Abstract

   Multiple sclerosis (MS) is a prevalent neurological disorder with a
   complex etiology, often associated with thyroid function. However, the
   causal relationship between these two conditions remains poorly
   understood. This study aimed to elucidate the causal relationship
   between thyroid function and MS using a bidirectional Mendelian
   randomization (MR) approach and to investigate the potential mediating
   role of immune cells. We conducted a two-sample MR analysis using
   summary statistics from large-scale genome-wide association studies
   (GWAS). We included results from sensitivity tests such as MR-Egger,
   weighted median, and leave-one-out analyses to support the robustness
   and reliability of the findings. The inverse variance-weighted (IVW)
   method was the primary approach, with sensitivity analyses conducted
   using seven additional MR methods. Furthermore, multivariable MR and
   mediation analysis were conducted to uncover potential mediating immune
   cells underlying the observed associations. The MR analysis showed that
   Hypothyroidism and elevated Thyroid-Stimulating Hormone (TSH)
   levels(normal) reduced the risk of MS (P = 0.012, OR (95%CI)
   :0.914(0.851, 0.98); P = 0.020, OR (95%CI) :0.88(0.789, 0.98)). Free
   thyroxine (FT4) increased the risk of MS (P = 0.020, OR (95%CI)
   :1.268(1.051, 1.53)). Mediation analysis showed evidence of indirect
   effect of FT4 on MS through “HLA DR on CD33br HLA DR + CD14” and “IgD-
   CD27- %B cell” with a mediated proportion of 39.16% (positive effect),
   78.53% (reverse effect) of the total effect. This study provided
   genetic evidence that FT4 may increase the risk of developing MS. “HLA
   DR on CD33br HLA DR + CD14” and “IgD- CD27- %B cell”, mediated the
   causal relationship between thyroid function and MS, highlighting the
   importance of further investigating their roles in these conditions.

Supplementary Information

   The online version contains supplementary material available at
   10.1038/s41598-025-89888-z.

   Keywords: Thyroid function, Multiple sclerosis, Mendelian
   randomization, Immune cells, Mediation analysis

   Subject terms: Multiple sclerosis, Autoimmunity

Introduction

   Multiple sclerosis (MS) is an autoimmune disorder of the central
   nervous system (CNS), characterized by inflammatory demyelination and
   axonal degeneration^[30]1. MS is a widespread neurological disease with
   varying prevalence across different regions and countries, and it
   brings about increasing healthcare demands and costs^[31]2–[32]4. The
   etiology of MS is multifactorial, involving genetic susceptibility as
   well as environmental factors^[33]5. Early studies demonstrated the
   role of T cells in the development of inflammation and demyelination in
   MS. Recent research has found the possibility that B cells play a
   pathogenic role by regulating T cells^[34]6.

   MS is frequently comorbid with autoimmune diseases, particularly
   thyroid disorders, which underscores the need to clarify their
   interplay^[35]7. Thyroid hormones are crucial factors regulating
   neuroglia generation, with their roles spanning multiple aspects such
   as cell proliferation, determination, differentiation, and cell
   death^[36]8. Additionally, Thyroid hormones play a crucial role in
   myelin formation, a process perturbed in MS^[37]9. However, it remains
   unclear whether there is a causal relationship between thyroid function
   and MS, as previous observational studies may be subject to selection
   bias, reverse causality, and residual confounding.

   TSH, as a key hormone regulating thyroid function, its synthesis and
   secretion are strictly regulated by the hypothalamic-pituitary-thyroid
   axis. In recent years, with the continuous development of GWAS, more
   and more genetic loci have been discovered to be associated with TSH
   levels^[38]10. To address this gap, we conducted a bidirectional
   Mendelian randomization (MR) study leveraging large-scale GWAS data. MR
   is an emerging method that leverages the advancements in large-scale
   GWAS to explore potential causal relationships. MR draws upon Mendel’s
   genetic laws, including the laws of segregation and independent
   assortment^[39]11. By using genetic variants as instrumental variables
   (IVs), this method can establish potential causal relationships between
   variables^[40]12. It circumvents the drawbacks of traditional methods.

   MR leverages the principle of random allocation of alleles at
   conception, thereby mimicking the randomization process of a randomized
   controlled trials (RCTs)^[41]13. By using genetic variants as IVs, we
   reduce confounding and the likelihood of reverse causation. Given the
   logistical and ethical challenges associated with large-scale RCTs on
   thyroid diseases and MS, MR provides a practical and robust alternative
   for inferring causality, offering valuable insight into disease
   mechanisms and potential therapeutic targets.

   Our primary objective was to elucidate the causal relationship between
   thyroid function and MS, with a focus on TSH, FT4, and hypothyroidism
   as exposures, and MS as the outcome. Furthermore, to explore the
   potential mediating role of immune cells in this relationship, we used
   the Multivariable Mendelian Randomization (MVMR) method. This study
   provided genetic evidence using IVs to investigate the causal
   relationship between thyroid function and MS.

Methods

Research design and data sources

   The data on hypothyroidism were sourced from FinnGen R9
   ([42]https://r9.finngen.fi/), encompassing 40,926 cases and 274,069
   controls^[43]14. Details of the data sources and basic information were
   provided in Supplementary file 2: Table [44]S1.

   The data on thyroid function levels within the normal range were
   derived from a large-scale meta-analysis of GWAS targeting thyroid
   function. This study tested up to 8 million genetic variants across as
   many as 72,167 individuals, revealing a total of 109 independent
   genetic variants associated with these traits^[45]15.

   The data on overall TSH levels were obtained from a meta-analysis of
   GWAS involving up to 119,715 individuals and 22.4 million genetic
   markers. This analysis identified 74 significant genome-wide TSH
   loci^[46]16.

   The MS data were sourced from a GWAS study that analyzed a total of
   32,367 MS cases and 36,012 controls across Australia, 10 European
   countries, and multiple states in the United States^[47]17. The study
   utilized a mixed linear model for association analysis implemented in
   GCTA, incorporating genotype principal components of common variants,
   chip type, and gender as covariates in the association analysis.
   Inverse variance-weighted(IVW) meta-analysis was employed to merge
   statistical data across layers. This research identified four novel
   SNPs that may influence MS risk independently of shared genetic
   signals.

   The aggregated summary statistics for 731 immune cell phenotype GWAS
   datasets were obtained from the GWAS catalog (IDs ranging from
   GCST0001391 to GCST0002121)^[48]18. Immune cell phenotypes encompassed
   absolute cell (AC) counts (n = 118), median fluorescence intensity
   (MFI) reflecting surface antigen levels (n = 389), morphological
   parameters [MP] (n = 32), and relative cell (RC) counts (n = 192). MFI,
   AC, and RC features included B cells, CDCs, T cell maturation stages,
   monocytes, bone marrow cells, TBNK, and Treg panels. MP features
   comprised CDC and TBNK panels. The immune trait GWAS was conducted on
   3,757 Europeans, with covariates including sex, age, and age squared.
   Approximately 22 million SNPs genotyped using high-density arrays were
   estimated using a reference panel based on the Sardinian sequence.

IV selection criteria

   To ensure the validity of our analysis, we conducted quality control on
   the SNPs, including the removal of non-biallelic SNPs, SNPs with
   ambiguous strand alleles, SNPs lacking reference SNP ID number (rsID),
   duplicated rsID or base pair positions, SNPs not in the Phase 3 of the
   1000 Genomes Project, SNPs with base pair positions or alleles
   mismatching with those in phase 3 of the 1000 Genomes Project, SNPs
   with imputation information score<0.9, and all SNPs on chromosomes X
   and Y. The number of selected IVs for FT4 (normal) is 23, for
   Hypothyroidism is 116, for TSH (normal) is 40, and for TSH (total) is
   75. Standard MR assessed the causal effect of a single exposure on an
   outcome, whereas Multivariable Mendelian Randomization (MVMR) allowed
   for simultaneous adjustment of multiple exposures, enabling the
   isolation of the independent causal effect of each exposure. This
   distinction led to differences in the selection of IVs: in standard MR,
   IVs were selected solely based on their association with the primary
   exposure, while in MVMR, IVs could be shared across exposures or
   specific to one exposure.

Statistical analysis

   We employed the clumping procedure in the PLINK software to filter out
   all significantly associated genetic loci with the phenotype as IVs
   (P < 5 × 10^^−8), with an r^2<0.001 and a window size of 10,000 kb.
   Some SNPs that were directly associated with the outcomes
   (P < 5 × 10^^−8) were excluded in the MR analysis to ensure the
   validity of the results. The number of index SNPs was 23 for FT4
   (normal), 116 for Hypothyroidism, 40 for TSH (normal), 75 for TSH
   (total), and 74 for MS. For genetic variants to serve as valid IVs, it
   requires that IVs meet three assumptions:^[49]1 IVs should be
   associated with the exposure;^[50]2 IVs should not be associated with
   confounding factors related to both the exposure and the outcome;^[51]3
   the effect of IVs on the outcome is entirely mediated through the
   exposure^[52]19,[53]20. When these assumptions hold, IVW is the most
   powerful method for MR. However, if certain instruments violate the IV
   assumptions, MR analysis may yield erroneous results. We conducted
   several sensitivity analyses.

   Firstly, the Q-test of IVW and MR-Egger was used, which allowed the
   detection of potential violations of assumption by assessing the
   heterogeneity across individual instrumental variables^[54]21.
   Secondly, MR-Egger was applied to estimate horizontal pleiotropy based
   on its intercept, ensuring that genetic variants were independently
   associated with the exposure and outcome. Additional analyses using MR
   methods with different modeling assumptions and strengths (weighted
   median and mode) were employed to enhance the stability and robustness
   of the results^[55]22. Thirdly, MR-PRESSO was used to detect outliers
   and correct for horizontal pleiotropy^[56]23. Fourthly, single SNP
   analysis and leave-one-out analysis were conducted to evaluate the
   possibility of individual SNPs driving the observed associations.

Multivariable mendelian randomization (MVMR)

   MVMR analysis further excluded potential horizontal pleiotropy. MVMR
   analysis utilized the IVW method to assess whether the association
   between multiple exposures and outcome risk was primarily influenced by
   other potential confounders. Effects of other traits were incorporated
   into the linear model based on the following formula, and instrumental
   variables present in all three datasets were retained:

   Inline graphic

   Here, Inline graphic represented the marginal effect size of the
   instrument, σ^2 denoted the variance of the residual term, Inline
   graphic signified the causal effect of exposure 1 on the outcome, and
   Inline graphic represented the causal effect of exposure 2 on the
   outcome. Subsequently, MVMR analysis was conducted to estimate the
   causal effects of multiple exposure factors on the outcome after
   adjusting for confounding factors.

Mediation mendelian randomization

   To explore the potential mediating role of immune cells in this
   relationship, we employed the Multivariable Mendelian Randomization
   (MVMR) method. This mediation effect estimation and hypothesis testing
   were performed in a multivariate MR Framework using aggregated
   association statistics of thyroid markers, immune cells, and
   MS^[57]24–[58]26.

   We employed the IVW method to estimate the causal effect of thyroid
   indicators on 731 immune cell phenotypes and conducted MVMR analysis to
   assess the causal impact of immune cells on MS (adjusting for causal
   thyroid indicators). To estimate the mediation effect, we applied the
   Bootstrap method, which was widely used for assessing the significance
   of indirect effects in mediation analyses. Specifically, we tested the
   null hypothesis (H0) that the product of the indirect path coefficients
   a and b equaled zero (H0: ab = 0). Here, a represented the effect of
   the exposure on the mediator, while b represented the effect of the
   mediator on the outcome, adjusted for the exposure. The Bootstrap
   approach involved resampling the data with replacement to create
   numerous pseudo-samples, from which the distribution of the product ab
   was derived. This allowed for the computation of confidence intervals
   and significance testing of the mediation effect without assuming
   normality of the sampling distribution^[59]27,[60]28.

Enrichment analysis of immune cell-related genes

   To further explore the significant immune cell-related SNPs identified
   in the Mendelian Randomization analysis, we conducted gene mapping and
   functional annotation using FUMA (Functional Mapping and Annotation)
   ([61]https://fuma.ctglab.nl/) and Metascape
   ([62]https://metascape.org/) tools. Lead SNPs associated with
   significant immune cell traits were mapped to their corresponding genes
   based on positional and expression quantitative trait locus (eQTL)
   annotations, with a focus on regulatory variants. Functional annotation
   included an assessment of the expression patterns of mapped genes
   across various tissues, particularly those relevant to the immune and
   CNS. Gene expression profiles were evaluated to identify potential
   tissue-specific roles using data from GTEx
   ([63]https://www.gtexportal.org/) and other publicly available
   datasets. We also performed pathway enrichment and protein-protein
   interaction (PPI) analyses to identify biological processes and
   molecular interactions associated with the mapped genes. Pathway
   enrichment was conducted using Metascape, incorporating Gene Ontology
   (GO)^[64]29, Kyoto Encyclopedia of Genes and Genomes
   (KEGG)^[65]30–[66]32, and Reactome databases, while PPI networks were
   constructed to highlight key functional hubs and potential gene
   interactions underlying the immune cell-mediated effects on MS.

   The statistical analyses were conducted in R version 4.4.1 software. MR
   analysis was performed using the MendelianRandomization package^[67]33,
   while MR-PRESSO utilized the MRPRESSO package.

Result

   First, a bidirectional causal association assessment was conducted. IVs
   are detailed in Supplementary file 2: Tables [68]S1 and [69]S2.
   Hypothyroidism and elevated TSH levels reduced the risk of MS, while
   increased FT4 levels increased the risk of MS. Patients with MS had an
   increased risk of hypothyroidism and elevated TSH levels, but MS did
   not significantly affect FT4 levels (Fig. [70]1). The reference line
   indicated protective factors on the left and risk factors on the right.

Fig. 1.

   [71]Fig. 1
   [72]Open in a new tab

   The forest plot for the causal relationship between thyroid function
   and MS. Left, MS was the outcome; Right, MS was the exposure.

The causal effect of thyroid function on MS

   The causal association determined by the IVW method was robust to other
   MR methods (Fig. [73]2 and Supplementary file 2: Table [74]S4),
   demonstrating the stability of the association. The intercept term of
   MR-Egger method ruled out potential horizontal pleiotropy.
   Hypothyroidism and elevated TSH levels(normal) reduced the risk of MS
   (P = 0.012, OR (95%CI) :0.914(0.851, 0.98); P = 0.020, OR (95%CI)
   :0.88(0.789, 0.98)). FT4 increased the risk of MS (P = 0.020, OR
   (95%CI) :1.268(1.051, 1.53)).

Fig. 2.

   [75]Fig. 2
   [76]Open in a new tab

   The forest plot for the sensitivity analysis of the causal relationship
   between thyroid function and MS.

   However, hypothyroidism and TSH failed to pass the MR-PRESSO global
   test, indicating potential horizontal pleiotropy between the two
   indicators. Only the relationship between FT4 and MS was a stable
   causal relationship. Both scatter plots and funnel plots excluded
   potential outliers and horizontal pleiotropy, further supporting the
   robustness of the study results (Fig. [77]3). LOOCV was shown in
   Supplementary file 2: Table [78]S5.

Fig. 3.

   [79]Fig. 3
   [80]Open in a new tab

   (A) Scatter plot of the causal effect of hypothyroidism on MS; (B) FT4
   on MS; (C) TSH (normal) on MS; (D) TSH (total) on MS; (E) Funnel plot
   of the causal effect of hypothyroidism on MS; (F) FT4 on MS; (G) TSH
   (normal) on MS; (H) TSH (total) on MS.

The causal effect of MS on thyroid function

   MS reduced the risk of Hypothyroidism, TSH levels(normal) and TSH
   (total) (P = 5.67E-06, OR (95%CI): 0.043 (0.025, 0.062); P = 0.019, OR
   (95%CI): 0.016 (0.003, 0.03); P = 0.002, OR (95%CI): 0.015 (0.005,
   0.025)). However, MS did not significantly affect FT4 levels
   (P = 0.990, OR (95% CI): 0 (-0.013, 0.013)).

   The reverse causal association also passed the majority of sensitivity
   analyses (Fig. [81]4 and Supplementary file 2: Table [82]S7); While TSH
   (normal) and TSH (total) did not pass the MR-Egger test, the causal
   association of MS with hypothyroidism also failed to pass the global
   test of MR-PRESSO. Furthermore, scatter plots and funnel plots excluded
   the possibility of potential outliers for the identified causal
   associations (Fig. [83]5). LOOCV was shown in Supplementary file 2:
   Table [84]S8.

Fig. 4.

   [85]Fig. 4
   [86]Open in a new tab

   The forest plot illustrated sensitivity analyses of the causal effect
   of MS on thyroid function.

Fig. 5.

   [87]Fig. 5
   [88]Open in a new tab

   (A) Scatter plot of the causal effect of MS on hypothyroidism; (B) MS
   on FT4; (C) MS on TSH (normal); (D) MS on TSH (total); (E) Funnel plot
   of the causal effect of MS on hypothyroidism; (F) MS on FT4; (G) MS on
   TSH (normal); (H) MS on TSH (total).

MVMR analysis

   We conducted MVMR analysis to correct for potential pleiotropic effects
   among nominally significant thyroid function and MS. The results were
   as follows (Fig. [89]6 and Supplementary file 2: Table [90]S6).

Fig. 6.

   [91]Fig. 6
   [92]Open in a new tab

   MVMR analysis for the causal effect of thyroid function on MS.

Mediation mendelian randomization

   We conducted comprehensive mediation analyses with thyroid function as
   the exposure variable, immune cells as mediators, and MS as the outcome
   variable. The results demonstrated that various immune cells could
   serve as intermediate factors mediating the causal association between
   thyroid function and MS (Table [93]1; Fig. [94]7). Further detailed
   results are presented in Supplementary file 2: Tables [95]S9–[96]S12.
   Additionally, comprehensive mediation analyses were performed with MS
   as the exposure variable, immune cells as mediators, and thyroid
   function as the outcome variable. The results indicated that various
   immune cells could act as intermediate factors mediating the causal
   association between MS and thyroid function (Table [97]1; Fig. [98]7).
   Additional detailed results can be found in Supplementary file 2:
   Tables [99]S13–[100]S16.

Table 1.

   Mediation analysis of immune cell subsets as mediating factors.
   Trait pairs Mediators Total effect Direct effect Indirect effect
   Proportion (%)
   Beta (95% CI) Beta (95% CI) Beta (95% CI)
   FT4 (normal)-MS HLA DR on CD33br HLA DR + CD14- 0.063 (− 0.094, 0.22)
   0.244 (− 0.584, 1.073) − 00.05 (− 0.113, − 0.004) 78.53
   FT4 (normal)-MS IgD- CD27- %B cell 0.063 (− 0.094, 0.22) 0.014
   (− 0.178, 0.207) 0.025 (0.001, 0.057) 39.16
   Hypothyroidism-MS HLA DR on CD33br HLA DR + CD14- − 0.135 (− 0.243,
   − 0.028) − 0.162 (− 0.413, 0.09) − 0.017 (− 0.038, − 0.001) 12.48
   Hypothyroidism-MS IgD- CD27- %B cell − 0.135 (− 0.243, − 0.028) − 0.117
   (− 0.223, − 0.012) − 0.013 (− 0.033, − 0.001) 9.80
   TSH (normal)-MS CD38 on IgD + CD38dim − 0.115 (− 0.263, 0.033) − 0.126
   (− 0.249, − 0.002) 0.012 (0, 0.028) 10.11
   MS-Hypothyroidism HLA DR on DC 0.047 (0.01, 0.084) 0.047 (0.025, 0.069)
   − 0.018 (− 0.033, − 0.007) 38.17
   MS-Hypothyroidism CD28 on CD45RA + CD4+ 0.047 (0.01, 0.084) − 0.021
   (− 0.052, 0.011) 0.007 (0.001, 0.015) 14.94
   MS-Hypothyroidism HLA DR on myeloid DC 0.047 (0.01, 0.084) − 0.001
   (− 0.026, 0.024) − 0.004 (− 0.009, − 0.001) 9.21
   MS-Hypothyroidism CD28 on resting Treg 0.047 (0.01, 0.084) 0.041
   (0.009, 0.072) 0.006 (0, 0.013) 12.86
   MS-TSH (normal) CD19 on IgD- CD27- 0.003 (− 0.008, 0.015) 0.001
   (− 0.01, 0.013) 0.002 (0, 0.004) 56.67
   MS-TSH (normal) CD19 on IgD- CD24- 0.003 (− 0.008, 0.015) 0.002
   (− 0.009, 0.013) 0.001 (0, 0.004) 41.68
   MS-TSH (total) CD19 on IgD- CD24- 0.007 (− 0.003, 0.017) 0.005
   (− 0.004, 0.014) 0.001 (0, 0.003) 15.48
   MS-TSH (total) CD19 on IgD- CD27- 0.007 (− 0.003, 0.017) 0.006
   (− 0.004, 0.015) 0.001 (0, 0.003) 16.86
   MS-TSH (normal) CD19 on IgD + CD38dim 0.003 (− 0.008, 0.015) 0.002
   (− 0.011, 0.014) 0.001 (0, 0.003) 37.53
   MS-TSH (normal) CD28 + CD45RA + CD8dim %CD8dim 0.003 (− 0.008, 0.015)
   0.003 (− 0.008, 0.015) 0.001 (0, 0.002) 25.94
   MS-TSH (normal) CD19 on sw mem 0.003 (− 0.008, 0.015) 0.002 (− 0.01,
   0.014) 0.001 (0, 0.003) 38.07
   MS-TSH (normal) Activated & resting Treg %CD4+ 0.003 (− 0.008, 0.015)
   − 0.002 (− 0.013, 0.01) − 0.001 (− 0.003, 0) 32.50
   MS-TSH (normal) CD33 on CD33br HLA DR + CD14- 0.003 (− 0.008, 0.015)
   0.004 (− 0.008, 0.015) − 0.001 (− 0.002, 0) 20.45
   MS-TSH (normal) TD CD8br %CD8br 0.003 (− 0.008, 0.015) 0.004 (− 0.008,
   0.015) − 0.001 (− 0.002, 0) 28.25
   [101]Open in a new tab

Fig. 7.

   [102]Fig. 7
   [103]Open in a new tab

   The immune cell subgroups that mediated the causal association between
   thyroid indicators and MS.

   We conducted a comprehensive mediation analysis utilizing immune cells
   as mediators. As depicted in Fig. [104]7, several immune cell subsets
   emerged as significant contributors to the observed causal
   associations. Key immune cells, such as CD27, CD28, CD19, CD33, and
   CD4, played pivotal roles in the association between thyroid function
   and MS. Specifically, our mediation analysis showed that “HLA DR on
   CD33br HLA DR + CD14-” and “IgD- CD27- %B cell” were key mediators of
   the effects of FT4 and hypothyroidism on MS risk. For FT4, a positive
   mediating effect of 39.16% was observed through “HLA DR on CD33br HLA
   DR + CD14-”, suggesting that elevated FT4 levels may increase MS risk
   via this immune cell pathway. Conversely, FT4 exerted a reverse
   mediating effect of 78.53% through “IgD- CD27- %B cell”, indicating a
   potential protective role of this B cell subset against MS in the
   context of altered FT4 levels.

Enrichment analysis of immune cell-related genes

   Supplementary file 1: Fig. [105]S1 depicted the expression profile of
   immune cell-related genes, providing a comprehensive view of their
   activities across 54 common tissues. It highlighted the specific
   expression patterns of these genes in tissues, with particular
   attention to those associated with the immune system and the CNS
   (Supplementary file 1: Fig. [106]S1).

   Supplementary file 1: Fig. [107]S2 presented the results of pathway
   enrichment analysis. The figure displayed a network of enriched
   pathways. The main roles of the pathway primarily included “positive
   regulation of the immune system”, “regulation of lymphocyte
   proliferation”, and “antigen processing and presentation”
   (Supplementary file 1: Fig. [108]S2).

   Supplementary file 1: Fig. [109]S3 exhibited the PPI network
   constructed for immune cell-related genes. Nodes represented proteins
   encoded by the identified genes, and edges denoted physical
   interactions between them. Prominent hub proteins in the network had
   high interaction counts with other proteins (Supplementary file 1: Fig.
   [110]S3).

Discussion

   In the present large-scale MR study, we used a bidirectional MR method
   and leveraged large-scale GWAS data to explore the causal relationship
   between thyroid function and MS, as well as to conduct a detailed
   analysis of the mediating role of immune cells in this relationship. MS
   is a chronic disease of the CNS, believed to be immune-mediated. It has
   been reported that there is an association between thyroid function and
   the prevalence and incidence of MS^[111]34. The association between
   autoimmune thyroid diseases and other autoimmune diseases emphasizes
   potential shared pathogenic mechanisms among these conditions^[112]35.
   Although the association between the two has been confirmed by multiple
   epidemiological studies, the causality and directionality have remained
   unclear. Our study showed that FT4 may be associated with an increased
   risk of developing MS.

   Patients with thyroiditis exhibit a significant increase in the
   incidence of MS^[113]36. Thyroid hormones are crucial for normal myelin
   formation, and they play an important role in regulating neurogenesis
   in the adult mammalian brain^[114]37. Elevated FT4 levels may disrupt
   normal myelin maintenance, thereby contributing to the onset of MS. It
   has been reported that normalizing thyroid hormone levels can reverse
   CNS symptoms, suggesting a direct causal relationship between altered
   thyroid hormone levels and CNS symptoms. Autoimmune reactions against
   antigens present in the thyroid may spread to the central nervous
   system^[115]38.

   The prevalence of thyroid diseases has increased among patients with MS
   and their relatives. Henderson et al. suggested that clinicians should
   examine patients who have not undergone recent thyroid function testing
   but present with unspecified symptoms to ascertain if they have thyroid
   dysfunction^[116]39. Our research emphasized the importance of
   monitoring FT4 levels in assessing the risk of MS. Early detection of
   FT4 abnormalities and the implementation of intervention measures may
   contribute to reducing the risk of developing MS. Further studies are
   warranted to delve deeper into the specific mechanisms by which FT4
   influences MS, particularly its regulatory role on immune function and
   its potential to promote the development of MS.

   Hypothyroidism caused by autoimmune thyroid disease is more severe and
   more common in MS patients^[117]40. Unfortunately, definitive results
   were not obtained in the Mendelian validation of the causal
   relationship between hypothyroidism and TSH factors with MS, and
   further research is needed to confirm this. We recognized that the
   effect of FT4 was no longer significant in our multivariable MR model
   after adjusting for TSH and hypothyroidism. Once FT4, hypothyroidism,
   and TSH were considered together in the multivariable model, the
   previously significant association for FT4 became attenuated, with its
   95% CI crossing the reference line. This indicated that the apparent
   effect of FT4 might have been partly explained by overlapping genetic
   influences or mediation through TSH, underscoring the complex interplay
   among these thyroid-related traits in MS risk.

   According to our results, TSH and hypothyroidism can decrease the risk
   of MS, while MS can increase the level of TSH and the risk of
   hypothyroidism. These results likely reflected differences in the
   direction and magnitude of causal effects in the bidirectional MR
   framework, possibly due to shared genetic pathways or pleiotropy. The
   observed associations could have been explained by the complex
   interactions between the immune and endocrine systems. Hypothyroidism
   and elevated TSH levels might have exerted immunosuppressive effects,
   potentially dampening autoimmune responses, including those involved in
   MS pathogenesis.

   The pathophysiology of MS appears to involve aberrant immune
   activation, leading to neuroinflammation involving both peripheral
   immune cells and resident cells of the central nervous system such as
   microglia and astrocytes^[118]41. Large-scale studies have uncovered
   multiple independent genetic variations that elevate the risk of MS.
   These variations are likely to play a functional role in the
   pathogenesis of MS, involving both adaptive and innate immune cells.
   Moreover, some of these MS-associated variations also mediate the risk
   of developing other autoimmune diseases^[119]42.

   We conducted gene enrichment analysis on genes identified through
   instrumental variable mapping and found that human leukocyte antigen
   (HLA)-related genes played an important role in the mediating effect
   between thyroid function and MS. Pathway analysis revealed that the
   major pathways of mediator immune cell-related genes in 54 tissues
   primarily included “positive regulation of the immune system”,
   “regulation of lymphocyte proliferation” and “antigen processing and
   presentation”. The effects of thyroid hormones are mainly mediated by
   thyroid-stimulating hormone receptor (TSHR)^[120]43. HLA-DR3 has a high
   affinity for the immunodominant peptide of the TSHR, which may have
   affected the binding and presentation of TSHR peptides, thereby
   influencing the regulation of thyroid function^[121]44. This is
   consistent with our findings. Among all the genetic factors associated
   with MS susceptibility, the association between the HLA and MS is the
   strongest^[122]45. Although the direct role of HLA alleles in MS has
   been confirmed, it is still difficult to understand the role of HLA-DQ
   alleles in the disease pathogenesis due to strong linkage
   disequilibrium^[123]46.

   Concerning the immune cell-mediated effects, we identified key immune
   cells such as CD27, CD28, CD19, CD33, and CD4 as pivotal players in the
   thyroid function-MS association. Specifically, “HLA DR on CD33br HLA
   DR + CD14-” and “IgD- CD27- %B cell” are identified as the primary
   mediators influencing the risk of MS associated with FT4 and
   hypothyroidism. FT4 exerts a 39.16% positive mediating effect on MS
   through “HLA DR on CD33br HLA DR + CD14”, while it has a 78.53%
   negative mediating effect through “IgD- CD27- %B cell”. In previous
   literature descriptions, localized infiltration of T lymphocytes and
   macrophages and death of oligodendrocytes, resulting in multiple small
   foci of inflammation, are the primary cause of myelin destruction,
   leading to the formation of lesions of inflammatory cells and their
   products, demyelination, and broken axons in the CNS^[124]47. This
   laterally confirms that variations in FT4 levels may regulate the risk
   of MS by affecting the activity and function of specific immune cell
   subsets.

   CD27 is associated with T-cell activation and memory cells, and
   research suggests that soluble CD27 in the cerebrospinal fluid serves
   as a biomarker to aid in the detection of early MS^[125]48. In our
   mediation analysis, FT4 mediated the protective effect of IgD- CD27- %B
   cells on the occurrence of MS, which is consistent with the conclusion
   of this literature. Many other studies have also emphasized the
   important role of other immune cells in the occurrence of MS. Research
   on both human MS and rodent models suggests that CD28 blockade can
   serve as a therapeutic approach for treating autoimmune
   demyelination^[126]49. MS is considered to be an autoimmune disease
   triggered by CD4 T cells, associated with specific Th cells targeting
   myelin antigens^[127]50. Monoclonal antibodies targeting CD20 and CD19
   hold potential advantages in MS therapy^[128]51. The neuroinflammatory
   protein CD33 is situated at the core of neuroinflammatory events
   associated with various neurodegenerative diseases and is linked to
   several neurodegenerative conditions, including MS^[129]52.

   These cells encompass diverse functions related to T-cell activation,
   B-cell function, and neuroinflammation, all intimately linked to the
   immunopathogenesis of MS. Our study not only provides genetic evidence
   for the causal relationship between thyroid function and MS through
   mediation analysis but also deeply reveals the key roles of specific
   immune cells in this process. These findings offer a new perspective
   for understanding the complex connection between thyroid diseases and
   autoimmune diseases and provide potential targets for future
   therapeutic interventions against these diseases. Future investigations
   should explore how these immune cells modulate MS onset and progression
   in the context of thyroid dysfunction, and how targeting these cells
   could lead to innovative MS interventions.

   This study exhibits several notable methodological strengths. It
   employs a bidirectional two-sample MR design based on GWAS data, which
   is a robust approach for investigating causal relationships.
   Furthermore, the study incorporates mediation MR analysis, enhancing
   its ability to unravel the underlying mechanisms. By utilizing various
   MR techniques, the study minimizes confounding factors and reduces bias
   in causal estimation, thereby ensuring the reliability of its findings.
   The application of MR methods in this study provides genetic evidence
   of a causal association between FT4 and MS. Notably, the study
   highlights the significant roles of specific immune cells, namely “HLA
   DR on CD33br HLA DR + CD14” and “IgD- CD27- %B cell”, in this causal
   relationship. Sensitivity analyses do not reveal heterogeneity,
   confirming the validity and robustness of the study’s findings. This
   underscores the importance of these immune cells in the pathogenesis of
   MS and suggests potential targets for future therapeutic interventions.

   However, there are also limitations to consider. Firstly, the majority
   of GWAS data originate from European populations, thus caution is
   warranted when extrapolating the results to other ethnic groups.
   Secondly, as mentioned earlier, thyroid function and MS are influenced
   by a combination of genetic variation and environmental factors,
   indicating that the study’s findings can only partially elucidate the
   causal impact of thyroid function on MS. Our bidirectional MR framework
   provides valuable initial insights, but the interpretation of binary
   exposures for complex traits such as hypothyroidism and MS must be
   approached with caution, as this approach may obscure underlying
   continuous or polygenic processes^[130]53. Future studies employing
   more refined phenotypic definitions or continuous indicators may better
   capture the nuances of how thyroid function influences MS risk.

Conclusion

   We have presented novel evidence suggesting that FT4 may serve as a
   causal determinant for the risk of developing MS. Furthermore, our
   findings reveal that specific immune cells, namely “HLA DR on CD33br
   HLA DR + CD14” and “IgD- CD27- %B cell”, play pivotal roles in both
   thyroid function and MS pathogenesis. These research outcomes imply an
   intricate interplay between MS and thyroid function, underscoring the
   necessity for further investigation to elucidate potential shared
   mechanisms underlying these two conditions.

Electronic supplementary material

   Below is the link to the electronic supplementary material.
   [131]Supplementary Material 1^ (4.2MB, zip)
   [132]Supplementary Material 2^ (1.2MB, xlsx)

Author contributions

   Y.R. and W.W. conceptualized the study, curated the data, and reviewed
   and edited the manuscript. X.W. handled the software and wrote the
   original draft. Z.W. contributed to the conceptualization, methodology,
   and reviewed and edited the manuscript. All authors reviewed the
   manuscript.

Funding

   None.

Data availability

   The data utilized in this paper were obtained from publicly available
   databases that comply with ethical standards and legal requirements.
   The data of this study can be found in the article or supplementary
   materials. To obtain the raw data from this study, please contact the
   corresponding author directly.

Declarations

Competing interests

   The authors declare no competing interests.

Ethical approval

   The data used in this paper were sourced from publicly accessible
   databases that adhere to ethical norms and legal regulations.

Footnotes

   Publisher’s note

   Springer Nature remains neutral with regard to jurisdictional claims in
   published maps and institutional affiliations.

   Yinghao Ren and Xin Wang have contributed equally to this work and
   share first authorship.

Contributor Information

   Weiliang Wang, Email: 983044268@qq.com.

   Zeyu Wang, Email: 850350243@qq.com.

References