Skip to main content
Download PDF

Sialic acid metabolism-based classification reveals novel metabolic subtypes with distinct characteristics of tumor microenvironment and clinical outcomes in gastric cancer

Cancer Cell International volume 25, Article number: 61 (2025) Cite this article

Abstract

Background

High heterogeneity in gastric cancer (GC) remains a challenge for standard treatments and prognosis prediction. Dysregulation of sialic acid metabolism (SiaM) is recognized as a key metabolic hallmark of tumor immune evasion and metastasis. Herein, we aimed to develop a SiaM-based metabolic classification in GC.

Methods

SiaM-related genes were obtained from the MsigDB database. Bulk and single-cell transcriptional data of 956 GC patients were acquired from the GEO, TCGA, and MEDLINE databases. Proteomic profiles of 20 GC samples were derived from our institution. The consensus clustering algorithm was applied to identify SiaM-based clusters. The SiaM-based model was established via LASSO regression and evaluated via Kaplan‒Meier curve and ROC curve analyses. In vitro and in vivo experiments were conducted to explore the function of ST3GAL1 in GC.

Results

Three SiaM clusters presented distinct patterns of clinicopathological features, transcriptomic alterations, and tumor immune microenvironment landscapes in GC. Compared with clusters A and B, cluster C presented elevated SiaM activity, higher metastatic potential, more abundant immunosuppressive features, and a worse prognosis. Based on the differentially expressed genes between these clusters, a risk model for six genes (ARHGAP6, ST3GAL1, ADAM28, C7, PLCL1, and TTC28) was then constructed. The model exhibited robust performance in predicting peritoneal metastasis and prognosis in four independent cohorts. As a hub gene in the model, ST3GAL1 promoted GC cell migration and invasion in vitro and in vivo.

Conclusions

Our study proposed a novel SiaM-based classification that identified three metabolic subtypes with distinct characteristics of tumor microenvironment and clinical outcomes in GC.

Introduction

Gastric cancer (GC) is a prevalent neoplasm with high mortality worldwide, and more than three-fifths of total cases occur in East Asia [1]. Despite the improved efficacy of multimodal treatments including surgical resection, chemotherapy, targeted therapy, and immunotherapy, GC patients with advanced stage still have a high tendency toward local relapse and distant metastasis as well as a low 5-year survival rate [2]. Considerable molecular heterogeneity among individuals remains an enormous challenge to current standard treatments and prognostic prediction, promoting the molecular classification of GC as a hotspot in intense research. The most representative molecular classification, proposed by The Cancer Genome Atlas Consortium (TCGA), categorized GC into four subtypes including microsatellite instability (MSI), Epstein-Barr virus (EBV), chromosomal instability (CIN), and genomically stable (GS) based on integrated genomic features [3]. The Asian Cancer Research Group (ACRG) identified four distinct GC molecular subtypes based on bulk transcription data, including mesenchymal-like, microsatellite-unstable, TP53-active, and TP53-inactive types [4]. These classical molecular classification systems provide a research rationale for understanding the molecular heterogeneity between individuals with GC. Recently, multiple protein posttranslational modifications, such as phosphorylation, succinylation, lactylation, and sialylation, have been linked with malignant transformation and therapeutic responsiveness in GC [5,6,7]. The emergence of these protein posttranslational modifications has increased the potential for developing novel GC molecular classifications to guide prognostic stratification and individualized treatment.

As one of the most critical posttranslational modifications, sialylation involves transferring sialic acid units from the donor (cytidine monophosphate N-acetylneuraminic acid, CMP-Neu5Ac) into the terminal ends of oligosaccharides and glycoproteins and shaping a sialylated glycan coat with negative charges on the cell surface, which plays a crucial role in cell proliferation, adhesion, migration, communication, and extracellular matrix (ECM) interactions [8]. Aberrant accumulation of sialic acid residues on the cell surface, namely hypersialylation, has been recognized as a hallmark across diverse cancer types, such as breast, ovarian, pancreatic, and colon cancers [9,10,11,12]. Mechanistically, the dysregulation of several key enzymes, such as CMP-Neu5Ac synthase, sialyltransferases, and neuraminidases, can promote hypersialylation of membrane glycans and then overload negative charges on the cell surface, eventually leading to cell–cell repulsion and cell detachment from primary tumor lesions [13]. Previous studies have provided in-depth mechanistic insights into the biological role of sialic acid metabolism (SiaM) in tumor progression, metastasis, immune evasion, and chemotherapy resistance [11, 14]. For instance, sialyl Lewis X oligosaccharides can promote tumor cells to attach and invade blood vessels by interacting with selectins on endothelial cells [15]. Hypersialylated glycans expressed on tumor cells can bind to sialic acid immunoglobulin-type lectins (SIGLECs) on tumor-infiltrating immune cells and then induce immunosuppression in the tumor microenvironment (TME), thereby promoting tumor immune evasion [14]. Given the essential role of SiaM in multiple malignant phenotypes, designing SiaM inhibitors that target sialyltransferase (ST) or the sialic acid-SIGLEC axis has emerged as a promising strategy in the field of antitumor drug development. Susceptibility tests on preclinical models preliminarily revealed that sialyltransferase inhibitors (STis) could suppress tumor proliferation and metastasis in multiple cancer types, such as breast, pancreatic, and lung cancers [16,17,18].

Despite these advances in SiaM in cancer research, there are still several key issues to be addressed: (1) the molecular landscape of SiaM with its immunological and clinical implications in GC remains unknown; (2) whether the heterogeneity of clinical outcomes between different GC patients is driven by specific SiaM phenotypes. In this study, based on multicohort clinical data and large-scale multi-omics profiles, we identified three GC subtypes with distinct clinical and molecular characteristics and sought to screen potential therapeutic targets, which could accelerate the development of novel therapeutic strategies for GC.

Materials and methods

Data source and sample collection

Public transcriptional data of 956 GC samples with gene expression profiles and annotated clinicopathological information were obtained from five datasets (GSE66229, GSE15081, GSE183904, TCGA-STAD, and the Kim cohort). Three datasets (GSE66229, n = 400; GSE15081, n = 56; and GSE183904, n = 40) were downloaded from the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/). Bulk RNA sequencing data and clinical information of the TCGA-STAD cohort containing 415 GC samples were acquired from the cBioportal database (https://www.cbioportal.org/). The Kim cohort contained 45 GC patients who received anti-PD-1 agents and provided RNA-seq data with annotated immunotherapeutic information, which was downloaded from the original publication and corresponding supplementary materials [19]. In addition, our previous study provided the proteomic profile of 20 GC samples with annotated clinical data (named the ZJUGC-A cohort) [20]. Additionally, a total of 36 tumor samples from GC patients were retrospectively collected from the Zhejiang University School of Medicine the First Affiliated Hospital for RT‒PCR analysis (named the ZJUGC-B cohort). Informed consent was provided by all the GC patients in this study. Human ethical approval was obtained from the ethical committee of our institution.

Consensus clustering analysis

The “REACTOME_SIALIC_ACID_METABOLISM” gene set, composed of 31 hub genes involved in the metabolic process of sialic acid, was downloaded from the MsigDB database (https://www.gsea-msigdb.org/gsea/msigdb/). After extracting the SiaM gene expression matrix from the ACRG cohort (GSE66229), unsupervised consensus clustering analysis was performed via the “ConsensusClusterPlus” R package to identify SiaM-based metabolic classification subtypes in GC. The “K-Means” clustering algorithm with default parameters was adopted as the clustering method. Based on the cumulative distribution function (CDF) and CDF delta area curves, we chose the optimal value of the k parameter from 2 to 10 for the number of clusters. Principal component analysis (PCA) was performed to evaluate the classification effect of the SiaM clusters.

Pathway enrichment and functional annotation

Differentially expressed genes (DEGs) among the three SiaM clusters were identified via the “limma” R package, with the parameters of P < 0.05 and |Fold change (FC)| > 1.5. Common genes identified from the DEGs between Cluster C and Clusters A/B were subjected to Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses. The cancer hallmark pathways were obtained from the MSigDB database and estimated via single-sample gene set enrichment analysis (ssGSEA). The ssGSEA method, a rank-based algorithm, is specifically devised to compute an overexpression metric for a predefined gene set in comparison to the rest of the genes throughout the entire genomic landscape [21]. It has been widely applied in bioinformatic studies to estimate the relative activation level of a specific signaling or pathway [22, 23]. In this study, the SiaM activity of each sample was also estimated using the ssGSEA method based on the bulk transcriptional expression of 31 SiaM-related genes. The value of the estimated score represents the relative activation degree of SiaM in each sample. Besides, Gene set enrichment analysis (GSEA) was performed via the “clusterProfiler” R package to compare the relative activity of sialic acid binding between the SiaM clusters. The accumulation of sialic acid was estimated using the methods described by Liu et al. [24]. Briefly, it depends on the rates of sialic acid synthesis, degradation, and transfer to the cell surface. The transcriptional expressions of the following genes were used to make the estimation: CMAS for sialic acid synthesis, sialyltransferase genes for sialic acid transfer (ST8SIA1, ST8SIA2, ST8SIA3, ST8SIA4, ST8SIA5, ST3GAL1, ST3GAL2, ST3GAL5, ST6GALNAC3, ST6GALNAC5, and ST6GALNAC6), NEU1 and CTSA for sialic acid degradation. The rate of sialic acid placement onto the cell surface is estimated using the geometric average of the transcriptional expression of CMAS and these sialyltransferase genes, whereas the rate of sialic acid degradation is estimated using the geometric average of the transcriptional expression of NEU1 and CTSA. The delta value between these two processes represents the accumulation of sialic acid in this study. Four types of cell migration-related activity, namely, cell adhesion, cell contraction, protrusion, motion and migration, were estimated via a set of marker genes as described by Sun et al. [25].

Estimation of stromal content and immune infiltration in the tumor microenvironment (TME)

TME scores, including stromal, immune, and stromal-immune comprehensive score (ESTIMATE) scores, were estimated for each GC sample via the “ESTIMATE” R package [26]. The abundance of endothelial cells and cancer-associated fibroblasts (CAFs) was inferred via the “MCPcounter” R package [27]. The relative infiltration levels of 22 types of tumor-infiltrating immune cells were quantified via the “CIBERSORT” algorithm [28]. A computational method named Tumor Immune Dysfunction and Exclusion (TIDE, http://tide.dfci.harvard.edu/) was utilized to evaluate the activation and inhibition levels of tumor-infiltrating T cells and the probability of resistance to immunotherapy in GC samples.

Single-cell transcriptomic analysis

Droplet-based scRNA-seq data (10X Genomics) of 59,072 cells from 40 GC samples were obtained from the GSE183904 dataset [29]. The “Seurat” R package was used to perform scRNA-seq analysis according to a previous study [29]. Based on the scRNA-seq data, SiaM activity was quantified by the AUC value in each cell, which was calculated via the “AUCell” R package [30]. The value of the estimated score represents the relative activation degree of SiaM in each cell. The pseudo-time trajectories of the cells were developed via the “Monocle” R package [31]. Different branches in the pseudo-time trajectory represented different cell subsets with specific differentiation states. Dynamic changes in gene expression in different cell states were visualized via the “plot_genes_branched_pseudo-time” function.

Selection of candidate SiaM-related DEGs and construction of a clinical prediction model

DEGs among three SiaM clusters (named SiaMGs) were screened via the “limma” R package in the ACRG cohort and matched to the proteomic profile of the ZJUGC-A cohort. Candidate SiaMGs were identified by comparing gene expression between the groups with no metastasis and those with peritoneal metastasis. Least absolute shrinkage and selection operator (LASSO) regression analysis was performed via the “glmnet” R package to identify candidate SiaMGs associated with peritoneal metastasis. The SiaM score of each GC patient was determined according to the following formula: SiaM score = Σcoefi * expi (where coefi and expi represent the regression coefficient and relative expression level of candidate SiaMGs, respectively). The prediction performance of the SiaM-based model for peritoneal metastasis was assessed via receiver operating characteristic (ROC) curves and area under the curve (AUC) values. Furthermore, the associations between the SiaM-based model and several biomarkers of targeted and immune checkpoint inhibitor (ICI) therapy were evaluated in the ACRG, TCGA-STAD, and Kim cohorts.

Quantitative real-time PCR

Total RNA was isolated from tumor tissues or cell lines via the AG RNAex Pro Reagent (#AG21101, Accurate Biology, China), and the RNA concentration was quantified by the A260/A280 ratio. Next, 5X Evo M-MLV RT Master Mix (#AG11706, Accurate Biology, China) was applied to generate cDNA. 2X SYBR Green Pro Taq HS Premix (High Rox Plus, #AG11740, Accurate Biology, China) was used to perform polymerase chain reactions on the StepOnePlus PCR System (Applied Biosystems, USA). The relative transcript level of each gene was quantified via the 2−ΔΔCt method, with GAPDH used as an endogenous control. All primers were synthesized by Tsingke Biotechnology (Beijing, China). All sequences of primers used in this study are listed in Table S1.

Cell culture and transfection

Three human GC cell lines (HGC27, MKN45, and AGS) purchased from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China) were cultured in Roswell Park Memorial Institute (RPMI)−1640 medium supplemented with 10% fetal bovine serum (FBS; Gibco, USA) and maintained in a 5% CO2 incubator at 37 °C. For gene silencing, two small interfering (si) RNAs targeting human ST3GAL1 (ST3GAL1-si1 sense sequence: 5′-GCGAUGAGGUGGACUUGUA-3′; ST3GAL1-si2 sense sequence: 5′-GGAGAUAAUGUCAGCAUGA-3′) were transfected into HGC27 and MKN45 cells via Lipofectamine 3000 transfection reagent (Invitrogen, Carlsbad, California, USA) and Opti-MEM medium (Gibco, Carlsbad, California, USA). For gene overexpression, human ST3GAL1 was upregulated in HGC27 and AGS cells via a lentiviral vector (PGMLV-6751, ST3GAL1 [NM_003003.4]) with puromycin resistance according to the manufacturer’s protocol (GenoMeditech, Shanghai, China).

Cell Counting Kit-8 (CCK-8) proliferation assay

Cell viability was assessed via the Cell Counting Kit-8 (CCK-8) assay, following the manufacturer’s protocol (#ab228554, Abcam, UK). Briefly, cells were seeded in 96-well plates at a density of 2 × 103 cells per 100 μL per well. After cell growth at different time points (0 h, 24 h, 48 h, and 72 h), the culture medium was removed, and 100 μL of fresh culture medium containing 10 μL of CCK-8 solution was added to each well, followed by 2 h of incubation. The absorbance at 450 nm was tested via a BioTek Synergy HT Microplate Reader (BioTek, USA).

Wound healing assay

The cells were seeded into 6-well plates at a density of 2 × 105 cells per well until confluency. Subsequently, the confluent monolayers were scratched with a 200 μL pipette tip, followed by washing three times with PBS. Then, the fresh culture medium was added to each well. The degree of wound closure was measured and photographed via a light microscope at different time points (0 h, 12 h, or 24 h).

Cell migration and invasion assays

Cell migration and invasion assays were performed in Transwell chambers (24-well format, 8.0 µm pore size; Millipore, Washington, DC, USA). A total of 1 × 105 cells in serum-free media were seeded in an uncoated upper chamber for migration assays, and 7 × 104 cells in serum-free media were cultured in a Matrigel (BD Biosciences, Lake Franklin, NJ, USA)-coated chamber for cell invasion assays. The lower chambers were filled with 600 μL of culture medium containing 10% FBS. After incubation at 37 °C for 24 h, the cells that had moved from the upper surface of the membrane to the lower side were washed three times with PBS, fixed in 4% paraformaldehyde for 15 min, and stained with 0.1% crystal violet for 30 min. The number of cells in five random fields (200×) under a light microscope was determined via ImageJ software.

In vivo peritoneal metastasis model

Four-week-old female BALB/C nude mice were purchased from the Nanjing Provincial Center for Disease Control and Prevention. The animal care and experimental procedures were reviewed and approved by the Institutional Animal Care and Use Committee at Zhejiang University. The mice were randomly divided into three groups (NC, n = 5; OV, n = 5; OV + STi, n = 5). After adaptive feeding for 1 week, 2 × 106 HGC27 cells stably transfected with the ov-ST3GAL1 vector or ov-NC were injected into the peritoneal cavities of the mice. During the second week, the OV + STi group received daily intraperitoneal injections of 60 mg/kg 3Fax-Peracetyl Neu5Ac (#566224, Merck, Germany), which is a classic STi. The other groups (NC and OV) received intraperitoneal injections of the same amount of PBS at the same time points. After 4 weeks, all the mice were sacrificed, and the metastatic tumors were weighed.

Immunohistochemistry (IHC)

Paraffin-embedded GC sections were deparaffinized in xylene, dehydrated with an ethanol gradient, and blocked with 3% H2O2 for 10 min. A citrate buffer (pH 6) solution was then applied to the slides for 15–20 min at 95 °C to retrieve antigens. The tissue sections were incubated overnight with MAL-II (#B-1265-1, 1:400 dilution, Vector Laboratories, USA), anti-E-cadherin antibody (#K011355P, 1:300 dilution, Solarbio, China), anti-vimentin antibody (#K200054M, 1:500 dilution, Solarbio, China), or anti-Ki67 antibody (#28074-1-AP, 1:200 dilution, Proteintech, USA) at 4 °C, followed by a 30 min incubation with an HRP-conjugated secondary antibody (ZSGB-bio, Beijing, China) at 37 °C. After washing three times with PBS, color was developed via DAB Chromogen (MXB Biotechnologies, Fuzhou, China). The slides were then rinsed in tap water and counterstained with hematoxylin. Five random fields per section were viewed under a light microscope, and a semiquantitative grading system (the H score) was applied to compare the immunohistochemical staining intensities.

Statistical analysis

All the statistical analyses were performed via R (version 4.2.1), GraphPad Prism (version 9.0), and SPSS software (version 26.0). Comparisons of nonparametric variables between two or multiple subgroups were performed via the Mann‒Whitney test or the Kruskal‒Wallis test, respectively. Differences in categorical variables among subgroups were examined via chi-square tests or Fisher’s exact tests under appropriate conditions. Comparisons of OS and DFS between different subgroups were performed via Kaplan‒Meier curves and Cox regression analyses, which are reported as hazard ratios (HRs) with corresponding 95% confidence intervals (CIs). The experimental data are presented as the means ± standard deviations (SDs). Differences with P < 0.05 were considered statistically significant.

Results

Study design

The workflow of this study is shown in Fig. 1. Firstly, we obtained five cohorts with 956 GC patients from the public databases and two cohorts with 56 GC patients from our institution. The ACRG cohort and the other cohorts were used as the training and validation cohorts, respectively. Besides, we obtained a geneset containing 31 SiaM-related genes from the MsigDB database. Secondly, a consensus cluster analysis was performed based on the SiaM-related gene expression to identify different metabolic clusters in the ACRG cohort. The associations of the SiaM-based classification with clinicopathological features and survival outcomes were further investigated. From the molecular perspective, we explored the differences in signaling pathways and TME characteristics among different SiaM clusters. Given the tight relevance between the metabolic classification and GC peritoneal metastasis, we developed and validated a six-gene model for predicting peritoneal metastasis, survival outcomes, and immunotherapy response in six GC cohorts. Considering that ST3GAL1 was one of the hub genes in the model, we explored the biological function of ST3GAL1 involved in GC progression and metastasis in vitro and in vivo.

Fig. 1
figure 1

The workflow of this study. Step 1: A total of five cohorts comprising 956 gastric cancer (GC) patients were retrieved from public databases, while an additional two cohorts with 56 GC patients were obtained from our institution. Step 2: A geneset containing 31 SiaM-related genes was downloaded from the MsigDB database and a consensus cluster analysis based on this geneset was conducted to establish a SiaM-based classification in the ACRG cohort. Then, we investigated the relationship between the SiaM-based classification with clinicopathological characteristics and survival outcomes. Step 3: The DEGs between different SiaM clusters were identified using limma analyses and then analyzed using multiple pathway enrichment analyses (GO, KEGG, GSEA, and ssGSEA). Step 4: TME features were characterized using multiple algorithms (ESTIMATE, MCPcounter, CIBERSORT, TIDE) and compared in different SiaM clusters. Step 5: A six-gene signature was established using LASSO regression to predict peritoneal metastasis of GC. The predictive and prognostic value of the model was evaluated in multiple independent cohorts, including ACRG, TCGA-STAD, Takeno, ZJUGC-A, and ZJUGC-B. The association between the model and immunotherapy biomarkers was investigated in the Kim cohort. Step 6: The biological function of ST3GAL1 in GC cell proliferation, migration, and invasion was investigated in vitro. Step 7: The biological function of ST3GAL1 in peritoneal metastasis of GC was explored in vivo. ICI, immune checkpoint inhibitor; SiaM, sialic acid metabolism; DEGs, differentially expressed genes; ST, sialyltransferase; GC, gastric cancer

Full size image

Identification of three SiaM clusters with different clinicopathological characteristics and prognoses in gastric cancer

First, a comparison in transcriptional expression of 31 SiaM-related genes was conducted between normal gastric tissues and GC samples in the ACRG cohort. As indicated in Fig. 2A, we identified 20 dysregulated genes (all P < 0.05), among which the 11 genes were significantly upregulated in the GC samples, mainly involved in the metabolic process of sialic acid synthesis (GNE, NANP, and NANS), degradation (CTSA, NEU1, NEU2, and NEU3), storage (SLC17A5 and NPL), and transfer (ST6GALNAC4 and ST8SIA4). Besides, a total of 9 genes were significantly down-regulated in the GC group, which are involved in sialic acid transfer (ST3GAL5, ST3GAL6, ST6GALNAC2, ST6GALNAC3, ST6GALNAC5, ST6GALNAC6, ST8SIA1, ST8SIA3, and ST8SIA5). When using P < 0.05 & |Log1.5(Fold change)| > 1 as the threshold, a total of 6 DEGs were identified (Table S2), including 3 upregulated genes (NANP, CTSA, and NPL) and 3 downregulated genes (ST6GALNAC3, ST6GALNAC6, ST3GAL6).

Fig. 2
figure 2

Identification of SiaM-based metabolic subtypes with distinct clinicopathological features and survival outcomes in patients with gastric cancer. A Differences in the transcript levels of SiaM-related genes between normal and tumor tissues in the ACRG cohort. The comparison between these two groups was examined using the Mann‒Whitney test. B Consensus matrix heatmap of co-occurrence proportions in the GC samples for k = 3. The matrix shows three kinds of expression patterns of SiaM-related genes in the GC samples. C Consensus clustering CDF for k values from 2 to 10. The horizontal and vertical axis represents the consensus index and CDF value, respectively. D Principal component analysis of three SiaM clusters. The weights of the first and second principal components are 13.69% and 8.97%, respectively. E Heatmap showing the distinct expression patterns of SiaM-related genes among the three SiaM clusters. The matrix in the middle panel shows transcriptional expression of the SiaM-related genes by tumor sample (column) and by gene (row). The top tracks show clinicopathological characteristics and survival data of tumor samples. Comparison of the proportions of patients with different clinicopathological features, including TNM stage (F), Lauren type (G), ACRG molecular subtype (H), and first metastatic site (I), among the three SiaM clusters. Differences among the three subgroups were examined via chi-square tests. J Distribution of patients with different metastatic sites in GC. The left pie chart shows the percentage of three clusters in patients with liver metastasis. The right pie chart shows the percentage of three clusters in patients with peritoneal metastasis. Kaplan‒Meier curves showing the differences in overall survival (K) and disease-free survival (L) among the three SiaM clusters. HRs with 95% CIs were calculated via log-rank tests. CDF, cumulative distribution function; DFS, disease-free survival; OS, overall survival; SiaM, sialic acid metabolism; * P < 0.05; ** P < 0.01; *** P < 0.001; ns, not significant

Full size image

Next, the unsupervised consensus clustering was conducted based on the expression of 31 SiaM-related genes to classify GC patients into different subgroups in the ACRG cohort. According to the consensus matrix (Fig. 2B), CDF plot (Fig. 2C), and delta area curve (Figure S1), k = 3 was set as the optimal parameter for clustering. The PCA plot revealed obvious intergroup differences among the three clusters (Fig. 2D). Therefore, GC patients were categorized into Cluster A (n = 110), Cluster B (n = 85), and Cluster C (n = 105) based on the expression matrix of SiaM-related genes (Fig. 2E). Interestingly, we observed that GC patients in different clusters presented dramatically different clinicopathological characteristics (Table S3). More specifically, Cluster C harbored greater proportions of patients with advanced TNM stage (III/IV) than Clusters A and B (75.2% vs. 44.5% vs. 53.0%, P < 0.001; Fig. 2F). From a pathological perspective (Fig. 2G), tumors of the intestinal type were enriched mainly in Clusters A (58.7%) and B (66.7%), whereas most tumors of the diffuse type were enriched in Cluster C (71.4%). According to the ACRG molecular classification, 50% of the tumors in Cluster A had an MSI status, and 62.4% of the tumors in Cluster B were classified into the microsatellite stable/TP53-inactive (MSS/TP53−) subtype. Notably, tumors characterized by epithelial-to-mesenchymal transition (EMT) were all categorized into Cluster C (P < 0.001, Fig. 2H). More importantly, Cluster C had a significantly greater metastasis rate than Clusters A and B (47.6% vs. 31.8% vs. 27.1%, P < 0.001; Fig. 2I). The peritoneum and liver are recognized as the most common organs involved in GC metastasis [32, 33]. In this study, comparative analysis of the first metastatic site further revealed that Cluster C was preferentially associated with peritoneal metastasis compared with Clusters A and B (63.0% vs. 22.2% vs. 14.8%, P < 0.001; Fig. 2J; Table S4). Nevertheless, the association between the SiaM clusters and liver metastasis was not significant (P = 0.166, Table S4). In addition, we observed that Cluster C occurred at a significantly younger age (P = 0.001), had poorer differentiation characteristics (P < 0.001), and preferred to develop the Bormann IV type (P < 0.001) than Clusters A and B (Table S4). K‒M analyses demonstrated that GC patients in Cluster C had significantly worse OS and DFS than those in Clusters A and B (both P < 0.001, Fig. 2K, L). These findings preliminarily support the successful establishment of a novel metabolic classification system to stratify GC patients with distinct clinicopathological characteristics and survival outcomes.

Pathway enrichment and functional annotation of three SiaM clusters

To elucidate the potential biological processes associated with SiaM, we first conducted differential gene analyses to identify the DEGs among the three SiaM clusters and screened common DEGs from the genes differentially expressed between Clusters C and A or B. A total of 1097 DEGs were included in the pathway enrichment analyses (Fig. 3A; Table S5). The results of the GO enrichment analysis revealed that the DEGs were involved in several tumor stroma-related biological processes, including cell adhesion, collagen fibril organization, ECM organization, inflammatory response, and angiogenesis (Fig. 3B; Table S6). At the level of molecular function and cellular component, these DEGs provided binding sites for integrins and proteins in the extracellular space and plasma membrane (Fig. 3B; Table S6). The KEGG enrichment analysis revealed that the DEGs were associated with several key signaling pathways, such as focal adhesion, cell adhesion molecules, and leukocyte transendothelial migration (Fig. 3C; Table S6). In addition, the cancer hallmark pathways significantly different among the three SiaM clusters were identified via the ssGSEA algorithm. We observed that several cancer metastasis-related pathways and immune response signaling pathways, such as EMT, angiogenesis, and the inflammatory response, were more activated in Cluster C than in Clusters A and B (Fig. 3D; Table S7). GSEA revealed that sialic acid binding signaling was significantly enriched in Cluster C (both FDRs < 0.001, Fig. 3E, F). Regression analysis confirmed that the tumors in Cluster C presented an increased accumulation rate of sialic acid compared with those in Clusters A and B (both P < 0.001, Fig. 3G). Furthermore, based on reference gene sets from previous studies, multiple cancer migration-related characteristics, including cell–cell adhesion, cell contraction, motion and migration, and protrusion [24, 25], were quantified via the ssGSEA algorithm and then compared among the three clusters. The results indicated that tumors in Cluster C presented increased metastatic potential compared with those in Clusters A and B (all P < 0.001, Fig. 3H). Correlation analyses further revealed that SiaM activity was positively correlated with cancer migration-related characteristics (all P < 0.001, Figure S2). These results suggest that aberrant SiaM activity might be one of the crucial metabolic characteristics contributing to aggressive biological behaviors and unfavorable clinical outcomes in patients with GC.

Fig. 3
figure 3

Pathway enrichment and biological features associated with sialic acid metabolism in gastric cancer. A Venn diagram showing the DEGs among the three SiaM clusters. Common genes were identified from the DEGs between Clusters C and A and between Clusters C and B. Gene Ontology (B) and Kyoto Encyclopedia of Genes and Genomes (C) pathway enrichment for the common DEGs. Gene Ontology pathways included biological processes (blue), cellular components (red), and molecular function (purple). D Heatmap showing the relative activity of cancer hallmark pathways estimated by ssGSEA analysis. Gene set enrichment analysis showing the enrichment score of sialic acid binding signaling in Cluster C compared with those in Clusters A (E) and B (F). G Comparison of the sialic acid accumulation rates among the three SiaM clusters in gastric cancer. The sialic acid accumulation rate was estimated using the delta value between the rate of sialic acid placement onto the cell surface and the rate of sialic acid degradation. The rate of sialic acid placement onto the cell surface was determined by the geometric average of the transcriptional expression of CMAS and specific sialyltransferase genes, whereas the rate of sialic acid degradation was determined by the geometric average of the transcriptional expression of NEU1 and CTSA. H Boxplot showing the differences in the relative activity of several cancer metastasis-related biological processes among the three SiaM clusters in gastric cancer. Cancer metastasis-related biological processes, including cell–cell adhesion, cell contraction, motion and migration, and protrusion, reflect the metastatic potential of tumor cells. *** P < 0.001. GSEA, gene set enrichment analysis; ssGSEA, single-sample gene set enrichment analysis

Full size image

TME landscape of three SiaM clusters

In addition to the oncogenic role of sialic acids in tumor cells, the relationship between sialic acids and immunosuppressive TME was also reported in colorectal and pancreatic cancers [34, 35]. However, few studies have linked sialic acids with TME in GC. In this study, KEGG pathway analyses preliminarily supported that common DEGs between cluster C and cluster A/B were significantly enriched in multiple TME-related pathways, such as ECM-receptor interaction, cell adhesion molecules, and complement and coagulation cascades (Fig. 3C). To further explore the differences in multiple TME characteristics between three SiaM clusters in GC, we conducted a comprehensive TME analysis using the bulk and single-cell transcriptional data. The TME comprises all the nonneoplastic cells in the tumor entity, including diverse immune cells, endothelial cells, cancer-associated fibroblasts, and other tissue-resident cells, as well as its noncellular contents, including ECM components and immune molecules produced and released by various cells [36]. Firstly, the ESTIMATE algorithm revealed that Cluster C had higher immune, stromal, and ESTIMATE scores than Clusters A and B (all P < 0.001, Fig. 4A). Among the cellular components, Cluster C had the highest levels of endothelial cells, fibroblasts, naïve B cells, resting DCs, monocytes, M2 macrophages, and resting mast cells (Fig. 4B, C). In contrast, Cluster A harbored more activated immune cells, such as memory CD4+ T cells, NK cells, and neutrophils. Furthermore, we focused on the abundance of various immune molecules and stromal contents among the three clusters (Fig. 4D; Table S8). Compared with those in Clusters A and C, ICG ligands (17/22, 77.3%) and ICG receptors (20/21, 95.2%) were significantly downregulated in Cluster B. Notably, Cluster C harbored the most abundant MHC molecules (17/20, 85.0%) and contained the highest levels of immunosuppressive chemokines, such as CXCL12, CCL2, and CCL22. Interestingly, we observed that many genes involved in the biological processes of ECM remodeling (28/31, 90.3%), the TGF-β response in CAFs (17/19, 89.5%), and angiogenesis (4/4, 100%) were consistently enriched in Cluster C. Three integrin pairs (integrin α2β1, integrin α6β1, and integrin α6β4), known as metastasis suppressors, were consistently downregulated in Cluster C. SIGLECs expressed on most immune cells serve as cell membrane receptors for SAs and participate in pathogen recognition, the immune response, and tumor progression [37]. Comparative analyses revealed that SIGLEC1, SIGLEC2, SIGLEC3, SIGLEC5, SIGLEC6, SIGLEC7, SIGLEC9, SIGLEC10, and SIGLEC15 were all overexpressed in Cluster C (all P < 0.05, Fig. 4E). We speculated that aberrant activation of the sialic acid-SIGLEC axis was correlated with immune dysregulation in the TME. Using the TIDE algorithm, we estimated the relative levels of tumor immune dysfunction and exclusion in each sample and compared these scores among the three clusters (Fig. 4F). The results indicated that Cluster C harbored the highest level of T-cell co-inhibition (P < 0.001). Cluster B had the highest level of T-cell exclusion (P < 0.001), the lowest level of the IFN-γ signature (P < 0.001), and the highest TIDE score (P < 0.001). Compared with Cluster C, Cluster A harbored a lower level of T-cell exclusion (P < 0.001) and a higher level of the IFN-γ signature than Cluster B (P < 0.001). Notably, patients in Cluster A had the lowest TIDE score (P < 0.001), indicating that they were the most sensitive to immunotherapy. Taken together, on the one hand, the TME landscape determined the distinct tumor biological and clinicopathological features among the three SiaM clusters; on the other hand, the large differences in immunophenotypes among the three clusters provided a molecular basis for screening GC patients who might benefit from immunotherapy.

Fig. 4
figure 4

Characterization of the tumor microenvironment landscape based on sialic acid metabolism in gastric cancer. A Comparison of tumor microenvironment scores, including stromal, immune, and ESTIMATE scores, in the SiaM clusters. These scores were determined using the ESTIMATE algorithm. Comparison of the infiltration levels of stromal (B) and immune cells (C) in the SiaM clusters. Stromal cells, including endothelial cells and fibroblasts, were quantified via the MCPcounter algorithm. The proportions of 22 types of tumor-infiltrating immune cells were estimated via the CIBERSORT algorithm. D Heatmap showing the gene expression matrix of several key immune (left panel) and stromal (right panel) components in the SiaM clusters, including ICG ligands, ICG receptors, MHC, chemokines, ECM remodeling-related genes, TGFβ response genes in CAFs, genes involved in angiogenesis, and integrins. E Comparison of transcriptional expression of the sialic acid-binding Ig-like lectin family among three SiaM clusters. The lower panel shows the cell-specific locations of the sialic acid-binding Ig-like lectin family. F Comparison of the relative levels of T-cell co-inhibition, T-cell exclusion, IFN-γ signature, and TIDE score among three SiaM clusters. These scores estimated by the TIDE algorithm represent the relative levels of immune infiltration and the potential of immunotherapy response. ICG, immune checkpoint gene; ECM, extracellular matrix; CAF, cancer-associated fibroblast; SIGLEC, sialic acid binding Ig-like lectin; TIDE, tumor immune dysfunction and exclusion. * P < 0.05; ** P < 0.01; *** P < 0.001; ns, not significant

Full size image

Deciphering the role of SiaM in tumor immune infiltration at single-cell resolution

The scRNA-seq data of 40 GC samples, including 10 normal gastric tissue samples, 26 primary tumor lesion samples, 1 normal peritoneum tissue sample, and 3 peritoneal metastatic lesion samples, were obtained from the GSE183904 dataset. After quality control and data standardization, the top 2000 highly variable genes were identified for cell clustering. PCA of the expression levels of these genes revealed that the data dimensionality was reduced to PC1-15. The UMAP algorithm was subsequently performed on PC1-15 to categorize all the cells into 19 clusters (Figure S3). A total of 10 cell types of those clusters were identified based on specific cell markers (Figure S4A–B), including endocrine cells (PROX1, GHRL, CHGA), mast cells (MS4A2, TPSB2, TPSAB1), endothelial cells (PECAM1, VWF, CDH5, CLDN5), fibroblasts (ACTA2, DCN, COL3A1, COL1A1, FAP), B cells (MS4A1, CD79A, DC19), macrophages (CD83, LYZ, CD14, CD163, CD68), epithelial cells (KRT18, KRT8, PGA3, PGC), plasma cells (MZB1, IGHA1, IGHG1), CD4+ T cells (CD4, CD3G, CD3E, CD3D), and CD8+ T cells (CCL5, CD8A, CD8B). The distributions of all the cells according to tissue origin and Lauren type are shown in Figure S4C–D. We subsequently investigated the associations between the Lauren type and immune infiltration in patients with GC. The results indicated that diffuse or mixed tumors contained more plasma cells, whereas intestinal tumors contained more CD8+ T cells (Figure S4E). Furthermore, the AUCell algorithm was applied to estimate the relative SiaM activity in each sample (Figure S4F). Compared with the group with low SiaM activity, the group with high SiaM activity harbored more plasma cells and macrophages, similar to the immune infiltrating feature of diffuse tumors (Figure S4G). Comparative analysis also revealed that diffuse or mixed tumors were characterized by higher SiaM activity than intestinal tumors (Figure S4H). In addition, we performed cell clustering for 4346 cells annotated with macrophages and then identified two macrophage clusters based on several cell markers, including M1-like (FCN1, S100A8, S100A9, and S100A12) and M2-like (CD163, APOE, FN1, C1QC, and SPP1) macrophages (Figures S4I, S5). Pseudo-time trajectory analysis was conducted to analyze the cell lineage compositions of the macrophages. The pseudo-time increased from the M1-like to the M2-like cell type (Figure S4J). The expression levels of SIGLEC3 and SIGLEC7 were significantly increased in M2-like macrophages (Figure S4K) and increased from the M1-like state to the M2-like state in the pseudo-time tree (Figure S4L). Nevertheless, the other members of the SIGLEC family were rarely expressed in M1-like and M2-like macrophages (Figure S6).

Construction and validation of a SiaM-based model for predicting the clinical outcomes of patients with GC

Our previous findings revealed the intrinsic connection between SiaM and peritoneal metastasis (Fig. 2J; Table S4). To establish a clinically actionable model for stratifying GC patients with different risks of peritoneal metastasis, we integrated and analyzed the transcriptional and proteomic profiles of GC samples from the ACRG and ZJUGC-A cohorts. First, we matched the SiaMGs from the ACRG cohort to the proteomic profile of the ZJUGC-A cohort. A total of 46 peritoneal metastasis-related differentially expressed proteins (DEPs), including 31 upregulated and 15 downregulated SiaMGs, were subsequently identified via limma analysis of the proteomic expression matrix (Fig. 5A). Using LASSO regression analysis (Fig. 5B; Figure S7), 6 peritoneal metastasis-related SiaMGs (ARHGAP6, ST3GAL1, ADAM28, C7, PLCL1, and TTC28) were screened for model construction as follows: SiaM score = 0.121 * ARHGAP6 + 0.318 * ST3GAL1 + 0.127 * ADAM28 + 0.129 * C7 + 0.310 * PLCL1 + 0.058 * TTC28. Correlation analysis revealed that the SiaM score was positively correlated with SiaM activity (R = 0.55, P < 2.2e−16; Fig. 5C), indicating that the simplified model retained its inherent biological features. Multiple cohort comparative analyses demonstrated that GC patients with peritoneal metastasis consistently presented higher SiaM scores than those with NM in the ZJUGC-A, ZJUGC-B, ACRG, and Takeno cohorts (Fig. 5D). ROC analyses revealed that the predictive performance of the SiaM score for peritoneal metastasis was effective and robust in the ZJUGC-B (AUC = 0.800, P < 0.001), ACRG (AUC = 0.796, P < 0.001), and Takeno (AUC = 0.681, P = 0.030) cohorts (Fig. 5E). In addition, Kaplan‒Meier analyses were performed to explore the prognostic feasibility of the SiaM-based model. We found that GC patients with high SiaM scores had poorer OS than those with low SiaM scores in the ZJUGC-A (HR = 4.90, P = 0.005; Fig. 5F), ZJUGC-B (HR = 4.12, P = 0.027; Fig. 5G), ACRG (HR = 1.76, P < 0.001; Fig. 5H), and TCGA-STAD (HR = 1.82, P < 0.001; Fig. 5I) cohorts. Similar results were observed for the assessment of DFS in the ACRG (HR = 1.89, P < 0.001; Fig. 5J) and TCGA-STAD (HR = 1.79, P = 0.003; Fig. 5K) cohorts. To explore the independent predictive and prognostic values of the SiaM-based model, multivariate logistic regression and Cox regression analyses were implemented in the ACRG cohort. The forest plot revealed that the SiaM score could independently predict peritoneal metastasis (OR = 3.12, P = 0.002), OS (HR = 1.44, P = 0.045), and DFS (OR = 1.61, P = 0.020) when adjusted for age, gender, Lauren type, and Bormann type (Fig. 5L). Similarly, the independent prognostic significance of the SiaM-based model was also validated in the TCGA-STAD cohort (Figure S8). To deepen the understanding of the SiaM-based model in clinical application, we further explored the associations between the SiaM score and other clinicopathological features. Significant differences in multiple clinicopathological characteristics, including age (P = 0.025), gender (P = 0.020), primary tumor site (P = 0.017), Lauren type (P < 0.001), WHO classification (P < 0.001), Borrmann type (P < 0.001), TNM stage (P < 0.001), ACRG molecular subtype (P < 0.001), and first metastatic site (P < 0.001), were observed between the groups with low and high SiaM scores (Table S9). These findings suggest that GC patients with high SiaM scores have a more advanced stage, greater peritoneal metastasis potential, and worse survival outcomes.

Fig. 5
figure 5

Construction and validation of a SiaM-based model for predicting clinical outcomes in gastric cancer. A Differentially expressed proteins identified from the common DEGs via proteomic limma analysis in the ZJUGC-A cohort. Red and blue points represent significantly up-regulated and down-regulated proteins, respectively. P < 0.05 & |Log1.5(Fold change)| > 1 was set as the threshold. B Identification of candidate variables for predicting peritoneal metastasis via LASSO regression analysis. The curves show that the parameters decreased to zero when the penalty (Lambda) increased. n = 6 was adopted as the best parameter. Ultimately, six candidate genes including ARHGAP6, ST3GAL1, ADAM28, C7, PLCL1, and TTC28 were enrolled in the model construction. C Correlation between the SiaM-based model (named the SiaM score) and SiaM activity in the ACRG cohort. The SiaM activity was estimated using the ssGSEA algorithm. Blue, yellow, and red dots represent cluster A, cluster B, and cluster C, respectively. D Association between the SiaM score and peritoneal metastasis in the ZJUGC-A, ZJUGC-B, ACRG, and Takeno cohorts. The comparison of the SiaM score was examined using the Mann‒Whitney test. E ROC curves showing the predictive performance of the SiaM score for peritoneal metastasis in the ZJUGC-B (blue), ACRG (red), and Takeno cohorts (green). Kaplan‒Meier curves showing the differences in overall survival between patients with low and high SiaM scores in the ZJUGC-A (F), ZJUGC-B (G), ACRG (H), and Takeno (I) cohorts. Kaplan‒Meier curves showing the differences in disease-free survival between patients with low and high SiaM scores in the ACRG (J) and TCG-STAD (K) cohorts. The HRs with 95% CIs were determined using log-rank tests. L Forest plot showing the independent predictive value of the SiaM score for peritoneal metastasis and survival outcomes via multivariate logistic and Cox regression analyses, respectively. The adjusted parameters included age, gender, Lauren type, and Bormann type. D, diffuse; M, mixed; I, intestinal; PM, peritoneal metastasis; NM, no metastasis; OS, overall survival; DFS, disease-free survival; * P < 0.05; ** P < 0.01; *** P < 0.001

Full size image

ST3GAL1 promotes GC cell migration and invasion in vitro and in vivo

The six candidate genes, including ARHGAP6, ST3GAL1, ADAM28, C7, PLCL1, and TTC28, have been screened out using LASSO regression analyses to develop a model for predicting peritoneal metastasis and survival outcomes in GC (Fig. 5B). The model named SiaM score was calculated according to the following formula: SiaM score = 0.121 * ARHGAP6 + 0.318 * ST3GAL1 + 0.127 * ADAM28 + 0.129 * C7 + 0.310 * PLCL1 + 0.058 * TTC28. In this model, ST3GAL had the largest coefficient compared with other genes (coefST3GAL1 = 0.318). These findings indicate that ST3GAL1 may not only serve as one of the most crucial markers for the model but also play a potential pro-oncogenic role in GC. Therefore, the biological function of ST3GAL1 in GC progression and metastasis was explored. The transcriptional expression of ST3GAL1 was successfully downregulated via siRNA (Figure S9A). CCK-8 assays revealed that the downregulation of ST3GAL1 did not affect the viability of HGC27 or MKN45 cells (Figure S9B–C). However, Transwell assays demonstrated that ST3GAL1 knockdown promoted the migration and invasion of HGC27 and MKN45 cells (Figure S9D–E). When using AGS and HGC27 cells stably overexpressing ST3GAL1 (Figure S9F), we did not observe an impact of ST3GAL1 on cell proliferation (Figure S9G–H). In contrast, the overexpression of ST3GAL1 enhanced the migration and invasion of AGS and HGC27 cells, whereas the application of STi suppressed this protumor effect (Figure S9I–J).

To examine whether ST3GAL1 promotes the peritoneal metastasis of GC cells in vivo, HGC27 cells overexpressing ST3GAL1 were intraperitoneally injected into nude mice, and a sialyltransferase inhibitor (3Fax-Peracetyl Neu5Ac, 60 mg/kg) was then applied to test its therapeutic efficacy (Figure S10A). The results revealed that tumor weights were significantly greater in the ST3GAL1-ov group than in the NC group (Figure S10B–C). The application of STi inhibited the peritoneal metastasis of GC cells overexpressing ST3GAL1 (Figure S10B–C). Furthermore, H&E staining revealed the histological features of peritoneal metastatic tumors, and IHC revealed enhanced staining of MAL-II and vimentin and weaker E-cadherin staining in the ST3GAL1-ov group than in the NC group (Figure S10D). STi partially reversed this tendency (Figure S10D). Ki67 staining was similar among the three groups (Figure S10D). Taken together, these experimental results suggest a potential oncogenic role of ST3GAL1 in GC progression and metastasis.

Discussion

Aberrant SiaM activity has been recognized as a metabolic hallmark of highly malignant phenotypes across diverse cancer types. For example, previous studies have demonstrated that overactivation of SiaM contributes to transcellular connectivity deconstruction, tumor proliferation, and immunosuppression in glioblastoma [38, 39]. In prostate cancer, aberrant sialylation was identified as a key promotor of tumor growth, invasion, and metastasis [40]. However, the molecular characteristics of SiaM and its immunological and clinical implications in GC remain unclear. This study provides the first comprehensive bioinformatic report to depict the molecular landscape of SiaM in large-scale GC cohorts with multi-omics data. Our metabolic classification based on SiaM-related genes could precisely stratify GC patients with different TME characteristics and clinical outcomes. Characterization of SiaM-related gene expression patterns could be a potentially applicable tool in clinical practice to guide prognostic stratification and individualized treatment. Thus, targeting SiaM might be a promising strategy to improve therapeutic sensitivity in patients with GC.

In this study, three distinct GC subtypes (clusters A–C) were established via consensus clustering based on the expression patterns of 31 SiaM-related genes in the ACRG cohort. Among the three clusters, Cluster C was identified as the most malignant subtype, characterized by elevated SiaM activity, aggressive clinicopathological features, high metastatic potential, and unfavorable prognosis. Conversely, Cluster A had more patients with an abundant MSI status and favorable prognosis, which might benefit from ICI therapy [41]. Mechanistically, various cancer-related pathways, such as cell adhesion, angiogenesis, PI3K-Akt signaling, the inflammatory response, and EMT, were enriched in Cluster C. Interestingly, we observed a significant association between ACRG molecular subtypes and SiaM-based metabolic subtypes in GC (Fig. 2H). All tumors characterized by the EMT subtype were enriched into Cluster C. A similar relationship between a specific gene or pathway and tumor subtype was also reported in pancreatic cancer [42,43,44,45]. For instance, Ohara et al. found that the activation of the SERPINB3-MYC axis could induce the basal-like/squamous subtype and promote disease progression in pancreatic cancer [45]. Similarly, Moreno et al. demonstrated that ADRA2A promoted the classical/progenitor subtype and reduced the disease aggressiveness of pancreatic cancer [44]. Brunton et al. observed that loss of HNF4A or GATA6 could drive a squamous-like metabolic profile in pancreatic cancer [42]. Besides, Daemen et al. stratified pancreatic ductal adenocarcinomas into three metabolic subtypes (slow proliferating, glycolytic, and lipogenic) which displayed distinct transcriptional landscapes associated with the metabolic ontologies [43]. Given these positive findings in previous studies, we speculate that dysregulation of SiaM may contribute to the EMT-like phenotype in GC, which deserves experimental validation in the future. Furthermore, Cluster C harbored high levels of cell adhesion, contraction, migration, and protrusion, indicating an increased metastatic potential of tumor cells [25]. From a pathological perspective, more than 70% of the tumors in Cluster C were characterized as diffuse under the Lauren classification system. Interestingly, a computational biology study demonstrated that sialic acid accumulation increased the number of negative charges on the cell surface and dictated the morphology and aggressiveness of diffuse GCs [24]. Apart from this biophysical reason, our findings further support that aberrant activation of those biological pathways induced by hypersialylation is another leading cause of poor differentiation and high metastatic potential in diffuse-type GC.

Sialylation contributes to immune modulation in the TME via interactions with tumor-infiltrating immune cells. The sialic acid residue of glycans has been identified as the ligand for SIGLECs, a subset of immune checkpoint proteins. The activation of the sialic acid-SIGLEC axis can inhibit the protumorigenic activities of various immune effector cells, such as NK cells [46], CD8+ T cells [47], and macrophages [48]. In this study, Cluster C had an immune-excluded phenotype characterized by abundant infiltration of naïve/resting and immunosuppressive cells (naïve B cells, resting DCs, M2 macrophages, and resting mast cells), high enrichment of immune checkpoints, activated programs of ECM remodeling and angiogenesis, inhibition of the T-cell response, and a high tendency toward immunotherapy resistance. At single-cell resolution, we found that the immune infiltration pattern in tumors with high SiaM activity contained abundant plasma cells and macrophages, which was similar to that in tumors with diffuse differentiation [29]. Furthermore, SIGLECs contribute to the polarization of macrophages from M1 to M2 in the TME, suggesting that M2-oriented macrophage polarization might play a vital role in sialic acid-SIGLEC axis-mediated immunosuppression [49]. Based on these findings, we believe that aberrant SiaM activity is a metabolic hallmark of tumor immune evasion in GC, especially in the diffuse type.

Sialylation-mediated resistance to multiple chemotherapy drugs has been observed in ovarian cancer [50] and multiple myeloma [51]. Notably, sialylation of ERBB2 dramatically decreased sensitivity to trastuzumab in GC cells [52]. Consistent with these previous findings, we also observed that GC patients with high SiaM scores might not be suitable for anti-ERBB2 therapy and ICI therapy (Figure S11). Nevertheless, CLDN18 was overexpressed in tumors with high SiaM scores (Figure S11), suggesting that these patients might benefit from anti-CLDN18.2 therapy [53]. Targeting sialylation has emerged as a promising strategy for preventing tumor metastasis and improving drug resistance, including target sialylation formation and degradation as well as blockade of the sialic acid-SIGLEC interaction [13]. The effect of STi is to target the ST enzyme binding site and interfere with enzyme activity. It avoids the transfer of sialic acid residues to the glycoconjugate, resulting in a decrease in hypersialylation on the cell surface [54]. In this study, we observed that the application of STi significantly reversed the oncogenic effect of ST3GAL1 on cell migration and metastasis both in vitro and in vivo. This study provides a preclinical foundation for the development of novel treatment strategies for GC patients.

Considering the high incidence and unfavorable prognosis of GC patients with peritoneal metastasis, previous studies have identified various predictive biomarkers based on the tissue-derived epigenome, transcriptome, metabolome, and exosomes [55,56,57,58]. Despite these encouraging findings, several limitations, including the use of noncomprehensive training methods, the limited number of patients, and the lack of independent validation, have restricted its application in clinical practice. Based on the close association between SiaM-related genes and peritoneal metastasis (Fig. 2J), we conducted a comprehensive bioinformatic analysis of transcriptomic and proteomic data to identify a clinically actionable model for predicting peritoneal metastasis and prognosis in GC patients. Ultimately, the predictive performance of the model, which included six SiaM-related genes (ARHGAP6, ST3GAL1, ADAM28, C7, PLCL1, and TTC28), was effective and robust in four independent cohorts. In addition, this model had an independent prognostic value for GC patients. Therefore, our findings provide a convenient and effective tool for stratifying GC patients with different risks of peritoneal metastasis and prognoses in clinical practice, which deserves further validation in prospective cohorts. However, the biological mechanism by which these SiaM-related genes contribute to peritoneal metastasis in GC remains unknown. ARHGAP6, a member of the Rho GTPase-activating protein family, has been identified as a tumor-suppressive biomarker in lung [59], bladder [60], and cervical [61] cancers. ADAM28 is involved in cell–cell and cell–matrix interactions. Yin et al. demonstrated that ADAM28 from both GC cells and endothelial cells could inhibit von Willebrand factor (vWF)-induced apoptosis of GC cells by cleaving vWF [62]. C7, a member of complement components, has been identified as a tumor suppressor in prostate cancer [63]. In contrast, high expression of C7 is correlated with unfavorable prognosis and chemotherapy resistance in breast cancer [64]. PLCL1 is involved in an inositol phospholipid-based intracellular signaling cascade. Khan et al. identified PLCL1 as one of the key necroptosis-related genes for predicting prognosis and immune infiltration [65]. TTC28 can control microtubule dynamics during embryonic cleavage [66]. Despite these research advances, the molecular mechanism by which these SiaM-related genes promote peritoneal metastasis in GC requires further exploration via in vitro/in vivo experiments. ST3GAL1 catalyzes the transfer of sialic acids to glycans, which are responsible for the formation of α2,3-sialylated glycans [13]. Previous studies have reported the pro-tumor effect of ST3GAL1 in diverse cancers [67, 68]. Fan et al. observed a reciprocal feedback loop of ST3GAL1 and GFRA1 signaling which promoted cell proliferation in breast cancer [68]. Lin et al. found that ST3GAL1-mediated sialylation of CD55 acted as an immune checkpoint and promoted tumor immune evasion in breast cancer [69]. Chong et al. reported that ST3GAL1-mediated self-renewal capacity was crucial to the sustenance of cell multiforme growth in glioblastoma [67]. To date, few studies have reported the role of ST3GAL1 in gastrointestinal cancers. To our knowledge, this is the first study to uncover the oncogenic effect of ST3GAL1 on the progression and metastasis of GC. Furthermore, bioinformatic analyses were performed to elucidate the underlying mechanisms of the biological function of ST3GAL1. GO enrichment analyses revealed that ST3GAL1 was involved in several immune-related pathways, such as the regulation of the inflammatory response, response to molecules of bacterial origin, and leukocyte migration (Figure S12). In vitro and in vivo experiments supported that ST3GAL1 promoted GC cell metastasis (Figures S9–S10). From the perspective of molecular mechanisms, we speculated that ST3GAL1 overexpression may induce aberrant accumulation of sialic acid on the tumor cell surface and increase negative charges, eventually leading to cell–cell repulsion and cell detachment from primary tumor lesions [15]. Besides, terminal sialic acid on the tumor cell membrane may bind to SIGLECs on tumor-infiltrating immune cells and then promote tumor immune evasion [14]. These potential mechanisms deserve more experimental validations in the future.

Nonetheless, several potential limitations need to be illustrated. First, the ACRG dataset was used as the training cohort, and other cohorts (TCGA-STAD, Takeno, ZJUGC-A, ZJUGC-B) were used as the validation cohorts in this study. The heterogeneity between these cohorts might introduce potential confounding factors. Second, the SiaM-related genes in the model were identified via bioinformatic analyses based on transcriptomic and proteomic data, and more functional experiments are needed to explore the biological mechanisms of candidate genes in GC peritoneal metastasis. Despite the preliminary results from pathway enrichment analyses, the molecular mechanisms by which ST3GAL1 promotes the migration and metastasis of GC cells deserve further investigation in the future.

In conclusion, this study proposed a novel metabolic classification system based on SiaM for stratifying GC patients with distinct TME patterns and clinical outcomes for the first time. A six-gene risk model with high efficacy and robustness was successfully established to predict peritoneal metastasis and prognosis in patients with GC. ST3GAL1 might be a potential therapeutic target for GC. These findings will guide prognostic stratification and individualized treatment for GC patients in clinical practice.

Availability of data and materials

The TCGA-STAD dataset was obtained from the cBioportal database (https://www.cbioportal.org/). The microarray (GSE66229, GSE15081) and scRNA-seq (GSE183904) datasets were acquired from the GEO database (http://www.ncbi.nlm.nih.gov/geo/). The RNA-seq data and clinical information of the Kim cohort were acquired from the original article with supplementary materials. The proteomic data of the ZJUGC-A cohort are available from the corresponding author upon reasonable request.

Abbreviations

GC:

Gastric cancer

SiaM:

Sialic acid metabolism

TME:

Tumor microenvironment

TCGA:

The Cancer Genome Atlas Consortium

GEO:

Gene Expression Omnibus

MSI:

Microsatellite instability

EBV:

Epstein-Barr virus

CIN:

Chromosomal instability

GS:

Genomically stable

ACRG:

Asian Cancer Research Group

CMP-Neu5Ac:

Cytidine monophosphate N-acetylneuraminic acid

ECM:

Extracellular matrix

SIGLECs:

Sialic acid immunoglobulin-type lectins

ST:

Sialyltransferase

STi:

Sialyltransferase inhibitor

CDF:

Cumulative distribution function

PCA:

Principal component analysis

DEGs:

Differentially expressed genes

GO:

Gene Ontology

KEGG:

Kyoto Encyclopedia of Genes and Genomes

GSEA:

Gene set enrichment analysis

ssGSEA:

Single-sample gene set enrichment analysis

CAF:

Cancer-associated fibroblast

LASSO:

Least absolute shrinkage and selection operator

SiaMGs:

DEGs among three SiaM clusters

EMT:

Epithelial-to-mesenchymal transition

ICI:

Immune checkpoint inhibitor

References

  1. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2021;71(3):209–49.

    Article  PubMed  Google Scholar 

  2. Alsina M, Arrazubi V, Diez M, Tabernero J. Current developments in gastric cancer: from molecular profiling to treatment strategy. Nat Rev Gastroenterol Hepatol. 2023;20(3):155–70.

    Article  CAS  PubMed  Google Scholar 

  3. Cancer Genome Atlas Research Network. Comprehensive molecular characterization of gastric adenocarcinoma. Nature. 2014;513(7517):202–9.

    Article  Google Scholar 

  4. Cristescu R, Lee J, Nebozhyn M, Kim KM, Ting JC, Wong SS, et al. Molecular analysis of gastric cancer identifies subtypes associated with distinct clinical outcomes. Nat Med. 2015;21(5):449–56.

    Article  CAS  PubMed  Google Scholar 

  5. Balmana M, Diniz F, Feijao T, Barrias CC, Mereiter S, Reis CA. Analysis of the effect of increased α2,3-sialylation on RTK activation in MKN45 gastric cancer spheroids treated with crizotinib. Int J Mol Sci. 2020;21(3):722.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  6. Yang D, Yin J, Shan L, Yi X, Zhang W, Ding Y. Identification of lysine-lactylated substrates in gastric cancer cells. iScience. 2022;25(7): 104630.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. Zhao C, Tao T, Yang L, Qin Q, Wang Y, Liu H, et al. Loss of PDZK1 expression activates PI3K/AKT signaling via PTEN phosphorylation in gastric cancer. Cancer Lett. 2019;453:107–21.

    Article  PubMed  Google Scholar 

  8. Dobie C, Skropeta D. Insights into the role of sialylation in cancer progression and metastasis. Br J Cancer. 2021;124(1):76–90.

    Article  CAS  PubMed  Google Scholar 

  9. Britain CM, Bhalerao N, Silva AD, Chakraborty A, Buchsbaum DJ, Crowley MR, et al. Glycosyltransferase ST6Gal-I promotes the epithelial to mesenchymal transition in pancreatic cancer cells. J Biol Chem. 2021;296: 100034.

    Article  CAS  PubMed  Google Scholar 

  10. Kvorjak M, Ahmed Y, Miller ML, Sriram R, Coronnello C, Hashash JG, et al. Cross-talk between colon cells and macrophages increases ST6GALNAC1 and MUC1-sTn expression in ulcerative colitis and colitis-associated colon cancer. Cancer Immunol Res. 2020;8(2):167–78.

    Article  CAS  PubMed  Google Scholar 

  11. Wu X, Zhao J, Ruan Y, Sun L, Xu C, Jiang H. Sialyltransferase ST3GAL1 promotes cell migration, invasion, and TGF-beta1-induced EMT and confers paclitaxel resistance in ovarian cancer. Cell Death Dis. 2018;9(11):1102.

    Article  PubMed  PubMed Central  Google Scholar 

  12. Yeo HL, Fan TC, Lin RJ, Yu JC, Liao GS, Chen ES, et al. Sialylation of vasorin by ST3Gal1 facilitates TGF-beta1-mediated tumor angiogenesis and progression. Int J Cancer. 2019;144(8):1996–2007.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  13. Al Saoud R, Hamrouni A, Idris A, Mousa WK, Abu IT. Recent advances in the development of sialyltransferase inhibitors to control cancer metastasis: a comprehensive review. Biomed Pharmacother. 2023;165: 115091.

    Article  CAS  PubMed  Google Scholar 

  14. Laubli H, Varki A. Sialic acid-binding immunoglobulin-like lectins (Siglecs) detect self-associated molecular patterns to regulate immune responses. Cell Mol Life Sci. 2020;77(4):593–605.

    Article  CAS  PubMed  Google Scholar 

  15. Pinho SS, Reis CA. Glycosylation in cancer: mechanisms and clinical implications. Nat Rev Cancer. 2015;15(9):540–55.

    Article  CAS  PubMed  Google Scholar 

  16. Miro L, Lopez J, Guerrero PE, Martinez-Bosch N, Manero-Ruperez N, Moreno M, et al. Sialyltransferase inhibitor Ac53FaxNeu5Ac reverts the malignant phenotype of pancreatic cancer cells, and reduces tumor volume and favors T-cell infiltrates in mice. Cancers. 2022;14(24):6133.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  17. Shu X, Li J, Chan UI, Su SM, Shi C, Zhang X, et al. BRCA1 insufficiency induces a hypersialylated acidic tumor microenvironment that promotes metastasis and immunotherapy resistance. Cancer Res. 2023;83(15):2614–33.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  18. Zhang X, Xu CH, Mo J, Zheng XJ, Chen YF, Yang AQ, et al. Self-assembled core-shell nanoscale coordination polymer nanoparticles carrying a sialyltransferase inhibitor for cancer metastasis inhibition. ACS Appl Mater Interfaces. 2023;15(6):7713–24.

    Article  CAS  PubMed  Google Scholar 

  19. Kim ST, Cristescu R, Bass AJ, Kim KM, Odegaard JI, Kim K, et al. Comprehensive molecular characterization of clinical responses to PD-1 inhibition in metastatic gastric cancer. Nat Med. 2018;24(9):1449–58.

    Article  CAS  PubMed  Google Scholar 

  20. Chen Y, Cai G, Jiang J, He C, Chen Y, Ding Y, et al. Proteomic profiling of gastric cancer with peritoneal metastasis identifies a protein signature associated with immune microenvironment and patient outcome. Gastric Cancer. 2023;26(4):504–16.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  21. Barbie DA, Tamayo P, Boehm JS, Kim SY, Moody SE, Dunn IF, et al. Systematic RNA interference reveals that oncogenic KRAS-driven cancers require TBK1. Nature. 2009;462(7269):108–12.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  22. Li Y, Yu J, Li R, Zhou H, Chang X. New insights into the role of mitochondrial metabolic dysregulation and immune infiltration in septic cardiomyopathy by integrated bioinformatics analysis and experimental validation. Cell Mol Biol Lett. 2024;29(1):21.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  23. Mannheimer JD, Tawa G, Gerhold D, Braisted J, Sayers CM, McEachron TA, et al. Transcriptional profiling of canine osteosarcoma identifies prognostic gene expression signatures with translational value for humans. Commun Biol. 2023;6(1):856.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  24. Liu D, Xing F, Wang Y, Xiao J, An Z, Xu Y. Molecular bases of morphologically diffused tumors across multiple cancer types. Natl Sci Rev. 2022;9(11):nwac177.

    Article  PubMed  PubMed Central  Google Scholar 

  25. Sun H, Zhou Y, Jiang H, Xu Y. Elucidation of functional roles of sialic acids in cancer migration. Front Oncol. 2020;10:401.

    Article  PubMed  PubMed Central  Google Scholar 

  26. Yoshihara K, Shahmoradgoli M, Martinez E, Vegesna R, Kim H, Torres-Garcia W, et al. Inferring tumour purity and stromal and immune cell admixture from expression data. Nat Commun. 2013;4:2612.

    Article  PubMed  Google Scholar 

  27. Becht E, Giraldo NA, Lacroix L, Buttard B, Elarouci N, Petitprez F, et al. Estimating the population abundance of tissue-infiltrating immune and stromal cell populations using gene expression. Genome Biol. 2016;17(1):218.

    Article  PubMed  PubMed Central  Google Scholar 

  28. Newman AM, Liu CL, Green MR, Gentles AJ, Feng W, Xu Y, et al. Robust enumeration of cell subsets from tissue expression profiles. Nat Methods. 2015;12(5):453–7.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  29. Kumar V, Ramnarayanan K, Sundar R, Padmanabhan N, Srivastava S, Koiwa M, et al. Single-cell atlas of lineage states, tumor microenvironment, and subtype-specific expression programs in gastric cancer. Cancer Discov. 2022;12(3):670–91.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  30. Aibar S, Gonzalez-Blas CB, Moerman T, Huynh-Thu VA, Imrichova H, Hulselmans G, et al. SCENIC: single-cell regulatory network inference and clustering. Nat Methods. 2017;14(11):1083–6.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  31. Qiu X, Hill A, Packer J, Lin D, Ma YA, Trapnell C. Single-cell mRNA quantification and differential analysis with Census. Nat Methods. 2017;14(3):309–15.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  32. Gwee YX, Chia DKA, So J, Ceelen W, Yong WP, Tan P, et al. Integration of genomic biology into therapeutic strategies of gastric cancer peritoneal metastasis. J Clin Oncol. 2022;40(24):2830.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  33. Markar SR, Mikhail S, Malietzis G, Athanasiou T, Mariette C, Sasako M, et al. Influence of surgical resection of hepatic metastases from gastric adenocarcinoma on long-term survival: systematic review and pooled analysis. Ann Surg. 2016;263(6):1092–101.

    Article  PubMed  Google Scholar 

  34. Boelaars K, Goossens-Kruijssen L, Wang D, de Winde CM, Rodriguez E, Lindijer D, et al. Unraveling the impact of sialic acids on the immune landscape and immunotherapy efficacy in pancreatic cancer. J Immunother Cancer. 2023;11(11): e007805.

    Article  PubMed  PubMed Central  Google Scholar 

  35. Egan H, Treacy O, Lynch K, Leonard NA, O’Malley G, Reidy E, et al. Targeting stromal cell sialylation reverses T cell-mediated immunosuppression in the tumor microenvironment. Cell Rep. 2023;42(5): 112475.

    Article  CAS  PubMed  Google Scholar 

  36. de Visser KE, Joyce JA. The evolving tumor microenvironment: from cancer initiation to metastatic outgrowth. Cancer Cell. 2023;41(3):374–403.

    Article  PubMed  Google Scholar 

  37. Gonzalez-Gil A, Li TA, Kim J, Schnaar RL. Human sialoglycan ligands for immune inhibitory Siglecs. Mol Aspects Med. 2023;90: 101110.

    Article  CAS  PubMed  Google Scholar 

  38. Kuliesiute U, Joseph K, Straehle J, Ravi VM, Kueckelhaus J, Kada Benotmane J, et al. Sialic acid metabolism orchestrates transcellular connectivity and signalling in glioblastoma. Neuro Oncol. 2023;25(11):1963–75.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  39. Schmassmann P, Roux J, Buck A, Tatari N, Hogan S, Wang J, et al. Targeting the Siglec-sialic acid axis promotes antitumor immune responses in preclinical models of glioblastoma. Sci Transl Med. 2023;15(705):eadf302.

    Article  Google Scholar 

  40. Scott E, Archer Goode E, Garnham R, Hodgson K, Orozco-Moreno M, Turner H, et al. ST6GAL1-mediated aberrant sialylation promotes prostate cancer progression. J Pathol. 2023;261(1):71–84.

    Article  CAS  PubMed  Google Scholar 

  41. Marabelle A, Le DT, Ascierto PA, Di Giacomo AM, De Jesus-Acosta A, Delord JP, et al. Efficacy of pembrolizumab in patients with noncolorectal high microsatellite instability/mismatch repair-deficient cancer: results from the phase II KEYNOTE-158 study. J Clin Oncol. 2020;38(1):1–10.

    Article  CAS  PubMed  Google Scholar 

  42. Brunton H, Caligiuri G, Cunningham R, Upstill-Goddard R, Bailey UM, Garner IM, et al. HNF4A and GATA6 loss reveals therapeutically actionable subtypes in pancreatic cancer. Cell Rep. 2020;31(6): 107625.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  43. Daemen A, Peterson D, Sahu N, McCord R, Du X, Liu B, et al. Metabolite profiling stratifies pancreatic ductal adenocarcinomas into subtypes with distinct sensitivities to metabolic inhibitors. Proc Natl Acad Sci USA. 2015;112(32):E4410–7.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  44. Moreno P, Ohara Y, Craig AJ, Liu H, Yang S, Dorsey TH, et al. ADRA2A promotes the classical/progenitor subtype and reduces disease aggressiveness of pancreatic cancer. Carcinogenesis. 2024;45(11):845–56.

    Article  CAS  PubMed  Google Scholar 

  45. Ohara Y, Tang W, Liu H, Yang S, Dorsey TH, Cawley H, et al. SERPINB3-MYC axis induces the basal-like/squamous subtype and enhances disease progression in pancreatic cancer. Cell Rep. 2023;42(12): 113434.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  46. Grabenstein S, Barnard KN, Anim M, Armoo A, Weichert WS, Bertozzi CR, et al. Deacetylated sialic acids modulates immune mediated cytotoxicity via the sialic acid-Siglec pathway. Glycobiology. 2021;31(10):1279–94.

    Article  CAS  PubMed  Google Scholar 

  47. Mei Y, Wang X, Zhang J, Liu D, He J, Huang C, et al. Siglec-9 acts as an immune-checkpoint molecule on macrophages in glioblastoma, restricting T-cell priming and immunotherapy response. Nat Cancer. 2023;4(9):1273–91.

    Article  CAS  PubMed  Google Scholar 

  48. Thiesler H, Gretenkort L, Hoffmeister L, Albers I, Ohlmeier L, Rockle I, et al. Proinflammatory macrophage activation by the polysialic acid-Siglec-16 axis is linked to increased survival of patients with glioblastoma. Clin Cancer Res. 2023;29(12):2266–79.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  49. Najafi M, Hashemi Goradel N, Farhood B, Salehi E, Nashtaei MS, Khanlarkhani N, et al. Macrophage polarity in cancer: a review. J Cell Biochem. 2019;120(3):2756–65.

    Article  CAS  PubMed  Google Scholar 

  50. Ou L, He X, Liu N, Song Y, Li J, Gao L, et al. Sialylation of FGFR1 by ST6Gal-I overexpression contributes to ovarian cancer cell migration and chemoresistance. Mol Med Rep. 2020;21(3):1449–60.

    CAS  PubMed  PubMed Central  Google Scholar 

  51. Natoni A, Bohara R, Pandit A, O’Dwyer M. Targeted approaches to inhibit sialylation of multiple myeloma in the bone marrow microenvironment. Front Bioeng Biotechnol. 2019;7:252.

    Article  PubMed  PubMed Central  Google Scholar 

  52. Duarte HO, Rodrigues JG, Gomes C, Hensbergen PJ, Ederveen ALH, de Ru AH, et al. ST6Gal1 targets the ectodomain of ErbB2 in a site-specific manner and regulates gastric cancer cell sensitivity to trastuzumab. Oncogene. 2021;40(21):3719–33.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  53. Shitara K, Lordick F, Bang YJ, Enzinger P, Ilson D, Shah MA, et al. Zolbetuximab plus mFOLFOX6 in patients with CLDN18.2-positive, HER2-negative, untreated, locally advanced unresectable or metastatic gastric or gastro-oesophageal junction adenocarcinoma (SPOTLIGHT): a multicentre, randomised, double-blind, phase 3 trial. Lancet. 2023;401(10389):1655–68.

    Article  CAS  PubMed  Google Scholar 

  54. Szabo R, Skropeta D. Advancement of sialyltransferase inhibitors: therapeutic challenges and opportunities. Med Res Rev. 2017;37(2):219–70.

    Article  PubMed  Google Scholar 

  55. Hiraki M, Kitajima Y, Koga Y, Tanaka T, Nakamura J, Hashiguchi K, et al. Aberrant gene methylation is a biomarker for the detection of cancer cells in peritoneal wash samples from advanced gastric cancer patients. Ann Surg Oncol. 2011;18(10):3013–9.

    Article  PubMed  Google Scholar 

  56. Kaji S, Irino T, Kusuhara M, Makuuchi R, Yamakawa Y, Tokunaga M, et al. Metabolomic profiling of gastric cancer tissues identified potential biomarkers for predicting peritoneal recurrence. Gastric Cancer. 2020;23(5):874–83.

    Article  PubMed  Google Scholar 

  57. Sawaki K, Kanda M, Miwa T, Umeda S, Tanaka H, Tanaka C, et al. Troponin I2 as a specific biomarker for prediction of peritoneal metastasis in gastric cancer. Ann Surg Oncol. 2018;25(7):2083–90.

    Article  PubMed  Google Scholar 

  58. Tokuhisa M, Ichikawa Y, Kosaka N, Ochiya T, Yashiro M, Hirakawa K, et al. Exosomal miRNAs from peritoneum lavage fluid as potential prognostic biomarkers of peritoneal metastasis in gastric cancer. PLoS ONE. 2015;10(7): e0130472.

    Article  PubMed  PubMed Central  Google Scholar 

  59. Wu Y, Xu M, He R, Xu K, Ma Y. ARHGAP6 regulates the proliferation, migration and invasion of lung cancer cells. Oncol Rep. 2019;41(4):2281–888.

    CAS  PubMed  Google Scholar 

  60. Chen W, Tan M, Yu C, Liao G, Kong D, Bai J, et al. ARHGAP6 inhibits bladder cancer cell viability, migration, and invasion via beta-catenin signaling and enhances mitomycin C sensitivity. Hum Cell. 2023;36(2):786–97.

    Article  CAS  PubMed  Google Scholar 

  61. Li J, Liu Y, Yin Y. Inhibitory effects of Arhgap6 on cervical carcinoma cells. Tumour Biol. 2016;37(2):1411–25.

    Article  CAS  PubMed  Google Scholar 

  62. Yin Q, Gu J, Qi Y, Lu Y, Yang L, Liu J, et al. ADAM28 from both endothelium and gastric cancer cleaves von Willebrand Factor to eliminate von Willebrand factor-induced apoptosis of gastric cancer cells. Eur J Pharmacol. 2021;898: 173994.

    Article  CAS  PubMed  Google Scholar 

  63. Chen Z, Yan X, Du GW, Tuoheti K, Bai XJ, Wu HH, et al. Complement C7 (C7), a potential tumor suppressor, is an immune-related prognostic biomarker in prostate cancer (PC). Front Oncol. 2020;10:1532.

    Article  PubMed  PubMed Central  Google Scholar 

  64. Zhang H, Zhao Y, Liu X, Fu L, Gu F, Ma Y. High expression of complement component C7 indicates poor prognosis of breast cancer and is insensitive to taxane-anthracycline chemotherapy. Front Oncol. 2021;11: 724250.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  65. Khan M, Lin J, Wang B, Chen C, Huang Z, Tian Y, et al. A novel necroptosis-related gene index for predicting prognosis and a cold tumor immune microenvironment in stomach adenocarcinoma. Front Immunol. 2022;13: 968165.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  66. Chen J, Castelvecchi GD, Li-Villarreal N, Raught B, Krezel AM, McNeill H, et al. Atypical cadherin Dachsous1b interacts with Ttc28 and Aurora B to control microtubule dynamics in embryonic cleavages. Dev Cell. 2018;45(3):376–91.e5.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  67. Chong YK, Sandanaraj E, Koh LW, Thangaveloo M, Tan MS, Koh GR, et al. ST3GAL1-associated transcriptomic program in glioblastoma tumor growth, invasion, and prognosis. J Natl Cancer Inst. 2016;108(2):djv326.

    Article  PubMed  Google Scholar 

  68. Fan TC, Yeo HL, Hsu HM, Yu JC, Ho MY, Lin WD, et al. Reciprocal feedback regulation of ST3GAL1 and GFRA1 signaling in breast cancer cells. Cancer Lett. 2018;434:184–95.

    Article  CAS  PubMed  Google Scholar 

  69. Lin WD, Fan TC, Hung JT, Yeo HL, Wang SH, Kuo CW, et al. Sialylation of CD55 by ST3GAL1 facilitates immune evasion in cancer. Cancer Immunol Res. 2021;9(1):113–22.

    Article  CAS  PubMed  Google Scholar 

Download references

Acknowledgements

We sincerely thank the research teams who provided the gene expression profiles with annotated clinicopathological information in the TCGA database, GEO database, and online publications.

Funding

This research was funded by the Construction Fund of Medical Key Disciplines of Hangzhou (OO20190001).

Author information

Author notes

  1. Junjie Jiang, Yiran Chen, and Yangyang Zheng have contributed equally to this work.

Authors and Affiliations

  1. Institute of Immunology, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China

    Junjie Jiang & Quan Zhou

  2. Department of Gastroenterology, Affiliated Hangzhou First People’S Hospital, Westlake University School of Medicine, 261 Huansha Road, Hangzhou, 310006, Zhejiang, China

    Junjie Jiang & Xiaofeng Zhang

  3. Department of Surgical Oncology, the First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China

    Yiran Chen, Haiyong Wang, Quan Zhou & Lisong Teng

  4. The Fourth School of Clinical Medicine, Zhejiang Chinese Medical University, Hangzhou, Zhejiang, China

    Yangyang Zheng

  5. Department of Medical Oncology, the First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China

    Yongfeng Ding

  6. Key Laboratory of Integrated Traditional Chinese and Western Medicine for Biliary and Pancreatic Diseases of Zhejiang Province, Hangzhou, Zhejiang, China

    Junjie Jiang & Xiaofeng Zhang

  7. Key Laboratory of Clinical Cancer Pharmacology and Toxicology Research of Zhejiang Province, Hangzhou, Zhejiang, China

    Junjie Jiang & Xiaofeng Zhang

  8. Hangzhou Institute of Digestive Disease, Hangzhou, Zhejiang, China

    Junjie Jiang & Xiaofeng Zhang

Authors
  1. Junjie Jiang
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Yiran Chen
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Yangyang Zheng
    View author publications

    You can also search for this author inPubMed Google Scholar

  4. Yongfeng Ding
    View author publications

    You can also search for this author inPubMed Google Scholar

  5. Haiyong Wang
    View author publications

    You can also search for this author inPubMed Google Scholar

  6. Quan Zhou
    View author publications

    You can also search for this author inPubMed Google Scholar

  7. Lisong Teng
    View author publications

    You can also search for this author inPubMed Google Scholar

  8. Xiaofeng Zhang
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

Conceptualization, X.Z. and J.J.; methodology, J.J. and Y.C.; software, Y.C.; validation, Y.Z. and Y.D.; formal analysis, Y.C. and J.J.; investigation, J.J., Y.Z. and Y.D.; resources, H.W.; data curation, Q.Z. and Y.C.; writing—original draft preparation, J.J. and Y.C.; writing—review and editing, H.W., Q.Z. and L.T.; visualization, Y.Z.; supervision, X.Z.; project administration, X.Z.; funding acquisition, X.Z. and L.T. All authors have read and agreed to the published version of the manuscript.

Corresponding author

Correspondence to Xiaofeng Zhang.

Ethics declarations

Ethics approval and consent to participate

The study was conducted in accordance with the Declaration of Helsinki. Human ethical approval was obtained from the Clinical Research Ethics Committee of the First Affiliated Hospital of Zhejiang University School of Medicine (No. 20220366B). The animal ethical approval was confirmed by the ethics board of the Laboratory Animal Center of Zhejiang University School of Medicine (No. 24905). Informed consent was obtained from all the patients.

Competing interests

The authors declare no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Jiang, J., Chen, Y., Zheng, Y. et al. Sialic acid metabolism-based classification reveals novel metabolic subtypes with distinct characteristics of tumor microenvironment and clinical outcomes in gastric cancer. Cancer Cell Int 25, 61 (2025). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12935-025-03695-0

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12935-025-03695-0

Keywords

Cancer Cell International

ISSN: 1475-2867

Contact us