PPP1R14A in male erectile dysfunction: key role in smooth muscle cells

Abstract

Background

Erectile dysfunction (ED) is a prevalent male health condition impairing quality of life. Existing therapies have limited efficacy, so novel pharmacological targets and clear molecular mechanisms of ED are needed.

Aim

This study aims to integrate single-cell RNA sequencing (scRNA-seq) and genome-wide association studies (GWAS) analysis to uncover potential therapeutic targets for ED and clarify its molecular mechanisms.

Methods

The study synthesized data from ED GWAS, scRNA-seq, protein quantitative trait loci, and expression quantitative trait loci datasets. Two-sample Mendelian randomization (MR) analysis, with instrumental variable weighted as the main method, was used to assess protein-ED causal relationships. Molecular docking screened candidate compounds, and murine penis experiments validated differential expression.

Outcomes

Reveal the key molecular targets and regulatory mechanisms of ED pathogenesis, screen potential targeted drugs, and provide theoretical and experimental basis for the precise treatment of ED.

Results

After scRNA-seq preprocessing, 2579 differentially expressed genes across five cell types were identified. MR analysis found 113 plasma proteins significantly associated with ED. Among overlapping genes, 4 eQTLs were ED-related, with PPP1R14A showing the strongest association (OR = 1.27; 95% CI, 1.10–1.45; P_IVW = 0.0007). PPP1R14A was predominantly expressed in smooth muscle cells (SMCs) and upregulated in ED patients. Pathway enrichment analysis revealed that PPP1R14A and its co-expressed genes are primarily involved in vascular smooth muscle contraction, actin phosphorylation, and calcium homeostasis. Progesterone and ampicillin were identified as potential binding candidates. Diabetic and elderly mice (with lower intracavernous pressure) had significantly higher PPP1R14A expression in cavernous tissue.

Clinical Translation

PPP1R14A can serve as a new candidate target for targeted therapy of ED. Potential binding compounds such as progesterone and ampicillin provide directions for subsequent drug development and lay the foundation for precise treatment of ED (especially diabetes-related and senile ED).

Strengths and Limitations

The advantage lies in the systematic analysis of the ED mechanism through multi-omics integration (scRNA-seq + MR + proteomics), clarifying cell-specific expression and causal relationships; the limitations are the limited sample size, the lack of direct in vivo and in vitro functional verification experiments, the need for further clinical verification of the safety and efficacy of drug candidates, and the failure to cover all ED subtypes and ethnic diversity.

Conclusion

PPP1R14A plays a key role in the pathogenesis of ED. It affects the erectile process by regulating the function of SMCs. This study provides new insights into the molecular mechanism of ED and also offers important support for the development of targeted therapeutic drugs.

Introduction

Erectile dysfunction (ED) is a prevalent public health concern associated closely with lifestyle factors, affecting more than 50% of men aged 35 to 75.¹ Although various treatment options, including phosphodiesterase-5 inhibitors, intracavernosal injections, shockwave therapy, hormone replacement therapy, and psychological counseling, have been developed, their therapeutic efficacy are not as satisfactory as expected.² Additionally, the high cost of drug development, lengthy development cycles, and low success rates highlight our insufficient understanding of biological mechanisms.³ Therefore, a deeper exploration of biological characteristics in ED patients holds significant clinical significance.

Proteins act as central regulators of physiological and pathological processes, playing crucial roles in both health and disease. Circulating proteins in human plasma not only perform vital functions within the blood but also support systemic physiological regulation by promoting communication between different tissues.^4,5 In many diseases, these circulating proteins exhibit notable dysregulation, making them important drug targets.⁶ Recent advances in large-scale proteomics have enabled researchers to simultaneously quantify thousands of circulating proteins.⁷ Through genome-wide association studies (GWAS), researchers can identify genetic variants associated with protein abundance, ie, protein quantity quantitative trait loci (pQTLs).⁸ By integrating pQTL with disease-associated variants and applying Mendelian randomization (MR), researchers can assess the causal role of proteins in diseases.

Single-cell RNA sequencing (scRNA-seq) provides an unbiased view of cellular composition and cell-type-specific gene expression patterns, enabling the capture of intrinsic heterogeneity within individuals.⁹ Through single-cell analysis, we can identify specific gene expression profiles in different cell types, which is crucial for improving disease stratification and treatment strategies.¹⁰

Considering the advantages of proteomic studies and scRNA-seq, this study aims to identify key proteins with significant roles in ED and explore their expression patterns in various cells as well as potential mechanisms. Based on existing research foundations and pre-analysis results, we propose a testable research hypothesis: Is the PPP1R14A protein a key molecule that regulates the function of smooth muscle cells (SMCs) and is involved in the pathogenesis of erectile dysfunction?

Materials and methods

ScRNA-seq data source and processing

We utilized scRNA-seq data from the Gene Expression Omnibus database (GSE206528). The original study included 3 controls and 5 ED patients. Among them, 3 normal tissue specimens were obtained from the tumor margins of penile carcinoma resection, and all these control patients reported good stimulated erections and morning erections. The corpus cavernosum (CC) tissues from 5 patients with ED were all collected from biopsy specimens obtained during inflatable penile prosthesis implantation. All ED patients were diagnosed with organic ED rather than psychogenic ED via nocturnal penile tumescence and intracavernosal injection (ICI) tests. Age and BMI were comparable between the two groups (see  Supplementary Table 1).

During quality control, we excluded individual cells with fewer than 500 detected genes (nFeature RNA). Additionally, each cell must have total RNA counts (nCount RNA) between 1000 and 20 000, with mitochondrial gene expression comprising less than 10% of total RNA to avoid the influence of stress or cell death (see Supplementary Figure 1). Data were normalized using LogNormalize method, followed by principal component analysis (PCA) to summarize genes in a lower-dimensional space. A two-dimensional uniform manifold approximation and projection (UMAP) plot was generated for visualization of single-cell clusters. Using Seurat's "FindAllMarkers" function, we identified marker genes for each cluster and annotated cell types based on known markers (see Supplementary Figure 2). For each cell type, we used the "FindMarkers" function to identify differentially expressed genes (DEGs) between ED patients and controls, with criteria set as logfc.threshold >0.3, minPct >0.25, and adjusted p-value (Padj) ≤ 0.05.

Genetic tool selection

From deCODE Health studies, we extracted summary statistics related to the levels of 4907 circular proteins and their genetic associations. These data originate from a large-scale pQTL study comprising 35 559 Icelanders. The eQTL data were obtained from the eQTLGen database (https://www.eqtlgen.org/), which is a comprehensive resource for genetic variants associated with gene expression levels.

ED SNP data were sourced from the FinnGen study (https://r11.finngen.fi/). FinnGen studies analyzed genetic information linked to health data in over 500 000 samples from the Finnish Biobank. In this study, we utilized the latest public ED data from FinnGen R11, which included 2548 cases and 196 451 controls.

In MR analysis, we used linkage disequilibrium (LD) parameters (r²) to assess SNP correlations. We selected pQTLs as IVs based on the following criteria: (1) SNPs within ±10 kb of the gene region; (2) statistical significance at genome-wide level (P < 5 × 10^-8) for selecting SNPs highly associated with plasma proteins or eQTLs; (3) LD threshold of r² = 0.001 and a distance of 10 000 kb to ensure independent SNPs and minimize LD effects; and (4) exclusion of weak IVs with F-value >10.

MR analysis

In MR analysis, we evaluated the causal relationship between proteins or eQTLs and ED using methods such as instrumental variable weighted (IVW), MR-Egger regression, Wald ratio, and weighted median. The IVW method served as the primary approach for analysis, with significant results defined by P < .05. Using MR-Egger intercept tests, we assessed directional pleiotropy, and a Cochran Q test evaluated heterogeneity, with P > 0.05 indicating no significant pleiotropy or heterogeneity.

Gene co-expression analysis, enrichment analysis, and phenotype network analysis

ARCHS4 is an integrated gene expression database. We retrieved datasets related to the target gene (PPP1R14A) from this database and calculated its correlations with other genes using Pearson correlation. Based on these correlations, we identified the top 30 and top 50 co-expressed genes. To further explore functional insights, we performed enrichment analyses using tools such as BioPlanet, the Kyoto Encyclopedia of Genes and Genomes (KEGG), Gene Ontology (GO), and GeneMANIA.

To evaluate potential drug targets' pleiotropy and off-target effects, we conducted a phenome-wide association study (PheWAS) on the AstraZeneca PheWAS portal (https://azphewas.com/), setting the threshold at 2E-8 to account for false positives.

Drug candidate prioritization and molecular docking

Using DSigDB database, we retrieved PPP1R14A and its related genes based on overlapping gene count, P-value, adjusted P-value, fold change, and combined score. From the Protein Data Bank (PDB) protein database (https://www.rcsb.org/), we downloaded high-resolution structures of PPP1R14A in pdb format. Small molecules' sdf files were obtained from PubChem (https://pubchem.ncbi.nlm.nih.gov/) and converted to mol2 format using Open Babel. Molecular docking between the protein and small molecules was performed using AutoDock 4.2 and AutoDock Tools 1.5.7. For visualization, PyMol software was employed to generate binding mode diagrams depicting interactions between the receptor and ligand.

Protein measurement in penile tissue

The intracavernous pressure (ICP) and mean arterial pressure (MAP) of 10-week-old C57 mice, 10-week-old db/db diabetic mice, and 20-week-old C57 mice were measured using a catheter-type pressure sensor. The ICP value is the stable value of mice after erectile response induced by electrical stimulation of the cavernous nerve. The electrical stimulation parameters for all three groups of mice were 1 volt (1 V) in intensity, 1 ms in pulse width, 13 Hz in frequency, and 60 s in stimulation duration. The erectile function of the three groups of mice was evaluated by calculating the ICP/MAP ratio. Penile tissues from the three groups of mice were collected for immunofluorescence staining, and the PPP1R14A antibody was purchased from Abcam. In addition, we obtained penile tissues resected during radical treatment from patients with penile tumors (informed consent was obtained) and performed immunohistochemical staining to observe the distribution of PPP1R14A protein in human penile tissues.

Instruments and software used

A catheter-type pressure sensor (Model: RWD-PT100, RWD Life Science Co., Ltd., Shenzhen, China) was adopted to determine the ICP and mean arterial pressure (MAP) of mice; an electrical stimulator (Model: RWD-S48, RWD Life Science Co., Ltd., Shenzhen, China) was used to perform electrical stimulation of the mouse cavernous nerve for erectile response induction. The primary antibody against PPP1R14A, α-SMA, and CD31 used in both immunofluorescence and immunohistochemistry experiments was purchased from Abcam plc, Cambridge, United Kingdom.

All statistical and bioinformatic analyses were implemented using R software (Version 4.4.1) with the following specialized analysis packages: Seurat (Version 5.0.1), TwoSampleMR (Version 0.5.6), biomaRt (Version 2.54.0), and ggplot2 (Version 3.4.4). Open Babel (Version 3.1.1) was employed for small molecule file format conversion (from SDF to MOL2). Molecular docking simulations between PPP1R14A and candidate ligands were conducted via AutoDock (Version 4.2.6) and AutoDock Tools (Version 1.5.7). PyMol software (Version 2.5.2) was used for three-dimensional visualization of the binding mode between PPP1R14A protein and small molecule ligands. ImageJ software (Version 1.54f) was adopted for the semi-quantitative analysis of fluorescence intensity in immunofluorescence-stained penile tissue samples.

Results

Single-cell transcriptomic analysis of CC tissue in ED and non-ED samples

Following rigorous quality control and preprocessing of scRNA-seq data, we obtained single-cell transcriptomes from 38 679 cells in ED samples and 25 174 cells in non-ED samples. PCA was employed to characterize cellular compositions, identifying five major clusters: fibroblasts, endothelial cells, SMCs, macrophages, and T cells (Figure 1A). The proportional distribution of each cell type across samples is depicted in Figure 1B, and the top five marker genes highly expressed in each cluster are shown in the heatmap (Figure 1C). Differential gene expression analysis was performed between ED and non-ED groups for each cell type, and the distribution of DEGs is presented as volcano plots in Figure 1D.

Figure 1.

(A) UMAP plot illustrating annotated cellular clusters. (B) Distribution of cellular clusters across samples. (C) Heatmap showing the top five genes expressed in each cluster. (D) Volcano plot of differentially expressed genes between ED and control groups across clusters.

Identification of causal relationships between plasma proteins and ED

A total of 3457 plasma proteins were included in the MR analysis. The F-statistics for the selected instrumental variables (IVs) ranged from 29.71 to 57932.93, indicating no evidence of weak instrument bias affecting the causal association. Our analysis revealed that 113 plasma proteins were significantly associated with the incidence of ED (P-IVW < 0.05; Figure 2A). Among these, the most notable finding was the phosphate regulating endopeptidase homolog X-linked gene (PHEX), which exhibited a significant positive association with ED risk (OR = 2.59, 95% CI: 1.53–4.39, P = 0.0004). Detailed results of other positive associations are presented in Supplementary Table 2.

Figure 2.

(A) Volcano plot illustrating the effects of plasma proteins on ED using inverse variance weighted (IVW) analysis. The red dashed line indicates the significance threshold, and the five most significant plasma proteins with positive and negative associations with ED risk are labeled. (B) GO enrichment analysis of significant plasma proteins, encompassing biological processes (BP), cellular components (CC), and molecular functions (MF). (C) KEGG pathway enrichment analysis of significant plasma proteins. (D) Venn diagram showing the overlap between significant plasma proteins (blue) and differentially expressed genes identified in single-cell analysis (orange).

Subsequently, KEGG and GO enrichment analyses were conducted on the genes encoding these positively associated plasma proteins, with the results summarized in Figure 2B and C. Notably, the KEGG enrichment analysis highlighted the “Cytoskeleton in Muscle Cells” pathway as both the most significantly enriched and the most frequent pathway. This finding aligns with the functional demands of CC smooth muscle, which must contract and relax properly during erection to facilitate blood engorgement. Additionally, GO pathway analysis revealed that these plasma protein genes are primarily involved in nerve regeneration, synaptic function, and intercellular interactions.

To further refine the analysis, we integrated the positively associated plasma protein genes with the differentially expressed genes identified from scRNA-seq analysis, uncovering a total of 24 overlapping genes.

Causal relationship between pQTL of overlapping genes and ED

We examined overlapping genes to characterize cell types and specific eQTLs potentially associated with ED. After stringent filtering, eQTLs with available SNPs were identified for 18 genes. Detailed information on the selected SNPs is provided in Supplementary Table 2. The analysis revealed that four eQTLs were significantly associated with ED occurrence (Figure 3A). Among these, PPP1R14A exhibited the most robust association (OR = 1.27, 95% CI: 1.10–1.45, P-IVW = 0.0007). These findings were further validated by weighted median and wald ratio analyses, confirming result robustness.

Figure 3.

(A) Forest plot of MR analysis for the four eQTLs associated with ED. Nsnp: Number of single nucleotide polymorphisms. (B, C) GeneMANIA-based enrichment analysis for the combined four genes and PPP1R14A.

Subsequently, enrichment analysis was conducted for the combined four genes as well as PPP1R14A individually using the GeneMANIA platform. The primary networks indicated significant physical interactions. Functional enrichment revealed that these genes are predominantly involved in protein modification, actin regulation, and phosphatase regulation.

Drug target analysis and molecular docking of PPP1R14A and co-expressed genes

We mapped the cellular expression of PPP1R14A in single-cell transcriptomic analysis (Figure 4A), revealing that PPP1R14A is primarily expressed in SMCs and exhibits significantly higher expression levels in ED patient samples compared to controls (Figure 4B, ^***P < 0.001). This finding is consistent with the plasma protein and eQTL MR analyses, which demonstrated that increased PPP1R14A expression is associated with elevated ED risk. This SMC-specific expression pattern was further validated by immunofluorescence co-staining of human CC tissue sections (Figure 4C): PPP1R14A (golden/brown staining) clearly co-localized with the SMC marker α-SMA (green), while showing minimal overlap with the endothelial cell marker CD31 (red); nuclei were counterstained with DAPI (blue).

Figure 4.

(A) UMAP plot showing the distribution of PPP1R14A expression, with the green dashed line delineating the smooth muscle cell cluster. (B) Differential expression of PPP1R14A between ED and control groups. ^***indicates P < .001. (C) Immunofluorescence co-staining of human corpus cavernosum tissue sections, labeling PPP1R14A (golden/brown), smooth muscle cells (α-SMA, green), endothelial cells (CD31, red), and nuclei (DAPI, blue). The merged image confirms co-localization of PPP1R14A with α-SMA-positive SMCs. (D) Molecular docking visualization of PPP1R14A with eight drug molecules.

Next, we performed gene–gene co-expression analysis for PPP1R14A using the ARCHS4 database, identifying the top 30 and top 50 co-expressed genes. Interaction networks of PPP1R14A are shown in Supplementary Figure 3. Enrichment analyses for PPP1R14A and its co-expressed genes were conducted using BioPlanet, KEGG, and GO databases (Supplementary Tables 3-5). The enrichment results for PPP1R14A highlighted its involvement in vascular smooth muscle contraction and its regulatory mechanisms, including PKC-mediated phosphorylation of myosin phosphatase inhibitor and integrin-linked kinase signaling. The co-expressed genes were enriched in biological processes such as smooth muscle contraction, vascular smooth muscle contraction, and angiogenesis regulation. In terms of molecular functions, they primarily exhibited oxidoreductase and alkaline phosphatase activities.

Furthermore, we explored drug candidates related to PPP1R14A and its co-expressed genes using the DSigDB database. Supplementary Table 6 lists the top three associated drugs for each group, including CHEMBL521784, carbachol, phenylephrine, 1-methyl-3-nitro-1-nitrosoguanidine, progesterone, ampicillin, decitabine, and cytarabine.

To investigate the interactions between PPP1R14A and these drug candidates, we performed molecular docking analysis using AutoDock. Visualization of the docking data is shown in Figure 4D. Among the identified drugs, progesterone and ampicillin exhibited binding affinities of -6.8 kcal/mol and -6 kcal/mol, respectively, indicating the potential for effective ligand-target interactions.

Differential expression of PPP1R14A in CC tissue

Subsequently, we compared the immunofluorescence staining of PPP1R14A in the CC tissue of 10-week-old C57 control mice, 20-week-old C57 aged mice, and 10-week-old db/db diabetic mice. Both the ICP of aged mice and diabetic mice was significantly lower than that of control mice, and the ICP/MAP values of these mice showed the same pattern (Figure 5B-C). The immunofluorescence of penile tissues from the three groups of mice is shown in Figure 5A. Semi-quantitative analysis of immunofluorescence revealed that the expression level of PPP1R14A in both the CC and corpus spongiosum of aged mice and diabetic mice was significantly increased compared with that of normal adult mice (Figure 5D-E).

Figure 5.

(A) Immunofluorescence of PPP1R14A (red) in penile tissues of 10-week-old C57, 20-week-old C57, and 10-week-old db/db mice. (B) (C) ICP and ICP/MAP values of the three groups of mice. ICP: Intracavernous pressure; MAP: Mean arterial pressure. (D) (E) Fluorescence intensity of PPP1R14A in the corpus cavernosum and corpus spongiosum tissues of the three groups of mice. (F) Immunohistochemical staining of PPP1R14A in human penile tissue.

Finally, we compared the tissue expression of PPP1R14A in human penile tissue obtained from a 64-year-old volunteer with penile tumor (informed consent obtained), who had moderate to severe erectile dysfunction. According to immunohistochemistry, it was observed that the expression of PPP1R14A in the CC of the penile shaft was higher than that in the CC near the glans (Figure 5F). In addition, the expression of PPP1R14A was significantly correlated with the morphology of vascular smooth muscle.

Discussion

To develop precise therapeutic targets for ED, it is essential to comprehensively understand the key molecular, genetic, and cellular differences between ED patients and healthy individuals. These differences encompass not only gene expression and protein regulation but also alterations in cellular functions and microenvironment. Through integrative multi-omics analysis, this study identifies the central role of PPP1R14A in ED pathogenesis. Specifically, PPP1R14A is highly expressed in SMCs of ED patients, and MR analysis based on pQTL and eQTL data confirms a causal relationship between its elevated expression and increased ED risk. These findings provide novel insights into ED pathology, suggesting that SMC dysfunction is a pivotal driver of the disease. PPP1R14A likely contributes to abnormal SMC contractility—such as excessive tension or relaxation dysfunction—by regulating calcium homeostasis and myosin (actin) phosphorylation pathways.^11,12

KEGG and GO enrichment analyses further indicate that PPP1R14A and its co-expressed genes are primarily involved in pathways related to vascular smooth muscle contraction, cytoskeletal regulation, and neural signal transduction. These results suggest that PPP1R14A may interfere with the hemodynamic changes necessary for erection by influencing the contractile function of CC SMCs. Notably, the high expression of PPP1R14A in ED patients is closely associated with aberrant activation of SMCs, which might be linked to its functional role in regulating actin dynamics and phosphatase activity.¹³

Previous studies have shown that PPP1R14A regulates smooth muscle contraction and relaxation by modulating phosphatase activity and affecting the phosphorylation state of myosin light chains (MLCs, Figure 6).¹⁴ Aberrations in PPP1R14A signaling have been implicated in various pathological conditions, including hypertension, asthma, inflammation, and diabetes. For instance, upregulation of PPP1R14A expression and phosphorylation has been observed in hypoxia-induced pulmonary hypertension, as well as in airway smooth muscle under inflammatory conditions and bladder smooth muscle in diabetic states.^15–17 Conversely, reduced PPP1R14A expression has been reported in intestinal smooth muscle under inflammatory conditions, accompanied by decreased muscle tone.¹⁸ The bidirectional regulation of PPP1R14A under inflammatory signals in different smooth muscle tissues warrants further investigation. Under hyperglycemic conditions, the upregulation of PPP1R14A may inhibit phosphatase activity, thereby reducing the dephosphorylation of MLCs.¹⁹ This could lead to excessive smooth muscle tension or relaxation dysfunction, ultimately affecting the engorgement and maintenance capacity of the CC. This mechanism not only explains the pathological link between metabolic abnormalities, such as diabetes, and ED but also provides a theoretical basis for therapeutic strategies targeting PPP1R14A.

Figure 6.

Mechanism of muscle relaxation mediated by calcium signaling and PPP1R14A. Calmodulin, CaM.

The PPP1R14A gene (also known as CPI-17), located on chromosome 19, encodes a 147-amino-acid polypeptide.²⁰ This protein contains three major domains: the N-terminal and C-terminal tails and the central PHIN domain spanning residues 35 to 120.²¹ In terms of drug development, molecular docking and drug target analyses conducted in this study identified several known drugs, such as progesterone and ampicillin, which may have potential binding affinity to PPP1R14A. These findings suggest that these drugs could potentially improve ED symptoms by acting on PPP1R14A. However, the clinical application of these drugs requires further validation regarding their safety and efficacy. Moreover, based on the PPP1R14A gene-drug interaction network, future research may explore more potential therapeutic strategies, such as developing specific inhibitors or agonists targeting PPP1R14A.

Despite its novel findings, this study has certain limitations. First, the research was based on a limited number of samples, which might not cover all possible ED subtypes or account for racial and ethnic diversity, potentially limiting the generalizability of the results. Second, although multi-omics analyses and genetic validation were conducted, the lack of direct experimental data (eg, animal models or in vitro experiments) limits the reliability of the conclusions. Additionally, while we hypothesized that PPP1R14A regulates phosphatase activity to influence smooth muscle function, the precise molecular mechanisms underlying this process remain to be elucidated. Lastly, molecular docking and drug target analyses only provided a theoretical basis for potential drug candidates, and the lack of experimental validation highlights the uncertainty in drug development. Further studies are necessary to confirm the safety and efficacy of these drug candidates.

Conclusion

In summary, this study, through integrative multi-omics analysis, systematically uncovered the critical role of PPP1R14A in ED pathogenesis and its potential molecular mechanisms for the first time. These findings provide not only new insights into the pathological basis of ED but also a theoretical and experimental foundation for the development of precise targeted therapies. Future research should delve deeper into the specific molecular mechanisms by which PPP1R14A regulates smooth muscle function and evaluate its clinical translational potential as a therapeutic target.

Author contributions

Yongdong Pan (Conceptualization, Methodology, Formal Analysis, Investigation, Writing—original draft, Visualization), Ruihang Zhang (Conceptualization, Software, Validation, Data curation, Formal analysis), Yubo Gu (Investigation, Resources, Data Curation), Lujie Song (Conceptualization, Resources, Supervision, Project administration, Funding acquisition, Writing—review & editing).

Yongdong Pan and Ruihang Zhang contributed equally.

Funding

We thank the Science and Technology Commission of Shanghai Municipality (22S31901700), the Discipline Leader of Shanghai Municipal Health (No. 2022XD015) and the Summit Plateau Program, Research Physician Program, Shanghai Jiao Tong University School of Medicine (No. 20240817).

Conflicts of interest

The authors declare that they have no conflict of interest.

Ethical approval

The research ethics review was conducted by Shanghai Sixth People's Hospital, with the number 2021-020 and 2025-KY-056. Informed consent was obtained from all participants in the original genome-wide association study and scRNA study.