Epidemiological research revealed that individuals with chronic liver disorders faced an increasing risk of developing osteoporosis as well as sarcopenia. Retrospective analysis indicated that sarcopenia was frequently observed among PBC patients, with a prevalence of 25.9%5. Although recent relevant studies have been conducted, the precise connection between PBC and sarcopenia is still not thoroughly understood6. Therefore, improving the long-term quality of life for patients with chronic liver disease accompanied by sarcopenia is a significant challenge. As genome sequencing technology advances, the establishment of correlations between traits and genetics is increasingly being prioritized as a practical strategy for overcoming the limitations of observational studies, randomized controlled trials, and traditional research paradigms. Genetic correlation (rg) is utilized to explore the genetic overlap. Theoretically, it can be investigated across the entire genome and represents the average of shared genetic effects among all causal loci within the genome7. Global and local genetic correlation analyses were first applied to identify the features associated with PBC and sarcopenia-related traits8. Subsequently, we employed a cross-trait genome-wide association study (GWAS) meta-analysis at the single nucleotide polymorphism (SNP) level to recognize pleiotropic genetic variations or loci. Consequently, our study delves into the common drug targets and pathogenic pathways between PBC and sarcopenia.
The data of PBC originated from 7 populations recently registered in the IEU OpenGWAS database (https://gwas.mrcieu.ac.uk/), containing 5 European groups and 2 East Asian groups, encompassing 5,004,018 SNPs. Based on the EWGSOP2 definition, the data comprised a broad spectrum of statistics regarding sarcopenia from various metrics, such as the usual walking pace (UWP), appendicular lean mass (ALM), hand grip strength for both left (HGSL) and right (HGSR), leg fat percentage for left (LFPL) and right (LFPR), as well as arm fat percentage for left (AFPL) and right (AFPR) (Table 1). To guarantee the dependability of our research results, we selected SNPs that demonstrated a notable genome-wide association with traits (P < 5 × 10) and eradicated SNPs with high linkage disequilibrium (r2 > 0.001 and kb < 10,000). Furthermore, it was important to discard palindromic SNPs, where the effective and the other alleles were complementary. Figure 1 illustrates the design of the experiment.
We conducted comprehensive genetic correlation analyses via linkage disequilibrium score regression (LDSC) and high-definition likelihood (HDL) to evaluate the shared polygenic architecture among traits. These two methodologies can integrate data from European ancestry samples within the 1000G EUR Project for calculating linkage disequilibrium (LD) scores. Simultaneously, rigorous quality control measures were implemented on the SNPs throughout this procedure. Additionally, we retained only those SNPs with a minor allele frequency (MAF) exceeding 0.01.
Local variation association analysis (LAVA) was enforced to explore the local genetic correlations between pairs of genes. Given the complexity of the genome's structure, LAVA can effectively examine the genetic sequence variations in diverse regions and accurately approximate the genetic correlations within smaller genome segments. Thus, compared to the global genetic correlation analyses, LAVA provides a more comprehensive insight into the genetic overlap between traits. To ensure the stability of the results, the Benjamini-Hochberg method was established to adjust the rate of false discoveries (FDR) (P < 0.05). SUPER GeNetic cOVariance Analyzer (SUPERGNOVA) utilizes a random-effects model for analyzing local genetic correlations and supplies a more accurate estimate of the similarity between pairs of traits compared to the fixed-effects model applied in LAVA.
Multi-trait analysis of GWAS (MTAG) presents an efficient way to detect genetic risk variations across traits, effectively leveraging the genetic structure of shared traits to enhance detection capabilities. The Cross-Phenotype Association (CPASSOC) was employed for cross-trait association analysis to recognize causal pleiotropic SNPs among multiple traits. We analyzed the traits jointly through statistical heterogeneity (SHet) to confirm the association of at least one genetic variant with a trait. Integrating MTAG and CPASSOC, the P-value of SNP was set less than 5 × 10 in both methods, and it met a criterion of less than 5 × 10 for the single-trait association between PBC and sarcopenia.
The Unified Test for Molecular Signatures (UTMOST) has the potential to determine the associations between genes and traits, which takes into account the combined impacts of SNPs within LD regions and integrating the Genotype -- Tissue Expression (GTEx) data of organisms. Subsequently, we validated the results of UTMOST by three methods: Multi-marker Analysis of Genomic Annotation (MAGMA), Functional Summarisation Attribution (FUSION), and Fine-mapping of Causal Gene Sets (FOCUS).
We conducted SNP genetic enrichment analyses on 54 different tissues using MAGMA with GTEx.V8 data. To guarantee the robustness of these associations, we implemented a Bonferroni correction to account for multiple tests. For pleiotropic genes, we established a significant threshold (P < 0.05/the total number of genes). Integrating the tissues detected by MAGMA, we enforced a cross-tissue transcriptome-wide association study (TWAS) analysis by Functional Summarisation Attribution (FUSION) to compute the genetic elements across diverse tissues. FUSION generates predictive models of functional and molecular phenotypic traits, which utilizes global genomic studies to summarize statistical predictions. As a probabilistic fine-mapping method, FOCUS integrates the correlation signals by simulating the association between TWAS signals. Genes with posteriori probability of causality (PIP) values of 0.9 or above were considered potential causal candidates.
Using the summary data-based Mendelian randomization (SMR) method, we have detected genes expressed across various tissues pinpointed by MAGMA. In our research for gene expression proxies, we chose expression quantitative trait loci (eQTLs) from various tissues acquired from the GTEx Consortium website. SMR was conducted to evaluate the influence of these genes on traits, centering on cis-eQTL and trans-eQTL with an MAF greater than 1% and a remarkable threshold of P < 5 × 10. To assess the robustness of our SMR findings, we performed the dependent instrumental heterogeneity (HEIDI) test, serving to highlight significant differences among diverse datasets. For this study, the P-value of HEIDI was greater than 0.05 and an FDR < 0.05 were considered significant.
We performed a colocalization analysis to evaluate the likelihood of shared causal variants between PBC and sarcopenia. This process involved removing SNPs with significant genome-wide associations (P < 5 × 10), eliminating duplicates and missing values, and extracting SNPs that appeared in both PBC and sarcopenia groups. Then, we assessed five potential scenarios to determine if the two traits could be collectively affected by a single genetic variant: PPH0, no correlation with either trait; PPH1, linked to PBC but not the sarcopenia-related traits; PPH2, associated with the sarcopenia-related traits but not PBC; PPH3, connected to both traits with different causal variants; PPH4, related to both traits, sharing a causal variant. Based on the criterion of PPH3 + PPH4 > 0.8, we established the evidence for colocalized genes.
To investigate the functional characteristics and biological connections of pre-selected potential drug genes, we comprehensively applied gene ontology (GO) analysis, Kyoto encyclopedia of genes and genomes (KEGG) pathway enrichment analysis, and gene set enrichment analysis (GSEA). Within the GO framework, we were dedicated to three essential dimensions: biological process (BP), molecular function (MF), and cellular component (CC), aiming to elucidate the activity patterns, biological roles, and cellular localization of these genes. KEGG provided a widespread understanding of gene-related metabolic pathways and signaling networks, revealing the underlying pathogenesis. GSEA comprehensively analyzed the complete gene sequence list, effectively detecting the biological functions of gene sets for gene complementation. While implementing GSEA, we conducted 10,000 alignments while setting the minimum and maximum gene set sizes to 10 and 200, respectively. Furthermore, to accurately identify significantly enriched gene groups, a variety of criteria were combined, including a normalized enrichment score (NES) with an absolute value greater than 1 (|NES|> 1) and an FDR less than 0.25.