Exonic variants of genes related to the vitamin D signaling pathway in the families of familial multiple sclerosis using whole‐exome next generation sequencing

Abstract Introduction Vitamin D (VD) deficiency has been associated with multiple sclerosis (MS) and other autoimmune diseases (AIDs). However, the effect of the genetics of VD on the risk of MS is subject to debate. This study focuses on genes linked to the VD signaling pathway in families with MS. The evaluation of gene variants in all the members of families could contribute to an additional knowledge on the information obtained from case‐control studies that use nonrelated healthy people. Material and Methods We studied 94 individuals from 15 families including at least two patients with MS. We performed whole‐exome next generation sequencing on all individuals and analyzed variants of the DHCR7, CYP2R1, CYP3A4, CYP27A1, GC, CYP27B1, LRP2, CUBN, DAB2, FCGR, RXR, VDR, CYP24A1, and PDIA3 genes. We also studied PTH, FGF23, METTL1, METTL21B, and the role of the linkage disequilibrium block on the long arm of chromosome 12, through analysis of the CDK4, TSFM, AGAP2, and AVIL genes. We compared patients with MS, other AIDs and unaffected members from different family types. Results The study described the variants in the VD signaling pathway that appear in families with at least two patients with MS. Some infrequent variants were detected in these families, but no significant difference was observed between patients with MS and/or other AIDs and unaffected family members in the frequency of these variants. Variants previously associated with MS in the literature were not observed in these families or were distributed similarly in patients and unaffected family members. Conclusion The study of genes involved in the VD signaling pathway in families that include more than one patient with MS did not identify any variants that could explain the presence of the disease, suggesting that VD metabolism could probably play a role in MS more as an environmental factor rather than as a genetic factor. Our study also supports the analysis of cases and unaffected individuals within families in order to determine the influence of genetic factors.


| INTRODUC TI ON
Multiple sclerosis (MS) is an inflammatory autoimmune disease of the central nervous system (CNS) that causes demyelination and axonal damage. MS etiology involves multiple factors, with environmental factors interacting with genetic predisposition. Although genome-wide association studies (GWAS) have validated the central role of the major histocompatibility complex in the genetics of MS, susceptibility has been found to be influenced by several genes outside the short arm of chromosome 6 ( Baranzini & Oksenberg, 2017;Cree, 2014;Sawcer et al., 2011). The disease is therefore considered to be polygenic. However, the variants identified through GWAS have little effect on the overall risk of MS; as a result, there is a considerable need for further research into genetic factors in MS. New techniques such as whole-exome sequencing (WES) offer a comprehensive view of the coding region of the genome, allowing us to study complex diseases like MS with more detail.
Vitamin D (VD) is a fat-soluble hormone which plays an essential role in calcium homeostasis and in skeletal development and maintenance. It is also thought to play a role in cancer, immune function, and autoimmune diseases (AIDs), including MS (Ascherio et al., 2014;Gianfrancesco et al., 2017;Løken-Amsrud et al., 2012;Mowry et al., 2012;Munger, Levin, Hollis, Howard, & Ascherio, 2006;Salzer et al., 2012;Simpson et al., 2010;Wang, Zeng, Wang, & Guo, 2018). There are several types of VD, including ergocalciferol (D 2 ) and cholecalciferol (D 3 ), formed from their respective previtamins, ergosterol, and 7-dehydrocholesterol (7-DHC). The main natural source of VD 3 in humans is production in the skin, where 7-DHC undergoes a 2-step reaction involving ultraviolet-B irradiation to form pre-VD 3 , followed by thermal isomerization, forming VD 3 . Both VD 2 and VD 3 can also be obtained in small amounts from a varied diet, and in larger amounts from specific foods or nutritional sup- . This step appears to be regulated by parathyroid hormone (PTH) and phosphaturic factor fibroblast growth factor 23 (FGF23) (Christakos, 2017).
It has also been suggested that this may in turn be influenced by intronic Low-frequency genetic variants in VD signaling pathway have been linked to MS; however, their role in disease pathogenesis is controversial. We are yet to determine their influence over familial forms of MS, whether the association is due to variations in a gene, or whether the presence of variants at different points in the pathway may be associated with the disease.
The aim of the present study is to analyze the genetic variations of genes related with the VD signaling pathway in families from familial forms of MS. By studying families, genetic factors may be more easily observable and less obscured by environmental factors than in sporadic cases.

| Study population
We studied 15 families with at least two patients with MS according to the McDonald criteria (Polman et al., 2011), in total 94 individuals. The study protocol included a specifically designed Conclusion: The study of genes involved in the VD signaling pathway in families that include more than one patient with MS did not identify any variants that could explain the presence of the disease, suggesting that VD metabolism could probably play a role in MS more as an environmental factor rather than as a genetic factor. Our study also supports the analysis of cases and unaffected individuals within families in order to determine the influence of genetic factors. Reporter™ (Thermo Fisher Scientific Inc.). Our analysis aimed to identify SNPs and indels located in the exons and splicing junctions F I G U R E 1 Vitamin D pathway. 7-DHC follows a two-stage reaction involving ultraviolet-B irradiation and thermal isomerization. Vitamin D is converted into 25-hydroxyvitamin D3 (25[OH]D 3 ) by a hydroxylation mechanism mediated by vitamin D 25 hydroxylase. Subsequently, 1α-hydroxylase, regulated by PTH and FGF23, transforms 25(OH)D 3 into its active form, 1,25-dihydroxyvitamin D3 (1,25[OH] 2 D 3 ). The proteins 25(OH)D 3 and 1,25 (OH) 2 D 3 mainly circulate bound to carrier proteins (DBP), but a small fraction circulates freely or bound to albumin. The internalization mechanism is produced mainly by DBP-megalin (LRP2). However, in some cells this process can occur by binding to other proteins and even by passive diffusion. Finally, 1,25(OH) 2 D 3 binds to the VDR nuclear receptor, interacting with the RXRA, forming a heterodimer that allows transport to the cell nucleus. Genes involved in the vitamin D pathway are shown in grey. The regulatory pathway of the genes and causing protein-level modifications (excluding synonymous variants), which were detected in over 40% of reads.

| Variant prioritization
Observing quality controls and bioinformatic filtering, we analyzed the coding and splicing regions of genes involved in the VD signaling pathway. We evaluated the list of variants identified against database information on previously described variants (http://www.ncbi.nlm. nih.gov/SNP/, http://www.1000genomes.org, http://gnomad.broadinstitute.org, and http://evs.gs.washington.edu/EVS). Firstly, we excluded variants located in introns, intergenic regions, and untranslated regions. We also removed synonymous variants; according to a polygenic pattern, we considered variants with a minor allele frequency (MAF) below 5% (http://gnomad.broadinstitute.org/). In order to understand the possible biological functions of the variants selected, we

| Sequencing data analysis
Sequencing results were evaluated using three analysis models: 1. Comparison of the variants detected in the overall cohort of family members, analysing the groups of individuals with MS, other AIDs, and, unaffected family members, taking into account the type of familial MS.

Prioritization of variants.
3. Analysis of pedigrees, assessing the role of the variants identified in each pedigree.

| Statistical analysis
Descriptive analysis results are expressed as absolute frequencies and percentages (n [%]), means ± standard deviation (SD), or medians (interquartile range). The Kolmogorov-Smirnov test was used to test quantitative data for normal distribution. The chi-square test was used to compare independent samples with qualitative variables; the Mann-Whitney U test was used for quantitative variables. Intergroup differences were evaluated with the Kruskal-Wallis H test and the Dunn post hoc test. More powerful quasi-likelihood score test (M QLS ) was used, which allows testing for case-control associations in samples with related individuals (Thornton & McPeek, 2007;Thornton, Zhang, Cai, Ober, & McPeek, 2012). Allele frequencies were tested to identify deviations from Hardy-Weinberg equilibrium. Bonferroni method was used to correct for multiple comparisons. Statistical significance was set at p < 0.05.

| Description of the families, population, and daylight exposition
We defined two types of family: type A and type B. The sample in-

| Exonic variants of genes involved in the vitamin D signaling pathway
Analysis of the genes associated with the VD signaling pathway revealed a total of 154 different variants in the 94 individuals tested.
Disregarding synonymous variants gives a final total of 77 different nonsynonymous variants (Table 1). All variants found followed the Hardy-Weinberg equilibrium.

| Variants affecting levels of circulating vitamin D
In the families, we detected single variants of CYP3A4, CYP27A1, CYP24A1, and five variants of the GC gene, with no significant differences between groups. We also analyzed the genes that regulate CYP27B1 expression (METTL21B, METTL1, FGF23, and PTH) ( Table 1).
Seven patients with MS (20%), one with an AID other than MS (7.7%), and eight unaffected individuals (17.4%) were carriers of FGF23 variant rs7955866. Regarding family type, this variant was present in six members of type-A families (13.6%) and 10 members of type-B families (29%). METTL21B variant rs141172155 was found in two patients with MS (5.7%) and in no other group. We observed two additional synonymous variants of the genes METTL21B (rs923829) and PTH (rs6256). Variant rs923829 was more frequent in type-A than in type-B families (16 [36.4%] vs. 8 [16%]). After prioritization and filtering, we detected rs6068812 variant of CYP24A1in one patient with RRMS, one with PPMS, and one unaffected individual from a type-B family. We analyzed the variants from genes potentially involved in the 12q LD block (CDK4, TSFM, AGAP2, AVIL, CYP27B1, METTL1, and METTL21B) and compared these against variant from CYP24A1 (rs6068812). All three individuals with rs6068812 also had a variant in AVIL gene (rs2172521), which was present in all members of our cohort. Therefore, no association was observed between the 12q LD block genes variants analyzed and CYP24A1 gene variant (rs6068812). None of these variants found in the families had a significant relationship to MS or MS with AID compared to unaffected family members.

| Variants affecting vitamin D internalization
We observed 18 nonsynonymous variants of LRP2, none of which was associated with MS. We also detected 29 variants of CUBN. The majority of variants observed after filtering and prioritization af-    Table 1). None of these variants found in the families had a significant relationship to MS or MS with AID compared to unaffected family members.

| VDR and RXRA variants
We observed only one variant of VDR and one variant (rs61751479) of

| PDIA3 variants
Two missense variants have been detected on the gene encoding PDIA3 in our cohort. One of them, rs139812953, is only observed in a MS case, in a case with other AID and in two unaffected cases.
The second one, positioned in chr15:44038766 is only present in two nonaffections. various potentially associated variants have been described. This theory is supported by some case-control studies which have analyzed variants at different points of the pathway; results are conflicting, however, with some studies finding no association (Agnello et al., 2017;Simon, Munger, Yang, & Ascherio, 2010). This hypothesis involves two possible underlying mechanisms: firstly, genomic variations may reduce the effectiveness of VDR function; and secondly, these variations may cause a change in gene expression in either immune or CNS cells (Lu, Taylor, & Körner, 2018). The most complex issue in the study of the VD metabolic pathway is related to the heterogeneity of the pathway depending on the cell or tissue type.

| D ISCUSS I ON
Accounting for the fact that different cell types internalize VD in different ways, our WES study of the different variants in the families of patients with familial MS included variations in all subpathways of the VD signaling pathway. VDR action is also variable after translocation to the nucleus, depending on the target cell (Zella, Kim, Shevde, & Pike, 2006;Zella et al., 2010). It has been suggested that this is due to different levels of local activators and coregulators and epigenetic mechanisms (Saccone, Asani, & Bornman, 2015); this issue lies beyond the scope of the present study. The influence of specific tissue or cell type prevents us from easily drawing conclusions as to the potential role of VD based only on levels of some of its metabolites in the blood.
No significant intergroup differences were observed for presence of these variants in our cohort. The results obtained are shown in Table   S1. We did locate the CYP24A1 variant rs6068812 in one patient with RRMS, one with PPMS, and one unaffected individual from a type-B family; this variant has not been associated with MS in the literature.
Given the very low MAF of the variant, we should consider a potential association with the disease and could be analyzed in further studies.
This stage of the VD metabolic pathway is one of those in which researchers have searched for variants influencing MS, especially in CYP27B1; it is therefore surprising that no association was found.
Associations have been described between VD deficiency and numerous genetic variants (Lafi, Irshaid, El-Khateeb, Ajlouni, & Hyassat, 2015;Li et al., 2014;Lu et al., 2012;Nissen et al., 2014;Signorello et al., 2011;Slater, Rager, Havrda, & Harralson, 2017;Wang et al., 2010;Zhang et al., 2012) (Table S1). However, these associations do not necessarily represent increased risk of MS, as no correlation has been demonstrated between the disease and plasma VD level (Ahn et al., 2010). This is probably due to the existence of confounding factors (Bu et al., 2010), including geographical (Elkum et al., 2014) or ethnic (Batai et al., 2014) variation, as well as environmental factors influencing plasma VD levels, such as sex, age, the season in which samples are taken (Engelman et al., 2013), the use of dietary vitamin D supplements, consumption of milk and cereals, obesity, and daily amount of time spent walking (Hansen et al., 2015). The biomarker used also represents a methodological bias that may affect study results: 25(OH)D and 1,25(OH) 2 D have been associated with different findings (Engelman et al., 2008). Regarding the regulation of CYP27B1 expression (genes METTL21B, METTL1, FGF23, and PTH), which is thought to influence VDR (Bouksila et al., 2018), METTL21B variant rs141172155 was observed only in two patients with MS from type-B families. We did not detect any METTL1 variant associated with MS (Alcina et al., 2013). This pathway is of interest, as CYP27B1 may be influenced by processes occurring in the kidneys in MS and other AIDs (Meyer et al., 2017). GWAS data suggest that there may be a specific locus for MS in a LD block affecting 17 genes on the long arm of chromosome 12. This area as a whole may therefore be considered a risk locus. However, we did not observe related variants in our analysis. Only two exonic and nonsynonymous variants (rs2172521, rs753181730) belonging to the AVIL gene were found in the families. Neither did we observe the proposed relationship with CYP24A1, located on chromosome 20 (Gandhi et al., 2010). CUBN (rs12259370). We also analyzed variants of DAB2, which, together with LRP2 and CUBN, is involved in vitamin D internalization. Variant rs3733801 was the most frequent, although no difference was observed in its frequency between patients with MS and the other individuals studied; it was more frequent in type-B than in type-A families, however. This could support the hypothesis that these families have more variants of genes related to VD internalization, which could represent a genetic predisposition to AIDs and could be analyzed in further studies.

| VDR and RXRA genes variants
Vitamin D receptor variants are one of the genetic biomarkers which have aroused the most interest in evaluating the risk of AID.
They have been associated with increased risk of a range of AIDs, including systemic lupus erythematosus (Carvalho et al., 2015), rheumatoid arthritis (Cavalcanti et al., 2016), autoimmune thyroid disease (Feng, Li, Chen, & Zhang, 2013), and ankylosing spondylitis (Cai et al., 2016). The most studied VDR variants are BsmI, FokI, ApaI, and TaqI, although other variants have been described (Dickinson et al., 2009). The ApaIand BsmI variants are located close to the 3′ end of the VDR gene, in the intron between exons 8 and 9, and cannot be analyzed in WES studies. The FokI (rs2228570) and TaqI (Tizaoui, Kaabachi, Hamzaoui, & Hamzaoui, 2015), although they do suggest that ethnicity, age, and geographic latitude may influence the associations between these variants and MS risk. We found no significant differences in rs2228570 and rs731236 frequency between patients with MS, patients with other AIDs, and unaffected family members. The only Spanish study into VDR variants in sporadic MS did find an association (García Martín et al., 2013). Variant rs731236 was significantly more frequent in members of type-A than type-B families, whereas rs2228570 prevalence was similar in both groups. No associations were detected for variants of the gene encoding RXRA, a transcription factor that together with VDR forms heterodimers needed for nuclear translation, similar to the findings of a recent study .

| PDIA3 gene variants
Protein disulfide isomerase family member 3 was recently described as a VD receptor (Doroudi, Olivares-Navarrete, Boyan, & Schwartz, 2015). Considering that its presence in the brain is greater than VDR, it has been suggested that would have a greater role in the CNS dis- it is present in practically all cell types and acts in nongenomic functions of the VD. There is no previous information on the influence of variants on the PDIA3 gene in MS (Landel, Sthefan, Cui, Eyles, & Feron, 2018). We observed only two variants of PDIA3 gene in our cohort but no associations were observed in MS or AID groups.
However, information on the signaling pathway of PDIA3 is scarce, so further studies will probably be needed.

| Analysis of variants in families
The hypothesis of genetic regulation of vitamin D in families has been proposed in studies of twins, which observed that certain variants, particularly of CYP27B1, may influence the risk of MS (Orton et al., 2008); however, very little information is available on this subject. A CYP27A1 mutation thought to be related to MS has been described in a family including three patients with the disease (Traboulsee et al., 2017).
Another published study design, using WES to study father-motherchild trio, includes 28 patients from eight families; however, the results are not known (García-Rosa et al., 2017). The only published study on the subject, evaluating VDR variants in 29 patients with familial MS and comparing them to unrelated controls, found a significant difference between groups for the TaqI variant of VDR (Yucel et al., 2018).
Our study is the first to analyze whole families and to compare patients to family members, thereby eliminating potential confounders.
Furthermore, families were divided into the two types described above further cases will be necessary to clarify its significance. WES has a number of disadvantages. The first of these is incomplete coverage: it does not encompass the whole genome, and yields a high number of rare variants of uncertain importance; we should also mention the ethical dilemma involved in the unexpected discov-

ery of incidental findings of variants associated with diseases that
were not the target of the study (Klein & Foroud, 2017 A limitation of our study is that the unaffected cases were classified according to the data obtained from direct interrogatory, but brain magnetic resonance was not performed. Thus, we cannot exclude the existence of subclinical MS.
As previously discussed, the great difficulty with WES is addressing and interpreting large numbers of variants. Our study was limited to exonic variants, which we can expect to be functional. Although this technique does allow the detection of intronic variants located close to exonic regions, causality is very difficult to interpret, particularly with small sample sizes. For this reason, we do not address some of the associations described in the literature.

| CON CLUS IONS
Multiple S is considered to be a polygenic disease and it is very doubtful that the risk may depend on one or few variants. It is more likely to depend on alterations on signaling pathways genes such as the ones related to VD pathway. In order to analyze the data of the studies, it seems necessary to determine the prevalence of the variants in these pathways in the families of patients with MS and especially in the different forms of MS families. Our study aimed to know which variants appear in families of MS. Table 1 includes the 77 nonsynonymous exonic variants in genes participating in the VD pathway found in the subjects families that present more than one patient with MS and that could be related to VD functionality. The genes with the greatest presence of variants correspond to those that affect the entry mechanisms in the cells, spatially CUBN and LRP2. However, the presence of all the variants found does not differ between the cases and the not affected members in the family.
When analyzing the pedigrees of each family included in the study, a relationship of these variants with MS has not been observed either.
This is the first study to address the whole family of patients with familial MS using WES and focusing their analysis into the VD metabolic pathway as a whole, analyzing the potential genetic influences over the pathway. Despite the number of studies reporting an association between genetic variants affecting the VD signaling pathway, MS and other AIDs, the information obtained is relatively limited.
After comparison with unaffected family members and analyzing the pedigrees, the majority of the variants published appear not to be specific to MS. This might be due to the fact that sporadic and familial forms involve different genetic factors, or that GWAS led to false positives in comparisons between patients and controls. Our study did not reproduce the association of MS risk with a CYP27B1 variant, or with exonic variants of VDR. We did detect several variants of interest in LRP2 and CUBN, as well as CYP24A1 and METTL21B variants; these should be confirmed by further studies. We have not observed any relationship between variants in the gene that encodes the PDIA3 receptor with the studied groups.
Furthermore, we did not identify any relationship with genetic variants previously associated with MS, including in type-B families, which we had hypothesized would display a greater influence due to genetic load; and there are two potential explanations for this. First, familial MS may not be directly influenced by the genetic alterations that affect VD signaling pathways in sporadic MS, which are detectable by GWAS. The second possible explanation is that vitamin D may be an environmental factor that influences the risk of MS but is not determined by genetic variants. This theory is supported by the variability of the cells and tissues in which the VD metabolic pathway occurs, and the fact that the pathway is influenced by local molecular and epigenetic factors, which makes it difficult for epidemiological and intervention studies to understand (uncover) VD's possible role in the disease. Although this study does not allow us to confirm either hypothesis, it indeed provides information on the variants found in the VD signaling pathway in MS families.

ACK N OWLED G M ENTS
The authors are grateful to our patients and their families for the information and for the DNA samples provided for the study. The authors also thank the Spanish Society of Neurology's Research Operations Office for helping in the English language revision of this paper.

CO N FLI C T S O F I NTE R E S T
The authors have no conflicts of interest to declare.