As a chronic demyelinating disorder of central nervous system (CNS), multiple sclerosis (MS) is characterized by multiple disseminated inflammatory lesions both temporally and spatially (1). Although ultimately the patients will suffer from irreversible disability, the course of disease development is highly variable, featuring a wide range of focal neurological signs and symptoms that affect mobility, sensation or cognition (2). MS usually clusters within families but does not follow the Mendelian inheritance pattern (3). Such familial recurrence pattern indicates a polygenic risk model for MS, in which the risk is likely conferred by an allele of moderate effect size and several minor-effect alleles (4).

Notably, the genes implicated by these identified loci cluster in key immunological pathways involving lymphocyte activation, receptor signaling and cytokine production (5–7). In accordance, among the approved disease-modifying therapies for relapsing-remitting MS, many have immune modulatory effects, such as directing immune cell trafficking, suppressing auto-reactive T cells, inducing regulatory T cells (Tregs) and regulating B cell activities (8–12). Although their efficacy is expected to diminish with time, which possibly reflects the dwindling roles played by inflammation as the disease develops, early treatment with drugs that reshape the immune environment can indeed limit the rate of evolving to the secondary progressive stage (13). Therefore, the genetic architecture of MS susceptibility provides some solid evidence supporting the view that MS is an immune-mediated disease and highlights the prominent roles of immune dysregulation in MS predisposition.

Since genetic variants are fixed and randomly allocated at conception, Two-sample Mendelian randomization (MR) can exploit these variants as unbiased proxies to approximate the effect of an exposure on the outcome of interest while minimizing the effects of confounding and reverse causation (14, 15). Although both body mass index (BMI) and serum 25-hydroxyvitamin D [25(OH)D] have been consistently shown causally associated with MS via MR approach (16–18), few MR study has addressed the causal relationship between circulating immune cell counts and the disease. To further explore the causal roles played by peripheral immunity on MS risk, we implemented a two-sample MR analysis using recently published data from the largest GWAS to date on blood cell phenotypes (19) and a high-resolution immune cell profiling GWAS (20).

The causal estimates of immune cell count on MS risk are as summarized graphically in Figure 2 . We observed that higher leucocyte count was robustly associated with increased MS susceptibility using IVW method [odds ratio (OR), 1.24; 95% confidence interval (CI), 1.07 – 1.43; p = 0.0039] after correcting for multiple testing (p < 0.05/6 for the six immune cell traits). Among the five major leucocyte subtypes, we also found suggestive evidence that lymphocyte count was positively associated with disease risk (OR, 1.17; 95% CI, 1.01 – 1.35; p = 0.0317). Sensitivity analyses provided consistent results and did not suggest bias from genetic pleiotropy. LOO analysis did not give evidence of SNPs disproportionally affecting the effect estimates. No obvious directional pleiotropy can be detected by visual inspection of funnel plots, and the MR-Egger regression intercept was also insignificant. The results from heterogeneity test, pleiotropy test and F-statistic is summarized in Supplementary Table 11 . Because significant heterogeneity was detected by Cochran Q test in some cell types, the random effect model was used to estimate the MR effect size (32). In contrast, no significant association was observed for neutrophil, basophil, eosinophil, or monocyte cell counts with MS, although a trend of positive correlation can be observed.

We next extended our analyses by further measuring the causal estimates of T cell and B cell on MS risk, the two major lymphocyte subpopulations implicated with the disease. T lymphocytes were gated based on the expression of CD3⁺ (total), CD45RA⁺CCR7⁺ (naive), CCR7⁺CD45RA⁻ (CM), CD45RA⁻CCR7⁻ (EM), CCR7⁻CD45RA⁺ (TD) and CD16/CD56⁺ (NKT). Tregs (CD25^hiCD127^lo) were subdivided into activated (CD25⁺⁺⁺CD45RA⁻), resting (CD25⁺⁺CD45RA⁺) and secreting (CD25⁺⁺CD45RA⁻). B lymphocytes were gated based on the expression of CD19⁺ (total), CD24⁺CD38^hi (transitional), CD24⁻CD38^−/dim (naive), CD27⁺IgD⁻ (switched memory) and CD27⁺IgD⁺ (unswitched memory). Due to the small GWAS sample size, currently we were only able to assess six of these lymphocyte subpopulations via MR methods, i.e., NKT, resting Treg, secreting Treg, TD T cell, B cell and unswitched memory B cell. Results of Wald ratio are presented instead of IVW in cases where less than two IVs is available, and only IVW and cML-BIC are shown when there are less than three IVs ( Figure 3 ). The IVW method revealed that an increase of NKT cell count was associated with a higher risk of MS (OR, 1.24; 95% CI, 1.06 - 1.45; p = 0.0082), which was supported by other MR methods. The causal effects of the other cellular subtypes on MS risk were attenuated to null. We further analyzed the two NKT subsets (CD8^br and CD8^dim) for which qualifying IV(s) is available. The results indicated that neither CD8^br NKT (OR, 1.05; 95% CI, 0.83 - 1.33; p = 0.70) nor CD8^dim NKT (OR, 1.05; 95% CI, 0.95 - 1.17; p = 0.34) subset was causally associated with MS. In addition, genetic predisposition to MS as exposure did not have causal impact on the absolute counts of leukocyte (OR, 1.01; 95% CI, 0.99 - 1.03; p = 0.38), lymphocyte (OR, 1.01; 95% CI, 0.98 - 1.04; p = 0.39) or NKT cell (OR, 1.00; 95% CI, 0.95 - 1.06; p = 0.88) in bi-directional MR analyses.

Epidemiologic and genetic studies have implied correlation between peripheral immunity and MS (3, 21, 33). Since most of the GWAS variants identified are noncoding variants located in the regulatory region of genes, elucidating their functional consequences can be extremely challenging (34, 35). Moreover, due to reverse causation and confounding, drawing causal inferences is also difficult and liable to bias. In comparison, the MR approach, which uses genetic variants robustly associated with the exposure as IVs, can be used to infer causality while controlling for various sources of confounder. In the present study, we first found evidence that peripheral leucocyte and lymphocyte counts were positively associated with the risk for MS using summary statistics from a recent large scale GWAS analyzing blood cell traits. This is in line with some recent MR studies, which indicate a causal relation between circulating interleukins (ILs) and MS (36), and that the association between BMI and MS is partly mediated via IL-6 signaling (37). Considering that the regulatory effects of variants are often dependent on cellular subtypes, it is to be expected that deciphering their pathologic roles will require analyzing the disease-relevant cell types (6, 38). We thus hypothesize that MS might be caused by the dysfunction of a relatively minor proportion of total lymphocytes, in which more prominent effects can be found. We further analyzed several subtypes of B cells and T cells, the two major cellular subtypes within lymphocyte population known to influence MS progression. We found that NKT cells were causally associated with an increased risk for MS, although no significant association was found by subdividing NKT cells based on CD8 expression.

NKT cells, which are innate-like T cells sharing features with both adaptive and innate cells, can be activated rapidly in response to self or foreign lipid antigens and adopt proinflammatory or immunoregulatory phenotypes in mice and human (39). In keeping with their modulatory functions, it has been noted that their levels are lower in the peripheral blood of MS patients (40), which can be restored during remission in patients undertaking IFN-β treatment (41). However, whether the lower percentage of circulating NKT cells indicates that they are protective in MS or whether it is only reflective of disease progression remains controversial. Such complexity can be partly explained by the heterogeneity of this cell population. NKT cells can be broadly divided into two distinct populations based on their expression of T cell receptor (TCR): the classical type I NKT (or invariant NKT, iNKT) and the non-classical type II NKT (42). Like CD4⁺ T cells, which can be categorized into Th1, Th2 or Th17 based on their cytokine secretion profile when activated, polarization towards distinctive functional subsets has also been described for iNKT cells (43).

iNKT cells can be induced by lipid and glycolipid antigens presented on the MHC Class I-related molecule CD1d, such as alpha-galactosylceramide (α-GalCer), to modulate experimental autoimmune encephalomyelitis (EAE) by secreting mixtures of cytokines typically associated with either pro-inflammatory [such as interferon gamma (IFN-γ)] or anti-inflammatory [such as interleukin 4 (IL-4)] responses (44). Of note, a synthetic glycolipid (OCH), which can induce a predominant production of IL-4 by iNKT cells, has been described, as it is particularly efficient for suppressing EAE via eliciting a Th2-biased cytokine production profile (45). Similarly, using transcriptomic and functional analysis, Carrion et al. have reported that the iNKT cells reactive to a human collagen type II self-peptide (hCII707-721) may constitute a protective iNKT cell subset, which is absent in primary and secondary progressive MS patients (46). On the other hand, iNKT cells also play a prominent role in B cell maturation and activation via CD40-CD40L, CD1d-TCR interaction and IFN-γ/IL-4 production (47). Given the efficiency of selective B cell depletion therapies in treating MS with anti-CD20 monoclonal antibodies (10, 11) and the increasingly recognized roles played by B cells in MS immunopathology (48), it is inconclusive whether iNKT cells may have predominantly protective or detrimental effects in MS.

Unlike in mice, circulating iNKT cells normally constitute less than 0.1% of total lymphocytes in human, which are outnumbered by type II NKT cells (49). It has been shown that type II NKT cells can react with sulfatide, a sulfated glycolipid enriched in the myelin sheaths that insulate nerve fibers, and that type II NKT cells, but not iNKT cells, accumulated in the CNS in the EAE model (50). Consistently, these sulfatide-reactive T cells were also more common in the peripheral blood of MS patients than controls (51), which further support the possibility that type II NKT cells contribute to MS pathogenesis. However, since type II NKT cells are harder to identify in human samples due to a lack of specific surface markers like α- GalCer/CD1d tetramers, they are much less studied than iNKT cells, and their roles in autoimmunity remain to be further validated (52).

Our study is subject to several limitations. First, as noted above, the NKT cells are a heterogenous population, whereas we were unable to distinguish between Type I and Type II NKT cells for our MR analyses, as the cells were gated as CD3⁺CD16⁺/CD56⁺ in their original flow panel, and only CD8 expression level was used to subdivide NKT population (20). In addition, NKT cells are not the only populations that co-express these molecules in blood, with Yδ T cells also sharing this co-expression (53). Therefore, it is necessary to refine cell populations with panels focusing on NKT cells to solidify and expand our findings. Second, given the relatively modest scale of the lymphocyte phenotyping GWAS, our second part of the MR analyses were limited by the availability of the variants that can be effectively used as IVs. Granted, relaxing the association threshold to a p value of < 10^-5, as was adopted by Orru ̀ et al. in their MR analyses (20), would undoubtedly provide more IVs, this would also compromise the first MR assumption (i.e., the variant being robustly associated with the exposure of interest). With the advent of larger scale phenotyping GWAS in future, we expect more SNPs passing GWAS significance threshold that can be reliably used as IVs for MR analysis. Third, although Mediterranean Sardinian population has been widely used for genetic studies to identify causal variants in complex diseases such as MS (54), it is relatively isolated from the mainland European population. Sardinia is colonized by the Neolithic Early European Farmers (EEF) with minor contributions from pre-Neolithic West European hunter-gatherers (WHG), two of the three ancestral populations that contribute ancestry to the present-day European populations (the other being the Ancient North Eurasians, ANE), and thus Sardinian population shares ancestral connection with part of Spanish and French populations (55, 56). Given the indicative causal association between NKT cell count and MS risk presented here, further genetic profiling using diverse European population is recommended. Finally, unlike the first MR assumption, the other two assumptions are not fully testable in practice. Because using variants that influence other traits outside the pathway of interest or that have a direct effect on the target outcome as IVs can potentially distort MR analysis and give false-positive causal inference, we applied several sensitivity tests to evaluate the robustness of the results, including MR-PRESSO and cML-MA-BIC to detect and adjust for horizontal pleiotropic outliers and violation of MR assumptions. By comparing results from different MR methods, we were able to minimize the potential bias resulting from pleiotropic effects, although careful interpretation of the causal inference is still warranted.

This study provides evidence that higher circulating leucocyte and lymphocyte counts increase the risk of MS. In addition, we also found causal association between higher NKT cell count and MS, which underscores the relevance of exploring the functional roles of NKT cells in MS. Given that type II NKT cells are more abundant and have a diverse immunoregulatory roles in human, we have great expectation that progress in understanding this T cell subset may hold promising immunotherapeutic potential.

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding authors.

Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

DH: study design and conceptualization, data interpretation, manuscript drafting for intellectual content. LL: data analysis and interpretation, manuscript revision. DS: data interpretation, manuscript revision. PZ: data interpretation, manuscript revision. LC: study design and conceptualization, data interpretation, manuscript revision. All authors contributed to the article and approved the submitted version.

This work was supported by the Strategic Priority Research Program (Pilot study) “Biological basis of aging and therapeutic strategies” of the Chinese Academy of Sciences (Grant number: XDB39040000).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.