Abstract

Introduction

Systemic lupus erythematosus (SLE) is a chronic, multi-system autoimmune disorder that profoundly affects the quality of life and imposes significant socio-economic burdens[1, 2]. The conventional treatment regimen for SLE includes nonsteroidal anti-inflammatory drugs (NSAIDs), antimalarials, corticosteroids, and immunosuppressants[3]. . Recent advancements have introduced biologic agents, offering new therapeutic avenues for SLE management. For instance, belimumab, approved by the U.S. Food and Drug Administration (FDA), targets the B-lymphocyte stimulator (BLyS) to curtail autoantibody production[4]. Despite the diversity of treatment options, therapeutic outcomes for SLE vary considerably among individuals, and most strategies prioritize symptom management over disease eradication[5] .Moreover, long-term use of existing medications can lead to serious adverse effects, including infections, osteoporosis, and retinal disorders[5].

Proteins play a crucial role in immune regulation and inflammatory responses, both of which are closely associated with the pathophysiological processes of SLE[6]. For example, complement activation is one of the key pathological mechanisms leading to tissue inflammation and damage in systemic lupus erythematosus (SLE)[7]. Furthermore, proteins are primary targets for pharmacological interventions[8]. In two Phase III trials, anifrolumab, which blocks the type I interferon receptor 1 (IFNAR1), received approval for treating SLE[9]. The circulating proteome, comprising proteins released both actively and passively into the bloods from various tissues and cells[10], has been confirmed its correlation with SLE. Yong Dai's team utilized data-dependent acquisition (DDA) and data-independent acquisition (DIA) proteomics techniques to identify three proteins as potential biomarkers for diagnosing SLE[11]. Liu et al. applied TMT-labeled quantitative proteomics alongside enzyme-linked immunosorbent assays (ELISA) to demonstrate notable differences in the serum levels of SAA1 and CD248 between SLE patients and healthy controls[12]. Nonetheless, current research is constrained by small sample sizes, a limited protein range, and potential confounding factors. Conducting randomized controlled trials to investigate the potential causal relationships between numerous proteins and SLE is also challenging.

Mendelian Randomization (MR) employs genetic variations as instrumental variables (IVs) to establish causal relationships between exposures and outcomes, effectively mitigating confounding factors and preventing reverse causality[13, 14]. In this study, we conducted a two-sample Mendelian randomization analysis using pQTL data derived from extensive proteomics studies and genome-wide association study (GWAS) data for SLE to explore the genetic causal relationship between these elements. Additionally, we constructed a protein-protein interaction (PPI) network and performed colocalization analysis for proteins statistically significant in the MR analysis. This effort aims to identify potential therapeutic targets and lay the foundation for future clinical applications (Figure 1).

Methods

Source of Exposure Data
We obtained single nucleotide polymorphism (SNP) data associated with plasma protein levels from the Fenland study.This research carried out a genome-wide proteomics association study involving 10,708 participants of European ancestry, assessing 4,775 plasma proteins utilizing the SomaScan v4 assay (http://www.omicscience.org/apps/pgwas)[15].

Source of Outcome Data
The outcome of interest, systemic lupus erythematosus (SLE), was studied using genome-wide association study (GWAS) data obtained from the Finnish Genetic Study (FinnGen). This study extracted over 500,000 samples from the Finnish biobank, integrating longitudinal phenotypic data and digital health records from the national health registry[16]. We accessed the publicly available FinnGen R10 dataset (https://r10.finngen.fi/), which includes data on 1,083 SLE cases and 306,504 controls[16].

Study Design
In this research, we performed a comprehensive two-sample Mendelian Randomization (MR) analysis to evaluate the causal relationship between circulating proteins and systemic lupus erythematosus (SLE). To ensure the validity of the results, the MR analysis mandated that the selected instrumental variables (IVs) satisfy three critical criteria: (1) the IVs must demonstrate a significant direct association with the exposure variable (circulating proteins); (2) the IVs must be independent of any confounders that could affect the exposure-outcome relationship; (3) the influence of the IVs on the outcome must be mediated exclusively through the exposure, with no alternative causal pathways[17].

Selection of Instrumental Variables
Informed by the three assumptions outlined previously and recent research findings,stringent criteria for SNP selection: (1) SNPs must be significantly associated with circulating proteins, adhering to a stringent significance threshold (P-value < 5 × 10-8)[18]; (2) To ensure the independence of the selected IVs and mitigate the effects of linkage disequilibrium (LD), SNP clustering methods were utilized (r² = 0.001, kb = 10,000)[19]; (3) IVs with an F-statistic less than 10, generally considered weak, were excluded to avoid instability and bias in effect estimation[20]; (4) SNPs strongly associated with the outcome variable (P-value < 5 × 10-8) were also excluded to prevent direct causal interference with the outcome[21]; (5) Cis-protein quantitative trait loci (cis-pQTLs) spanning a gene range of ±1Mb were selected. Located at or near the gene encoding the target proteins, cis-QTLs are favored for their substantial contribution to explaining protein expression, compared to variants in other genomic regions[22].

Mendelian Randomization Analysis and Sensitivity Analysis
In this study, we conducted two-sample MR analysis using R software (version 4.3.3) and the "TwoSampleMR" package, employing various statistical methods to assess the potential causal relationship between circulating proteins and the risk of systemic lupus erythematosus (SLE). We employed the inverse variance weighted (IVW) method as the primary analytical tool when two or more instrumental variables (IVs) were available. Additionally, when the number of IVs was three or more, the weighted median (WM) and Mendelian randomization-Egger (MR-Egger) methods were also implemented. For proteins represented by a single IV, the Wald ratio method was applied to estimate the change in the log odds ratio of SLE risk per one standard deviation (SD) increase in protein levels[23].

Additionally, we performed sensitivity analysis using Cochran's Q test, MR-Egger, and MR-PRESSO methods. The Cochran's Q statistic assessed heterogeneity among the selected instrumental variables, with a P-value < 0.05 denoting significant heterogeneity. A significant intercept in the MR-Egger method indicated the potential pleiotropy, which was determined using a P-value < 0.05[24]. The MR-PRESSO method, utilized via the "MR-PRESSO" package, aimed to identify and remove SNP outliers with horizontal pleiotropy. However, when the number of SNPs is small, it is insufficient for effective heterogeneity and pleiotropy analysis[25]. Lastly, the MR-Steiger test was applied to evaluate directional causality by comparing the proportion of variance explained by the instrumental variables in relation to both the exposure and outcome variables, thus assessing the suitability of the instrumental variables[26].

pQTL - GWAS Colocalization Analysis
To determine whether protein expression levels and SLE risk share causal variants within the same genomic region, we performed a colocalization analysis using the "coloc" R package with default prior probabilities[27]. Bayesian methods were applied to each cis-gene locus of the proteins to evaluate five mutually exclusive hypotheses: (1) No significant association with either trait (H0); (2) Associated only with protein levels (H1); (3) Associated only with SLE risk (H2); (4) Association with both traits, driven by distinct causal variants (H3); (5) Both traits driven by the same causal variant (H4). In our analysis, co-localization was deemed supported when the posterior probability of sharing pathogenic variants (PP.H4) exceeded 0.6[28].

Protein-Protein Interaction Network
To enhance our understanding of the interactions between proteins, we constructed a protein-protein interaction (PPI) network. Using the STRING database (https://cn.string-db.org/), we developed the PPI network with a minimum interaction threshold of 0.4, maintaining other parameters at their default settings[29].

Results

Circulating proteins and SLE risk mendelian analysis results
In this study, due to the stringent selection criteria for instrumental variables, only 1445 circulating proteins were included in the analysis (Supplementary Table 1). Mendelian randomization (MR) analysis revealed that 82 plasma proteins were potentially causally associated with SLE risk (P < 0.05) as shown in Figure 2. Specifically, high expression of 46 proteins and low expression of 36 proteins are positively correlated with an increased risk of systemic lupus erythematosus (SLE). Additionally, for proteins with at least three SNPs (nSNP ≥ 3), 13 proteins were further validated by the Weighted Median (WM) method as being associated with SLE risk (P < 0.05), while no specific associations were identified by the MR Egger method (Supplementary Table 2).

The minimum F-statistic for the selected instrumental variables (IVs) for each protein was 23.926, confirming their strength as robust instruments. The p-values of the Steiger test ranged from 0 to 9.01E-06, affirming that the directionality of the instrumental variables is consistent with the fundamental assumptions of the MR analysis. Additionally, the MR-Egger intercept tests for each protein indicated no evidence of horizontal pleiotropy (P > 0.05), with further details available in Supplementary Table 2.

Colocalization Analysis
Bayesian co-localization analysis was conducted on 82 circulating proteins that showed statistically significant associations, encompassing both upstream and downstream regions (detailed results are presented in Supplementary Table 3). The analysis revealed co-localization between systemic lupus erythematosus (SLE) and several proteins, namely PDHX (PP.H4 = 0.67), CTSF (PP.H4 = 0.64), and TNFAIP3 (PP.H4 = 0.66), suggesting that these proteins may share common genetic variants with SLE (Figure 3).

PPI networks
A total of 82 proteins (P < 0.05) were entered into the STRING database to construct a protein network. Given the study's threshold for minimum interaction strength of 0.4, only 50 imported proteins successfully formed a network with other supplementary proteins, collectively comprising 82 nodes and 54 edges (Figure 4). Subsequent network functional enrichment analysis (Supplementary Figure 1) identified several proteins significantly involved in biological processes such as inflammatory response and immune system regulation, with PFDR values of 1.1e-04 and 4.8e-04, respectively. These processes are intimately connected to the pathophysiology of SLE.[30, 31].

Discussion

This proteomics Mendelian Randomization study investigated the causal relationships between 4775 plasma proteins and the risk of systemic lupus erythematosus (SLE). Employing stringent selection criteria, we identified 82 plasma proteins significantly associated with SLE risk. Notably, proteins such as TNFAIP3, PDHX, and CTSF appear to share the same causal variants as the disease. We further elucidated the intricate biological interactions among these proteins using the Protein-Protein Interaction (PPI) network.

TNFAIP3 (TNF Alpha Induced Protein 3) is an ubiquitin-editing enzyme extensively documented to act as an endogenous negative feedback regulator of inflammatory responses by inhibiting NF-κB signaling pathway activity[32, 33]. However, TNFAIP3 also promotes the phosphorylation of receptor-interacting protein 3 (RIP3) through deubiquitination[34], thereby activating the NLRP3 inflammasome pathway, which contributing to the development of lupus nephritis[35].Additionally, various GWAS studies across different populations have demonstrated that single nucleotide polymorphisms (SNPs) at the TNFAIP3 locus are associated with susceptibility to SLE[36]. For instance, the rs2230926 SNP alters the amino acid sequence of the A20 protein (from phenylalanine to cysteine), diminishing its capacity to inhibit TNF-induced NF-κB activation. This alteration compromises the inflammatory control in individuals carrying this risk allele, thereby heightening their susceptibility to SLE[37, 38]. Additionally, cohort studies have associated the TNFAIP3 rs5029939 genetic polymorphism with SLE susceptibility and potential impacts on its clinical phenotype[38]. Notably, this study identifies for the first time that rs59693083 (located in the TNFAIP3 promoter[39]) escalates the risk of SLE, offering new insights into the pathogenesis and clinical treatment of the disease.

PDHX, a crucial component of the pyruvate dehydrogenase complex (PDC), primarily facilitates the conversion of pyruvate into acetyl-CoA, bridging glycolysis to the Krebs cycle[40]. Research indicates that the 11p13 locus, situated between PDHX and CD44, is linked to genetic susceptibility to SLE. This region contains multiple regulatory sites, which potentially affect the expression and function of both PDHX and CD44, consequently influencing immune regulation and inflammatory responses in SLE[41]. This aligns with our MR results. Intriguingly, high PDHX expression has also been associated with diminished immune cell infiltration in the tumor microenvironment[42], suggesting that PDHX's role in SLE pathogenesis might be more complex than previously understood. These insights necessitate further basic and clinical studies to elucidate the underlying mechanisms.

CTSF (Cathepsin F) is a lysosomal cysteine protease integral to various physiological and pathological processes, including immune responses and antigen processing[43]. However, its impact on the immune system is subject to ongoing debate. Evidence indicates that inhibiting CTSF activity may block MHC class II processing in macrophages, reducing antigen presentation by MHC class II, and potentially ameliorating diseases associated with inappropriate or excessive immune responses[44]. However, In clear cell renal carcinoma (ccRCC) studies, CTSF expression inversely correlated with both the infiltration of diverse immune cells and the expression of MHC molecules such as TAP1 and TAP2[45]. Our MR results indicate a negative correlation between CTSF and SLE risk. Additionally, some studies provide indirect evidence. For example, cysteine cathepsins (Cts) have been implicated in the hydrolysis of the extracellular matrix (ECM), involving the processing of cytokines, chemokines, and cell adhesion molecules, which are pivotal in inflammatory responses[46].. Cts may mitigate inflammation triggered by cellular debris by efficiently degrading damaged organelles and proteins. Although the link between CTSF and SLE risk has been less extensively explored, the causal relationship identified in this study invites further investigation into its specific role and mechanisms in SLE and immune system regulation, potentially unveiling new therapeutic targets.

This study employed a large-scale dataset for Mendelian Randomization analysis to enhance result reliability. In the initial phases, we selected cis-pQTLs for inclusion due to their typical proximity to or within genes encoding proteins, thus directly regulating protein expression by affecting transcription, translation, degradation, stability, or activity. Additionally, our team utilized various statistical methods such as the Wald ratio, Inverse Variance Weighted (IVW), Weighted Median (WM), and Mendelian Randomization-Egger (MR-Egger) to bolster the robustness of the MR analysis. Finally, we performed protein-protein interaction (PPI) network analysis to validate the links between protein-related biological processes and the disease's pathological processes.

Despite the strengths of our study, it has several limitations. Firstly, it relies on population data from individuals of European ancestry, thereby constraining the generalizability of the findings to other ancestral groups. Secondly, while the Fenland study included a broad range of circulating proteins, our stringent selection criteria for instrumental variables might have excluded other relevant target proteins. Moreover, protein expression and function are subject to influences from various factors, such as environmental interactions and epigenetic modifications, which were not comprehensively accounted for in our analysis. These complex influences extend beyond genetic variations and were not thoroughly addressed in this study.

Conclusion

Overall, this study offers a comprehensive assessment of the causal relationships between circulating proteins and systemic lupus erythematosus (SLE), further confirming their critical roles in the initiation and progression of the disease. Notably, proteins such as TNFAIP3, PDHX, and CTSF are identified as likely candidates for new therapeutic targets. However, additional research is essential to unravel the complex mechanisms linking these candidate proteins with SLE risk, a crucial step in validating their potential clinical relevance.

Abbreviations

B-lymphocyte stimulator: BLyS; cysteine cathepsins: Cts; Cis-protein Quantitative Trait Loci: cis-pQTLs; data-dependent acquisition: DDA; data-independent acquisition: DIA; Extracellu-lar Matrix: ECM; enzyme-linked immunosorbent assays: ELISA; Food and Drug Administration: FDA; Genome - Wide Associa-tion Study: GWAS; I interferon receptor 1: IFNAR1; Instrumen-tal Variables: IVs; Inverse Variance Weighted: IVW; Linkage Disequilibrium: LD; Mendelian Randomization: MR; Mendelian Randomization-Egger: MR-Egger; protein Quantitative Trait Loci: pQTL; Protein-Protein Interaction: PPI; Receptor-Interact-ing Protein 3: RIP3; Single Nucleotide Polymorphisms: SNPs; Systemic Lupus Erythematosus: SLE; Trans-protein Quantita-tive Trait Loci: trans-pQTLs; Weighted Median: WM.

Supplementary Materials

Declarations

Author contributions

Xinzhen Zhao: Writing – review & editing, original draft, Conceptualization, Visualization, Formal analysis. Yinying Chai: Writing – review & editing, Validation, Supervision, Data curation, Formal analysis. Qianran Hong: Writing – review & editing, Writing – original draft, Validation, Funding acquisition, Formal analysis. Yuxuan Song: Writing – review & editing, Validation, Funding acquisition, Data curation,. Yibo He: Writing – review & editing, Conceptualization, Supervision.

Acknowledgements

Not Applicable.

Funding information

This work was supported by Innovation Fund for Qutstanding Doctoral Candidates of Peking University Health Science Center (BMU2024BSS001).

Ethics approval and consent to participate

Not Applicable.

Competing Interests

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

The data that supports the findings of this study are available in the supplementary material of this article

References