Integrating human plasma proteomes with genome-wide association data implicates novel proteins and drug targets for Rheumatoid arthritis

doi:10.21203/rs.3.rs-4974860/v1

Download PDF

Research Article

Integrating human plasma proteomes with genome-wide association data implicates novel proteins and drug targets for Rheumatoid arthritis

https://doi.org/10.21203/rs.3.rs-4974860/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Background

Genome-wide association studies (GWAS) have uncovered over 100 loci associated with Rheumatoid arthritis (RA) risk. However, how these loci contribute to RA risk remains largely unknown, which has hampered the development of new therapeutics. To identify genes contributing to RA risk through their effects on protein abundance, we conducted the first large-scale proteome-wide association study (PWAS) by integrating the largest up-to-date RA GWAS results with human plasma proteomes.

Methods

The PWAS was perform using RA GWAS summary statistics from discovery (22,350 RA cases and 74,823 controls) and replication (31,313 RA cases and 995,377 controls) cohorts, by leveraging precomputed protein expression weights generated from ARIC (N = 7,213) and INTERVAL (N = 3,301) studies. Then, Mendelian randomization (MR) and colocalization analyses were employed to investigate causal relationships between PWAS proteins and RA. Druggable targets exploration were finally conducted to prioritize potential therapeutic targets for RA.

Results

We identified 21 genes whose genetically regulated protein abundances were associated with RA risk. Of note, 10 genes were potentially causal and were prioritized as candidate RA genes. Among the 10 causal genes, six genes (OLFML3, PAM, ICOSLG, FCRL3, ERAP2, IL6R) were also associated to RA at transcriptome level, including the three novel genes (ICOSLG, FCRL3, ERAP2) that were not implicated in the original GWAS, which were regarded as novel candidate genes for RA. Druggable targets exploration identified 120 drug-gene interactions involving in 7 causal PWAS genes, including seven drugs or compounds targeting novel PWAS gene ERAP2 and ICOSLG, which possesses superior anti-inflammatory and anti-rheumatic activity in autoimmune diseases, hence might be candidates for treating RA.

Conclusions

Our results provide novel insights into RA pathogenesis and promising targets for further mechanistic investigations and drug development of RA.

Rheumatoid arthritis (RA) is the most prevalent autoimmune disease characterized by persistent synovitis and systemic inflammatory responses, which affects up to 1% of the population[1]. As the pathogenesis has not been fully clarified, RA can only be treated by relieving pain and reducing inflammation, and it generally necessitates lifelong therapy[2]. The etiology of RA involves a variety of risk factors, such as genetics, gender, and environment, of which genetic factors account for more than 60%[3]. Since there is a clear causal relationship between genetic factors and RA, it is significant to understand the pathologic process from the genomic perspective. Over the last decade, genome-wide association studies (GWASs) have identified over 100 loci robustly associated with RA, making significant progress and invaluable utility in initially revealing the genetic architecture of RA[4]. Nevertheless, these variants only possess a small effect size on RA, and it remains challenging to decipher the molecular mechanisms underlying the identified genetic effects, which has impeded the advancement of new therapeutics[5].

Over 80% risk variants associated with RA are located in non-coding regions of the genome[4], which are enriched in various elements related to the regulation of gene function, indicating that these variants likely contribute to the underlying pathogenesis by regulating gene expression[6]. Recently, the transcriptome-wide association studies (TWAS) have been extensively proposed as an improved approach to investigate the associations between genetically predicted gene expression and disease risk, showing elevated power in prioritizing disease risk genes[7–10]. Taking advantage of TWAS, some studies have prioritized RA candidate genes[11, 12], but these mediating effects of gene expression can only explain a small fraction of GWAS heritability[13]. As the ultimate products of gene expression, proteins are closely related to the biological process of pathogenesis, which act as a major source of efficient biomarkers and therapeutic targets[14]. Gene expression is regulated at multiple levels including pre-transcriptional, translational, and post-translational, and studies have revealed that there is a notoriously poor correlation between mRNA and protein expression levels[15–17]. Furthermore, genetic variation can affect protein abundance by altering the rate and stability of gene expression, but these effects tend to be significantly reduced during translation process, indicating that their potential effects on downstream phenotypes may be further attenuated or buffered[18]. Therefore, it is critical to investigate whether the identified risk variants contribute to the pathogenesis of RA by regulating protein abundance.

With the widespread utilization of proteome sequencing technology, numerous protein quantitative trait loci (pQTLs) have been identified[19–21], which are more likely to exert effects on protein abundance by directly influencing the processes of translation, providing novel clues for exploring the causal associations between proteins and phenotypic variations from a genetic perspective. Inspired by TWAS, the proteome-wide association study (PWAS) has been established to explore the associations between protein expressions and phenotypes[22]. PWAS leverages an external reference panel that measures both genetic variation and protein expression to train a predictive model of protein expression, enabling the estimation of association statistics between predicted protein expression and phenotype. PWAS has been empirically verified and shown to be reliable in various complex diseases and traits, such as Alzheimer's disease, heart failure, and colorectal cancer[23–25].

In the current study, to identify candidate proteins associated with RA, we performed PWAS by integrating the largest up-to-date RA GWAS[26] with two independent human plasma pQTL datasets[19, 20]. Then, Mendelian randomization and colocalization analysis were employed to investigate causal relationships between specific proteins and RA. In addition, we performed TWAS and cell-type-specific expression to reveal the expression patterns of potential RA proteins identified by PWAS in different cell types. Finally, we explored the potential drug targets for RA by analyzing PWAS findings using the Drug-Gene Interaction Database (DGIdb). Our findings provide a novel perspective to identify RA-associated risk proteins as well as promising targets for the further research of pathogenesis and the development of new therapeutics.

GWAS summary statistics of RA

The GWAS summary data of RA was derived from the most recent large-scale GWAS meta-analysis of RA reported by Ishigaki and colleagues[26], which included 22,350 RA cases and 74,823 controls from 25 independent cohorts with European descent. All RA cases in each cohort satisfied the 1987 American College of Rheumatology (ACR) criteria[27] or the 2010 ACR/European League Against Rheumatism criteria[28], or were diagnosed with RA by a specialist rheumatologist. All cohort were obtained following protocols approved by their institutional ethics committees and with the informed consent of all participants. Detailed information of genotyping, imputation and association test can be found in the original literature[26].

To increase confidence of PWAS results, the replication PWAS of RA was performed using proteomes and independent GWAS summary data. The replication GWAS summary data was obtained from a resent published GWAS of RA, which contains 31,313 RA cases and ~ 1 million controls from six countries in Northwestern Europe[29]. RA cases were diagnosed by rheumatologists and/or captured through the nationwide Scandinavian rheumatology quality registries and/or the 10th revision of the International Statistical Classification of Diseases (ICD-10) code-based registration of all inpatient and outpatient healthcare visits. The detailed methodological description can be found in the published study[29].

Blood genetic and proteomic data

The blood proteomes integrated with genotyping data were derived from the Atherosclerosis Risk in Communities (ARIC) study[30] and the INTERVAL study[20]. The relative concentrations of plasma proteins or protein complexes from the blood samples of two studies were measured by the slow off-rate modified aptamer (SOMAmer), a proteomic profiling platform. Genotyping for individuals in ARIC was performed on the Affymetrix 6.0 DNA microarray and imputed to the TOPMed reference panel, while genotyping in INTERVAL was conducted using Affymetrix Axiom UK Biobank genotyping array and imputed to 1000 Genomes Phase 3-UK10K reference panel. After quality control, a total of 4,483 unique proteins from 7,213 European individuals in ARIC and 3,170 proteins from 3,301 European individuals in INTERVAL were included in the PWAS analyses.

Proteome-wide association studies (PWAS)

The associations between plasma protein levels and RA were estimated by the adapted version of functional summary-based imputation (FUSION) pipeline[7] using GWAS summary statistics of RA and pQTLs data from ARIC and INTERVAL studies. For each protein, FUSION trained the predictive models to estimate the effects of genetic variants on protein abundance using a cohort with both genotype and protein abundance data. The predictive model in the ARIC study were fitted using the elastic net algorithm, and penalized linear model was used in the INTERVAL study[31]. Next, using the protein predictive models, FUSION simulated the protein profile of large-scale GWAS summary data, and perform PWAS to estimate the associations between predicted protein levels and RA. The proteins within the MHC region were excluded in PWAS analyses. A strict significance threshold with the Bonferroni correction were used for PWAS associations (P < 2.29×10^− 5, 0.05/2,188, the total number of proteins tested in PWAS). To determine the novelty of the identified PWAS genes, we counted the significant SNPs of RA (P < 5×10 − 8) within ± 1 Mb for each PWAS gene. The genes located out 1Mb of any genome-wide-significant SNP were regarded as novel gene for RA.

Meta-analysis of the discovery and replication PWAS analyses

To increase the statistical power of the PWAS results, we performed a meta-analysis of the discovery and replication PWAS analyses using the sample size weighted analysis in METAL[32]. The proteins were declared significant in meta-analysis if the meta-analysis P values passed Bonferroni corrected threshold (P < 2.29×10^− 5) and demonstrated the same directions of association in both the discovery and replication datasets.

Causal effect inference of PWAS proteins

To investigate whether the proteins identified by PWAS were causally associated with RA, we applied two independent but complementary approaches, Bayesian colocalization (COLOC) and Mendelian randomization (MR). COLOC was utilized to assess the posterior probability that PWAS protein and RA risk loci shared the same genetic variant, rather than the variant shared coincidentally due to linkage disequilibrium (LD). Five mutually exclusive hypotheses were tested in COLOC analysis[33]: H0, no causal association with either GWAS or pQTL (PPH0); H1, association with GWAS only (PPH1); H2, association with pQTL only (PPH2); H3, association with GWAS and pQTL in two independent SNPs (PPH3); and H4, association with GWAS and pQTL in one shared SNP (PPH4). The PWAS proteins with posterior probability PPH4 > 0.8 were considered as strong evidence for colocalization.

In addition, to investigate whether the proteins identified by PWAS have causal relationships with RA, the two-sample MR analysis were performed using GWAS summary data of RA and independent pQTL data. The pQTL data was derived from the largest GWASs of plasma protein levels that measured using 4,907 protein aptamers in 35,559 Icelanders[34]. MR analysis was executed with the “TwoSampleMR” package in R platform[35], with the phenotypic trait (RA) serving as the outcome, the expression level of PWAS proteins acting as the exposure, and significant pQTL SNPs of PWAS proteins using as instrumental variables. To ensure the robustness of the MR results, four models of MR were employed, Wald ratio, inverse variance weighted (IVW), MR Egger, and weighted median, depending on the number of instrumental variables (IVs). The Wald ratio method was used to estimate the causality from exposure to outcome when only one independent IV was accessible, while IVW method was utilized in cases that multiple IVs were available. The MR Egger intercept and Cochran’s Q statistic were further used to examine horizontal pleiotropy and heterogeneity respectively (P > 0.05 indicating the absence of heterogeneity or horizontal pleiotropy). The P value of MR association was Bonferroni-adjusted for the number of proteins with IV available (P < 2.63×10^− 3, 0.05/19). Proteins have evidence with a causal role by either COLOC or MR were regarded as causal PWAS associations.

Examination of the potential causal proteins at mRNA level

Given that DNA is transcribed into messenger RNA and ultimately translated into protein, we further detected whether the causal PWAS genes had similar evidence at the transcript level by leveraging TWAS and gene expression profile at bulk and single-cell levels. TWAS analysis was conducted using joint-tissue imputation (JTI) approach[10], which leveraged genetic regulation shared by different tissues to improve prediction performance in a tissue-dependent manner. Specific to this study, the joint-tissue TWAS was performed using three precomputed gene expression weights of RA-relevant tissues, including whole blood, lymphocytes and fibroblasts from GTEx cohort[36]. The genes with P-value < 1.67×10^− 3 (0.05/30) by Bonferroni corrections were considered as significant.

Then, differential expression analyses were employed to verify the evidence of the causal relationship between PWAS proteins and RA. We selected three largest bulk RNA-seq data related to RA from Gene Expression Omnibus (GEO) database, including GSE93777 (315 RA and 133 healthy controls), GSE45291 (493 RA and 20 healthy controls), GSE120178 (127 RA and 20 healthy controls). The differently expressed genes were conducted using the student t-test with a significance threshold of P-value < 1.67×10^− 3 (0.05/30) by Bonferroni corrections. In addition, we examined cell-type-specific

expression of the PWAS genes using human synovium single-cell RNA sequencing data from Zhang et al.[37]. The study applied scRNA-seq integrated with mass cytometry and flow cytometry to fibroblasts, T cells, B cells, and monocytes from synovial tissue from 36 patients with RA. After quality control, 5,265 cells were clustered into 18 unique cell populations, including four types of fibroblasts, four types of monocytes, six types of T cells, and four types of B cells.

Gene set enrichment analyses

To explore the possible biological functions involved in RA-associated proteins identified by PWAS, we performed gene set enrichment analyses in GO, KEGG and Reactome databases using the R package “clusterProfiler”[38]. For each pathway in these databases, the P-value was calculated based on the hypergeometric test and corrected for multiple testing using the Benjamini-Hochberg procedure. The pathway reached a significance threshold of FDR < 0.05 was regarded as significant.

Protein-protein interaction network

To investigate the potential protein-protein interactions (PPIs) among RA-associated proteins identified by PWAS, a PPI network was constructed using the STRING database (https://string-db.org/)[39]. The proteins in PPI network were further clustered by the Markov clustering algorithm. The STRING webtool also provides gene set enrichment analyses, which aims to detect GO terms and pathways that are significantly enriched with genes in the network versus random genes in genome.

Druggable targets exploration

To assess if the proteins identified by PWAS can serve as potential therapeutic targets for RA, we explored the interactions between these proteins and drugs using Drug-Gene Interaction Database (DGIdb) (https://www.dgidb.org/)[40]. DGIdb contains more than 20,000 drugs and 10,000 genes involved in over 70,000 drug-gene interactions, which has been have been mined from30 distinct sources, such as DrugBank, ChEMBL, PharmGKB, Therapeutic Target Database, Drug Target Commons. DGIdb can not only provide searches for established drug-gene interactions but also query whether the genes are ‘potentially’ druggable according to their membership in gene categories associated with druggability (e.g., kinases).

PWAS identified proteins associated with RA

To identify candidate proteins for RA, we performed the first large-scale PWAS by integrating the most recent GWAS of RA with human blood protein expression weights generated by the ARIC and INTERVAL studies. A total of unique 2,188 proteins (1,259 in ARIC and 929 in INTERVAL) were significantly heritable (P < 0.01) and included in the PWAS analyses. Among the total of 2,188 associations, we identified 17 genes (comprising 20 associations) whose genetically regulated protein abundances were significantly associated with RA risk after Bonferroni corrections (P < 2.29×10^− 5, 0.05/2,188) (Table S1). To bolster the robustness of PWAS results, we performed replication PWAS of RA using independent GWAS summary data. Out of the 17 PWAS proteins, 12 proteins were replicated for significant associations with RA in replication PWAS (Table S2), confirming the validity of our PWAS results. To increase the statistical power of the PWAS results, we performed a meta-analysis of the discovery and replication PWAS analyses. After Bonferroni corrections, 21 proteins (comprising 27 associations) showed proteome-wide significant with RA (Table 1; Fig. 1; Table S3). For example, several well-known risk genes or drug targets for RA, CCL21 (P_Meta =2.09×10^− 13), FCGR2A (P_Meta =9.62×10^− 11), FCGR2B (P_Meta =1.12×10^− 10) enjoyed the most significant PWAS results for RA. Furthermore, six genes were implicated by both ARIC and INTERVAL proteomes, including FCGR2A, FCGR3B, IL6R, ERAP2, FCRL3, and PMEL, indicating the repeatability of PWAS results across proteomes. Notably, among these PWAS identified genes, we found nine genes (ICOSLG, ERAP2, TGFBR3, TNFRSF1B, ERAP1, PCBD2, FCRL3, CCL22, MAPK3) were not implicated in the original GWAS (not proximal to any genome-wide-significant SNP for RA), which were regarded as novel candidate genes for RA.

Table 1

Meta-analysis of the discovery and replication PWAS identified 21 candidate genes in human plasma proteomes associated with rheumatoid arthritis
PANEL	Gene	Chr	Z_discovery	Z_replication	Z_meta	P_discovery	P_replication	P_meta	Novelty^a	Direction
INTERVAL	CCL21	9	-6.189	-5.778	-7.343	6.07×10^− 10	7.54×10^− 9	2.09×10^− 13	Known	--
INTERVAL	FCGR2A	1	-5.223	-5.165	-6.473	1.76×10^− 7	2.40×10^− 7	9.62×10^− 11	Known	--
ARIC	FCGR2A	1	-5.283	-5.122	-6.449	1.27×10^− 7	3.02×10^− 7	1.12×10^− 10	Known	--
ARIC	FCGR2B	1	-4.901	-5.214	-6.424	9.55×10^− 7	1.85×10^− 7	1.32×10^− 10	Known	--
INTERVAL	PTPN11	12	-3.785	-5.526	-6.395	1.54×10^− 4	3.27×10^− 8	1.61×10^− 10	Known	--
INTERVAL	OLFML3	1	5.203	5.012	6.320	1.96×10^− 7	5.39×10^− 7	2.61×10^− 10	Known	++
INTERVAL	ICAM5	19	5.120	4.040	6.098	3.06×10^− 7	5.34×10^− 5	1.07×10^− 9	Known	++
INTERVAL	IL6R	1	-5.013	-4.700	-5.966	5.36×10^− 7	2.60×10^− 6	2.43×10^− 9	Known	--
INTERVAL	FCGR3B	1	4.941	4.325	5.586	7.75×10^− 7	1.53×10^− 5	2.32×10^− 8	Known	++
ARIC	FCGR3A	1	4.539	4.136	5.287	5.65×10^− 6	3.54×10^− 5	1.24×10^− 7	Known	++
INTERVAL	ICOSLG	21	3.790	3.617	5.160	1.51×10^− 4	2.98×10^− 4	2.47×10^− 7	Novel	++
ARIC	IL6R	1	-4.050	-4.097	-5.107	5.12×10^− 5	4.18×10^− 5	3.27×10^− 7	Known	--
INTERVAL	ERAP2	5	2.075	4.658	5.062	3.80×10^− 2	3.19×10^− 6	4.14×10^− 7	Novel	++
INTERVAL	TGFBR3	1	-5.584	-3.515	-5.001	2.35×10^− 8	4.40×10^− 4	5.69×10^− 7	Novel	--
INTERVAL	PMEL	12	-3.435	-4.119	-4.947	5.93×10^− 4	3.81×10^− 5	7.55×10^− 7	Known	--
ARIC	TNFRSF1B	1	-1.778	-1.892	-4.828	7.54×10^− 2	5.85×10^− 2	1.38×10^− 6	Novel	--
ARIC	PMEL	12	-3.242	-4.051	-4.825	1.19×10^− 3	5.10×10^− 5	1.40×10^− 6	Known	--
INTERVAL	IL27	16	-2.981	-4.113	-4.808	2.87×10^− 3	3.91×10^− 5	1.53×10^− 6	Known	--
ARIC	ERAP1	5	2.255	4.270	4.745	2.41×10^− 2	1.95×10^− 5	2.09×10^− 6	Novel	++
ARIC	ERAP2	5	1.467	4.494	4.726	1.42×10^− 1	7.00×10^− 6	2.29×10^− 6	Novel	++
INTERVAL	PAM	5	4.474	3.553	4.711	7.70×10^− 6	3.81×10^− 4	2.46×10^− 6	Known	++
INTERVAL	PCBD2	5	4.173	3.540	4.611	3.01×10^− 5	4.00×10^− 4	4.02×10^− 6	Novel	++
ARIC	FCGR3B	1	4.848	3.223	4.506	1.25×10^− 6	1.27×10^− 3	6.62×10^− 6	Known	++
ARIC	FCRL3	1	4.786	3.213	4.479	1.70×10^− 6	1.31×10^− 3	7.50×10^− 6	Novel	++
INTERVAL	FCRL3	1	4.756	3.179	4.437	1.97×10^− 6	1.48×10^− 3	9.14×10^− 6	Novel	++
ARIC	CCL22	16	-1.353	-4.170	-4.383	1.76×10^− 1	3.05×10^− 5	1.17×10^− 5	Novel	--
INTERVAL	MAPK3	16	-4.295	-3.151	-4.274	1.75×10^− 5	1.63×10^− 3	1.92×10^− 5	Novel	--
^a Genes not within a 1-Mb window of SNPs with P-value < 5×10^− 8 identified in discovery and replication GWAS.

COLOC and MR prioritize 10 proteins causally associated with RA

To investigate whether the genes identified by PWAS were causally associated with RA, we performed two independent but complementary analyses, COLOC and MR, using GWAS summary and pQTL data. COLOC analysis showed strong causal evidences for seven genes that the same risk variants drove the GWAS and pQTL signals (PPH4 > 0.8, Fig. 2). We further performed MR analysis to examine the horizontal pleiotropy and heterogeneity of the casual estimates. After Bonferroni corrections, MR analysis revealed four genes with strong evidence of causal associations (Fig. 2; Figure S3), among which FCGR3B was also prioritized as causal gene in COLOC analysis. Taken together, we identified 10 potential causal genes (FCGR3B, OLFML3, PAM, ICOSLG, CCL21, FCGR3A, FCRL3, FCGR2A, ERAP2, IL6R) by COLOC and MR analyses (Fig. 2) and were prioritized as candidate RA susceptibility genes, including the three novel genes (ICOSLG, FCRL3, ERAP2). The horizontal pleiotropy and heterogeneity were not detected for the ten causal genes (Table S4).

Examination of the potential causal proteins at mRNA level

Given that DNA is transcribed into messenger RNA and ultimately translated into protein, we further investigated whether the 10 causal PWAS genes had similar evidence at transcriptional level by leveraging differential expression analysis and TWAS. First, we conducted differential expression analysis between RA and healthy control from three largest bulk RNA-seq data. After comparing causal PWAS genes with the differential expression genes, we found all the 10 causal genes had similar evidence at the transcriptional level after Bonferroni corrections, of which four genes (FCGR3B, FCGR3A, FCRL3, IL6R) were significant in all three datasets (Fig. 3A). The joint-tissue TWAS of RA identified six genes (OLFML3, PAM, ICOSLG, FCRL3, ERAP2, IL6R) whose cis-regulated gene expression was associated with RA (Fig. 3B), including the three novel genes. Novel genes ERAP2 was were significantly associated with RA in all three tissues (whole blood, fibroblasts and lymphocytes), while FCRL3 and ICOSLG were significant in whole blood and lymphocytes. We next investigated the cell-type-specific expression of the causal PWAS genes using human synovium single-cell RNA sequencing data (Fig. 3C). Genes OLFML3 and PAM were highly expressed in Fibroblasts (CD45⁻ Podoplanin⁺), while FCGR2A and FCGR3A exhibited high expression in Monocytes (CD45⁺ CD14⁺). Among the three novel causal genes, ICOSLG was highly expressed in IGHD⁺ CD270 naive B cells, while FCRL3 and ERAP2 showed high expression in FOXP3⁺ Tregs cells and HLA⁺ sublining fibroblasts respectively, indicating the potential functions of these genes in RA pathogenesis.

Gene set enrichment analyses

To figure out the relevant biological pathways and molecular mechanisms of the PWAS genes, 21 PWAS significant genes were subjected to gene set enrichment analyses in Gene Ontology, KEGG, and Reactome database respectively. GO enrichment analyses identified several GO terms for PWAS genes, such as regulation of leukocyte proliferation (P = 3.66×10^− 5), regulation of lymphocyte proliferation (P = 1.55×10^− 4), regulation of dendritic cell antigen processing and presentation (P = 4.66×10^− 3), and negative regulation of B cell receptor signaling pathway (P = 4.66×10^− 4) (Fig. 4A). We also detected 23 significant KEGG functional pathways for PWAS genes. The most significant KEGG pathways included Fc gamma R − mediated phagocytosis (P = 2.47×10^− 5), cytokine-cytokine receptor interaction (P = 7.34×10^− 4), and natural killer cell mediated cytotoxicity (P = 6.49×10^− 4) (Fig. 4B). Furthermore, enrichment analyses in the Reactome database identified lots of pathways associated with RA, including MAPK3 (ERK1) activation (P = 6.88×10^− 5), signaling by interleukins (P = 2.58×10^− 3), and signaling by FGFR1 (P = 3.05×10^− 3) (Fig. 4C). Identification of these pathways contributed to a better understanding of the regulation of RA pathogenesis.

Connectivity among the identified proteins

To explore the connectivity among the 21 RA-associated proteins identified by PWAS, we constructed a PPI network using the STRING database, which contained 15 nodes and 31 edges (Fig. 4D). The edges in the PPI network were colored according to the protein-protein interactions based on different data sources. Particularly, several well-known RA risk genes identified by PWAS, such as IL6R, FCGR2A, and FCGR2B, were present in the center of PPI network and form more protein-protein interactions with other PWAS genes, suggesting that these genes play important roles in the regulation of RA pathogenesis. Furthermore, the novel genes identified by PWAS were closely interconnected with known risk genes of RA. For example, novel PWAS genes FCRL3 and ICOSLG were experimentally determined to interact with known RA risk gene FCGR3A, which were enriched for genes involved in regulation of lymphocyte activation and proliferation. Novel gene ERAP2 was interact with ERAP1 and enriched in antigen processing and presentation of peptide antigen.

Druggable targets exploration for RA therapy

As most drugs exert their therapeutic effects through targeting proteins, we finally explored whether the 10 causal PWAS genes can serve as potential therapeutic targets using DGIdb database. In DGIdb, we observed 120 drug-gene interactions involving in 7 genes (CCL21, ICOSLG, FCGR3B, FCGR2A, ERAP2, IL6R, FCGR3A) (Table S5), of which several drugs have been approved for the treatments of autoimmune diseases, such as methotrexate (targeting FCGR3B), tocilizumab (targeting IL6R), and indomethacin (targeting FCGR3A). Of note, we identified six and one drug-gene interactions for novel PWAS gene ERAP2 and ICOSLG respectively, although none of these drugs has been approved for RA treatment. For example, AMG-557, a human antibody against T cell co-stimulator ligand (ICOSL) in subjects with systemic lupus erythematosus (SLE), exhibited the highest interaction score with ICOSLG. Furthermore, we evaluated the potential druggability of the 10 causal PWAS gene. We found three well-known drug targets of RA (FCGR2A, IL6R, FCGR3A) coupled with novel PWAS gene ICOSLG fell within the clinically actionable category (Table S6). Given that enzymes predominantly form the bulk of therapeutic targets, the novel PWAS gene ERAP2 was classified as enzyme category. These data suggest that the novel PWAS genes ICOSLG and ERAP2 may be sever as potential therapeutic targets for RA.

In this study, we performed the first PWAS by integrating the largest up-to-date RA GWAS and human plasma pQTL data to identify candidate proteins associated with RA pathogenesis and find potential therapeutic targets for RA treatment. Collectively, 21 genes were identified whose genetically regulated protein abundances were significantly associated with RA risk. After causal effect inference, 10 out of 21 genes were potentially causal and prioritized as candidate RA genes. Of note, among the 10 causal genes, six (OLFML3, PAM, ICOSLG, FCRL3, ERAP2, IL6R) also show evidence of association with RA at transcriptional level, reflecting consistent findings at the mRNA and protein levels. The rest four causal genes, FCGR3B, CCL21, FCGR3A, FCGR2A, as the well-known risk genes and drug targets for RA, were only identified by the PWAS alone, highlighting the importance and necessity of integrating proteomes with GWAS. These high-confidence risk genes may serve as promising targets for further mechanistic investigations and drug development.

Of note, among the 10 causal PWAS genes, three genes (ICOSLG, ERAP2, FCRL3) were not nominated by the original GWAS, thus were regarded as novel candidate genes for RA, underscoring the effectiveness of PWAS in identifying risk genes for RA. The gene FCRL3 encodes the one of several Fc receptor-like glycoproteins, a member of the immunoglobulin receptor superfamily, which possesses tyrosine-based immunoregulatory function. According to our results of TWAS and single-cell transcriptome, FCRL3 were significant in lymphocytes, especially high expressed in FOXP3⁺ Tregs cell of synovium from RA patients. Studies have showed that regulatory T-cells expressing FCRL3 exhibited a memory phenotype and high levels of PD-1, thus reducing capacity to suppress the proliferation of effector T cells, which contribute to the progression of autoimmunity[41, 42]. In addition, FCRL3 has been proved to enhances activation of NF-kappa-B and MAPK signaling pathways in TLR9 stimulated B-cells, ultimately promoting proliferation, activation, and survival of B cells[43]. Gene ICOSLG encodes ligand for the T-cell-specific cell surface receptor ICOS, which has been proved to not only co-stimulate the T cells proliferation and cytokine secretion, but also induce B cells proliferation and differentiation into plasma cells[44]. Endoplasmic reticulum aminopeptidase 2 (ERAP2) is a key enzyme that generate antigenic epitopes, which bind to major histocompatibility complex class I (MHC-I)[45]. A recent study has indicated that ERAP2 inhibits the Hedgehog signaling pathway and upregulates the expression of NLRP3, cleaved Caspase-1, and Gasdermin D to promote pyroptosis in CD4⁺ T cells[46]. These findings provided the genetic and biological evidences that these novel PWAS genes are more likely functional associated with RA pathogenesis, which also emphasized the significance of linking genetic factors to RA etiology via specific proteins.

Exploring potential therapeutic targets is particularly vital for RA treatment, since some individuals do not respond adequately to currently pharmacologic therapies[2]. The druggability exploration analysis observed 120 drug-gene interactions involving in 7 genes, including several drugs have been commonly used as first-line treatment for rheumatoid arthritis, such as methotrexate (targeting FCGR3B), tocilizumab (targeting IL6R), and indomethacin (targeting FCGR3A), demonstrating the effectiveness of PWAS genes in exploration of novel drugs for RA. Notably, AMG-557, a human antibody against T cell co-stimulator ligand, were found closely interacted with novel PWAS gene ICOSLG. In a randomized, double-blind, placebo-controlled study, AMG-557 has showed safety and potential efficacy in subjects with SLE accompanied by arthritis[47], which supports further studies of AMG 557 as a potential therapeutic for autoimmune diseases. Furthermore, six drugs or compounds were identified to interacted with PWAS novel gene ERAP2, including scopoletin, tosedostat, and esculetin. Scopoletin and tosedostat is a coumarin synthesized by diverse medicinal and edible plants, which possesses promising medicinal properties, including anti-bacterial, anti-oxidative, anti-inflammatory, and anti-neoplastic activity. Resent study has identified scopoletin as major anti-rheumatic components that may bind and inhibit tyrosine kinases on fibroblast-like synoviocytes to block NF-κB signaling, and thus combat the progression of rheumatoid arthritis[48]. Analogously, esculetin treatment could suppress NF-κB and MPAK pathway activation, and inflammatory cytokine production, which are the main causes of joint deterioration in RA[49]. Further pharmacological experiments and clinical trials are required to confirm the clinical safety and efficacy of these drugs in RA treatment.

Our study possesses several strengths. Primarily, we performed the first PWAS for RA using the largest and most comprehensive reference human proteomes integrated with the latest GWAS summary statistics of RA. After causal effect inference using MR and COLOC, 10 genes were prioritized as candidate causal genes whose proteomic abundance was associated with RA, which reduced the false-positive of our study. Furthermore, we spotlighted three potential novel candidate genes for RA, presenting prospective avenues for further mechanistic investigations and drug development of RA. Additionally, targeting PWAS genes, we identified several drugs that possesses superior anti-inflammatory and anti-rheumatic activity in autoimmune diseases, which could be serve as the potential therapeutics for RA.

Admittedly, the results of this study should be carefully interpreted in conjunction with its limitations. First, though over 4,000 proteins or protein complexes were examined in the present study, it does not provide coverage for the entire plasma proteome. More comprehensive and larger proteomic datasets will be required in the future to estimate the genetic effects on protein abundance of the rest proteins. Second, owing to the absence of proteomes in other tissues, the present PWAS were only conducted using plasma proteome as reference. Assessing the role of genetically regulated protein in other tissues, especially in synovium, will provide more insight into RA pathogenesis. Third, the present findings were restricted to European participants to avoid population stratification confounding, which limits the generalizability of our results. Further expansion of the scale and diversity of PWAS studies can help with more precise estimates and enable its broader applications. Despite these limitations, PWAS genes are key to understand disease etiology, facilitate biological interpretation of GWAS results, and prioritize follow-up functional studies.

In summary, using PWAS, we identified ten causal genes whose genetically regulated protein abundances were significantly associated with RA risk, including three novel genes (ICOSLG, FCRL3, ERAP2). Our results illuminated the causal pathways and potential drug of these PWAS genes, which highlight promising targets for future mechanistic and therapeutic studies to find effective treatments for RA.

Rheumatoid arthritis

pQTL

Protein quantitative trait loci

GWAS

Genome-wide association studies

TWAS

Transcriptome wide association studies

MHC

Major histocompatibility complex

Linkage disequilibrium

SNP

Single nucleotide polymorphism

Mendelian randomization

Odds ratio

Confidence interval

IVW

Inverse Variance Weighted

PPH4

Posterior probability of hypothesis 4

GEO

Gene Expression Omnibus

scRNA-seq

Single-cell RNA sequencing

Gene ontology

KEGG

Kyoto Encyclopedia of Genes and Genomes

PPI

Protein–protein interaction

STRING

Search Tool for the Retrieval of Interacting Genes

Availability of data and materials

The RA GWAS summary statistics utilized in the discovery and replication PWAS can be accessed from GWAS Catalog (https://www.ebi.ac.uk/gwas/summary-statistics, accession ID: GCST90132223) and deCODE Genetics (https://www.decode.com/summarydata/) respectively. The bulk RNA-seq data of RA were obtained from GEO database (accession ID: GSE93777, GSE45291, and GSE120178). The human synovium single-cell RNA sequencing data of RA patients can be visualized at https://portals.broadinstitute.org/single_cell/study/amp-phase-1.

Funding

This study is supported by the Key Program of National Natural Science Foundation of China (92169211), the Key-Area Research and Development Program of Shaanxi Province (Grant 2023-ZDLSF-41) and the “Special Support Program for High-level Talents” in Shaanxi Province.

Author information

Xin Ke and Shi Yao contributed equally to this work.

Authors and Affiliations

Department of Clinical Immunology, Xijing Hospital, and National Translational Science Center for Molecular Medicine, Fourth Military Medical University, Xi'an, Shaanxi, P. R. China, 710032.

Xin Ke, Xi Zheng, Tian-Yue Liu, Feng-Fan Yang, Yi-Fan Li, Kui Zhang, Zhao-Hui Zheng & Ping Zhu

Guangdong Key Laboratory of Age-Related Cardiac and Cerebral Diseases, Affiliated Hospital of Guangdong Medical University, Zhanjiang, Guangdong, China, 524001.

Shi Yao

Key Laboratory of Biomedical Information Engineering of Ministry of Education, Biomedical Informatics & Genomics Center, School of Life Science and Technology, Xi'an Jiaotong University, Xi'an, Shaanxi, P. R. China, 710049

Hao Wu

Contributions

All authors read and approved the final version of the manuscript. X. Ke, P. Zhu and Z. Zheng conceived and designed the study. X. Ke, X. Zheng, and K. Zhang contributed to literature search and collected data. X. Ke, H. Wu, T. Liu, F. Yang and Y. Li were responsible for data analysis and results interpretation. X. Ke, S. Yao, P. Zhu and Z. Zheng wrote and revised the manuscript. P. Zhu and Z. Zheng supervised all the processes in this manuscript.

Corresponding authors

Correspondence to P. Zhu or Z. Zheng.

Ethics declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

All the authors have consented for publication.

Competing interests

The authors declare that they have no competing interests in this study.

Finckh A, Gilbert B, Hodkinson B, Bae SC, Thomas R, Deane KD, Alpizar-Rodriguez D, Lauper K. Global epidemiology of rheumatoid arthritis. Nat Rev Rheumatol. 2022;18(10):591–602.
Burmester GR, Pope JE. Novel treatment strategies in rheumatoid arthritis. Lancet. 2017;389(10086):2338–48.
Dedmon LE. The genetics of rheumatoid arthritis. Rheumatology (Oxford). 2020;59(10):2661–70.
Okada Y, Eyre S, Suzuki A, Kochi Y, Yamamoto K. Genetics of rheumatoid arthritis: 2018 status. Ann Rheum Dis. 2019;78(4):446–53.
Cannon ME, Mohlke KL. Deciphering the Emerging Complexities of Molecular Mechanisms at GWAS Loci. Am J Hum Genet. 2018;103(5):637–53.
Okada Y, Wu D, Trynka G, Raj T, Terao C, Ikari K, Kochi Y, Ohmura K, Suzuki A, Yoshida S, et al. Genetics of rheumatoid arthritis contributes to biology and drug discovery. Nature. 2014;506(7488):376–81.
Gusev A, Ko A, Shi H, Bhatia G, Chung W, Penninx BW, Jansen R, de Geus EJ, Boomsma DI, Wright FA, et al. Integrative approaches for large-scale transcriptome-wide association studies. Nat Genet. 2016;48(3):245–52.
Gamazon ER, Wheeler HE, Shah KP, Mozaffari SV, Aquino-Michaels K, Carroll RJ, Eyler AE, Denny JC, Consortium GT, Nicolae DL, et al. A gene-based association method for mapping traits using reference transcriptome data. Nat Genet. 2015;47(9):1091–8.
Hu Y, Li M, Lu Q, Weng H, Wang J, Zekavat SM, Yu Z, Li B, Gu J, Muchnik S, et al. A statistical framework for cross-tissue transcriptome-wide association analysis. Nat Genet. 2019;51(3):568–76.
Zhou D, Jiang Y, Zhong X, Cox NJ, Liu C, Gamazon ER. A unified framework for joint-tissue transcriptome-wide association and Mendelian randomization analysis. Nat Genet. 2020;52(11):1239–46.
Wu C, Tan S, Liu L, Cheng S, Li P, Li W, Liu H, Zhang F, Wang S, Ning Y, et al. Transcriptome-wide association study identifies susceptibility genes for rheumatoid arthritis. Arthritis Res Ther. 2021;23(1):38.
Ni J, Wang P, Yin KJ, Yang XK, Cen H, Sui C, Wu GC, Pan HF. Novel insight into the aetiology of rheumatoid arthritis gained by a cross-tissue transcriptome-wide association study. RMD Open 2022, 8(2).
Hormozdiari F, Gazal S, van de Geijn B, Finucane HK, Ju CJ, Loh PR, Schoech A, Reshef Y, Liu X, O'Connor L, et al. Leveraging molecular quantitative trait loci to understand the genetic architecture of diseases and complex traits. Nat Genet. 2018;50(7):1041–7.
Fugger L, Jensen LT, Rossjohn J. Challenges, Progress, and Prospects of Developing Therapies to Treat Autoimmune Diseases. Cell. 2020;181(1):63–80.
Vogel C, Marcotte EM. Insights into the regulation of protein abundance from proteomic and transcriptomic analyses. Nat Rev Genet. 2012;13(4):227–32.
Liu Y, Beyer A, Aebersold R. On the Dependency of Cellular Protein Levels on mRNA Abundance. Cell. 2016;165(3):535–50.
Buccitelli C, Selbach M. mRNAs, proteins and the emerging principles of gene expression control. Nat Rev Genet. 2020;21(10):630–44.
Battle A, Khan Z, Wang SH, Mitrano A, Ford MJ, Pritchard JK, Gilad Y. Genomic variation. Impact of regulatory variation from RNA to protein. Science. 2015;347(6222):664–7.
Zhang J, Dutta D, Kottgen A, Tin A, Schlosser P, Grams ME, Harvey B, Consortium CK, Yu B, Boerwinkle E, 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.
Sun BB, Maranville JC, Peters JE, Stacey D, Staley JR, Blackshaw J, Burgess S, Jiang T, Paige E, Surendran P, et al. Genomic atlas of the human plasma proteome. Nature. 2018;558(7708):73–9.
Pietzner M, Wheeler E, Carrasco-Zanini J, Cortes A, Koprulu M, Worheide MA, Oerton E, Cook J, Stewart ID, Kerrison ND, et al. Mapping the proteo-genomic convergence of human diseases. Science. 2021;374(6569):eabj1541.
Brandes N, Linial N, Linial M. PWAS: proteome-wide association study-linking genes and phenotypes by functional variation in proteins. Genome Biol. 2020;21(1):173.
Wingo AP, Liu Y, Gerasimov ES, Gockley J, Logsdon BA, Duong DM, Dammer EB, Robins C, Beach TG, Reiman EM, et al. Integrating human brain proteomes with genome-wide association data implicates new proteins in Alzheimer's disease pathogenesis. Nat Genet. 2021;53(2):143–6.
Levin MG, Tsao NL, Singhal P, Liu C, Vy HMT, Paranjpe I, Backman JD, Bellomo TR, Bone WP, Biddinger KJ, et al. Genome-wide association and multi-trait analyses characterize the common genetic architecture of heart failure. Nat Commun. 2022;13(1):6914.
Sun J, Zhao J, Jiang F, Wang L, Xiao Q, Han F, Chen J, Yuan S, Wei J, Larsson SC, et al. Identification of novel protein biomarkers and drug targets for colorectal cancer by integrating human plasma proteome with genome. Genome Med. 2023;15(1):75.
Ishigaki K, Sakaue S, Terao C, Luo Y, Sonehara K, Yamaguchi K, Amariuta T, Too CL, Laufer VA, Scott IC, et al. Multi-ancestry genome-wide association analyses identify novel genetic mechanisms in rheumatoid arthritis. Nat Genet. 2022;54(11):1640–51.
Arnett FC, Edworthy SM, Bloch DA, McShane DJ, Fries JF, Cooper NS, Healey LA, Kaplan SR, Liang MH, Luthra HS, et al. The American Rheumatism Association 1987 revised criteria for the classification of rheumatoid arthritis. Arthritis Rheum. 1988;31(3):315–24.
Aletaha D, Neogi T, Silman AJ, Funovits J, Felson DT, Bingham CO 3rd, Birnbaum NS, Burmester GR, Bykerk VP, Cohen MD, et al. 2010 Rheumatoid arthritis classification criteria: an American College of Rheumatology/European League Against Rheumatism collaborative initiative. Arthritis Rheum. 2010;62(9):2569–81.
Saevarsdottir S, Stefansdottir L, Sulem P, Thorleifsson G, Ferkingstad E, Rutsdottir G, Glintborg B, Westerlind H, Grondal G, Loft IC, et al. Multiomics analysis of rheumatoid arthritis yields sequence variants that have large effects on risk of the seropositive subset. Ann Rheum Dis. 2022;81(8):1085–95.
The Atherosclerosis Risk in Communities (ARIC). Study: design and objectives. The ARIC investigators. Am J Epidemiol. 1989;129(4):687–702.
Wang G, Sarkar A, Carbonetto P, Stephens M. A simple new approach to variable selection in regression, with application to genetic fine mapping. J R Stat Soc Ser B Stat Methodol. 2020;82(5):1273–300.
Willer CJ, Li Y, Abecasis GR. METAL: fast and efficient meta-analysis of genomewide association scans. Bioinformatics. 2010;26(17):2190–1.
Giambartolomei C, Vukcevic D, Schadt EE, Franke L, Hingorani AD, Wallace C, Plagnol V. Bayesian test for colocalisation between pairs of genetic association studies using summary statistics. PLoS Genet. 2014;10(5):e1004383.
Ferkingstad E, Sulem P, Atlason BA, Sveinbjornsson G, Magnusson MI, Styrmisdottir EL, Gunnarsdottir K, Helgason A, Oddsson A, Halldorsson BV, et al. Large-scale integration of the plasma proteome with genetics and disease. Nat Genet. 2021;53(12):1712–21.
Hemani G, Zheng J, Elsworth B, Wade KH, Haberland V, Baird D, Laurin C, Burgess S, Bowden J, Langdon R et al. The MR-Base platform supports systematic causal inference across the human phenome. Elife 2018, 7.
Consortium GT. The GTEx Consortium atlas of genetic regulatory effects across human tissues. Science. 2020;369(6509):1318–30.
Zhang F, Wei K, Slowikowski K, Fonseka CY, Rao DA, Kelly S, Goodman SM, Tabechian D, Hughes LB, Salomon-Escoto K, et al. Defining inflammatory cell states in rheumatoid arthritis joint synovial tissues by integrating single-cell transcriptomics and mass cytometry. Nat Immunol. 2019;20(7):928–42.
Yu G, Wang LG, Han Y, He QY. clusterProfiler: an R package for comparing biological themes among gene clusters. OMICS. 2012;16(5):284–7.
Szklarczyk D, Gable AL, Lyon D, Junge A, Wyder S, Huerta-Cepas J, Simonovic M, Doncheva NT, Morris JH, Bork P, et al. STRING v11: protein-protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets. Nucleic Acids Res. 2019;47(D1):D607–13.
Cannon M, Stevenson J, Stahl K, Basu R, Coffman A, Kiwala S, McMichael JF, Kuzma K, Morrissey D, Cotto K, et al. DGIdb 5.0: rebuilding the drug-gene interaction database for precision medicine and drug discovery platforms. Nucleic Acids Res. 2024;52(D1):D1227–35.
Swainson LA, Mold JE, Bajpai UD, McCune JM. Expression of the autoimmune susceptibility gene FcRL3 on human regulatory T cells is associated with dysfunction and high levels of programmed cell death-1. J Immunol. 2010;184(7):3639–47.
Nagata S, Ise T, Pastan I. Fc receptor-like 3 protein expressed on IL-2 nonresponsive subset of human regulatory T cells. J Immunol. 2009;182(12):7518–26.
Li FJ, Schreeder DM, Li R, Wu J, Davis RS. FCRL3 promotes TLR9-induced B-cell activation and suppresses plasma cell differentiation. Eur J Immunol. 2013;43(11):2980–92.
Yong PF, Salzer U, Grimbacher B. The role of costimulation in antibody deficiencies: ICOS and common variable immunodeficiency. Immunol Rev. 2009;229(1):101–13.
Lopez de Castro JA. How ERAP1 and ERAP2 Shape the Peptidomes of Disease-Associated MHC-I Proteins. Front Immunol. 2018;9:2463.
Zhang J, Cai H, Sun W, Wu W, Nan Y, Ni Y, Wu X, Chen M, Xu H, Wang Y. Endoplasmic reticulum aminopeptidase 2 regulates CD4(+) T cells pyroptosis in rheumatoid arthritis. Arthritis Res Ther. 2024;26(1):36.
Cheng LE, Amoura Z, Cheah B, Hiepe F, Sullivan BA, Zhou L, Arnold GE, Tsuji WH, Merrill JT, Chung JB. Brief Report: A Randomized, Double-Blind, Parallel-Group, Placebo-Controlled, Multiple-Dose Study to Evaluate AMG 557 in Patients With Systemic Lupus Erythematosus and Active Lupus Arthritis. Arthritis Rheumatol. 2018;70(7):1071–6.
Chen Q, Zhou W, Huang Y, Tian Y, Wong SY, Lam WK, Ying KY, Zhang J, Chen H. Umbelliferone and scopoletin target tyrosine kinases on fibroblast-like synoviocytes to block NF-kappaB signaling to combat rheumatoid arthritis. Front Pharmacol. 2022;13:946210.
Garg SS, Gupta J, Sahu D, Liu CJ. Pharmacological and Therapeutic Applications of Esculetin. Int J Mol Sci 2022, 23(20).

No competing interests reported.

SupplementaryTable.xlsx

Download PDF

Reviewers agreed at journal
06 Nov, 2024
Reviewers agreed at journal
06 Nov, 2024
Reviewers agreed at journal
05 Nov, 2024
Reviews received at journal
30 Oct, 2024
Reviewers agreed at journal
22 Oct, 2024
Reviews received at journal
22 Oct, 2024
Reviewers agreed at journal
14 Oct, 2024
Reviewers agreed at journal
07 Sep, 2024
Reviewers agreed at journal
06 Sep, 2024
Reviewers invited by journal
04 Sep, 2024
Editor invited by journal
29 Aug, 2024
Editor assigned by journal
27 Aug, 2024
Submission checks completed at journal
27 Aug, 2024
First submitted to journal
25 Aug, 2024

You are reading this latest preprint version

Integrating human plasma proteomes with genome-wide association data implicates novel proteins and drug targets for Rheumatoid arthritis

Status:

Version 1

Abstract

Background

Methods

Results

Conclusions

Figures

Introduction

Methods

GWAS summary statistics of RA

Blood genetic and proteomic data

Proteome-wide association studies (PWAS)

Meta-analysis of the discovery and replication PWAS analyses

Causal effect inference of PWAS proteins

Examination of the potential causal proteins at mRNA level

Gene set enrichment analyses

Protein-protein interaction network

Druggable targets exploration

Results

PWAS identified proteins associated with RA

COLOC and MR prioritize 10 proteins causally associated with RA

Examination of the potential causal proteins at mRNA level

Gene set enrichment analyses

Connectivity among the identified proteins

Druggable targets exploration for RA therapy

Discussion

Abbreviations

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1