Expression of PDZ domain-containing proteins is correlated with prognosis and immune infiltration in hepatocellular carcinoma

Introduction

Primary liver cancer was the sixth most prevalent cancer and the third leading cause of cancer-related deaths worldwide in 2022, with approximately 865,269 new cases and 757,948 deaths (1). The National Cancer Center of China reported 316,500 primary liver cancer-related deaths in 2022, accounting for 12.3% of all malignant tumor deaths in the country. Specifically, both the number of deaths and mortality rates ranked second (2). Hepatocellular carcinoma (HCC) is the most frequent primary liver cancer, accounting for 75–85% of all diagnosed cases (1). Current treatment approaches for HCC include maximal surgical resection, chemotherapy, molecular targeted therapy, immunotherapy, transarterial chemoembolization, and radiotherapy. However, HCC heterogeneity leads to varied treatment responses, high relapse rates, and low 5-year survival rates (3,4). Therefore, accurate assessment of patient prognosis and treatment responsiveness is essential for personalized precision therapy, alongside the identification of new treatment targets and strategies.

Proteins perform diverse functions and are involved in almost all biological processes. Protein-protein interactions (PPIs) are crucial in various physiological cellular pathways, including gene expression, cell growth, proliferation, nutrient absorption, metabolism, movement, intercellular signaling, and apoptosis (5). Protein structural domains serve as fundamental units mediating PPIs, and numerous critical domains [e.g., the Tudor domain (6) and coiled-coil domain (7)] have been utilized as focal points for identifying key genes and developing prognostic models. The PDZ domain is a common PPI module with 1,163 PDZ domains across 484 proteins in the human genome (8). The PDZ domains are relatively small protein modules that typically contain 80–110 amino acid residues (9), and despite low sequence identity among PDZ domains, their secondary structure is highly conserved. A PDZ domain usually comprises six β-strands and two α-helices (10). Furthermore, the PDZ domain exhibits a complex dynamic allostery, facilitating its binding with multiple partners, with the C-terminal PDZ-binding motif being the most common partner. The C-terminal motif has three types based on its sequence: type I, type II, and type III (11). PDZ domains exist in multidomain scaffolds and anchoring proteins, helping organize intracellular enzymes and receptors for signal transduction (9). PDZ domain-containing proteins (PDZ proteins) participate in numerous signaling pathways and have been associated with various diseases and disorders, including neurodegenerative diseases, psychiatric disorders, hearing and visual impairments, metabolic disorders, and cardiovascular conditions. Specifically, 145 of the 151 PDZ proteins are implicated in various cancers, playing critical roles in cancer initiation and metastasis (9).

While the role and prognostic significance of individual PDZ proteins in HCC has been widely studied (12,13), systematic analyses of multiple PDZ proteins remain scarce. For instance, DLG5 expression is significantly reduced in HCC tissues compared with adjacent normal tissues and inhibits invadopodia formation via Girdin and TKS5 (14,15). However, another study has reported a tumor-promoting role for DLG5 (16). Similarly, PARD3 has been shown to promote tumorigenesis in HCC by activating the Sonic Hedgehog signaling and Akt pathways, and its high expression is associated with poor prognosis (17,18). These findings underscore the complex roles of PDZ proteins in HCC progression

Recent studies have highlighted the crucial role of the tumor immune microenvironment in HCC progression (19,20). Immune cell infiltration, particularly by T cells, natural killer (NK) cells, and tumor-associated macrophages, significantly influences tumor growth, metastasis, and patient prognosis (21). For instance, hepatitis B virus (HBV)-associated HCCs can inhibit antitumor CD8⁺ T cell responses via the long noncoding RNA HDAC2-AS2, thereby promoting immune evasion and tumor progression (22). Additionally, AKR1D1 has been shown to suppress liver cancer progression by promoting bile acid metabolism, which may enhance NK cell activity and modulate the immune microenvironment (23). Moreover, PDZ domain-containing proteins have been implicated in regulating immune signaling pathways (24,25), suggesting their potential roles in modulating immune responses in HCC. Understanding the interplay between PDZ proteins and immune infiltration may provide novel insights into therapeutic strategies targeting both tumor cells and the immune microenvironment.

Accurate prediction of HCC prognosis is crucial for guiding treatment decisions. Currently, serum alpha-fetoprotein (AFP) levels, tumor-node-metastasis (TNM) staging, and histological grading are commonly used to assess prognosis (26). However, these traditional indicators are limited in capturing tumor heterogeneity. To improve prediction accuracy and clinical applicability, various prognostic models based on protein expression, mRNA, lncRNA, miRNA, ctDNA, and single nucleotide polymorphisms (SNPs) have been proposed (27). Nonetheless, existing models face two major limitations: first, the prognostic genes identified through high-throughput screening often lack clear biological significance, hindering clinical translation; second, most models do not focus on specific functional modules, making it difficult to uncover core mechanisms driving HCC progression. To address these issues, recent studies have shifted towards two main strategies: focusing on specific gene subgroups [such as HBV-related genes (28), T-cell exhaustion-related genes (29), and chromosomal instability-related genes (30)] and targeting key structural domains [such as the Tudor domain (6) and coiled-coil domain (7)].

In this study, we plan to develop a prognostic model centered on the PDZ domain. By leveraging bioinformatics methods and large-scale data from public databases, we aim to identify key PDZ proteins in HCC and validate a multivariable prediction model. Additionally, we plan to explore the correlation between these proteins and immune infiltration, as well as their potential prognostic and therapeutic significance. By focusing on a specific structural domain and integrating multi-omics data, our model is expected to not only improve the accuracy of HCC prognosis prediction but also enhance clinical applicability by identifying biologically meaningful prognostic factors. This approach may provide a more reliable basis for understanding HCC heterogeneity and guiding personalized treatment strategies. We present this article in accordance with the TRIPOD reporting checklist (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2024-1018/rc).

Methods

HCC tissue specimen collection

In total, 27 pairs of HCC and corresponding adjacent normal tissue samples were collected from patients diagnosed with different clinical stages of HCC at Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, China. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. Ethical approval for this study was obtained from the Ethics Committee of Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, China (approval No. TJ-IRB202412023). Informed consent was waived due to the use of archived anonymized samples.

Data acquisition

RNA sequencing (RNA-seq) expression profiles, clinical data, and gene mutation data were obtained from The Cancer Genome Atlas (TCGA, https://www.cancer.gov/tcga). The GSE57957 dataset was obtained from the Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo/), and liver cancer data were sourced from The International Cancer Genome Consortium (ICGC, https://dcc.icgc.org/). Patients with incomplete data were excluded. TCGA and ICGC provided data of 346 and 232 HCC specimens, respectively, whereas the GSE57957 dataset included 39 HCC samples and 39 matched adjacent non-tumor tissues. Detailed clinical characteristics (including age, sex, TNM stage, and survival outcomes) of the TCGA cohort are summarized in https://cdn.amegroups.cn/static/public/jgo-2024-1018-1.xlsx. Clinical data for the ICGC and GSE57957 cohorts were not available.

Identification of differentially expressed PDZ proteins (DEPs) in HCC and normal tissues

In the TCGA cohort, raw read count data of HCC samples were retrieved from the TCGA database. All reads were aligned to the human reference genome using the STAR software (31). Mapped reads were quantified using the featureCounts software (32). The raw counts were then normalized to transcripts per million (TPM) values. Differentially expressed genes (DEGs) between tumor and adjacent normal tissues were identified using the limma R package. In addition, differential expression analysis was also performed in the GSE57957 microarray dataset using the limma R package (33). DEGs were defined based on the following criteria: |log₂ fold change (FC)| >1 and P<0.05. Finally, the intersection of DEGs from the TCGA cohort, DEGs from the GSE57957 dataset, and a curated list of 149 PDZ domain-containing proteins from the literature was taken to identify the DEPs.

Functional enrichment analysis

Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed to explore the functions of DEPs using the clusterProfiler package in R. The results were visualized using the SRPlot platform.

Development and validation of the DEPs-related prognostic model

Survival data with missing values or times <30 days were excluded. TCGA and ICGC datasets were used for training and validation, respectively. Cox regression analyses (univariate and multivariate) were conducted to examine the correlation between DEP expression in HCC tissues and overall survival (OS) in the training cohort, with statistical significance set at P<0.05. Subsequently, a random survival forest (RSF) model was constructed for model reduction using the rfsrc function in the randomForestSRC R package (v2.9.3), as random forest offers a robust measure of variable importance. Nine genes with a relative importance greater than 0.2 were selected for further analysis. Model prediction was based on a linear combination of the expressions of the nine DEPs, with weights determined based on their relative coefficients in multivariate Cox regression. The risk score was calculated as: Risk score = (β1 × DEP1 expression) + (β2 × DEP2 expression) + … + (βn × DEPn expression). The Survminer R package was used to determine the risk score threshold and to divide the patients into high- and low-risk groups. Survival curves were plotted for each set (training set, validation set, and multiple clinical subsets of the training set) based on grouping using the survival package in R. These curves were used to assess the prognostic differences between the two groups of patients with HCC. Finally, the discriminatory ability of the model was validated using receiver operating characteristic (ROC) curves.

Independence of the prognostic model from clinical indicators

Cox regression analyses (univariate and multivariate) were conducted to evaluate whether the predictive ability of the prognostic model was independent of other clinical factors (such as age, sex, grade, and stage) using the risk score and other clinical variables as independent factors and OS as the dependent variable.

Optimization and visualization of the prognostic model

Multivariate Cox regression analysis was performed to evaluate survival at 1, 3, and 5 years in the training cohort by incorporating the aforementioned clinical indicators into the new prognostic model. The findings were visualized using a nomogram. ROC curves were used to assess the discriminatory power of the model, and calibration curves were used to evaluate its accuracy. Decision curve analysis (DCA) was performed to assess the clinical applicability of the model at 1, 3, and 5 years.

Immune infiltration analysis

Single-sample gene set enrichment analysis was performed to quantify immune cell infiltration in tumors using the gene set variation analysis R package. Subsequently, differences in immune cell infiltration between high- and low-risk groups were evaluated. Additionally, eight immune evaluation algorithms—CIBERSORT, EPIC, MCPcounter, TIMER, XCELL, ESTIMATE, quanTIseq, and IPS—were used to analyze the immune infiltration characteristics of the two groups.

Immune therapy analysis

We directly compared the compositions of immune responders and non-responders between the high- and low-risk groups. Additionally, tumor immune dysfunction and exclusion (TIDE) scores and exclusion and dysfunction scores were determined using the TIDE algorithm, and differences in immunotherapy responses between the high-risk and low-risk groups were analyzed.

Analysis of SNPs and copy number variations (CNVs)

Information on SNPs and CNVs was sourced from TCGA, and frequently mutated genes in the high- and low-risk groups based on the SNP data were identified using the Maftools package. CNV segment analysis was performed using Genomic Identification of Significant Targets in Cancer 2.0 on GenePattern, with default parameters.

Expression and survival analysis of DEPs

The Wilcoxon rank-sum test was used to assess the differential expression of DEPs between patients with HCC and healthy controls. Patients were classified into high- and low-risk groups based on a risk score threshold using the Survminer R package. Survival curves were generated for each group using the survival package in R.

RNA extraction and quantitative polymerase chain reaction (qPCR)

Total RNA was extracted from human adjacent tumor tissues and HCC samples using the TRIzol reagent (Thermo Fisher Scientific, Waltham, MA, USA). Reverse transcription was performed using the PrimeScript™ RT Reagent Kit (Takara, Japan). qPCR was performed using the SYBR qPCR Master Mix (Takara, Japan) to determine the expressions of DEPs. GAPDH served as the internal control for normalization. The relative expressions of DEPs were determined using the 2^−ΔΔCT method. qPCR was performed using HCC and adjacent tissue samples from 27 patients, and data were analyzed using a paired t-test. The primer sequences used are listed in Table S1.

Statistical analysis

R software (version 4.3.0) and GraphPad Prism 9 (GraphPad Software, San Diego, CA, USA) were used for data analysis and visualization. The log-rank test was used for plotting survival curves, and the Wilcoxon rank-sum test was used for differential expression analysis. Furthermore, t-test and Chi-squared test were used for continuous and categorical variables, respectively. Batch effects were corrected using the “ComBat” function from the “sva” R package (34). Statistical significance was set at two-tailed P value <0.05, unless otherwise specified.

Results

DEPs between HCC and normal tissues

A total of 13,647 and 9,247 DEGs were identified between HCC and normal tissues in TCGA and GSE57957 datasets, respectively, and the intersection of DEGs and PDZ proteins yielded 36 DEPs (Figure 1).

Figure 1 Venn diagram of DEPs. The green and purple circles represent DEGs from the GSE57957 and TCGA datasets, respectively. The yellow circle indicates PDZ proteins. The overlapping region of all three circles represents the final set of DEPs. DEG, differentially expressed gene; DEP, differentially expressed PDZ protein; TCGA, The Cancer Genome Atlas.

GO and KEGG analyses of DEPs

GO analysis of the DEPs revealed that they were primarily enriched in the establishment and maintenance of cell polarity, including apical/basal polarity, bipolar cell polarity, and cell polarity maintenance in biological processes; cell-cell junctions, adherens junctions, the apical part of the cell, postsynaptic density, and asymmetric synapses in cellular components; and PDZ domain binding, actin binding, protein-macromolecule adaptor activity, β-catenin binding, and frizzled binding in molecular functions (Figure 2A). KEGG pathway analysis revealed enrichment of DEPs in the Hippo signaling pathway, tight junctions, human papillomavirus infection, and the cytoskeleton in muscle cells (Figure 2B).

Figure 2 GO and KEGG analyses. (A) GO analysis of DEPs. The outer circle displays GO terms identified by unique identifiers, with color-coded blocks representing each category: biological processes (blue), cellular components (yellow), and molecular functions (green). Bar height represents the number of genes associated with each GO term, and color intensity indicates enrichment significance based on −log₁₀(P value), with darker colors representing higher statistical significance. The inner circles represent the rich factor (0–1) and the number of selected genes within each GO term, providing insight into the enrichment strength and gene distribution. (B) KEGG analysis of DEPs. The size of the bubbles in the right panel corresponds to the gene count, and a more intense red color indicates a smaller P value. DEP, differentially expressed PDZ protein; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes.

Key DEPs identification and prognostic model construction

Univariate Cox regression analysis of DEPs in the training set revealed 15 DEPs significantly associated with the OS of patients with HCC (Table 1). Subsequently, these 15 DEPs were used to construct a RSF model in the training set (Figure 3A). The error rate of the model was stabilized via parameter tuning when the number of trees reached 200. The importance of the DEPs was ranked using the VIMP method in the following order: HTRA2, LIN7B, MPP6, SLC9A3R1, RHPN1, PARD3, GORASP1, DLG5, and SNX27 (Figure 3B). The predictive model was characterized as follows:

Figure 3 Analysis of DEPs using RSF modeling. (A) Error rate of the random survival forest. (B) DEPs ranked by importance. Nine DEPs with a relative importance score of >0.2 were selected using the RSF algorithm. DEP, differentially expressed PDZ protein; RSF, random survival forest.

Risk score = (−0.101 × SNX27 expression) + (−0.068 × DLG5 expression) + (0.093 × GORASP1 expression) + (0.162 × PARD3 expression) + (0.104 × RHPN1 expression) + (0.074 × SLC9A3R1 expression) + (0.162 × MPP6 expression) + (0.174 × LIN7B expression) + (0.166 × HTRA2 expression).

Model validation

Predictive performance of the model evaluated through ROC curves revealed an area under the curve (AUC) value of >0.65 for 1-, 3-, and 5-year survival in the training set (Figure 4A) and an AUC value of >0.7 at 1 year and approximately 0.7 at 3 years in the test set (Figure 4B), demonstrating the strong predictive ability. Survival curves (Figure 4C-4L) generated based on the risk classification for the two groups in each dataset demonstrated significantly worse prognosis for the high-risk group than the low-risk group (P<0.05) in each set, except for certain clinical subsets of the training set (including females, stage III–IV, and G3–4), indicating that the nine DEPs in the model are strong predictors of patient prognosis.

Figure 4 ROC curve and survival curve of the model. ROC curves of the model in the training (A) and validation (B) sets, with red, blue, and orange representing 1-, 3-, and 5-year survival, respectively. Risk scores were used to categorize the population, and Kaplan-Meier survival curves were plotted for the training set (C), validation set (D), age <65 years (E), age ≥65 years (F), female sex (G), male sex (H), stage I–II (I), stage III–IV (J), G1–2 (K), and G3–4 (L). Blue and red represent the low- and high-risk groups, respectively. AUC, area under the curve; CI, confidence interval; ROC, receiver operating characteristic.

Independence and optimization of the model

Univariate analysis indicated that both the stage and risk scores exhibited certain prognostic value (Figure 5A), and both were identified as independent prognostic predictors in a multivariate Cox regression model (Figure 5B). These factors were reintroduced into the Cox model for further fitting, and the results were visualized using a nomogram (Figure 6). Furthermore, calibration curves demonstrated good performance of the model at 1, 3, and 5 years (Figure 7A-7C). Additionally, DCA revealed a high clinical benefit across all datasets (Figure 7D-7F), which may aid in patient consultation, decision-making, and follow-up.

Figure 5 Independent predictive capability of the prognostic model. Univariate (A) and multivariate (B) regression analyses of the prognostic model and clinical indicators with overall survival. Statistical significance was set at P<0.05. CI, confidence interval.

Figure 6 Nomogram for predicting 1-, 3-, and 5-year survival rates. The nomogram was constructed by summing the points for each variable on the point scale. The total score projected on the bottom scale represents the probability of 1-, 3-, and 5-year OS. ***, P<0.001. OS, overall survival.

Figure 7 Calibration and DCA curves of the model. Calibration curves for 1-year (A), 3-year (B), and 5-year (C) survival. The horizontal and vertical axes represent the predicted and observed survival status, respectively. The red line indicates the nomogram-predicted survival, and the gray line represents the ideal reference line. Calibration was performed using 1,000 bootstrap resamples (B=1,000); n=346, events =123. DCA curves for the model at 1 year (D), 3 years (E), and 5 years (F). DCA, decision curve analysis; OS, overall survival.

Immune infiltration analysis based on risk scores

The relationship between DEP-related risk scores and immune cell infiltration levels as well as the overall immune characteristics of patients with HCC was assessed. Notably, patients in the high- and low-risk groups based on their risk scores exhibited significant differences in immune cell infiltration levels, including activated CD4 T cells, activated CD8 T cells, effector memory CD8 T cells, type 1 T helper cells (TH1 cells), type 2 T helper cells (TH2 cells), CD56bright NK cells, CD56dim NK cells, eosinophils, and neutrophils (P<0.05) (Figure 8). Accumulation of activated CD4 T cells and TH2 cells and low intratumoral accumulation of other cells was observed in the high-risk group. Additionally, all eight algorithms revealed notable differences in immune infiltration among the different risk groups (Figure 9). Collectively, these findings suggest that DEPs play a crucial role in immune cell infiltration within HCC.

Figure 8 Immune cell infiltration in the high- and low-risk groups. Analysis of immune cell infiltration patterns in the high- and low-risk groups. *, P<0.05; ***, P<0.001. MDSC, myeloid-derived suppressor cell.

Figure 9 Heatmap illustrating immune infiltration characteristics. Immune infiltration characteristics were assessed using eight different algorithms. *, P<0.05; **, P<0.01; ***, P<0.001.

Immune therapy analysis

An association between the risk score and the effectiveness of immunotherapy was observed in our study. In particularly, patients with lower risk scores exhibited a higher response rate to immunotherapy (P<0.001; Figure 10A). Immune evasion contributes to resistance to immunotherapy. The TIDE score, which assesses T-cell dysfunction and immune rejection across various tumor types, along with the exclusion and dysfunction scores, were significantly higher in the high-risk group than in the low-risk group (Figure 10B-10D).

Figure 10 DEP-related risk score and response to immunotherapy. (A) Sensitivity analysis of the high- and low-risk groups for immunotherapy. (B-D) TIDE, exclusion, and dysfunction scores for the high- and low-risk groups. ***, P<0.001. DEP, differentially expressed PDZ protein; TIDE, tumor immune dysfunction and exclusion.

Genomic alteration analysis based on the risk scores

The effects of the DEP-related risk scores on genetic variations, including SNPs and CNVs, were evaluated. We identified 8 genes with SNPs and 38 CNVs between the high- and low-risk groups (Figures 11,12). The most commonly altered SNPs in HCC were observed in TP53, TTN, OBSCN, PXDNL, and MUC4. The most frequent CNVs in HCC were 1q21.3 amplification, 17p13.1 deletion, 13q14.2 deletion, 4q21.3 deletion, and 4q24 deletion.

Figure 11 SNP profiles of patients in the high- and low-risk groups. The histogram at the top shows the total number of SNPs across the eight genes for each case, and the histogram on the right indicates the number of samples with SNPs in these eight genes. In the heatmap, blue represents the presence of SNPs and gray represents their absence. SNP, single nucleotide polymorphism.

Figure 12 CNV profiles of genes in patients in the high- and low-risk groups. The histogram at the top shows the total number of CNVs across the 38 genes for each case, and the histogram on the right represents the number of samples with CNVs in these genes. In the heatmap, red indicates copy number amplification, blue indicates copy number deletion, and gray indicates the absence of CNV. Amp, amplification; CNV, copy number variation; Del, deletion.

Validation of DEPs in the prognostic model

TCGA data were used to validate the mRNA expression of the DEPs. All nine DEPs were upregulated in HCC tissues (Figure 13A). Additionally, we validated the mRNA expression of the signature genes in HCC and adjacent tumor tissues using qPCR. Compared with adjacent tumor tissues, the expression of SNX27, DLG5, PARD3, and RHPN1 was significantly upregulated in HCC tissues, whereas no notable differences were found in the expression of GORASP1, SLC9A3R1, MPP6, LIN7B, and HTRA2 (Figure 13B). Notably, in the TCGA database, high expression of DEPs was associated with a poor prognosis (Figure 14), confirming the prognostic value of DEPs.

Figure 13 Validation of key DEPs through TCGA and qPCR. (A) HTRA2, LIN7B, MPP6, SLC9A3R1, RHPN1, PARD3, GORASP1, DLG5, and SNX27 expression in HCC and normal tissues from TCGA. (B) Expression of these nine DEPs between HCC tissues and adjacent tumor tissues validated using qPCR. *, P<0.05; **, P<0.01; ***, P<0.001; ns, not significant. DEP, differentially expressed PDZ protein; HCC, hepatocellular carcinoma; qPCR, quantitative polymerase chain reaction; TCGA, The Cancer Genome Atlas.

Figure 14 Survival curves for the nine DEPs in TCGA. Blue and red represent the low- and high-expression groups, respectively. CI, confidence interval; DEP, differentially expressed PDZ protein; TCGA, The Cancer Genome Atlas.

Discussion

The PDZ domain—a highly conserved PPI module—has been implicated in the onset and progression of various cancers, including breast cancer, cervical cancer, colon cancer, prostate cancer, liver cancer, and glioblastoma (35). Multiple PDZ inhibitors have also been explored for cancer treatment (36). We identified nine key DEPs (HTRA2, LIN7B, MPP6, SLC9A3R1, RHPN1, PARD3, GORASP1, DLG5, and SNX27) between HCC and normal tissues and constructed a prognostic model for HCC, with the risk score for a comprehensive characterization of these nine genes. The model demonstrated good predictive performance using the risk score alone or in combination with clinical stage. We also validated the differential expression of these DEPs in liver cancer using TCGA database and qPCR. Notably, only four DEPs were found to be overexpressed in HCC using qPCR, possibly because of the small sample size.

Several studies have confirmed the role of these nine DEPs in HCC. For example, DLG5 exhibits a dual effect in HCC (14), acting both as an inhibitor of tumor invasion and as a potential promoter of tumorigenesis, depending on its functional context. Increased levels of DLG5 in HCC cells have been shown to inhibit the formation of invadopodia, actin-rich membrane structures involved in extracellular matrix degradation and cancer cell invasion. This inhibitory effect occurs through DLG5’s binding with Girdin, preventing its interaction with TKS5.The Girdin-TKS5 interaction is crucial for TKS5 phosphorylation and invadopodia-mediated invasion in HCC cells (15). In contrast, another study indicated that DLG5 overexpression promotes tumorigenesis (16), which may involve the regulation of the Hippo signaling pathway (14). PARD3 promotes tumorigenesis in HCC by activating sonic hedgehog and the Akt pathways (17,18), with its high expression correlating with poor prognosis (37). Mechanistically, PARD3 interacts with aPKC to enhance sonic hedgehog signaling, supporting the self-renewal of CD133⁺ tumor-initiating cells (17). Additionally, DCAF1 binds to PARD3 to activate Akt signaling, promoting tumor growth and metastasis (18). High PARD3 expression is also linked to advanced tumor stages, vascular invasion, and immune evasion, suggesting its potential as a prognostic biomarker and therapeutic target in HCC (37). HTRA2 is an independent prognostic factor for poor survival and is associated with immune cell infiltration in HCC (38). HTRA2 is a potential biomarker for HCC diagnosis (39). SLC9A3R1—a member of the solute carrier protein family and autophagy-related gene—may predict HCC prognosis (40,41). Furthermore, SLC9A3R1 is a candidate gene for the prognostic CNV at 17q25.1-25.3 in HCC (42). Elevated MPP6 expression has been linked to a worse prognosis, angiogenesis, and immune evasion in HCC (43). These findings suggest that DEPs not only serve as prognostic markers but may also contribute to the underlying mechanisms in HCC.

We comprehensively analyzed the correlation between the DEP-related risk scores and immune cell infiltration. The high-risk group exhibited high intratumoral accumulation of activated CD4⁺ T and TH2 cells, along with low intratumoral accumulation of activated CD8⁺ T cells, effector memory CD8⁺ T cells, TH1 cells, CD56bright NK cells, CD56dim NK cells, eosinophils, and neutrophils. CD8⁺ T cells, key effectors of acquired cellular immunity, have been the focus of clinical cancer therapies for over two decades (44-46). Activated CD8⁺ T cells are a uniform population of cytotoxic cells that secrete interferon-γ (IFN-γ) and effectively kill tumor cells (47). Cancer is a chronic condition, and a prolonged T cell response is imperative for effective cancer immunity. Unlike activated CD8⁺ T cells, memory CD8⁺ T cells can persist over time, maintaining their functionality across host tissues and tumors. Memory CD8⁺ T cells consist of circulating subsets, including central memory T cells, effector memory T cells (TEMs), stem cell memory T cells, and non-circulating resident memory T cells. In particularly, TEMs play a crucial role in anti-tumor immunity and are associated with improved prognosis in patients with cancer (46). In tumor immunity, CD4⁺ T cells primarily regulate CD8⁺ T cell function, optimize the amplitude and quality of cytotoxic T lymphocyte responses (as T helper cells or TH), and potentially exhibit direct cytotoxic functions (as cytotoxic CD4⁺ T cells) (48,49). CD4⁺ TH cells differentiate into several effector subtypes, including TH1, TH2, TH17, TH9, Treg, and Tfh, each characterized by distinct cytokine profiles and immune functions (50). TH1 cells stimulate anti-tumor immune responses. Specifically, activated TH1 cells induce apoptosis, senescence, and cell cycle arrest in tumor cells by secreting IFN-γ and tumor necrosis factor alpha. Additionally, TH1 cells amplify the anti-tumor immune response by activating NK and B cells. TH2 cells induce an immunosuppressive pro-tumorigenic response. TH2 cells facilitate tumor growth and metastasis by secreting interleukin (IL)-4, -5, and -13 (50). Innate immunity also plays a role in tumor immunity, and NK cell infiltration is associated with a better prognosis. NK cells can eliminate tumor cells directly by releasing cytotoxic granules or engaging death receptors and boost the anti-tumor immune response by secreting FLT3L, CCL5, and X-C motif chemokines (XCL1/2) (51). Furthermore, tumor-associated tissue eosinophilia is linked to an improved prognosis, with its positive effects in various tumors being independent of traditional prognostic factors such as stage, age, sex, and other clinical variables (52). Neutrophils in the tumor microenvironment contribute to cancer cell growth, angiogenesis, and immunosuppression. However, leveraging their adaptability, neutrophils can also exert an anti-tumor effect by blocking transforming growth factor beta signaling (51). Our findings revealed greater infiltration of pro-tumor immune cells and lower anti-tumor immune cell levels in the high-risk group, indicating a poor prognosis, which is consistent with the low response rate to immunotherapy and the high TIDE score in high-risk patients.

We noted that some DEPs have been reported to be closely associated with the immune microenvironment in HCC. For example, high PARD3 expression has been associated with reduced infiltration of dendritic cells and cytotoxic cells, which may contribute to immune evasion (37). Additionally, HTRA2 has been shown to influence immune cell infiltration in HCC by negatively correlating with the infiltration of B cells, CD8⁺ T cells, Th17 cells, and dendritic cells, while positively correlating with TH cells and Th2 cells, thereby potentially promoting immune evasion (38). Furthermore, elevated MPP6 expression has been linked to poor immune surveillance in HCC by inversely correlating with immune cell infiltration, such as CD8⁺ T cells, B cells, dendritic cells, and cytotoxic cells, and by promoting the accumulation of TH cells, Tcm cells, and Th2 cells. MPP6 expression was also positively associated with tumor mutation burden (TMB) and immune checkpoint gene expression, which may facilitate immune escape and contribute to a worse prognosis (43). Moreover, hScrib and hDlg, as typical PDZ domain-containing proteins, are closely related to the establishment of the immunological synapse, T cell activation, reactive oxygen species (ROS) production in macrophages, and cytokine production by dendritic cells (25). Their roles in regulating immune responses suggest that these PDZ proteins might also influence immune evasion mechanisms in HCC, warranting further investigation.

We also assessed the relationship between DEP-related risk scores and genetic variations, and 8 genes with SNPs and 38 CNVs were identified between the high- and low-risk groups (Figures 11,12). The identified genetic variations differed significantly from previously reported SNPs (rs1800562 in HFE, rs17868323 and rs11692021 in UGT1A7, rs2279744 in MDM2, rs1143627 in IL-1B, and rs4880 in MnSOD) (53), as well as CNVs (gains at 1q21-23, 8q13-21, 12q23-24, 17q12-21, 17q25, and 20q11 and deletions at 4q13, 8p12-21, 13q14, and 17p13) in HCC (54). Several high-frequency mutations identified in this study have been previously in reported HCC. TP53 is a crucial tumor suppressor gene involved in regulating the cell cycle and maintaining genomic stability (55). The TP53 locus spans 21,451 kb on 17p13.1 and contains approximately 100 naturally occurring SNPs. Among them, the non-synonymous substitution SNP at codon 72 in exon 4 (rs1042522) has been extensively studied for its impact on p53 tumor-suppressing activity (56). Similarly, high-frequency mutations in TTN (57), OBSCN (58), and MUC4 (59,60) have been consistently observed in HCC across different cohorts. In contrast, our study provides the first report of PXDNL mutations in HCC. Notably, PXDNL mutations were more frequent in the low-risk group, suggesting a potential protective role in HCC. PXDNL encodes a peroxidasin-like protein implicated in extracellular matrix remodeling and oxidative stress regulation (61). Given that oxidative stress plays a pivotal role in tumor progression and immune modulation, mutations in PXDNL may influence tumor microenvironment remodeling and immune surveillance. Additionally, PXDNL has been implicated in urothelial carcinoma of the bladder, where its overexpression promotes tumor cell motility via the Wnt/β-catenin pathway and is associated with poor prognosis (62). However, the effects and underlying mechanisms of PXDNL and its mutations in HCC remain to be explored. The CNVs identified in this study are commonly observed in HCC, exemplified by the amplification of 1q21.3 and the deletion of 17p13.1. Amplification at 1q21.3-44 and loss of heterozygosity at 17p13.1-13.3 are frequently observed in early-stage HCC (63). Amplifications in 1q21.3 have been linked to the upregulation of oncogenes such as RIT1, a Ras-family GTPase implicated in tumor proliferation and survival (64). Meanwhile, deletions in 17p13.1 often target tumor suppressors such as TP53, further contributing to genomic instability and HCC progression (62). Moreover, MYH10, which is involved in actin cytoskeleton remodeling and cell migration, was located in a CNV region identified in this study and may influence tumor invasion and metastasis (65). Our findings highlight the distinct genetic landscapes between high- and low-risk HCC patients. These observations warrant further functional validation to delineate their precise contributions to HCC pathogenesis.

Despite conducting an extensive and systematic analysis of the PDZ proteins, our study has several limitations that warrant consideration. First, our findings were derived from public databases and clinical samples without validation through in vivo or in vitro functional assays, such as knockdown or overexpression experiments, to directly establish the roles of PDZ proteins in HCC progression or immune modulation. Future studies will focus on functional experiments to further elucidate these mechanisms. Second, our analysis was limited to 149 previously reported key PDZ proteins (9), representing only a fraction of the extensive PDZ protein family. Additionally, while our study systematically analyzed PDZ proteins in HCC, our experimental validation was limited to qPCR-based expression analysis of selected DEPs. The variability in DEP expression and immune infiltration across different HCC subtypes or patient populations warrants further investigation to determine their broader applicability and clinical relevance. Third, upstream regulatory mechanisms and downstream target genes of the PDZ proteins, as well as potential synergistic or competitive interactions among different PDZ proteins, were not explored in this study. Future research should aim to clarify these regulatory networks. Finally, although our prognostic model was validated internally using TCGA and ICGC datasets, the lack of external independent cohort validation, coupled with a relatively limited sample size, raises concerns about its generalizability and clinical utility. Further validation using large, multicenter, prospective cohorts is essential to confirm its robustness and applicability in diverse populations.

In conclusion, while our study provides a preliminary exploration of the expression patterns and potential prognostic significance of PDZ proteins in HCC, future studies integrating functional assays, external validation, and comprehensive analyses of regulatory mechanisms are necessary to deepen our understanding and advance clinical translation.

Conclusions

This study involved a comprehensive analysis of PDZ proteins and their roles in HCC prognosis, immune infiltration, and genetic variations. We identified nine key DEPs that could predict patient outcomes in HCC and may serve as potential therapeutic targets and constructed a prognostic model. Our findings indicate that high DEP-related risk scores correlate with increased cancer-promoting immune cell levels and genetic alterations that contribute to poor prognosis. Nevertheless, further in vitro and in vivo studies are essential to confirm the clinical applicability of these markers. Future studies should investigate the complex regulatory mechanisms of PDZ proteins in HCC progression and explore PDZ-targeted therapies to improve HCC treatment outcomes.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2024-1018/rc

Data Sharing Statement: Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2024-1018/dss

Peer Review File: Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2024-1018/prf

Funding: This work was supported by National Natural Science Foundation of China-Youth Science Fund (No. 82300689) and the Key Research and Development Projects of Hubei Province, China (No. SCZ202111).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2024-1018/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, China (approval No. TJ-IRB202412023) and individual consent for this retrospective analysis was waived.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024;74:229-63. [Crossref] [PubMed]
2. Zheng RS, Chen R, Han BF, et al. Cancer incidence and mortality in China, 2022. Zhonghua Zhong Liu Za Zhi 2024;46:221-31. [PubMed]
3. Wang Y, Jin B, Wu X, et al. Exploration of prognostic and treatment markers in hepatocellular carcinoma via GPCR-related genes analysis. Heliyon 2024;10:e29659. [Crossref] [PubMed]
4. Wang H, Cai H, Li L. Comprehensive analysis of m6A reader YTHDF2 prognosis, immune infiltration, and related regulatory networks in hepatocellular carcinoma. Heliyon 2024;10:e23204. [Crossref] [PubMed]
5. Nardella C, Visconti L, Malagrinò F, et al. Targeting PDZ domains as potential treatment for viral infections, neurodegeneration and cancer. Biol Direct 2021;16:15. [Crossref] [PubMed]
6. Jiang Y, Liu L, Shan W, et al. An integrated genomic analysis of Tudor domain-containing proteins identifies PHD finger protein 20-like 1 (PHF20L1) as a candidate oncogene in breast cancer. Mol Oncol 2016;10:292-302. [Crossref] [PubMed]
7. Hong WF, Zhu DX, Chen YJ, et al. Coiled-coil domain-containing 154 promotes colorectal cancer proliferation and metastasis via interacting with minichromosome maintenance complex component 2. Cancer Lett 2023;578:216460. [Crossref] [PubMed]
8. Pelaseyed T, Hansson GC. CFTR anion channel modulates expression of human transmembrane mucin MUC3 through the PDZ protein GOPC. J Cell Sci 2011;124:3074-83. [Crossref] [PubMed]
9. Christensen NR, Čalyševa J, Fernandes EFA, et al. PDZ Domains as Drug Targets. Adv Ther (Weinh) 2019;2:1800143. [Crossref] [PubMed]
10. Stevens AO, He Y. Allosterism in the PDZ Family. Int J Mol Sci 2022;23:1454. [Crossref] [PubMed]
11. Gardiner J, Overall R, Marc J. PDZ domain proteins: 'dark matter' of the plant proteome? Mol Plant 2011;4:933-7. [Crossref] [PubMed]
12. Shibata T, Chuma M, Kokubu A, et al. EBP50, a beta-catenin-associating protein, enhances Wnt signaling and is over-expressed in hepatocellular carcinoma. Hepatology 2003;38:178-86. [Crossref] [PubMed]
13. Zhang G, Liu T, Wang Z. Downregulation of MAGI1 associates with poor prognosis of hepatocellular carcinoma. J Invest Surg 2012;25:93-9. [Crossref] [PubMed]
14. Andolfi C, Tiribelli C, Pascut D. Recent Hints on the Dual Role of Discs Large MAGUK Scaffold Protein 5 in Cancers and in Hepatocellular Carcinoma. Front Biosci (Landmark Ed) 2022;27:164. [Crossref] [PubMed]
15. Ke Y, Bao T, Zhou Q, et al. Discs large homolog 5 decreases formation and function of invadopodia in human hepatocellular carcinoma via Girdin and Tks5. Int J Cancer 2017;141:364-76. [Crossref] [PubMed]
16. Sarathi A, Palaniappan A. Novel significant stage-specific differentially expressed genes in hepatocellular carcinoma. BMC Cancer 2019;19:663. [Crossref] [PubMed]
17. Wu J, Tan HY, Chan YT, et al. PARD3 drives tumorigenesis through activating Sonic Hedgehog signalling in tumour-initiating cells in liver cancer. J Exp Clin Cancer Res 2024;43:42. [Crossref] [PubMed]
18. Zhang J, Shi Y, Ding K, et al. DCAF1 interacts with PARD3 to promote hepatocellular carcinoma progression and metastasis by activating the Akt signaling pathway. J Exp Clin Cancer Res 2024;43:136. [Crossref] [PubMed]
19. Zhang YZ, Ma Y, Ma E, et al. Sophisticated roles of tumor microenvironment in resistance to immune checkpoint blockade therapy in hepatocellular carcinoma. Cancer Drug Resist 2025;8:10. [Crossref] [PubMed]
20. Donne R, Lujambio A. The liver cancer immune microenvironment: Therapeutic implications for hepatocellular carcinoma. Hepatology 2023;77:1773-96. [Crossref] [PubMed]
21. Ruf B, Heinrich B, Greten TF. Immunobiology and immunotherapy of HCC: spotlight on innate and innate-like immune cells. Cell Mol Immunol 2021;18:112-27. [Crossref] [PubMed]
22. Gao Y, Zhang Z, Huang X, et al. HBV-associated hepatocellular carcinomas inhibit antitumor CD8(+) T cell via the long noncoding RNA HDAC2-AS2. Nat Commun 2025;16:2055. [Crossref] [PubMed]
23. Wei H, Suo C, Gu X, et al. AKR1D1 suppresses liver cancer progression by promoting bile acid metabolism-mediated NK cell cytotoxicity. Cell Metab 2025;37:1103-1118.e7. [Crossref] [PubMed]
24. Chockley PJ, Ibanez-Vega J, Krenciute G, et al. Synapse-tuned CARs enhance immune cell anti-tumor activity. Nat Biotechnol 2023;41:1434-45. [Crossref] [PubMed]
25. Barreda D, Gutiérrez-González LH, Martínez-Cordero E, et al. The Scribble Complex PDZ Proteins in Immune Cell Polarities. J Immunol Res 2020;2020:5649790. [Crossref] [PubMed]
26. Fong ZV, Tanabe KK. The clinical management of hepatocellular carcinoma in the United States, Europe, and Asia: a comprehensive and evidence-based comparison and review. Cancer 2014;120:2824-38. [Crossref] [PubMed]
27. Wang X, Zhao M, Zhang C, et al. Establishment and Clinical Application of the Nomogram Related to Risk or Prognosis of Hepatocellular Carcinoma: A Review. J Hepatocell Carcinoma 2023;10:1389-98. [Crossref] [PubMed]
28. Shen J, Shu M, Xie S, et al. A Six-Gene Prognostic Risk Prediction Model In Hepatitis B Virus-Associated Hepatocellular Carcinoma. Clin Invest Med 2021;44:E32-44. [Crossref] [PubMed]
29. Zhou Y, Wu W, Cai W, et al. Prognostic prediction using a gene signature developed based on exhausted T cells for liver cancer patients. Heliyon 2024;10:e28156. [Crossref] [PubMed]
30. Hu Y, Tang C, Zhu W, et al. Identification of chromosomal instability-associated genes as hepatocellular carcinoma progression-related biomarkers to guide clinical diagnosis, prognosis and therapy. Comput Biol Med 2022;148:105896. [Crossref] [PubMed]
31. Dobin A, Davis CA, Schlesinger F, et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 2013;29:15-21. [Crossref] [PubMed]
32. Liao Y, Smyth GK, Shi W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics 2014;30:923-30. [Crossref] [PubMed]
33. Ritchie ME, Phipson B, Wu D, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 2015;43:e47. [Crossref] [PubMed]
34. Leek JT, Johnson WE, Parker HS, et al. The sva package for removing batch effects and other unwanted variation in high-throughput experiments. Bioinformatics 2012;28:882-3. [Crossref] [PubMed]
35. Subbaiah VK, Kranjec C, Thomas M, et al. PDZ domains: the building blocks regulating tumorigenesis. Biochem J 2011;439:195-205. [Crossref] [PubMed]
36. Liu X, Fuentes EJ. Emerging Themes in PDZ Domain Signaling: Structure, Function, and Inhibition. Int Rev Cell Mol Biol 2019;343:129-218. [Crossref] [PubMed]
37. Li S, Huang J, Yang F, et al. High expression of PARD3 predicts poor prognosis in hepatocellular carcinoma. Sci Rep 2021;11:11078. [Crossref] [PubMed]
38. Feng L, Li Z, Xiong Y, et al. HtrA2 Independently Predicts Poor Prognosis and Correlates with Immune Cell Infiltration in Hepatocellular Carcinoma. J Oncol 2023;2023:4067418. [Crossref] [PubMed]
39. Wang H, Jiang F, Hao F, et al. The expression of HtrA2 and its diagnostic value in patients with hepatocellular carcinoma. Medicine (Baltimore) 2018;97:e0128. [Crossref] [PubMed]
40. Xiong L, Luo Y, Yuan T, et al. Prognostic 7-SLC-Gene Signature Identified via Weighted Gene Co-Expression Network Analysis for Patients with Hepatocellular Carcinoma. J Oncol 2023;2023:4364654. [Crossref] [PubMed]
41. Chen W, Hu MJ, Zhong XL, et al. Screening of a novel autophagy-related prognostic signature and therapeutic targets in hepatocellular carcinoma. J Gastrointest Oncol 2021;12:2985-98. [Crossref] [PubMed]
42. Zhang J, Wen B, Cong W, et al. Association of chromosome 17q copy number variation with overall survival of patients with hepatocellular carcinoma and screening of potential target genes. Zhonghua Yi Xue Yi Chuan Xue Za Zhi 2015;32:615-9. [PubMed]
43. Cheng Q, Wang W, Liu J, et al. Elevated MPP6 expression correlates with an unfavorable prognosis, angiogenesis and immune evasion in hepatocellular carcinoma. Front Immunol 2023;14:1173848. [Crossref] [PubMed]
44. Khan MN, Mao B, Hu J, et al. Tumor-associated macrophages and CD8+ T cells: dual players in the pathogenesis of HBV-related HCC. Front Immunol 2024;15:1472430. [Crossref] [PubMed]
45. Chen Y, Yu D, Qian H, et al. CD8(+) T cell-based cancer immunotherapy. J Transl Med 2024;22:394. [Crossref] [PubMed]
46. Han J, Khatwani N, Searles TG, et al. Memory CD8(+) T cell responses to cancer. Semin Immunol 2020;49:101435. [Crossref] [PubMed]
47. St Paul M, Ohashi PS. The Roles of CD8(+) T Cell Subsets in Antitumor Immunity. Trends Cell Biol 2020;30:695-704. [Crossref] [PubMed]
48. Borst J, Ahrends T, Bąbała N, et al. CD4(+) T cell help in cancer immunology and immunotherapy. Nat Rev Immunol 2018;18:635-47. [Crossref] [PubMed]
49. Miggelbrink AM, Jackson JD, Lorrey SJ, et al. CD4 T-Cell Exhaustion: Does It Exist and What Are Its Roles in Cancer? Clin Cancer Res 2021;27:5742-52. [Crossref] [PubMed]
50. Basu A, Ramamoorthi G, Albert G, et al. Differentiation and Regulation of T(H) Cells: A Balancing Act for Cancer Immunotherapy. Front Immunol 2021;12:669474. [Crossref] [PubMed]
51. Maiorino L, Daßler-Plenker J, Sun L, et al. Innate Immunity and Cancer Pathophysiology. Annu Rev Pathol 2022;17:425-57. [Crossref] [PubMed]
52. Davis BP, Rothenberg ME. Eosinophils and cancer. Cancer Immunol Res 2014;2:1-8. [Crossref] [PubMed]
53. Jin F, Xiong WJ, Jing JC, et al. Evaluation of the association studies of single nucleotide polymorphisms and hepatocellular carcinoma: a systematic review. J Cancer Res Clin Oncol 2011;137:1095-104. [Crossref] [PubMed]
54. Midorikawa Y, Tang W, Sugiyama Y. High-resolution mapping of copy number aberrations and identification of target genes in hepatocellular carcinoma. Biosci Trends 2007;1:26-32. [PubMed]
55. Peng Q, Lao X, Chen Z, et al. TP53 and MDM2 gene polymorphisms, gene-gene interaction, and hepatocellular carcinoma risk: evidence from an updated meta-analysis. PLoS One 2013;8:e82773. [Crossref] [PubMed]
56. Ortiz-Cuaran S, Cox D, Villar S, et al. Association between TP53 R249S mutation and polymorphisms in TP53 intron 1 in hepatocellular carcinoma. Genes Chromosomes Cancer 2013;52:912-9. [Crossref] [PubMed]
57. Zheng J, Chen J, Wang S, et al. Genomic and immune landscape in hepatocellular carcinoma: Implications for personalized therapeutics. Environ Toxicol 2024;39:1601-16. [Crossref] [PubMed]
58. Kassem PH, Montasser IF, Mahmoud RM, et al. Genomic landscape of hepatocellular carcinoma in Egyptian patients by whole exome sequencing. BMC Med Genomics 2024;17:202. [Crossref] [PubMed]
59. Pulakuntla S, Syed K, Reddy VD. Analysis of Somatic Mutations in the TCGA-LIHC Whole Exome Sequence to Identify the Neoantigen for Immunotherapy in Hepatocellular Carcinoma. Curr Issues Mol Biol 2023;46:106-20. [Crossref] [PubMed]
60. Ke L, Shen J, Feng J, et al. Somatic Mutation Profiles Revealed by Next Generation Sequencing (NGS) in 39 Chinese Hepatocellular Carcinoma Patients. Front Mol Biosci 2021;8:800679. [Crossref] [PubMed]
61. Péterfi Z, Tóth ZE, Kovács HA, et al. Peroxidasin-like protein: a novel peroxidase homologue in the human heart. Cardiovasc Res 2014;101:393-9. [Crossref] [PubMed]
62. Lu M, Liu B, Li D, et al. PXDNL activates the motility of urothelial bladder carcinoma cells through the Wnt/β-catenin pathway and has a prognostic value. Life Sci 2023;312:121270. [Crossref] [PubMed]
63. Midorikawa Y, Yamamoto S, Tsuji S, et al. Allelic imbalances and homozygous deletion on 8p23.2 for stepwise progression of hepatocarcinogenesis. Hepatology 2009;49:513-22. [Crossref] [PubMed]
64. Li JT, Liu W, Kuang ZH, et al. Amplification of RIT1 in hepatocellular carcinoma and its clinical significance. Ai Zheng 2003;22:695-9. [PubMed]
65. Jin Q, Cheng M, Xia X, et al. Down-regulation of MYH10 driven by chromosome 17p13.1 deletion promotes hepatocellular carcinoma metastasis through activation of the EGFR pathway. J Cell Mol Med 2021;25:11142-56. [Crossref] [PubMed]