KLHL4 is part of a family of 42 proteins, each characterized by a BTB/POZ domain at the N-terminus, a BACK domain in the middle and 5–6 Kelch domains at the C-terminus. Most KLHL proteins associate with Cullin 3 to form a Cullin-E3 ubiquitin ligase complex, acting as adapters that recognize target proteins via the Kelch domains during ubiquitination. These proteins are crucial for various cellular processes, including cytoskeletal organization, ion channel gating, transcriptional suppression and protein targeting for ubiquitination (20). Furthermore, the KLHL4 gene has also been linked to the synthesis and transport of long-chain fatty acids (Fig. 4A; pink cluster), a process implicated in diabetes and heart diseases. However, the mechanism of fatty acid entry into cells remains poorly understood and is thought to involve protein-mediated transport (21). KLHL4 is associated with the kinesin superfamily proteins (KIFs) family (Fig. 4A; light green cluster), which is essential for intracellular transport and fundamental for cellular function, survival and tissue morphogenesis. KIFs act as molecular motors that directionally transport cargo, including organelles, protein complexes and mRNAs, and play crucial roles in tumor suppression (22).

The nuclear factor erythroid 2-related factor 2 (NRF2) is a transcription factor that regulates the cellular antioxidant response and strongly correlates with KLHL4 expression (Fig. 4A; yellow cluster). NRF2 regulates genes that protect cells against oxidative stress, which is significant in cancer. The primary regulator of NRF2 activity is its interaction with Kelch-like ECH-associated protein 1 (Keap1). Under normal conditions, Keap1 binds to NRF2, promoting its degradation, but oxidative stress disrupts this interaction, allowing NRF2 to activate protective genes (23). For instance, during the carcinogenesis of GC (24), oxidative stress promotes the transcription of genes that protect against oxidative and electrophilic stress (25–27). In addition, NRF2 expression is associated with tumor-associated macrophages (TAMs) M2 polarization, which is well-known for exerting a pro-tumorigenic effect. TAMs are critical in the tumor microenvironment (TME) for eliminating tumor cells by creating a toxic environment. Polarization is linked to the NRF2 target protein Cu/Zn-superoxide dismutase, associated with M2 polarization through a redox-sensitive mechanism. Oxidative stress and reactive oxygen species are vital for M2 macrophage activation, promoting an immunosuppressive TME in GC (28–30). Therefore, pharmacological activation of NRF2 is a promising therapeutic approach for chronic diseases underlined by oxidative stress and inflammation (31) (Fig. 4A; yellow cluster).

Based on the TCGA-STAD data, EO-GC cases exhibited a higher KLHL4 gene expression than L-GC (Fig. 4Ba), principally in the 21–40-year age range (Fig. 4Bb). Concerning the methylation KLHL4 score, no differences between EO-GC and L-GC were observed (Fig. 4Ca), even when subjects were categorized by age (Fig. 4Cb). Lastly, the Kaplan-Meier analysis, using the median KLHL4 gene expression showed a non-significant tendency where a high expression of KLHL4 would be associated with lower survival of patients with EO-GC from the TCGA-STAD dataset (Fig. 4D). The expression data for KLHL4 in Fig. 4 highlight its significantly higher expression in EO-GC compared to L-GC. This observation aligns with the Kaplan-Meier survival analysis, which, although not statistically significant, suggests a trend where higher KLHL4 expression is associated with poorer survival.

In summary, the upregulation of KLHL4 in EO-GC suggests its potential involvement in enhancing intracellular transport and oxidative stress responses, both of which may contribute to the aggressive clinical phenotype observed in these patients. These processes likely shape the TME by supporting immune evasion and promoting tumor cell survival, thereby linking increased KLHL4 expression to poorer prognosis.

Following the identification of KLHL4, MAGEL2, another gene found to be upregulated in EO-GC, was next examined. MAGEL2 acts as a tissue-specific regulator of the retromer-dependent endosomal protein recycling pathway, important for secretory granule formation and maturation (32). The retromer complex, composed of VPS26, VPS29 and VPS35, facilitates the recycling of proteins from the endocytic pathway back to the plasma membrane and is critical in secretion regulation (33).

In addition to its role in the endosomal recycling pathway, MAGEL2 is involved in ubiquitination, interacting with and stimulating E3 RING ubiquitin ligases (34,35). This interaction highlights its significant role in ubiquitin processes.

Data mining of the TCGA-STAD platform revealed an overexpression of MAGEL2 in GC, with associations to the retromer multimeric protein complex and ubiquitination system (Fig. 5A; red cluster).

MAGEL2 additionally exhibits interactions with the structural maintenance of chromosome (SMC) protein complexes, which play crucial roles in chromatin structure reorganization, chromosome segregation and DNA repair. The interaction between SMC5-SMC6 proteins and MAGEL2, identified in the TCGA-STAD database, suggests a potential role in DNA double-strand break repair through homologous recombination in patients with GC (36) (Fig. 5A; yellow cluster). Further protein-protein association networks and functional enrichment analyses using the Search Tool for the Retrieval of Interacting Genes and proteins (STRING; http://string-db.org/) analysis revealed that MAGEL2 interacts with various transcription factors and RNA-binding proteins, particularly those involved in mRNA metabolic processes. Notably, MAGEL2 co-immunoprecipitates with YTHDF2, reducing its nuclear accumulation after heat shock (37) (Fig. 5A; green cluster).

Based on the TCGA-STAD data, EO-GC cases exhibited significantly higher MAGEL2 gene expression than L-GC (Fig. 5Ba); this difference was not as evident when patients were categorized by age (Fig. 5Bb). Concerning the methylation score of MAGEL2, no statistically significant differences were observed between EO-GC and L-GC (Fig. 5Ca); however, a non-significant decreasing trend in methylation levels was noted with increasing patient age (Fig. 5Cb). Lastly, the Kaplan-Meier survival analysis demonstrated that higher MAGEL2 gene expression is associated with low survival in EO-GC (Fig. 5D). This comprehensive analysis underscores the importance of MAGEL2 in EO-GC, highlighting its overexpression and involvement in critical cellular pathways and carcinogenic processes.

The expression patterns of MAGEL2 in Fig. 5 reveal its upregulation in EO-GC, particularly among younger age groups. The Kaplan-Meier analysis indicated that elevated MAGEL2 expression is associated with unfavorable survival outcomes. This suggests that MAGEL2′s role in protein recycling and chromatin remodeling may contribute to tumor progression and poor prognosis in patients with EO-GC.

In summary, the overexpression of MAGEL2 in EO-GC may exacerbate disruptions in protein recycling and chromatin structure, leading to cellular dysfunctions that support cancer progression. Its role in endosomal recycling and ubiquitination highlights its potential as a mediator of poor prognosis through tumor-promoting pathways.

Data mining of KLHL4 and MAGEL2 revealed an association with RNLS, a gene involved in oxidative stress and immune regulation. RNLS is an FAD-dependent amine oxidase that metabolizes water-soluble vitamins and nicotine. This enzymatic hormone, secreted by the kidneys and circulating in the bloodstream, oxidizes the less abundant forms of 1,2-dihydro-beta-NAD(P) and 1,6-dihydro-beta-NAD(P) to beta-NAD(P) (+) (38,39). RNLS impacts various cell types, suggesting similar roles in cancer. Its transcript levels are increased in pancreatic cancer, melanoma and other malignancies. In these cancers, higher RNLS levels are associated with shorter survival (40). Treatment with anti-RNLS antibodies at the single-cell level resulted in increased tumor density of macrophages, neutrophils and lymphocytes and increased expression of IFN-γ and granzyme B in natural killer cells and T cells in murine melanoma models (41). The presence of RNLS in both cancer and immune cells suggests that multiple cell types may contribute to the effects of RNLS on cancer cell growth (42,43).

STRING analysis, as a functional protein association network, showed a remarkable association between RNLS and divalent cation transporting channels, such as ATPase plasma membrane Ca²⁺ transporting 4 (ATP2B4) (Fig. 6A; green cluster), which facilitates divalent cation transport, and ATP4B, which is crucial for gastric acid secretion (44). RNLS also has a significant relationship with zinc finger protein 148 (ZNF148), a member of the Kruppel family of zinc finger DNA-binding proteins (Fig. 6A; green cluster). Increased ZNF148 expression has been linked to lower survival in colorectal cancer. ZNF148 influences the expression of multiple matrix metalloproteinases, which have protective and damaging effects during inflammation and are crucial for health maintenance (45–48).

ZNF148 protein directly engages with two transcription factors, STAT3 and SP1, which control gene transcription. It also interacts with several histone-coding genes, including H1-4, H2AC8 and H4C6. Furthermore, ZNF148 is associated with genes that contribute to desmosome formation, such as desmoplakin, filaggrin, hornerin and Annexin A2 (ANXA2). ANXA2 encodes a member of the annexin family, which includes calcium-dependent phospholipid-binding proteins that play roles in cellular growth regulation and signal transduction pathways (Fig. 6A; red cluster).

Another group of genes related to RNLS (Fig. 6A; yellow cluster) includes proteins like galectin 7B (LGALS7B), which are involved in cell-cell and cell-matrix interactions necessary for normal growth control. LGALS7B has a tumor-suppressive function, with gene down-regulation in GC (49). Stratifin is an adapter protein regulating general and specialized signaling pathways, playing a significant role in cell proliferation and metastasis in GC (50,51).

Based on the TCGA-STAD data, EO-GC cases exhibited a higher RNLS gene expression than L-GC (Fig. 6Ba), principally in the 41–50-year age range (Fig. 6Bb). Concerning the methylation score of RNLS, an insignificant increase in the methylation score was found in L-GC (Fig. 6Ca), but certain changes in methylation levels were observed in the 41–50-year age range (Fig. 6Cb). Lastly, the Kaplan-Meier survival analysis demonstrated that higher RNLS gene expression showed an insignificant trend toward lower overall survival in patients with EO-GC (Fig. 6D).

Fig. 6 demonstrates the increased expression of RNLS in EO-GC, particularly in the 41–50-year age group. The Kaplan-Meier analysis further shows a trend where higher RNLS expression is linked to reduced survival, emphasizing its potential involvement in oxidative stress regulation and immune evasion, key factors in EO-GC progression.

In summary, the increased expression of RNLS in EO-GC highlights its role in oxidative stress regulation and immune modulation. These functions may contribute to the establishment of an immunosuppressive TME, promoting cancer cell survival and explaining its association with poor prognosis.

In addition to KLHL4, MAGEL2 and RNLS, cytochrome P450 family 8 subfamily B member 1 (CYP8B1) was also identified to be upregulated in EO-GC. This gene's role in bile acid metabolism and steroid hormone synthesis suggests a potential connection between metabolic reprogramming and EO-GC progression.

CYP8B1 plays a crucial role in metabolic pathways such as steroid hormone synthesis, bile acid metabolism, cholesterol metabolism and lipid homeostasis. In steroid hormone synthesis, CYP8B1 collaborates with enzymes such as hydroxy-delta-5-steroid dehydrogenase, 3 beta- and steroid delta-isomerase 1,2,7 (HSD3B1, HSB3B2 and HSB3B7) (52,53) (Fig. 7A; yellow cluster). L-GC has been linked to androgens, estrogens, progesterone, their receptors and related signals (54). Cellular responses to steroid hormones, including ESR1, ESR2 and AR, are facilitated by hormone-receptor binding. ESR1 has been implicated in the cancer-promoting effects of estrogen in various cancers, including breast, colon, prostate and gastric cells. However, the association between these receptors and GC has yielded inconsistent results in several studies (55–57).

Additionally, the CYP8B1 pathway produces bile acids, which serve as potent signaling molecules that influence various metabolic processes, such as lipid homeostasis, glucose regulation and microbiota composition (Fig. 7A; red cluster). These bile acids are synthesized from cholesterol in the liver through the action of key enzymes, including CYP8B1 and CYP7A1, which are transcriptionally regulated by NR1H4 (nuclear receptor subfamily 1 group H member 4), a nuclear receptor also known as the bile acid receptor. Lastly, research has demonstrated a correlation between the presence of gastric intestinal metaplasia and an increased risk of gastric cancer, particularly for the intestinal subtype, which follows a well-established carcinogenic cascade (58). A retrospective study showed that high levels of bile acids in the stomach were associated with a higher incidence of GC (59,60).

Analysis with STRING revealed a direct relationship between CYP8B1 and hepatocyte nuclear factor 4α (HNF4α) involving bile acid homeostasis (Fig. 7A; blue circle). HNF4α is a transcription factor that binds DNA as a homodimer and regulates genes preferentially expressed in the liver. It plays a central role in bile acid homeostasis by controlling genes involved in bile acid biosynthesis, including hydroxylation and beta-oxidation of the cholesterol side chain in vivo (61).

Using TCGA-STAD data, it was verified that CYP8B1 transcript expression in EO-GC and L-GC individuals did not exhibit any statistically significant differences (Fig. 7Ba), nor was any change observed when patients were categorized by age (Fig. 7Bb). The methylation levels of the gene did not show any significant differences between both groups of patients (Fig. 7Ca). Subsequently, when subjects were categorized into different age groups, a statistically significant difference in the CYP8B1 methylation score between the 51–60-year age range and the 61–80-year age range was found (Fig. 7Cb). Lastly, Kaplan-Meier survival analysis revealed a non-significant trend toward better overall survival in patients with EO-GC with lower CYP8B1 gene expression (Fig. 7D).

The data in Fig. 7 show no significant differences in CYP8B1 expression between EO-GC and L-GC. However, the Kaplan-Meier analysis indicated a trend where lower CYP8B1 expression is associated with better survival outcomes. This finding highlights its potential role in EO-GC pathogenesis.

In summary, the role of CYP8B1 in bile acid metabolism and lipid homeostasis suggests that its upregulation in EO-GC could contribute to metabolic reprogramming in tumor cells. By influencing bile acid signaling and microbiota composition, CYP8B1 may drive gastric carcinogenesis and affect patient outcomes.

Tight junctions (TJ) are critical for the functioning of epithelial and endothelial cells, maintaining cell polarity, adhesion and permeability. Reduced TJ integrity leads to increased tissue permeability, a characteristic of tumors and inflamed tissues. During the initial stage of tumor metastasis, the disconnection between tumor and endothelial cells makes the TJ the first barrier cancer cells must overcome in metastasis (62).

TJs comprise three essential membrane proteins: Occludin, claudin and junctional adhesion molecules. The CLDN family is vital for TJ functions, including regulating defense and barrier functions, differentiation and polarity in epithelial and endothelial cells (Fig. 8A; blue cluster). The loss of CLDNs contributes to the disruption of cell junctions in a tissue-dependent manner and plays an essential role in cancer cell migration, invasion and metastasis (63). The distribution patterns of various claudins in GC differ between tumor tissue and adjacent tissue (64–66). Specifically, CLDN6 expression is higher in GC tissues than in adjacent tissues (67). However, certain studies suggest that lower levels of CLDN6 expression in GC tissues compared to adjacent tissues are associated with factors such as age, lymph node metastasis, pathological stage and tumor metastasis. Furthermore, several studies have reported that the upregulation of CLDN6 expression is associated with decreased survival rates in GC (68–71). Known for its role in TJ integrity, its reduced expression may contribute to increased tissue permeability and metastasis in EO-GC.

Analysis of protein interactions using STRING revealed an association between CLDN6 and proteins implicated in mesenchymal cell proliferation (72) (Fig. 8A; green cluster). Another group of proteins interacting with CLDN6 includes mitofusin proteins and mitochondrial outer membrane GTPases mediating mitochondrial clustering and fusion (Fig. 8A; red cluster). CLDN6 interacted with dynamins (DNMs), which catalyze the hydrolysis of GTP and utilize this energy to mediate vesicle scission (Fig. 8A; green cluster). These proteins participate in various forms of endocytosis, including clathrin-mediated synaptic vesicle and rapid endocytosis (73). Also, DNM2 is part of the machinery responsible for vesicle formation and regulates the cytoskeleton, facilitating intracellular vesicle transport (74).

Based on the TCGA-STAD data, no general differences in CLDN6 gene expression were seen between EO-GC and L-GC (Fig. 8Ba); there was also no significant difference in protein levels when patients were categorized by age (Fig. 8Bb). In terms of the gene methylation levels, no significant changes were observed when comparing both groups of patients (Fig. 8Ca). However, the 21–40-year age range exhibited a significantly higher methylation score than the other age ranges (Fig. 8Cb). Kaplan-Meier survival analysis did not indicate any survival differences between EO-GC and L-GC according to the median CLDN6 expression (Fig. 8D).

In summary, the downregulation of CLDN6 in EO-GC suggests a weakening of TJ integrity, potentially facilitating cancer cell invasion and metastasis. This highlights the biological importance of CLDN6 in maintaining epithelial barriers and its potential role as a prognostic biomarker in EO-GC.

Following the identification of CLDN6, MIOX gene expression was next examined. The initial committed step in mammalian inositol catabolism is catalyzed by MIOX, which performs the unique four-electron dioxygen-dependent ring cleavage of myo-inositol to D-glucuronate. This enzyme facilitates the binding of ferric iron and inositol oxygenase activity, playing a significant role in the inositol catabolic process, primarily located in the cytoplasm and inclusion bodies (75).

STRING analysis involving the MIOX gene revealed a direct relationship with uridine 5′-diphosphate-glucuronosyltransferase (UGT) genes (Fig. 9A; red cluster), which are membrane proteins of the endoplasmic reticulum expressed in a tissue-specific manner. MIOX has been identified as a regulatory gene of tumor ferroptosis in several cancer types, such as clear cell renal cell carcinoma (ccRCC). In ccRCC, a significant downregulation of MIOX in tumor tissues relative to adjacent renal tissues has been observed, with a negative correlation between MIOX expression levels in ccRCC tissues and the malignant behavior, as well as poor prognosis of ccRCC (76).

Additionally, MIOX has been implicated in bladder (77) and prostate cancer progression (78) and lung squamous cell carcinoma, where it is part of a gene signature indicative of the connection with glycolysis (79). Furthermore, it has been established that certain gene isotypes of the UGT family have differential expressions between normal and tumor stomach tissue, whose expression changes would affect the progression of GC (80). However, no relationship between MIOX and GC, particularly EO-GC, has been reported.

STRING analysis identified a gene cluster, including the ABCG2 transporter and two UGT genes (Fig. 9A; green cluster). Additionally, relationships were observed with genes involved in ATP-dependent ABC-type transporters (Fig. 9A; yellow cluster). A connection was also established between two genes related to E3 ubiquitin ligases (Fig. 9A; blue cluster).

Based on the TCGA-STAD dataset, it was found that EO-GC exhibited a significantly lower MIOX gene expression and a relative increase in gene expression in L-GC (Fig. 9Ba); this was confirmed by an increase in transcripts in older age groups, between 51 and 100 years (Fig. 9Bb). Furthermore, EO-GC displayed a higher methylation score than L-GC (Fig. 9Ca), particularly in the 21–40-year age range (Fig. 9Cb). Kaplan-Meier survival analysis did not indicate any association of MIOX expression with survival in patients with EO-GC using the median MIOX expression level as a cut-off (Fig. 9D).

The lower expression of MIOX in EO-GC (Fig. 9) suggests that an alteration in inositol metabolism may contribute to tumor progression. While Kaplan-Meier analysis did not show a significant difference in survival, the methylation changes observed in EO-GC warrant further studies on the role of MIOX in metabolic reprogramming and the regulation of ferroptosis. In summary, the significant downregulation of MIOX in EO-GC suggests a disruption in inositol metabolism, potentially impairing ferroptosis-a cell death pathway crucial for tumor suppression. This alteration may create metabolic vulnerabilities that cancer cells exploit for progression.

PNMA5 gene, also downregulated in EO-GC, has been implicated in apoptosis and cancer metastasis. The PNMA family members have been identified as onconeural antigens exhibiting aberrant expression in cancer cells in patients with paraneoplastic disorders. This protein family is closely associated with autoimmunity, neurodegeneration and cancer, with several PNMA family members characterized by their involvement in apoptosis and cancer-related signaling pathways (81). Studies have shown that PNMA5 is deregulated in patients with CRC; it contributes to CRC metastasis by potentially facilitating cancer cell migration and invasion. Cellular markers related to epithelial-mesenchymal transition (EMT) revealed that PNMA5 promotes EMT in CRC, promoting cell migration and invasion (82,83). Given that metastases are more detrimental to cancer-associated mortality than primary tumors, understanding the role of PNMA5 in EMT and metastasis in GC is crucial for developing targeted therapies.

In particular, PNMA5 is associated with ZBTB8A, which facilitates DNA-binding activity specific to RNA polymerase II transcription regulatory regions, potentially playing a role in transcriptional regulation. Another noteworthy gene is ARPC4, identified as a potential biomarker or drug target in metastatic GC (Fig. 10A; green cluster). In addition, a group of PNMA5-related genes is involved in DNA replication, single-strand DNA binding, repair and homologous recombination (Fig. 10A; red cluster). Two smaller clusters relate to genes regulating transcription (Fig. 10A; blue cluster). Furthermore, PNMA5 is related to genes that initiate transcription, such as GTF3C1 (Fig. 10A; yellow cluster), which activates polymerases to initiate gene transcription (84–86).

Based on the TCGA-STAD dataset, it was found that EO-GC exhibited a lower PNMA5 gene expression than L-GC (Fig. 10Ba). A tendency toward increased transcript levels was observed in older age groups, particularly between 51 and 80 years of age, although no significant differences were detected across groups (Fig. 10Bb). Gene methylation levels did not show any significant differences when only the two groups of EO-GC vs. L-GC were compared (Fig. 10Ca). Regarding methylation, EO-GC displayed a higher PNMA5 methylation score, especially in the 21–40-year age group. While differences between several age groups were observed (Fig. 10Cb). Kaplan-Meier survival analysis indicated a non-significant trend toward lower survival among EO-GC patients with higher PNMA5 gene expression (Fig. 10D).

Based on the TCGA-STAD dataset, PNMA5 gene expression was significantly lower in EO-GC compared to L-GC. However, when patients were stratified into age groups, this difference was not statistically significant, indicating that the observed expression difference is more clearly captured when using a binary classification (EO-GC vs. L-GC) rather than categorical age groupings. Kaplan-Meier survival analysis showed a tendency for poorer outcomes in patients with higher PNMA5 expression, suggesting its potential involvement in EMT and cancer metastasis.

In summary, the downregulation of PNMA5 in EO-GC may reflect alterations in apoptosis and EMT, processes critical for cancer metastasis. Its association with EMT markers highlights its potential role in driving invasive behavior and poor prognosis in patients with EO-GC.

Lastly, ACTL8 was also down-regulated in EO-GC compared to L-GC cases. ACTL8 has been implicated in the differentiation of epithelial cells and is thought to be located within the cytoplasm of the dynactin complex, facilitating the activation of the dynein molecular motor for ultra-processive transport along microtubules (Fig. 11A; green cluster). The dynactin complex, a critical component of the ARP2/3 complex, is crucial for cell shape and movement by forming actin filaments on the lamellipodial cell surface (87). Furthermore, the ARP2/3 complex plays a role in the cytoplasmic cytoskeleton by promoting actin polymerization in the nucleus, which regulates gene transcription and DNA repair (88).

The relationship between this cluster network and histone-related genes, including EP400, H4C6 and MYSM1, is noteworthy (Fig. 11A; red cluster). These genes involve essential processes, such as chromatin remodeling (89). Furthermore, high ACTL8 expression has been associated with poor prognosis in head and neck cancer (90) and metastasis in CRC (91).

Based on the TCGA-STAD data, it was found that EO-GC exhibited a lower PNMA5 gene expression as compared with L-GC (Fig. 11Ba), particularly in the 21–40-year age range (Fig. 11Bb). Unfortunately, no methylation data for ACTL8 in GC were found. Lastly, Kaplan-Meier survival analysis based on the median ACTL8 gene expression did not indicate any significant influence of ACTL8 on survival outcomes in EO-GC cases (Fig. 11C).

The expression data for ACTL8 in Fig. 11 highlight its downregulation in EO-GC compared to L-GC, particularly in younger patients. Although the Kaplan-Meier analysis did not reveal any significant survival association, ACTL8′s role in cytoskeletal organization and chromatin remodeling suggests its potential importance in EO-GC pathogenesis.

In summary, the downregulation of ACTL8 in EO-GC suggests impaired cytoskeletal dynamics and chromatin remodeling. These disruptions may hinder normal cell differentiation while facilitating cancer cell motility, contributing to EO-GC progression.