Evidence of association with type 1 diabetes in the SLC11A1 gene region

Type 1 diabetes (T1D) is a heritable polygenic autoimmune disease in which both genetic and environmental factors contribute to pathogenesis. To date over 50 loci have been identified that affect risk of T1D [1–3]. The causal genes, variants and haplotypes involved within many of these regions have yet to be identified.

The nonobese diabetic (NOD) mouse develops autoimmune insulin-dependent diabetes spontaneously, which resembles human T1D, and has been widely used to study and map non-MHC T1D loci. The loci identified in the NOD mouse as high priority candidates for human T1D studies include solute carrier family 11 member 1 (Slc11a1), which was formerly known as natural resistance-associated macrophage protein 1 (Nramp1) [4]. Slc11a1 is encoded in the Idd5.2 region on mouse chromosome 1. The NOD strain with the disease-predisposing allele expresses the functional protein, whereas the T1D-resistant B10 strain does not [5, 6]. The NOD allele was originally identified as providing resistance to bacterial infections in mice. Kissler et al. used RNA interference to reduce Slc11a1 expression in vivo in NOD mice, and found that this reduced the frequency of T1D, mimicking the protective Idd5.2 T1D-resistant haplotype [7]. Furthermore, Slc11a1 was found to augment activation of a diabetogenic T-cell clone by enhancing the processing and presentation of pancreatic islet antigens, such as glutamic acid decarboxylase GAD₆₅, in dendritic cells (DCs) [8].

In humans, SLC11A1 is 14 kb in length with 15 exons. The gene is located in an approximately 400 kb region of high linkage disequilibrium (LD) on chromosome 2q35. The human and mouse SLC11A1 protein sequences have a high degree of conservation, with 88% identity and 93% overall sequence similarity [9]. SLC11A1 is expressed in monocytes [10, 11], which are the circulating precursors of macrophages and DCs, the major antigen-presenting cells in the immune system. SLC11A1 has pleiotropic effects on macrophage function, all of which are important in resistance to intracellular pathogens. These include release of nitric oxide, L-arginine flux, oxidative burst, tumouricidal and antimicrobial activities, as well as upregulation of CXC chemokine KC, tumour necrosis factor-α, interleukin-1β, inducible nitric oxide synthase and MHC class II expression [12, 13]. Roles of monocytes and macrophages have been implicated in the pathogenesis of T1D [14, 15]. Recently, macrophages have been shown to be one of the major immune cell populations in infiltrated pancreatic islets of autopsy tissues from patients with T1D [15, 16], suggesting that macrophages may contribute to the early phase of beta-cell destruction. Together with the evidence showing that over-activation of SLC11A1 could potentially induce and maintain autoimmune diseases, these data make SLC11A1 a candidate gene for autoimmune and immune-mediated disorders, such as T1D. Interestingly, SLC11A1 has been shown to suppress IL-10 production [17], and the gene that encodes IL-10 has recently been associated with T1D susceptibility [1, 18], as well as with risk of ulcerative colitis [19] and of systemic lupus erythematosus [20].

Genetic variants in the SLC11A1 gene region have been reported to be associated with various infectious [21–29] and chronic immune diseases, such as T1D [30–33], rheumatoid arthritis (RA) [34–37], juvenile RA (also known as juvenile idiopathic arthritis; JIA) [38, 39], sarcoidosis [40], inflammatory bowel disease (IBD) [41–45], Kawasaki disease [46] and multiple sclerosis [47]. More recently, SLC11A1 has also been claimed to be associated with Behcet's syndrome in a Turkish population [48] and esophageal cancer in a South African population [49]. Blackwell et al. identified a potentially functional polymorphic microsatellite with Z-DNA forming dinucleotide repeats in the promoter region of the human SLC11A1 gene [50, 51]. Allele 3 with the apparent stronger promoter activity that drives higher expression of SLC11A1 relative to allele 2 may result in chronic macrophage hyperactivation, thus predisposing to autoimmune diseases, but protecting against infectious diseases (see additional file 1 for microsatellite allele sequences) [51]. However, not all studies have replicated the association of the promoter polymorphism in T1D [31, 32]. Indeed, associations at other polymorphisms, such as rs17235409 (D543N), rs17235416 (3'UTR) and rs3731865 (INT4), have been reported in immune-related and infectious diseases [22, 23, 25–28, 35–37, 39, 43], implying that alternative variants may be causal. In IBD, Zaahl et al. have shown that the SLC11A1 association involves a protective effect of the promoter SNP, rs7573065 (-237 C→T) [44]. They previously found that in the presence of allele 3 of the 5' microsatellite, the change from allele C to T at rs7573065 (-237 C→T) lowered the expression of SLC11A1, to a similar level as observed with allele 2 of the microsatellite [52]. Combined, these data suggest that both rs7573065 (-237 C→T) and the microsatellite (r² = 0.02 and D' = 0.98 in controls; see additional file 2) together might be associated with disease.

Previously, Maier et al. found no evidence of association of SLC11A1 with T1D using four tag SNPs (rs2276631, rs2279015, rs1059823 and rs1809231), but only 1,709 T1D cases, 1,829 controls and 1,632 families were studied [4]. They also genotyped the non-synonymous SNP (nsSNP) rs17235409 (D543N) and the microsatellite (GT)n in 1,632 families and did not obtain any evidence of association [4]. Nor did they obtain evidence of association with the nsSNP rs17235409 (D543N), which was genotyped in an additional 1,995 cases and 2,101 controls [4]. The genome-wide association study (GWAS) performed by the Wellcome Trust Case Control Consortium (WTCCC) identified a SNP, rs4674297, in MGC50811 (also known as C2orf62), located within the same LD block as SLC11A1 that showed some evidence of association with T1D (P = 0.0070) [53]. Barrett and colleagues performed a meta-analysis of three GWAS totalling 7,514 cases and 9,045 controls [1]. They found rs4674297 was one of the four most T1D-associated SNPs in the region with P-values of around 10^-4 [1]. Following the additional functional support obtained using the NOD mouse model [7, 8], we have performed a comprehensive association analysis of sequence polymorphisms in the SLC11A1 region in a maximum of 8,863 unrelated T1D cases and 10,841 controls as well as up to 5,696 T1D families in order to identify the most associated, potentially causal, T1D variant(s) and its effect on expression and splicing of the SLC11A1 gene.

We genotyped and analysed three of the Maier et al. tag SNPs (rs2276631, rs2279015 and rs1809231 [4]), the nsSNP rs17235409 (D543N), the indel rs17235416 (3'UTR), rs3731865 (INT4), rs7573065 (-237 C→T) and rs4674297 in up to 5,878 cases and 6,406 controls (Figure 1 and Table 1). As the promoter microsatellite (GT)n has been suggested to be a functional variant, we genotyped this polymorphism in the maximum number of samples in the collection available at the time, which was 7,894 cases and 7,560 controls (Figure 1 and Table 1). Since rs3731865 (INT4), the indel rs17235416 (3'UTR) and rs4674297 showed the most evidence of association with T1D (P ≤ 0.005; Figure 1), they were genotyped in our extended case-control collection in an additional 2,985 cases and 4,435 controls to test whether their effects were independent.

We had over 96% power to detect an effect size of 0.90, at an alpha level of 0.005, assuming a multiplicative allelic effects model and a minor allele frequency (MAF) of 0.29, with a sample size of 8,863 cases and 10,841 controls. rs3731865 (INT4) showed the most evidence of association with T1D (P = 1.55 × 10^-6; OR = 0.90 (95% confidence interval (C.I.) 0.86-0.94); Table 1). The support for association with T1D at rs4674297, a SNP identified from the Barrett et al. meta-analysis study (P = 2.9 × 10^-4) [1], was maintained (P = 1.57 × 10^-4; OR = 0.91 (95% C.I. 0.86-0.96); Table 1; 5,897 cases and 5,461 controls in our full dataset of 8,863 cases and 10,841 controls overlapped with 7,514 cases and 9,045 controls in the Barrett et al. meta-analysis study). Our samples were not genotyped at the other three SNPs, rs12471773, rs2290708 and rs3816560, found in the meta-analysis as they are in high LD with the associated SNPs (rs3731865 and rs4674297; r² > 0.8 in controls) and so are unlikely to significantly improve the T1D association signal in this region. We found no evidence of associations that were independent of rs3731865 (INT4) (P > 0.007).

The 2004 IBD study by Zaahl et al. suggested that the promoter SNP, rs7573065 (-237 C→T), and the microsatellite (GT)n might, in combination, be associated with disease [52]. However, these variants are not associated with T1D, either together in a joint effects model (P = 0.33) or individually (P > 0.01; Table 1). We also assessed whether rs7573065 (-237 C→T) and the microsatellite were involved with T1D susceptibility in a haplotypic manner. Haplotype 3.T (microsatellite.rs7573065) was grouped with all the other haplotypes consisting of the protective allele 2 of the microsatellite, since they were found to exert similar effects on transcriptional activity [52]. Using the most 'susceptible haplotype', 3.C, as the reference, the grouped haplotypes did not show an association with T1D (P = 0.15; OR = 0.96 (95% C.I. 0.90-1.02)). Furthermore, haplotype 3.T by itself did not confer protection against T1D (P = 0.32; OR = 0.94 (95% C.I. 0.83-1.07)). Together these analyses suggest that rs7573065 (-237 C→T) does not confer protection against T1D, either by itself or in combination with allele 3 of the microsatellite.

SLC11A1 has several known alternative splice transcripts of which the two major isoforms expressed are full length transcripts with or without exon 4a (74 bp in length). Previous studies have shown that exon 4a, which is located between exon 4 and 5, encoded by an Alu element, is transcribed in vivo but would introduce two termination codons in exon 5 resulting in severely truncated, and thus non-functional, SLC11A1 protein [9]. Interestingly, the most T1D-associated SLC11A1 polymorphism, rs3731865, is located in intron 4, 13 bp 3' of exon 4 and 167 bp 5' of the alternatively spliced exon 4a. Although bioinformatics analyses failed to predict known splice elements around or at rs3731865, due to its location within the gene, we hypothesised that rs3731865 might affect elements for transcription or splicing of SLC11A1 exons. This splicing could cause changes in functional message levels and the amount of functional SLC11A1 protein expressed. Therefore, measuring the ratio of transcripts with or without exon 4a in the cell population of interest and correlating the expression of SLC11A1 with genotypes of rs3731865 could be informative for determining the effect of this SNP on gene expression.

SLC11A1 mRNA is expressed specifically in whole blood, CD14⁺ monocytes, CD33⁺ and CD66b (granulocytes) myeloid cells according to the microarray expression data available from BioGPS and HaemAtlas [10, 11]. Since whole blood samples were available in the laboratory, we investigated if the genotypes of rs3731865 correlated with the overall expression of SLC11A1 and also with the isoform ratio in whole blood samples collected in PAXgene RNA tubes from 42 healthy donors recruited from Cambridge BioResource (CBR) [54]. It was hypothesised that in donors homozygous for the T1D susceptibility allele, there would be increased levels of transcripts without exon 4a, which codes for functional protein, and less of the alternative transcript, which encodes premature stop codons and, thus, a non-functional protein. Thereby an overall higher splice isoform ratio is expected in individuals homozygous for the susceptibility allele compared to individuals homozygous for the protective allele. However, no genotype effect was detected for expression of SLC11A1 overall or for isoform ratio differences (P = 0.92 and 0.70, respectively; Figure 2).

The failure to detect SLC11A1 expression differences based on rs3731865 genotypes could be due to various reasons, such as there is actually no correlation between its transcriptional splicing and genotypes of rs3731865 in whole blood, or there is, but the quantitative PCR (qPCR) assays are not sensitive enough to detect it, or perhaps the current sample size may not be large enough to have sufficient power to detect a small difference. It is also possible that a purified cell subset should be studied instead of analysing overall expression in whole blood, such as monocyte-derived macrophages, where the gene is known to function.

The incidence of T1D is predicted to double in children under 5 years by 2020 [55], with the likely cause being the changing environment. These environmental changes could impact our association study, with gene-environment interactions becoming increasingly important. Our control cohort is of similar age to the parents of our cases. This single generational difference could still reduce our power to find effects in our case-control study in the presence of gene-environment interactions, owing to their differences in early life environmental exposures. However, this does not affect the significance of the findings obtained. Taking the data collectively, we identified the strongest allelic association with T1D at rs3731865 in intron 4 of SLC11A1. This SNP is also most associated with JIA in Finnish families in the same direction where the minor allele is also protective for disease [39]. Although the promoter microsatellite remains the only variant reported to date for which there is evidence of a correlation with function in the SLC11A1 gene region [43, 51], alleles 2 and 3 of the microsatellite are in strong LD with rs3731865 (r² = 0.79 in controls; see additional file 2). Therefore, any T1D association detected for the microsatellite might be tagging the association of rs3731865. The extensive prior evidence linking SLC11A1 function and expression with autoimmune, immune-related and infectious diseases in mice and humans, and the confirmed polygenic model of T1D inheritance of numerous small effects, suggest that SLC11A1 variation may well have a very small effect on human T1D risk [56, 57].

A search for additional rare variants via next generation sequencing, as well as results coming from the 1000 Genomes Project [58], could provide further information regarding the possibility of an association of SLC11A1 with T1D. Recently, Zeller et al. and Heinig et al. performed a large-scale investigation of the transcriptome of circulating monocytes and monocyte-differentiated macrophages using the Illumina Human HT-12 expression BeadChips [14, 59]. The two microarray probes (ILMN_1741165 and ILMN_1735737) in SLC11A1 mapped to the 3'UTR of the gene. Using the data from Heinig et al. a stronger expression signal is detected by the ILMN_1741165 probe compared to the other probe [14]. Although expression quantitative trait loci (eQTL) and eSNPs for SLC11A1 were identified from these studies for the ILMN_1741165 probe, including the T1D-associated SNP, rs4674297 (P = 1.03 × 10^-65) [59], this probe maps to repeat sequences [14, 59]. Therefore the eQTL identified for SLC11A1 is highly questionable and could be an artefact. Only about 39% of eQTLs are likely to be due to "simple effect" SNPs, those that altered transcription levels [60]. The majority of eQTL variations are, however, owed to other alterations in gene expression, namely altered initiation/termination and/or splicing [60]. Hence, analysis of microarrays with multiple probes per gene and per exon e.g. Affymetrix Gene Array, and, in the future, direct RNA sequencing using the next generation technologies should be more informative. Investigating the correlation of SLC11A1 RNA expression and protein level in genotyped human macrophage/stimulated monocyte or DC samples based on the genotype/haplotype of the most T1D-associated variant(s) would help to prove causality of the genes in human T1D and help to identify biological disease mechanisms involved in pathogenesis of T1D.

We conclude that genetic variation of the SLC11A1 gene could possibly be associated with T1D susceptibility. We did not observe a difference in SLC11A1 expression at the RNA level based on the genotypes of the SNP most associated with T1D, rs3731865 (INT4), in whole blood samples. However, a potential correlation cannot be ruled out in purified blood cell subsets.