T cell composition and polygenic multiple sclerosis risk: A population‐based study in children

Abstract

Background and purpose

Patients with multiple sclerosis (MS) have altered T cell function and composition. Common genetic risk variants for MS affect proteins that function in the immune system. It is currently unclear to what extent T cell composition is affected by genetic risk factors for MS, and how this may precede a possible disease onset. Here, we aim to assess whether an MS polygenic risk score (PRS) is associated with an altered T cell composition in a large cohort of children from the general population.

Methods

We included genotyped participants from the population‐based Generation R study in whom immunophenotyping of blood T cells was performed at the age of 6 years. Analyses of variance were used to determine the impact of MS‐PRSs on total T cell numbers (n = 1261), CD4⁺ and CD8⁺ lineages, and subsets therein (n= 675). In addition, T‐cell‐specific PRSs were constructed based on functional pathway data.

Results

The MS‐PRS negatively correlated with CD8⁺ T cell frequencies (p = 2.92 × 10⁻³), which resulted in a positive association with CD4⁺/CD8⁺ T cell ratios (p = 8.27 × 10⁻⁹). These associations were mainly driven by two of 195 genome‐wide significant MS risk variants: the main genetic risk variant for MS, HLA‐DRB1*15:01 and an HLA‐B risk variant. We observed no significant associations for the T‐cell‐specific PRSs.

Conclusions

Our results suggest that MS‐associated genetic variants affect T cell composition during childhood in the general population.

Multiple sclerosis (MS)‐associated genetic variants affect T cell composition during childhood (n = 675). Higher MS polygenic risk was associated with lower total and naive CD8⁺ T cell numbers at the age of 6 years, resulting in a higher CD4⁺/CD8⁺ ratio. Genetic MS risk may contribute to overall MS risk later in life by altering the immune system at a young age.

INTRODUCTION

The exact pathophysiology of the autoimmune response in multiple sclerosis (MS) remains to be elucidated, and results from a complex interplay between genetic and environmental risk factors that contribute to disease risk. Migration studies suggest a pivotal time window during childhood in which environmental risk factors (e.g., Epstein–Barr virus [EBV] or cytomegalovirus [CMV] seropositivity, and serum 25‐hydroxyvitamin‐D [25(OH)D]) contribute to MS risk [1, 2].

Genome‐wide association studies (GWASs) have identified 233 genetic variants (single nucleotide polymorphisms [SNPs]) that significantly affect disease susceptibility in adult patients [3]. These studies have shown that the genetic risk of MS is polygenic, with a large number of genetic variants that each have a small effect on disease risk [3]. However, the majority of people with high genetic risk for MS are never diagnosed with the disease [4]. How these risk variants together affect biological mechanisms culminating in a higher risk of developing MS and at which time points during the lifespan are unknown. To capture this combined effect of genetic risk variants, polygenic risk score (PRS) analyses can be used [5].

T cells are suggested to be the main immune lineage that are directly or indirectly influenced by genetic MS risk variants [3]. Several studies have described altered T cell homeostasis in MS patients, with increased CD4⁺/CD8⁺ ratios that decrease during treatment [6, 7].

Here, we aimed to determine how PRSs for MS are associated with the distribution of naive, memory, and activated CD4⁺ and CD8⁺ T cell populations in 6‐year‐old children from the general population. We assessed the main genetic variants driving these associations and the relationship between overall genetic MS risk and EBV and CMV seropositivity, as well as serum 25(OH)D levels.

MATERIALS AND METHODS

Study sample

This study was performed in the context of the population‐based Generation R study (n = 9749), located in the Netherlands, which investigates various aspects of childhood development [8]. At the age of 6 years, absolute numbers of total T cells were measured in 3465 participants, and detailed phenotyping of T cell subsets was performed in 1079 participants [9]. In the current study, we included unrelated participants of European ancestry, who had high‐quality genotype data available (Figure 1).

FIGURE 1.

Flowchart of the participant selection in the current study

The study protocol of the Generation R study has been approved by the Medical Ethical Committee of the Erasmus Medical Center in Rotterdam. The legal representatives of the children provided written informed consent.

Genotype data

Sample collection of the genetic data and genotype calling procedures have been reported in previous work [10]. Genotype data were collected at birth, derived from cord blood, or during a visit to the research center, and genotyped on Illumina 610K and 660K SNP arrays. Information on the quality control of the genetic data, the imputation method, and the calculation of principal components can be found elsewhere [11]. To summarize, we selected subjects of European ancestry based upon the first four principal components inside the range of the HapMap Phase II Northwestern European founder population [12]. Furthermore, we imputed the genotype data using 1000 Genomes (Phase I, Version 3) data to calculate our PRSs [13].

Polygenic scoring

We used a large discovery GWAS for MS (N = 41,505; 14,802 cases/26,703 controls), carried out by the International Multiple Sclerosis Genetics Consortium (IMSGC; https://imsgc.net/), using imputed genotype data to estimate weighted PRSs [3]. We used PRSice 2 [14], an R‐script to calculate clumped (r ² < 0.10, window = 250 kB) and p‐value thresholded polygenic scores in PLINK (v1.9) [15], for the computation of our PRSs. We used various p‐value thresholds (P _T < 0.001, 0.005, 0.01, 0.05, 0.1, 0.5, 1) in the calculation of the PRSs to include the effect of suggestive variants and capture the polygenic architecture of MS [3]. In addition, we computed a MS‐PRS only incorporating the genome‐wide risk variants (P _T < 5 × 10⁻⁸) from the final meta‐analysis of the IMSGC (47,429 cases/68,374 controls).

Because of the strong effect of the major histocompatibility complex (MHC) region on MS pathophysiology [3], three sets of PRSs were calculated: one with all SNPs included, one excluding SNPs from the MHC region, and one only including SNPs from the MHC. To characterize the presence of HLA‐DRB1*15:01, we used rs3135388 as tag SNP [16]. Rs9266629 was included in our analyses as tag SNP for HLA‐B genetic risk variant rs3819284. In addition, we used rs9268839, rs1057149, and rs2187688 as tag SNPs for HLA‐DRB variants of importance in rheumatoid arthritis, diabetes mellitus type 1, and systemic lupus erythematosus to investigate possible confounding of the T cell distribution [17, 18].

Due to the low minor allele frequency (<0.20) of rs3135388 and rs9266629, we dichotomized their carriership into two groups, homozygous non‐risk and at‐risk participants. At‐risk participants consisted of heterozygous and homozygous carriers of the effect allele.

Construction of T‐cell‐specific PRSs

In a final analysis, we used cis expression quantitative trait loci (cis‐eQTL) effect data to associate the MS genome‐wide risk variants with particular T cell subsets, providing us with the possibility to construct PRSs specific for these subsets.

CD4⁺ and CD8⁺ T‐cell‐specific PRSs were constructed by assessing whether a locus implicated by an SNP [3] had high mRNA expression in a specific subset [19, 20, 21, 22]. Additionally, eQTL were assessed for all MS risk SNPs and assigned to specific T cell subsets [20]. Lastly, gene ontology analyses were performed on all MS risk SNPs to assign loci to functional pathways in the aforementioned lymphocyte subsets [23, 24, 25].

EBV and CMV serology

To identify the presence of immunoglobulin G antibodies against CMV and EBV‐viral capsid antigen (VCA), enzyme‐linked immunoassays were used, as described before [9]. Seropositivity was defined by a sample‐threshold ratio greater than 0.6 (CMV) and 0.8 (EBV capsid antigen).

Vitamin D assessment

Measurements of 25(OH)D were conducted in blood sera of children at the age of 6 years (110 μl per sample) using the liquid chromatography/tandem mass spectrometry method, which is a highly sensitive method for measuring 25(OH)D levels, commonly used in epidemiological studies [26]. Serum 25(OH)D levels were residualized for season of blood draw to adjust for the nonlinear effect of season on serum 25(OH)D levels.

T cell phenotyping

The presence and phenotype of CD4⁺ and CD8⁺ T cell populations were analyzed in whole blood using six‐color flow cytometry [9]. Absolute T cell numbers were determined with a routine diagnostic lyse‐no‐wash protocol [9]. We analyzed the following subsets within the CD4⁺ and CD8⁺ T cell lineage [9]: naive (CD45RO⁻CCR7⁺), central memory (Tcm; CD45RO⁺CCR7⁺), CD45RO‐positive effector memory (CD45RO⁺CCR7⁻), CD45RA‐positive effector memory (TemRA; CD45RO⁻CCR7⁻), terminally differentiated (CD57⁺), and activated (HLA‐DR⁺) T cells. CD56⁺ natural killer T cells were excluded from our analysis. The above subsets were analyzed as part of the Generation R study [6, 7]. CD4⁺/CD8⁺ T cell ratios were calculated by dividing the total number of CD4⁺ and CD8⁺ T cells. Flow cytometric data were obtained using an LSRII flow cytometer (BD Biosciences) with standardized instrument settings and analyzed using FACSDiva analysis software [27, 28].

Statistical analyses

We performed our analyses using R statistical software (v3.5.1) [29]. Before our analyses, we divided the MS‐PRSs into quartiles. In the first analyses, we analyzed the association between environmental risk factors for MS (EBV, CMV, and vitamin D) and our PRSs, to investigate possible confounding in our subsequent T cell analyses. To determine the association between the MS‐PRSs' quartiles and EBV and CMV seropositivity, logistic regression was performed across the different p‐value thresholds. We used analyses of variance (ANOVAs) to investigate the relationship between the MS‐PRSs' quartiles and serological 25(OH)D levels.

All T cell outcomes of interest were log‐transformed to ensure normally distributed residuals of our ANOVA models in the subsequent T cell analyses. At first, the MS‐PRSs' quartiles, based on different thresholds, were all associated with absolute T cell counts using ANOVAs. Next, the PRS threshold with the strongest association (lowest p‐value) was used in subsequent ANOVAs investigating the associations with different T cell subpopulations of interest. In addition, we performed ANOVAs with the MS‐PRS including only the genome‐wide significant risk variants, to compare the effects of these variants on T cell populations with the effects of suggestive variants. Tukey tests were used to investigate post hoc differences in significant ANOVA associations.

In all our analyses, we adjusted for age at blood withdrawal, sex, and the first 10 genetic principal components. Effect modification by sex was tested in a separate sensitivity analysis. We used false discovery rate (FDR) per research question to account for multiple testing [30].

RESULTS

Participant characteristics

Of the participants for whom absolute T cell counts were determined, 2286 had genotype data available. After selection on relatedness, genotype quality, and European ancestry, 1261 participants remained eligible for analyses using absolute T cell numbers (Figure 1). Participants had an even distribution of sex (50.8% male) and a median age of 6.0 years (interquartile range = 5.9–6.2; Table 1). No difference was found in mean PRS for MS compared to the participants who had no absolute T cell counts available (n = 1569; p = 0.62).

TABLE 1.

Demographics and laboratory measurements in study participants

Characteristic	Absolute T cell counts, n = 1261	T cell phenotyping, n = 675
Age, years, median (IQR)	6.0 (5.9–6.2)	6.0 (5.9–6.2)
Male, n (%)	641 (50.8)	340 (50.4)
Level of maternal education, n (%)
High	818 (64.8)	419 (62.1)
Middle	405 (32.1)	239 (35.4)
Low	17 (1.3)	9 (1.3)
Unknown	21 (1.7)	8 (1.2)
Season of blood draw, n (%)
Spring	282 (22.4)	171 (25.3)
Summer	290 (23.0)	150 (22.2)
Autumn	351 (27.8)	174 (25.8)
Winter	338 (26.8)	180 (26.7)
EBV seropositivity, n (%)	512/1225 (41.8)	270/649 (41.6)
CMV seropositivity, n (%)	353/1227 (28.8)	191/649 (29.4)
Serum 25(OH)D levels, nmol/L, median (IQR)	74.0 (58.7–88.0)	74.0 (57.0–88.0)
Reported presence of paternal MS, n (%)	2/1036 (0.2)	0/572 (0.0)
Reported presence of maternal MS, n (%)	2/1144 (0.2)	0/611 (0.0)

Abbreviations: 25(OH)D, 25‐hydroxyvitamin‐D; CMV, cytomegalovirus; EBV, Epstein–Barr virus; IQR, interquartile range; MS, multiple sclerosis.

Detailed immunophenotyping data were available from 1079 participants. Selection on relatedness, genotype quality, and ancestry left us with 675 participants eligible for the detailed T cell subset analyses (Figure 1). Descriptive characteristics of these participants were comparable to the participants with absolute T cell counts available (Table 1), and the MS‐PRS did not differ from participants who did not take part in the detailed immunophenotyping (n = 2155; p = 0.42).

EBV and CMV serology and serum 25(OH)D levels

EBV and CMV serology was measured in 4464 participants. Of these participants, 1551 had good‐quality genotype data available. We found no relationship in these participants between the PRSs' quartiles for MS and EBV and CMV seropositivity (Tables S1 and S2).

Serum 25(OH)D levels were measured in 3983 participants, of whom 1442 had good‐quality genotype data available. The known association of the rs7041 A‐allele with lower 25(OH)D levels was replicated in our dataset (p = 5.19 × 10⁻¹⁰) [31]. The genotype data did not include rs5688. The MS risk SNP located in the CYP24A1 gene (rs2248137) and the MS‐PRSs' quartiles across all thresholds were not associated with the level of serum 25(OH)D in these participants (Table S3).

Effects on absolute T cell counts

We found no significant association between the MS‐PRSs' quartiles and absolute T cell counts across all thresholds (Table S4). The PRS with a threshold of P _T < 0.005 revealed the strongest association with absolute total T cell counts (p = 0.29) and was used in our subsequent detailed immunophenotyping analyses.

Associations with T cell subpopulations

The MS‐PRS (P _T < 0.005) had a significant association with total (p = 2.92 × 10⁻³) and naive (p = 1.85 × 10⁻⁴) CD8⁺ T cell numbers and CD4⁺/CD8⁺ T cell ratios (p = 8.27 × 10⁻⁹). A complete overview of the associations between the MS‐PRS (P _T < 0.005) and T cell populations can be found in Table 2. Post hoc Tukey tests revealed significant negative associations of the first and second MS‐PRSs' quartiles with the fourth quartile considering both total and naive CD8⁺ T cell numbers (Figure 2a,b). Accordingly, this resulted in a significant positive association of the PRS with the CD4⁺/CD8⁺ ratio across multiple quartiles (Figure 2c). In the sensitivity analyses, we found no evidence for interaction effects by sex (Table S5).

TABLE 2.

Analysis of variance results of the multiple sclerosis polygenic risk score (P _T < 0.005) and T cell subsets of interest

Cell subset	Unadjusted p	FDR‐adjusted p
Total CD4+ T cells	0.17	0.71
CD4+ Naive	0.14	0.70
CD4+ Tcm	0.54	0.77
CD4+ TemRO	0.49	0.77
CD4+ TemRA	0.46	0.77
Total CD8+ T cells	2.92 × 10−3	0.02a,*
CD8+ naive	1.85 × 10−4	3.14 × 10−3a,*
CD8+ Tcm	0.25	0.72
CD8+ TemRO	0.32	0.72
CD8+ TemRA	0.61	0.77
CD3+CD8− HLA‐DR+	0.53	0.77
CD3+CD8+ HLA‐DR+	0.27	0.72
CD3+CD8+CD57−CD56−	0.55	0.77
CD3+CD8+CD57+CD56−	0.56	0.77
CD3+CD8−CD57−CD56−	0.59	0.77
CD3+CD8−CD57+CD56−	0.83	0.91
CD4+/CD8+ ratio	8.27 × 10−9	2.81 × 10−7a,**

Included: n = 675 children; data are corrected for age, sex, and 10 genetic principal components.

Abbreviations: FDR, false discovery rate; Tcm, central memory T cells; TemRA, RA‐positive effector memory T cells; TemRO, RO‐positive effector memory T cells.

^aSignificant values after FDR multiple testing correction.

*p < 0.05, **p < 0.001.

FIGURE 2.

Post hoc Tukey test results of the different polygenic risk scores (PRSs) associated with the CD4⁺/CD8⁺ ratio and other T cell subsets. Numbers on the x‐axis correspond to the different quartiles of the multiple sclerosis (MS)‐PRS. Based on increasing MS‐PRS, the participants were divided into four quartiles of similar sample size. (a) Association between total CD8⁺ cells and the quartiles of the MS‐PRS (P _T < 0.005). (b) Association between naive CD8⁺ cells and the quartiles of the MS‐PRS (P _T < 0.005). (c) Association between the CD4⁺/CD8⁺ ratio and the quartiles of the MS‐PRS (P _T < 0.005). (d) Association between the CD4⁺/CD8⁺ ratio and the quartiles of the MS‐PRS (P _T < 5 × 10⁻⁸)

To investigate the MHC dependency of these significant associations, we excluded the HLA‐DRB1*15:01 tag variant rs3135388 and the MHC region from our MS‐PRS (Tables S6–S9). The removal of rs3135388 (and variants with high linkage disequilibrium to this variant within a 1‐Mb region) attenuated the significant CD8⁺ results, but these associations remained significant and the association involving the CD4⁺/CD8⁺ ratio increased in significance. Additional adjustment for HLA‐DRB1*15:01 status within this model still resulted in a significant association between the MS‐PRS and the CD4⁺/CD8⁺ ratio (Table S7). When adding the status of several other HLA‐DRB alleles to our model as covariates, we again observed a significant association of the MS‐PRS with the CD4⁺/CD8⁺ ratio (Table S8). After removal of the total MHC region, the association between the MS‐PRS and CD4⁺/CD8⁺ ratio became weaker, but remained significant (Table S9). The PRS with only MHC risk variants showed comparable results to our PRS including both MHC and non‐MHC risk variants (Table S10). Due to the different linkage disequilibrium structure of the MHC region, we constructed several PRSs using various clumping parameters for the MHC region. However, altered clumping of the MHC did not significantly change our results (Table S11).

In addition, we calculated an MS‐PRS only incorporating risk variants that were genome‐wide significant (P _T < 5 × 10⁻⁸) in the MS GWAS performed by the IMSGC [3], and analyzed the association with the T cell subsets (Table S12). We observed a similar positive association of the first and second quartiles of the MS‐PRS with the fourth quartile concerning the CD4⁺/CD8⁺ ratio (Figure 2d and Table S12).

Individual risk variant analyses

Subsequently, we investigated the relationship between an increased CD4⁺/CD8⁺ ratio and individual genetic MS risk variants. Of the 195 of 233 MS genome‐wide significant risk SNPs that were available in our study population, two risk variants showed a significant association with the CD4⁺/CD8⁺ ratio after FDR correction (Table S13, Figure 3a). A positive association was found for rs3135388 (HLA‐DRA, a tag variant for HLA‐DRB1*15:01), whereas a negative association was observed for rs9266629 (intergenic, between ZDHHC20P2 and FGFR3P1, a tag variant for HLA‐B variant rs3819284).

FIGURE 3.

Associations of rs3135388 and rs9266629 with T cell subsets and the CD4⁺/CD8⁺ ratio. Carriership of rs3135388 and rs9266629 was dichotomized into two groups, homozygous non‐risk and at‐risk. At‐risk participants consisted of heterozygous and homozygous carriers of the effect allele. (a) Left: Associations of rs3135388 with the CD4⁺/CD8⁺ ratio and CD4⁺ naive cells. Right: Associations of rs9266629 with the CD4⁺/CD8⁺ ratio and CD8⁺ naive cells. (b) Combined associations of rs3135388 and rs9266629 with the CD4⁺/CD8⁺ ratio

Next, we associated these two risk variants with the previous T cell subsets of interest (Tables S14 and S15). Rs3135388 status (MHC Class II) was positively associated with CD4⁺ total and naive numbers, whereas rs9266629 (MHC Class I) showed positive associations with various CD8⁺ subsets (e.g., total CD8⁺, CD8⁺ naive, and CD8⁺ Tcm; Figure 3a). Figure 3b shows the combined effects of the two risk variants on the CD4⁺/CD8⁺ ratio.

T cell PRS

In a final analysis, the autosomal genome‐wide significant MS risk variants from the IMSGC were classified into specific MS‐associated T cell PRSs (Figure 4, Table S16). We observed no significant associations between the different PRSs and the T cell outcomes, including the CD4⁺/CD8⁺ ratio.

FIGURE 4.

Flowchart describing the selection process of specific T cell polygenic risk scores using cis expression quantitative trait loci data. Risk variants were grouped using a combination of mRNA expression data, expression quantitative trait loci, and gene ontology analyses as described in the Materials and Methods section. IMSGC, International Multiple Sclerosis Genetics Consortium; MS, multiple sclerosis; SNP, single nucleotide polymorphism

When incorporating rs3135388 and rs9266629 status into the T cell PRSs, we observed several dose‐dependent significant associations (Tables S17–S19). A higher CD8⁺ PRS quartile was associated with a lower count of CD8⁺ T cell subsets and a higher CD4⁺/CD8⁺ ratio (p = 8.55 × 10⁻⁷). The PRS involving Th1‐, Th17‐, Th17.1‐, and CD8⁺‐associated variants showed positive associations with CD4⁺ TemRA counts (p = 0.01) and the CD4⁺/CD8⁺ ratio (p = 2.34 × 10⁻⁶; Table S18). Similar associations were observed for the PRS incorporating variants associated with Treg and Th2 cells (Table S19). All these associations were driven by rs3135388 and rs9266629 status.

DISCUSSION

In this study, we found that PRSs for MS are associated with T cell numbers in the peripheral blood at the age of 6 years in children from the general population. A higher genetic risk for MS was associated with an increased CD4⁺/CD8⁺ ratio, which was the result of a negative association between the MS‐PRS and total and naive CD8⁺ T cell numbers. Two genome‐wide significant MS risk variants, rs3135388 and rs9266629, were individually associated with the changes in the CD4⁺/CD8⁺ ratio and absolute numbers of several CD4⁺ and CD8⁺ T cell subsets. No associations were found between T‐cell‐specific MS‐PRSs and T cell numbers.

We found no associations between MS‐PRSs and seropositivity for EBV and CMV, and serum 25(OH)D levels. We conclude that MS‐PRSs are associated with T cell composition in children of a general population, and that this mechanism likely contributes to the overall risk of developing MS later in life.

Our findings are consistent with immunological alterations reported in adult MS patients. An increased CD4⁺/CD8⁺ ratio has been described in MS, as well as in other autoimmune disease patients and their relatives [6, 7, 32]. We now found that genetic predisposition for MS alters the CD4⁺/CD8⁺ ratio in children from the general population. This is relevant because the CD4⁺/CD8⁺ ratio is a relatively stable immunological parameter from childhood to young adulthood [33]. Our reported observed negative correlations of the MS‐PRS with CD8⁺ and not CD4⁺ T cells are consistent with previously described reductions of CD8⁺ and not CD4⁺ T cells in the peripheral blood of MS patients [6]. In late stage adult MS, there is a local enrichment of memory CD8⁺ T cells in the central nervous system (CNS) [34, 35]. Accordingly, in MS‐discordant monozygotic twins, a clonally expanded memory CD8⁺ T cell population was found in the cerebrospinal fluid of twins asymptomatic for MS [36]. In children, however, it is not likely that genetically determined reductions in CD8⁺ T cells are the indirect result of an increased influx of memory cells into the CNS, because both central and effector memory T cell populations did not correlate with the MS‐PRS. Instead, genetic variation may impair thymic output of naive CD8⁺ T cells, a phenomenon described in MS patients at an early age possibly before disease onset, which is supported by our observed negative association between the MS‐PRS and naive CD8⁺ T cells [6, 37]. That the positive correlation of the MS‐PRS with the CD4⁺/CD8⁺ ratio is mainly driven by major MHC risk alleles could imply that in children, the development of CD4 and CD8 double‐ into single‐positive thymocytes is influenced by genetically altered HLA Class I‐ and II‐expressing thymic epithelial cells [38, 39].

The major risk variant for MS, HLA‐DRB1*15:01, was found to be associated with higher CD4⁺ T cell numbers, whereas the rs9266629 variant that is protective for MS and a tag‐variant for HLA‐B SNP rs3819284 associated positively with CD8⁺ T cell numbers. The same associations were observed when incorporating these two variants in our MS‐associated T cell PRSs. For CD4⁺ T cells, it has been reported that naive populations are genetically affected and more activated and thus prone to differentiate into effector subsets during early MS [40]. Additionally, it was found that HLA‐DRB1*15:01 status increases autoproliferation and brain‐homing of the CD4⁺ T cells [41]. Disturbed T cell homeostasis in favor of CD4⁺ T cells, influenced by rs3135388, the absence of rs9266629, and overall genetic risk for MS, may consequently be one of the first steps in the immunopathogenesis before possible onset of symptoms in MS. An increased proportion of these cells in peripheral blood could lay the foundation for the activation of autoreactive CD4⁺ T cells later in life under the influence of environmental risk factors and HLA‐DRB1*15:01, as suggested in the immunopathogenesis of MS [42]. Subsequent migration across the blood–brain barrier could initiate the MS disease process, where autoreactive T cells cause inflammatory demyelination of the CNS [42].

The entire MHC region contributed substantially to our results. Upon investigation of an MS‐PRS with only MHC risk variants, our results were comparable to our PRS with all variants. Excluding the MHC region from the MS‐PRS attenuated our results; however, they remained significant. This finding is in line with the widely known substantial role of the MHC region in MS pathogenesis. Due to this strong association of MHC alleles with MS (especially HLA‐DRB1*15:01), we kept these variants in our PRSs, as this is the closest reflection of overall genetic MS risk; however, PRSs are not able to fully capture the effect of the MHC due to the epistatic interactions within this region of the genome.

We did not find an association between MS‐PRSs and known environmental risk factors for MS (i.e., EBV and CMV serology and serum 25[OH]D). Low levels of serum 25(OH)D have been associated with increased MS risk in adults and children [43, 44]. Our results suggest that serum 25(OH)D is not correlated with polygenic risk scores for MS at this age. This environmental risk factor could exert its effect on MS pathophysiology independently, a hypothesis also demonstrated by Mendelian randomization studies [45].

Persistent EBV and CMV seropositivity affects MS susceptibility in adults, but also in children [46, 47]. The observed nonsignificant association in our study between polygenic risk for MS and EBV status could be explained by the EBV status assessment used in our study. EBV seropositivity in the Generation R study was tested by using EBV‐VCA antibodies, which are not associated with MS in adults [48]. However, in pediatric MS, EBV‐VCA positivity is significantly more present compared to controls, pleading for the validity of this marker in children [49]. Another possible explanation is that genetic MS risk does not associate with seropositivity for these viruses at this young age, and that EBV and CMV exert their effect at a later age in MS pathophysiology, as supported by earlier work [50].

Our study has several strengths. First, we were able to include a large number of young children, allowing us to detect robust effects of genetic MS risk even before possible disease onset. Furthermore, we incorporated several suggestive genetic risk variants for MS in our PRSs to reflect the polygenic architecture of MS as much as possible. Additionally, we tested for CMV positivity and serum 25(OH)D levels using accurate methods. Lastly, we performed additional analyses incorporating specific T cell PRSs constructed with functional pathway data.

Limitations are also present. Due to the cross‐sectional design of our study, we are unable to investigate the temporal dynamics between genetic risk for MS and the T cell compartment of the immune system. Second, it was not possible to investigate the functional properties of T cells in this study, such as their proinflammatory and transmigration capacity. Future studies are needed to validate our results, test the reported relationships longitudinally, and incorporate environmental risk factors for MS to address their effect on T cells and to better understand the pathophysiology of MS. Longitudinal well‐powered population‐based study designs could potentially translate the altered CD4⁺/CD8⁺ ratio to a clinically usable biomarker for MS.

In conclusion, we report an association between genetic risk scores for MS and alterations in the T cell lineages in school‐age children from the general population. This shows that MS genetics change the composition of the adaptive immune system during childhood, which possibly contributes to overall risk of developing MS later in life.

CONFLICT OF INTEREST

R.F.N. participates in trials with Sanofi Genzyme and Novartis. None of the other authors has any conflict of interest to disclose.

AUTHOR CONTRIBUTIONS

Casper de Mol: Conceptualization (equal), formal analysis (equal), investigation (equal), methodology (equal), project administration (equal), visualization (equal), writing–original draft (equal), writing–review & editing (equal). Kirsten I. M. Looman: Conceptualization (equal), data curation (equal), formal analysis (equal), methodology (equal), writing–original draft (equal), writing–review & editing (equal). Marvin van Luijn: Conceptualization (equal), investigation (equal), methodology (equal), supervision (equal), writing–original draft (equal), writing–review & editing (equal). Karim L. Kreft: Conceptualization (equal), formal analysis (equal), investigation (equal), methodology (equal), writing–original draft (equal), writing–review & editing (equal). Philip R. Jansen: Conceptualization (equal), formal analysis (equal), investigation (equal), methodology (equal), writing–review & editing (equal). Menno van Zelm: Conceptualization (equal), data curation (equal), investigation (equal), methodology (equal), writing–review & editing (equal). Joost Smolders: Conceptualization (equal), methodology (equal), writing–original draft (equal), writing–review & editing (equal). Tonya White: Conceptualization (equal), methodology (equal), supervision (equal), writing–review & editing (equal). Henriette A. Moll: Conceptualization (equal), data curation (equal), formal analysis (equal), investigation (equal), methodology (equal), supervision (equal), writing–review & editing (equal). Rinze F. Neuteboom: Conceptualization (equal), formal analysis (equal), funding acquisition (equal), investigation (equal), methodology (equal), resources (equal), supervision (equal), writing–original draft (equal), writing–review & editing (equal).

Supporting information

 

de Mol CL, Looman KIM, van Luijn MM, et al. T cell composition and polygenic multiple sclerosis risk: A population‐based study in children. Eur J Neurol. 2021;28:3731–3741. 10.1111/ene.15019

DATA AVAILABILITY STATEMENT

The data used in this study are not publicly available due to legal and informed consent restrictions. However, qualified researchers can request to access the data by contacting the Generation R study (datamanagementgenr@erasmusmc.nl).