HGF knockdown suppresses ovarian cancer malignancy: insights from a transient receptor potential-related gene model

Abstract

Background

Ovarian carcinoma (OV) is a prevalent gynecologic malignancy. While transient receptor potential (TRP) channels are substantially correlated with tumor growth, their role in OV remains unclear. This study therefore aims to construct a comprehensive TRP-related prognostic model and analyze its association with the cell infiltration and checkpoint expression.

Methods

Based on transcriptomic data from public OV databases, TRP activity scores were calculated using single-sample gene set enrichment analysis (ssGSEA), and correlated gene modules were identified through Weighted Gene Co-expression Network Analysis (WGCNA). Functional enrichment analysis was performed to elucidate underlying biological pathways. A prognostic signature was developed via machine learning algorithms, and its clinical utility was validated through construction of a nomogram integrating key clinicopathological parameters. Comprehensive immune characterization was conducted to compare microenvironmental features between risk subgroups. Finally, functional assays including cell counting kit-8 (CCK-8), wound healing, and Transwell were employed to experimentally validate the role of a candidate gene in OV progression.

Results

Of the 21 co-expression modules identified, the brown module demonstrated the strongest positive correlation with TRP scores. Functional enrichment analysis revealed that TRP-related genes were predominantly involved in immune system pathways. Using the nine identified genes, a risk model was developed, and Riskscore was employed to categorize patients into high- and low-risk groups. Overall survival (OS) was notably higher for patients in the OV low-risk group than for those in the high-risk group. The nomogram’s findings demonstrated that the Riskscore had an independent impact on prognosis. According to immunological characterization, the low-risk group had higher levels of cellular infiltration, including activated B cells and activated CD4 T cells. High-risk group had low expressions of immune checkpoint genes, including LAG3, CD274, and CD27. In vitro tests showed that HGF knockdown markedly reduced the viability, motility, and invasion of OV cells.

Conclusion

The TRP-related gene signature constructed in this study predicts the prognosis and immune microenvironment status of OV patients, providing a new perspective for prognosis assessment and targeted therapy in OV.

Introduction

Ovarian cancer (OV) ranks as the most lethal gynecological malignancy [1–3], with cervical cancer and uterine body cancer following it, and both its morbidity and mortality rates continue to rise [4, 5]. In primary lesions, epithelial ovarian carcinoma is the most common type of malignant ovarian tumor, accounting for 70% of all ovarian malignancies [6]; other types include ovarian stromal tumors, germ-cell tumors, sex-cord stromal tumors, and other rarer subtypes [7]. Only 50% of ovarian cancer cases survive for 5 years because 75% of cases are detected at an advanced stage due to a lack of reliable diagnostic and surveillance markers. Within 5 years, about 80% of patients with advanced stages either recur or have a dismal prognosis [8–10]. Overall survival (OS) is not prolonged by approved maintenance therapy with PARP inhibitors or bevacizumab, despite the fact that it is beneficial in extending progression-free survival [11]. As a result, it is still vital to look for biomarkers linked to OV development and aid in diagnostic prediction.

It has been demonstrated that alterations in the cell cycle, which block cell death pathways and encourage cellular proliferation, are closely involved in tumor development [12]. A malignant tumor phenotype can arise as a result of abnormalities in Ca²⁺ homeostasis, which are frequently linked to these changes in cells [13]. By modifying intracellular Ca²⁺ concentration or membrane potential, transient receptor potential (TRP) channels are believed to be ion channels permeable to mono- and divalent cations and function as signal transducers [14]. Based on topological differences and amino acid sequence homology, TRP superfamily contains seven common subfamilies (TRPN, TRPP, TRPML, TRPC, TRPM, TRPV, TRPA) [15]. An eighth species, TRPY, was discovered in yeast [16]. In the past 10 years, it was demonstrated that TRP channels mediated aberrant differentiation, defective cell proliferation, and death in response to external stimuli, which results in unchecked cancer invasion and growth [17–19]. For instance, esophageal squamous carcinoma patients with high expression of TRPV2, a member of the TRPV subfamily, have a poor prognosis [20], and breast cancer patients with TRPV6 have tumor cell growth and basal calcium influx inhibition [21], suggesting that TRPV6 may be a target for treating breast cancer. Furthermore, it was discovered that melanoma metastasis was linked to downregulation of TRPM1 of the TRPM family [22]. Additionally, it was shown that TRP channels are crucial for the activation of T and B cells, the bactericidal activity of neutrophils and macrophages, the presentation of antigens by dendritic cells, and the degranulation of mast cells [23]. Therefore, a comprehensive investigation into the role of TRP channels in OV is imperative to elucidate their potential as prognostic biomarkers and therapeutic targets. Currently, the systemic role of TRP channel-related genes in OV prognosis assessment and immune microenvironment regulation remains unclear, hindering their clinical translation. Thus, this study calculated the TRP-correlated gene scores of each OV patient in public databases by bioinformatics. TRP-related modular genes in OV were identified and enriched using WGCNA. OV patients were categorized by the median Riskscore into high-risk and low-risk groups, and clinicopathological features and Riskscore were integrated into a nomogram. The variations in biological pathways and immunological profiles among the various risk categories were then investigated. Overall, the TRP-related gene analysis system established in this study not only provides a novel approach for prognostic assessment in OV patients but also lays a theoretical foundation for further development of relevant biomarkers and targeted therapeutic strategies.

Methods

Data collection and pre-processing

RNA-seq data in fragments per kilobase per million (FPKM) format were obtained from The Cancer Genome Atlas (TCGA)-OV database (https://portal.gdc.cancer.gov/). The FPKM values were converted to transcripts per million (TPM) values for normalization. We retained only samples with complete survival information (overall survival time > 0 days) and corresponding clinical annotations. After filtering, 373 OV samples with protein-coding genes were included for subsequent analysis. Additionally, the GSE32062 dataset was retrieved from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.govgeo/). Ensembl identifiers were converted to gene symbols, and expression values for genes with multiple mappings were averaged. After excluding samples lacking overall survival data, 260 OV samples were retained. The TCGA-OV cohort served as the training set, while the GSE32062 dataset functioned as an independent validation cohort throughout this study. Finally, we obtained TRP-related genes based on previous studies for subsequent analysis [24].

Weighted gene co-expression network analysis (WGCNA)

First, this study employed the single-sample gene set enrichment analysis (ssGSEA) algorithm to calculate the TRP score for each OV patient [25]. This score is derived by evaluating the rank ordering pattern of TRP-related genes within individual sample expression profiles, with its numerical value directly reflecting the overall activation level of the TRP channel-related gene set in the sample. Next, to identify co-expression networks and screen genes from different clusters, we next created weighted gene co-expression networks using the "WGCNA" package [26]. After clustering samples and filtering missing genes, the "pickSoftThreshold" R function was applied to decide the optimal soft threshold power (β = 9) to better detect important correlations between modules. Next, we performed a hierarchical cluster analysis to identify gene modules with at least 60 genes in a module (minModuleSize = 60). Next, we used the "WGCNA" package to compute the values of gene importance and module affiliation to assess the relation between feature scores and gene modules [27]. Lastly, we used the "Heatmap" package to select different module signature genes based on the first main component of the module expression [28]. The link between module and feature scores was examined according to the correlation between the clinical feature diagnosis and the module signature genes. The modules with the strongest correlations were filtered in order to retrieve the genes they contained.

Functional enrichment analysis

KEGG and GO analysis was conducted with the R package "clusterProfiler" [29, 30] under the condition of p-value < 0.05. We created bubble plots for each of the Top10 functions enriched in the three terms of GO analysis and the Top10 pathways in the KEGG pathway enrichment results to identify the enrichment pathways and biological processes of modular genes. We employed the R package “clusterProfiler” to perform GSEA to analyze the pathways of various biological processes in high- and low-risk groups of TCGA-OV, with KEGG database as a reference for pathway enrichment analysis.

Screening TRP-related signature genes to develop a diagnostic model

In our investigation, we searched for brown module genes that were associated with a p-value of less than 0.05 and a TRP score more than 0.5. Univariate Cox proportional risk regression was then performed using the R package "survival" [31] to identify significant prognostic genes. To increase genes in the model, we compressed the genes using LASSO Cox regression analysis with the "glmnet" package [32] and used tenfold cross-validation to improve the model’s generalization. Significant genes and their corresponding correlation coefficients associated with prognostic outcomes in the TCGA-OV cohort were screened by multifactorial stepwise regression analysis. The prognostic model was defined by the formula: Riskscore = Σβi × Expi, where Expi represents the normalized expression level of each selected gene, i is the gene expression, and β denotes the Cox regression coefficient for that gene. After standardizing the gene expression values using z-score normalization, TCGA-OV patients were divided by the optimal cutoff value of the Riskscore into high- and low-risk groups. Survival differences between these groups were evaluated using the Kaplan–Meier method with the "survminer" R package [33], and statistical significance was assessed with the log-rank test. Additionally, ROC curves were generated using the "timeROC" R package [34] to assess the model performance. We then computed the AUC for the 1-, 2-, 3-, 4-, and 5-year periods.

Establishment and validation of the clinical characteristics and prognostic nomogram of OV

To determine whether the Riskscore could serve as an independent prognostic factor for survival in TCGA-OV patients, we integrated it with key clinical variables, including age, stage, and grade, and conducted both univariate and multivariate Cox regression analyses. Subsequently, a predictive nomogram was developed by incorporating the Riskscore along with clinically significant parameters identified through multivariate regression [35]. The "caret" package plotted calibration curve to assess the prediction of the nomogram. The model’s predictive ability was assessed using DCA using the "rmda" tool in R package.

Immunological characterization of OV

The StromalScore, ImmuneScoreh, and ESTIMATEScore linked to the tumor immune microenvironment in TCGA-OV were visualized using the R package “estimate” [36] based on the transcriptome expression patterns of the samples. The “MCPcounter” program was used to examine the connection between the 10 immune cell scores and the TCGA-OV dataset’s Riskscore [37]. The ssGSEA function of the “GSVA” package was then used to examine the scores of 28 tumor-infiltrating immune cells [38]. Additional analysis was performed to calculate the expression of immune checkpoint genes in the two risk groups.

Cell culture and plasmid transfection

BeNa Culture Collection (Xinyang, China) provided two OV cell lines, SK-OV-3 (BNCC310551, RRID: CVCL_0532) and A2780 (BNCC351906, RRID: CVCL_0134), while Yage Biotech (Shanghai, China) provided the human normal ovarian cell line IOSE-80 (YS2273C, RRID: CVCL_5546). DMEM (Hyclone, Logen, UT, USA) and RPMI-1640 medium (11875093, Gibco, Waltham, MA, USA) were added with 10% FBS (10099141, Gibco, USA), 100 μg/mL streptomycin, and 100 μg/mL penicillin G (15140122, Gibco, USA). A2780 cells were cultivated in DMEM medium, while IOSE-80 and SKOV3 cells were cultivated in RPMI-1640 medium. Each cell type was grown with 5% CO₂ at 37 °C in an incubator.

HGF-knockdown (si-HGF) plasmid and si-NC (control plasmid) ordered from GenePharma (Shanghai, China) were transfected into 2 × 10⁴ cells/well logarithmic growth phase A2780 cells and SK-OV-3 cells, respectively, according to the guidelines provided with the Lipo3000 Liposome Transfection Reagent (L3000-001, ThermoScientific, Waltham, MA, USA). si-HGF sequences were as follows: antisense: GAAGAAUGGUACAAAUCCAAG, sense: UGGAUUUGUACCAUUCUUCUG.

RNA extraction and qRT-PCR

Using the RNA Extraction Kit (TRIzol, Invitrogen, USA), total RNA was extracted from IOSE-80, SK-OV-3, and A2780 cells in accordance with the instruction. Next, the isolated total RNA’s purity and concentration were evaluated. The process of creating cDNA templates were initiated using the HiScript II kit (Vazyme, China). Specific primers and the KAPA SYBR® FAST kit (Sigma Aldrich, San Luis, MO, USA) were utilized in qRT-PCR. With GAPDH as an internal control, data were calculated examined using the 2^−∆∆CT approach. Table 1 lists the primer sequences for specific genes.

Table 1.

The sequences of primers for RT–qPCR used in this study

Gene name	Forward primer	Reverse primer
CD40LG	5' GCGGCACATGTCATAAGTGAGG 3'	5' GTCCTTGTCTTTTAACGGTCAGC 3'
HGF	5' GAGAGTTGGGTTCTTACTGCACG 3'	5' CTCATCTCCTCTTCCGTGGACA 3'
NAAA	5' ATTACGACCACTGGAAGCCAGC 3'	5' GGAAAAGTGCCTCCAGGCTGAG 3'
S1PR4	5' GCTACATCCTCTTCTGCCTGGT 3'	5' CATCAGCACCGTCTTCAGCAGG 3'
STAT4	5' CAGTGAAAGCCATCTCGGAGGA 3'	5' TGTAGTCTCGCAGGATGTCAGC 3'
STAT5A	5' GTTCAGTGTTGGCAGCAATGAGC 3'	5' AGCACAGTAGCCGTGGCATTGT 3'
TAP1	5' GCAGTCAACTCCTGGACCACTA 3'	5' CAAGGTTCCCACTGCTTACAGC 3'
TRPM2	5' GGCAGCCTTGTACTTCAGTGAC 3'	5' GAGGCAGAACAGGATGAAGTCC 3'
VSIG4	5' GATGGCAACCAAGTCGTGAGAG 3'	5' CCTGGCATTGAAGGCTAATCCTC 3'
GAPDH	5' GTCTCCTCTGACTTCAACAGCG 3'	5' ACCACCCTGTTGCTGTAGCCAA 3'

Cell viability

To assess the effect of HGF on SK-OV-3 and A2780 cell viability, colorimetric assays were conducted using CCK-8 (Dojindo, Kumamoto, Japan) in compliance with the instruction. In brief, cells (2000 cells/well) were cultured for 24, 48, and 72 h in 96-well microtiter plates. After two PBS washes, each well contained 100 μL of fresh medium and 10 μL of CCK-8 solution. The cells were then incubated for 3 h at 37 °C with 5% CO₂. A SPECTROstar® Nano (BMG LABTECH GmbH, Ortenberg, Germany) was employed to detect absorbance at 450 nm [39].

Wound healing assay

For migration experiments, transfected cells were added to 6-well plates (2 × 10⁴/mL). 6-well plates were added with 2 mL of cell suspension, and then they were incubated with 5% CO₂ at 37 °C. After adhering the cells to the wall, the monolayer was scraped into a homogeneous wound using a 10 μL plastic pipette tip. The monolayers were incubated in a medium free of FBS after being washed with PBS. The lengths of the wound edges between two edges of migrating cell sheet were measured by taking pictures at 0 and 48 h. The experiment was conducted in triplicate [40].

Cell invasion assay

Cell suspension of SK-OV-3 and A2780 cells was prepared in serum-free medium at a density of 2 × 10⁴ cells/mL. The upper chamber of a Transwell insert (Corning, Beijing, China) was pre-coated with Matrigel (30 μg/well; BD, San Jose, CA, USA) and filled with 100 μL of the cell suspension. The lower chamber was supplemented with 600 μL of medium containing 10% FBS. Following incubation, invaded cells were fixed using 4% paraformaldehyde and colored with crystal violet. Six different fields of view were used to count the invaded or migrated cells in the bottom compartment under a microscope [41].

Statistical analyses

All statistical analyses employed GraphPad Prism 8 (GraphPad Software, San Diego, USA) and R software (version 3.6.0, R Foundation, Vienna, Austria). KM curves were used to display the produced data. The Student’s t-test, one-way ANOVA, or two-way ANOVA with Bonferroni correction were employed. For all analyses, the p-value < 0.05 was the threshold for statistical significance.

Result

Construction of co-expression networks and identification of OV-related genes

In TCGA-OV samples, we observed that 26 out of 33 TRP-related genes were mutated in 436 (30.96%) OV patients, of which the top 4 mutated genes were PKDREJ (4%), PKD2L1 (3%), TRPM2 (3%) and PKD1L1 (3%) (Fig. 1A). We thereafter used the WGCNA technique to build a gene co-expression network in order to more precisely mine the key genes linked to the TCGA-OV phenotype. In order to satisfy the network’s scale-free topology, the soft threshold was set at 9 (Fig. 1B), followed by dynamic module identification in the TCGA-OV cohort with no less than 60 genes per module, and a total of 21 comparable gene modules were discovered (Fig. 1C). Genes that could not be grouped into other modules were found in the gray module. Out of the 21 modules, the dark green module has the most genes, followed by the gray module (Fig. 1D). The brown module with 471 genes had the largest positive correlation with TRP scores, according to our subsequent analysis of the relationship between module eigenvalues and TRP scores (cor = 0.57, p = 2.38e−33, Fig. 1E). In the brown module, the scatterplot revealed a substantial correlation between module membership and gene significance (cor = 0.84, p < 1e−200, Fig. 1F).

Fig. 1.

TCGA-OV cohort’s co-expression network construction. A TRP-related gene mutation status. B Scale-free fitting index analysis for different soft threshold powers (β). Different soft threshold powers by average connectivity analysis. C Gene dendrogram based on clustering using the dissimilarity measure (1-TOM). D Each module’s gene count. E Module eigenvector correlation with each module’s features. F Module membership versus gene significance scatter diagram for the brown module’s TRP-related gene score

Functional enrichment analysis of TRP-related genes in OV

To look into the biological activities and potential pathways linked to TRP in TCGA-OV, we first performed GO and KEGG enrichment analysis of the genes in the brown module using the R package "clusterProfiler." This made it possible for us to look at how TRP regulates the pathophysiology of TCGA-OV. T cell receptor signaling pathway, NF-kappa B signaling pathway, B cell receptor signaling pathway, cytokine–cytokine receptor interaction were the KEGG signaling pathways that mainly affected by TRP-related genes (Fig. 2A). GO enrichment analysis showed that TRP-related genes were mostly implicated in BPs relevant to leukocyte differentiation, leukocyte activation, leukocyte cell–cell adhesion, T cell activation, and other immunological pathways (Fig. 2B). The receptor complex, vesicle lumen, and other structural outgrowths were the CCs where TRP-related genes were primarily found in TCGA-OV (Fig. 2C). The three primary TRP-related gene alterations linked to MF were cytokine receptor binding, cytokine activity, and tumor necrosis factor receptor binding (Fig. 2D).

Fig. 2.

TRP-correlated genes in TCGA-OV: GO and KEGG enrichment analysis. A Functional annotation of TRP-related genes’ KEGG signaling pathway. B–D TRP-related gene GO functional annotation

Development of clinical prognostic models and validation

The 471 genes in the brown module were subjected to univariate Cox regression analysis to identify TRP-related prognostic features in TCGA-OV. Employing the LASSO Cox regression method via the "glmnet" package, we further compressed the gene number. We next employed tenfold cross-validation to enhance the model’s capacity for generalization (Fig. 3A, B). Then, using multifactorial stepwise regression analysis, we identified nine genes (CD40LG, HGF, NAAA, S1PR4, STAT4, STAT5A, TAP1, TRPM2, and VSIG4) that were associated independently with the prognosis of TCGA-OV (Fig. 3C). A Riskscore to evaluate the prognosis of TCGA-OV was formulated: Riskscore = (−0.391*CD40LG) + 0.233*HGF + 0.165*NAAA + (−0.262*S1PR4) + (−0.275*STAT4) + 0.187* STAT5A + (−0.195*TAP1) + 0.191*TRPM2 + 0.195*VSIG4. TCGA-OV patients were categorized by the Riskscore’ ideal threshold into low- and high-risk groups. The Kaplan–Meier curves showed that the OS for TCGA-OV low-risk patients was better (p < 0.0001, Fig. 3D).

Fig. 3.

TRP-related prognostic model construction and validation for OV patients. A, B LASSO cox reduction gene count. C Prognostic gene signature coefficient distribution. D KM survival curves for TCGA training cohort. E Riskscore ROC curves for TCGA training cohort. F Survival distributions and Riskscore curves for the TCGA cohort. G Riskscore KM survival curves for the GSE32062 cohort. H Riskscore’s ROC curve for the GSE32062 cohort. I Survival status of low- and high-risk groups’ in the GSE32062 cohort

Employing the “timeROC” R package, ROC analysis demonstrated that the prognostic model based on the Riskscore achieved AUC values of 0.67, 0.70, 0.68, 0.70, and 0.70 for predicting 1-, 2-, 3-, 4-, and 5-year OS, respectively, in the TCGA-OV training cohort (Fig. 3E). These results affirmed the model’s performance in predicting OS. Furthermore, analysis of Riskscore distributions in relation to survival outcomes revealed a significantly higher mortality rate among high-risk patients (p < 0.001, Fig. 3F). We further assessed the robustness of the clinical prognostic model using the GSE32062 validation set, using similar models and equivalent coefficient used for the identified training set. Consistent with the training set, validation results confirmed that high-risk OV patients experienced markedly worse results (p < 0.0001, Fig. 3G). The prognostic performance of the model was further supported by time-dependent ROC analysis in the GSE32062 cohort, with AUC values of 0.67, 0.70, 0.63, 0.66, and 0.70 for 1‑, 2‑, 3‑, 4‑, and 5‑year survival predictions, respectively (Fig. 3H). Similarly, a markedly elevated mortality rate was observed among high-risk patients in this validation set (p < 0.001, Fig.3I).

Enrichment pathway differences between the two risk subgroups of OV

KEGG enrichment analysis showed that the synthesis of glycosaminoglycans (chondroitin sulfate/dermatan sulfate), ECM-receptor interaction, and the pathways for protein digestion and absorption were all differentially expressed in the OV high-risk group (Fig. 4A). The OV low-risk group focused mostly on linoleic acid metabolism and mineral absorption routes (Fig. 4B).

Fig. 4.

GSEA analysis of DEGs in different OV risk subgroups. A, B KEGG enrichment analysis of the two groups of OV

Development of a nomogram combining Riskscore and key clinical characteristics and validation

We also extended the algorithm to include the Riskscore and three clinical parameters (age, stage, and grade) for a more comprehensive prognostication of OV patients. Multivariate analysis [age: HR (95% CI) = 1.33 (1.03,1.72), p = 0.031; Riskscore: HR (95% CI) = 2.75 (2.13,3.54), p < 0.001] and univariate analysis [age: HR (95% CI) = 1.31 (1.01,1.69), p = 0.043; Riskscore: HR (95% CI) = 2.72 (2.11,3.5), p < 0.0001] were employed for OV prognosis evaluation. Cox regression analysis revealed that age and risk score were significant predictive factors for OV patients (Fig. 5A, B). A nomogram was created to further quantify the risk and survival assessment. The Riskscore exerted the most substantial influence on OS in OV patients (Fig. 5C) Calibration curves showed that the predictive accuracy of the nomogram closely aligned between the predicted and ideal outcomes at 1, 3, and 5 years, indicating excellent prognostic performance (Fig. 5D). DCA analysis verified that both the Riskscore and nomogram exhibited superior reliability in prognostic prediction (Fig. 5E).

Fig. 5.

Nomogram creation for OV patient prognosis prediction. A One-way Cox analysis of risk score and clinical characteristics. B Multifactor Cox analysis of clinical characteristics and risk score. Nomogram modeling (C). D, E Nomogram calibration and decision curves

Immune characterization among different risk subgroups of OV patients

ESTIMATE analysis showed higher immune infiltration in the OV low-risk group (Fig. 6A). The analysis of the "MCPcounter" tool then showed that Riskscore was closely negatively associated with the scores of NK cells, T cells, cytotoxic lymphocytes, B lineage, and CD8 T cells (p < 0.001, Fig. 6B). The OV low-risk group showed notably higher infiltration score of immature B cells, activated dendritic cells, and active CD4 T cells, activated B cells, activated CD8 T cells compared to the high-risk group, as indicated by the findings of the ssGSEA analysis (Fig. 6C). This implies that the immune systems of those in the OV low-risk category are more active, which may aid in the destruction of tumor cells and the prevention of cancer’s spread. Next, patients in the high-risk group were more vulnerable to immune escape as they had lower expressions of immune checkpoint genes, such as CD27, CD274, and LAG3, than the low-risk patients (Fig. 6D).

Fig. 6.

The connection between the OV immune microenvironment and Riskscore. A ESTIMATE evaluation of immune infiltration in comparison to Riskscore groups. B Timer evaluation of the relationship between immune cell scores and Riskscore. C Expression of immune cell scores for 28 ssGSEA in Riskscore groups. D Immune checkpoint expressions in groups with high and low risks. ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05, and Ns represents p > 0.05

Downregulation of HGF impaired the migratory and invasive ability of OV cells

The qPCR showed that HGF, NAAA, STAT5A, TRPM2, and VSIG4 gene expressions were considerably higher in SK-OV-3 and A2780 cells than in IOSE-80 cells, while IOSE-80 cells exhibited notably higher expression for the genes CD40LG, S1PR4, STAT4, and TAP1 (Fig. 7A). Numerous cellular processes, including migration, angiogenesis, and proliferation, as well as the spread of several human malignancies, have been linked to HGF, according to studies [42]. When OV patients were first diagnosed, their circulating serum HGF levels were significantly greater than those of patients with benign disease or borderline tumors, whereas elevated sHGF levels indicated worse prognostic outcomes [43]. To explore the role of HGF in OV progression, we constructed two distinct HGF-knockdown plasmids and transfected them into SK-OV-3 and A2780 cell lines. The transfection group exhibiting the most efficient knockdown (si-HGF #1) was selected for subsequent functional analyses (Fig. 7B, C). Next, the CCK8 test showed that A2780 and SK-OV-3 cell viability was demonstrated to be significantly decreased by HGF knockdown (Fig. 7D, E). The results of Transwell and wound healing demonstrated that the migration and metastasis of A2780 and SK-OV-3 cells were HGF notably suppressed by knockdown (Fig. 7F–I).

Fig. 7.

The biological functions of HGF in OV. A qPCR of the expressions of HGF, NAAA, STAT5A, TRPM2, VSIG4, CD40LG, S1PR4, STAT4, and TAP1 in IOSE-80, SK-OV-3, and A2780 cells. B, C Validation of HGF knockdown. D, E Validation of the effect of HGF knockdown on SK-OV-3 and A2780 cell viability. F–I Representative images of wound healing assay of SK-OV-3 and A2780 cells (magnification, × 40; scale bars = 200 μm) and Transwell assay (magnification, × 100; scale bars = 200 μm). Data are expressed as SD ± mean, ****p < 0.0001, ***p < 0.001; **p < 0.01; *p < 0.05

Discussion

Because of its high recurrence rate and absence of distinct symptoms, OV is one of the most deadly gynecologic cancers in the world [44, 45]. Biomarkers could be effective in the prognostic assessment of OV. Numerous biological pathways that contribute to the progression of cancer require Ca²⁺ signaling [46]. Ion channels, especially those in the TRP superfamily, have been shown to affect intracellular Ca²⁺ levels [47]. It has been suggested that genes linked to TRP channels function as indicators of immunological traits and cancer development in a range of tumor types, such as triple-negative breast cancer [48] and cervical cancer [49]. It is yet unknown, though, how TRP channel-related genes link to OV. Here, by constructing a novel prognostic model based on nine TRP-related genes, this study successfully stratifies OV patients into distinct risk subgroups with significant survival differences and reveals their contrasting immune microenvironment profiles. Experimental validation further identifies HGF as a key promoter of tumor malignancy. These findings provide a valuable tool for prognostic stratification and may help identify potential beneficiaries of immunotherapy, while also suggesting new targets for therapeutic intervention against OV.

The nine TRP-related signature genes identified in our prognostic model (CD40LG, HGF, NAAA, S1PR4, STAT4, STAT5A, TAP1, TRPM2, and VSIG4) are implicated in diverse mechanisms underlying OV pathogenesis. CD40LG potentiates apoptosis and enhances cisplatin sensitivity through immune activation [50], while HGF receptor inhibition suppresses tumor progression and metastasis [51]. Of particular interest is NAAA, which hydrolyzes N-acylethanolamines that can activate PPARα and exert anti-inflammatory effects; its downregulation in high-risk tumors might promote an immunosuppressive microenvironment through accumulation of these lipid mediators, though this potential mechanism requires experimental validation in OV [52]. S1PR4, which ligands the lipid second messenger S1P, modulates acquired immunity by promoting immune cell polarization and activation [53], and shows positive correlation with CD8+ T cell infiltration in the ovarian tumor microenvironment [54]. STAT4 overexpression promotes OV cell metastasis and correlates with poor patient outcomes [55], while elevated STAT5A expression enhances invasion and DNA repair capacity in high-risk OV patients [56]. Notably, while STAT4 overexpression in tumor cells has been linked to enhanced metastasis and poor prognosis, its negative coefficient in our risk model suggests a more complex, context-dependent role. We hypothesize that this association with favorable outcome may reflect the predominant contribution of STAT4-mediated anti-tumor immunity within the tumor microenvironment, highlighting the critical balance between its pro-metastatic and immuno-stimulatory functions in determining patient survival. TAP1 facilitates metastasis and associates with unfavorable prognosis [57], and TRPM2 overexpression correlates with poor prognosis, immune checkpoint expression (ICOSLG, CD40, CD86), and M2 macrophage infiltration [58]. VSIG4, highly expressed in OV tissues, promotes tumor progression and recurrence [59]. These findings collectively demonstrate that TRP signature genes regulate critical processes including invasion, metastasis, and immunomodulation in OV.

Our findings that these TRP-related signature genes are closely associated with immune regulation logically prompted us to investigate their potential impact on the tumor immune microenvironment. Tumorigenesis and development are caused by the tumor microenvironment, which also somewhat affects how well immunotherapy works [60]. Adaptive immune cells such as T cells and B cells are tumor-killing effector cells, whereas certain innate immunity cells are thought to have immunoregulatory roles in the tumor microenvironment [61]. TRP-related genes have recently been found to enhance immunological responses and promote T cell activation and proliferation in the TME of acute myeloid leukemia [62]. Given that TRP channel-related signature genes were discovered to be substantially positively linked with immune infiltration and increased expression of the immune checkpoints CD86 and PD-1, immune checkpoint inhibitor therapy may be more advantageous for high-risk colon cancer patients [63]. In this investigation, we found that there was more immune cell infiltration in the OV low-risk patients. This indicates that the immune system is more active in OV low-risk individuals, which could aid in the destruction of tumor cells and slow the spread of cancer. These studies imply that the tumor microenvironment may be connected to the various regulatory effects of TRP channel-related signature genes on immune infiltration. In cancer treatment, immunotherapy has garnered public interest [60]. For instance, immunotherapy of anti-PD-L1, anti-PD-1, or anti-CTLA4 drugs have been extensively used for OV [64, 65]. Additionally, immunological checkpoints and the immunotherapeutic response in microenvironment of acute myeloid leukemia (AML) can be enhanced by TRP-related genes [62]. Our findings therefore suggest that the TRP-based risk score reflects the dynamic equilibrium between anti-tumor immunity and compensatory immunosuppression within the TME, where the high-risk phenotype may represent an immune-dysfunctional state characterized by inadequate effector cell infiltration despite elevated checkpoint expression, potentially informing stratified immunotherapeutic strategies.

This study has several limitations that should be acknowledged. First, the prognostic model was derived exclusively from public databases. Although independent training and validation sets were used, the absence of multi-center, prospective clinical cohorts may constrain the generalizability of our findings to real-world clinical settings. Future work should validate the model’s robustness through multi-center, prospective studies with long-term follow-up. Second, while our bioinformatic analyses revealed significant associations between the TRP-related signature and the immune microenvironment, and functional experiments confirmed the role of HGF, the mechanistic contributions of other key genes remain unclear. Subsequent research should systematically elucidate the functions of these genes through both in vitro and in vivo experiments, particularly focusing on their roles in TRP channel-mediated calcium signaling and immunoregulatory pathways. Furthermore, the characterization of immune cell infiltration within the tumor microenvironment relied primarily on computational deconvolution algorithms. Although multiple complementary methods were employed, these inferences lack experimental validation. Future studies should incorporate techniques such as flow cytometry, immunohistochemistry, or multiplex immunofluorescence to quantitatively assess immune cell infiltration and verify the computational predictions.

Conclusions

We successfully developed and verified a risk characterization model containing nine TRP-related signature genes (CD40LG, HGF, NAAA, S1PR4, STAT4, STAT5A, TAP1, TRPM2, and VSIG4) based on a public database. In OV patients, the signature demonstrated high predictive potential for OS, biological traits, immunotherapy, and immune infiltration. These results advance our knowledge of how TRP channel-associated genes affect the immunological profile and prognosis of OV patients and could offer fresh insights into the creation of tailored treatment plans and targeted immunotherapies.

Abbreviations

OV

Ovarian cancer

TRP

Transient receptor potential

TCGA

The cancer genome atlas

GEO

Gene expression omnibus

WGCNA

Weighted gene co-expression network analysis

GO

Gene ontology

KEGG

Kyoto Encyclopedia of Genes and Genomes

BP

Biological process

MF

Molecular function

CC

Cellular component

AUC

Area under ROC curve

DCA

Decision curve analysis

LASSO

Least absolute shrinkage and selection operator

OS

Overall survival

ROC

Receiver operating characteristic analysis

GSVA

Gene set variant analysis

ssGSEA

Single-sample gene set enrichment analysis

Author contributions

All authors contributed to this present work: [MYH] and [TSZ] designed the study, [LW] acquired the data, [TSZ] interpreted the data. [MYH] drafted the manuscript, [WWZ] revised the manuscript. All authors read and approved the manuscript.

Funding

This study was supported by National Superior Specialty of Traditional Chinese Medicine (Oncology), Ningbo Institute of Traditional Chinese Medicine, Oncology Research Institute with Tingsu Zhang, and Key Discipline of Haishu District (Internal Medicine of Traditional Chinese Medicine) with Wenwei Zhou.

Data availability

The datasets generated and/or analyzed during the current study are available in the [GSE32062] repository [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE32062].

Declarations

Ethics approval and consent to participate

Ethical approval was not required for this study because it is not involved any human experiments.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.