Statistics from Altmetric.com
Multiple sclerosis (MS) is a complex disorder of the central nervous system (CNS) where gene–environment interactions are considered a key part of disease susceptibility.1 Among the environmental factors thought to be associated with MS, low ultraviolet radiation (UVR) exposure and low vitamin D levels are among the strongest and most consistent.2 ,3 Studies investigating vitamin D for its role in MS clinical course have found levels of the major circulating metabolite of vitamin D, 25-hydroxyvitamin D (25(OH)D), are lower during relapse relative to remission4 ,5 and inversely associated with relapse risk.6–8 The inverse relationship between vitamin D and relapse is likely mediated by the immunomodulatory effects of the active metabolite of vitamin D, 1,25-dihydroxyvitamin D (1,25(OH)2D), acting to up-regulate regulatory T-cell function and depress inflammatory immune activity.9
Genome-wide association studies (GWAS) have successfully identified more than 60 loci as associated with MS risk,10 ,11 and other studies have suggested a predictive gene–environment interaction between either childhood UVR exposure or vitamin D intake with MS risk and vitamin D receptor (VDR) polymorphisms.12 ,13 Less success has been received in the search for genetic modulators of MS clinical course14 and disease severity.11 This may be due to the narrow time frame during which relapses occur and during which genes may exert their effects on relapse risk, directly and via interaction with environmental factors.
We were interested in whether the magnitude of the association between 25(OH)D and relapse was dependent on the genotype. Therefore, using a well-validated prospective cohort study design, we evaluated whether there was interaction between serum 25(OH)D, a number of genetic predictors involved in the vitamin D metabolism and VDR/retinoid X receptor (RXR) transcription factor formation pathway and subsequent hazard of relapse in MS.
Materials and methods
The Southern Tasmanian MS Longitudinal Study was designed as a prospective cohort study to evaluate the role of UV exposure and 25(OH)D on the clinical course of MS.6 ,15 Briefly, this study followed a cohort of 198 persons with clinically definite MS (2001 McDonald criteria16) living in southern Tasmania, Australia between 2002 and 2005. Of these, 145 participants of the relapsing–remitting MS (RRMS) phenotype were followed beyond one review and 141 participants had genotype data. When participants discontinued participation or were lost to follow-up (8/198; 4%), they were censored at the date of study exit or their last attended review. Ethics approval was obtained from the Southern Tasmania Human Research Ethics Committee. All participants provided informed consent.
Measurement of relapses and 25(OH)D
These measurements have been described in detail elsewhere.6 Briefly, relapses were defined according to established criteria16 and were reported by phone or at each 6-monthly review. All relapse reports were validated by the study physician and study neurologist.
At each summer and winter review (January–April and June–September), blood samples were taken. All samples were stored at −80°C and shielded from light. Since 25(OH)D is the major circulation form of vitamin D, provides the best estimate of a patient’s long-term vitamin D status17 and has been associated with MS onset and relapse, free and bound serum 25(OH)D concentrations were measured using a commercially available radioimmunoassay (Stillwater, Minnesota-DiaSorin Inc). Assays were performed following the conclusion of the study. Consequently, neither participants nor study personnel were aware of participants’ 25(OH)D concentrations during the study.
A total of 164 MS cases were genotyped on the Illumina Infinium Hap370CNV array as a part of the Australia and New Zealand Multiple Sclerosis Genetics Consortium (ANZgene).18 An additional 29 MS cases were genotyped using the Illumina HumanOmniExpress-12v1_A array. All the samples were previously identified as being from persons of European descent,18 and a conservative quality control was conducted with PLINK19: individuals with call rates less than 0.90, single nucleotide polymorphisms (SNPs) with call rates less than 0.95 or in Hardy–Weinberg equilibrium (p<10−7), or duplicates were excluded, leaving 189 cases with 290 536 SNPs.
Vitamin D pathway analysis and SNP selection
It has long been recognised that genes do not work alone, but in an intricate network of interaction. As our a priori hypothesis was that there are gene–vitamin D interactions which modulate the clinical course of MS, we therefore generated a canonical vitamin D metabolism and VDR/RXR transcription factor formation pathway using Ingenuity Pathways Analysis (IPA) (http://www.ingenuity.com), which includes an extensive Knowledge Base derived from published interactions between gene products and the different forms of vitamin D. The pathway of vitamin D metabolic processes and formation of the VDR/RXR transcription factor complex is shown in figure 1. In total, there were 21 genes involved in this pathway. For each of these genes, we selected all SNPs from the Hg19 genome and 1000 genome data20 with minor allele frequency (MAF) ≥1% located within the physical boundaries, including 1 kb upstream or downstream, and the genotyped SNPs for each of these genes were filtered further based on the genotype dataset. Finally, 276 genotyped SNPs from the 21 genes were tested (see online supplementary table S1), for those SNPs with high linkage disequilibrium (LD) in the same gene, TagSNPs were selected using the r2-based tagger tool SNAP.21
Height (m) and weight (kg) were measured at study entry, and body mass index (BMI) was calculated as weight divided by height squared. Also at study entry, skin melanin density was measured on the upper inner arm using a spectrophotometer.22 Clinical disability was measured every winter review by a single physician, including the Expanded Disability Status Scale (EDSS).
Predictors of 25(OH)D were evaluated by multilevel mixed-effects linear regression (SEs inflated to allow for unknown covariance structure), with SNPs as the predictors and 25(OH)D as the outcome adjusted for season of measure (summer, winter), age, BMI, melanin density (%), current sun exposure (‘less than 0.5 h a day; ‘1/2–1 h a day’; ‘1–2 h a day’; ‘2–3 h a day’; ‘3–4 h a day’; greater than 4 h a day’), vitamin D supplementation (‘never’; ‘less than 200 IU a day’; ‘200–400 IU a day’), fish intake (‘never’; ‘less than once a week’; ‘1–2 serves a week’; ‘2–4 serves a week’; ‘greater than 4 serves a week’) and smoking (No, Yes). Dependent variables were transformed as required to make the residuals less heteroskedastic; however, all regression coefficients are reported on the scale of the original variable.
The hazard of relapse was modelled as a function of 25(OH)D and other covariates on time-to-relapse and was calculated using Cox proportional hazards models for repeated events, using the gap-time model described by Prentice and colleagues.23 SEs were adjusted to reflect multiple events per person. All covariates satisfied the proportional hazards assumption with the exception of the binary variable for sex and the categorical variable for baseline EDSS (0–2.5, 3.0–5.0, 5.5–7.0, 7.5–9.0). For this reason, all models are stratified to allow the baseline hazards to differ by sex and baseline EDSS category. 25(OH)D was estimated at monthly intervals between summer and winter measurements using methods described previously.6 Briefly, a sinusoidal curve was fitted to the measured 25(OH)D to yield a sample average value of 25(OH)D over the year, this used to predict values at monthly intervals between measures for each person. The sinusoidal regression model waswhere t denotes the day of the year the sample was collected, and βj(j=0, 1, 2) are estimated regression coefficients. These modelled 25(OH)D values were then used as the primary predictor variable in survival analysis models. To examine whether there was an interaction between modelled 25(OH)D and a SNP on relapse, a product term was included in the model, and then, the association between 25(OH)D and hazard of relapse can be estimated at each SNP allele level.
To estimate cumulative effects of the significant and other top SNPs involved in the vitamin D pathway on relapse or on 25(OH)D levels, we created a variable that provided values for the number of risk genotypes to represent the ‘genetic risk score’. Briefly, a person was designated as having a risk variant of a SNP if they carried a SNP that was independently associated with an increased hazard of relapse. For instance, a person for whom the heterozygote and minor homozygote were each associated with an increased hazard of relapse, all persons heterozygous or homozygous for the minor allele would be considered carriers of the risk variant, whereas homozygotes for the major allele would be considered non-carriers of the risk variant. The total number of risk variants was summed and the risk score included as a covariate in the model.
Significance threshold was adjusted by Bonferroni correction for the number of genes evaluated. SNPs with a p value <0.00238 (0.05/21) or the adjusted p value (padj)<0.05 after adjustment for covariates and multiple comparisons for all 21 genes were considered to be significant.
All data analyses were performed using STATA/IC V.12.1 (StataCorp LP, College Station, Texas, USA).
The total of 141 participants who were of RRMS course and followed beyond one review, and for whom SNP data were obtained, comprised the cohort for analysis. This group was followed up for an average of 2.3 years, included 75.2% females, and had mean age 45.9 years (SD, 10.2), and 82.3% (116/141) used immunomodulatory therapy during the study. The mean EDSS at study entry was 2.8 (SD, 1.5), and the mean MS duration from diagnosis was 6.95 years (SD, 7.08). A total of 122 confirmed relapses occurred in 70 participants.
We found that the relationship between 25(OH)D and hazard of relapse was significantly different for different alleles of two genotyped SNPs (rs908742 in PRKCZ and rs3783785 in PRKCH), even after adjustment for covariates (age, sex and baseline EDSS) and this interaction persisted after adjustment for multiple comparisons (pinteraction=0.001, padj=0.021, respectively; table 1). For example, a significant protective association was found between 25(OH)D and relapse when subjects carried the major homozygous genotype of rs3783785 (HR=0.76, p=4.65×10−6) and rs908742 (HR=0.77, p=3.63×10−5). However, it was not significant when subjects carried at least one copy of the rare allele of rs3783785, and an increased risk was observed when individuals carried the minor homozygous genotype of rs908742 (HR=1.18, p=0.017; table 2). The modulating effect of the two SNPs also persisted after adjustment for immunomodulatory medication use (data not shown). The two SNPs were not associated with hazard of relapse or levels of 25(OH)D (table 1).
We also found two genotyped SNPs (rs1993116 in CYP2R1 and rs7404928 in PRKCB) that were significantly associated with lower levels of 25(OH)D; persisting after adjustment for sun exposure and other relevant confounders and after adjustment for multiple comparisons (ptrend=0.001, padj=0.021, respectively; table 1). A clear dose response was observed with the CYP2R1 SNP rs1993116, compared to those homozygous for the minor allele heterozygotes had levels of 25(OH)D that were significantly lower by 7.1 nmol/L (p=0.005) and major allele homozygotes were significantly lower by 12.4 nmol/L (p=0.004). Dose response could not be examined reliably for rs7404928 because of the low numbers in the minor allele group (table 3). These two SNPs were not associated with hazard of relapse and did not modify the association between 25(OH)D and relapse (table 1).
We then examined the combined effect on 25(OH)D levels of the two genotyped SNPs associated with low 25(OH)D (rs1993116 in CYP2R1 and rs7404928 in PRKCB) by creating a ‘genetic risk score’. We found that 25(OH)D levels were significantly lower by 13.58 nmol/L (p=0.00025) for those individuals carrying two ‘risk’ genotypes compared to those carrying less than one ‘risk’ genotype (table 3). This cumulative effect of these two SNPs accounted for 3.5% of the variation in 25(OH)D levels with rs1993116 in CYP2R1 accounting for 2.6% of the variation.
We did not identify any SNPs that were significantly associated with hazard of relapse after adjusting for covariates and multiple comparisons. However, four genotyped SNPs (rs281508 and rs6740453 in PRKCE, rs3733359 in GC and rs3818740 in RXRA; table 4) were significantly associated with the hazard of relapse after adjustment for covariates but not multiple comparisons. When we examined the combined effect on relapse of these SNPs by creating a ‘genetic risk score’, we found a dose–response effect with increasing number of risk genotypes (ptrend=8.86×10−6). For example, compared to subjects carrying less than two risk genotypes, those with three and four risk genotypes had an increased relapse HR of 1.95 (95% CI 1.17 to 5.25) and 4.21 (95% CI 2.27 to 7.81; table 4), respectively.
In a prospective MS cohort designed to assess the effects of vitamin D and personal UVR exposure on MS clinical course, we have shown that two loci within the genes involved in the vitamin D pathway interact with an environmental factor, serum 25(OH)D, to influence the clinical course of MS. Two SNPs (rs908742 in PRKCZ and rs3783785 in PRKCH) were found to significantly modify the association between serum 25(OH)D and hazard of relapse. In line with this finding, we identified a number of other genes involved in the vitamin D pathway that were associated with hazard of relapse, although these associations were not significant after adjustment for multiple comparisons. We also identified two SNPs (rs1993116 in CYP2R1 and rs7404928 in PRKCB) that were significantly associated with 25(OH)D levels.
Previously, we found that higher 25(OH)D levels were associated with lower relapse risk in this same cohort.6 Interestingly, the inverse association between 25(OH)D and relapse was observed among those with the major homozygous genotypes of rs3783785 and rs908742, both located within introns of the protein kinase C (PKC) family genes PRKCH and PRKCZ. PKC family genes have been shown to be associated with other neurological disorders including Alzheimer's disease, status epilepticus and cerebellar ataxia.24 PKC family genes have been shown to be regulated by 1,25(OH)2D in chondrocytes25 and mediated by VDR in downstream signalling pathways.26 Our findings suggest that PRKCH and PRKCZ do not have strictly independent roles but work in conjunction with each other, as we found that other genes in the PKC family were associated with 25(OH)D levels (PRKCB), and hazard of relapse (PRKCE), although some of these associations were not significant after adjustment for multiple comparisons. For those associated with relapse, we demonstrated a cumulative effect with increasing number of risk genotypes, with those with three and four risk genotypes having hazard of relapse of 1.95 (95% CI 1.17 to 5.25) and 4.21 (95% CI 2.27 to 7.81; ptrend=8.86×10−6), respectively.
MS is believed to be the result of a misdirected immune response by autoreactive T cells against as yet undetermined CNS antigens.27 PRKCH, PRKCZ, PRKCB as well as other genes in the 11-member PKC family have been found to affect T-cell activation.28–30 It is possible therefore that abnormalities in PKC activity induced by one or more of these polymorphisms may lead to altered T-cell function. PRKCA has been reported to be associated with MS risk in UK, Finish and Canadian populations,31 ,32 and PRKCB and PRKCH had been associated with the risk of similar autoimmune diseases including systemic lupus erythematosus (SLE)33 and rheumatoid arthritis.34 ,35 In rheumatoid arthritis, PRKCH messenger RNA was expressed at high levels in T cells and was significantly down-regulated during immune response.34
We also identified two SNPs within the CYP2R1 and PRKCB genes that were inversely associated with levels of 25(OH)D. A CYP2R1 SNP (rs10741657) has previously been associated with vitamin D levels in a vitamin D GWAS.36 Notably, the SNP (rs1993116) identified in this cohort is in complete LD with rs10741657 (LD=1), which supports the association identified in our cohort. Interestingly, the SNP (rs7404928) in PRKCB was suggested to be associated with risk of rheumatoid arthritis34 ,35 and was identified in this cohort as a novel polymorphism associated with low 25(OH)D levels, but was not identified in the vitamin D GWAS. PRKCB is a member of the PKC gene family that has been demonstrated to be regulated by 1,25(OH)2D in chondrocytes25 and mediated by VDR in downstream signalling pathways.26 Studies have shown that PRKCB increases phosphorylation of VDR37 and is a key element in normal T-cell migration,29 and 25(OH)D has been demonstrated to modulate TReg (regulatory T cell) and TH cell function in vivo,38 suggesting that PRKCB may mediate some of the effect of 25(OH)D on T-cell function. Notably, the cumulative effects of these two SNPs accounted for 3.5% of the variation in 25(OH)D concentrations, in which rs1993116 in CYP2R1 accounted for 2.6%, which was similar to that identified from the vitamin D GWAS.36
We now consider most common diseases to be genetically influenced by the combined effects of many loci and/or a number of rare variants that interact with environmental factors in multiple ways.39 In this study, we observed a cumulative effect of several SNPs on either the risk of relapse or 25(OH)D levels, supporting the theory that pathway analysis may increase the overall effect within known signalling pathways and enhance the genetic susceptibility to disease. Such an approach has proved useful in examining the contribution of T-cell function, cell adhesion and cellular communication in the pathogenesis of MS.11 ,40 In our study, using pathway analyses, we have been able to demonstrate that several genes potentially work together to influence 25(OH)D levels and/or the hazard of relapse. Consequently, modulation of these genes and/or pathways is an important potential avenue of investigation in the treatment of MS and other immunological disorders.
A strength of our study was the availability of detailed environmental data, including potential confounders such as sun exposure and vitamin D supplementation. A weakness of our study is the sample size. Even though this is one of the largest and most well-studied MS cohorts available, the difficulty of reaching statistical significance reflects the challenge of undertaking genetic studies of clinical course, in comparison with aetiology studies. These questions of gene–environment interactions on MS clinical course can only be answered in longitudinal studies of this nature. Therefore, as discussed above, we have used other methodologies to overcome potential type 1 error, including allele dose responses and cumulative genotype risk scores. While some of the associations identified here are not significant after adjustment for multiple comparisons, we believe that these results still provide important evidence for real associations. Our results demonstrate one methodology that can bridge the gap between the need for larger and larger genetic studies and the realities of undertaking MS longitudinal studies.
In conclusion, our data support the hypothesis that gene–vitamin D interactions may influence MS clinical course. We found two novel polymorphisms within the PRKCH and PRKCZ genes that significantly modified the relationship between 25(OH)D levels and relapse. We found other SNPs within the PKC family genes that were associated with serum 25(OH)D and MS relapse. These findings indicate that the PKC family genes may play a role in the pathogenesis of MS relapse through modulating the association between 25(OH)D and relapse. Replication studies and investigations into the molecular mechanisms by which PKC members alter relapse are required.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online table
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.