Inherited Variation in Vitamin D Genes and Type 1 Diabetes Predisposition

The etiology and pathophysiology of type 1 diabetes remain largely elusive with no established concepts for a causal therapy. Efforts to clarify genetic susceptibility and screening for environmental factors have identified the vitamin D system as a contributory pathway that is potentially correctable. This review aims at compiling all genetic studies addressing the vitamin D system in type 1 diabetes. Herein, association studies with case control cohorts are presented as well as family investigations with transmission tests, meta-analyses and intervention trials. Additionally, rare examples of inborn errors of vitamin D metabolism manifesting with type 1 diabetes and their immune status are discussed. We find a majority of association studies confirming a predisposing role for vitamin D receptor (VDR) polymorphisms and those of the vitamin D metabolism, particularly the CYP27B1 gene encoding the main enzyme for vitamin D activation. Associations, however, are tenuous in relation to the ethnic background of the studied populations. Intervention trials identify the specific requirements of adequate vitamin D doses to achieve vitamin D sufficiency. Preliminary evidence suggests that doses may need to be individualized in order to achieve target effects due to pharmacogenomic variation.


Introduction
The growing incidence of type 1 diabetes (T1D) is understood to result from an interplay of several factors including environment, nutrition and genetics. One of the environmental and nutritional factors may be a vitamin D deficiency that is highly prevalent and increases the risk for T1D as well as other autoimmune disorders [1,2]. In vitro studies could show a protective effect of active vitamin D for cytokine treated human pancreatic islets [3]. This field of research is of continuing interest with a steady increase in publications over recent years. Due to the lack of a causal therapy in T1D, vitamin D research intended to pave the way for novel immune modulatory concepts both for prevention as well as therapy.
Vitamin D is structurally related to the steroid hormones and its mechanism of action also involves a nuclear receptor similar to thyroid, gonadal and adrenal hormones. The physiological effects extend from the classical bone and calcium/phosphate regulation to muscle, vasculature, immunity, skin, gut and brain which explains that the vitamin D receptor (VDR) is expressed on a vast number of cells which respond to vitamin D. The immune effects of vitamin D on dendritic cells, macrophages and T lymphocytes have attracted major attention since they hold promise for novel therapies [4,5].

Vitamin D Pathways
There are two major forms of vitamin D: vitamin D 2 and vitamin D 3 . While D 2 (ergocalciferol) is of exogenous origin via food intake, vitamin D 3 (cholecalciferol) comes primarily from skin production  The coexpression of the vitamin D system genes (e.g., VDR and CYP27B1) in multiple cell types including lymphocytes, antigen-presenting cells and pancreatic islet cells [11][12][13] highlights the importance of the vitamin D pathway in T1D.
The 25(OH)D 3 concentration as a marker of the vitamin D status is influenced by environmental and genetic factors. Both sunlight exposure and variants in vitamin D pathway genes affect circulating 25(OH)D 3 levels. Low 25(OH)D 3 levels as well as specific vitamin D system gene polymorphisms enhance T1D susceptibility. Since vitamin D biosynthesis is regulated by genes, their polymorphisms (e.g., VDR, CYP2R1, CYP27B1, CYP24A1, DBP and cubilin) may alter the bioavailability as well as target effects of vitamin D metabolites.

Vitamin D Receptor Gene
The VDR belongs to the nuclear receptor family of transcription factors composed of three domains: a modulating N-terminal dual zinc finger DNA-binding domain, a C-terminal ligand-binding domain and an unstructured region that links the two functional domains [14]. The human VDR gene spans over 100 kilobases (kb) of genomic DNA, located at chromosome 12q13.11, contains eight protein-coding exons 2-9, six untranslated exons 1a-1f, introns and 3 UTR 1 exons [15]. Exon 1 encodes the 5 untranslated region; exons 2 and 3 encode the DNA-binding domain, important for the interaction with the VDRE in target genes. The exons 5-9 encode the ligand-binding region responsible for 1,25(OH) 2 D 3 binding.
Several single nucleotide polymorphisms (=SNPs; more than 5000) have been described for the human VDR gene (https://www.ncbi.nlm.nih.gov/snp/) and four of them have been intensively studied in relation to T1D susceptibility (Table 1).
The rs10735810 (FokI) SNP consists of a T to C substitution eliminating the first start codon (ATG) by generation of an alternative start site (ACG) leading to a differently sized protein. The shorter form (424 aa; C allele or F allele, methionine at the fourth position) is considered to be more active than the long form (427 aa; T allele or f allele, methionine at first position) [17,18]. The SNPs rs1544410 (BsmI), rs7975232 (ApaI) and rs731236 (TaqI) are located at the 3 untranslated region (UTR) of the gene and are without consequences for the VDR protein structure, however, they are strongly linked to a poly(A) microsatellite repeat in the 3 UTR. The poly(A) sequence in the 3 UTR region of genes regulates gene expression, especially through the modulation of mRNA stability.
In contrast, Pani et al. (2000) [21] demonstrated other VDR haplotypes "At", "Bt" and "BAt" to confer T1D susceptibility in a German family study. Likewise, the same allele constellation "BBAAtt" was found in a cohort from South Croatia as reported by Skrabic et al. 2003 [29]. Furthermore, the association of T1D with the B allele-as risk enhancing-was confirmed in two case-control Chinese studies (Taiwanese and Han population of Beijing) [22,40], two Japanese studies [30,44] and one genetically heterogeneous Brazilian study [48]. Nevertheless, also the "AA" genotype in Taiwanese and a "t" allele in Iranian populations have been suggested as risk conferring [22,60]. Accordingly, Ramos-Lopez et al. (2006) demonstrated a higher frequency of alleles "A" and "t" within the haplotype composed of the SNPs rs9729, rs731236 (TaqI), rs7975232 (ApaI) and rs757343 (Tru9I) in another family study [38].
Recently, Miettinen et al. (2015) analyzed the genotype distributions of 13 VDR SNPs in a Finnish population consisting of families whose offspring had T1D (cases) and families with healthy offspring (controls) [71] where all VDR SNPs were associated with the 25(OH)D 3 levels. Two VDR SNPs (rs1544410 BsmI and rs731236 TaqI) differed in the genotype distributions: "Bb" and "Tt" genotypes were more prevalent (corresponding to "B" and "t" allele) than in the control mothers. The investigators suggest that maternal VDR SNPs enhance a child's risk for T1D independent of the child's genotype. The maternal vitamin D status and VDR genotype may hereby regulate in utero development and have an influence on the later T1D risk in conjunction with environmental factors.
In addition, in Chilean T1D patients where the population is characterized by a heterogeneous admixture of people from European and South American Indian descent, a further haplotype "BAT" [43] conferred susceptibility.
Fassbender et al. confirmed the study from Mcdermott et al. (1997) with "T" as risk allele for the development of T1D [19] in a small German cohort [27], a finding later corroborated by Garcia et al. (2007) [43].
Notably, the different haplotypes associated with T1D as reported by Mcdermott et al. (1997) [19] (bAT), Pani et al. (2000) [21] (BAt) and Garcia et al. (2007) [43] (BAT) indicate a variable genetic predisposition to T1D depending on the ethnic origin. This was also shown by Audi et al. (2004) [34] by analyzing the SNP in the start codon of exon 2 (rs10735810 FokI) additional to the rs1544410 (BsmI) SNP in two Spanish populations with different genetic backgrounds (Barcelona and Navarra). A combined genotype showed that the homozygous "bbFF" genotype was more prevalent in T1D patients from Barcelona whereas the homozygous "BBFF" genotype was more frequent in Navarra. Another study conducted by San Pedro et al. (2005) included families of Basque origin where a risk-associated four-locus haplotype (fBAt) was identified [36] confirming the same profile ("BAt") as described by Pani et al. (2000) [21]. Furthermore, haplotype analysis performed in North India showed that the haplotypes "FBAt" and "fBAT" were significantly more frequent in T1D patients [52]. Moreover, those haplotypes differed in comparison to those from South India (bAT) but were found in concordance with the "BAT" haplotype present in Chile [19,43]. In an Iranian population, the haplotypes "tAbf", "tabF" and "tAbF" conferred an increased risk for T1D [60].
The genotype and allele distribution of rs10735810 (FokI) VDR SNP differs between patients and controls in many studies, however, the risk allele (F or f) also does so among the studies. The "FF" genotype and/or "F" allele predispose to T1D in Japanese, Rumanian, Uruguayan, Turkish and Iranian populations [23,24,42,55,59,61]. The "F" allele and the combination of vitamin D gene "BBFFAATt" are even considered to enhance the risk for diabetic complications, particularly diabetic nephropathy (DN) [62]. In contrast, studies from Egypt, Italy and Croatia [35,41,68] observed an association with the "ff" genotype and T1D risk. Mory et al. (2016) [73] found homozygous "ff" to be more frequent in T1D subjects with thyroid antibodies (Abs) and thyroid dysfunction in Brazil. T1D patients carrying the rs10735810 (FokI) SNP with thyroid peroxidase Abs showed an 18-fold risk to develop thyroid dysfunction.
Additionally, the rarely analyzed SNP rs757343 (Tru9I) showed an overtransmission of the allele G (corresponding to "U") from parents to affected children as shown by Ramos- Lopez et al. 2006 andBoraska et al. 2008 [38,46]. Furthermore, in the study from Gyorffy et al. (2002) the haplotype "bau" was found more frequently in patients than in controls [26].
Interestingly, one of the lowest T1D incidence rates in Europe was described for the Greek island Crete: here, two haplotypes of the four VDR SNPs confer the highest risk (aBFT and aBFt) for T1D [50]. This underscores that an interplay of genetic and environmental factors modulates T1D susceptibility.

Vitamin D Receptor and Meta-Analysis, Diabetes Complications and Monogenetic Vitamin D Disorders
On the basis of the diverging results of VDR SNPs and T1D susceptibility, nine meta-analyses have been performed [75][76][77][78][79][80][81][82][83]. Eight out of nine meta-analyses on the VDR gene and T1D published between 2006 and 2017 agree on the conclusion that rs10735810 (FokI) and/or rs1544410 (BsmI) SNPs play an important role in the development of T1D [76][77][78][79][80][81][82][83] (Table 3).  Table because only an abstract was available. "B" allele of the rs154410 (=BsmI) SNP was associated with an increased risk for the development of T1D especially in Asians. rs10735810 (=FokI) and rs731236 (=TaqI). Ponsonby et al. (2008) [76] suggest that the association between the VDR SNPs and T1D should be seen as depending on the environment and not being responsible for T1D by itself. Therefore, the authors conducted a meta-analysis of 16 case-controlled studies from 19 regions and four additionally analyzed SNPs (rs7975232 ApaI, rs731236 TaqI, rs1544410 BsmI, and rs10735810 FokI) under the aspect of ambient winter UV radiation. The study centres were located across a range of latitudes from 33 • S to 65 • N corresponding to a winter UV radiation range from 1.0 mW/m 2 to 133.8 mW/m 2 . The authors observed that the allele "B" of rs1544410 (BsmI) and the allele "F" of rs10735810 (FokI) were more likely risk factors for T1D under high-winter-UV radiation exposures. Four years later, Zhang et al. (2012) [77] published a meta-analysis based upon 23 case-control studies covering Asians, European and Latino populations and evaluating the ethnic-specific effects for an association with T1D. The main inclusion criteria for this meta-analysis were publications in English or Chinese; available data for genotype distributions in cases and controls; the genotype distribution of the tested controls was in Hardy-Weinberg equilibrium (HWE). Hereby, the "BB" genotype of the rs1544410 (BsmI) SNP was associated with an increased risk for the development of T1D, especially in Asians. This finding was also confirmed in the meta-analysis from Wang et al. 2012 [78], where 3854 cases and 6498 controls were included and an increased T1D risk for the "B" allele a particularly in East Asian population was found. In another meta-analysis, Wang et al. (2014) [79] selected 13 case-control studies (1973 T1D and 1986 controls) from the Asian region and evaluated two VDR SNPs (rs1544410 BsmI and rs10735810 FokI). Interestingly, the regional stratification analysis indicates that the rs1544410 (BsmI) "B" allele conferred an enhanced T1D risk in East Asia but the rs10735810 (FokI) allele "F" in the West Asian population. An additional meta-analysis covering the four VDR SNPs in Asian, European and Latino populations concluded that "BAT" was a significant T1D risk factor [81]. Furthermore, a recent meta-analysis on the basis of nine studies comprising 1053 children with T1D (Asian, European and Latino origin) confirmed the "BB"genotype of rs1544410 (BsmI) as risk marker for T1D and also for the "tt" genotype of the rs731236 (TaqI) SNP [83]. Liu et al. (2014) focused on the diabetes complications (diabetic nephropathy (DN) and diabetic retinopathy (DR)) and studied four variants of the VDR [82]. Hereby, the rs10735810 (FokI) SNP was associated with nephropathy risk in Caucasian diabetes patients, represented in a dominant model.
Apart from association studies, there are also informative case reports on genetic vitamin D disorders in T1D. One case report describes the development of T1D in a child with pre-existing hereditary vitamin D-resistent rickets (VDRR) due to a compound heterozygous mutation of the VDR (L263R and R391S) that led to dissociated responses of the CYP24A1 and RELB promoters to 1,25-Dihydroxyvitamin D 3 action [84]. Another case of VDRR was reported from India, where a 10-year-old girl had developed T1D [85]. An additional case report with an inborn error of vitamin D metabolism was published recently. A new missense mutation of the PHEX gene has been described in a T1D patient from a Han Chinese pedigree over four generations that caused X-linked hypophosphatemic rickets manifesting with renal phosphate wasting, a bone mineralisation and vitamin D metabolism defect [86].
These experiments of nature underline that a vitamin D defect syndrome may have the potential for additional disease such as β-cell autoimmunity resulting in T1D. A systematic investigation of the acquired and the innate immune system in fifteen patients with VDRR showed some impairments of the innate immunity, particularly lower cathelicidin production and a proinflammatory cytokine profile of lymphocytes [87]. This illustrates the enormous plasticity of the immune system adapting to a genetic defect and that only few patients with hereditary vitamin D defect syndromes will develop an autoimmune disease such as T1D.
The CYP2R1 gene is located on chromosome 11p15.2 with a length of 15 kb and contains five exons separated by four introns. Two SNPs (rs10741657 and rs12794714) were investigated in T1D [102]. The SNP rs10741657 (G/C) maps to a 2 kb mRNA transcript and rs12794714 (C → T, Ser → Ser) is a synonymous SNP in exon 2. Our investigations revealed that the allele G of the CYP2R1 rs10741657 SNP is more often transmitted to T1D affected offspring. Additionally, the case-control analysis shows a higher frequency of the GG genotype in T1D patients. The latter correlated with a lower 25(OH)D 3 concentration [Ramos- ] [88]. The subsequent analysis of a large number of samples from case/control (n = 8517/10,438) and T1D families (n = 1933) in the British population revealed an association between the two SNPs (rs10741657 and rs12794714) with T1D in a combined dataset [57].
The next enzyme in the vitamin D cascade, CYP27B1, is coded by a gene situated on chromosome 12p13.1-q13.3. The gene contains nine exons and eight introns and extends over 4.8 kb. Mutations in the CYP27B1 gene can lead to an inactive protein unable to bind 25(OH)D 3 as found in vitamin D dependent rickets [103]. Two SNPs within the CYP27B1 were investigated in relation to T1D, rs10877012 SNP (−1260 C/A) located in the promoter region and the rs4646536 SNP (+2838 C/T) in intron 6.
We originally observed that allele "C" and genotype "CC" were more frequent in patients with T1D than in controls [90]. Later studies, one from Egypt with 240 subjects and another one with a large collective (population different countries: British, Finland, USA, Norway, Romania) confirmed these findings [91,93]. Additionally, Bailey et al. (2007) showed also an association of T1D with the rs4646536 SNP. Cooper et al. (2011) confirmed the protective effect of the allele "A" of the rs10877012 SNP [57,91].
The last enzyme in the vitamin D cascade, CYP24A1 is capable of hydroxylating both metabolites (25(OH)D 3 , and 1,25(OH) 2 D 3 ). However, the preferred substrate is 1,25(OH) 2 D 3 . CYP24A1 catabolizes 1,25(OH) 2 D 3 in a complex of steps resulting in the production of water-soluble calcitroic acid [104]. Major alterations in the enzymatic activity of CYP24A1 can be due to mutations of the CYP24A1 gene located on chromosome 20p13 (20.5 kb, 12 exons) that cause idiopathic infantile hypercalcemia [105]. The CYP24A1 gene was investigated in relation to T1D susceptibility: sixteen tag SNPs for CYP24A1 that were analyzed by Bailey et al. (2007) [91] as well as two further SNPs (rs6013897 and rs2296241) did not show any association with T1D [57,89]. This gene is of potential clinical relevance because an undiagnosed CYP24A1 mutation may cause hypercalcemia also in adults if these are exposed to high vitamin D dosages.
A further essential component of the vitamin D system is the DBP, also called group-specific component (GC) that belongs to the proteins of the albumin family and transports vitamin D in the circulation. DBP is a single chain glycoprotein with a molecular weight of 52 kDa, predominantly synthesized in the liver. The DBP gene maps to chromosome 4q11-q13 and contains 13 exons and 12 introns and extends over 42.5 kb. Among the many characterized DBP variants, two known SNPs in exon 11 were investigated for T1D susceptibility (rs7041 and rs4588). While the rs7041 SNP results in a T to G substitution (Aps to Glu in codon 416), rs4588 SNP leads to a C to G substitution (threonine to lysine in codon 420). The majority of the studies including the SNPs rs4588 and rs7041 and rs2282679 SNP failed to prove an association with T1D [57,67,89,94,95,98]. Two studies, however, originating from the same laboratory showed an association with rs7041 SNP and T1D [96,97]. Nevertheless, it has to be pointed out that several DBP/GC combinations of SNPs are conserved in the population forming a diverse profile of haplotypes. Such DBP haplotypes give rise to low or high affinity DBP/GC protein structures with a different binding of the free vitamin D metabolite and also affecting the monocyte production of cathelicidin [106].
Another molecule with a crucial role in vitamin D trafficking is cubilin. This 460-kDa long protein is mainly localized in the proximal renal tubule, but has been identified in other tissues including placenta, intestinal epithelium among others.
A crucial role of cubilin is the formation of an endocytic receptor complex with megalin. That complex is capable of binding DBP/25(OH)D 3 with high affinity, mediating its uptake into the cells. The loss of functional cubilin in patients leads to loss of the 25(OH)D 3 in urine and subsequent decrease in vitamin D metabolites plasma levels. Hence, our group examined SNPs within the cubilin gene as potential risk markers for T1D [99]. The cubilin gene is located on chromosome 10p12.33-p13. We analysed five cubilin SNPs (rs3740168, rs3740165, rs1801233, rs1801229 and rs2796835) in a case-control study (200 T1D and 200 controls). Out of these, the rs3740165 SNP was found to be associated with increased T1D risk. The genotype "AA" of the rs3740165 was more prevalent in T1D patients than in control but without correlation neither with 25(OH)D 3 , nor with 1,25(OH) 2 D 3 concentration. It has to be pointed out that the SNP does not change the coding sequence in this position (Pro → Pro). Therefore, functional susceptibility may develop by a linked gene variant or a regulatory SNP.

Major Susceptibility to Type 1 Diabetes by HLA and Other Immune Genes: Vitamin D
The strongest susceptibility to type 1 diabetes is conferred by high risk HLA DR-DQ alleles present as hetero-or homozygous combinations most patients. There are up to 40 additional risk loci identified and some of them have been shown to affect lymphoid enhancer sequence in T and B cells, thymus and CD34 + stem cells [107]. Vitamin D regulates several immune genes as identified through genome wide studies by VDR chromatin immunoprecipitation followed by mass DNA sequencing (CHIP-seq) [108,109] where VDR binding to autoimmune susceptibility loci was identified amongst them type 1 diabetes sites. Hereby, the VDR-enhanced susceptibility to T1D may form a genetically determined proinflammatory cytokine pattern [110].

Vitamin D Intervention in Type 1 Diabetes and Pharmacogenomics
Vitamin D deficiency is a worldwide problem [111]. It enhances the risk for various conditions including T1D [112] and provides the rationale for many intervention trials (clincaltrials.gov currently-as of 11 April 2017-3122 trials listed). The potential to modify the development of T1D was reported in a case-control study and a birth cohort follow-up study from Finland: it strongly indicated that vitamin D supplementation in infancy decreases the risk of T1D [113,114]. The therapeutic benefit of vitamin D onT1D was tested in some clinical trials [2] but only few studies examined the effect of the vitamin D SNPs in the context of vitamin D supplementation for T1D [115][116][117]. We recently performed a randomized, double-blind, placebo-controlled trial with cross-over design in which thirty-nine patients with T1D received 4000 IU/day cholecalciferol for three months followed by placebo or in reverse sequence. Hereby, besides an improvement of the vitamin D status (median 25(OH)D 3 increased to 38 ng/mL), the regulatory T cells (Treg) demonstrated a differential response to vitamin D to three months' treatment according to VDR SNPs. Furthermore, this trial also showed an improvement of glycemic parameters under vitamin D treatment. Patients carrying the genotypes aa, TT and bb (rs7975232 ApaI, rs731236 TaqI and rs1544410 BsmI) were capable of raising their Treg cell number [115]. A further study tested in vitro the functional role of the VDR rs10735810 (FokI) SNP in T-helper (CD3 + CD4 + ) from twenty patients with T1D. The stimulation of CD3 + CD4 + cells with 25(OH)D 3 [62.4 nM] and 1,25(OH) 2 D 3 [1 × 10 −8 M] for 72 h revealed a reduced percentage of CD4 + cells isolated from T1D patient carrying "FF", suggesting a beneficial balance in the T cell compartment [116].
In a prior in vitro study, Mauf et al. (2015) [117] had explored the immunomodulatory effects of vitamin D supplementation on 25(OH)D 3 levels, on dendritic cells in twelve patients with T1D and the role of the vitamin D SNPs. Remarkably, the 25(OH)D 3 treatment (50 ng/mL) for seven days inhibited the differentiation of monocytes into dendritic cells, promoting the formation of intermediate cells (IC). The increase of IC under supplementation with 25(OH)D 3 was related to the genotypes of two VDR SNPs (rs731236 TaqI and rs1544410 BsmI) and one SNP of the CYP24A1 gene (rs927650), illustrating that the immune effects of vitamin D supplementation can depend on genetic variants of the vitamin D system.

Conclusions
Vitamin D deficiency is a risk factor for T1D and genes of the vitamin D system show robust associations with T1D. The vitamin D system appears to affect the immune regulatory pathways, leading to the final β-cell destruction. Several experimental lines of evidence suggest islet protection by vitamin D. Exploiting this potential will be a challenge for future studies, including larger controlled trials with different doses of vitamin D and functional studies to elucidate mechanistic actions in the immune system and also for metabolic end points.