Establishment and validation of a prognostic model based on liquid-liquid phase separation-related genes in gastric cancer

Introduction

Gastric cancer (GC) has developed into a global disease, ranking fifth among the most common cancers in the world and fifth among cancer-related mortality, with about one million new cases of GC each year (1). East Asia, South America and Eastern Europe are currently the regions with relatively high concentrations of GC incidence and mortality. In China, GC has the second-highest incidence rate among all cancers and the second-highest mortality rate among cancer-related deaths (2). Over the recent years, the incidence of GC has shown a trend towards a younger age group in both high- and low-risk countries, with the risk of GC gradually increasing in people younger than 50 years of age (3). In addition to family aggregation of genetic predisposition and Helicobacter pylori infection, the development of GC is also inextricably linked to individual living habits and personal dietary factors (4,5).

Despite the high incidence of GC, GC is usually detected at an advanced stage with a poor prognosis due to the hidden clinical signs of early GC (6). The standard treatment for advanced GC is chemotherapy, but the median overall survival (OS) with standard chemotherapy is usually only about 8 months (7). Among the causes of poor prognosis in GC is intratumoral and intertumoral heterogeneity. However, histologic classification alone is not sufficient to effectively stratify patients and is not conducive to improving their clinical prognosis (8). In order to realize the precise treatment of GC, it is essential to explore the potential molecular processes involved in the development of GC and to develop novel targeted drugs (9).

Intracellular biochemical processes require precise spatiotemporal regulation to maintain normal homeostatic equilibrium, and disruption of this equilibrium is a key factor in cancer development. Recent research advances have revealed the role of liquid-liquid phase separation (LLPS) in the formation of biomolecular condensates, a key mechanism for the spatiotemporal orchestration of biological activities within cells. These biomolecular condensates have been widely observed to be instrumental in regulating a variety of key pathological processes in cancer cells, and aberrant regulation of LLPS is increasingly recognized as a potential driver of carcinogenesis (10). It is widely recognized that cancer development is closely related to genetic abnormalities. Some scholars have proposed that LLPS may contribute to the mutation of normal cells to form a condenser of oncogenic transcription factors, which may be associated with the occurrence of a variety of malignant tumors. Intrinsically disordered region (IDR) in the NUP98-HOXA9 fusion protein is LLPS-competent and may activate oncogenic gene expression by enhancing the binding of transcription factors to genomic targets or facilitating long-range interactions between enhancers and oncogene promoters (11). Moreover, it has been shown that the compartmentalization and condensation of LLPS enable cancer cells to obtain a large number of super-enhancers (SEs) when driving oncogenes, thereby ensuring robust transcription of oncogenes (12). LLPS also plays an important role in regulating a variety of biological pathways associated with cancer, including DNA damage repair, metabolic reprogramming, and immune responses (10,13).

Previous studies have primarily relied on endoscopic findings to predict GC risk and establish predictive models or have constructed predictive models based on a single gene combined with clinical factors (14,15). Research findings indicated that age and Helicobacter pylori infection are established high-risk factors for GC (16). Based on these findings, this study employed a whole-genome factor to construct a model for predicting survival rates in GC patients.

In this study, we constructed an LLPS-related prognostic model for GC using univariate Cox analysis and the least absolute shrinkage and selection operator (LASSO). Subsequently, we validated the model using a cohort of GC patients from the Gene Expression Omnibus (GEO) database. The value of LLPS-related genes (LLPSGs) in prognosis was explored by analyzing the immune infiltration and functional enrichment of GC patients based on LLPSGs, aiming to provide a new prognostic prediction tool for GC patients. Our study contributes to the understanding of the link between LLPS and GC prognosis, and LLPSGs may be potential therapeutic targets for GC patients. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1060/rc).

Methods

Data acquisition and processing

We obtained 5,854 LLPSGs from Data resource of LLPS (DrLLPS) website (https://llps.biocuckoo.cn/). We extracted The Cancer Genome Atlas of Stomach Adenocarcinoma (TCGA-STAD) dataset from TCGA database, as a training cohort, covering RNA sequencing data from 375 GC samples and 32 normal samples, as well as detailed clinical information from 443 GC patients. Differentially expressed genes (DEGs) between tumor and normal groups were identified in the TCGA-STAD dataset based on P values <0.05 [after false discovery rate (FDR) correction] and absolute log₂-fold changes >1 using the software package “limma”. Resulting sets of DEGs associated with OS of GC patients and overlapping genes in LLPS from the DrLLPS database were obtained. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

LLPS-associated prognostic model construction

First, we performed a univariate Cox regression analysis of overlapping LLPSGs to select genes associated with prognosis. Subsequently, to refine the filtering of LLPSGs with prognostic value, we employed the LASSO Cox regression algorithm in the training set using 10-fold cross-validation and the “glmnet” package of R, which integrates the survival time, survival status, and gene expression data for the Regression analysis. The published risk scores for predicting the prognosis of GC patients were calculated as follows:

Riskscore=∑i=0N(βi×Expi)[1]

where N is the total number of selected prognosis-related LLPSGs, βi is the regression coefficient determined by the LASSO regression model, and Expi is the log2-transformed value of the gene expression level of the corresponding LLPSGs. Based on the median risk score, we categorized GC patients into high-risk score groups and low-risk score groups. To assess the association between the LLPSGs prognostic model and OS, we used Kaplan-Meier (K-M) survival analysis. In addition, we utilized time-dependent subject work characteristics [receiver operating characteristic (ROC)] curves and their corresponding area under the curve (AUC) values to assess the prognostic efficacy of this risk model (17). To further evaluate the performance of the prognostic model, we used two R packages, “survminer” and “timeROC”, to predict the prognosis of GC patients. In addition, we also constructed a nomogram using the “rms” package to predict the survival rate of GC patients at 1-, 3-, and 5-year. The predictive accuracy of the model is evaluated based on the AUC value.

Functional enrichment analysis

With the aim of understanding the underlying molecular mechanisms and biological pathways between high- and low-risk GC patients, we performed Gene Ontology (GO) and Kyoto Genes and Genomes Encyclopedia (KEGG) analyses on the LLPSGs screened by univariate Cox analysis, using R’s “clusterProfiler” package.

Immune infiltration analysis

We used the CIBERSORT algorithm to perform an in-depth analysis of the differences in immune cell infiltration between GC patients in the high-risk scoring group and those in the low-risk scoring group. CIBERSORT is a sophisticated bioinformatics tool that enables the use of gene expression data to quantitatively assess the relative abundance and composition of 22 different immune cell subpopulations in clinical samples. Furthermore, we combined multiple immune-infiltration methods, including xCell and Microenvironment Cell Populations-counter (MCP-counter), to enhance the accuracy and reliability of the results.

Validation of the prognostic model

To validate the accuracy of the LLPSGs prognostic model in GC prognosis prediction, we used the GEO15459 dataset from the GEO database, RRID:SCR_005012. This dataset contains clinical information and gene expression data of 192 GC patients as our independent validation cohort.

GSEA analysis of KEGG enrichment

For gene set enrichment analysis (GSEA), we downloaded the GSEA software version 3.0 from the official GSEA website provided by Broad Institute (http://software.broadinstitute.org/gsea/index.jsp). We divided the samples into high and low expression groups based on the median gene expression level. In addition, we downloaded the c2.cp.kegg.v7.4.symbols.gmt gene set from the Molecular Signatures Database website, which was used to evaluate specific biological pathways and molecular mechanisms. In our analysis, we set a P value of less than 0.05 and an FDR of less than 0.25 as the criteria for statistical significance.

Statistical analysis

We analyzed all data using R software (R ×64 version 4.1.2) and Sangerbox 3.0 (18). Multiple groups of data were statistically analyzed by analysis of variance (ANOVA), and comparisons between two groups were made by Student’s t-test or Wilcoxon rank sum test (19). We considered P value <0.05 (two-sided) to be statistically significant (20).

Results

Differential expression of LLPS-associated genes

We incorporated 435 GC patients in the TCGA-STAD cohort in the TCGA database as a training cohort for model construction (Table 1). Based on the GC gene expression profile, 1,410 DEGs were identified, of which 806 were up-regulated and 604 were down-regulated (Figure 1A). We obtained 417 genes by taking the intersection of 1,410 DEGs with 5,854 LLPSGs (Figure 1B). The expression of the top 60 LLPSGs in GC patients and controls is shown in the heatmap (Figure 1C).

Figure 1 Differentially expressed LLPS-associated genes in GC. (A) Volcano mapping of differentially expressed genes in GC in the TCGA-STAD dataset. (B) Intersection of 1,410 GC differentially expressed genes and 5,854 LLPS-associated genes. (C) Heatmap of the top 60 LLPS-associated genes in the GC patients and controls. GC, gastric cancer; LLPS, liquid-liquid phase separation; N, normal; T, tumor; TCGA-STAD, The Cancer Genome Atlas of Stomach Adenocarcinoma.

Construction of a prognostic risk model and functional enrichment analysis for LLPSGs in GC

Using univariate Cox analysis, we identified 48 genes associated with GC prognosis from 417 differential genes associated with LLPS. These 48 genes may become prognostic marker genes for GC (Figure 2A). To reveal the potential biological functions and pathways associated with LLPSGs, we analyzed the 48 LLPSGs differential genes identified by univariate Cox analysis using GO and KEGG enrichment analyses. GO enrichment revealed pathways associated with cell proliferation and DNA transcription, such as maintenance of DNA methylation, cell cycle and transcription regulatory region DNA binding (Figure 2B-2D). KEGG pathway analysis was also enriched for tumor-associated pathways such as Cell cycle, Mismatch repair and cAMP signaling pathway (Figure 2E).

Figure 2 Screening and functional analysis of LLPSGs. (A) Univariate Cox analysis identified LLPSGs associated with the prognosis of GC patients. (B) Top 10 were significantly enriched in BP. (C) Top 10 were significantly enriched in CC. (D) Top 10 were significantly enriched in MF. (E) Top 10 were significantly enriched in KEGG pathway. Gene ratio = count/set size. BP, biological process; CC, cellular component; CI, confidence interval; GC, gastric cancer; GO, Gene Ontology; KEGG, Kyoto Genes and Genomes Encyclopedia; LLPS, liquid-liquid phase separation; LLPSG, LLPS-related gene; MF, molecular function.

We used LASSO-Cox regression analysis to further refine the screening genes. We chose a λ of 0.0786440234805017 to construct a risk prediction model with 6 LLPSGs (Figure 3A,3B). The risk score was calculated as follows: risk score =0.00333634465856454 × APOA4 + 0.0215638208527231 × CDKN2A + 0.000180393652330056 × LYAR − 0.0109702774866363 × PRKAR2B − 0.0390709339615644 × PTPRU + 0.158902735342236 × SOX4. In the characterization of the six LLPSGs, the expression of all six genes in the GC patients was significantly different from that of the normal group (Figure 3C). Survival analysis showed that the expression of all 6 LLPSGs was able to significantly affect the prognosis of GC patients (Figure 3D-3I).

Figure 3 Construction of LLPS-related prognostic models. (A,B) Screening of characterized genes by LASSO regression analysis. (C) Differences in six genes in the TCGA-STAD dataset in GC patients and controls. (D-I) Survival analyses of APOD (D), BRI3BP (E), F5 (F), MYB (G), SNCG (H), VCAN (I). ****, P<0.0001. CI, confidence interval; GC, gastric cancer; H, high; HR, hazard ratio; L, low; LASSO, least absolute shrinkage and selection operator; LLPS, liquid-liquid phase separation; TCGA-STAD, The Cancer Genome Atlas of Stomach Adenocarcinoma.

K-M survival curves indicated a significant difference in survival between the high-risk score group and the low-risk score group (P<0.01), with the high-risk group having a lower survival rate (Figure 4A). Gene expression heatmap showed that APOD, SNCG, VCAN, and F5 were more highly expressed in the high-risk scoring group, whereas BRI3BP and MYB were more highly expressed in the low-risk scoring group. Risk map distributions and survival status of the GC patients showed that high-risk scoring patients had a significantly lower survival time than low-risk scoring patients (Figure 4B). For 1-, 3-, and 5-year survival, the AUC predictive values of the LLPSGs signature were 0.63, 0.63, and 0.70 (Figure 4C). In addition, multivariate Cox analysis showed that age and risk score could be used as independent and efficient indicators to assess the prognosis of GC patients (Figure 4D).

Figure 4 Risk score analysis of the LLPS-related prognostic model in the training cohort. (A) Survival curves for the high-risk and the low-risk score groups in the model. (B) Distribution of risk scores, survival status, and heat maps of the 6 genes. (C) ROC curves predicting survival at 1-, 3-, and 5-year. (D) Multivariate Cox regression analysis combining age, sex, and risk scores in the training cohort. AUC, area under the curve; CI, confidence interval; H, high; HR, hazard ratio; L, low; LLPS, liquid-liquid phase separation; ROC, receiver operating characteristic.

In order to establish a quantitative method for prognostic prediction of GC patients, we constructed a predictive nomogram based on the six LLPSGs (Figure 5A). The expression of each gene in the nomogram has a corresponding score at the top of the nomogram, and the 1-, 3-, and 5-year survival rates of GC patients can be predicted after summing the six scores to calculate the total score (Figure 5B).

Figure 5 Construction of nomogram for predicting survival at 1-, 3-, and 5-year for patients with GC. (A) Nomogram for predicting survival at 1-, 3-, and 5-year for patients with GC. (B) Application of column line plots to prediction. GC, gastric cancer.

Validation of the LLPS-associated prognostic model

A total of 192 GC patients from GEO15459 were included in the validation set cohort (Table 1). In the validation cohort, K-M survival analysis showed that patients in the high-risk group had shorter OS than those in the low-risk group (P<0.001) (Figure 6A). We plotted a heatmap showing the expression of 6 LLPSGs. In the validation cohort, GC patients were also categorized into high-risk score and low-risk score groups. High-risk score patients exhibited lower survival status than low-risk score patients (Figure 6B). The AUC values of ROC curves in the validation cohort were 0.61, 0.64, and 0.66 for 1-, 3-, and 5-year, respectively (Figure 6C). Multifactorial Cox analysis in the validation cohort demonstrated that risk score could still be considered as independent and valid indicators for assessing the prognosis of GC patients (Figure 6D).

Figure 6 Risk score analysis of the LLPS-related prognostic model in the validation cohort. (A) Survival curves for the high-risk score group and the low-risk score group of the model. (B) Distribution of risk scores, survival status, and heat maps for the 6 genes. (C) ROC curves predicting survival at 1-, 3-, and 5-year. (D) Multivariate Cox regression analysis combining age, sex, and risk scores in the validation cohort. AUC, area under the curve; CI, confidence interval; H, high; HR, hazard ratio; L, low; LLPS, liquid-liquid phase separation; ROC, receiver operating characteristic.

Relationship between LLPS-associated prognostic model and immune cell infiltration in GC

We calculated the differences in the levels of 22 immune cell infiltration between the high- and low-risk groups using the CIBERSORT algorithm (Figure 7A). GC patients in the high-risk group had higher levels of naive B cells, monocytes, macrophages M2 and resting mast cells than those in the low-risk group, and GC patients in the high-risk group had lower levels of CD4 memory-activated T cells, follicular helper T cells, resting NK cells and activated mast cells were lower than those in the low-risk scoring group. Subsequently, we explored the relationship between the eight immune cells mentioned above and the six LLPSGs in the prognostic model. GC patients with high APOD expression had higher levels of naive B cells, monocytes, and resting mast cells, as well as lower levels of CD4 memory-activated T cells, resting NK cells, and activated mast cells (Figure 7B). GC patients with high BRI3BP expression had higher levels of T cells follicular helper and lower levels of monocytes, and resting mast cells (Figure 7C). F5 high-expressing GC patients had lower levels of CD4 memory-activated T cells (Figure 7D). MYB high-expressing GC patients had higher levels of CD4 memory-activated T cells, higher levels of follicular helper T cells, and lower levels of naive B cells (Figure 7E). GC patients with high SNCG expression had higher levels of naive B cells, monocytes, resting mast cells, and lower levels of CD4 memory-activated T cells, follicular helper T cells, resting NK cells, and activated mast cells (Figure 7F). GC patients with high VCAN expression had higher levels of naïve B cells, follicular helper T cells, and lower levels of M2 macrophages (Figure 7G). In addition, we performed a correlation analysis between LLPSGs and specific immune cells. Statistically significant correlations were found between all six LLPSGs and specific immune cells (Figure 8). Additionally, both xCell and MCP-counter methods demonstrated significant differences in T cell and macrophage levels between the high- and low-risk groups (Figure S1).

Figure 7 Relationship between LLPS-associated prognostic model and immune cell infiltration in GC. (A) Differences in 22 immune cells between high-risk scoring and low-risk scoring groups. (B-G) Comparison of the infiltration levels of 8 immune cells according to the expression levels of APOD (B), BRI3BP (C), F5 (D), MYB (E), SNCG (F), and VCAN (G). *, P<0.05; **, P<0.01; ***, P<0.001; ****, P<0.0001. GC, gastric cancer; LLPS, liquid-liquid phase separation.

Figure 8 Correlation analysis of LLPSGs and immune cells (A-X). LLPSGs, liquid-liquid phase separation-related gene.

Enriched KEGG pathways in GSEA analysis

We intended to identify the biological pathways enriched in each of the 6 LLPSGs. A total of 187 KEGG pathways were enriched in the TCGA-STAD database, and we performed GSEA analysis on the 6 LLPSGs. By combining the main biological functions of the pathways and their FDR value and P value, we selected 5 KEGG pathways sorted by FDR value. We found that all these 6 LLPSGs could be enriched in tumorigenesis-related pathways, including prostate cancer [Normalized Enrichment Score (NES) =2.4615, P<0.001], pathways in cancer (NES =2.7152, P<0.001), cell cycle (NES =−2.2829, P<0.001), and basic repair (NES =−2.2221, P<0.001). This may be the molecular mechanism by which LLPSGs lead to GC occurrence (Figure 9A-9F).

Figure 9 Single-gene GSEA analysis of six LLPS-associated genes. (A-F) GSEA patterns of five KEGG pathways significantly enriched for APOD (A), BRI3BP (B), F5 (C), MYB (D), SNCG (E), and VCAN (F). Screening criteria for selected KEGG pathways: FDR value <0.25, and P value <0.05. FDR, false discovery rate; GSEA, gene set enrichment analysis; H, high; KEGG, Kyoto Genes and Genomes Encyclopedia; L, low; LLPS, liquid-liquid phase separation; NES, Normalized Enrichment Score.

Discussion

GC is a major challenge to global health, with approximately 70% of patients diagnosed at advanced stages due to low early diagnosis rates, which leads to missing the optimal time for surgical treatment (21). And half of the GC patients who receive adjuvant therapy develop local recurrence and systemic metastases (22). Therefore, the development of a simple and effective diagnostic method is crucial for the early detection and treatment of GC. In recent years, serum tumor markers have received extensive attention in the field of medical research. Carcinoembryonic antigen (CEA), carbohydrate antigen 19-9 (CA19-9), CA125, CA724, alpha-fetoprotein (AFP), CA242 and CA50 are commonly used biomarkers for GC diagnosis. However, the sensitivity and specificity of these biomarkers are poor (23). Consequently, there is an urgent requirement for the identification of new predictive biomarkers to enhance the accuracy of early diagnosis of GC and to improve the prognosis of patients.

In our study, we used TCGA and GEO datasets to build and validate our prognostic models. In addition, we used the Human LLPS Database website to identify LLPSGs. We first identified LLPSGs associated with GC prognosis. Subsequently, we constructed an LLPS-associated prognostic model consisting of 6 genes using LASSO-Cox regression analysis. K-M survival curves and ROC curves indicated that the predictive model had good predictive ability and was validated in the validation dataset. Then, nomogram based on age and the expression of VCAN, APOD, MYB, SNCG, F5, and BRI3BP was constructed to effectively predict the prognosis of GC patients.

VCAN is a member of the aggrecan/versican proteoglycan family. As a chondroitin sulfate proteoglycan, VCAN forms a key component of the extracellular matrix (24). Tumor cells and their stroma play a crucial role in the expression and secretion of VCAN during tumor development (25). Aberrant expression of VCAN proteins has been strongly associated with poor prognosis in a variety of tumors. A wide range of experimental studies, including in vitro and in vivo models, have demonstrated that VCAN is able to regulate a variety of cellular processes involved in tumor phenotypes and characteristics, the generation of drug resistance, and angiogenesis in the tumor stroma (26,27). Notably, high expression of VCAN was found to be positively associated with advanced stage, low differentiation, high metastasis rate, and poor prognosis of GC (28). Some scholars identified VCAN as a potential prognostic marker for GC by bioinformatics analysis, and its expression level was up-regulated in GC patients with poorer prognosis, which is consistent with our findings (29). In the present study, we observed that the expression level of VCAN was reduced in GC patients and was recognized as a risk-promoting factor by our prognostic risk model. Accordingly, upregulation of VCAN may play a key role in the development and progression of GC.

Apolipoprotein D (APOD) is a glycoprotein belonging to the APO family. APOD plays a multifaceted role in lipid transport, food intake, inflammatory response, antioxidant defense, and development, and has been strongly implicated in the development of several cancers (30). Aberrant expression of APOD has been shown to be associated with poor prognosis in a variety of malignant tumors, including breast cancer (31), hepatocellular carcinoma (32), and malignant melanoma (33). Dysregulation of lipid metabolism is a well-recognized feature of cancer. Tumor cells typically rely on lipid metabolism to obtain energy, maintain survival, promote proliferation and metastasis (34). As a member of the APO family, APOD plays a crucial role in lipid metabolism and has an influential role in a broad range of cancer-related biological processes (35). Studies have demonstrated that APOD protein expression in GC samples was significantly higher than that in normal tissues adjacent to the cancer. Moreover, univariate and multivariate Cox analysis indicated that high APOD expression was an independent prognostic risk factor for GC patients. (36). In our LLPS-associated prognostic model, elevated APOD levels were associated with poor prognosis in GC. Our study further confirms the utility of APOD in promoting GC progression, emphasizing its potential as an important molecular target in prognostic assessment and therapeutic intervention in GC.

Activation of the coagulation and fibrinolytic systems is a common phenomenon during the development of malignant tumors. A growing body of research evidence supports that activation of the coagulation system is associated with increased aggressiveness of tumors and increased risk of metastasis, which may ultimately lead to a poor prognosis for patients (37). F5, also known as coagulation factor V, is a high molecular weight (330 kDa) procoagulant factor in the circulation that plays a key role in the cascade of blood coagulation. When it is activated, F5 acts as a cofactor to promote the activation of coagulation factor X, which in turn converts plasminogen to the active prothrombin enzyme (38). Recent studies have shown that multiple factors in the coagulation system may have an impact on the prognosis of cancer patients (39,40). It was noted that the messenger RNA (mRNA) level of F5 was significantly up-regulated in GC tissues, and that the high expression of F5 was closely associated with poor prognosis and advanced disease stage in GC patients (41). These findings echo the results of our study, further confirming that F5 may be a potential prognostic biomarker for GC and providing a new perspective for the diagnosis and treatment of GC.

MYB proteins are multifunctional gene expression regulators consisting of three distinct functional domains responsible for sequence-specific DNA binding, transcriptional activation and negative protein regulation, and are critical regulators of cell growth and differentiation (42). As a proto-oncogene in a diverse range of malignant tumors, elevated expression levels of MYB have also been observed in GC (43). MYB has a dual cell cycle checkpoint function, on the one hand regulating the G2-M transition, on the other hand preventing DNA re-replication and maintaining the diploid state of cells (44). This may be a pivotal mechanism for promoting tumorigenesis after MYB mutation. In addition, MYB is involved in multiple GC-related pathways driving GC proliferation, migration, and invasion, leading to poor prognosis of GC patients (45). In some studies, it has been found that CD36 recognizes Helicobacter pylori, leading to a decrease in the BATF2-MYB protein complex, which in turn inhibits the activation state of CD4⁺ memory T cells (46). Alterations in the immune microenvironment lead to GC occurrence. In our model, high expression of MYB was associated with better prognosis and longer OS in GC patients. Our immune cell infiltration analysis also revealed that high MYB expression was positively correlated with CD4⁺ memory T cell scores. This may be one of the potential mechanisms by which MYB improves patient prognosis.

SNCG, also called γ-synuclein, is a member of the family of synuclein proteins (47). SNCG is strongly associated with neurodegenerative diseases such as Alzheimer’s disease and Parkinson’s disease (48), as well as with the development and progression of a variety of human tumors (49). SNCG expression has been reported to be significantly elevated in breast cancer samples and to promote distant metastasis of breast cancer cells (50). Furthermore, inhibition of SNCG reduced the proliferation and invasiveness of bladder cancer cells (51). SNCG is also considered to have potential as a prognostic biomarker for biliary tract cancer (52). There are also studies confirming the correlation between SNCG and the depth of invasion of cancer cells and lymph node metastasis in GC (53). The expression level of SNCG was found to be lower than that of normal controls in GC, and its expression level was significantly correlated with the prognosis of GC (54). This finding is consistent with the results observed in our study.

BRI3-binding protein (BRI3BP) is an allele of BRI3 whose mRNA is highly expressed in brain, kidney and liver (55). It has been proposed that overexpression of BRI3BP enhances mitochondrial cytochrome c release and caspase-3 activity, thereby promoting apoptosis in animal cells. In addition, this effect is significantly enhanced in the presence of certain anticancer drugs (like etoposide) (56). In our model, we found that high expression of BRI3BP had a favorable impact on the prognosis of GC patients. This phenomenon may be attributed to the fact that high BRI3BP expression promotes the apoptotic process of GC cells, although this needs to be verified by further studies.

The complex interactions between immunotherapy and tumor-infiltrating immune cells (TIIC) have become a fascinating and rapidly advancing field of research. It has been considered that the biological behavior of tumors can be predicted by detecting the presence and distribution of various immune cells (57). In our research, we found a significant difference in immune cell infiltration between the high-risk and low-risk groups. High density of T lymphocytes in tumor tissue correlates with good survival in human GCs (58). Tumor-associated macrophages (TAM) are the most numerous TIICs in the tumor environment. They can be categorized into M0 macrophages, M1 macrophages, and M2 macrophages. M1 macrophages are traditionally considered to have the ability to destroy tumor cells, whereas M2-type macrophages inhibit anti-tumor immune responses by expressing a variety of immunosuppressive factors and chemokines, such as reduced antigen presentation and suppression of T-cell function (59). It has been identified that in the tumor microenvironment of GC, activated mast cells are able to produce and secrete neutrophil chemotactic factor, which promotes neutrophil recruitment to exert anti-tumor effects (60). Based on these findings, we hypothesized that the genes in the LLPS-associated prognostic model may be involved in the development of GC by regulating the function of these immune cells. In the future, it may be possible to effectively improve the prognosis of GC patients by combining targeted therapy with immunotherapy.

Conclusions

In this study, an LLPS-related prognostic model based on age, VCAN, APOD, MYB, SNCG, F5, and BRI3BP was constructed and validated. In addition, it was revealed that these genes probably influence the progression of GC through the regulation of the function of immune cells and are poised to be considered as potential therapeutic targets in the treatment of GC.

However, our study has some limitations. First, we were unable to directly verify experimentally the expression differences of LLPSGs in GC tissues and normal gastric tissues, as well as the distribution of immune cells. Therefore, we need further functional and mechanistic experiments to explore the potential mechanism of carcinogenesis of LLPSGs. Second, we need to follow up in a larger sample size population to verify the accuracy and generalizability of LLPS-related prognostic models. In addition, our study has not yet considered drug use in GC patients, which may have an impact on patient prognosis. In future studies, we will explore and analyze these limitations in greater depth.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1060/rc

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1060/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1060/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.

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. Cao W, Chen HD, Yu YW, et al. Changing profiles of cancer burden worldwide and in China: a secondary analysis of the global cancer statistics 2020. Chin Med J (Engl) 2021;134:783-91. [Crossref] [PubMed]
3. Lu L, Mullins CS, Schafmayer C, et al. A global assessment of recent trends in gastrointestinal cancer and lifestyle-associated risk factors. Cancer Commun (Lond) 2021;41:1137-51. [Crossref] [PubMed]
4. Lazăr DC, Tăban S, Cornianu M, et al. New advances in targeted gastric cancer treatment. World J Gastroenterol 2016;22:6776-99. [Crossref] [PubMed]
5. Li L, Zhu X, Shou T, et al. MicroRNA-28 promotes cell proliferation and invasion in gastric cancer via the PTEN/PI3K/AKT signalling pathway. Mol Med Rep 2018;17:4003-10. [Crossref] [PubMed]
6. Digklia A, Wagner AD. Advanced gastric cancer: Current treatment landscape and future perspectives. World J Gastroenterol 2016;22:2403-14. [Crossref] [PubMed]
7. Arnold M, Abnet CC, Neale RE, et al. Global Burden of 5 Major Types of Gastrointestinal Cancer. Gastroenterology 2020;159:335-349.e15. [Crossref] [PubMed]
8. Körfer J, Lordick F, Hacker UT. Molecular Targets for Gastric Cancer Treatment and Future Perspectives from a Clinical and Translational Point of View. Cancers (Basel) 2021;13:5216. [Crossref] [PubMed]
9. Joshi SS, Badgwell BD. Current treatment and recent progress in gastric cancer. CA Cancer J Clin 2021;71:264-79. [Crossref] [PubMed]
10. Mehta S, Zhang J. Liquid-liquid phase separation drives cellular function and dysfunction in cancer. Nat Rev Cancer 2022;22:239-52. [Crossref] [PubMed]
11. Ahn JH, Davis ES, Daugird TA, et al. Phase separation drives aberrant chromatin looping and cancer development. Nature 2021;595:591-5. [Crossref] [PubMed]
12. Sabari BR, Dall’Agnese A, Boija A, et al. Coactivator condensation at super-enhancers links phase separation and gene control. Science 2018;361:eaar3958. [Crossref] [PubMed]
13. Kilic S, Lezaja A, Gatti M, et al. Phase separation of 53BP1 determines liquid-like behavior of DNA repair compartments. EMBO J 2019;38:e101379. [Crossref] [PubMed]
14. Eom BW, Joo J, Kim S, et al. Prediction Model for Gastric Cancer Incidence in Korean Population. PLoS One 2015;10:e0132613. [Crossref] [PubMed]
15. Lan Y, Sun W, Zhong S, et al. A risk prediction model for gastric cancer based on endoscopic atrophy classification. BMC Cancer 2025;25:518. [Crossref] [PubMed]
16. Gu J, Chen R, Wang SM, et al. Prediction Models for Gastric Cancer Risk in the General Population: A Systematic Review. Cancer Prev Res (Phila) 2022;15:309-18. [Crossref] [PubMed]
17. Hua Y, Liu X, Lv J, et al. Untargeted metabolomics integrated with SHAP analysis identifies novel biomarkers of oxaliplatin induced peripheral neurotoxicity in gastric cancer. Transl Cancer Res 2025;14:4621-37. [Crossref] [PubMed]
18. Shen W, Song Z, Zhong X, et al. Sangerbox: A comprehensive, interaction-friendly clinical bioinformatics analysis platform. Imeta 2022;1:e36. [Crossref] [PubMed]
19. Zhang Z, Zhao X, Chen X, et al. Extracellular volume fraction based on dual-layer spectral computed tomography in assessment of gastric cancer: feasibility analysis of low-dose equilibrium phase scanning. Quant Imaging Med Surg 2025;15:8529-40. [Crossref] [PubMed]
20. Yan B, Zhang P, Zu Q, et al. NKAP overexpression promotes gastric cancer immune escape by inducing IL-10 secretion from mature dendritic cells during anti-PD-L1 therapy. J Gastrointest Oncol 2025;16:1443-60. [Crossref] [PubMed]
21. Song Z, Wu Y, Yang J, et al. Progress in the treatment of advanced gastric cancer. Tumour Biol 2017;39:1010428317714626. [Crossref] [PubMed]
22. Niccolai E, Taddei A, Prisco D, et al. Gastric cancer and the epoch of immunotherapy approaches. World J Gastroenterol 2015;21:5778-93. [Crossref] [PubMed]
23. Pan Y, Zheng Y, Yang J, et al. A new biomarker for the early diagnosis of gastric cancer: gastric juice- and serum-derived SNCG. Future Oncol 2022;18:3179-90. [Crossref] [PubMed]
24. Wight TN. Provisional matrix: A role for versican and hyaluronan. Matrix Biol 2017;60-61:38-56. [Crossref] [PubMed]
25. Calon A, Lonardo E, Berenguer-Llergo A, et al. Stromal gene expression defines poor-prognosis subtypes in colorectal cancer. Nat Genet 2015;47:320-9. [Crossref] [PubMed]
26. Asano K, Nelson CM, Nandadasa S, et al. Stromal Versican Regulates Tumor Growth by Promoting Angiogenesis. Sci Rep 2017;7:17225. [Crossref] [PubMed]
27. Theocharis AD, Karamanos NK. Proteoglycans remodeling in cancer: Underlying molecular mechanisms. Matrix Biol 2019;75-76:220-59. [Crossref] [PubMed]
28. Wang L, Feng L, Liu L, et al. Joint effect of THBS2 and VCAN accelerating the poor prognosis of gastric cancer. Aging (Albany NY) 2023;15:1343-57. [Crossref] [PubMed]
29. Jiang K, Liu H, Xie D, et al. Differentially expressed genes ASPN, COL1A1, FN1, VCAN and MUC5AC are potential prognostic biomarkers for gastric cancer. Oncol Lett 2019;17:3191-202. [Crossref] [PubMed]
30. Rassart E, Desmarais F, Najyb O, et al. Apolipoprotein D. Gene 2020;756:144874. [Crossref] [PubMed]
31. Jankovic-Karasoulos T, Bianco-Miotto T, Butler MS, et al. Elevated levels of tumour apolipoprotein D independently predict poor outcome in breast cancer patients. Histopathology 2020;76:976-87. [Crossref] [PubMed]
32. Utsunomiya T, Ogawa K, Yoshinaga K, et al. Clinicopathologic and prognostic values of apolipoprotein D alterations in hepatocellular carcinoma. Int J Cancer 2005;116:105-9. [Crossref] [PubMed]
33. Miranda E, Vizoso F, Martín A, et al. Apolipoprotein D expression in cutaneous malignant melanoma. J Surg Oncol 2003;83:99-105. [Crossref] [PubMed]
34. Bian X, Liu R, Meng Y, et al. Lipid metabolism and cancer. J Exp Med 2021;218:e20201606. [Crossref] [PubMed]
35. Broadfield LA, Pane AA, Talebi A, et al. Lipid metabolism in cancer: New perspectives and emerging mechanisms. Dev Cell 2021;56:1363-93. [Crossref] [PubMed]
36. Wang Z, Chen H, Sun L, et al. Uncovering the potential of APOD as a biomarker in gastric cancer: A retrospective and multi-center study. Comput Struct Biotechnol J 2024;23:1051-64. [Crossref] [PubMed]
37. Tas F, Ciftci R, Kilic L, et al. Clinical and prognostic significance of coagulation assays in gastric cancer. J Gastrointest Cancer 2013;44:285-92. [Crossref] [PubMed]
38. Cramer TJ, Gale AJ. The anticoagulant function of coagulation factor V. Thromb Haemost 2012;107:15-21. [Crossref] [PubMed]
39. Bazzarelli AK, Scheer AS, Tai LH, et al. Tissue Factor Pathway Inhibitor Gene Polymorphism -33T → C Predicts Improved Disease-Free Survival in Colorectal Cancer. Ann Surg Oncol 2016;23:2274-80. [Crossref] [PubMed]
40. Koizume S, Jin MS, Miyagi E, et al. Activation of cancer cell migration and invasion by ectopic synthesis of coagulation factor VII. Cancer Res 2006;66:9453-60. [Crossref] [PubMed]
41. Liu Y, Liao XW, Qin YZ, et al. Identification of F5 as a Prognostic Biomarker in Patients with Gastric Cancer. Biomed Res Int 2020;2020:9280841. [Crossref] [PubMed]
42. Ness SA. The Myb oncoprotein: regulating a regulator. Biochim Biophys Acta 1996;1288:F123-39. [Crossref] [PubMed]
43. Claerhout S, Lim JY, Choi W, et al. Gene expression signature analysis identifies vorinostat as a candidate therapy for gastric cancer. PLoS One 2011;6:e24662. [Crossref] [PubMed]
44. Katzen AL, Jackson J, Harmon BP, et al. Drosophila myb is required for the G2/M transition and maintenance of diploidy. Genes Dev 1998;12:831-43. [Crossref] [PubMed]
45. Xie Y, Rong L, He M, et al. LncRNA SNHG3 promotes gastric cancer cell proliferation and metastasis by regulating the miR-139-5p/MYB axis. Aging (Albany NY) 2021;13:25138-52. [Crossref] [PubMed]
46. Jiang Q, Chen Z, Meng F, et al. CD36-BATF2\MYB Axis Predicts Anti-PD-1 Immunotherapy Response in Gastric Cancer. Int J Biol Sci 2023;19:4476-92. [Crossref] [PubMed]
47. Liu C, Qu L, Shou C. Role and Characterization of Synuclein-γ Unconventional Protein Secretion in Cancer Cells. Methods Mol Biol 2016;1459:215-27. [Crossref] [PubMed]
48. Surgucheva I, Newell KL, Burns J, et al. New α- and γ-synuclein immunopathological lesions in human brain. Acta Neuropathol Commun 2014;2:132. [Crossref] [PubMed]
49. Strohl A, Mori K, Akers S, et al. Synuclein-γ (SNCG) expression in ovarian cancer is associated with high-risk clinicopathologic disease. J Ovarian Res 2016;9:75. [Crossref] [PubMed]
50. Cirak Y, Furuncuoglu Y, Yapicier O, et al. Predictive and prognostic values of BubR1 and synuclein-gamma expression in breast cancer. Int J Clin Exp Pathol 2015;8:5345-53.
51. Chen Z, Zhang F, Zhang S, et al. The down-regulation of SNCG inhibits the proliferation and invasiveness of human bladder cancer cell line 5637 and suppresses the expression of MMP-2/9. Int J Clin Exp Pathol 2020;13:1873-9.
52. Takemura Y, Ojima H, Oshima G, et al. Gamma-synuclein is a novel prognostic marker that promotes tumor cell migration in biliary tract carcinoma. Cancer Med 2021;10:5599-613. [Crossref] [PubMed]
53. Zheng S, Shi L, Zhang Y, et al. Expression of SNCG, MAP2, SDF-1 and CXCR4 in gastric adenocarcinoma and their clinical significance. Int J Clin Exp Pathol 2014;7:6606-15.
54. Li Y, Pan Q, Cheng M, et al. Identification and validation of anoikis-associated gene SNCG as a prognostic biomarker in gastric cancer. Aging (Albany NY) 2023;15:2541-53. [Crossref] [PubMed]
55. Lin L, Wu Y, Li C, et al. Cloning, tissue expression pattern, and chromosome location of a novel human gene BRI3BP. Biochem Genet 2001;39:369-77. [Crossref] [PubMed]
56. Yamazaki T, Sasaki N, Nishi M, et al. Augmentation of drug-induced cell death by ER protein BRI3BP. Biochem Biophys Res Commun 2007;362:971-5. [Crossref] [PubMed]
57. Chen Y, Jia K, Sun Y, et al. Predicting response to immunotherapy in gastric cancer via multi-dimensional analyses of the tumour immune microenvironment. Nat Commun 2022;13:4851. [Crossref] [PubMed]
58. Oya Y, Hayakawa Y, Koike K. Tumor microenvironment in gastric cancers. Cancer Sci 2020;111:2696-707. [Crossref] [PubMed]
59. Noy R, Pollard JW. Tumor-associated macrophages: from mechanisms to therapy. Immunity 2014;41:49-61. [Crossref] [PubMed]
60. Yang Y, He W, Wang ZR, et al. Immune Cell Landscape in Gastric Cancer. Biomed Res Int 2021;2021:1930706. [Crossref] [PubMed]