D). Down-regulation of ATF-2 also resulted in a slightly decreased protein binding to the DMBT probe (Fig. 4D). Silencing of CREB1 and ATF-2 by the respective siRNAs was confirmed by in western blot experiments (Fig. S4A, B). Together, our data suggest that CREB1 and ATF-2 are essential parts of a larger multi-protein complex that binds strongly to the DMBT1 probe containing the IBD risk allele of rs2981804 but substantially less to the sequence with the protective allele.

The Th17 Cytokine IL-22 Stimulates DMBT1 Expression in Intestinal Epithelial Cells Dependent on the Transcription Factors CREB1 and ATF-2
Next, we aimed to confirm that ATF2 and CREB1 are involved in the transcriptional regulation of DMBT1. As we have shown that IL-22 is an inducer of DMBT1 gene expression in IEC ( Fig. 2A), we next analyzed DMBT1 expression in IL-22stimulated DLD-1 cells in which expression of CREB1 or ATF-2 was silenced by siRNA transfection. In control siRNAtransfected cells, expression of DMBT1 was significantly upregulated by IL-22 after 6 hours of stimulation (p = 0.01; Fig. 5A). In DLD-1 cells transfected with CREB1 siRNA, no increase in DMBT expression was observed following IL-22 stimulation (Fig. 5A). Silencing of ATF-2 prior to IL-22 stimulation had an intermediate effect as IL-22-induced DMBT1 up-regulation was still detectable but weaker than that of control siRNA-transfected cells stimulated with IL-22 (Fig. 5A). Western blot experiments with a DMBT1 antibody confirmed that the IL-22-induced DMBT1 protein expression is inhibited by silencing of CREB1 or ATF-2 expression (Fig. 5B).

The Genomic Region Comprising DMBT1 SNP rs2981804 Differentially Influences Promoter Activity
To analyze whether the genomic region comprising the DMBT1 SNP rs2981804 directly influences gene transcription in vitro, we cloned a PCR-amplified DMBT1 fragment with either the rs2981804 A or G allele into two luciferase reporter vectors containing either a minimal promoter (minP) or a strong SV40 promoter/enhancer as described in the methods and methods S1 section. The inserts were cloned either immediately upstream, 2 kb downstream or up-and downstream of the promoterluciferase gene region to analyze potential position-dependent short-range and long-range effects on gene expression.
Luciferase assay in transfected DLD-1 cells revealed that in the minP vector with low basal expression, the DMBT1 region suppresses gene transcription, especially when cloned both up-and downstream of the promoter-luciferase gene (Fig. 6A). When cloned into a strong SV40 promoter-driven luciferase expression vector, the DMBT1 fragment was able to further increase luciferase gene expression (Fig. 6B). However, there were no significant differences between the A and the G allele of rs2981804. Similar results were obtained in HT-29 cells (data not shown).

Intestinal DMBT1 Expression is Lower in Homozygous Carriers of the DMBT1 rs2981804 IBD Risk Allele
To determine whether rs2981804 is associated with differential DMBT1 expression in CD patients, DMBT1 mRNA expression levels measured in human intestinal biopsies (Fig. 3B) were subdivided according to the rs2981804 genotype [AA: n = 11 patients, 36 biopsies, 21 not inflamed, 15 inflamed; GA: n = 14 patients, 37 biopsies, 22 not inflamed, 15 inflamed; GG: n = 2 patients, 2 biopsies (both not inflamed)]. Overall, a trend towards higher DMBT1 expression in carriers of the protective G allele was observed (Fig. 7A, p = 0.06, AA vs. GA+GG). An inflammation-induced increase in DMBT1 expression was observed for both rs2981804 AA and GA carriers (Fig. 7B). However, basal DMBT1 expression was lower in homozygous carriers of the IBD risk allele (AA) than in GA and GG carriers (p = 0.052 AA vs. GA, p = 0.03 AA vs. GA+GG) (Fig. 7B). Similar results were obtained    1-7), the addition of 50-fold excess of unlabelled DMBT1 probe inhibited protein binding (lane 7) as well as did unlabelled CREB1 probe or a CREB1 antibody (lanes 4, 6). The protein complex bound to a labelled DMBT1 probe (lanes 8-9) is migrating slower (and therefore larger) than that of the CREB1 probe. (D) Silencing of ATF-2 or CREB1 expression diminished protein binding to the DMBT1 probe. DLD-1 cells were transfected with siRNA against CREB1, ATF-2 or a non-specific control siRNA 48 h prior to nuclear protein isolation. EMSA was performed as in Fig. 2A  when NOD2 SNP carriers (see table 2) were excluded from analysis (Fig. 7C).

Discussion
In this study, we performed a detailed functional analysis of the DMBT1 gene. First, we investigated the effect of the Th17 cytokine IL-22 on DMBT1 expression. In these detailed experiments, we confirmed DMBT1 as a target gene of the cytokine IL-22. Our results are supported by a recent study [22]. IL-22 is produced by IL-23R-expressing Th17 cells [36] and we have recently shown that variants in IL23R influencing the CD risk are associated with differential IL-22 expression [34]. The same IL23R variants are also associated with differential intestinal DMBT1 expression suggesting a functional link between IL23R genotype, IL-22 expression and DMBT1 expression. A very recent study demonstrated that DMBT1 expression in intestinal inflammation is influenced by antibody treatment against IL-23R or IL-23 [24], further supporting this hypothesis. Interestingly, we recently demonstrated that the clinical response to infliximab, an anti-TNF-a antibody used for treatment of IBD, is associated with certain IL23R genotypes in UC patients [41].
In this study, we show that several previously not analyzed DMBT1 SNPs and their haplotypes are associated with the susceptibility to IBD. The two most strongly IBD-associated DMBT1 SNPs were the non-coding SNPs rs2981745 and rs2981804 for which the minor allele frequencies were significantly higher in CD and UC patients compared to controls. Associations remained significant even after conditioning the analysis on the most strongly-associated SNP for each disease. Considering the weak LD between both SNPs (r 2 = 0.24, 0.18 and 0.17 for CD, UC, and controls, respectively), our data suggest largely independent associations with IBD susceptibility. We also demonstrated for a large number of DMBT1 haplotypes highly significant associations with CD and UC, with P-values as low as 10 218 . A recent study demonstrated that a deletion variant in the repetitive region of DMBT1 encoding the SRCR domains is associated with CD [20]. Although it can not completely be excluded that the SNPs analyzed in our study are in LD with this variant, it is rather unlikely. SNP rs2277244 analyzed in our study is localized within the repetitive SRCR genomic region and linkage analysis proved that this SNP is not in LD with any other SNP examined here (r 2 #0.02) suggesting that the SNPs analyzed in our study are associated with IBD susceptibility independent of the SRCR region. However, it has to be mentioned that our association results need to be confirmed by replication in other large casecontrol cohorts for both CD and UC. In recent GWAS studies and meta-analyses, the DMBT1 gene region did not appear as major risk factor for IBD [9,32,42,43]. One reason might be ethnic differences between study population as it has been described for other genes such as PHOX2B, NCF4 and FAM92B or DLG5 [44,45,46,47]. Another fact that might contribute to that lack is the poor coverage of the DMBT1 gene region on the first available GWAS chips, likely due to the many repetitive genomic regions within the gene. Moreover, even though the most recent and so far largest GWAS meta-analysis identified 163 IBD-associated variants, these genes explain only a small part of the IBD risk variation observed. Rare haplotypes conferring a highly significant IBD risk (as we have observed in our study) might be missed.
Interestingly, a very recent study by Dinu et al. demonstrated that a considerable number of genes or chromosomal regions contribute to CD risk through SNP-SNP interactions [48]. Many of those genes were not identified in recent GWAS analyses including more than 20,000 patients and controls as these studies analyzed only single SNP associations [48]. Remarkably, Dinu et al. identified several so far unknown CD risk genes (FGFR2, FOXI2, GLRX3) on chromosome 10 q26, surrounding the chromosomal region that also harbours the DMBT1 gene [48].
Interestingly, both most strongly IBD-associated DMBT SNPs in our study (rs2981745 and rs2981804) are located in non-coding genomic regions, thereby not altering DMBT1 protein structure or function. Therefore, we hypothesized that these SNPs might be located within recognition sequences of transcription factors. For the first time, we demonstrated that the transcription factors CREB1 and ATF-2 differentially bind to DNA probes containing either the IBD risk allele A or the protective G allele of SNP rs2981804. Moreover, we show that CREB1 and ATF-2 are involved in the transcriptional regulation of IL-22-induced DMBT1 expression. We show in our study that the IBD risk allele of rs2981804 is associated with a lower DMBT1 gene expression in colonic tissue from CD patients identifying for the  There was a trend towards higher DMBT1 expression in carriers of the G allele (GA+GG) in comparison to AA carriers (p = 0.06). Each dot represents one biopsy and the black horizontal bar is the median of each group. (B) When biopsies were subdivided according to inflammation status, homozygous carriers of the AA risk allele of SNP rs2981804 had lower basal and inflammationinduced DMBT1 expression levels in comparison to carriers of the G allele (p = 0.03 AA vs. GA+GG). (C) DMBT1 mRNA expression is independent of NOD2 genotype status. DMBT1 mRNA expression was determined in biopsies from CD patients irrespective of NOD2 genotype (n = 75; black bars) or excluding biopsies from NOD2 SNP carriers (n = 45 NOD2 wild-type carriers; grey bars). Data are presented as median DMBT1 expression in the respective groups. There were no significant differences in DMBT1 expression between all biopsies and NOD2 wildtype carriers for the different DMBT1 rs2981804 genotypes. n.s. = not significant;`: no standard deviation is given for GG NOD2 wild-type carriers as this group comprised only one biopsy. doi:10.1371/journal.pone.0077773.g007 first time a link between rs2981804 alleles and DMBT1 gene expression. These results were independent of the NOD2 genotype.
Our luciferase assays revealed that the genomic region comprising SNP rs2981804 can act either as a transcriptional activator or repressor, depending on the promoter context. Given that enhancers or repressors usually do not induce or repress gene expression directly but interact with promoter-specific transcription factors, such differences can be explained. Similar results have been described for example for the c-myc enhancer [49]. There is the possibility that the respective genomic region comprising DMBT1 SNP rs2981804 not solely influences DMBT1 expression. As enhancers/repressors can be located several hundred kb away from genes which they regulate [50], the expression of other genes located within this chromosomal region might be influenced as well. Moreover, CREB1 and ATF-2 are transcription factor that are ubiquitously expressed and are involved in the regulation of many genes and cellular processes such as immunity, cell proliferation, differentiation, and cell survival [51]. Therefore, further analysis is necessary to determine whether the genomic region comprising DMBT1 SNP rs2981804 influences expression of other genes in vivo. Moreover, the differential influence of this SNP on DMBT1 expression, especially in the presence of different transcription factors and in the context of different cytokine environments that have a strong influence on DMBT1 expression [12,22,23], should be aims of future analyses.
In conclusion, we demonstrated novel significant associations of DMBT1 variants and haplotypes with the susceptibility to CD. Moreover, we provide for the first time functional evidence that the non-coding DMBT1 SNP rs2981804 modifies the binding sites for the transcription factors CREB1 and ATF-2. This DMBT1 SNP is associated with decreased overall DMBT1 expression in the colon thereby probably contributing to increased CD susceptibility. As DMBT1 encodes -like NOD2 -an antibacterial pattern recognition receptor [11], our results support the hypothesis that a dysregulated antibacterial response of the innate immune system might contribute to the pathogenesis of CD. Figure S1 Overview of the DMBT1 luciferase reporter constructs in a vector with a minimal promoter-driven luciferase expression. For cloning details, see supplementary methods. AmpR, ampicillin resistance gene; luc2, luciferase gene; ori, origin of replication; minP, minimal promoter. (DOC) Figure S2 Overview of the used DMBT1 luciferase reporter constructs with an SV40 promoter driven luciferase expression. For cloning details, see supplementary methods. AmpR, ampicillin resistance gene; luc2, luciferase gene; ori, origin of replication. (DOC) Figure S3 IL23R CD risk-increasing variants (OR.1) are associated with higher DMBT1 expression in minor allele carriers (minor vs. WT.1) independent of inflammation status. Intestinal DMBT1 expression and IL23R genotypes were determined in 75 biopsies from 27 CD patients. DMBT1 expression is presented as a quotient derived from dividing expression in minor allele carriers of the respective IL23R SNP by the expression in WT carriers. The ORs for the respective SNPs were available from a previous study [8].   Table S8 Association between DMBT1 rs2981745 genotypes and CD disease characteristics in the subcohort of the Munich IBD center (n = 628) for which detailed phenotypic data based on the Montreal classification were available. For each variable, the number of patients included is given. P T , P-value for testing for differences between carriers and non-carriers of the T allele. OR T : corresponding odds ratios and 95% confidence intervals (95% CI). For age at diagnosis, P-values are given based on a median split. Significant P-values are depicted in bold. However, after Bonferroni correction for multiple testing, significance was lost. 1 Disease behaviour was defined according to the Montreal classification. A stricturing disease phenotype was defined as presence of stenosis without penetrating disease. The diagnosis of stenosis was made surgically, endoscopically, or radiologically (using MRI enteroclysis). 2 Immunosuppressive agents included azathioprine, 6-mercaptopurine, 6-thioguanin, methotrexate, infliximab and/or adalimumab. 3 Only surgery related to CD-specific problems (e.g. fistulectomy, colectomy, ileostomy) was included. (DOC) Table S9 Association between DMBT1 rs2981745 genotypes and UC disease characteristics in the subcohort of the Munich IBD center (n = 283) for which detailed phenotypic data based on the Montreal classification were available. For each variable, the number of patients included is given. P T , P-value for testing for differences between carriers and non-carriers of the T allele. OR T : corresponding odds ratios and 95% confidence intervals (95% CI). For age at diagnosis, age and BMI P-values are given based on a median split. Significant association is highlighted in bold. However, after Bonferroni correction for multiple testing, this significance was lost.

(DOC)
Table S10 Association between DMBT1 rs2981804 genotypes and CD disease characteristics in the subcohort of the Munich IBD center (n = 626) for which detailed phenotypic data based on the Montreal classification were available. For each variable, the number of patients included is given. P G , P-value for testing for differences between carriers and non-carriers of the G allele. OR G : corresponding odds ratios and 95% confidence intervals (95% CI). For age at diagnosis P-values are given based on a median split. 1 Disease behaviour was defined according to the Montreal classification. A stricturing disease phenotype was defined as presence of stenosis without penetrating disease. The diagnosis of stenosis was made surgically, endoscopically, or radiologically (using MRI enteroclysis). 2 Immunosuppressive agents included azathioprine, 6-mercaptopurine, 6-thioguanin, methotrexate, infliximab and/or adalimumab. 3 Only surgery related to CD-specific problems (e.g. fistulectomy, colectomy, ileostomy) was included. (DOC) Table S11 DMBT1 gene markers in CD -Haplotype frequencies (HF), P -values, and odds ratios (OR) with 95% confidence intervals (CI). P-values for individual haplotypes are presented for all haplotypes with a frequency of at least 1% in the whole sample and with an omnibus haplotype Pvalue ,0.05. Significant P-values (,0.05) are highlighted in bold and significant P-values robust to multiple testing (P,2.5610 23 for omnibus P-values, P,4.8610 24 for detailed haplotype P-values) are highlighted in bold italic.

(DOC)
Table S12 DMBT1 gene markers in UC -Haplotype frequencies (HF), P-values, and odds ratios (OR) with 95% confidence intervals (CI). P-values for individual haplotypes are presented for all haplotypes with a frequency of at least 1% in the whole sample and with an omnibus haplotype Pvalue ,0.05. Significant P-values (,0.05) are highlighted in bold and significant P-values robust to multiple testing (P,2.5610 23 for omnibus P-values, P,4.810 24 for detailed haplotype P-values) are highlighted in bold italic.
(DOC) Table S14 Analysis for epistasis between SNPs rs2066844 = p.Arg702Trp, rs2066845 = p.Gly908Arg and rs2066847 = p.Leu1007fsX1008 in the NOD2 gene and the SNPs DMBT1 rs2981745 and rs2981804 within the DMBT1 gene regarding CD/UC susceptibility. All Pvalues given are uncorrected for multiple comparisons.