Skip to main content

Integrative multi-omics analysis identifies genetically supported druggable targets and immune cell specificity for myasthenia gravis



Myasthenia gravis (MG) is a chronic autoimmune disorder characterized by fluctuating muscle weakness. Despite the availability of established therapies, the management of MG symptoms remains suboptimal, partially attributed to lack of efficacy or intolerable side-effects. Therefore, new effective drugs are warranted for treatment of MG.


By employing an analytical framework that combines Mendelian randomization (MR) and colocalization analysis, we estimate the causal effects of blood druggable expression quantitative trait loci (eQTLs) and protein quantitative trait loci (pQTLs) on the susceptibility of MG. We subsequently investigated whether potential genetic effects exhibit cell-type specificity by utilizing genetic colocalization analysis to assess the interplay between immune-cell-specific eQTLs and MG risk.


We identified significant MR results for four genes (CDC42BPB, CD226, PRSS36, and TNFSF12) using cis-eQTL genetic instruments and three proteins (CTSH, PRSS8, and CPN2) using cis-pQTL genetic instruments. Six of these loci demonstrated evidence of colocalization with MG susceptibility (posterior probability > 0.80). We next undertook genetic colocalization to investigate cell-type-specific effects at these loci. Notably, we identified robust evidence of colocalization, with a posterior probability of 0.854, linking CTSH expression in TH2 cells and MG risk.


This study provides crucial insights into the genetic and molecular factors associated with MG susceptibility, singling out CTSH as a potential candidate for in-depth investigation and clinical consideration. It additionally sheds light on the immune-cell regulatory mechanisms related to the disease. However, further research is imperative to validate these targets and evaluate their feasibility for drug development.


Myasthenia gravis (MG) is a chronic autoimmune disorder of the neuromuscular junction [1]. The clinical hallmark of MG is muscle weakness associated with fatigability, which can lead to potentially life-threatening exacerbations, such as myasthenic crisis, which affects around 15% of individuals with MG, remains the leading cause of mortality among patients [2, 3]. MG is a rare disease with an annual incidence of roughly 10–29 cases per 1 million people and a prevalence ranging from 100 cases to around 350 cases per 1 million people [2]. Unlike congenital myasthenic syndromes, characterized by mutations in different genes encoding molecules important in the neuromuscular junction cause major changes in function and are inherited in classic mendelian patterns [4], it can be broadly stated that MG is a complex disorder resulting from the interplay of genetic and environmental factors, triggering autoimmune responses [5]. The underlying genetic pathogenesis is evidenced by the high disease concordance among identical twins [6], and associations with genes in the major histocompatibility complex (MHC) locus have been recognized for more than 30 years [7]. Over the past few decades, the emergence of genome-wide association studies (GWAS) has identified multiple susceptibility variants beyond the MHC with MG risk, yielding an estimated heritability of 25.6% [8]. In a recent GWAS encompassing 1,873 MG patients and 36,370 healthy individuals, identified significant associations in the CHRNA1 and CHRNB1 genes, as well as confirmed the previous association signals at PTPN22, HLA-DQA1/HLA-B and TNFRSF11A [9]. These discoveries shed light on the intricate genetic landscape of MG and provide valuable insights into its underlying mechanisms.

The goal of MG treatment is to achieve complete remission or minimal manifestation status with minimal side effects and eventually to avoid a myasthenic crisis [10]. Despite the availability of standard therapies, including acetylcholinesterase inhibitors, steroids, steroid-sparing immunosuppressants, and thymectomy, symptoms of MG are unsatisfactorily treated in up to half of individuals over the course of their disease [11]. A significant proportion of patients heavily rely on corticosteroid administration, which can result in severe side effects, including infections, osteoporosis, diabetes, glaucoma, and other complications [11]. Furthermore, some patients exhibit inadequate response to conventional treatment, with approximately 10–20% of MG patients classified as having "refractory" MG, emphasizing the pressing demand for innovative therapeutic solutions. Although there are several novel treatment options for MG, the therapeutic aim of complete remission only be achieved in a subset of patients [12], indicating that new safe and effective immunotherapies are desperately needed.

The conventional process of drug discovery and development is a time-consuming and costly endeavor. The integration of genomics into the drug discovery process has become indispensable, providing a vital avenue for expediting the development of novel therapeutic targets [13]. The combination of molecular quantitative trait locus (molQTL) studies, such as gene expression or protein quantitative trait loci (eQTLs or pQTLs), with genome-wide association (GWAS) data allows for the identification of target genes associated with risk variants through causal inference [14]. One approach is through drug target Mendelian randomization (MR), a statistical genetic methodology that leverages genetic variants as instrumental variables to assess the causal relationship between an exposure (like genetically predicted druggable gene expression or protein levels) and a specific outcome (such as MG risk). This approach employs genetic data to simulate the design of a randomized controlled trial (RCT) without requiring a drug intervention (Additional file 1: Figure S1) [15]. By synergistically amalgamating diverse data sources [16,17,18,19,20,21,22] (Additional file 2: Table S1) and employing rigorous MR and colocalization analyses, this study endeavor strives to identify potential repurposing opportunities for MG, delving into their potential implications in MG susceptibility. Subsequently, further investigation was conducted on the MR associations that exhibited statistical significance and provided evidence for colocalization, aiming to identify immune-cell-specific effects.


Identification of actionable druggable genes

The druggable genome was defined as described in Finan et al. as a selection of genes [23], which included 4479 genes and divided into 3 Tiers based on druggability levels: Tier 1 contains genes encoding targets of approved or clinical trial drugs, Tier 2 genes encoding targets with high sequence similarity to Tier 1 proteins or targeted by small drug-like molecules, and Tier 3 contains genes encoding secreted and extracellular proteins, genes belonging to the main druggable gene families, and genes encoding proteins with more restricted similarity to Tier 1 targets. After removing duplicate and non-autosomal genes, 4300 of 4479 druggable genes were retained in the subsequent analysis (Additional file 2: Table S2).

Selection of eQTL genetic instruments for drug target gene expression

To simulate exposure to the corresponding drugs, we sought publicly accessible eQTL data for gene expression that encode these druggable proteins. Publicly available data from the eQTLGen consortium (, n = 31,684) was utilized to identify common (minor allele frequency > 1%) single-nucleotide variants (SNVs) associated with the expression of drug target genes in blood [16]. We retrieved the complete cis-eQTL results (distance between SNV and gene is < 1 Mb, FDR < 0.05) and allele frequency data from the consortium. The SNPs that were associated with gene expression (cis-eQTL P < 5 × 10–8) were selected as genetic instrumental variables (IVs). To obtain independent IVs, we conducted LD clumping using the TwoSampleMR R package with genotype data of Europeans from the 1000 Genomes were used as a reference panel. In a 10 Mb window, if LD values (r2) of two or more SNPs were smaller than 0.01, these SNPs were considered independent IVs.

Deriving pQTL genetic determinants of circulating protein levels

We obtained pQTL data from six large-scale genome-wide proteomic GWAS studies, namely the ARIC study [17], the INTEVAL study [18], the KORA F4 study [19], the IMPROVE study [20], the AGES-Reykjavik study [21], and the Framingham Heart study (FHS) [22]. Each of these studies undertook proteomic profiling using either SomaLogic SomaScans or O-link proximal extension assays. We restricted proposed instrumental variants to cis-pQTLs for druggable proteins, used a P value threshold of 5 × 10–8. For proteins derived from the ARIC study [17], we utilized the sentinel cis-pQTL specific to each protein as it was available. Regarding proteins from the other five studies [18,19,20,21,22], we employed lead variants categorized as tier 1 instrumental variants according to Zheng et al [24], which were associated with fewer than five proteins and exhibited no heterogeneity across the studies.

Immune-cell-type-specific eQTL data

The immune-cell-specific RNA expression and eQTL data were acquired from the Database of Immune Cell Expression (DICE, [25], which included eQTLs from 15 different immune cell types from 91 healthy subjects. The presented cell types account for over 60% of all circulating mononuclear cells, consisting of three innate immune cell types (Classical monocytes, Non-classical monocytes, NK cells), four adaptive immune cell types that have not encountered cognate antigen in the periphery (Naive B cells, Naive CD4+ T cells, Naive CD8+ T cells, and Naive TREG T cells), six CD4+ memory or more differentiated T cell subsets (TH1, TH1/17, TH17, TH2, follicular TFH, and Memory TREG cells), and two activated cell types (Activated Naive CD4+ cells and Activated Naive CD8+ T cells) [25]. We used these datasets specifically for follow-up analyses of genetically predicted effects identified to evaluate cell-type specificity.

GWAS summary statistics of MG

For the primary analysis, the largest MG GWAS reported by Chia et al. was used in this study. Briefly, Chia et al. performed a large-scale GWAS analysis on MG, which included 1,873 patients and 36,370 controls [9] (, GWAS Catalog ID: GCST90093061). In this study, the diagnosis of MG relied on standard clinical criteria, including characteristic fatigable weakness, supported by electrophysiological and/or pharmacological abnormalities. Notably, the study is limited to anti-acetylcholine receptor antibodies (anti-AChR) positive cases, and individuals testing positive for antibodies to muscle-specific kinase (anti-MuSK) were excluded. We summarized the genome-wide significant loci identified in the MG GWAS conducted by Chia et al. in Additional file 2: Table S3. To explore age-dependent genetic heterogeneity in MG, we utilized summary statistics from early-onset MG (GWAS Catalog ID: GCST90093465; 595 cases vs. 2,718 controls, aged 40 years or younger) and late-onset MG (GWAS Catalog ID: GCST90093466; 1,278 cases vs. 33,652 controls) separately. For external validation, summary statistics were obtained from the UK Biobank ( (224 MG cases vs. 417,332 controls) and FinnGen Biobank ( (461 MG cases vs. 408,430 controls). The MG phenotype was identified through questionnaires completed by the participants, and data on MG subtypes were not applicable. Detailed information on various GWAS datasets is provided in Table 1. To ensure data integrity, we removed SNPs with duplicate or missing identification (rsID) from the dataset for subsequent analysis.

Table 1 Details of GWAS summary data and LDSC of single-trait heritability

SNP-based heritability calculation

We used Linkage Disequilibrium Score (LDSC) Regression ( to estimate the SNP-based heritability (h2) of each of the trait, representing the proportion of phenotypic variance that is explained by all common genetic variants included in the analysis. Also, we used single-trait LDSC to estimate genomic inflation factors λGC, used to evaluate the polygenicity and confounding because of population stratification or cryptic relatedness. To minimize bias caused by low interpolation quality, we restricted this analysis to haplotype map 3 SNPs.

Mendelian randomization analysis

The two-sample MR approach was based on the following assumptions: (i) the genetic variants used as an instrumental variable are associated with target exposure, i.e. gene expression levels and protein levels; (ii) there are no unmeasured confounders of the associations between genetic variants and outcome; (iii) the genetic variants are associated with the outcome only through changes in the exposure, i.e. no pleiotropy. We therefore used a curated genotype–phenotype database (PhenoScanner) [26] to search for associations between variants used to instrument each drug target and other traits that may represent pleiotropic pathways. We used fixed-effects, inverse-variance-weighted MR for proposed instruments that contain more than one variant, and Wald ratio for proposed instruments with one variant. Steiger filtering was applied to exclude variants that were potentially influenced by reverse causation [27]. For proposed instruments with multiple variants, we assessed the heterogeneity across variant-level MR estimates using the Cochrane Q method (mr_heterogeneity option in TwoSampleMR package). Benjamini–Hochberg false discovery rate (FDR) of 0.05 was applied to select best MR estimates with robust signals. The findings of our MR studies are presented as MR estimates (β) or odds ratio (OR) and 95% confidence interval (CI) for the risk of MG per genetically predicted 1-standard deviation (SD) increase in blood gene expression or circulating protein level. We conducted MR analyses using the TwoSampleMR ( package.

Colocalization analysis

While MR largely mitigates bias from confounding, linkage disequilibrium (LD) between SNPs might be an important source of noncausal associations. For statistically significant MR results, coloc ( was used to evaluate the probability of QTL and MG loci sharing a single causal variant [28], which assess potential confounding by LD. To determine the posterior probability of each genomic locus containing a single variant affecting both the gene/protein and the MG risk, we analyzed all SNPs within 1 Mb of the cis-eQTL/cis-pQTL. Moreover, we evaluated whether each genomic locus harbored a causal variant that influenced both disease risk and the variability in gene expression across the 15 cell-type-specific datasets individually. Assuming a solitary causal variant, four hypotheses can be outlined: H0, proposing the lack of causal variants for both traits; H1, positing the existence of a causal variant for trait 1; H2, suggesting a causal variant for trait 2; H3, postulating two distinct causal variants for traits 1 and 2; and H4, proposing a shared causal variant between the two traits. Statistically significant MR hits with posterior probability for hypothesis 4 (PPH4) > 0.8 (the probability of a shared causal variant) were investigated. Visualization of colocalization results was performed using the LocusCompareR R package [29].


Overall analysis plan

The study design is illustrated in Fig. 1. Initially, we identified druggable proteins as described in Finan et al. [23]. These proteins include targets of approved and clinical-phase drugs, proteins resembling approved drug targets and proteins accessible to monoclonal antibodies or drug-like small molecules in vivo. Next, we selected independent genetic variants that act locally on gene expression from eQTLGen Consortium or specifically influence plasma levels of the proteins from six large proteomic GWASs of individuals of European ancestry. The primary analysis involved retrieving summary statistics from the largest MG GWAS dataset of European ancestry [9]. In addition, we conducted a subgroup analysis using GWAS summaries for early-onset MG and late-onset MG separately. Employing the single-trait LDSC method, the SNP heritability for MG was estimated at 0.075 (SE = 0.015), while early-onset MG showed a heritability of 0.3649 (SE = 0.135), and late-onset MG exhibited a heritability of 0.078 (SE = 0.015) (Table 1). Then, genetic colocalization was conducted on MR results that surpassed our significance threshold after accounting for multiple testing, to mitigate potential confounding by LD. Afterward, we performed external replication using MG GWAS summary statistics from the FinnGen Consortium and UK Biobank. Subsequently, we explored whether putative genetic effects may be cell type specific by employing genetic colocalization to assess the interplay between immune-cell-specific eQTLs and MG risk.

Fig. 1
figure 1

Flow diagram of study design. Using a variety of data sources, this study examined the instruments proposed for actionable druggable proteins, specifically cis-pQTL and cis-eQTL, against MG GWAS summary statistics. Subsequently, further investigation was conducted on the MR associations that exhibited statistical significance and provided evidence for colocalization, aiming to identify immune-cell-specific effects

MR analysis with blood gene expression and MG outcome

We used two-sample MR to systematically evaluate the evidence for the causal effects of druggable gene expression on MG outcome. Using cis-eQTLs as proposed instruments available from the eQTLGen Consortium, four genes (CDC42BPB, TNFSF12, CD226 and PRSS36) showed significant MR results (Table 2, Fig. 2). Specifically, we found that a 1 standard deviation (SD) increased blood expression of CDC42BPB (OR = 1.694; 95% CI 1.361–2.108; P = 2.347 × 10–6), TNFSF12 (OR = 1.433; 95% CI 1.214–1.691; P = 2.119 × 10–5), and PRSS36 (OR = 3.186; 95% CI 1.805–5.624; P = 6.401 × 10–5) were significantly associated with increased MG susceptibility, whereas increased CD226 gene expression (OR = 0.652; 95% CI 0.528–0.804; P = 6.205 × 10–5) in blood were associated with decreased MG risk (cis-eQTL instruments in Additional file 2: Table S4, full MR results in Additional file 2: Table S5, MR scatter plot shown in Additional file 1: Figure S2). We employed Cochran's Q test to assess potential heterogeneity in the IVW results, estimating MR effects across each eQTL and none of the MR signals exhibited evidence of significant heterogeneity. In the replication phase, our study employed GWAS summary data from the UK and FinnGen Biobank datasets. Although MR analysis did not identify any genetically predicted gene expression causally linked to MG risk after FDR correction (full MR results in Additional file 2: Table S5), we observed similar pattern for CD226 expression with MG susceptibility (OR = 0.595; 95% CI 0.403–0.877; P = 8.754 × 10–03) (Additional file 1: Figure S3).

Table 2 MR-identified genes/proteins with MG risk
Fig. 2
figure 2

Miami plot with circles representing the MR results for gene/protein on MG. FDR, P value (FDR adjust); Black dashed line indicates the threshold for significance (FDR < 0.05) threshold. The x axis is the chromosome and gene start position of each MR finding in the cis region. The y axis represents the − log10 FDR of the MR findings. MR findings with positive effects (increased level of gene expression/protein associated with increasing the MG risk) are represented by filled circles on the top half of the plot; Conversely, MR findings suggesting a negative effect, implying a reduced level of gene expression or protein linked to an elevated risk of MG, are illustrated in the lower half of the plot

In the subgroup analysis discerning early-onset and late-onset MG, coherent patterns emerged (Fig. 3), For instance, a 1 SD increase in genetically predicted CDC42BPB expression levels was associated with elevated risk for both early-onset MG (OR = 2.065; 95% CI 1.322–3.225; P = 1.433 × 10–3) and late-onset MG (OR = 1.578; 95% CI 1.191–2.092; P = 1.506 × 10–3) risk. Furthermore, pertaining to CD226 and TNFSF12, subgroup MR findings revealed a relationship between CD226 levels and late-onset MG (OR = 0.621; 95% CI 0.451–0.854; P = 3.412 × 10–3), as well as an association between TNFSF12 levels and late-onset MG (OR = 1.533; 95% CI 1.243–1.891; P = 6.570 × 10–5), while no statistically significant MR estimates were observed in the context of early-onset MG.

Fig. 3
figure 3

Forest plot showing MR estimate for genetically proxied gene and protein expression on MG and its subgroups. Forest plot showing MR estimate (95% CI) from two sample MR analyses. P are unadjusted. CI confidence interval. MR estimates of significant MR results used cis-eQTL instruments on MG and its subgroups. MR estimates of significant MR results used cis-pQTL instruments on MG and its subgroups

We then performed genetic colocalization analyses to evaluate the probability of shared single causal variants between MG loci and eQTLs. Three loci (CDC42BPB, CD226, and PRSS36) exhibited profiles suggesting causal relationships, except for TNFSF12 (Fig. 4, Additional file 2: Table S6). Furthermore, we investigated the associations between index cis-eQTLs (specifically, rs1790974 for CD226 levels, rs10143668 for CDC42BPB levels, and rs78924645 for PRSS36 levels) and more than 5,000 other diseases, traits or proteins levels as documented in PhenoScanner [26] (Additional file 1: Figure S4), which suggest potential horizontal pleiotropic effects in our MR analysis.

Fig. 4
figure 4

LocusCompare plot depicting colocalization of the top SNP associated with eQTL surrounding CDC42BPB (A), PRSS36 (B) and CD226 (C) and MG GWAS. The top right plots show the association results in the MG GWAS; the bottom right plots represent the corresponding eQTL results; the left plot shows the colocalization of genetic association and eQTL signals. The SNP indicated by the purple diamond is the SNP for which the European LD information is shown

MR analysis of circulating proteome identifies actionable targets in MG

Using proposed cis-pQTL instruments, MR analyses revealed significant associations between the levels of three circulating proteins (PRSS8, CTSH, CPN2) and MG susceptibility in the primary analysis, after FDR correction (cis-pQTL instruments in Additional file 2: Table S7, and full MR results in Additional file 2: Table S8), which exhibit compelling evidence of colocalization with MG risk (Fig. 5, Additional file 2: Table S9). Notably, among these proteins, circulating CTSH exhibited associations that were replicated in two independent proteomic studies (Fig. 2, Table 2). The estimated associations with MG per 1 SD increase in the genetically predicted circulating CTSH level were consistent between the INTERVAL (OR = 1.257; 95% CI 1.125–1.405; P = 5.49 × 10–4) and ARIC (OR = 1.218; 95% CI 1.101–1.348; P = 1.330 × 10–4) studies and displayed strong evidence of colocalization (Fig. 5). Despite the MR analysis failing to identify any genetically predicted circulating protein levels associated with MG susceptibility using the proposed cis-pQTL instruments and GWAS summary data from the FinnGen and UK Biobank datasets after FDR correction, a persistent association for Cathepsin H (CTSH) abundance abundance with MG was evident. This sustained association suggests an elevated risk of MG across three MG outcomes. A SD increase in genetically predicted circulating CTSH abundance was found to be associated with an elevated risk of MG (minimum OR = 1.218; 95% CI 1.101–1.348; P = 1.329 × 10–4), and this trend was consistently observed in both the UK Biobank (OR = 1.275; 95% CI 1.003–1.620; P = 4.713 × 10–2) and FinnGen Biobank (OR = 1.185; 95% CI 1.017–1.381; P = 2.942 × 10–2) datasets (Additional file 1: Figure S3).

Fig. 5
figure 5

LocusCompare plot depicting colocalization of the top SNP associated with pQTL surrounding CPN2 (A), CTSH (B) and PRSS8 (C)and MG GWAS. The top right plots show the association results in the MG GWAS; the bottom right plots represent the corresponding pQTL results; the left plot shows the colocalization of genetic association and eQTL signals. The SNP indicated by the purple diamond is the SNP for which the European LD information is shown

Additionally, subgroup analyses can provide further insights into the nuanced relationships between these proteins and MG susceptibility (Fig. 3). Genetically predicted circulating CTSH abundance was associated with an increased risk of late-onset MG (OR = 1.256; 95% CI 1.110–1.421; P = 2.957 × 10–4). Also, genetically predicted elevation in circulating PRSS8 abundance is associated with an increased risk of MG (OR = 4.252; 95% CI 2.275–7.945; P = 5.69 × 10–6), as well as heightened risks for both early-onset MG (OR = 8.916; 95% CI 2.558–31.083; P = 5.948 × 10–6) and late-onset MG (OR = 3.477; 95% CI 1.662–7.275; p = 9.379 × 10–4).Conversely, a genetically proxied increase in circulating CPN2 levels demonstrated an association with decreased MG risk (OR = 0.819; 95% CI 0.739–0.908; P = 1.444 × 10−4) and also late-onset MG (OR = 0.798; 95% CI 0.706–0.903; P = 3.282 × 10–04). Due to the limited number of cis-acting pQTLs available for each protein, applying classic sensitivity methods to test MR assumptions becomes challenging.

We also searched in the PhenoScanner database to assess potentially pleiotropic effects of the cis-pQTLs of MR-prioritized proteins by testing the association of cis-pQTLs with other diseases or traits [26] (Additional file 1: Figure S4). We observed no evidence of pleiotropic effects for the cis-pQTL associated with CPN2 (rs11711157). We discovered an association between the CTSH cis-pQTL (rs34593439) and narcolepsy, a condition for which immune-mediated dysregulation has been considered as one of the potential causes [30]. Also, we found that PRSS8 cis-pQTL (rs1060506) was associated with a broad spectrum of weight-related traits, including whole body fat mass and body fat percentage, among others, which was consistent with the findings for cis-eQTL for PRSS36. This suggests that our MR estimate of the effect of the above-mentioned proteins on MG susceptibility may have been biased by the stated confounders.

Identifying immune-cell-specific effects

Given the pivotal role of immune cells in the pathogenesis of MG, we aim to investigate whether potential genetic effects exhibit specificity toward certain cell types. Leveraging immune cell eQTL datasets from the DICE database, we conducted a comprehensive genetic colocalization analysis for each of the six identified loci (CDC42BPB, CD226, PRSS36, PRSS8, CTSH, and CPN2) from the preceding analysis. Remarkably, we identified robust evidence of colocalization, with a posterior probability of 0.854, linking CTSH expression in TH2 cells and MG risk (Additional file 2: Table S10, Additional file 1: Figure S5). TH2 cells are renowned for their capacity to produce cytokines that stimulate B cell activation and differentiation of B cells, essential for antibody production [31]. Furthermore, while analyzing data from other immune cell subsets, we observed less pronounced evidence specific to individual cell types (Fig. 6). The identification of a colocalization signal within a cell type relevant to MG pathogenesis suggests a potential direct contribution of the genetic variant to disease development by influencing gene expression within that specific cell subset.

Fig. 6
figure 6

Colocalization results between GWAS and eQTLs across immune cell types. PPH4 of shared genetic signal between GWAS and eQTLs for MG across different immune cell types


Gaining an in-depth understanding of the intricate relationship between genetic discoveries and pharmaceutical targets holds paramount importance in effectively translating GWAS findings into clinical applications [32]. Our study assumes pioneering significance as the first to center on actionable druggable genes, unlocking the potential of drug repurposing strategies within the domain of MG treatment. Notably, we identified three genes (CDC42BPB, CD226, and PRSS36) and three proteins (PRSS8, CTSH, and CPN2) as having significant MR results using cis-eQTL and cis-pQTL genetic instruments, which also exhibit compelling evidence of colocalization with MG. Additionally, through a meticulous exploration of immune cell-specific effects, this study may shed light on mechanistic insights underlying the loci associated with MG.

Our MR analysis establishes a compelling link between genetic variants associated with increased CDC42BPB gene expression and an elevated MG susceptibility. This finding underscores the potential of CDC42BPB inhibitors as a promising avenue for therapeutic intervention. CDC42BPB is a serine/threonine protein kinase intricately involved in regulating actin cytoskeleton dynamics and cell contraction. Several small molecules or biological inhibitors of CDC42BPB, such as SR-7826, BDP8900 and BDP9066 [33], have exhibited antitumor activity. However, a comprehensive understanding of the mechanistic underpinnings through which elevated CDC42BPB contributes to the pathophysiology of MG necessitates further investigation. Conversely, we observed a protective effect associated with increased CD226 gene expression against MG. CD226, also known as DNAX accessory molecule-1 (DNAM-1), is a member of the immunoglobulin superfamily and is expressed on various immune cells, including natural killer (NK) and T cells [34]. CD226, along with the inhibitory receptors TIGIT and CD96, constitutes cell-surface receptor family 3, binding to nectin and nectin-like proteins [35]. Prior research emphasizes the potent roles of this receptor family in regulating tumor immunity [36]. These findings underscore the importance of maintaining the expression of the activating receptor CD226, in orchestrating effective immune responses. Genetic polymorphisms within the CD226 gene have been associated with serval autoimmune diseases, including multiple sclerosis [37]. Building on these insights, we hypothesize that targeted manipulation of CDC42BPB and CD226 expression could potentially serve as a potent therapeutic strategy for managing MG. Nonetheless, a comprehensive validation of these proposed therapeutic targets necessitates further rigorous investigation.

Our study emphasizes a consistent and robust association between genetically predicted circulating Cathepsin H abundance and an elevated risk of MG across diverse datasets, including the UK Biobank and FinnGen Biobank. The subgroup analyses further revealed nuanced insights, particularly the association between genetically predicted circulating CTSH abundance and an increased risk of late-onset MG. Additionally, the posterior probability that CTSH expression levels in TH2 cells and MG outcomes shared a single causal signal in the 1-Mb locus around the cis-pQTL, rs12148472, was 0.8854 for MG susceptibility (Additional file 1: Figure S5). This observation implies that CTSH abundance might serve as a biomarker or contribute to the biological mechanisms underlying MG. CTSH encodes cathepsin H, a member of the papain-like cysteine proteases that are involved in major histocompatibility complex class II antigen presentation. Notably, CTSH has garnered attention for its involvement in type 1 diabetes [38] and narcolepsy risk [39], both of which prominently feature autoimmune components. Worth noting is the fact that other members of the cathepsin family, such as cathepsin S and cathepsin K, have received more attention as drug targets for autoimmune diseases. For example, cathepsin S inhibitors have been investigated for their potential in treating diseases like and multiple sclerosis [40].

Furthermore, our study provided robust genetic evidence supporting a potential causal role of increased CPN2 protein levels in reducing the risk of MG. CPN2, also known as Carboxypeptidase N Subunit 2, forms a complex with enzymatically active small subunits (CPN1). CPN plays a pivotal role as a zinc metalloprotease responsible for cleaving and partially inactivating anaphylactic peptides, specifically complement component 3a (C3a) and C5a, within the classical and lectin pathways of complement activation [41]. Protein–protein interaction studies using the STRING database ( have shown interactions between CPN2 and C5 (Additional file 1: Figure S6). It is widely recognized that dysregulated complement activation is a primary pathogenic mechanism in MG. Notably, C5 inhibitors, such as Eculizumab, have proven effective in preventing complement-dependent membrane attacks at the neuromuscular junction, presenting a promising avenue for the treatment of MG [42]. Future research is warranted to investigate the potential therapeutic implications of CPN2 in MG.

A significant hurdle in our analytical pursuit lay in discerning whether the association with MG risk stemmed from PRSS36, PRSS8, or both entities, given the substantial correlation between the cis-eQTL instrument for PRSS36 (rs78924645) and cis-pQTL instrument (rs1060506) for PRSS8 (r2 = 0.93 in 1000 Genomes Project European ancestry participants). Considering that both PRSS36 and PRSS8 belong to the serine protease family, a group of proteolytic enzymes involved in diverse biological processes, we regard them as potential drug targets for MG. It is worth noting that PRSS8 inhibits TLR4-mediated inflammation in human and mouse models of inflammatory bowel disease [43], thereby implicating its plausible relevance to TLR4-mediated inflammation pathways in the context of MG. As such, the identification of PRSS36 and PRSS8 as putative targets holds promise; however, the trajectory towards their validation as credible therapeutic avenues warrants further inquiry, coupled with a comprehensive assessment of their viability for subsequent drug development endeavors.

Genetic variation controls transcriptional regulation in a cell type-specific manner to regulate immune pathways [44, 45]. The cell-type specificity of the effects elicited by common genetic variants on both gene and protein expression engenders a pronounced reliance on the precise cellular contexts. In navigating this intricacy, the relationships underpinning eQTL and pQTL manifest a profound dependence on the distinct cell types under scrutiny. In this study, we embarked on a comprehensive exploration, utilizing genetic colocalization analysis, to delineate the interplay between immune-cell-specific eQTLs and MG susceptibility, thereby unearthing the underlying molecular mechanisms. This approach yielded valuable insights into the genetic orchestration governing CTSH expression across an array of distinct cell types. Of note is the robust colocalization unveiled between CTSH expression within TH2 cells and the propensity for MG risk, signifying a putative involvement of CTSH expression within this cellular subtype in driving the genesis and advancement of MG. This revelatory aspect of our study underscores the imperative of acknowledging and integrating cell type heterogeneity within the realm of future research pursuits aimed at identifying potential therapeutic targets for MG. This newfound awareness stands poised to wield transformative influence in guiding the trajectory of future scientific explorations and therapeutic innovations targeting MG and other akin disorders.

The current study notably complements and extends previous efforts by employing key approaches to protect against potential biases, strengthen causal inference and enhance understanding of potential mechanisms. With our rigorous instrument selection process that used comprehensive datasets on gene expression and plasma protein levels, we have facilitated a thorough evaluation of actionable drug targets, notably including the previously unexplored CTSH. Our study, enriched with multi-omics data, has successfully overlooked targets that do not align with the potentially druggable gene targets prioritized in the Chia et al. study, as indicated by the Priority Index analysis [9, 46]. It is crucial to underscore that the peak cis-eQTL or cis-pQTL identified at each locus did not attain the genome-wide significance level threshold. Consequently, these loci were not deemed significantly associated with MG in Chia et al.'s study. This constitutes a noteworthy extension to previous study, emphasizing our pivotal role in contributing novel insights that surpass the confines of earlier investigations.

Several limitations need to be considered in our study. Firstly, only a small proportion of genes/proteins may be effectively instrumented by multiple SNPs, with the majority being instrumented by only two or one SNP. This limitation restricts our ability to conduct sensitivity analyses. Secondly, there is a potential concern related to the presence of epitope-binding artifacts caused by coding variants. These artifacts may introduce spurious signals, potentially leading to false-positive cis-pQTLs. Thirdly, our study is confined to AchR-positive MG cases, so caution is advised when extrapolating our results to patients with anti-MuSK and other autoantibodies. Additionally, although the GWAS summary data integrated into our study already constitute the most extensive MG GWAS dataset currently available, the limited number of cases in both the primary and replication datasets constrained our ability to replicate the MR estimates observed in the discovery cohort. Therefore, it is crucial to conduct further research with larger and more diverse populations, especially including non-European individuals and patients with anti-MuSK and other autoantibodies.


Overall, this study contributes valuable insights into the genetic and molecular factors associated with MG susceptibility, with CTSH emerging as a potential candidate for further investigation and clinical consideration. Moreover, by laying bare the intricacies of cell-specific genetic influences on gene and protein expression, our investigation serves as a clarion call for a nuanced consideration of cellular contexts in the quest for unraveling the underlying mechanisms of complex diseases like MG. It is imperative to underscore, however, that a requisite next step involves undertaking further studies, encompassing rigorous functional analyses, to definitively validate the viability of these targets and to ascertain their appropriateness for subsequent drug development endeavors.

Availability of data and materials

The eQTL data from eQTLGen consortium were obtained from The eQTL data from DICE were obtained from The ARIC pQTL summary statistics used in the paper are freely downloaded from ARIC website ( Summary statistics for the MG GWAS are publicly available for download from the GWAS catalog. Data processing was completed using R software (version 3.6.3), with packages TwoSampleMR (version 0.5.4), coloc (version 4.0.4), dplyr (version 1.0.0), readr (version 1.3.1), tidyverse (version 1.3.0), forestplot (version 1.9), plyr (version 1.8.6), devtools (version 2.3.0), remotes (version 2.1.1).


  1. Gilhus NE, Tzartos S, Evoli A, et al. Myasthenia gravis. Nat Rev Dis Primers. 2019;5(1):30.

    Article  PubMed  Google Scholar 

  2. Punga AR, Maddison P, Heckmann JM, et al. Epidemiology, diagnostics, and biomarkers of autoimmune neuromuscular junction disorders. Lancet Neurol. 2022;21(2):176–88.

    Article  PubMed  Google Scholar 

  3. Zhang C, Wang F, Long Z, et al. Mortality of myasthenia gravis: a national population-based study in China. Ann Clin Transl Neurol. 2023;10(7):1095–105.

    Article  PubMed  PubMed Central  Google Scholar 

  4. Finsterer J. Congenital myasthenic syndromes. Orphanet J Rare Dis. 2019;14(1):57.

    Article  PubMed  PubMed Central  Google Scholar 

  5. Avidan N, Le Panse R, Berrih-Aknin S, et al. Genetic basis of myasthenia gravis - a comprehensive review. J Autoimmun. 2014;52:146–53.

    Article  CAS  PubMed  Google Scholar 

  6. Ramanujam R, Pirskanen R, Ramanujam S, et al. Utilizing twins concordance rates to infer the predisposition to myasthenia gravis. Twin Res Hum Genet. 2011;14(2):129–36.

    Article  PubMed  Google Scholar 

  7. Zhong H, Zhao C, Luo S. HLA in myasthenia gravis: from superficial correlation to underlying mechanism. Autoimmun Rev. 2019;18(9):102349.

    Article  PubMed  Google Scholar 

  8. Renton AE, Pliner HA, Provenzano C, et al. A genome-wide association study of myasthenia gravis. JAMA Neurol. 2015;72(4):396–404.

    Article  PubMed  PubMed Central  Google Scholar 

  9. Chia R, Saez-Atienzar S, Murphy N, et al. Identification of genetic risk loci and prioritization of genes and pathways for myasthenia gravis: a genome-wide association study. Proc Natl Acad Sci U S A. 2022.

    Article  PubMed  PubMed Central  Google Scholar 

  10. Iorio R. Myasthenia gravis: the changing treatment landscape in the era of molecular therapies. Nat Rev Neurol. 2024.

    Article  PubMed  Google Scholar 

  11. Verschuuren JJ, Palace J, Murai H, et al. Advances and ongoing research in the treatment of autoimmune neuromuscular junction disorders. Lancet Neurol. 2022;21(2):189–202.

    Article  CAS  PubMed  Google Scholar 

  12. Narayanaswami P, Sanders DB, Wolfe G, et al. International consensus guidance for management of myasthenia gravis: 2020 Update. Neurology. 2021;96(3):114–22.

    Article  PubMed  PubMed Central  Google Scholar 

  13. Namba S, Konuma T, Wu KH, et al. A practical guideline of genomics-driven drug discovery in the era of global biobank meta-analysis. Cell Genom. 2022;2(10):100190.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  14. Kreitmaier P, Katsoula G, Zeggini E. Insights from multi-omics integration in complex disease primary tissues. Trends in genetics : TIG. 2023;39(1):46–58.

    Article  CAS  PubMed  Google Scholar 

  15. Davey SG. Capitalizing on Mendelian randomization to assess the effects of treatments. J R Soc Med. 2007;100(9):432–5.

    Article  Google Scholar 

  16. Võsa U, Claringbould A, Westra H-J, et al. Large-scale cis- and trans-eQTL analyses identify thousands of genetic loci and polygenic scores that regulate blood gene expression. Nat Genet. 2021;53(9):1300–10.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  17. Zhang J, Dutta D, Köttgen A, et al. Plasma proteome analyses in individuals of European and African ancestry identify cis-pQTLs and models for proteome-wide association studies. Nat Genet. 2022;54(5):593–602.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  18. Sun BB, Maranville JC, Peters JE, et al. Genomic atlas of the human plasma proteome. Nature. 2018;558(7708):73–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  19. Suhre K, Arnold M, Bhagwat AM, et al. Connecting genetic risk to disease end points through the human blood plasma proteome. Nat Commun. 2017;8:14357.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  20. Folkersen L, Fauman E, Sabater-Lleal M, et al. Mapping of 79 loci for 83 plasma protein biomarkers in cardiovascular disease. PLoS Genet. 2017;13(4):e1006706.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  21. Emilsson V, Ilkov M, Lamb JR, et al. Co-regulatory networks of human serum proteins link genetics to disease. Science. 2018;361(6404):769–73.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  22. Yao C, Chen G, Song C, et al. Genome-wide mapping of plasma protein QTLs identifies putatively causal genes and pathways for cardiovascular disease. Nat Commun. 2018;9(1):3268.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  23. Finan C, Gaulton A, Kruger FA, et al. The druggable genome and support for target identification and validation in drug development. Sci Trans Med. 2017.

    Article  Google Scholar 

  24. Zheng J, Haberland V, Baird D, et al. Phenome-wide Mendelian randomization mapping the influence of the plasma proteome on complex diseases. Nat Genet. 2020;52(10):1122–31.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  25. Schmiedel BJ, Singh D, Madrigal A, et al. Impact of genetic polymorphisms on human immune cell gene expression. Cell. 2018;175(6):1701-15.e16.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  26. Staley JR, Blackshaw J, Kamat MA, et al. PhenoScanner: a database of human genotype-phenotype associations. Bioinformatics. 2016;32(20):3207–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  27. Hemani G, Tilling K, Davey SG. Orienting the causal relationship between imprecisely measured traits using GWAS summary data. PLoS Genet. 2017;13(11):e1007081.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  28. Wallace C. Eliciting priors and relaxing the single causal variant assumption in colocalisation analyses. PLoS Genet. 2020;16(4):e1008720.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  29. Liu B, Gloudemans MJ, Rao AS, et al. Abundant associations with gene expression complicate GWAS follow-up. Nat Genet. 2019;51(5):768–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  30. Faraco J, Lin L, Kornum BR, et al. ImmunoChip study implicates antigen presentation to T cells in narcolepsy. PLoS Genet. 2013;9(2):e1003270.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  31. Walker JA, McKenzie ANJ. T(H)2 cell development and function. Nat Rev Immunol. 2018;18(2):121–33.

    Article  CAS  PubMed  Google Scholar 

  32. Carss KJ, Deaton AM, Del Rio-Espinola A, et al. Using human genetics to improve safety assessment of therapeutics. Nat Rev Drug Discovery. 2023;22(2):145–62.

    Article  CAS  PubMed  Google Scholar 

  33. East MP, Asquith CRM. CDC42BPA/MRCKα: a kinase target for brain, ovarian and skin cancers. Nat Rev Drug Discov. 2021;20(3):167.

    Article  CAS  PubMed  Google Scholar 

  34. Du X, de Almeida P, Manieri N, et al. CD226 regulates natural killer cell antitumor responses via phosphorylation-mediated inactivation of transcription factor FOXO1. Proc Natl Acad Sci U S A. 2018;115(50):E11731–40.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  35. Bi J. CD226: a potent driver of antitumor immunity that needs to be maintained. Cell Mol Immunol. 2022;19(9):969–70.

    Article  CAS  PubMed  Google Scholar 

  36. Weulersse M, Asrir A, Pichler AC, et al. Eomes-dependent loss of the co-activating receptor CD226 restrains CD8(+) T cell anti-tumor functions and limits the efficacy of cancer immunotherapy. Immunity. 2020;53(4):824-39.e10.

    Article  CAS  PubMed  Google Scholar 

  37. Piédavent-Salomon M, Willing A, Engler JB, et al. Multiple sclerosis associated genetic variants of CD226 impair regulatory T cell function. Brain. 2015;138(Pt 11):3263–74.

    Article  PubMed  Google Scholar 

  38. Fløyel T, Brorsson C, Nielsen LB, et al. CTSH regulates β-cell function and disease progression in newly diagnosed type 1 diabetes patients. Proc Natl Acad Sci U S A. 2014;111(28):10305–10.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  39. Ollila HM, Sharon E, Lin L, et al. Narcolepsy risk loci outline role of T cell autoimmunity and infectious triggers in narcolepsy. Nat Commun. 2023;14(1):2709.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  40. Link JO, Zipfel S. Advances in cathepsin S inhibitor design. Curr Opin Drug Discov Devel. 2006;9(4):471–82.

    CAS  PubMed  Google Scholar 

  41. Morser J, Shao Z, Nishimura T, et al. Carboxypeptidase B2 and N play different roles in regulation of activated complements C3a and C5a in mice. J Thromb Haemost. 2018;16(5):991–1002.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  42. DeHart-McCoyle M, Patel S, Du X. New and emerging treatments for myasthenia gravis. BMJ Med. 2023;2(1):e000241.

    Article  PubMed  PubMed Central  Google Scholar 

  43. Sugitani Y, Nishida A, Inatomi O, et al. Sodium absorption stimulator prostasin (PRSS8) has an anti-inflammatory effect via downregulation of TLR4 signaling in inflammatory bowel disease. J Gastroenterol. 2020;55(4):408–17.

    Article  CAS  PubMed  Google Scholar 

  44. Yazar S, Alquicira-Hernandez J, Wing K, et al. Single-cell eQTL mapping identifies cell type-specific genetic control of autoimmune disease. Science. 2022;376(6589):eabf3041.

    Article  CAS  PubMed  Google Scholar 

  45. Onuora S. Single-cell RNA sequencing sheds light on cell-type specific gene expression in immune cells. Nat Rev Rheumatol. 2022;18(7):363.

    Article  CAS  PubMed  Google Scholar 

  46. Chia R, Saez-Atienzar S, Drachman DB, et al. Implications of CHRNB1 and ERBB2 in the pathobiology of myasthenia gravis. Proc Natl Acad Sci U S A. 2022;119(36):e2209096119.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

Download references


The authors gratefully thank the International Myasthenia Gravis Genomics Consortium, the UK Biobank and the FinnGen study for providing publicly accessible MG GWAS summary statistics for this analysis. Special thanks are extended to the Chinese Institute for Brain Research (Beijing) for their assistance. Thanks for the support from the Chinese Institutes for Medical Research, Beijing. We extend our sincere appreciation to the authors of the referenced QTL projects for their generous and open sharing of data. We express our gratitude to all the patients and families whose voluntary decision to donate samples has made our research possible.


This study was supported by the National Natural Science Foundation of China (81825008), the National Key Research and Development Program of China (2021YFA1101403), the Beijing Municipal Public Welfare Development and Reform Pilot Project for Medical Research Institutes (JYY2023-7), the Project for Innovation and Development of Beijing Municipal Geriatric Medical Research Center (11000023T000002041657), and the Project of Construction and Support for High-level Innovative Teams of Beijing Municipal Institutions (BPHR20220112), Youth Beijing Scholar NO.020, Dengfeng Talent Program (DFL20220701), the China Postdoctoral Science Foundation (2023M732407), and the Beijing Postdoctoral Research Foundation (2023-ZZ-004). The funders played no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Author information

Authors and Affiliations



JH and GC had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. JL and FW contributed equally to this work, and contributed equally to this work. Concept and design: JL, GC, JH. Drafting of the manuscript: JL, FW, ZL. Critical revision of the manuscript for important intellectual content: JH, YM. Statistical analysis: JL, FW, JF, JX. Obtained funding: JL, GC, JH. Administrative, technical, or material support: CZ, YH, TC, YZ, SZ. Supervision: YD, GC, JH. All authors approved the final version of the manuscript.

Corresponding authors

Correspondence to Guoliang Chai or Junwei Hao.

Ethics declarations

Ethics approval and consent to participate

This study is covered under the Xuanwu Hospital, Capital Medical University, Human Research Ethics Committee approval ([2022]-148).

Consent for publication

Not applicable.

Competing interests

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

Additional information

Publisher's Note

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

Supplementary Information

Additional file 1: Figure S1.

Mendelian Randomization (MR) vs. Randomized controlled trial (RCT). Figure S2. Scatter plots of significant results from genetically predicted gene expression on MG risk in the primary analysis. Figure S3. Forest plot showing MR estimate for genetically proxied gene and protein expression on MG outcome across different datasets. Figure S4. PhenoScanner disease/trait annotation of the index eQTL/pQTL instrument with other traits. Figure S5. LocusCompare plot depicting colocalization of the top SNP associated with eQTL surrounding CTSH in TH2 cell and MG GWAS. Figure S6. Protein–protein interaction of CPN2 using the STRING database (

Additional file 2: Table S1.

Basic Characteristics of eQTL Databases, pQTL Studies, and GWAS Datasets in the study. Table S2. An overview of druggable proteins and the coverage of genes/proteins in eQTLGen and pQTL studies. Table S3. Genome-wide significant loci in MG GWAS from Chia et al. Table S4. The cis-eQTLs instruments used for drug target gene expression on MG risk in the primary analysis. The outcome is MG GWAS from Chia et al. (1,873 patients and 36,370 controls). Table S5. MR full results using cis-eQTLs on MG risk across different datasets. Table S6. Colocalization results using cis-eQTLs on MG across different datasets. Table S7. The cis-pQTLs instruments used for protein expression on MG risk in the primary analysis. The outcome is MG GWAS from Chia et al. (1,873 patients and 36,370 controls). Table S8. MR full results of proteins using sentinel cis-pQTLs from six large proteomic studies on MG risk across different datasets. Table S9. Colocalization results using cis-pQTLs on MG risk across different datasets. Table S10. Colocalization results for genes/proteins using eQTL from the DICE database.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit The Creative Commons Public Domain Dedication waiver ( applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Li, J., Wang, F., Li, Z. et al. Integrative multi-omics analysis identifies genetically supported druggable targets and immune cell specificity for myasthenia gravis. J Transl Med 22, 302 (2024).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: