Inflammation-metabolite crosstalk in multiple sclerosis: A mediation Mendelian randomization study of plasma inflammatory proteins,iron,and serum/cerebrospinal fluid metabolites

Abstract

Objective

To investigate the causal relationships between inflammatory proteins, iron metabolism, blood/CSF metabolites, and multiple sclerosis (MS) risk using genetic evidence.

Methods

We performed a two-sample, two-step Mendelian randomization (MR) analysis using European-ancestry genome-wide association study data. The exposures comprised 91 inflammatory proteins, while potential mediators included 1091 blood metabolites, 309 metabolite ratios, 233 circulating metabolic traits, and 338 cerebrospinal fluid metabolites. For the outcome, we assessed MS risk using two independent datasets: International Multiple Sclerosis Genetics Consortium (IMSGC) and UK Biobank. Our primary analysis utilized inverse-variance weighted regression. To ensure robust results, we conducted comprehensive sensitivity analyses including MR-Egger, weighted median, MR-PRESSO, and Bayesian Weighted MR approaches to evaluate potential pleiotropy and strengthen causal inference.

Results

We observed a statistically significant but modest elevation in MS risk associated with interleukin-7 (IL-7; OR = 1.40, 95% CI: 1.07–1.83, p = 0.016) in the IMSGC cohort, with a weaker effect in the UK Biobank (OR = 1.001, 95% CI: 1.000–1.002, p = 0.047). The IL-7 was causally linked to six blood metabolic traits (taurocholenate sulfate, anthranilate, taurodeoxycholate, albumin, sphingomyelin (d18:1/24:1, d18:2/24:0), leucine-to-phosphate ratio), all influencing MS risk. No significant interactions between iron metabolism and inflammatory proteins were found.

Conclusions

This MR study establishes IL-7 as a potential causal risk factor for MS, partially mediated by blood metabolites. The findings prioritize IL-7 and associated metabolic pathways (bile acids/kynurenine) for therapeutic targeting.

Keywords

IL-7 multiple sclerosis cytokine iron metabolism Mendelian randomization metabolite

Introduction

Inflammatory proteins, especially cytokines, play a pivotal and multifaceted role in the pathogenesis of multiple sclerosis (MS), orchestrating immune responses that drive neuroinflammation, demyelination, and neurodegeneration. Pro-inflammatory cytokines such as TNF-α, interleukin (IL)-1β, IL-6, IL-12, IL-17, IL-23, and IFN-γ promote pathogenic Th1/Th17 responses, blood–brain barrier disruption, and excitotoxic damage, while anti-inflammatory cytokines such as IL-4, IL-10, and TGF-β exert regulatory and neuroprotective functions.^1–6 Previous research has highlighted the IL-6/IL-23-Th17 axis, IL-7 signaling, and B cell-derived cytokine imbalances (e.g. elevated IL-6 and reduced IL-10) as key drivers of disease progression.^1,3–5 Although therapies targeting cytokines, such as IFN-β and glatiramer acetate, show immunomodulatory benefits by altering IL-7 and IL-10 levels,^3,5 other interventions (e.g. TNF-α or IL-12/23 blockade) have failed due to cytokine pleiotropy and context-specific effects.¹ Emerging investigations are examining the potential roles of dietary bioactive compounds, cytokine gene polymorphisms, and cytokine-mediated synaptic modulation, particularly involving TNF-α and IL-1β signaling pathways, in the inflammatory processes underlying neurodegenerative pathogenesis.^2,6,8 Overall, understanding the complex cytokine network is essential for developing targeted and personalized MS therapies.

Iron metabolism is crucial in the pathogenesis of MS, affecting both CNS function and neurodegeneration. Inflammatory responses can cause astrocytes to retain excess iron, worsening neuroinflammatory damage.⁷ Furthermore, paramagnetic rim lesions, which are associated with persistent inflammation and disease progression, have been observed in MS patients.^8,9 Pro-inflammatory cytokines have also been shown to increase iron uptake by macrophages, further exacerbating CNS iron dysregulation.^10,11 These findings indicate a complex interplay between inflammatory cytokines and iron metabolism in the pathology of MS.

Altered metabolomic profiles are believed to influence immuno-inflammatory responses that play a crucial role in the progression of MS.¹² Recent advancements in this area offer significant opportunities for developing targeted therapeutic interventions.¹³ Mendelian randomization (MR) studies have identified certain metabolites, such as elevated serum serine levels, that are associated with increased disability progression in MS.¹⁰ Additionally, higher levels of sphingolipids, including sphingosines, sphingomyelins, and ceramides, have been detected in MS lesions and are associated with immune cell infiltration, further implicating these metabolites in the disease's pathology.¹⁴ Furthermore, metabolomic analyses of CSF from MS patients have revealed increased levels of bile acid precursors and 25-hydroxycholesterol, indicating underlying demyelination and neuronal damage.¹⁵ However, current metabolomic studies often encounter challenges, such as small sample sizes and limited metabolite coverage.¹⁶ Despite these advancements, the specific interactions between inflammatory factors and metabolites in MS remain inadequately understood.

In this study, we employed a bidirectional two-sample MR analysis to investigate causal relationships between 91 inflammatory proteins (including cytokines, chemokines, growth factors, and other inflammatory mediators) and MS. By utilizing large-scale genome-wide association study (GWAS) datasets, we aimed to uncover potential causal links between these inflammatory traits and the risk of developing MS. Subsequently, a two-step MR analysis was conducted to examine the impact of inflammatory proteins on metabolic traits, including iron metabolism and a diverse array of blood and CSF metabolites. Our findings support the hypothesis that inflammatory proteins may influence specific metabolite levels, potentially mediating the pathogenesis of MS.

Methods

Study design

This study employed a mediation MR approach to explore the causal relationships between plasma inflammatory proteins, metabolic traits, and MS using publicly available GWAS summary data (Figure 1). Analyses proceeded in three stages: (1) assessing effects of 91 inflammatory proteins on MS; (2) identifying MS-associated metabolic traits; and (3) quantifying mediation effects of metabolites in protein-MS pathways. Ethical approvals were obtained for all original GWAS.

Figure 1.

Flowchart.

Data source

For our data source, we utilized genome-wide summary-level statistics for single-nucleotide polymorphisms (SNPs) from publicly accessible GWAS datasets, with detailed descriptions of these datasets available in Table S1.

Exposures

Regarding exposures, we relied on genetic data from the largest GWAS meta-analysis conducted by the SCALLOP Consortium,¹⁷ which included data from 14,824 participants of European ancestry across 11 cohorts, as outlined in Table S2.

Outcomes

The primary outcome, MS, was examined using data from two large-scale GWAS datasets: the discovery phase (GWAS) of International Multiple Sclerosis Genetics Consortium (IMSGC), which comprised 41,505 individuals (14,802 cases and 26,703 controls) of European ancestry,¹⁸ and the UK Biobank cohort, which included 396,565 participants¹⁹ (1356 cases and 395,209 controls). These two MS GWAS datasets provide extensive genotype–phenotype data with rigorous quality control. The GWAS summary statistics for these datasets were sourced from their respective repositories, specifically the IMSGC at https://gwas.mrcieu.ac.uk/datasets/ and the UK Biobank at https://www.leelabsg.org/resources.

Mediators

Summary statistics for metabolic traits were obtained from multiple sources. Iron metabolism data were sourced from FinnGen (R10 release, phenocode: finngen_R10_E4_ IRON_MET), which included 410,970 individuals. Data for 1091 blood metabolites and 309 metabolite ratios were sourced from CLSA cohort with 39,656 Canadians²⁰ (Table S3). Additional data for 233 circulating metabolic traits came from research involving 136,016 participants across 33 cohorts²¹ (Table S4). Furthermore, data for 338 CSF metabolites were derived from a study with 689 participants²² (Table S5). The selection of mediators was based on several criteria: first, there needed to be a causal association between inflammatory plasma proteins and the mediator with unidirectional effects; second, we required consistent directional associations throughout the mediation pathway (from inflammatory proteins to the mediator and subsequently to MS), with the mediator-MS link remaining robust regardless of inflammatory protein adjustment; and third, the associations between inflammatory plasma proteins and the mediator, as well as between the mediator and MS, had to align in direction. Ultimately, six MS risk factors that fulfilled these criteria were included in the mediation analyses.

Genetic instrument selection

Genetic instruments were selected based on three key criteria: a strong association with the exposure (P < 1 × 10⁻⁵),^23,24 no association with confounding factors, and no direct effect on the outcome independent of the exposure.²⁵ Variants with a minor allele frequency below 0.01 were excluded, and linkage disequilibrium (LD) was controlled using an r² threshold of <0.001 and a distance threshold of 10,000 kb.^23–25 Instruments with F statistics ≤ 10, palindromic alleles (A/T or C/G), or missing outcome data were excluded, as were inflammatory plasma proteins and metabolites associated with fewer than three SNPs to ensure robustness.^23,24 Additionally, beta values were aligned to maintain consistent effect directions of SNPs on exposure and outcome.²⁵

Removing confounders

To minimize confounding in our analysis, we identified and removed SNPs significantly associated with potential confounders using the LDlink database (https://ldlink.nih.gov/?tab = ldtrait) specifically for European populations. We considered four potential confounding factors: immune-related traits, such as IL-6, TNF-α, and C-reactive protein levels; vitamin D metabolism²⁶; Epstein-Barr virus infection²⁷; and autoimmune diseases, including rheumatoid arthritis,²⁷ type 1 diabetes,²⁸ and inflammatory bowel disease.¹²

Based on these criteria, four SNPs were excluded from the analysis: rs4738684 and rs1047891, both associated with taurocholenate sulfate levels and involved in vitamin D metabolism; rs11591147, linked to sphingomyelin and also relevant to vitamin D metabolism; and rs1260326, associated with albumin, as well as vitamin D metabolism and inflammatory bowel disease.

Statistical analysis

Two-sample MR

We conducted a bidirectional two-sample MR analysis to explore the causal relationships between inflammatory plasma proteins and metabolites, to determine if metabolites might influence inflammatory plasma proteins. In the reverse direction, genetic instruments for MS were obtained from the IMSGC consortium and clumped (r² < 0.001, kb = 10,000) to ensure independence. These SNPs were then mapped to IL-7, with harmonization ensuring allele alignment.

The inverse-variance weighted (IVW) method, utilizing a multiplicative random-effects model, served as the primary analytical approach for causal inference.²⁹ Prior to MR analysis, we assessed the strength of genetic instruments to ensure their validity. For each exposure-associated SNP, we calculated the proportion of variance explained (R²) in the exposure using the following formula:

R^{2} = \frac{2 \times β^{2} \times MAF \times (1 - MAF)}{2 \times β^{2} \times MAF \times (1 - MAF) + 2 \times N \times MAF \times (1 - MAF) \times S E^{2}},

where β represents the SNP-exposure effect size, MAF is the minor allele frequency, N is the sample size of the exposure GWAS, and SE is the standard error of the SNP-exposure association. We then computed the F-statistic to evaluate instrument strength:

F = \frac{R^{2} \times (N - 2)}{1 - R^{2}},

Only SNPs with F > 10 were retained to minimize weak instrument bias.³⁰

Sensitivity analysis

We assessed robustness and pleiotropy using multiple methods: (1) MR-Egger regression to detect directional pleiotropy (intercept p < 0.05)^31,32; (2) MR-PRESSO to identify outlier variants³¹; (3) Cochran's Q statistic for heterogeneity (p < 0.05)³³; (4) leave-one-out analysis to evaluate single-SNP influence; and (5) Bayesian Weighted MR (BWMR) to model pleiotropy probabilistically.³⁴

Mediation analysis

We employed the product-of-coefficients mediation MR approach:

1. Total effect estimation: IVW-MR derived the overall effect of inflammatory proteins on MS (Total effect β).

2. Mediation decomposition:

Exposure-mediator effects (Direct A β): IVW-MR estimated protein-metabolite associations.

Mediator-outcome effects (Direct B β): IVW-MR assessed metabolite-MS effects.

3. Mediated effect (β): Calculated as product of coefficients (Direct A β × Direct B β), with variance estimated via delta method:

SE (β_{med}) = \sqrt{β_{A}^{2} \times SE (β_{B})^{2} + β_{B}^{2} \times SE (β_{A})^{2}}

Proportion mediated: β_med/β_total

Statistical significance was evaluated using Wald tests (Z = β/SE) and 95% CIs.^35,36

Analytical tools

Data analysis was performed using R statistical software (v4.4.0), specifically employing the TwoSample MR package (v 0.6.2) for MR analyses, MRPRESSO (v 1.0) for testing pleiotropy, and ggplot2 (v 3.5.2) for visualizing the data.

Results

Causal effects of inflammatory plasma proteins on MS

A two-sample MR analysis was conducted to assess the causal effects of inflammatory plasma proteins on MS. The IVW analysis revealed that four inflammatory proteins were significantly associated with MS in the IMSGC cohort, while six were identified in the UK Biobank cohort, with a significance level of p < 0.05 (Table S6). Notably, IL-7 demonstrated a statistically significant but modest elevation of MS risk in both cohorts, with an IVW-derived OR of 1.40 (95% CI: 1.07–1.83, p = 0.016) in the IMSGC cohort and an OR of 1.001 (95% CI: 1.000–1.002, p = 0.047) in the UK Biobank cohort (Table 1). These findings were supported by MR-PRESSO and weighted median analyses in the IMSGC cohort, as well as MR-PRESSO results in the UK Biobank cohort. Consistent with primary IVW results, IL-7 exhibited modest causal evidence for MS in both IMSGC and UK Biobank (p < 0.05) with BWMR (Table 1). Among the metabolites analyzed, taurocholenate sulfate (OR = 1.16, 95% CI: 1.00–1.35, p = 0.047), anthranilate (OR = 1.19, 95% CI: 1.03–1.38, p = 0.020), and sphingomyelin (d18:1/24:1, d18:2/24:0) (OR = 1.001, 95% CI: 1.000–1.002, p = 0.028) demonstrated nominally significant associations with MS risk. In contrast, no significant causal effects were observed for taurodeoxycholate (TDC; OR = 1.24, 95% CI: 0.96–1.60, p = 0.104), leucine-to-phosphate ratio (OR = 0.91, 95% CI: 0.77–1.08, p = 0.27), or albumin (OR = 1.39, 95% CI: 0.92–2.09, p = 0.12) (Figure 2, Table S7).

Figure 2.

Bidirectional Mendelian randomization analysis of interleukin-7 (IL-7) and multiple sclerosis. The forest plot shows Mendelian randomization estimates (inverse-variance weighted [IVW], weighted median, Mendelian randomization [MR]-Egger, MR-PRESSO, and Bayesian Weighted MR [BWMR]) for the six metabolites that demonstrated consistent causal associations with both IL-7 and multiple sclerosis (MS).

Table 1.

Mendelian randomization estimates of the effect of IL-7 on MS in independent replication populations.

Exposure	Outcome	Method	Number of SNP	F-statistic	OR (95%CI)	p-Value	Heterogeneity	P_h	Egger intercept	P_intercept
IL-7	MS (IMSGC)	iVW	9	21.92	1.40 (1.07–1.83)	0.016	14.954	0.037	0.019	0.698
		MR Egger	9		1.12 (0.37–3.38)	0.846	15.302	0.054
		Weighted median	9		1.43 (1.07–1.92)	0.016
		PRESSO	9		1.40 (1.31–1.49)	0.042
		BWMR	9		1.40 (1.05–1.86)	0.020
IL-7	MS (UK biobank)	iVW	23	21.76	1.001 (1.000–1.002)	0.047	16.411	0.795	<0.001	0.579
		MR Egger	23		1.000 (0.998–1.003)	0.814	16.093	0.764
		Weighted median	23		1.001 (0.999–1.004)	0.173
		PRESSO	23		1.001 (1.001–1.001)	0.032
		BWMR	23		1.001 (1.000–1.002)	0.047

BWMR: Bayesian Weighted MR; IL-7: interleukin-7; IMSGC: International Multiple Sclerosis Genetics Consortium; MR: Mendelian randomization; MS: multiple sclerosis; SNP: single nucleotide polymorphism.

Scatter plots depicting the effects of SNPs on IL-7 and MS risk across both cohorts are illustrated in Figure 3, alongside forest plots (Figure 2) that present the individual and combined MR-estimated effect sizes. A total of 9 SNPs in the IMSGC cohort and 23 SNPs in the UK Biobank cohort were significantly associated with MS, as detailed in Table S8 for replication purposes. F-statistics indicated no weak instrument bias (see Table S9), and sensitivity analyses revealed no evidence of pleiotropy (Table S10). Furthermore, leave-one-out analyses showed that no single SNP had a disproportionate impact on the results (Figure 4). The funnel plots of SNP distribution were symmetric, indicating minimal bias in the causal estimates (Figure S1).

Figure 3.

Mendelian randomization results for effects of interleukin-7 (IL-7) on multiple sclerosis (MS). (a) Scatterplot for the effect of IL-7 on MS in cohort of International Multiple Sclerosis Genetics Consortium (IMSGC), with the slope of each line corresponding to estimated Mendelian randomization effect per method. Data are expressed as raw β values with 95% CI. (b) Forest plot of individual and combined single nucleotide polymorphism (SNP); Mendelian randomization (MR)-estimated effect sizes for IL-7 on MS in cohort of IMSGC. Scatterplot (c) and Forest plot (d) show the effects of IL-7 on MS in cohort of UK biobank.

Figure 4.

Mendelian randomization sensitivity analysis for the effect of interleukin-7 (IL-7) on multiple sclerosis (MS). Leave-one-out sensitivity analysis for the effect of IL-7 on MS in (a) the International Multiple Sclerosis Genetics Consortium (IMSGC) cohort and (b) the UK biobank.

Causal effects of MS on IL-7

Reverse MR analysis showed no significant causal effect of MS on IL-7 levels (p > 0.05), supporting unidirectional causality as detailed in Table 2 and Figure 2. These results were consistent across various MR methods, and the F-statistics indicated no weak instrument bias (refer to Table 2). Additionally, sensitivity analyses did not reveal any pleiotropic effects, as illustrated in Figure S2 and Table S11.

Table 2.

Bidirectional MR results for the relationship between MS and IL-7.

Exposure	Outcome	Method	Number of SNP	F-statistic	p-Value	OR (95%CI)
MS (IMSGC)	IL-7	MR Egger	63			1.00 (0.95–1.04)
		Weighted median	63		0.84	0.98 (0.95–1.01)
		iVW	63	88.88	0.18	1.00 (0.98–1.03)
		Simple mode	63		0.85	0.98 (0.92–1.05)
		Weighted mode	63		0.61	0.98 (0.95–1.02)
MS (UK biobank)	IL-7	MR Egger	117	23.21	0.36	0.00 (0.00–1.22)
		Weighted median	117			0.03 (0.00–12.42)
		iVW	117		0.06	0.33 (0.01–8.08)
		Simple mode	117		0.25	0.51 (0.00–82886.61)
		Weighted mode	117		0.50	0.03 (0.00–39.13)

IL-7: interleukin-7; IMSGC: International Multiple Sclerosis Genetics Consortium; MR: Mendelian randomization; MS: multiple sclerosis; SNP: single nucleotide polymorphism.

Mediation analyses

Two-step MR revealed that iron metabolism was potentially associated with increased MS risk (IMSGC: OR = 1.034, 95% CI: 1.01–1.06, p = 0.012; UK Biobank: OR = 1.000, 95% CI: 1.00–1.00, p = 0.066), with 92 and 82 metabolites linked to MS in each cohort, respectively (Table S12). For IL-7, six metabolites mediated its effect on MS, including taurocholenate sulfate, anthranilate, TDC, leucine:phosphate ratio, albumin (IMSGC), and sphingomyelin (d18:1/24:1, d18:2/24:0) (UK Biobank) (Figure 5, S3–S6). Sensitivity analyses showed no evidence of pleiotropy (Table S13, Figure S7); SNP details are provided in Tables S14–S20.

Figure 5.

Causal effects of interleukin-7 (IL-7)-associated metabolites on multiple sclerosis (MS) risk. The forest plot shows Mendelian randomization estimates (inverse-variance weighted [IVW], weighted median, Mendelian randomization [MR]-Egger, MR-PRESSO, and Bayesian Weighted MR [BWMR]) for the six metabolites that demonstrated consistent causal associations with both IL-7 and MS.

Finally, the mediation analysis revealed that IL-7 has an indirect effect on MS by regulating specific serum metabolites (Table S21), while it does not influence CSF metabolites. Each metabolite's mediation effects and their respective proportions were statistically significant (p < 0.05). Specifically, taurocholenate sulfate mediated 8.7% of IL-7's effect on MS, anthranilate accounted for 8.4%, TDC contributed 16.5%, the leucine-to-phosphate ratio mediated 8.4%, albumin represented 3.9%, and sphingomyelin (d18:1/24:1, d18:2/24:0) mediated 10.0% of the effect (Table 3). However, most mediators exhibited wide confidence intervals in their mediation effects that included zero, though anthranilate showed a more precise positive association (β = 0.028, 95%CI: 0.01–0.06). These findings highlight the need for larger studies to validate the observed effects.

Table 3.

The mediation effect of IL-7 on MS via affecting metabolites.

Mediator	Total effect β	Direct effect A β	Direct effect B β	Mediation effect β (95% CI)	Mediated proportion (%) (95% CI)
Taurocholenate sulfate levels	0.333	0.128	0.226	0.029 (−0.01 to 0.07)	8.7% (−2.8% to 20.2%)
Anthranilate levels	0.333	0.177	0.161	0.028 (0.01 to 0.06)	8.4% (−1.6% to 18.7%)
Taurodeoxycholate levels	0.333	0.189	0.291	0.055 (−0.03 to 0.14)	16.5% (−8.4% to 41.3%)
Leucine-to-phosphate ratio	0.333	−0.138	−0.201	0.028 (−0.01 to 0.07)	8.4% (−4.3% to 20.9%)
Albumin levels	0.333	0.040	0.333	0.013 (−0.08 to 0.10)	3.9% (−23% to 31%)
Sphingomyelin (d18:1/24:1, d18:2/24:0)	0.001	0.135	0.001	0.0001 (−0.02 to 0.02)	10.0% (−17.3% to 17.6%)

IL-7: interleukin-7; MS: multiple sclerosis.

Discussion

In this study, we conducted a bidirectional two-sample MR analysis to investigate the causal relationships between inflammatory plasma proteins, specifically IL-7, metabolites, and the risk of MS. Our results indicate a modest causal link between increased serum IL-7 levels and a higher risk of developing MS, with consistent findings across both the IMSGC and UK Biobank cohorts. The strength of these results was further supported by sensitivity analyses, including MR-PRESSO, weighted median methods, and BWMR, which enhanced our confidence in the reliability of the causal estimates. Notably, we found no evidence of reverse causality; genetic predisposition to MS did not significantly affect IL-7 levels, indicating a one-way causal relationship from IL-7 to MS risk.

Our findings are consistent with previous research highlighting the critical role of IL-7 and its receptor (IL-7R) in autoimmune diseases, including MS. Earlier studies have identified genetic links between IL-7R variants and susceptibility to MS^37,38 along with functional evidence suggesting that IL-7 may facilitate the growth of autoreactive immune cells, potentially worsening disease severity.^39,40 The involvement of IL-7 in promoting neuroinflammation, particularly through T cell expansion, further underscores its role in the development of MS. Additionally, variants in IL-7Rα have been associated with changes in IL-7 bioavailability, which may elevate autoimmune risk in individuals with higher circulating IL-7 levels.⁷ However, there is conflicting evidence regarding the role of IL-7 in MS, especially concerning treatment responses. For example, a study has reported increased IL-7 levels in MS patients undergoing interferon beta therapy, indicating that the regulation of IL-7 may be complex and differ between treated and untreated individuals.⁴¹ Our findings provide genetic evidence supporting the role of IL-7 in the development of MS, suggesting that IL-7 could be a promising target for therapeutic interventions. This aligns with prior MR studies implicating immune-related proteins (e.g. CCL25, CXCL10, and CD40 ligand receptor) in MS pathogenesis,^20,42–45 though the specific proteins identified vary, likely due to differences in GWAS sources (e.g. IMSGC vs. UK Biobank) or SNP selection criteria.

Notably, our mediation analysis further elucidates that IL-7's effect on MS is partially mediated by serum metabolites (e.g. taurocholenate sulfate, anthranilate), accounting for 3.9%–16.5% of the total effect, while no mediation was observed for CSF metabolites. Taurodeoxycholate and taurocholenate sulfate, both derivatives of bile acids, emerged as significant mediators. These bile acids play important roles in lipid metabolism and immune regulation, which are critical factors in the pathophysiology of MS,³⁹ supporting the idea that dysregulation in bile acid metabolism could contribute to the disease. Furthermore, the synergistic roles of TDC and taurocholenate sulfate in lipid peroxidation and immune signaling suggest potential pathways through which these metabolites may drive neurodegeneration in MS.⁴⁶ Discrepancies with other metabolome-wide MR studies, such as associations for serine,^13,47 lipid species,⁴⁸ and CSF diglycerol and methylsuccinate⁴⁹ may stem from variations in metabolite coverage (plasma vs. CSF) or cohort heterogeneity. Despite these differences, our two-step MR approach strengthens potential causal inference by integrating inflammatory and metabolic pathways, highlighting IL-7's role in modulating MS risk through peripheral (but not central) metabolic intermediates.

Another important pathway identified in our analysis is the tryptophan-kynurenine (KYN) pathway, which has long been associated with immune regulation and neuroinflammation.^7,40 Anthranilate, a metabolite within this pathway, was found to mediate the effect of IL-7 on MS risk, aligning with previous research that emphasizes the role of kynurenine metabolites in MS.⁵⁰ The association of anthranilic acid (AA) with MS may vary by disease stage: while brain AA depletion reflects active demyelination in advanced MS models,⁵¹ our observed elevation in MS risk could indicate compensatory mechanisms in early disease. This dichotomy parallels other redox-active metabolites (e.g. uric acid) showing stage-dependent effects. Future longitudinal studies should clarify AA's temporal dynamics across MS phases.

Additionally, sphingomyelin, a lipid essential for maintaining myelin structure and CNS function, was identified as a mediator of IL-7's effect on MS risk. The balance between sphingomyelins and ceramides, two critical components of sphingolipid metabolism, is vital for preserving neuronal integrity. While previous studies have reported decreased sphingomyelin levels in MS patients due to upregulated sphingomyelinase activity,⁵² our MR analysis identified a positive association between a specific sphingomyelin species (d18:1/24:1, d18:2/24:0) and MS risk (OR = 1.001). This discrepancy may reflect: (1) differential effects of distinct sphingomyelin molecular species; (2) compensatory upregulation of specific sphingomyelins during early disease stages; or (3) limitations in cross-sectional versus genetic causal estimates. Further studies should clarify the temporal dynamics of sphingomyelin metabolism in MS progression.

Interestingly, our results indicated that higher serum albumin levels were associated with an increased risk of MS in the IMSGC cohort; however, reverse MR analysis did not confirm a significant relationship between MS and albumin levels. These findings contrast with previous studies that reported the reduced albumin levels in MS patients, potentially due to impaired iron metabolism.⁵³ Given these inconsistencies, further research is necessary to clarify the role of albumin in MS and determine whether it plays a direct or indirect role in disease progression.

Additionally, the leucine-to-phosphate ratio emerged as a potential mediator in the relationship between IL-7 and MS. Leucine is known for its influence on neuroinflammation and excitability, with studies suggesting its involvement in neurodegenerative conditions.^54,55 Phosphate homeostasis is also crucial for maintaining neuronal signaling and energy metabolism.⁵⁶ Notably, some initially identified metabolites, including TDC, albumin levels, and leucine-to-phosphate ratio, failed to reach significance in BWMR, possibly due to residual pleiotropy or limited statistical power. This highlights the importance of employing complementary MR methods or experiments to prioritize high-confidence associations for translational research.

Emerging evidence suggests the kynurenine-AA axis may influence neurological disorders through gut microbiota-mediated metabolite crosstalk. Probiotic interventions have been shown to modulate AA levels through the kynurenine pathway,⁵⁷ while AA supplementation directly alters neurotransmitter synthesis and behavioral outcomes in a microbiome-dependent manner.⁵⁸ Although our study does not directly validate anthranilate's microbial origins, these findings support IL-7's potential role in promoting neuroinflammation through microbiota–tryptophan interactions. The observed kynurenine metabolite alterations imply a cytokine-metabolite network, suggesting microbial tryptophan metabolism modulation as a potential therapeutic strategy for IL-7-driven neuroinflammation.

Our findings on IL-7 and related metabolites suggest several clinically relevant possibilities. First, therapeutic modulation of IL-7 signaling through approaches like monoclonal antibodies could potentially influence pathogenic metabolites, including taurocholenate sulfate and anthranilate. Second, with proper clinical validation, IL-7-associated metabolites may have utility as diagnostic or prognostic biomarkers. Third, integrating IL-7R genetic screening with metabolite profiling could offer opportunities for earlier risk stratification. While these possibilities represent promising translational directions, they will require additional investigation to determine their potential for improving MS clinical management.

Limitations

This study has several limitations. First, despite the use of multiple sensitivity analyses, pleiotropy remains a concern, as unaccounted pleiotropic effects could bias the causal estimates. While F-statistics indicated strong genetic instruments, weak instrument bias cannot be entirely ruled out, particularly for proteins like IL-7, where effect sizes were modest in the UK Biobank cohort. Second, while our SNP selection threshold (p < 1 × 10^⁻5) balanced discovery power and specificity, the lack of p-value correction in initial screening may increase false-positive risks. However, the application of BWMR, which provides posterior probabilities independent of frequentist p-values, strengthened confidence in our primary findings (e.g. IL-7 and sphingomyelin (d18:1/24:1, d18:2/24:0)). Future studies may prioritize stricter thresholds (p < 5 × 10⁻⁸) for hypothesis-free discovery. Third, although our mediation analysis identified potential metabolic mediators, the modest mediation proportions indicate that additional, unidentified mediators may also play a role in the inflammatory-metabolic-MS axis. Moreover, the generalizability of our findings is constrained by the focus on European populations, which underscores the necessity for replication in more diverse cohorts.

Conclusion

In conclusion, our MR study presents modest genetic evidence indicating that IL-7 potentially contributes to the risk of MS, with its effects being partially mediated by specific metabolites. These results enhance our understanding of the inflammatory and metabolic pathways involved in the development of MS and suggest that targeting IL-7, along with the associated metabolic changes, could lead to innovative therapeutic strategies for preventing or managing the disease. Future research should aim to further clarify the intricate relationships between inflammatory factors and metabolic pathways in MS, as well as investigate therapies that specifically target IL-7 to help slowdown disease progression.

Supplemental Material

sj-pdf-1-sci-10.1177_00368504251378619 - Supplemental material for Inflammation-metabolite crosstalk in multiple sclerosis: A mediation Mendelian randomization study of plasma inflammatory proteins, iron, and serum/cerebrospinal fluid metabolites

Supplemental material, sj-pdf-1-sci-10.1177_00368504251378619 for Inflammation-metabolite crosstalk in multiple sclerosis: A mediation Mendelian randomization study of plasma inflammatory proteins, iron, and serum/cerebrospinal fluid metabolites by Wu Yan, Wang Jianhong, Gan Ping and Zhang Linming in Science Progress

Supplemental Material

sj-pdf-2-sci-10.1177_00368504251378619 - Supplemental material for Inflammation-metabolite crosstalk in multiple sclerosis: A mediation Mendelian randomization study of plasma inflammatory proteins, iron, and serum/cerebrospinal fluid metabolites

Supplemental material, sj-pdf-2-sci-10.1177_00368504251378619 for Inflammation-metabolite crosstalk in multiple sclerosis: A mediation Mendelian randomization study of plasma inflammatory proteins, iron, and serum/cerebrospinal fluid metabolites by Wu Yan, Wang Jianhong, Gan Ping and Zhang Linming in Science Progress

Footnotes

Acknowledgements

The authors thank the investigators of the original studies for sharing the GWAS summary statistics, SCALLOP Consortium (2023), CLSA cohort (2023), 33 cohorts (2021), MWAS study (2021), FinnGen Cohort, IMSGC (2019), and UK Biobank (2015). AI tools were used solely for language polishing during manuscript preparation.

ORCID iDs

Wu Yan

Wang Jianhong

Ethics approval and consent to participate

The requirement for ethics approval was waived by our Institutional Review Board since this study is a secondary analysis of publicly available data. Ethical approval had been obtained for each of the original GWAS studies, and no individual-level data were used in our analyses.

Authors’ contribution

Wu Yan and Zhang Linming contributed to conceptualization, investigation, project administration, and supervision. Wu Yan and Wang Jianhong contributed to data curation, methodology, software, visualization, and writing—original draft. Wu Yan and Gan Ping contributed to validation. Zhang Linming and Gan Ping contributed to formal analysis and resources. Wu Yan contributed to funding acquisition. Wu Yan, Wang Jianhong, Zhang Linming, and Gan Ping contributed to writing—review &editing.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Wu Yan is supported by the Yunnan Clinical Medical Center for Neurocardiac Diseases (grant nos. 2024YNLCYXZX0053) and Yunnan Fundamental Research Projects (grant nos. 202201AT070291 and 202301AY070001-197).

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

All data used in the present study were obtained from GWAS summary statistics which were publicly released by genetic consortia.

Supplemental material

Supplemental material for this article is available online.

References

Göbel

Ruck

Meuth

. Cytokine signaling in multiple sclerosis: lost in translation. Mult Scler 2018; 24: 432–439.

Miller

Dziedzic

Saluk-Bijak

, et al. A review of various antioxidant compounds and their potential utility as complementary therapy in multiple sclerosis. Nutrients 2019; 11: 1528.

Ireland

Monson

Davis

. Seeking balance: potentiation and inhibition of multiple sclerosis autoimmune responses by IL-6 and IL-10. Cytokine 2015; 73: 236–244.

Jana

Mondal

Jana

, et al. Interleukin-12 (IL-12), but not IL-23, induces the expression of IL-7 in microglia and macrophages: implications for multiple sclerosis. Immunology 2014; 141: 549–563.

Lundström

Hermanrud

Sjöstrand

, et al. Interferon beta treatment of multiple sclerosis increases serum interleukin-7. Mult Scler 2014; 20: 1727–1736.

Grunwald

Krętowska-Grunwald

Adamska-Patruno

, et al. The role of selected interleukins in the development and progression of multiple sclerosis-A systematic review. Int J Mol Sci 2024; 25: 2589.

Duarte-Silva

Meuth

Peixoto

. The role of iron metabolism in the pathogenesis and treatment of multiple sclerosis. Front Immunol 2023; 14: 1137635.

Tham

Frischer

Weigand

, et al. Iron heterogeneity in early active multiple sclerosis lesions. Ann Neurol 2021; 89: 498–510.

Hofmann

Krajnc

Dal-Bianco

, et al. Myeloid cell iron uptake pathways and paramagnetic rim formation in multiple sclerosis. Acta Neuropathol 2023; 146: 707–724.

10.

Lin

Yang

Fuh-Ngwa

, et al. Genetically determined serum serine level has a novel causal effect on multiple sclerosis risk and predicts disability progression. J Neurol Neurosurg Psychiatry 2023; 94: 526–531.

11.

Gillen

Mubarak

Nguyen

, et al. Significance and in vivo detection of iron-laden microglia in white matter multiple sclerosis lesions. Front Immunol 2018; 9: 55.

12.

Yang

Musco

Simpson-Yap

, et al. Investigating the shared genetic architecture between multiple sclerosis and inflammatory bowel diseases. Nat Commun 2021; 12: 5641.

13.

Rispoli

Valentinuzzi

De Luca

, et al. Contribution of metabolomics to multiple sclerosis diagnosis, prognosis and treatment. Int J Mol Sci 2021; 22: 11112.

14.

Ladakis

Pedrini

Reyes-Mantilla

, et al. Metabolomics of multiple sclerosis lesions demonstrates lipid changes linked to alterations in transcriptomics-based cellular profiles. Neurol Neuroimmunol Neuroinflamm 2024; 11: e200219.

15.

Crick

Griffiths

Zhang

, et al. Reduced plasma levels of 25-hydroxycholesterol and increased cerebrospinal fluid levels of bile acid precursors in multiple sclerosis patients. Mol Neurobiol 2017; 54: 8009–8020.

16.

Fitzgerald

Smith

Kim

, et al. Multi-omic evaluation of metabolic alterations in multiple sclerosis identifies shifts in aromatic amino acid metabolism. Cell Rep Med 2021; 2: 100424.

17.

Zhao

Stacey

Eriksson

, et al. Genetics of circulating inflammatory proteins identifies drivers of immune-mediated disease risk and therapeutic targets. Nat Immunol 2023; 24: 1540–1551.

18.

International Multiple Sclerosis Genetics Consortium . Multiple sclerosis genomic map implicates peripheral immune cells and microglia in susceptibility. Science 2019; 365: eaav7188.

19.

Sudlow

Gallacher

Allen

, et al. UK Biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLoS Med 2015; 12: e1001779.

20.

Chen

Pettersson-Kymmer

, et al. Genomic atlas of the plasma metabolome prioritizes metabolites implicated in human diseases. Nat Genet 2023; 55: 44–53.

21.

Karjalainen

Karthikeyan

Oliver-Williams

, et al. Genome-wide characterization of circulating metabolic biomarkers. Nature 2024; 628: 130–138.

22.

Panyard

Kim

Darst

, et al. Cerebrospinal fluid metabolomics identifies 19 brain-related phenotype associations. Commun Biol 2021; 4: 63.

23.

Huang

Jiang

, et al. Investigating shared genetic architecture between inflammatory bowel diseases and primary biliary cholangitis. JHEP Rep 2024; 6: 101037.

24.

Zhang

Zhu

Wang

, et al. Plasma metabolites as mediators in immune cell-pancreatic cancer risk: insights from Mendelian randomization. Front Immunol 2024; 15: 1402113.

25.

Thompson

Burgess

Thompson

. Mendelian Randomization: Methods for Using Genetic Variants in Causal Estimation. Boca Raton: CRC Press, Biometrics, Volume 73, Issue 1, March 2017, Page 356, Doi: 10.1111/biom.12674.

26.

Sangha

Quon

Pfeffer

, et al. The role of vitamin D in neuroprotection in multiple sclerosis: An update. Nutrients 2023; 15: 2978.

27.

Kwiatkowski

Stewart

Cho

, et al. Nano and microparticle emerging strategies for treatment of autoimmune diseases: multiple sclerosis and type 1 diabetes. Adv Healthc Mater 2020; 9: e2000164.

28.

Emdin

Khera

Kathiresan

. Mendelian randomization. JAMA 2017; 318: 1925–1926.

29.

Yao

Han

Guo

, et al. The causal relationships between gut microbiota, brain volume, and intelligence: a two-step Mendelian randomization analysis. Biol Psychiatry 2024; 96: 463–472.

30.

Burgess

Thompson

. CRP CHD genetics collaboration. Avoiding bias from weak instruments in Mendelian randomization studies. Int J Epidemiol 2011; 40: 755–764.

31.

Bowden

Davey Smith

Burgess

. Mendelian randomization with invalid instruments: effect estimation and bias detection through egger regression. Int J Epidemiol 2015; 44: 512–525.

32.

Bowden

Del Greco

Minelli

, et al. A framework for the investigation of pleiotropy in two-sample summary data Mendelian randomization. Stat Med 2017; 36: 1783–1802.

33.

VanderWeele

. Mediation analysis: a practitioner's guide. Annu Rev Public Health 2016; 37: 17–32.

34.

Zhao

Ming

, et al. Bayesian weighted Mendelian randomization for causal inference based on summary statistics. Bioinformatics 2020; 36: 1501–1508.

35.

Carter

Gill

Davies

, et al. Understanding the consequences of education inequality on cardiovascular disease: Mendelian randomisation study. Br Med J 2019; 365: l1855.

36.

Galarza-Muñoz

Briggs

FBS

Evsyukova

, et al. Human epistatic interaction controls IL7R splicing and increases multiple sclerosis risk. Cell 2017; 169: 72–84.e13.

37.

Gregory

Schmidt

Seth

, et al. Interleukin 7 receptor alpha chain (IL7R) shows allelic and functional association with multiple sclerosis. Nat Genet 2007; 39: 1083–1091.

38.

Calzascia

Pellegrini

Lin

, et al. CD4 T cells, lymphopenia, and IL-7 in a multistep pathway to autoimmunity. Proc Natl Acad Sci U S A 2008; 105: 2999–3004.

39.

Lee

Axtell

, et al. IL-7 promotes T(H)1 development and serum IL-7 predicts clinical response to interferon-β in multiple sclerosis. Sci Transl Med 2011; 3: 93ra68.

40.

Lundström

Highfill

Walsh

, et al. Soluble IL7Rα potentiates IL-7 bioactivity and promotes autoimmunity. Proc Natl Acad Sci U S A 2013; 110: E1761–E1770.

41.

Bhargava

Smith

Mische

, et al. Bile acid metabolism is altered in multiple sclerosis and supplementation ameliorates neuroinflammation. J Clin Invest 2020; 130: 3467–3482.1.

42.

Ding

, et al. Genetically predicted association of 91 circulating inflammatory proteins with multiple sclerosis: a Mendelian randomization study. Brain Sci 2024; 14: 33.

43.

Lin

. Causal association of circulating inflammatory proteins on neurodegenerative diseases: insights from a Mendelian randomization study. J Cell Mol Med 2024; 28: e70176.

44.

Mao

Liu

, et al. Exploring the mediating role of cerebrospinal fluid metabolites in the pathway from circulating inflammatory proteins to multiple sclerosis: a Mendelian randomization study. Mult Scler Relat Disord 2025; 98: 106440.

45.

Gong

Cao

Zhu

. Exploring the causal association between 91 circulating inflammatory proteins and neurodegenerative diseases: a bidirectional two-sample Mendelian randomization and bioinformatics analysis. Brain Behav 2025; 15: e70586.

46.

Sreejayan

von Ritter

. Effect of bile acids on lipid peroxidation: the role of iron. Free Radic Biol Med 1998; 25: 50–56.

47.

Sun

Kiker

, et al. A metabolome-wide Mendelian randomization study prioritizes potential causal circulating metabolites for multiple sclerosis. J Neuroimmunol 2023; 379: 578105.

48.

Gilchrist

Mutz

Hysi

, et al. Evaluating metabolome-wide causal effects on risk for psychiatric and neurodegenerative disorders. BMC Med 2025; 23: 26.

49.

Zhang

Xiao

, et al. Mendelian randomization study of causal link from cerebrospinal fluid metabolomics to neurodegenerative diseases. Neurogenetics 2024; 26: 15.

50.

Fathi

Vakili

Yaghoobpoor

, et al. Dynamic changes in kynurenine pathway metabolites in multiple sclerosis: a systematic review. Front Immunol 2022; 13: 1013784.

51.

Polyák

Galla

Nánási

, et al. The tryptophan-kynurenine metabolic system is suppressed in cuprizone-induced model of demyelination simulating progressive multiple sclerosis. Biomedicines 2023; 11: 45.

52.

Momchilova

Pankov

Alexandrov

, et al. Sphingolipid catabolism and glycerophospholipid levels are altered in erythrocytes and plasma from multiple sclerosis patients. Int J Mol Sci 2022; 23: 7592.

53.

Oliveira

Kallaur

Reiche

EMV

, et al. Albumin and protein oxidation are predictors that differentiate relapsing-remitting from progressive clinical forms of multiple sclerosis. Mol Neurobiol 2017; 54: 2961–2968.

54.

Hartman

Santos

O'Riordan

, et al. Potent anti-seizure effects of D-leucine. Neurobiol Dis 2015; 82: 46–53.

55.

Hegdekar

Lipinski

Sarkar

. N-Acetyl-L-leucine improves functional recovery and attenuates cortical cell death and neuroinflammation after traumatic brain injury in mice. Sci Rep 2021; 11: 9249.

56.

Arredondo

Cefaliello

Dyrda

, et al. Excessive release of inorganic polyphosphate by ALS/FTD astrocytes causes non-cell-autonomous toxicity to motoneurons. Neuron 2022; 110: 1656–1670.e12.

57.

Rudzki

Ostrowska

Pawlak

, et al. Probiotic Lactobacillus Plantarum 299v decreases kynurenine concentration and improves cognitive functions in patients with major depression: a double-blind, randomized, placebo controlled study. Psychoneuroendocrinology 2019; 100: 213–222.

58.

Castells-Nobau

Puig

Motger-Albertí

, et al. Microviridae bacteriophages influence behavioural hallmarks of food addiction via tryptophan and tyrosine signalling pathways. Nat Metab 2024; 6: 2157–2186.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.44 MB

2.33 MB