Pan‐Cancer Analysis of the Immunomodulatory Roles and Prognostic Value of PDXP and Experimental Verification in SKCM

ABSTRACT

Multiple public databases, including TCGA, GEO, and TARGET, were utilized to comprehensively analyze pyridoxal phosphatase (PDXP) expression profiles across 34 cancer types. We further explored the associations of PDXP expression with clinicopathological features, tumor immune microenvironment, genomic heterogeneity, functional signaling pathways, and therapeutic response. In addition, in vitro and in vivo experiments were performed to preliminarily validate the potential mechanism linking PDXP expression to melanoma progression. PDXP exhibited aberrant differential expression in most cancer types and tended to be highly expressed in malignant tumor cells. PDXP expression was correlated with overall survival outcomes in 14 cancer types. In most investigated tumors, PDXP expression was significantly associated with the expression of immunoregulatory and immune checkpoint genes, as well as the general landscape of the tumor immune microenvironment. Correlation analysis of genomic characteristics indicated that PDXP may serve as a potential immune‐related biomarker in skin cutaneous melanoma (SKCM). Functional enrichment and signaling pathway analysis showed that elevated PDXP expression in SKCM was correlated with enhanced tumor cell proliferative activity. In vitro and in vivo functional experiments demonstrated that PDXP knockdown significantly suppressed colony formation and migration of melanoma cells, induced G2/M phase cell cycle arrest, and increased cell apoptosis. These findings suggest that PDXP is associated with malignant phenotypes of melanoma cells, which are accompanied by changes in protein kinase B (AKT)/mammalian target of rapamycin (mTOR) signaling activity. Moreover, we established a prognostic model with moderate predictive performance for melanoma using least absolute shrinkage and selection operator (LASSO)‐Cox regression, providing a preliminary prognostic reference for SKCM patients.

1. Introduction

Cancer represents a major societal, public health, and economic burden in the 21st century, accounting for approximately one in six deaths (16.8%) globally and nearly one in four deaths (22.8%) attributable to noncommunicable diseases (NCDs) [1]. The occurrence and progression of cancer are highly complex and involve the combined effects of multiple factors. In recent years, accumulating evidence has highlighted the critical role of the tumor microenvironment (TME) across all stages of cancer development, ranging from tumor initiation, progression and invasion to metastatic dissemination and outgrowth. The TME includes diverse immune cell types, endothelial cells, cancer‐associated fibroblasts, pericytes, as well as multiple other tissue‐resident cell types [2]. Meanwhile, a growing number of studies have highlighted the pivotal role of the immune system in tumor drug resistance and metastasis [3]. The immune system is capable of recognizing and clearing abnormal cells such as malignant tumor cells [4]. However, tumor cells can develop multiple mechanisms to escape immune surveillance, thereby strengthening their drug resistance and metastatic potential [5, 6, 7]. Immune checkpoints act as key regulators in modulating immune responses, and tumor cells frequently hijack these pathways to inhibit the function of immune cells and avoid immune elimination [8]. Despite continuous advances in therapeutic strategies, numerous limitations still exist in the early diagnosis, targeted therapy and prognostic prediction of cancer. Owing to the continuous development and improvement of public databases including The Cancer Genome Atlas (TCGA), it has become increasingly feasible to identify novel diagnostic and prognostic biomarkers for cancer through pan‐cancer analysis of a single gene, as well as evaluation of their associations with clinical prognosis, immune infiltration and relevant signaling pathways [9].

Pyridoxal phosphatase (PDXP), also known as PLPP, chronophin or CIN [10], has been poorly investigated in previous studies, with most research focusing on its role in vitamin B6 metabolism. Vitamin B6, which is involved in more than 150 reactions in lipid, amino acid, glucose, and DNA metabolism [11], comprises three natural forms: pyridoxine (PN), pyridoxal (PL), and pyridoxamine (PM). It is their phosphorylated derivatives that account for the actual biological activity: pyridoxine 5′‐phosphate (PNP), pyridoxal 5′‐phosphate (PLP), and pyridoxamine 5′‐phosphate (PMP) [12]. The homeostasis of vitamin B6 is coordinately regulated by multiple enzymatic reactions: pyridoxal kinase (PDXK) catalyzes the phosphorylation of PL, PM, and PN, while PDXP converts them back to their non‐phosphorylated forms [13]. PLP acts as an essential coenzyme in cellular metabolism, and the catalytic activity of more than 140 enzymes relies on it [14]. PLP is increasingly studied for its substantial role in regulating carcinogenesis. Tumors exhibit divergent demands for PLP: cancer cells such as pancreatic ductal adenocarcinoma (PDAC) and clear cell renal cell carcinoma (ccRCC) rely on PLP to drive one‐carbon metabolism, nucleotide synthesis, sulfur metabolism and other processes, thereby meeting their proliferative and antioxidant needs. In contrast, in cancers including non‐small cell lung carcinoma (NSCLC) and colorectal cancer (CRC), PLP exerts anti‐tumor effects by inhibiting inflammatory pathways and maintaining genomic stability [15, 16, 17, 18].

Limited studies investigating the direct association of PDXP with diseases have mainly focused on neuropsychiatric disorders, glioblastoma (GBM), and acute myeloid leukemia (AML). PDXP plays a vital role in the latent phase of epileptogenesis through several signaling pathways, thus serving as a potential therapeutic target for epilepsy [19, 20]. Glioblastoma is the most aggressive primary brain tumor in adults [21]. PDXP acts as a glial tumor modifier, and its downregulation caused by aberrant promoter methylation is associated with shorter overall survival in glioblastoma patients. PDXP expression may regulate the interplay between glioma cell proliferation and invasion [22, 23]. In addition, PDXP promotes the proliferation of AML cells and inhibits apoptosis and differentiation by activating the PI3K/AKT/mTOR (phosphoinositide 3‐kinase/protein kinase B/mammalian target of rapamycin) signaling pathway [24, 25].

However, we found that comprehensive pan‐cancer analyses of PDXP, as well as studies on its prognostic significance, immunomodulatory roles and biological functions, are still lacking to date. Through preliminary pan‐cancer investigations, we identified that PDXP is highly expressed in melanoma, for which relevant research remains scarce. Melanoma arises from the malignant transformation of melanocytes, which are derived from the neural crest [26]. Melanoma has become a global health concern, with more than 330,000 new cases and 58,000 deaths reported worldwide in 2022 [27]. Immunotherapy has completely reshaped the therapeutic landscape of advanced melanoma: the combination of programmed death‐1 (PD‐1) inhibitors and the novel anti‐cytotoxic T‐lymphocyte‐associated antigen 4 (CTLA‐4) heavy chain‐only antibody HBM4003 has achieved a 40.0% objective response rate (ORR) in PD‐1 treatment‐naive mucosal melanoma patients, with factors such as the ratio of regulatory T cells (Treg) to CD4⁺ T cells in the tumor microenvironment serving as key predictive biomarkers for therapeutic efficacy [28]. This motivated us to explore predictive biomarkers for the efficacy of immunotherapy in melanoma. Our study investigated the correlations between PDXP expression and clinicopathological characteristics, the tumor immune microenvironment (TIME), genomic heterogeneity, functional pathways and therapeutic responses. Our findings may inform future research on PDXP‐based anti‐tumor therapies.

2. Materials and Methods

2.1. Data Retrieval and Preprocessing

The uniformly standardized pan‐cancer dataset, comprising TCGA, TARGET (Therapeutically Applicable Research to Generate Effective Treatments), and GTEx (Genotype‐Tissue Expression) (PANCAN, N = 19,131, G = 60,499), was obtained from the University of California, Santa Cruz (UCSC) Xena database (https://xenabrowser.net/). Subsequently, we extracted expression data for the gene ENSG00000241360 (PDXP) across various samples. We further filtered samples based on their origins: Solid Tissue Normal, Primary Solid Tumor, Primary Tumor, Normal Tissue, Primary Blood‐Derived Cancer‐Bone Marrow, and Primary Blood‐Derived Cancer‐Peripheral Blood. Additionally, we excluded samples with an expression level of 0 and applied a log2(x + 1) transformation to all expression values. To eliminate systematic variations and potential batch effects between the TCGA and GTEx cohorts, batch correction was performed using the ComBat algorithm of the R package sva (v3.46.0) during data preprocessing. Single‐cell transcriptome data of skin cutaneous melanoma (SKCM, GSE277165) were obtained from the Gene Expression Omnibus (GEO), with top 10 samples (GSM8515675‐GSM8515684) selected. Cancer type abbreviations followed standard TCGA nomenclature.

2.2. Differential Expression and Clinical Correlation Analysis

The Sangerbox database (http://vip.sangerbox.com/login.html) was used to analyze PDXP differential expression and clinical correlation across 34 cancer types (TCGA + GTEx cohorts). Expression data were log2(x + 1)‐transformed, and unpaired Student's t‐test (p < 0.05) was used to compare tumor‐normal expression differences. Unpaired Wilcoxon tests were applied to analyze PDXP expression differences among different clinical stages (T/N/TNM stages, p < 0.05). Violin plots and grouped bar charts were generated using the ggplot2 package in R. Results derived from unadjusted nominal p < 0.05 in pan‐cancer comparisons are regarded as exploratory and hypothesis‐generating observations.

2.3. Single‐Cell Sequencing and Spatial Transcriptome Analysis

Single‐cell RNA‐seq analyses were performed using R software (version 4.4.3) and the Seurat package. For quality control (QC), eligible cells were screened according to strict filtering criteria: the number of detected genes per cell (nFeature_RNA) was limited between 300 and 7000; the unique molecular identifier (UMI) count per cell (nCount_RNA) was greater than 1000, while the top 3% of cells with the highest UMI counts were excluded; the proportion of mitochondrial gene expression (mt_percent) was less than 10%, and the proportion of erythrocyte gene expression (HB_percent) was maintained below 3%. After removing low‐quality cells and genes, gene expression matrices were standardized and normalized for subsequent analysis.

Principal component analysis (PCA) was applied for dimensionality reduction, and the top 20 principal components were retained for downstream clustering analysis. Batch effect correction was performed using Harmony on single‐cell sequencing data from 10 melanoma samples. Cell clustering was conducted via the FindClusters function in Seurat with a resolution of 0.3, and cell clusters were visualized using Uniform Manifold Approximation and Projection (UMAP). The Wilcoxon rank‐sum test was used to screen differentially expressed genes (DEGs) among cell populations. The DEG screening criteria were set as follows: absolute log2 fold change > 1, gene expression detected in no less than 25% of cells in each group, and adjusted p‐value < 0.05. These rigorous filtering standards ensured the robustness and biological reliability of the identified DEGs. Biologically distinct cell populations were identified and manual cell type annotation of the clusters was carried out using well‐characterized marker genes. The AverageExpression function was used to calculate the average expression level of PDXP in each cell cluster.

Spatial transcriptome analysis was performed using an online bioinformatics analysis platform (https://www.grswsci.top/login/). 10x Genomics Visium spatial transcriptome data (SKCM1: Human Melanoma, IF Stained (FFPE), SKCM2: GSE179572‐GSM5420750) were imported into the platform, and valid spatial spots were retained after quality control. The single‐cell annotation results were mapped to spatial spots using the cell type mapping function built into the platform. Finally, the spatial expression profile of the PDXP gene was plotted using the visualization tools of the platform to visualize its spatial expression characteristics in tissues.

2.4. Prognostic Analysis

Sangerbox was used to evaluate PDXP's association with overall survival (OS) via Cox proportional hazards regression. Hazard ratios (HR) and their 95% confidence intervals (95% CI) were calculated, and the log‐rank test was used to determine prognostic significance. Patients in the most significant cancer cohort were divided into high/low PDXP groups by median expression, and Kaplan–Meier curves with the log‐rank test were used to compare survival differences. Unadjusted survival metrics across pan‐cancer cohorts (p < 0.05) serve as exploratory indicators for hypothesis generation.

2.5. Tumor Immune Microenvironment (TIME) Analysis

We used the Sangerbox database to evaluate the correlation between PDXP expression and immunoregulatory genes (chemokines, chemokine receptors, major histocompatibility complex (MHC) molecules, immunoinhibitors, immunostimulators) as well as immune checkpoint genes via Pearson correlation analysis. The CIBERSORT algorithm was applied to estimate the relative infiltration abundance of 22 immune cell subsets in tumor tissues, and the correlation between PDXP expression and infiltration levels of each immune cell subset was determined using Pearson correlation analysis. The ESTIMATE (Estimation of STromal and Immune cells in MAlignant Tumor tissues using Expression data) package was used to calculate StromalScore, ImmuneScore, and ESTIMATEScore. Immune correlation analyses across multiple cancer types based on nominal p < 0.05 are interpreted as exploratory observations.

2.6. Genomic Heterogeneity Analysis

We used Guangre Biotech (https://www.grswsci.top/) to determine the correlations between PDXP expression and genomic heterogeneity parameters (Aneuploidy Score, Ploidy, homologous recombination deficiency (HRD) Score, Silent Mutation Rate, Nonsilent Mutation Rate, single‐nucleotide variant (SNV) neoantigens) across 33 cancer types via Spearman correlation analysis. Additionally, in the Gene Expression and Mutation Landscape module of the Sangerbox database, nearly 100 mutation‐tested samples were selected for each cancer type, and the top 20 genes with the highest mutation frequencies were identified. The chi‐squared test was used to compare the differences in gene mutation frequencies between the PDXP high‐expression and low‐expression groups. All genomic heterogeneity associations identified via nominal p < 0.05 in pan‐cancer evaluations are considered hypothesis‐generating.

2.7. Functional Enrichment Analysis

We used the limma package to identify differentially expressed genes between the PDXP high‐expression and low‐expression groups across 33 cancer types, with data obtained from Guangre Biotech. Based on the Kyoto Encyclopedia of Genes and Genomes (KEGG) metabolic pathways and Hallmark signature pathways from the Molecular Signatures Database (MSigDB), gene set enrichment analysis (GSEA) was performed on these differentially expressed genes. For SKCM samples, 402 co‐expressed genes significantly correlated with PDXP expression were identified using the cor function in R, and subsequent KEGG pathway enrichment analysis and Gene Ontology (GO) functional enrichment analysis were conducted based on these genes. For enrichment analyses, the Benjamini‐Hochberg false discovery rate (FDR) correction was applied, and adjusted p‐values (q‐values) < 0.05 were considered statistically significant.

2.8. Cell Culture and Treatment

Human melanoma cell lines A2058 (cat. no. CBP60328) and A375 (cat. no. CBP60329) [29], as well as mouse melanoma cell line B16‐F10 (cat. no. CBP60335), were obtained from COBIOER Biosciences (Beijing, China). Cells were cultured in Dulbecco's modified Eagle's medium (DMEM) (cat. no. 11995‐065; Thermo Fisher Scientific Inc.) supplemented with 10% fetal bovine serum (FBS; cat. no. 16000‐044; Gibco; Thermo Fisher Scientific Inc.) and 1% penicillin–streptomycin (cat. no. 15140‐122; Gibco; Thermo Fisher Scientific Inc.) at 37°C in a humidified atmosphere containing 5% CO₂. Cells were seeded into 6‐well plates and cultured to 60%–70% confluence [30]. All cell lines utilized were authenticated by an expert before being used for experimentation, and were free from mycoplasma contamination.

2.9. Cell Transfection

Recombinant lentiviruses for PDXP knockdown (pLenti‐U6shRNA‐CMV‐GFP‐2A‐Puro vector, cat. no. CS5006368) and the corresponding empty vector were purchased from Applied Biological Materials Inc. The target sequences of lentivirus‐mediated short hairpin RNA (shRNA) were as follows: shControl: 5′‐GGGTGAACTCACGTCAGAAT‐3′; shPDXP #1: 5′‐GCTCTGTTTGTGAGCAACAAC‐3′; shPDXP #2: 5′‐AGCTCTTCAGCTCCGCGCTGTG‐3′; shPDXP #3: 5′‐CATCACGGAGAACTTCAGCAT‐3′. A2058, A375, and B16‐F10 cells were infected with the lentiviruses plus virus infection enhancer (cat. no. REVG005; Genechem) for 2 days [24]. Cells were selected with puromycin (2 μg/mL, cat. no. HY‐15695; MCE) for 2 weeks to ensure the establishment of stable cell lines [31]. The knockdown efficiency of PDXP was verified by Western blot.

2.10. Functional Assays

2.10.1. CCK‐8 Assay

A2058 and A375 cells, which had been transfected with shControl or shPDXP#3 lentiviral vectors, were plated in triplicate in 96‐well plates at a seeding density of 5 × 10³ cells per well, then cultured for 24 h to allow complete attachment. 10 μL of Cell Counting Kit‐8 (CCK‐8) reagent was added to each well. Following a 2‐h incubation at 37°C, the absorbance at 450 nm was measured using an Agilent BioTek 800 TS Absorbance Reader.

2.10.2. Colony Formation Assay

Transfected cells (500–1000/well, 6‐well plates) were incubated for 10 days, fixed with 4% paraformaldehyde for 15 min, stained with 0.5% crystal violet solution (cat. no. C3886; MilliporeSigma; Merck KGaA) for 30 min, and colonies (≥ 50 cells) were manually counted under a microscope at × 20 magnification. The experimental results were normalized to the control group.

2.10.3. Transwell Migration Assay

Cell migration ability was assessed using Transwell inserts with 8‐μm‐pore chambers (cat. no. 3422; Corning Inc.). Transfected cells were seeded in serum‐free DMEM in the upper chamber at a density of 2 × 10⁴ cells/well, while the lower chamber was filled with complete DMEM. After 24 h of incubation at 37°C, the cells that had migrated to the lower surface of the membrane were fixed and stained. Migrated cells were observed under an Olympus inverted microscope (model IX73, Olympus Corporation, Tokyo, Japan), and the number of migrated cells was counted in 5 randomly selected 200 × fields.

2.11. Flow Cytometry Assay

2.11.1. Cell Apoptosis Detection

For detection of cell apoptosis, 5 × 10⁵ A2058 and A375 cells were resuspended in 500 μL of diluted 1 × Annexin V Binding Buffer from the Annexin V‐APC/DAPI Apoptosis Kit (cat. no. E‐CK‐A258; Elabscience). Subsequently, 5 μL of Annexin V‐APC Reagent and 5 μL of DAPI Reagent from the same kit were added to the cell suspension. The mixture was incubated for 20 min in the dark. Apoptosis was assessed by Annexin V‐APC/DAPI double staining. Data were acquired on a Beckman Coulter Cytoflex LX flow cytometer and analyzed using FlowJo software.

2.11.2. Cell Cycle Analysis

To evaluate cell‐cycle distribution, 1 × 10⁶ A375 and A2058 cells were fixed in 500 μL of precooled 70% anhydrous ethanol (PBS/absolute ethanol ratio = 3:7) at 4°C overnight. After rinsing with PBS, cells were sequentially incubated with RNase A Reagent and Cycle‐Blue Reagent (Cell Cycle Assay Kit, cat. no. E‐CK‐A353; Elabscience) in the dark. Cell‐cycle profiles were acquired on a Beckman Coulter Cytoflex LX flow cytometer, and data were processed using ModFit LT software.

2.12. Western Blot Analysis

The total proteins from A2058 and A375 cells were extracted using RIPA buffer (cat. no. 89901; Thermo Fisher Scientific Inc.) containing protease inhibitors (cat. no. 36978; Thermo Fisher Scientific Inc.). The protein concentration was quantified by the Pierce BCA Protein Assay Kit (cat. no. 23225; Thermo Fisher Scientific Inc.). The proteins were separated by 10% SDS‐PAGE and transferred onto PVDF membranes (cat. no. 88518; Thermo Fisher Scientific Inc.). The membranes were blocked with 5% non‐fat milk in Tris‐buffered saline with Tween 20 (TBST) for 1.5 h and then incubated overnight at 4°C with the corresponding primary antibodies. The primary antibodies used were as follows: β‐actin (cat. no. TA‐09, 1:1000; ZSGB‐BIO), PDXP (cat. no. 27490‐1‐AP, 1:500; Proteintech), Bax (cat. no. 200958, 1:1000; ZENBIO), Bcl2 (cat. no. 250198, 1:1000; ZENBIO), Bcl‐xl (cat. no. R23603, 1:1000; ZENBIO), AKT (cat. no. 342529, 1:1000; ZENBIO), Phospho‐AKT (cat. no. 310021, 1:500; ZENBIO), mTOR (cat. no. R380411, 1:500; ZENBIO), Phospho‐mTOR (cat. no. 381557, 1:500; ZENBIO), and PTEN (cat. no. 321015, 1:500; ZENBIO). After incubation with primary antibodies, the membranes were incubated with the corresponding secondary antibodies (1:4000; Biosharp) at room temperature for 1 h. Protein signals were visualized using an enhanced chemiluminescence (ECL) kit (cat. no. WP20005; Thermo Fisher Scientific Inc.) and detected with an ECL visualization system.

2.13. In Vivo Experiments

Male C57BL/6J mice (7–9 weeks old, n = 5 per group) were housed under specific pathogen‐free (SPF) conditions with free access to food and water. Mice were randomly assigned to either the shControl or shPDXP#3 group using a random number table. Subcutaneous xenograft models were established by injecting 5 × 10⁵ B16‐F10 cells transfected with shControl or shPDXP#3 into the right flank of each mouse. Investigators measuring tumor volumes were blinded to group allocations. Tumor dimensions were measured every 2 days using digital calipers, and tumor volume was calculated using the formula: (tumor length × width²)/2 [32]. On day 14 post‐injection, mice were humanely euthanized, and excised tumors were weighed and photographed. All animal protocols were approved by the Animal Ethics Committee of Wannan Medical University (WNMC‐AWE‐2026223) and performed in strict accordance with the Guidelines for the Care and Use of Laboratory Animals.

2.14. Construction of Prognostic Model

The TCGA‐SKCM cohort was randomly partitioned into a training set (N = 314) and an internal test set (N = 135) in a 7:3 ratio using caret package in R with a fixed random seed to ensure reproducibility. To avoid survival information leakage, the 402 PDXP‐correlated candidate genes (identified via expression correlation in Section 2.7) were first evaluated by univariate Cox regression analysis within the training set (p < 0.05). Subsequently, LASSO‐Cox regression analysis (using glmnet and survminer packages) was applied to these prognostic genes to identify the key prognostic genes and establish the prognostic risk score model.

Patients in both cohorts were stratified into high‐ and low‐risk groups based on the median risk score of the training set. Kaplan–Meier (K‐M) survival analysis (using the survival package) and log‐rank tests were performed to compare overall survival (OS) between the groups. Time‐dependent receiver operating characteristic (ROC) curves (using the timeROC package) were plotted to calculate 1‐, 3‐, and 5‐year area under the curve (AUC) values. Univariate and multivariate Cox regression analyses were performed to evaluate whether the risk score was an independent prognostic factor after adjusting for clinicopathological variables. Finally, a clinical nomogram was constructed using the rms package, and its performance was assessed using calibration curves and the Concordance Index (C‐index).

2.15. Diagnostic Analysis

Data retrieved from the TCGA database were used to evaluate the potential diagnostic value of PDXP in cancer via ROC curves. An AUC value greater than 0.7 was considered to indicate high diagnostic value.

2.16. Therapeutic Sensitivity Analysis

We used Guangre Biotech to perform drug sensitivity analysis of PDXP in the PRISM and CTRP databases. Differentially expressed genes between PDXP high‐expression and low‐expression groups in each cancer type were submitted to the Connectivity Map (CMAP) database to identify drugs that can reverse the PDXP signature.

2.17. Statistical Analysis

Statistical associations and intergroup differences were evaluated using Spearman rank correlation analysis, Pearson correlation analysis, and the Wilcoxon rank‐sum test. Hazard ratios and prognostic factors were estimated using univariate and multivariate Cox proportional hazards regression models. All statistical computations and visualizations were performed using GraphPad Prism 10 software and RStudio (version 4.4.3). For high‐throughput pan‐cancer screening analyses where multi‐testing FDR correction was not strictly applicable, p < 0.05 was defined based on unadjusted nominal p‐values, and the corresponding findings were interpreted strictly as exploratory and hypothesis‐generating observations. Otherwise, an FDR/q‐value of less than 0.05 was considered statistically significant.

3. Results

3.1. Differential Expression of PDXP in Pan‐Cancer

We performed an exploratory analysis of PDXP expression profiles based on the TCGA combined with GTEx pan‐cancer datasets. PDXP was differentially expressed between tumor and normal tissues in 27 cancer types (Figure 1A). Compared with adjacent normal tissues, PDXP was significantly upregulated in 15 tumor types, including SKCM (p = 1.3e‐61), BRCA (p = 2.1e‐46), WT (p = 2.3e‐29), UCS (p = 3.6e‐15), ALL (p = 1.0e‐64), LAML (p = 2.2e‐88), and PCPG (p = 5.7e‐3). In contrast, PDXP was significantly downregulated in 12 cancer types, such as GBM (p = 9.4e‐51), GBMLGG (p = 3.6e‐84), LGG (p = 1.2e‐56), TGCT (p = 6.4e‐17), KICH (p = 2.0e‐13), and CHOL (p = 1.1e‐10). We further explored the correlation between PDXP expression and clinical stage across cancers. For T stage, PDXP was highly expressed in high T stage (T3 + T4) and lowly expressed in low T stage (T1 + T2) in BRCA and STES. An opposite pattern was observed in KIPAN and THCA, where PDXP levels were lower in high T stage than in low T stage (Figure 1B). For N stage, PDXP was elevated in high N stage in STES, KIRP, and HNSC, but reduced in high N stage in BRCA (Figure 1C). For overall TNM stage, PDXP differed significantly in four cancers, with higher expression in advanced stages in KIPAN (Figure 1D).

FIGURE 1.

PDXP showed significant differential expression in pan‐cancer. (A) Compared with normal tissues, PDXP was highly expressed in 15 tumors and low expressed in 12 tumors. (B) PDXP expression was associated with T stage in four tumor types. In BRCA and STES, PDXP was highly expressed in patients with high T stage (T3 + T4) and expressed at lower levels in patients with low T stage (T1 + T2). This expression relationship was reversed in KIPAN and THCA. (C) PDXP expression was associated with N stage in six tumor types. (D) In four tumor types, PDXP expression was correlated with TNM stage. Red font indicates significant differences. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

We then evaluated the diagnostic performance of PDXP using ROC curve analysis (Figure S1A–H). The AUC values reached 0.969 in SKCM (Figure S1A), 0.991 in PCPG (Figure S1C), and 0.941 in CHOL (Figure S1G). Moderate AUC values were also observed in KICH (Figure S1D) and TGCT (Figure S1H). Diagnostic analyses based on TCGA and GTEx may be affected by batch effects in data processing pipelines. We compared standalone TCGA datasets with TCGA datasets integrated with GTEx, and the results demonstrated that their diagnostic values were essentially similar, with several cancer types showing high diagnostic efficacy (Figure S1I). The above findings support its potential to serve as a diagnostic biomarker.

3.2. Single‐Cell Sequencing and Spatial Transcriptome Analysis of PDXP in Pan‐Cancer

To determine whether PDXP upregulation in tumors originates from malignant cells rather than stromal or immune cells, we analyzed pan‐cancer single‐cell RNA‐seq datasets. A heatmap (Figure 2A) showed the expression level of PDXP across diverse cell types in multiple cancers. PDXP was specifically enriched in malignant cells in SKCM, MCC, BRCA, LIHC, MB, NPC, OS, and PAAD.

FIGURE 2.

PDXP was specifically highly expressed in malignant cells. (A) PDXP was specifically highly expressed in malignant cells in SKCM, MCC, BRCA, LIHC, MB, NPC, OS, and PAAD. (B, C) SKCM samples (GSM8515675–GSM8515684) were clustered into 11 clusters by single‐cell sequencing analysis. (D) Bubble diagram of marker genes in single‐cell analysis annotation. (E) PDXP expression in the SKCM cell cluster and cancer stem cell cluster was significantly stronger and more widespread than in other clusters. (F) The average expression of PDXP in the cancer stem cell cluster was higher than that in the SKCM cell cluster. (G, H) Regions with high PDXP transcriptional activity were highly consistent with histologically defined tumor‐infiltrated areas.

Focusing on SKCM, dimensionality reduction and clustering of single‐cell data identified 11 distinct cell clusters (Figure 2B). The composition and proportion of these cell subpopulations across different samples were illustrated in Figure 2C. Distinct cell types were annotated using canonical marker genes, as depicted in the dot plot (Figure 2D).

The UMAP feature plot demonstrated that PDXP expression was notably stronger and more widespread in malignant SKCM cell clusters and cancer stem cell clusters compared with other stromal and immune cell populations (Figure 2E). Further quantitative analysis of relative average expression confirmed that PDXP levels were significantly elevated in cancer stem cell clusters and SKCM cell clusters, whereas its expression remained relatively low in non‐malignant cell types (Figure 2F). This preferential expression pattern suggests a potential link between PDXP and tumor stemness, malignant progression, and therapy resistance.

To validate these findings in the spatial context, we performed spatial transcriptomic profiling on two clinical sections of SKCM (SKCM‐1 and SKCM‐2) (Figure 2G,H). Spatial mapping of cell types revealed distinct histological architectures, where regions with prominent PDXP transcriptional activity were highly consistent with histologically defined tumor‐infiltrated areas, directly verifying the enrichment of PDXP in the core tumor regions.

3.3. Prognostic Value of PDXP in Pan‐Cancer

Univariate Cox regression analysis was performed to assess the prognostic value of PDXP. PDXP expression was significantly associated with overall survival (OS) in 14 cancer types (Figure 3A). High PDXP expression acted as a survival‐associated risk factor for poor OS in SKCM‐M (HR = 1.33, 95% CI = 1.14–1.56, p = 3.6e‐4), SKCM (HR = 1.26, 95% CI = 1.09–1.46, p = 1.7e‐3), KICH (HR = 3.09, 95% CI = 1.42–6.73, p = 3.4e‐3), BRCA (HR = 1.33, 95% CI = 1.07–1.65, p = 9.0e‐3), TARGET‐ALL (HR = 1.58, 95% CI = 1.09–2.30, p = 0.02), ACC (HR = 1.49, 95% CI = 1.06–2.09, p = 0.02), and TARGET‐ALL‐R (HR = 1.32, 95% CI = 1.04–1.67, p = 0.02). Conversely, high PDXP expression was identified as a survival‐associated protective factor in GBMLGG (HR = 0.58, 95% CI = 0.51–0.67, p = 3.9e‐13), LGG (HR = 0.67, 95% CI = 0.54–0.83, p = 1.8e‐4), PAAD (HR = 0.74, 95% CI = 0.60–0.91, p = 3.9e‐3), KIRC (HR = 0.79, 95% CI = 0.66–0.94, p = 7.7e‐3), STES (HR = 0.83, 95% CI = 0.70–0.98, p = 0.03), THYM (HR = 0.60, 95% CI = 0.37–0.99, p = 0.03), and PCPG (HR = 0.37, 95% CI = 0.14–0.96, p = 0.04).

FIGURE 3.

Correlation of PDXP expression level with overall survival (OS) in tumor patients. (A) PDXP expression was significantly correlated with OS in 14 cancers. In SKCM, KICH, BRCA, TARGET‐ALL, and ACC, high PDXP expression was a survival‐associated risk factor for poor OS. In contrast, in GBMLGG, LGG, PAAD, KIRC, STES, THYM, and PCPG, low PDXP expression predicted poor OS. (B–D) There were significant survival differences between high and low PDXP expression groups in most cancers. Low PDXP expression was significantly associated with longer survival time in ACC, KICH, BRCA and SKCM‐M. In contrast, in GBMLGG (p = 1.5e‐8), STES (p = 0.03), LGG (p = 0.01), and KIRC (p = 0.04), high PDXP expression was associated with longer survival time.

Kaplan–Meier survival curves further validated these findings (Figure 3B–D). Low PDXP expression was significantly correlated with longer survival in ACC (p = 6.4e‐3), KICH (p = 4.6e‐3), BRCA (p = 2.3e‐3), and SKCM‐M (p = 0.02). In contrast, high PDXP expression predicted better survival in GBMLGG (p = 1.5e‐8), STES (p = 0.03), LGG (p = 0.01), and KIRC (p = 0.04). These exploratory observations indicate that PDXP is associated with overall survival outcomes in multiple cancers, potentially serving as an unfavorable prognostic indicator in certain malignancies while showing protective tendencies in others.

3.4. Relationship Between PDXP Expression and Tumor Immune Microenvironment

We systematically analyzed the correlation between PDXP and the tumor immune microenvironment. PDXP expression was generally positively correlated with immunoregulatory genes in ACC, SKCM, KICH, WT, KIRP, KIRC, KIPAN, UVM, UCEC, and ALL, but negatively correlated in THYM, DLBC, GBMLGG, LGG, CHOL, and CESC (Figure 4A). For immune checkpoint genes, PDXP was positively correlated in ACC, WT, KIRP, KIRC, KIPAN, UVM, KICH, and UCEC, and negatively correlated in DLBC, GBMLGG, LGG, THCA, CESC, BLCA, and LUAD (Figure 4B).

FIGURE 4.

Correlation analysis between PDXP expression and immune‐related gene expression. (A) In ACC, SKCM, KICH, WT, KIRP, KIRC, KIPAN, UVM, UCEC, and ALL, the expression of PDXP was generally positively correlated with the expression of immunoregulatory genes. (B) In ACC, WT, KIRP, KIRC, KIPAN, UVM, KICH, and UCEC, high PDXP expression was significantly positively correlated with high expression of immune checkpoint genes. (C) PDXP expression was closely related to the infiltration abundance of immune cells across cancer types. (D) The infiltration abundance of M0 macrophages and resting NK cells was significantly positively correlated with PDXP expression in SKCM, while the infiltration abundance of CD8⁺ T cells was negatively correlated.

PDXP expression was also closely associated with immune cell infiltration (Figure 4C). In SKCM, PDXP was positively correlated with M0 macrophage and resting NK cell infiltration, but negatively correlated with CD8+ T cell abundance (Figure 4D). Furthermore, PDXP was negatively correlated with ImmuneScore, StromalScore, and ESTIMATEScore in THCA, SKCM, SKCM‐M, SARC, PRAD, LUAD, LGG, GBMLGG, BRCA, ACC, WT, NB, and TARGET‐ALL‐R (Figure 5A–D), suggesting that high PDXP expression may be accompanied by reduced immune and stromal cell infiltration.

FIGURE 5.

Association of PDXP expression level with pan‐cancer immune infiltration scores. (A) The correlation heat map showed that the expression level of PDXP was significantly correlated with ImmuneScore, StromalScore, and ESTIMATEScore. (B–D) In THCA, SKCM, SKCM‐M, SARC, PRAD, LUAD, LGG, GBMLGG, BRCA, ACC, WT, NB, and TARGET‐ALL‐R, the expression of PDXP was significantly negatively correlated with ESTIMATE, Immune, and Stromal scores. In THYM, PDXP expression was positively correlated with ESTIMATE and Immune scores. *p < 0.05, **p < 0.01, ***p < 0.001.

Notably, multiple analytical results indicated that PDXP is predominantly enriched in malignant tumor cells. Accordingly, the negative correlations of PDXP with ImmuneScore and StromalScore observed in our study may partially stem from elevated tumor purity rather than a direct immunomodulatory effect of PDXP. To address potential confounding effects of tumor purity, we validated the correlations between PDXP and specific immune cells using EPIC and MCP‐counter across the pan‐cancer cohorts (Figure S4A,B). In gliomas (GBMLGG and GBM), we observed an apparent discrepancy: PDXP expression was negatively correlated with the overall ESTIMATE ImmuneScore (Figure 5A–D) but showed strong positive correlations with CD4⁺ T cells and general T cell lineages (Figure S4A,B). This discrepancy likely reflects the severe tumor‐purity dilution in highly aggressive, PDXP‐high gliomas.

Unlike gliomas, SKCM showed consistent results across all methodologies: PDXP expression was significantly negatively correlated with CD8⁺ T cells and cytotoxic lymphocytes. These multi‐algorithm analyses consistently suggest a potential association between PDXP and an immunosuppressive microenvironment in SKCM.

3.5. Correlation Analysis Between PDXP Expression and Pan‐Cancer Genomic Characteristics

We evaluated the correlation between PDXP and six genomic instability indicators across 33 cancers (Figure 6A). PDXP was positively correlated with nonsilent mutation rate, SNV neoantigens, and HRD in BRCA, UCEC, STAD, SKCM, SARC, PRAD, KICH, and LUAD, but negatively correlated in THYM and PAAD, indicating a broad role in genome stability regulation. We further constructed mutation landscapes of the top 20 frequently mutated genes in four cancers (Figure 6B–E). In SKCM and STAD, the high PDXP group exhibited higher mutation density and tumor mutation burden (TMB) than the low PDXP group. In SKCM (Figure 6B), the high PDXP subgroup showed frequent mutations in PCLO (49.3%) and HYDIN (38.4%). In STAD (Figure 6E), PTEN mutation frequency was significantly increased in the high PDXP group (57.9%, p = 0.02). These observations support a potential association between aberrant PDXP expression and genomic instability as well as tumor malignant progression. Collectively, PDXP may serve as a potential immune‐related biomarker, though further clinical validation is still required to confirm its clinical applicability.

FIGURE 6.

Correlation between PDXP expression levels and tumor genomic stability. (A) The radar chart shows that in BRCA, UCEC, STAD, SKCM, SARC, PRAD, KICH, and LUAD, the expression level of PDXP was significantly positively correlated with Nonsilent Mutation Rate, SNV Neoantigens, and HRD, while it was significantly negatively correlated in THYM and PAAD. (B–E) In SKCM and STAD, the high PDXP expression group showed significantly higher mutation density and tumor mutation burden (TMB) than the low expression group, which was consistent with the results of the radar chart. (B) In SKCM, the high expression group was enriched with frequent mutations of genes such as PCLO (49.3%) and HYDIN (38.4%). (E) In STAD, the mutation frequency of PTEN (57.9%, p = 0.02) was significantly increased in the high expression group. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Drug sensitivity analyses based on the PRISM (Figure S2A) and CTRP (Figure S2B) cell line databases revealed that PDXP expression was significantly correlated with the sensitivity of multiple clinical drugs. Specifically, PDXP expression showed significant correlations with responses to epidermal growth factor receptor (EGFR) inhibitors, including afatinib, lapatinib, and gefitinib (p < 0.001). These results suggest that high PDXP expression may be associated with a decreased response to EGFR‐targeted therapy. Furthermore, Connectivity Map analysis (Figure S2C) identified MS‐275 and AH‐6809 as candidate small‐molecule compounds capable of reversing the oncogenic transcriptomic signature associated with elevated PDXP expression.

3.6. Functional Enrichment and Signaling Pathway Analysis of PDXP in Pan‐Cancer

GSEA was performed to explore biological functions and pathways associated with PDXP. A bubble heatmap (Figure 7A) showed that PDXP was positively correlated with DNA Repair, G2M Checkpoint, and E2F Targets in most cancers, supporting a role in cell proliferation. Meanwhile, PDXP was negatively correlated with interferon response and inflammatory pathways, implying immune suppression. Pathway activity analysis (Figure 7B) indicated that PDXP was positively associated with the cell cycle pathway across cancers. In SKCM, PDXP expression positively correlated with cell cycle and tuberous sclerosis complex (TSC)/mTOR signaling activities, suggesting a pro‐tumor role via enhanced proliferation and survival.

FIGURE 7.

Association of PDXP expression with multiple signaling pathways and biological functions. (A) PDXP expression was significantly positively correlated with DNA Repair, G2M Checkpoint, and E2F Targets in most cancer types. (B) PDXP expression was highly enriched in the Cell Cycle pathway across most cancer types. In SKCM, PDXP expression was potentially associated with the activation of cell cycle and TSC/mTOR pathways, while inversely associating with the Apoptosis pathway. (C) A total of 402 genes significantly correlated with PDXP expression in SKCM were selected for enrichment analysis. (D) KEGG pathway analysis revealed potential enrichment in the mTOR signaling pathway, lysosome, and cell cycle pathways. (E) GO functional enrichment analysis showed significant enrichment in mitochondrial function and protein metabolic processes. ***p < 0.001.

We selected 402 PDXP‐correlated genes in SKCM (Figure 7C) for enrichment analysis. KEGG analysis (based on nominal p < 0.05, Figure 7D) revealed potential trends toward enrichment in mTOR signaling, lysosome, and cell cycle pathways. These findings suggest a potential exploratory link between PDXP and cellular proliferation pathways. GO analysis (Figure 7E) showed statistically significant enrichment in cellular components related to mitochondrial function and protein metabolism (adjusted p < 0.05), indicating additional roles in cellular homeostasis.

3.7. Functional Verification of PDXP in Melanoma

A series of in vitro and in vivo experiments was performed to validate the bioinformatic findings. Western blot confirmed efficient knockdown of PDXP in A375 and A2058 melanoma cells (Figure 8A,B). CCK‐8 assays showed that cell viability was significantly reduced in the shPDXP#3 group compared with the negative control (NC) group (Figure 8C,D), indicating inhibited proliferation. Colony formation assays revealed fewer and smaller colonies in the knockdown group (Figure 8E). Transwell assays demonstrated suppressed melanoma cell migration (Figure 8F). In a mouse syngeneic model, PDXP knockdown markedly reduced tumor volume and weight (Figure 8G,H).

FIGURE 8.

PDXP knockdown suppressed the proliferation and migration of melanoma cells. (A, B) Western blot results showed that PDXP was effectively knocked down in human melanoma cell lines A375 and A2058. (C, D) CCK‐8 assays showed that the OD value in the shPDXP#3 group decreased significantly over time compared with the NC group. (E) Colony formation assay showed the number and size of colonies in the shPDXP#3 group were significantly lower than those in the control group. (F) The Transwell assay showed that the number of cells passing through the chamber in the shPDXP#3 group was significantly reduced. (G, H) In the syngeneic mouse model, PDXP knockdown significantly reduced tumor volume and weight compared with the shControl group. **p < 0.01, ***p < 0.001.

Flow cytometry revealed G2/M phase arrest in shPDXP#3 cells (Figure 9A) and increased early and late apoptosis (Figure 9B), supporting a role for PDXP in melanoma cell survival. Western blot showed downregulation of Bcl‐2 and Bcl‐xL and upregulation of Bax (Figure 9C), indicating activation of the intrinsic apoptotic pathway. Moreover, PDXP knockdown was accompanied by decreased phosphorylation levels of AKT and mTOR and increased PTEN expression (Figure 9D), suggesting a potential association between PDXP loss and suppressed AKT/mTOR signaling activity.

FIGURE 9.

PDXP promotes melanoma cell proliferation and suppresses apoptosis potentially via the AKT/mTOR signaling pathway. (A) Flow cytometry analysis showed abnormal accumulation of cells in the G2/M phase in the shPDXP#3 group. (B) The proportions of early apoptotic (Q3) and late apoptotic (Q2) cells increased significantly in the shPDXP#3 group. (C) Western blot results showed that PDXP knockdown downregulated the expression of Bcl‐2 and Bcl‐xL, and upregulated the expression of Bax. (D) PDXP knockdown significantly inhibited the phosphorylation of AKT and mTOR and increased PTEN expression. **p < 0.01, ***p < 0.001.

3.8. Construction and Validation of the PDXP‐Related Prognostic Signature

To construct a robust prognostic model, LASSO‐Cox analysis was performed on candidate genes in the TCGA training cohort (N = 314), identifying 14 key prognostic genes (Figure 10A,B). A multivariate Cox proportional hazards model was then established to calculate risk scores (Figure 10C).

FIGURE 10.

PDXP‐related genes were utilized to construct a prognostic model in SKCM. (A–B) LASSO‐Cox regression analysis and determination of the optimal penalty parameter (λ) for candidate prognostic genes in the TCGA training cohort. (C) Forest plot illustrating the multivariate Cox proportional hazards regression analysis of the 14 key prognostic genes. (D–G) Validation of the prognostic signature in the TCGA training cohort (N = 314): (D) Kaplan–Meier survival curve showing significantly worse prognosis in the high‐risk group compared with the low‐risk group (p < 0.001). (E) Time‐dependent ROC curves predicting 1‐, 3‐, and 5‐year overall survival (OS), with AUC values of 0.807, 0.711, and 0.748, respectively. (F, G) Distribution of risk scores and survival status, illustrating that mortality rates increased as risk scores rose. (H–K) Validation of the prognostic signature in the independent TCGA test cohort (N = 135): (H) Kaplan–Meier survival curve demonstrating a significantly shorter OS in high‐risk patients (p = 0.007). (I) Time‐dependent ROC curves predicting 1‐, 3‐, and 5‐year OS, with AUC values of 0.662, 0.685, and 0.693, respectively. (J, K) Distribution of risk scores and patient survival status in the test cohort.

Based on the median risk score, patients in both the training (N = 314) and internal test (N = 135) cohorts were divided into high‐ and low‐risk groups (Figure 10D–K). Kaplan–Meier curves showed that high‐risk patients had significantly worse overall survival (OS) in both the training (Figure 10D) and test cohorts (Figure 10H). ROC analysis confirmed high predictive accuracy, with 1‐, 3‐, and 5‐year AUCs of 0.807, 0.711, and 0.748 in the training cohort (Figure 10E), and 0.662, 0.685, and 0.693 in the internal test cohort (Figure 10I). High risk scores consistently correlated with increased mortality (Figure 10F,G,J,K).

For clinical application, a nomogram integrating the risk score and key clinical variables was constructed (Figure S3A), with 1‐, 3‐, and 5‐year calibration curves showing consistency between predicted and actual survival (Figures S3B–D). Finally, univariate (Figure S3E) and multivariate (Figure S3F) Cox analyses confirmed that the risk score was an independent prognostic predictor for SKCM patients after adjusting for clinical factors.

4. Discussion

In the present study, we conducted a comprehensive pan‐cancer analysis combined with in vitro and in vivo experiments to systematically explore the expression characteristics, potential clinical significance, biological functions, and underlying molecular mechanisms of PDXP. Bioinformatic results preliminarily indicated that PDXP exhibited aberrant differential expression between tumor and normal tissues across 27 cancer types, and its expression level was potentially correlated with clinical tumor stage in multiple malignancies. ROC analysis suggested the potential diagnostic value of PDXP for tumor discrimination. Pan‐cancer single‐cell RNA‐sequencing and spatial transcriptomic analyses consistently indicated that PDXP was specifically enriched and highly expressed in malignant tumor cells, especially in melanoma.

Prognostic correlation analysis revealed that PDXP expression was associated with overall survival in 14 cancer types. In most cancer types, PDXP expression showed potential associations with the expression of immunoregulatory genes, immune checkpoint genes, and the infiltration abundance of diverse immune cell subsets, and was negatively correlated with ESTIMATE, immune, and stromal scores. Pan‐cancer genomic instability analysis further showed that PDXP plays a broad role in genome stability regulation. Mutation landscapes of the top 20 frequently mutated genes in SKCM further support a potential association between aberrant PDXP expression and genomic instability as well as tumor malignant progression. Database‐based drug sensitivity analysis suggested a potential trend toward decreased EGFR‐targeted therapy response in tumors with high PDXP expression.

GSEA results revealed that high PDXP expression was closely enriched in cell proliferation and immunosuppressive signaling pathways. In SKCM, PDXP‐related genes showed potential enrichment trends in mTOR, lysosome, and cell cycle pathways, suggesting that PDXP is associated with melanoma cell cycle and AKT/mTOR signaling regulation. Consistent with bioinformatic predictions, our in vitro functional experiments showed that PDXP knockdown reduced proliferation, colony formation, and migration of melanoma A375 and A2058 cells, induced G2/M phase cell cycle arrest, and promoted apoptosis, which was accompanied by downregulation of AKT/mTOR signaling activity. Moreover, in vivo B16‐F10 xenograft mouse models further confirmed that PDXP deficiency effectively restrained melanoma tumor growth. Additionally, we constructed a PDXP‐related prognostic model for SKCM, providing a preliminary exploratory tool for melanoma prognostic evaluation. It should be noted that due to the unadjusted nominal p‐values (p < 0.05) in multi‐cohort screening, all pan‐cancer correlation analyses in this study serve as exploratory and hypothesis‐generating observations, warranting further clinical and biological validation.

Notably, the biological functions and clinical implications of PDXP may exhibit context‐dependent dualities and tissue‐specific inconsistencies across different malignancies, which is closely linked to the divergent demands of various tumors for its substrate, pyridoxal 5′‐phosphate (PLP). As outlined in our introduction, while certain cancers (such as PDAC and ccRCC) highly rely on PLP to drive crucial metabolic processes for proliferation, others (like NSCLC and CRC) show susceptibility to PLP, where PLP exerts anti‐tumor effects by inhibiting inflammatory pathways and maintaining genomic stability [15, 16, 17, 18]. Consequently, the regulatory impact of PDXP—which converts PLP back to its non‐phosphorylated form—could be highly divergent. This duality is further supported by previous literature: while PDXP downregulation is associated with shorter survival in glioblastoma patients [22, 23], acting potentially as a glial tumor modifier, it functions as an oncogenic driver in acute myeloid leukemia by promoting cell proliferation [24, 25]. Our findings primarily substantiate the tumor‐promoting role of PDXP in cutaneous melanoma, but we fully acknowledge that its systemic effects across the entire pan‐cancer landscape remain complex and potentially contradictory, necessitating cautious and cancer‐specific interpretations in future functional studies.

We conducted a comprehensive analysis to elucidate and validate the role of PDXP across pan‐cancer, using approaches spanning bioinformatics and in vitro and in vivo experiments. This study has several strengths: (1) it is the first pan‐cancer analysis focused on the PDXP gene; (2) it explores the potential roles of PDXP from multiple perspectives, including clinical relevance, immunity, molecular mechanisms, and targeted therapy; (3) it combines extensive in vitro and in vivo experiments in melanoma to verify the molecular mechanisms and biological functions of PDXP; and (4) it constructs a prognostic model with moderate predictive performance for SKCM.

Nevertheless, our study has certain limitations. First, we performed a large number of statistical analyses across 33 or 34 cancer types. Given the scale of the pan‐cancer analyses, there is a substantial risk of generating false‐positive results. Moreover, the statistical methods mainly rely on p < 0.05 as the significance threshold, and all conclusions drawn solely on this basis should be regarded as exploratory. Second, the negative correlations of PDXP expression with ImmuneScore and StromalScore may partially reflect higher tumor purity rather than direct immunomodulatory activity. Although PDXP expression exhibits strong correlations with genomic instability indicators and various immune checkpoints, due to the absence of verification in independent melanoma cohorts with available clinical data on immune checkpoint inhibitor (ICI) therapy response, these findings remain associative and exploratory. Third, the half‐maximal inhibitory concentration (IC50) prediction and CMap analyses were computationally simulated using pharmacogenomics databases; these cell‐line‐based predictions cannot account for critical clinical pharmacology variables, and the predicted candidates therefore remain associative and exploratory. Fourth, the precise mechanism underlying the relationship between PDXP and the tumor immune microenvironment remains unexplored and lacks sufficient experimental validation; further investigations are warranted to clarify its potential biological functions in tumor immune regulation. Fifth, although functional in vitro and in vivo assays were performed to evaluate the role of PDXP in melanoma, the primary loss‐of‐function experiments relied on a single shRNA sequence, and rescue experiments were not conducted; thus, further validation with multiple targeting sequences and gene‐restoration assays is required to establish definitive causal mechanisms. Sixth, while an association between PDXP expression and AKT/mTOR pathway phosphorylation was observed, detailed upstream and downstream signaling cascades remain to be elucidated. Seventh, we only validated the expression pattern, underlying mechanism, and correlation of PDXP with tumor cell proliferation and metastasis in melanoma cell lines; future studies can extend these validations to multiple tumor types. Finally, our bioinformatics analyses were mainly based on public databases (TCGA, TARGET, and GTEx); these database‐based results should be interpreted with caution due to potential batch effects and heterogeneous sample backgrounds in public transcriptomic datasets. Additional external validations and clinical samples are required to validate the prognostic and diagnostic value of PDXP. Furthermore, pharmacological experimental verification, prospective clinical trials, and independent ICI‐treated patient cohorts are critically needed to validate the predictive capacity and translational value of PDXP.

5. Conclusion

In summary, we performed a comprehensive pan‐cancer analysis to elucidate the potential role of PDXP, with further mechanistic exploration and functional validation conducted specifically in melanoma. Additionally, exploratory diagnostic and prognostic risk models were constructed based on PDXP expression, offering preliminary frameworks for clinical patient stratification. These findings provide new insights into the role of PDXP in tumorigenesis and cancer progression and suggest that PDXP could serve as a potential immune‐related prognostic biomarker and a candidate therapeutic target. Nevertheless, further extensive studies and independent clinical cohorts are still warranted to validate these preliminary findings and clarify the precise mechanisms underlying these associations.

Author Contributions

Yingjian Hu: investigation, conceptualization, writing – original draft, writing – review and editing, visualization, validation, methodology, software, formal analysis, project administration, resources, supervision, data curation. Xinzhu Zeng: methodology, software, investigation, data curation, writing – original draft, visualization. Wei Cao: conceptualization, methodology, software, investigation, writing – review and editing, writing – original draft, visualization, validation. Jianhao Jiao: conceptualization, formal analysis, data curation, supervision, writing – review and editing. Yuan Zhao: software, data curation, writing – original draft, visualization. Jun Wang: conceptualization, formal analysis, investigation, project administration, writing – review and editing, funding acquisition, supervision.

Funding

This work was funded by the Open Project of the Key Laboratory of Dermatology (Anhui Medical University), Ministry of Education (AYPY2024‐10).

Ethics Statement

The animal experiment was approved by the Animal Ethics Committee of Wannan Medical University and was performed in accordance with the Guidelines for the Care and Use of Laboratory Animals (WNMC‐AWE‐2026223). This study also analyzed publicly available, anonymized human genomic and clinical datasets (e.g., TCGA, GTEx, and GEO databases), which do not require additional institutional ethical approval or direct patient informed consent. Our study is not a clinical trial, and it does not involve any human samples. Therefore, no clinical trial ethical registration or related registration information is required.

Consent

All authors gave their consent for publication.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: ROC analysis for evaluating the potential diagnostic performance of PDXP in multiple cancers. (A) The AUC value of PDXP in SKCM was 0.969. (B) The AUC value of PDXP in UCS was 0.901. (C) The AUC value of PDXP in PCPG was 0.991. (D) The AUC value of PDXP in KICH was 0.802. (E) The AUC value of PDXP in GBM was 0.864. (F) The AUC value of PDXP in DLBC was 0.867. (G) The AUC value of PDXP in CHOL was 0.941. (H) The AUC value of PDXP in TGCT was 0.797. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. (I) Comparison of diagnostic values between standalone TCGA datasets and TCGA datasets integrated with GTEx.

Figure S2: Exploratory analysis of the correlation between PDXP expression and predicted clinical drug sensitivity. (A, B) Analyses of the PRISM and CTRP databases showed that PDXP expression was highly correlated with the drug‐sensitivity AUC of EGFR inhibitors (e.g., afatinib and lapatinib), with high statistical significance in both datasets (p < 0.001). (C) Connectivity Map analysis identified MS‐275 and AH‐6809 as candidate compounds with gene‐expression reversal profiles.

Figure S3: Construction of a clinical nomogram and assessment of independent prognostic value. (A) A prognostic nomogram integrating the 14‐gene risk score and key clinical variables (age, T stage, N stage, and stage) to predict 1‐, 3‐, and 5‐year OS probability for SKCM patients. (B–D) Calibration curves of the nomogram for predicting 1‐, 3‐, and 5‐year OS. (E‐F) Forest plots of univariate (E) and multivariate (F) Cox regression analyses. The multivariate analysis adjusting for clinicopathological factors confirmed that the risk score remained an independent prognostic predictor for patients with SKCM.

Figure S4: Multi‐algorithm validation of the correlation between PDXP expression and tumor‐infiltrating immune/stromal cells across pan‐cancer cohorts. (A) Heatmap illustrating the correlation between PDXP expression and eight immune/stromal cell types deconvoluted by the EPIC algorithm. Red tiles indicate positive correlations, while blue tiles indicate negative correlations. (B) Heatmap illustrating the correlation between PDXP expression and ten immune/stromal cell populations estimated by the MCP‐counter algorithm. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

Data S1: Supporting Information.

Data Availability Statement

Data supporting the findings are available from the GEO database (https://www.ncbi.nlm.nih.gov/geo/) and UCSC Xena Browser (https://xenabrowser.net/). Bioinformatic analyses were performed using Sangerbox and Guangre Biotech platforms. The original contributions generated in this study are included in the article and Supporting Information. Further inquiries can be directed to the corresponding authors.