Human intelectin-1 (ITLN1) genetic variation and intestinal expression

Intelectins are ancient carbohydrate binding proteins, spanning chordate evolution and implicated in multiple human diseases. Previous GWAS have linked SNPs in ITLN1 (also known as omentin) with susceptibility to Crohn's disease (CD); however, analysis of possible functional significance of SNPs at this locus is lacking. Using the Ensembl database, pairwise linkage disequilibrium (LD) analyses indicated that several disease-associated SNPs at the ITLN1 locus, including SNPs in CD244 and Ly9, were in LD. The alleles comprising the risk haplotype are the major alleles in European (67%), but minor alleles in African superpopulations. Neither ITLN1 mRNA nor protein abundance in intestinal tissue, which we confirm as goblet-cell derived, was altered in the CD samples overall nor when samples were analyzed according to genotype. Moreover, the missense variant V109D does not influence ITLN1 glycan binding to the glycan β-D-galactofuranose or protein–protein oligomerization. Taken together, our data are an important step in defining the role(s) of the CD-risk haplotype by determining that risk is unlikely to be due to changes in ITLN1 carbohydrate recognition, protein oligomerization, or expression levels in intestinal mucosa. Our findings suggest that the relationship between the genomic data and disease arises from changes in CD244 or Ly9 biology, differences in ITLN1 expression in other tissues, or an alteration in ITLN1 interaction with other proteins.


Results
The intelectin-1 locus and Crohn's disease. To gain insight into the potential influence of IBD-associated allelic variants on expression and protein function, we genotyped surgical specimens from a previously reported cohort of individuals requiring surgical resection for Crohn's disease or ulcerative colitis (n = 134) 29 . Accordingly, we found by direct sequence analysis of PCR products that the GWAS-identified SNP, rs2274910 (T/C), located in intron 3 of ITLN1, was associated with rs2274907 (A/T), which encodes a missense variant in exon 4, where valine is substituted with aspartic acid at amino acid position 109 (V109D) 1,30 . Utilizing the 1000 Genomes Project Phase 3 of the Ensembl database we found rs2274910 and rs2274907 are in linkage disequilibrium (LD) across human populations, consistent with findings of association in our U.S. cohort (Supplemental Table 1) 31,32 . We sequenced the entire coding region of ITLN1 cDNA in all specimens of this cohort, exploring the potential for additional exon variants in IBD subpopulations. We found rs2274908 (C/T: H89H/Q) exclusively as a synonymous (silent) variant associated with rs2274907 30 . No additional missense variants were identified.
To display allelic variants associated with disease, we constructed a locus map of ITLN1 (Fig. 1A) annotated with published SNPs as outlined by the National Human Genome Research Institute-European Bioinformatics Institute (NHGRI-EBI) Catalog and NCBI-LitVar database, including variants within ITLN1 and its flanking regions intergenic to CD244 (3') and ITLN2 (5') 33,34 . ITLN1 and CD244 (rs1333062 and rs4656940, respectively) have been annotated together as risk loci for IBD, due in part to their close proximity (~ 14,000 nt) 23,24 . The SNP rs133062 is located in the 3'-flanking region of ITLN1, 46 nt intergenic to exon 8, where rs4656940 is located within intron 1 of CD244. The odds ratio (OR) for Crohn's disease of rs4656940 is approximately equivalent (1.15; 95% confidence interval [CI] = 1.09-1.21) to that of rs2274910 (1.14; 95% CI = NR) of ITLN1 1,23 . Several annotated SNPs in CD244 (rs4656942 and rs11265498) and in the adjacent gene LY9 (rs540254 and rs560681), but not any in neighboring SLAMF7, are in LD with GWAS-identified rs2274910 and rs4656940 (Supplemental Table 2) 35 . In addition, rs4656958 located in the 5' flanking region of ITLN1 (2004 nt 5' to exon 1) has also been identified by GWAS as a risk allele for IBD (OR = 1.06; 95% CI = 1.03-1.09) [25][26][27] . Thus, GWAS-identified IBD-associated SNPs are located within ITLN1 and its flanking regions, and these SNPs are in LD with SNPs in CD244 and Ly9, which are genes located 3' to ITLN1. Data in the GTEx browser revealed that quantitative trait loci (QTL) for expression (eQTL) in blood for ITLN1 and CD244 are both affected by several disease risk SNPs, including rs1333062, rs2274907 and rs2297559 (Supplemental Table 3) 36 . In addition, QTL for splicing (sQTL) of USF1 was similarly affected by these SNPs, as was sQTL for Ly9 by rs465940. No significant eQTLs or sQTLs were identified in small intestine or colon tissue 36 . Although other genes located at 1q23.3 may be responsible for the disease association, we chose to investigate ITLN1 as the strongest candidate based on our prior knowledge of the function of intelectins.
Further review of ITLN1 rs2274910 (C/T) revealed a distinction in the frequency of the GWAS-identified risk allele (C) between African (28%) and non-African (67%) superpopulations (SP) 31,32 . This prompted us to further explore allele frequency in the context of different SP groups. The study populations for most IBD-focused GWAS are comprised primarily of non-African SP, including investigations identifying ITLN1 as a risk locus 1,23,[25][26][27]37 . Using the Ensembl database, we performed pairwise LD analyses of disease-associated ITLN1 variants (Fig. 1a, Table 1) in African and European SP using rs2274907 (A/T: V109D) as a focus point simply due to it being a missense variant. In addition to being in complete LD (D' = 1.0, r 2 = 1.0) with rs2274910, rs2274907 (V109D) was found also to be in LD with GWAS-identified alleles across the ITLN1 locus, where risk alleles are the major alleles in European SP and minor alleles in African SP (Table 1, Fig. 1b) 31,32,35 .
The allele genotype frequencies in European SP: VV (7.6%), VD (49.3%), and DD (43.1%) contrasts sharply with that in the African SP VV (45.7%), VD (43.1%), and DD (11.2%) (Supplemental Fig. 1). In our small cohort of surgical specimens, allele frequencies were not significantly different for those diagnosed as either Crohn's disease or ulcerative colitis as compared to respective controls, but there was a suggestion of skewing in specimens diagnosed as ileal compared to colonic Crohn's disease, worthy of follow-up analysis in a larger cohort (Supplemental Table 1). Our findings clarify that rs2274907 (T, D109), which is in complete LD with GWAS-identified rs2274910 risk allele (C), is the major allele in non-African SP; in Crohn's disease multiple GWAS report a further skewing of this high frequency of the risk allele observed in the studied populations 1,23-27 .   www.nature.com/scientificreports/ therapy-resistant Crohn's disease cohort 38 . Their analysis indicated eQTLs for ITLN1 in intestine (but not blood) are linked to GWAS-identified risk alleles rs4656958 and rs2274910, and eQTLs for CD244 in blood (but not intestine) are linked to the risk alleles rs4656958, rs4656940 and rs2274910. In both cases, a statistically significant decrease in expression is associated with the GWAS-identified SNPs. To better clarify expression patterns of ITLN1 within intestinal tissue, we generated an ITLN1-specific quantitative real-time PCR assay to enumerate the absolute mRNA transcript levels. ITLN1 mRNA was abundant in each anatomic segment across both small and large intestine (Supplemental Fig. 2a). Similarly, high levels of ITLN1 gene expression were observed in surgical specimens of both ileum and colon (Fig. 2a). Crohn's disease diagnosis did not alter transcript levels in either small or large intestine (Fig. 2a). Of note, ITLN1 expression in the colon was elevated (p < 0.05) in specimens from individuals diagnosed with ulcerative colitis (Fig. 2a). To determine whether ITLN1 expression is modified by risk-associated haplotype, we stratified the data by genotype. In ileum, colon, and combined sample sets, the risk associated haplotype (queried by rs2274907 genotype) did not modify ITLN1 expression ( Fig. 2b and Supplemental Fig. 2b). Together, these findings suggest that the mucosal transcript level of ITLN1 is unaltered in Crohn's disease. Moreover, the risk-associated haplotype does not modify mucosal ITLN1 expression ( Table 1). The distinguishing chemical properties of valine and aspartic acid provided impetus to further explore the potential of missense variant V109D to modify ITLN1 oligomeric structure or function. Using PyMOL to visualize the published crystal structure of human ITLN1 (PBD: 4WMY), residue 109 was found to be located at the protein surface; however, it is not located at the sites annotated for glycan binding and/or calcium coordination ( Fig. 1c) 20,39 . Bio-layer interferometry analysis found that recombinant V109-ITLN1 reversibly bound to immobilized β-D-galactofuranose with kinetics and affinity that mirror recombinant D109-ITLN1 (117 nM versus 107 nM, respectively) ( Fig. 2c). Without finding discernable influence on glycan binding, we next tested if the V109D change altered ITLN1 oligomerization state and stability. For other secreted lectins (e.g., ficolins and mannose-binding lectin), oligomerization is important for avidity in binding cell-surface glycans as well as subsequent effector functions [40][41][42] . Previous studies report that human ITLN1 forms stable homotrimers, consisting of ~ 35 kDa monomers, connected through intermolecular disulfides (C31 and C48) 6,11,43 . Analysis of human ileal lysates confirmed that ITLN1 forms a stable homotrimer at approximately 110 kDa; however, a consistent ITLN1 band near the ~ 220 kDa marker was also identified using a gradient SDS-PAGE (4-20%). This suggests that human ITLN1 forms higher molecular weight oligomers (likely dimers of homotrimers), which are stable under denaturing (but not reducing) conditions (Fig. 2d). These findings agree with those reported for an amphibian ITLN1 ortholog, where dimers of homotrimers were shown to mediate bacterial agglutination 40 . We also performed immunoblots of ileal specimens isolated from V109 and D109 homozygote individuals to determine the potential influence on oligomer formation. Neither VV or DD genotype affected the formation of homotrimers or hexamers (Fig. 2d). These data suggest that that it is unlikely that the role of V109D variation in Crohn's disease pathogenesis results from alterations in ITLN1 glycan binding or oligomerization.
Alternatively, we reasoned that D109 might influence protein stability, and that such differences would be observed as changes in protein abundance. In human ileal samples, we observed that under reducing conditions ITLN1 was present primarily as a monomer at ~ 40 kDa, with the exception of a minor fraction migrating at ~ 70 kDa (Fig. 2d). This minor band may represent a covalent non-reducible dimer but further study is needed to fully characterize this entity. Human ITLN1 possesses a single glycosylation site at N163, which slightly increases the molecular weight from that predicted (33 kDa) by the deduced amino acid sequence 43 . Ileal samples from individuals diagnosed with Crohn's disease demonstrated equivalent relative proportions of protein oligomers and no additional degradation products compared to control specimens (Fig. 2d). Thus, ITLN1 was found to be an abundant protein in the small intestine, where expression was comparable in specimens from control and Crohn's disease biopsies, consistent with mRNA expression data ( Fig. 2a, d).
To further investigate the possible significance of V109D variation in Crohn's disease, we performed a sequence analysis (NCBI BLAST) of the neighboring amino acid region across various ITLN1-like orthologs. In all Eutherian mammals queried (n = 63), aspartic acid (D) was reported as the amino acid at the corresponding position (Fig. 1d). Even amongst primates, valine at position 109 is unique to Homo sapiens. Nevertheless, other chordates were found to encode hydrophobic amino acids like valine (e.g., leucine) at this position, including ancient ascidians (Fig. 1d). While aspartic acid marks the risk-allele associated with Crohn's disease, the V109D variation is not a "mutation", despite its location in a conserved domain of the protein and differences in chemical properties of these residues. Furthermore, V109, or the genetic variants in LD with this allele, could be considered "protective" in the context of Crohn's disease.
Human intelectin-1 is a goblet cell product. In mice, where mammalian intelectins were first discovered, Itln1 is expressed in Paneth cells 8 . In contrast, a recent single-cell RNA sequence analysis investigation identified ITLN1 mRNA in goblet cells of the ileum and colon 44 . We sought to localize ITLN1 protein in this these tissues to confirm and extend this cell-type expression pattern in humans. Both ITLN1 and ITLN2 are expressed in the human small intestine and cannot be distinguished using most commercially available antibodies; therefore, we generated paralog-specific antibodies to clarify the cellular origin of ITLN1. Colon biopsies of non-IBD subjects, preserved in fixatives to retain mucus (either Carnoy's fixative or HistoChoice™), were incubated with either antisera targeting a domain highly conserved in intelectins (ITLN CONS ) or targeting the N-terminal specific domain (ITLN1 NTERM ). Both ITLN CONS and ITLN1 NTERM immunoreactivity were localized to colonic goblet cells (Fig. 2e, Supplemental Fig. 2c). To confirm goblet cell localization, samples were stained with a mucin-2 (MUC2) antibody generated against a peptide sequence located within the C-terminal domain (adjacent to the autocatalytic sequence) of the mature, glycosylated MUC2 protein (MUC2 C3 ); this antibody targets goblet cell granules and secreted protein 45  www.nature.com/scientificreports/ www.nature.com/scientificreports/ granule, the cellular staining pattern of ITLN1 was different. Specifically, punctate staining of ITLN1 adjacent to the granule and relatively weak staining within the granule was a feature of all ITLN1-positive samples tested (n = 6 colon, n = 3 ileum). To further characterize the staining pattern of ITLN1 within goblet cells, we used another MUC2 antibody that targets the tandem repeat region of the apo-MUC2 protein (PH1900, MUC2 APO ). This antibody identifies non-O-glycosylated MUC2, which resides within the endoplasmic reticulum region of goblet cells 46,47 . MUC2 APO staining of the cytoplasmic region in goblet cells mimicked the staining pattern for ITLN1 NTERM (Fig. 2e, Supplemental Fig. 2C). At higher magnification, ITLN1 NTERM staining was identified in both cytoplasmic and granule compartments, and readily apparent in both longitudinally and transversely orientated tissue sections (Fig. 2e). To better localize ITLN1 within goblet cells, we generated a chicken anti-ITLN antibody (conserved, ITLN CONS-FLOR ) for dual labeling with the rabbit-derived MUC2 C3 reagent in fluorescence microscopy. As observed by conventional microscopy, ITLN1 and MUC2 largely colocalized within colonic goblet cells, where ITLN1 was also found occasionally in the cytosolic compartments adjacent to nuclear (DAPI) staining (Fig. 2f). Immunohistochemistry of human ileum using the ITLN1 NTERM antibody also identified goblet cells, where the aforementioned cytosolic and granular staining pattern was also evident (Fig. 2g). It is worth noting that as a feature distinct from findings in mice, human Paneth cells remained unstained with the ITLN1 NTERM antibody (Supplemental Fig. 2D). Although staining of human ileum with the ITLN CONS antibody identified Paneth cells in addition to goblet cells, paralog-specific antibodies determined Paneth cell staining to be human ITLN2 (Fig. 2g,

Discussion
Various GWAS investigations have identified SNPs at the ITLN1 locus as risk alleles for Crohn's disease 1,23-27 . We observed that these SNPs are in LD, forming a risk haplotype that extends into the adjacent CD244 and Ly9 genes. Given the abundant expression of ITLN1 in intestinal tissues and the likely role of this lectin in innate immunity, we focused our investigation on ITLN1 in the intestine. The first GWAS-identified SNP at this locus (rs2274910) is located in intron 3 of ITLN1 1 . We determined that this variant was associated with an adjacent SNP in exon 4 (rs2274907), which encodes a valine to aspartic acid amino change at residue 109. Although valine is the most common allele in African SP, aspartic acid is both the most common and the disease-associated allele in European SP-the primary population studied in IBD GWAS 1,23-27 . Our results clarify that V109D is in LD with other GWAS-identified Crohn's disease-associated risk SNPs, resulting in a risk haplotype. Our data suggest that mRNA transcript and protein abundance of mucosal ITLN1, now confirmed to be goblet-cell derived, is unaltered in the Crohn's disease samples overall, as well as in samples analyzed according to rs2274907 (A/T, V109D) genotype in both disease and non-disease specimens. We clarify that ITLN1 D109 is the ancestral amino acid in mammalian orthologs, supporting that D109 is not a mutation. Furthermore, our data indicate that any role(s) of V109D in Crohn's disease pathogenesis is unlikely due to changes in ITLN1 carbohydrate recognition or protein oligomerization.
Further investigation is required to determine the mechanism by which genetic variation at this locus affects risk for Crohn's disease. In this study, we tested relevant possibilities, and our data support that it is unlikely to be due to ITLN1 expression in intestinal goblet cells. It is possible that expression of ITLN1 in another tissue could underlie disease risk. Moreover, expression of other genes at this locus, notably either CD244 or Ly9, which include SNPs in LD with the risk haplotype might explain disease risk [33][34][35] . Based on available QTL data, expression of USF1 or CD244 in whole blood would also be reasonable candidates for future investigation 36 .

Methods
Human specimens and ethics statement. All methods were carried out in accordance with relevant guidelines and regulations. Surgical specimens were obtained at The Cleveland Clinic Foundation (Cleveland, OH) with informed consent from all patients, and deidentified under protocols approved by the Institutional Review Board at The Cleveland Clinic Foundation as previously described 29 . These specimens were collected from subjects undergoing surgery for Crohn's disease (CD), ulcerative colitis (UC), or non-CD/UC-related (control, CTRL) surgical indications (i.e., colon cancer, bowel obstruction, and familial adenomatous polyposis). Race was recorded for these specimens based on medical record reporting. Diagnoses were based on standard criteria using clinical, radiological, endoscopic and histopathological findings 29 . Exclusion criteria included the diagnoses of backwash ileitis, indeterminate colitis, concurrent CMV or C. difficile infection. Samples were immediately either snap frozen in liquid nitrogen (for RNA, DNA, and protein analysis) or placed into tissue fixative for histology. The mucosa, utilized for mRNA and protein expression, was dissected away from the intestinal/colonic serosa and associated adipose tissue. The overview of regional mRNA expression (Suppl. Figure 1) used commercially-obtained RNAs of human tissue (Biochain, Newark, CA). Genotyping. Genomic DNA was isolated and purified using the QIAamp DNA minikit (Qiagen, Germantown, MD) according to the manufacturer's protocol. Isolated DNA was quantified by ultraviolet absorbance spectroscopy (260 nm) using a NanoDrop spectrophotometer (Thermo Scientific/NanoDrop Products, Wilmington, DE). Exon variants in each specimen of our IBD cohort were explored by Sanger sequencing of cDNA corresponding to the entire coding region of ITLN1 amplified using specific primers (

RNA isolation.
The general procedures for RNA isolation and synthesis of cDNA were previously described by our group 51,52 . For the current study, RNA was freshly isolated for all specimens from frozen tissue 29 . Briefly, frozen tissue was dispersed using Brinkmann Polytron homogenizer in guanidinium thiocyanate (GTC) buffer  Software. Real-time PCR was performed on cDNA template corresponding to 10 ng of RNA using a Roche Diagnostics Lightcycler 2.0 (Roche, Indianapolis, IN) as previously described 51,52 . Briefly, each reaction included 4 mM MgCl2, 0.5 μM forward and reverse primers, and 1 × LightCycler FastStart SYBR Green I mix (Roche, Mannheim, Germany). A control without cDNA template was included in the procedural assay. The PCR conditions were: initial denaturation at 95 °C for 10 min, followed by 45 cycles with each cycle consisting of denaturation, 95 °C for 15 s; annealing at 60 °C for 5 s; and extension at 72 °C for 10 s. Following the cycle runs, samples were denatured to establish the melting temperature(s) of the PCR product. The sample melt temperatures were compared to that of the internal standard to confirm template specificity. Absolute quantification of target gene within each sample was obtained by inclusion of an external cloned cDNA standard of known concentration, where crossing point (i.e., cycle number) is matched to a standard curve. Data was plotted in Prism version 9.0.0 (GraphPad).
Rabbit anti-ITLN1 polyclonal antibodies. Analysis of the pre-ITLN1 deduced protein sequence using the SignalP4.0 algorithm indicated signal cleavage would occur between G16 and W17 53  www.nature.com/scientificreports/ nating 19-amino acid sequence (W17 to P35) of ITLN1 is a domain beginning at amino acid S36 that has very high sequence identity to ITLN2 as well as other intelectin orthologs. BLASTp search of the non-redundant protein database with the 19-amino acid W17 to P35 peptide sequence yielded numerous primate ITLN1 orthologs with highly similar sequence (> 80%), but no other mammalian proteins. This 19-amino acid peptide was synthetically prepared, purified and characterized to confirm proper mass, as previously described 54 . The peptide was coupled to ovalbumin (OVA, Thermo Fischer Scientific, Waltham, MA) at a ratio of 20 molecules peptide per molecule OVA using two strategies. First, one equivalent volume of 0.2% glutaraldehyde was added to the 20:1 peptide/ovalbumin mixture in PBS; the mixture was then stirred for 1 h at room temperature. 1 M glycine was added to a final concentration of 200 mM and stirred for an additional 1 h. Second, the peptide was reacted with maleimide-activated-ovalbumin (Imject™ Maleimide-Activated Ovalbumin, Thermo Fischer Scientific, Waltham, MA) according to the suppliers' protocol. Briefly, the peptide was mixed with the activated ovalbumin at a 20:1 molar ratio in PBS and allowed to react at room temperature for 2 h. Each of the two peptide/OVA conjugates were dialyzed against PBS for 24 h at 4 °C, changing dialysis buffer every 6-8 h. The protein concentration was adjusted to 750 µg/ml with PBS and then equal portions of the two peptide/OVA conjugates were mixed to then use as antigen. In a separate set of reactions, the peptide was coupled to keyhole limpet hemocyanin (KLH, Thermo Fischer Scientific, Waltham, MA) at a ratio of 200 molecules peptide per molecule KLH using the identical protocols (glutaraldehyde and maleimide-activated-KLH (Imject™ Maleimide-Activated KLH, Thermo Fischer Scientific, Waltham, MA) and equal portions of the resulting two peptide/KLH conjugates were mixed to use as antigen. Rabbits were immunized by Antibodies Inc. (Davis, CA) with the ITLN1 peptide/OVA antigen on days 1, 14, 21, and then boosted by immunization with ITLN1 peptide/KLH antigen on days 35 and 49 to produce the N-terminal polyclonal antisera (termed ITLN1 NTERM ). Preimmunization sera was collected and used as control antisera. All methods for these procedures were in accordance with relevant guidelines and regulations, and approved by the U.S. Department of Health and Human Services Public Health Service Animal Welfare Assurance Committee (Assurance ID D16-00576). Rabbit polyclonal antiserum generated by immunization with human ITLN1-derived synthetic peptides was described and characterized previously (termed ITLN CONS ) 11,55 .
Chicken anti-ITLN1 IgY. The ITLN1 protein sequence was analyzed using the online algorithm for design of peptide-directed antibodies provided by the NHLBI (https:// hpcwe bapps. cit. nih. gov/ AbDes igner/) 56 . A 28 aa peptide (CTVGDRWSSQQGSKADYPEGDGNWANYN) was identified as an attractive antigen candidate. This peptide spanned the V108D polymorphic site, and was highly similar in sequence to human ITLN2 and C57BL/6 mouse Itln1 (27/28 identical residues to both of these orthologs). This peptide was synthetically prepared, purified and characterized 54 . In two separate reactions, the peptide was coupled to KLH, using the two chemical coupling strategies (glutaraldehyde and maleimide) and protocols, as described above. Each reaction product was dialyzed against PBS and protein concentration adjusted to 750 µg/ml. Equal portions of the two peptide/KLH conjugates were mixed to use as antigen. Leghorn chickens were given a primary inoculation with the KLH conjugate in complete Freund's adjuvant, followed by two booster inoculations at 3 and 6 weeks (with incomplete Freund's adjuvant) by Aves Labs, Inc. (Davis, CA). An enriched IgY fraction from egg yolk of immunized chickens (termed ITLN CONS-FLOR ) was obtained by organic extraction de-lipidation, high-salt precipitation and then extensive dialysis against PBS.
Bio-layer interferometry. Dissociation constants for recombinant ITLN1 with biotin-β-D-galactofuranose were determined using a bio-layer interferometry OctetRed instrument (ForteBio, Fremont, CA). Biotin-β-Dgalactofuranose was diluted to 5 μM in PBS, pH 7.5, and loaded (120 s) onto Streptavidin biosensors (Forte Bio www.nature.com/scientificreports/ Protein Standard (ThermoFisher Scientific). Membranes were blocked in PBS-T containing 5% w/v skim milk for 30 min, and incubated with primary antibodies at 4 °C. Following overnight incubation, membranes were washed with PBS-T for 1 h, probed with donkey anti-rabbit IgG horseradish peroxidase (HRP) linked secondary antibody (GE Healthcare Amersham, Pittsburgh, PA) for 3 h at room temperature, rinsed in PBS-T, and visualized using Femto and ECL West chemiluminescent substrates (ThermoFisher Scientific). Chemiluminescent signal was detected using a Biospectrum AC Imaging System (UVP, Upland, CA).