Transcriptomic data, methylation profiles, and clinical information for PRAD were obtained from the Cancer Genome Atlas (TCGA, https://portal.gdc.cancer.gov). Normal prostate tissue expression data were sourced from the Genotype-Tissue Expression (GTEx, https://www.gtexportal.org) project. Protein expression levels and immunohistochemistry images were retrieved from the Human Protein Atlas (HPA). Two independent gene expression datasets (GSE150692 and GSE141445) were downloaded from the Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo) for validation and co-expression analysis^[21,22].

Both datasets contain prostate cancer samples with matched clinical information. GSE150692 includes RNA-seq data from a large cohort with long-term follow-up, suitable for survival analysis. GSE141445 provides gene expression profiles from both tumor and adjacent normal tissues, allowing differential expression analysis. Both datasets are publicly available on GEO and have been used in previous prostate cancer bioinformatics studies. A curated list of 830 GPCR genes was compiled from existing literature and genomic databases for focused screening^[13,23-25].

Differential expression analysis was performed using three widely accepted algorithms: edgeR, DESeq2, and limma^[26-28]. Raw counts were normalized and transformed using log2(TPM + 0.1). Genes with |log2(Fold Change)| > 1.5 and false discovery rate (FDR) < 0.05 were considered significantly differentially expressed. The final set of robust differentially expressed genes (DEGs) was defined as the intersection of DEGs identified by all three methods. In detail, differential expression analysis using edgeR and DESeq2 was performed on raw count data. For limma, normalized TPM values were log2-transformed (log2 (TPM + 0.1)). The expression matrix used in the main results section (e.g., heatmaps) was based on log2 (TPM + 0.1). The intersection of DEGs from the three methods was used for downstream analysis to minimize false positives.

Kaplan-Meier survival curves were generated based on gene expression levels (high vs. low, stratified by median expression). The log-rank test was used to assess the statistical significance of differences in disease-free survival. Associations between gene expression and clinical parameters (Gleason score, T stage, N stage) were evaluated using the Kruskal-Wallis test or Mann-Whitney U test for categorical variables.

Methylation levels in the promoter region of LGR6 were compared between tumor and normal samples using beta values from the Illumina Infinium Methylation Assay. Statistical significance was assessed with the Mann-Whitney U test. Major transcript isoforms of LGR6 were identified using transcript-level TPM values from TCGA and GTEx.

Somatic mutation data were processed using maftools in R. Mutation frequencies and their associations with expression were visualized and tested. Pearson correlation coefficients were calculated between LGR6 and key genes (e.g., PD-L1, RNA modification genes). A correlation coefficient > 0.5 was used to define strongly co-expressed genes.

Genes strongly co-expressed with LGR6 were subjected to Gene ontology and Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis using the clusterProfiler R package^[29]. The RNA stemness score (RNAss) was calculated based on transcriptome data, and its correlation with LGR6 expression was assessed across multiple cancer types.

The amino acid sequences of LGR6 (UniProt ID: Q9HBX8) and TP53 (UniProt ID: P04637) were used as inputs. The three-dimensional structures of LGR6 and TP53 were predicted using AlphaFold3^[30]. AlphaFold3 was run with default parameters: number of recycles = 3, ensemble = 8, relaxation = yes. The predicted models with the highest predicted local distance difference test (pLDDT) scores were selected. Protein-protein docking was performed using Rosetta to identify potential interaction interfaces^[31]. Meanwhile, we used Discovery Studio software to further identify potential ligand-binding pockets on the predicted LGR6 structure.

For docking, we used AlphaFold3’s multimer mode and Rosetta v3.13 with the following parameters: docking protocol = local_dock, interface score threshold = -1.0. The docking protocol consisted of global docking followed by local refinement. The Rosetta energy score and interface score were used as scoring functions, 1,000 poses were generated, and the final model was selected based on the lowest interface energy and highest cluster density. The lowest interface score decoy from the largest cluster was chosen as the final model. Interface residues were defined as those with any atom within 5 Å of the partner chain. Hydrogen bonds and pi-pi stacking were analyzed using Rosetta’s InterfaceAnalyzer.

Conserved domains in LGR6 were identified via Pfam (https://pfam.xfam.org) and NCBI CDD (https://www.ncbi.nlm.nih.gov/Structure/cdd). The LGR6 protein sequence was submitted to Pfam (https://pfam.xfam.org) and NCBI CDD (https://www.ncbi.nlm.nih.gov/Structure/cdd). Default parameters were used. The identified domains (e.g., leucine-rich repeat, 7-transmembrane region) were recorded. We searched DrugBank (https://go.drugbank.com) and the FDA Orange Book (https://www.fda.gov/drugs) for drugs targeting LGR6 or its homologous receptors LGR1-LGR3. Search terms included “LGR6”, “LGR1”, “LGR2”, “LGR3”, and “leucine-rich repeat GPCR”. Only FDA-approved or investigational drugs with reported activity against these receptors were included. Then, existing drugs targeting these domains or structurally similar receptors (LGR1-3) were retrieved from DrugBank (https://go.drugbank.com/) and the FDA Orange Book (https://www.fda.gov/).

The 3D structure of protein was shown by Pymol^[32]. Structural similarity between LGR6 and LGR1, LGR2, LGR3 was quantified by calculating the root-mean-square deviation (RMSD) of Cα atoms after structural alignment. The predicted LGR6 structure was aligned to each of the LGR1-3 structures (obtained from the AlphaFold Protein Structure Database or predicted using AlphaFold3 as for LGR6) using PyMOL’s "align" command (Needleman-Wunsch algorithm). RMSD values (in Å) were reported; lower RMSD indicates higher structural similarity.

All statistical analyses were conducted in R 4.2.1 and Python 3.9, with two-sided P < 0.05 as significance threshold. Differential expression was analyzed using edgeR, DESeq2 on raw counts and limma on log2 (TPM + 0.1) data. Clinical correlations were assessed by Kruskal-Wallis and Mann-Whitney U tests. Survival analysis used Kaplan-Meier method and log-rank test. Pearson correlation was applied for gene co-expression and stemness index analysis. Methylation differences were tested by Mann-Whitney U test. For the initial screening of 830 GPCR genes, the Benjamini-Hochberg FDR was applied to correct for multiple testing. All subsequent analyses including survival, mutation, methylation, stemness, WNT pathway correlation, and pan-cancer comparisons were similarly FDR-corrected within each analysis domain. Statistical methods were annotated in figure legends accordingly.

In detail, the Mann-Whitney U test (also known as Wilcoxon rank-sum test) was used to compare gene expression levels between two independent groups (e.g., tumor vs. normal, or high vs. low expression groups). The test does not assume a normal distribution and ranks all observations from both groups together. The U statistic was calculated using the scipy.stats.mannwhitneyu function in Python (v3.9). A P-value < 0.05 was considered statistically significant.

Based on the TCGA database, we downloaded and obtained the transcriptome data for prostate cancer. Principal component analysis (PCA) of the log2-transformed transcriptomic data revealed a clear separation between the disease group and the control group along the PC1 and PC2 dimensions, indicating global transcriptomic differences between prostate cancer tissues and normal tissues [Figure 1A]. Using the mRNA expression matrix, we performed differential expression analysis with edgeR, DESeq2, and limma (log2 (TPM + 0.1)), respectively (screening thresholds: |log2(Fold Change)| > 1.5 and FDR < 0.05). A substantial number of DEGs were identified, with a considerable amount of both up-regulated and down-regulated genes [Figure 1B and C]. These three differential expression algorithms identified 688, 705, and 1,009 DEGs, respectively. By taking the intersection of the three DEG sets, we obtained 352 common DEGs, comprising 138 up-regulated [Figure 1D] and 214 down-regulated genes [Figure 1E]. Subsequently, a further intersection with the GPCR gene set (containing 830 genes) was performed, ultimately screening 19 significantly differentially expressed GPCR genes. These 19 genes are LGR6, CXCR2, NPFFR2, ADRA1A, GPR149, HCAR3, ADGRD2, HTR1E, ADRB3, OR51A7, OR51T1, OR51G2, OR52I1, KISS1R, OR51S1, OR2B6, OR52R1, F2RL2, and OR51E2.

Survival prognosis analysis of the aforementioned 19 differentially expressed GPCR genes identified 7 GPCR genes significantly associated with the disease free survival of prostate cancer patients (log-rank P < 0.05) [Figure 2A and B]. Integrated with the differential expression results, OR51E2 was found to be the most significantly up-regulated gene in tumor tissues (log2FC = 2.93, P = 1.14 × 10^-16) [Figure 2B], while LGR6 showed the most pronounced down-regulation (log2FC = -2.69, P = 1.95 × 10^-46) [Figure 2C]. Survival correlation analysis revealed that although OR51E2 was up-regulated in the disease group, patients with low expression of this gene did not show a significant survival advantage, and the P-value of 0.037 indicated a marginally significant association with prognosis [Figure 2B]. In contrast, the prognostic significance of LGR6 was consistent with its down-regulated expression pattern, with a significant P-value of 0.0058.

Clinical correlation analysis demonstrated that LGR6 expression levels were significantly associated with Gleason score (P = 2.4 × 10^-6), T stage (Kruskal-Wallis P = 1 × 10^-4), and N stage (P = 4.3 × 10^-4) [Figure 2D]. For OR51E2, its expression was significantly correlated with Gleason score (P = 2.4 × 10^-6) and T stage (Kruskal-Wallis P = 1 × 10^-4), but not with N stage (P = 0.16) [Figure 2E]. These figures show differences in LGR6 expression across Gleason score groups, suggesting a potential link between LGR6 and tumor aggressiveness. The trend of changing LGR6 expression with advancing T stage may reflect its role in tumor progression. N stage indicates lymph node metastasis status (N0: no metastasis; N1: metastasis present). The differential expression of LGR6 in the N1 subgroup may imply its involvement in metastatic regulation. These results suggested that LGR6 is a potential biomarker for prostate cancer.

To further evaluate whether LGR6 is an independent prognostic factor, we performed multivariable Cox regression adjusting for age, Gleason score, T stage, and N stage. LGR6 high expression remained significantly associated with favorable disease-free survival (HR = 0.54, 95%CI: 0.34-0.87), indicating that LGR6 is an independent prognostic biomarker in PRAD [Supplementary Table 1]. Subgroup analysis further confirmed the consistency of LGR6 prognostic value across different clinical strata [Supplementary Table 2]. Meanwhile, analysis of an independent dataset (GDS2546) revealed an LGR6 expression pattern in healthy adjacent prostate tissue and primary prostate tumor that was consistent with the TCGA-PRAD data [Supplementary Table 3, Supplementary Figure 1]. However, the function of LGR6 in prostate cancer remains unclear and warrants further investigation.

To further clarify the expression of LGR6 in prostate cancer, we downloaded data from the HPA database (https://www.proteinatlas.org) to determine the protein expression and transcriptional levels of LGR6 in normal prostate tissues and cancerous tissues. Subsequently, we found that LGR6 exhibits a relatively high expression level in the prostate (Staining: Medium) [Figure 3A]. The mRNA expression level in normal prostate tissues was approximately 6 nTPM (sample 59: 8.6; sample 207: 6) [Figure 3B]. However, in prostate cancer tissues, the level of LGR6 staining was low (Staining: weak) [Figure 3C, Supplementary Figure 2]. These results demonstrate that LGR6 expression, both at the protein and transcriptional levels, is higher in normal prostate tissues than in cancerous tissues.

To investigate the impact of alternative splicing on LGR6, we analyzed prostate cancer data from TCGA and found that ENST00000255432 and ENST00000367278 are the predominant isoforms of LGR6 [Figure 3D]. According to GTEx (genotype-tissue expression) data, these two isoforms are also dominant in normal prostate tissues [Figure 3E]. This indicates that alternative splicing is not the reason for the altered expression of LGR6 in prostate cancer. Subsequently, we analyzed methylation data from prostate cancer (TCGA-PRAD) and discovered that the methylation level in the promoter region of LGR6 is significantly higher in prostate cancer tissues than in normal tissues (P value = 1.46 × 10^-10) [Figure 3F]. Meanwhile, the difference in methylation levels is primarily distributed in the promoter region [Figure 3G]. Therefore, we hypothesize that a significant reason for the low expression of LGR6 in prostate cancer is the hypermethylation of the gene's promoter region.

To further investigate the function of LGR6 in prostate cancer, we analyzed the association between LGR6 mutations and its expression, as well as the correlation between LGR6 expression and gene mutations. Subsequently, we found that mutations in LGR6 do not affect its expression level [Figure 4A], and its mutation rate is relatively low in TCGA-PRAD patients, with a mutation rate of only 0.4% [Figure 4B]. Further analysis revealed significant differences in the mutation rates of 13 genes between the high and low LGR6 expression groups, namely FOXA1, SPOP, LRP1B, PIK3CA, APC, CDH12, XIRP1, ZAN, SCN5A, TRPM6, SSPO, NEB, and UBR4 [Figure 4C]. Among these 13 genes, nine (FOXA1, SPOP, PIK3CA, APC, LRP1B, CDH12, TRPM6, UBR4, SCN5A) have well-established associations with cancer, with FOXA1, SPOP, PIK3CA, and APC being particularly critical and core genes in cancer biology. Additionally, gene expression correlation analysis showed that while LGR6 showed no significant correlation with PD-L1 expression levels (R = 0.087, P = 0.053) [Figure 4D], it was closely related to the expression levels of these 13 genes, exhibiting an overall significant positive correlation (R = 0.15, P = 6.4 × 10^-4) [Figure 4E]. Meanwhile, it was found that LGR6 expression levels show significant correlations with the expression levels of RNA modification genes across various cancers, particularly in PRAD [Figure 4F]. Thus, we speculate that LGR6 may influence the expression of these cancer-related genes by regulating RNA modification genes.

To further investigate the potential functions of LGR6, we downloaded the datasets GSE150692 and GSE141445 from the GEO database. Subsequently, correlation analysis revealed 105 genes with a correlation coefficient greater than 0.5 with LGR6; all had P < 0.05. Among these, 102 genes showed a positive correlation, while three exhibited a negative correlation. Functional enrichment analysis of these genes indicated that the cancer-related signaling pathway most significantly enriched was the WNT signaling pathway [Figure 5A]. Meanwhile, we calculated the correlation between the pan-cancer stemness index RNAss and LGR6 expression and found that LGR6 exhibited the most significant correlation in PRAD [Figure 5B]. Additionally, we performed correlation analysis on six gene sets related to the WNT signaling pathway, namely: genes related to the WNT signaling pathway, regulatory factors controlling the initiation and inhibition of the WNT signaling pathway, receptors of the WNT signaling pathway, regulators of the WNT signaling pathway, kinases and phosphatases involved in the regulation of the WNT signaling pathway, and other genes playing roles in cell proliferation, differentiation, and apoptosis. The results demonstrated that LGR6 showed significant correlations with these genes in PRAD [Figure 5C]. These findings collectively suggest that LGR6 may influence the progression of prostate cancer by modulating the activity of the WNT signaling pathway.

Furthermore, we investigated the regulatory network of LGR6 and its significantly correlated genes, revealing that LGR6 has predicted regulatory relationships with multiple oncogenes, such as TP53 [Figure 5D]. TP53 may regulate the expression of LGR6 by binding to its promoter region. Meanwhile, we first used AlphaFold3 to predict the three-dimensional structures of LGR6 and TP53, followed by computational docking analysis using both AlphaFold3 and Rosetta software. The results suggest that TP53 and LGR6 may potentially interact, with relatively diverse interaction sites (the primary interaction region of LGR6 is 140-280, while that of TP53 is 10-180), primarily mediated by hydrogen bonds and pi-pi stacking [Figure 5E]. These findings suggest that LGR6 may be associated with TP53, although a physical interaction has not been experimentally validated. These computational predictions suggest a potential interaction, which requires experimental validation (e.g., co-IP or pull-down assays).

Currently, there are no commercially available targeted drugs for LGR6. Since no targeted drugs against LGR6 have been approved or developed by the FDA, our study focused on drug repurposing for LGR6-targeted therapy. To explore potential therapeutic agents, we investigated evolutionary relationships and structural pharmacology. Through sequence alignment, we identified several conserved domains in LGR6, including PLN00113, LRR, PPP1R42, and LRR_8 [Figure 6A]. Based on these domains, we searched drug databases such as the FDA and DrugBank and identified homologous targeted drugs. Specifically, drugs associated with the LRR domain included Dapansutrile and Selnoflast, while drugs matching the PPP1R71 domain included Selnoflast, Defactinib, and Lorlatinib [Figure 6B].

Simultaneously, we observed that LGR6 shares significant structural similarity with LGR1, LGR2, and LGR3 in the LGR family. The RMSD values from structural alignment fell within 15 Å, indicating a high degree of structural conservation among LGR family members [Figure 6C] As LGR6 exhibits high structural similarity with LGR1-3 in the LGR family, we can infer the putative active site and druggable pocket of LGR6 based on the conserved domains and known ligand-binding regions of LGR1-3. This suggests that targeted drugs developed for LGR1-3 might potentially bind with LGR6. Further investigation revealed that agonists for LGR1 include Corifollitropin alfa and Follitropin; agonists for LGR2 include Choriogonadotropin alfa, Goserelin, and Lutropin alfa; and an agonist for LGR3 is Thyrotropin alfa. These agonists are speculative candidates requiring experimental validation for targeting LGR6 [Figure 6D]. These findings suggest that LGR6 may have multiple potential agonists that could be explored for the treatment of prostate cancer.