Effect of the immune cells and plasma metabolites on rheumatoid arthritis: a mediated mendelian randomization study

Background Increasing evidence indicates a close relationship between alterations in human immune cells and plasma metabolites with Rheumatoid Arthritis (RA). However, limited studies have left the causal relationships behind these links unclear. Methods A bidirectional Mendelian Randomization (MR) study was conducted, combined with mediation analysis, using data from genome-wide association study database covering 731 immune cell phenotypes and 1,400 plasma metabolite traits to explore their causal relationships with RA and potential mediating effects. The primary method used for MR analysis was inverse-variance weighted and False Discovery Rate (FDR) correction was applied to verify the robustness of our results. Results HLA DR on CD33- HLA DR+ (myeloid cell group) (OR, 1.422; 95% CI, 1.194–1.694; P < 0.001; PFDR = 0.012) increased the risk of developing RA. CD19 on IgD+ CD38- naive (B cell group) (OR, 0.969; 95% CI, 0.954–0.985; P < 0.001; PFDR = 0.021) reduced the risk of developing RA. RA was a risk factor for HLA DR on CD14- CD16+ monocytes (monocyte group) (OR, 1.242; 95% CI, 1.102–1.401; P < 0.001; PFDR  = 0.047). RA was a protective factor for memory B cell %lymphocyte (B cell group) (OR, 0.861; 95% CI, 0.795–0.933; P < 0.001; PFDR = 0.050), CD4+ CD8dim T cell %lymphocyte (TBNK group) (OR, 0.802; 95% CI, 0.711–0.904; P < 0.001; PFDR  = 0.043), CD4+ CD8dim T cell %leukocyte (TBNK group) (OR, 0.814; 95% CI, 0.726–0.913; P < 0.001; PFDR  = 0.046), CD24 on IgD+ CD24+ B cells (B cell group) (OR, 0.857; 95% CI, 0.793–0.927; P < 0.001; PFDR  = 0.038), and CD24 on unswitched memory B cells (B cell group) (OR, 0.867; 95% CI, 0.797–0.942; P < 0.001; PFDR  = 0.050). Increasing levels of docosatrienoate (22:3n3) (OR, 0.886; 95% CI, 0.838–0.936; P < 0.001; PFDR = 0.023) significantly reduced the risk of developing RA. The mediating effect of plasma metabolites in this context was not established. Conclusion This study provides genetic evidence for the intricate relationships between immune cells, plasma metabolites, and RA, highlighting the potential mechanisms involved. This will contribute to future directions in precision medicine and research.


Introduction
Rheumatoid Arthritis (RA) is a chronic autoimmune disease characterized by the disruptions of multiple metabolic pathways (1).It features immune cell infiltration, systemic inflammatory responses, neovascularization, and synovial hyperplasia, leading to bone and cartilage erosion and degradation, which subsequently results in loss of function and disability in patients (1,2).Annually, 0.46% of the global population is affected by RA, with women being approximately three times more likely to develop the condition than men (3).About 2.3 million Europeans are diagnosed with RA annually, with a prevalence of approximately 1% (4,5).The direct and indirect socioeconomic costs of RA in European countries exceed €45 billion annually (4).RA has a significant negative impact on patients' economic burden, mental health, and quality of life, making it crucial to enhance research for early diagnosis and treatment.
The immune mechanisms in patients with RA are complex, involving not only immune diseases but also potentially affecting the respiratory system, bone and soft tissues, psychology, and cardiovascular areas (6,7).For instance, immune cells, such as B cells and T cells, in patients with RA produce pro-inflammatory factors that stimulate fibroblast polarization, activating osteoclasts and releasing matrix metalloproteinases, ultimately driving bone and cartilage destruction (8).Recent studies have indicated that the aging of immune cells leads to protein homeostasis failure, mitochondrial dysfunction, and lysosomal failure, resulting in impaired cellular functions that trigger autoimmunity and ultimately induce RA (9).
Metabolomics enables high-throughput characterization of cells, body fluids, and tissues within an organism, thereby revealing physiological and pathological changes and enhancing the understanding of diseases (10, 11).In systemic diseases such as rheumatoid arthritis, abnormal changes in circulating metabolomics may refl ect the patient' s microbiome, complications, genetic susceptibility, drug responses, systemic inflammation, and various environmental factors such as smoking, alcohol consumption, and diet (12).Metabolites serve as reference indicators for the effects of internal and external factors on a biological organism, such as elevated plasma cholesterol levels in patients with hypercholesterolemia. Metabolic changes in patients with RA are complex and diverse.For example, the upregulation of branched-chain amino acids and alanine and downregulation of glucose and glycylglycine play direct or indirect roles in inducing immune responses and driving inflammatory mechanisms in patients (13).Enhancing metabolomic research in patients with RA can help us better understand drug mechanisms and aid clinicians in improving diagnosis and treatment prognosis by recognizing variations in internal metabolites among patients.
Mendelian Randomization (MR) is an epidemiological method based on Mendel's laws, which uses genetic variations as tools to explore causal relationships between risk factors and outcomes (14).Compared with other statistical methods, MR can minimize the influence of potential confounders and reverse causation, making the results more reliable (15).Therefore, to explore the potential causal relationships between immune cell phenotypes, plasma metabolites, and RA, two-sample bidirectional MR analysis and mediation analysis were performed using Genome-Wide Association Study (GWAS) data and FinnGen database.

Study design
First, the possible bidirectional causal relationships between immune cells and RA were explored.Next, the potential bidirectional causal relationships between metabolites and RA were investigated.Plasma metabolites were then introduced as mediators to examine their mediating effects on the relationship between immune cell phenotypes and RA.All included Instrumental Variables (IVs) met three basic assumptions: first, IVs are significantly associated with exposure; second, IVs are not related to confounders or outcome; third, the included IVs influence the outcome solely through their effect on exposure (16).

Data source GWAS data sources for immune cells
The GWAS data for the 731 immune cell phenotypes used in this study were obtained from a public repository (GCST0001391 to GCST0002121) (17,18).The dataset included 22 million genetic variants from 3,757 individuals of Italian Sardinian ancestry.The dataset was categorized into 118 absolute cell counts, 389 median fluorescence intensities representing surface antigen levels, 32 morphological phenotypes, and 192 relative cell counts.

GWAS data sources for plasma metabolites
The GWAS data for the 1400 plasma metabolites used in this study were obtained from a public repository (GCST90199621 to GCST90201020) (18,19).The dataset included more than 8,000 individuals of European ancestry and plasma metabolite data consisting of 1,091 blood metabolites and 309 metabolite ratios.

Data source for RA
Data related to RA was sourced from the tenth edition of the FinnGen database (R10) (20, 21).The dataset finngen_R10_M13_ RHEUMA included 276,565 study subjects of European descent, comprising 13,621 cases and 262,844 controls.

Selection of instrumental variables
First, for immune cells and plasma metabolites, the significance threshold was set to P < 1*10 ^-5 , in line with previous studies (22).For reverse MR, a more stringent threshold (P < 5*10 ^-8 ) was applied to select RA-related Single Nucleotide Polymorphisms (SNPs) as instrumental variables.Second, strongly linked variants were excluded to avoid linkage disequilibrium issues among SNPs (R2 = 0.001, clumping distance = 10,000 kb).Third, to avoid weak instrument bias, F-statistics were calculated for all IVs included in the study and the IVs with F < 10 were excluded (23).

Statistical analysis
The Inverse-Variance Weighted (IVW) method with a randomeffects model was used as the primary analysis method for the MR analysis.To ensure the reliability of the results, the P-values obtained from the IVW method were corrected using the Benjamini-Hochberg method to control the False Discovery Rate (FDR) due to multiple tests (24).Considering that the accuracy of the IVW method is based on the assumption of no horizontal pleiotropy, MR-Egger regression was performed throughout the MR process to calculate the intercept and assess the significance of horizontal pleiotropy (25).In the absence of horizontal pleiotropy, a P FDR <0.05 indicates significant positive results, and a P FDR <0.2 represents suggestive positive results (26).A two-step MR design was used to perform mediation analysis to determine whether plasma metabolites mediate the relationship between immune cells and RA.Heterogeneity between IVs was assessed by Cochran's Q test based on IVW and MR-Egger methods.Funnel plots, Leave-One-Out (LOO) analysis, and additional methods such as MR-Egger, weighted median, weighted mode, and simple mode were included as part of our sensitivity analysis throughout the study to enhance the robustness of the results.
All data analyses were performed using the "TwoSampleMR" package (version 0.5.10) in R statistical software (version 4.3.3).

Results of selection of instrumental variables
For detailed information on IVs and their association with immune cell phenotypes, blood metabolites, and rheumatoid arthritis, please refer to Supplementary Data S1.

Effects of plasma metabolites on RA risk
In the MR analysis of 1400 metabolites and RA risk, 59 metabolites were significant in IVW method (P < 0.05) and showed no horizontal pleiotropy in MR-Egger regression (Supplementary Figure S11).After FDR correction, an increase in docosatrienoate (22:3n3) levels (OR, 0.886; 95% CI, 0.838-0.936;P < 0.001; P FDR = 0.023) significantly reduced the risk of developing RA.The four other Supplementary Methods showed the same trend (Supplementary Figure S12).The robustness of these results was further validated by LOO analysis and funnel plots (Supplementary Figure S12).The heterogeneity and horizontal pleiotropy results are shown in Supplementary Data S2.

Impact of RA on plasma metabolites causation
In the MR analysis of RA and docosatrienoate (22:3n3) levels, no significant causal relationship was found (OR, 0.997; 95% CI, 0.949-1.048;P = 0.907) (Supplementary Figure S13).The heterogeneity and horizontal pleiotropy results are shown in Supplementary Data S2.

Discussion
This is the first study to utilize a large volume of publicly available genetic data to investigate the causal associations between 731 immune cell phenotypes, 1400 plasma metabolites, and RA using MR analysis.Ultimately, it was observed that RA was causally associated with 6 immune cell phenotypes, whereas 2 immune cells phenotypes were causally associated with RA (P FDR < 0.05).Meanwhile, RA was suggestively causally associated with 29 immune cell phenotypes, whereas 10 immune cell phenotypes were suggestively causally Frontiers in Endocrinology frontiersin.orgassociated with RA (P FDR < 0.2).One metabolite was significantly causally associated with the risk of RA (P FDR < 0.05).no mediation effect of the metabolite between immune cell traits and RA was observed.A significant decrease in the risk of RA was observed with a reduction in the proportion of CD19 on IgD+ CD38-naïve cells.B cells are involved in RA through the production of autoantibodies antigen presentation to T cells, and secretion of cytokines (such as tumor necrosis factor alpha, Interleukin-1b) (6, 27).Chimeric antigen receptor T cells have been widely used to treat autoimmune diseases by inducing rapid and sustained depletion of circulating B cells (28).Despite the improvement in inflammatory markers in patients with RA following B cell depletion therapy, the risk of infection and cancer has increased (29).Therefore, it is crucial to distinguish between protective and pathogenic B cells for targeted intervention.Currently, there is no literature on the association between CD19 on IgD+ CD38-naïve cells and the risk of RA.This study is the first to identify a negative causal relationship between CD19 on IgD+ CD38-naïve cells and RA risk, potentially providing new insights into the mechanism of B cell involvement in RA.Our results indicated that an increased proportion of HLA DR on CD33-HLA DR+ cells is associated with an increased risk of RA.The Human Leukocyte Antigen-DR (HLA-DR) molecule presents exogenous antigens to CD4+ T cells, promoting CD40L expression and specific Interferon-g (IFN-g) activation on CD4+ T cells, thereby inducing immune responses and contributing to RA development (30,31).Recent studies have confirmed the Shared Epitope (SE) hypothesis, suggesting that HLA-DR molecules with SE motifs increase the risk of RA by specifically binding to citrullinated peptides and triggering autoimmune processes (32).The HLA-DRB1 *0404 gene is considered to have the greatest impact on increasing the risk of RA (31).Additionally, the presence of SE alleles specifically increases the risk of anti-citrullinated protein antibodies in patients with RA, but the underlying mechanism remains unclear and warrants further investigation (31).Forest plot of mediation effect analysis of immune cells, plasma metabolites, and rheumatoid arthritis.This study revealed a positive correlation between RA and the levels of HLA DR on CD14-CD16+ monocytes.Monocytes, as part of the mononuclear phagocyte system, play a crucial role pathogen recognition and clearance and are important targets for RA treatment (33,34).HLA DR on CD14-CD16+ monocytes, a subset of nonclassical monocytes, is involved in promoting the resolution or disease progression in chronic inflammation (35).The role of non-classical monocytes in arthritis remains controversial.While Serum Transfer-Induced Arthritis (STIA) models show that non-classical monocytes are not essential for arthritis development and may even alleviate it, other studies have shown a positive correlation between non-classical monocytes and markers of joint destruction (36,37).Previous research has indicated that there are fewer nonclassical monocytes in patients with RA than in the general population (38), but recent studies have shown an increase in these cells in the peripheral blood of patients with RA (39)(40)(41), which is consistent with our findings.This suggests the transformation of non-classical monocytes to classical monocytes in patients with RA.Analysis of synovial tissue biopsies from patients with RA revealed the accumulation of CD4+ CD8dim T cells in the synovial tissue (42).Our results revealed a negative correlation between RA and the expression of CD4+ CD8dim T cells as a percentage of lymphocytes and leukocytes, possibly suggesting that these cells are recruited from circulating peripheral blood to the joint fluid and synovium in patients with RA.There are no current studies on these cell levels in patients with RA, warranting further investigation.Previous studies have debated the B cell subtypes in RA, with some indicating a lower frequency of memory B cells in patients with RA than in healthy individuals, while others found no difference (43).Naïve mature B cells differentiate into memory B cells upon antigen contact and further differentiate into plasma cells through somatic hypermutation, class switch recombination, and other mechanisms resulting in antibody secretion (44).Our results revealed a negative correlation between RA and percentage of memory B cells among lymphocytes, suggesting the accumulation of memory B cells in the synovial tissue of patients with RA.The significant increase in circulating memory B cells after infliximab treatment in patients with RA further supports the idea of memory B cell migration to inflamed tissue, which can be corrected by tumor necrosis factor blockade (45).The decrease in B cells in patients with RA compared with healthy individuals may be due to the migration of B lymphocytes from peripheral blood to joint fluid, which promotes local inflammation through cytokine, chemokine secretion, and antigen presentation (43).This finding is consistent with our reverse MR results, indicating a negative correlation between RA and three B cell types (Memory B cell %lymphocyte, CD24 on IgD+ CD24+ B cell, and CD24 on unswitched memory B cell).
It has been demonstrated that higher levels of docosatrienoate (22:3n3) are associated with a reduced risk of RA.Docosatrienoate (22:3n-3), an omega-3 unsaturated fatty acid with 22 carbon atoms and 3 double bonds, has not been extensively studied clinically.Omega-3 fatty acids have been found to have anti-inflammatory and immunomodulatory effects through various mechanisms, such as altering the composition of fatty acids in cell membrane phospholipids and inhibiting the activation of pro-inflammatory factor kappa B, thereby suppressing the expression of inflammatory genes (46).Clinical studies have shown that omega-3, as an anti-inflammatory nutrient, has a positive effect on symptoms, cardiovascular events, and the progression of rheumatoid disease in patients with RA (46)(47)(48).Our findings provide genetic evidence for future research on the metabolic mechanisms of docosatrienoate (22:3n3) in RA.
Firstly, this study is the first to use the latest large-scale GWAS cohorts and multiple MR analysis methods to explore the causal relationships between immune cells, metabolites, and RA, using a series of sensitivity analyses to ensure the robustness of our results.Secondly, our study strictly followed the STROBE-MR guidelines and conclusions were based on MR-Egger regression horizontal pleiotropy negativity and FDR correction.However, while acknowledging the strengths of this study, its limitations must also be considered.First, to avoid population bias, all the groups involved in the study were of European ethnicity, which may affect the generalization of the conclusions.Second, due to the lack of personal information in the source data, this study was unable to stratify the analysis according to factors such as sex and age.Lastly, given the exploratory nature of our study, subsequent related studies based on our findings could introduce techniques such as single-cell RNA sequencing and fine-tuned localization.

Conclusion
Bidirectional MR and mediation analyses were utilized to provide genetic evidence for elucidating the complex relationships between immune cells, plasma metabolites, and RA.This study provided genetic evidence for future researchers and clinicians to explore the prevention and treatment of RA.

FIGURE 1 Forest
FIGURE 1Forest plot of positive causal associations between immune cell phenotypes and rheumatoid arthritis after false discovery rate correction.

FIGURE 3 Forest
FIGURE 3Forest plot of positive causal associations between rheumatoid arthritis and immune cell phenotypes after false discovery rate correction.

FIGURE 4 Forest
FIGURE 4Forest plot of suggestive positive causal associations between rheumatoid arthritis and immune cell phenotypes after false discovery rate correction.