Under oxygen-sufficient conditions, normal cells primarily satisfy their energy requirements through the sequential biological processes of 'glucose uptake → glycolysis → pyruvate entry into the tricarboxylic acid (TCA) cycle → OXPHOS'. By contrast, tumor cells preferentially utilize aerobic glycolysis, consuming large amounts of glucose and converting pyruvate into lactate, yielding a net gain of only two ATP molecules per glucose molecule (27). This metabolic shift is partly driven by mitochondrial reprogramming, wherein tumor cells maintain TCA cycle activity to support biosynthetic precursor synthesis, partially compensating for the reduced ATP yield from OXPHOS (28).

Glycolytic reprogramming occurs predominantly within the cytosol. In highly migratory tumor cells, glycolytic enzymes are recruited to the cell cortex, where they propagate as self-organizing waves along the membrane/cortex and interact with actin filaments (29). Key regulatory enzymes, including hexokinase (HK), pyruvate dehydrogenase kinase (PDK), phosphofructokinase-1 (PFK1) and the pyruvate kinase M2 (PKM2) isoform, orchestrate a series of catalytic reactions (30). These protein-protein interactions among metabolic enzymes lead to altered levels of glucose-metabolizing proteins and glycolytic intermediates such as fructose, glyceraldehyde and pyruvate. Most pyruvate is subsequently converted into lactate, which is exported extracellularly via monocarboxylate transporters (MCTs). Extracellular lactate accumulation further drives histone lysine lactylation, influencing transcriptional regulation, post-transcriptional modifications and post-translational processes in tumor cells (31). By accelerating glycolytic flux, tumor cells rapidly generate energy, thereby promoting proliferation, invasion, metastasis and therapy resistance, a mechanism critically involved in GC pathogenesis (32,33). Consequently, targeting key glycolytic enzymes represents a promising strategy to suppress the Warburg effect and impede GC progression (Fig. 2A).

The HK family comprises four isoforms (HK1-4), each encoded by distinct genes and exhibiting unique functional and distribution profiles. Among these, HK2 is upregulated in most tumor cells, including GC cells, and serves as the key rate-limiting enzyme catalyzing the first committed step of glycolysis (34). HK2 transfers a phosphate group from ATP to glucose, generating glucose-6-phosphate (G6P), which provides a critical metabolic intermediate for downstream reactions (35). PFK1, a major regulatory node in the glycolytic pathway, acts as another essential rate-limiting enzyme that directs G6P flux into glycolysis; it catalyzes the conversion of fructose-6-phosphate to fructose-1,6-bisphosphate. Under conditions of high energy demand or hypoxia, PFK1 activity is enhanced, accelerating G6P consumption and favoring glycolytic ATP production (36). The PDK family also consists of four isoforms (PDK1-4). Elevated expression of PDK isoforms in GC cells promotes S-phase entry and serves as an independent prognostic factor in patients with GC (37). PDK plays a decisive role in directing pyruvate metabolism, where it phosphorylates and inactivates pyruvate dehydrogenase, thereby preventing the conversion of pyruvate to acetyl-CoA and entry into the TCA cycle. This suppression of mitochondrial OXPHOS shifts metabolism toward anaerobic glycolysis, leading to lactate accumulation. Concomitantly, altered histone acetylation patterns and suppressed programmed cell death-ligand 1 (PD-L1) expression contribute to immune evasion, further enhancing the pro-tumorigenic effects in GC (38). Pyruvate kinase, the terminal rate-limiting enzyme of glycolysis, transfers a high-energy phosphate group from phosphoenolpyruvate to ADP, generating pyruvate. The metabolic fate of pyruvate is closely linked to the oligomeric state of the predominant tumor isoform PKM2. In its tetrameric form, PKM2 exhibits high catalytic activity, facilitating pyruvate conversion to acetyl-CoA and supporting sustained energy production via OXPHOS. By contrast, the dimeric form of PKM2 displays reduced activity, leading to the accumulation of glycolytic intermediates such as phosphoribosyl pyrophosphate and 3-phosphoglycerate. This metabolic configuration prioritizes biosynthetic precursor supply, thereby supporting rapid tumor proliferation and metastasis (39,40).

Fatty acids (FAs), essential components of various cellular structures, play critical roles in biological processes such as membrane biogenesis and energy provision (41). Mitochondrial β-oxidation serves as the core pathway for FA breakdown and energy production. However, long-chain FAs cannot freely traverse the mitochondrial membrane. Instead, they are first converted to acyl-CoA in the cytosol and then transported into the mitochondrial matrix via the carnitine palmitoyltransferase (CPT) system located on the inner mitochondrial membrane (42). Subsequently, through repeated cycles of dehydrogenation, hydration and thiolysis, FAs are degraded to generate acetyl-CoA, NADH and NADPH, which enter the TCA cycle to produce ATP (42). To rapidly adapt to hypoxic and nutrient-deprived microenvironments, tumor cells reprogram lipid metabolism by modulating FA uptake, synthesis, storage and β-oxidation. They utilize lipids for membrane construction, to support progression, and upregulate factors such as CD36 in regulatory T cells and CD8⁺ T cells. This enhances lipid uptake and alters immune cell recruitment, activation and function, thereby facilitating immune evasion and accelerating tumor proliferation and metastasis (43,44) (Fig. 2B).

The transport of long-chain FAs into cells is a critical step in lipid metabolic reprogramming, primarily mediated by transport proteins. The transmembrane glycoprotein CD36, expressed on various cells within the TME, including endothelial, stromal and immune cells, facilitates long-chain FA uptake via endocytosis (45). Internalized FAs can be directed to lipid droplets for storage or undergo β-oxidation. This process involves activation of the Lck/Yes novel tyrosine kinase, which phosphorylates and inactivates zinc finger DHHC-type palmitoyltransferase 5 at tyrosine 91. Subsequently, acyl-protein thioesterase 1 mediates CD36 depalmitoylation, initiating a new cycle of CD36-mediated endocytosis and FA transport. This dynamic palmitoylation-depalmitoylation cycle of CD36 forms a closed loop crucial for remodeling the tumor lipidome and generating oncogenic lipid species, thereby promoting tumor growth and progression (46). Furthermore, FAs can upregulate O-GlcNAcylation, which in turn modifies CD36 and enhances its transcriptional expression. This positive feedback loop increases FA uptake in GC cells, accelerates metastasis and is closely associated with poor prognosis in patients with GC (47). While transporter-dependent pathways are important, they are not the sole mechanism for long-chain FA uptake. The acidic TME can neutralize negatively charged FAs, promoting their protonation and passive diffusion into cells. This alternative pathway further drives FA accumulation and lipid droplet formation, supporting cancer cell metabolism and rapid proliferation (48).

The CPT system, a key regulator of FA β-oxidation, is central to lipid metabolic reprogramming. It consists of two isoforms: i) CPT1, located on the outer mitochondrial membrane and serving as the primary rate-limiting enzyme for FA translocation into mitochondria (49); and ii) CPT2, residing on the inner mitochondrial membrane, which facilitates the entry of FAs into the mitochondrial matrix for β-oxidation (50). In GC, CPT1A is highly expressed and interacts with S100A10, promoting its succinylation at lysine 47 (51). CPT1A expression also correlates with CD44, a key regulator of cancer metastasis. CD44 activates intercellular signaling pathways that target CPT1A to modulate FA oxidation (FAO), thereby enhancing the invasive and metastatic potential of GC cells (51,52). Similarly, CPT2 is notably upregulated in GC cells (53). Nuclear factor of activated T cells c3, a member of the NFAT transcription factor family, can be activated by reactive oxygen species (ROS) and translocate to the nucleus. There, it binds to the CPT2 promoter to enhance its transcription, elevating β-oxidation levels. This increase in FA catabolism raises the NADPH/NADP⁺ ratio, protecting GC cells from ROS-induced damage and maintaining cellular homeostasis (53).

Through metabolic reprogramming of amino acid networks, tumor cells continuously adapt to survival pressures. Glutamine (Gln), the most abundant free amino acid in systemic circulation, plays a particularly critical role. As tumor cells are unable to synthesize Gln de novo, they compete with normal cells for exogenous Gln, developing a metabolic dependency to sustain proliferation and metastasis (54). Given its hydrophilic nature, Gln requires specific solute carriers, such as solute carrier family 1 member 5 (SLC1A5; also known as ASCT2), to traverse the plasma membrane. Once intracellular, Gln is sequentially deamidated by glutaminase (GLS) and glutamate dehydrogenase to produce glutamate (Glu), which is further converted to α-ketoglutarate (α-KG). This metabolite replenishes the TCA cycle, provides biosynthetic precursors, enhances mitochondrial activity and ultimately fuels cancer proliferation (55-57).

Due to their rapid energy consumption, tumor cells frequently face glucose starvation in the TME. To address this, they reconfigure intrinsic metabolic pathways to maintain NADPH homeostasis, thereby meeting the high energy demands and preserving the NADPH/NADP⁺ balance (58). In GC, upregulation of malic enzyme 1 (ME1) helps replenish intracellular NADPH pools and maintain redox homeostasis. Conversely, ME1 knockdown depletes NADPH, lowers the glutathione (GSH)/GSH disulfide ratio, elevates ROS, exacerbates oxidative stress and promotes GC cell death (58,59). A portion of the Glu generated from Gln in mitochondria is transported back to the cytosol, where it serves as a substrate for GSH synthesis. Glu-cysteine ligase catalyzes the conjugation of Glu and cysteine to form γ-glutamylcysteine, which is subsequently combined with glycine by GSH synthase to produce GSH. This pathway assists tumor cells in countering oxidative stress (60-62) and modulates host immune surveillance and suppression. When Gln is extensively depleted by tumors, impaired GSH synthesis triggers Fas/CD95-mediated apoptosis. However, concurrent upregulation of PD-L1 expression can suppress CD8⁺ T cell antitumor activity. Thus, simultaneously targeting Gln deprivation and PD-L1 activity, by blocking PD-L1/PD-1 interactions, presents a promising clinical strategy to alleviate immunosuppression in the TME (63). The oncogenic transcription factor c-Myc further supports Gln metabolism by binding to E-box sequences in target gene promoters. This upregulates GLS expression, enhances Gln uptake and catabolism, supplies TCA cycle intermediates, promotes nucleotide synthesis and sustains the malignant proliferative phenotype of tumors (64,65) (Fig. 2C).

In addition to Gln, other amino acids contribute to metabolic remodeling in GC, enabling tumor cells to compete for energy resources and support disease progression. Branched-chain amino acids (BCAAs), such as leucine, isoleucine and valine, are non-polar aliphatic amino acids characterized by carboxyl groups, amino groups and isopropyl side chains. They can be directly incorporated into proteins or catabolized into various metabolites to fuel related metabolic pathways (66). BCAAs enter cells via L-type amino acid transporters (LATs) and are shuttled into mitochondria by SLC25A44. Within mitochondria, they are transaminated to branched-chain α-keto acids (BCKAs), which subsequently transfer amino groups to α-KG. BCKAs are further metabolized to acetyl-CoA and succinyl-CoA, while generating Glu. The resulting metabolites ultimately enter the TCA cycle to support energy production (67).

In addition, an inter-amino acid compensatory mechanism exists. Under Gln-deficient conditions, elevated expression of the PPM1K protein promotes BCAA catabolism to sustain tumor cell survival (68). Previous studies suggest that elevated circulating BCAA levels may be associated with several human cancer types (69,70). However, a large-scale prospective cohort study by Yu et al (71) reached a contrasting conclusion, indicating that higher BCAA levels correlate with a reduced risk of GC, particularly among elderly populations. Given this dual role of BCAAs in tumor development, further investigation is needed to determine whether these context-dependent effects stem from distinct components within BCAA catabolic pathways.

Nucleotides, composed of a nucleoside and a phosphate group, are essential for DNA and RNA synthesis, enzyme regulation, substrate activation and various metabolic processes in cells. Their synthesis is energy-intensive and requires carbon and nitrogen donors for support (72). This process relies on the cytosolic one-carbon (1C) metabolic pathway. To meet the demands of proliferation and metastasis, tumor cells upregulate the expression of enzymes related to 1C metabolism, altering their metabolic patterns to support the de novo synthesis of purines and pyrimidines (73). Pyrimidines and purines are the core structural components of nucleotides, and nucleotide synthesis involves two pathways: i) De novo synthesis; and ii) salvage pathways (74). In the de novo synthesis process, precursor molecules generated by various metabolic pathways are utilized. For example, Gln is deaminated by GLS to produce Glu, which is then converted by glutamate dehydrogenase, releasing ammonia in the process. In glucose metabolism, the pentose phosphate pathway produces G6P, which then undergoes dehydrogenation, decarboxylation and isomerization reactions to form ribose-5-phosphate. Together, these intermediates provide essential raw materials for nucleotide synthesis.

The salvage pathway allows cells to rapidly generate nucleoside monophosphates from free nucleobases or nucleosides through phosphorylation or phosphoribosyl transfer reactions. This process avoids base wastage, is relatively energy-efficient and helps replenish the nucleotide pool (75). Normal cells primarily rely on the salvage pathway for nucleotide synthesis. However, the aggressive behavior of tumor cells depends on higher nucleotide synthesis and metabolic activity, requiring faster and greater acquisition of nucleotides to support rapid proliferation, drug resistance, immune evasion and metastasis. Therefore, tumor cells mainly utilize the de novo synthesis pathway for nucleotide production (76). U2AF homology motif kinase 1 (UHMK1) is a nuclear-localized serine/threonine kinase and acts as an effective oncogenic factor. In GC, UHMK1 expression is upregulated; it can phosphorylate nuclear receptor coactivator 3 at serine 1062 and threonine 1067, promoting its interaction with activating transcription factor 4 (ATF4) and inducing the nuclear translocation of ATF4. This activates transcriptional activity, mediates oligomeric nucleotide synthesis pathways and simultaneously enhances the de novo purine synthesis pathway, providing a material basis for GC cell proliferation and driving GC progression (77,78) (Fig. 2D).

Cells can secrete nucleotides and nucleosides to initiate paracrine and autocrine signaling, thereby driving cellular metabolic networks, regulating immune responses and influencing the development and progression of tumor cells (79). Due to genomic mutations, the expression level of major histocompatibility complex (MHC) class I polypeptide-related sequence A (MICA) increases in tumors. MICA binds to the natural killer group 2 member D receptor, thereby triggering NK cell-mediated immune surveillance. This process is associated with the high levels of nucleotide metabolism observed in tumor cells (80,81). A study found that certain nucleotide metabolism-related genes are associated with levels of immune cell infiltration (82). Patients with GC exhibiting high nucleotide metabolism have poorer overall survival, disease-specific survival, disease-free interval and progression-free interval. Additionally, these patients show increased expression of immunosuppressive factors and MHC molecules (82).

Adenosine can receive a phosphate group under the catalysis of adenosine kinase, which is an important pathway for nucleotide synthesis. Conversely, nucleotides can be dephosphorylated by phosphatases to generate adenosine; this bidirectional conversion is a key node in intracellular purine metabolism (83). Adenosine receptors are present on the surface of various cell types including human dendritic cells (DCs), which mainly express A1 and A3 receptors (84). Adenosine can mediate the migration of DCs to lymphoid organs, promote antigen presentation and T cell activation, thereby driving immune regulation (85,86). Additionally, regulatory B cells can regulate their own function through adenosine signaling derived from ATP enzymatic degradation, maintaining an immunosuppressive cell phenotype and inhibiting T cell activation, thus participating in the regulation of adaptive immunity. Excessive adenosine signaling can impede immune surveillance and thereby promote tumor progression (84).

Early transcriptome studies mainly focused on the competing endogenous RNA (ceRNA) interactions between mRNA and long ncRNA. With advances in high-throughput sequencing and RNA interaction analysis technologies, it was discovered that circRNAs retain only the circularized regions of exons, introns or exon-intron sequences, and naturally lack a 5' cap and 3' poly(A) tail. This makes them more stable within the ceRNA network and enables them to have specific regulatory functions (87). circRNAs can bind to miRNA through sequence-complementary miRNA response elements, competitively inhibiting the degradation or translational repression of target mRNAs by miRNAs. Therefore, circRNAs regulate the expression level of target gene transcripts and are extensively involved in various cellular biological processes (88).

circDNMT1 can act as a molecular sponge for miR-576-3p, thereby relieving its inhibitory effect on the hypoxia-inducible factor 1-α (HIF-1α) 3' untranslated region (3'UTR). This accelerates glucose uptake in GC cells, increases lactate production, suppresses pyruvate and ATP generation and promotes tumor progression (89). Similarly, inhibition of circSLAMF6 can enhance the targeted regulation of myosin heavy chain 9 (MYH9) by miR-204-5p, thereby weakening the glycolytic capacity of GC cells under hypoxia and further suppressing tumor metastasis and invasion (90). As shown in Table I, numerous circRNAs have been reported to participate in GC glycolysis via the ceRNA network in previous years. The majority of circRNAs linked to GC are upregulated and act as miRNA sponges to abolish the inhibition of key glycolytic enzymes or transcription factors, consequently enhancing glucose uptake and lactate production as well as promoting GC proliferation, migration and invasion.

GC cells can recruit mesenchymal stem cells (MSCs) by secreting microvesicles. Under the combined induction of immune cells and Helicobacter pylori, these MSCs are transformed into GC-MSCs, which retain mesenchymal lineage and stem cell properties, possess multipotent differentiation potential and strongly promote GC cell proliferation, migration and angiogenesis (91). circ_0024107 derived from GC-MSCs can act as a molecular sponge for miR-5572 and miR-6855-5p, negatively regulating their expression, upregulating CPT1A mRNA and protein expression, participating in FA transport and activating FAO, thus promoting GC proliferation and metastasis (92). Additionally, circRNAs can enhance the expression of amino acid metabolism-related transporters by sponging miRNAs through sequence complementarity. A study has shown that circAKT3 targets miR-515-5p, relieving its suppression of SLC1A5 by binding to the SLC1A5 3'UTR, thereby increasing the expression of the SLC1A5 transporter and enhancing Gln uptake in GC cells, which in turn promotes malignant proliferation (93). circB3GNTL1 and circSFMBT2 can also participate in amino acid metabolism in GC by sponging miR-598 and miR-665, respectively (94,95). Although direct research focusing on circRNA-mediated regulation of nucleotide metabolic reprogramming in GC via the ceRNA network is limited, a study has found that hsa_circ_0008434 in GC can sponge miR-6838-5p, relieving its inhibition of ubiquitin-specific peptidase 9X (USP9X), resulting in USP9X protein accumulation and promoting GC cell proliferation, migration and invasion (96). USP9X, through multiple synergistic mechanisms, regulates the stability of Gln synthetase, increases nucleotide availability, supports nucleotide synthesis and stemness maintenance in tumor cells, and participates in tumor growth and proliferation (97).

The binding of circRNAs to RBPs relies on the complementary pairing or spatial recognition between specific binding sites within circRNA sequences and the RNA-binding domains of RBPs. This interaction forms distinct circRNA-RBP complexes, which regulate the activity, conformational changes and subcellular localization of target proteins, thereby modulating circRNA functions in tumor cells (98). Additionally, circRNAs can regulate the expression of key metabolic dependency enzymes or transporters and participate in multiple signaling pathways to directly or indirectly influence metabolic reprogramming in GC cells.

Polypyrimidine tract-binding protein 1 (PTBP1), a shuttling protein between the cytoplasm and nucleus with distinct functions in each compartment, is involved in tumor proliferation and invasion (99). circ_0001756 binds to PTBP1, facilitating its interaction with phosphoglycerate kinase 1 (PGK1) via the RRM1 domain. This interaction enhances the stability of PGK1 mRNA, promotes its cytoplasmic expression and participates in the first step of ATP production through the glycolytic pathway, thereby contributing to GC initiation and progression (100). Stearoyl-CoA desaturase 1 (SCD1), a key rate-limiting enzyme that desaturates saturated FAs to produce monounsaturated FAs, controls lipid metabolic flux. Aberrant transcriptional regulation and epigenetic modifications of SCD1 induce abnormal lipid accumulation, driving tumor progression (101). circTFRC directly binds to SCD1 mRNA, recruits ELAV-like RNA binding protein 1 and interacts with the 3'UTR of SCD1 mRNA. This protects SCD1 from nuclease degradation, enhances its stability and sustains SCD1 mRNA translation. Consequently, this promotes oncogenic lipid metabolic remodeling, thereby accelerating GC progression (102).

Rapidly dividing tumor cells exhibit methionine (Met) dependence due to their high demand for Met in processes such as histone modification, nuclear division and GSH synthesis (103,104). Met deprivation induces DNA damage in tumor cells, leading to proliferation arrest. A study has shown that eukaryotic initiation factor 4A3 (EIF4A3) binds to and interacts with the flanking sequences downstream of circ-CDK13. Met degradation downregulates EIF4A3 expression, restricting circ-CDK13 circularization, inhibiting GC cell proliferative activity, increasing intracellular cisplatin concentrations and enhancing chemosensitivity (105). The oncogenic transcription factor myelocytomatosis oncogene (MYC) has been extensively studied in metabolic contexts, with its expression consistently elevated across all stages of intestinal-type GC development (106,107). MYC provides sugar backbones for purine and pyrimidine biosynthesis and activates phosphoribosyl pyrophosphate synthetase 2 (PRPS2) via eukaryotic translation initiation factor 4E. Through specific cis-regulatory elements within its 5'UTR, PRPS2 couples protein and nucleotide biosynthesis (108), enabling tumor cells to reprogram nucleotide metabolic pathways to adapt to nutrient-deprived environments. Research has confirmed that circ-TNPO3 binds to the KH domain at the C-terminus of insulin-like growth factor 2 mRNA-binding protein 3 (IGF2BP3), impairing IGF2BP3-mediated stabilization of MYC mRNA and thereby inhibiting GC proliferation and metastasis (109). Additionally, circ-hnRNPU can also directly bind to non-POU domain-containing octamer-binding protein (NONO), inducing the cytoplasmic accumulation of NONO, thereby abolishing the activating effect of nuclear NONO on c-Myc and downregulating c-Myc expression (110). c-Myc is well established as a transcriptional activator of PRPS2, a rate-limiting enzyme in de novo purine and pyrimidine synthesis (108). Whether circ-hnRNPU-mediated c-Myc downregulation translates into altered nucleotide metabolic flux in GC cells, however, has not been investigated. Direct evidence linking circ-hnRNPU to changes in nucleotide synthesis rates, PRPS2 activity, or intracellular metabolite levels is currently absent, and this putative relationship awaits experimental validation.

Conventional translation initiation typically involves the recognition of the 5' cap by the small ribosomal subunit and its associated initiation factors. However, circRNAs lack this structure and cannot initiate translation through the canonical pathway. IRESs, which are structured RNA elements capable of directly recruiting the small ribosomal subunit and initiation factors, enable a subset of circRNAs to undergo cap-independent translation, yielding functional short peptides (111).

Emerging evidence indicates that peptides encoded by endogenous circRNAs play potential roles in tumorigenesis (112). For example, circ_0008035 encodes a novel peptide, EXT1-219aa, which suppresses EXT1 phosphorylation. This promotes the nuclear localization of PKM2 and reduces its EXT1-dependent glycosylation, thereby enhancing pyruvate metabolism and accelerating glycolysis in GC cells (113). By contrast, circUBE2G1 can encode circ_UBE2G1-99aa, which binds to enolase 1, thereby promoting glycolysis in GC (114). Similarly, circFNDC3B may encode a 267-amino-acid peptide that facilitates c-Myc protein degradation via the proteasomal pathway, inhibiting GC cell proliferation and migration (115). These findings underscore the importance of c-Myc in amino acid metabolic remodeling and suggest that circFNDC3B may participate in GC progression through this metabolic axis. c-Myc serves as a key transcriptional driver of PRPS2 in the de novo synthesis pathways of purines and pyrimidines (108). Whether the negative regulatory mechanism of FNDC3B-267aa on c-Myc (115) establishes a link between this peptide and nucleotide metabolism remains to be determined, as direct evidence demonstrating that FNDC3B-267aa alters PRPS2 activity, nucleotide precursor levels or pyrimidine/purine pools in GC cells is currently lacking.

To date, studies on circRNA-encoded micropeptides regulating metabolic reprogramming in GC remain relatively limited, with most focusing on glycolytic pathways. However, research on other solid tumors, such as glioblastoma and colorectal cancer, has revealed that circRNA-derived micropeptides influence malignant progression by modulating key processes such as nucleotide metabolism and FAO (112,116). Given the high conservation of these metabolic pathways in cancer, it is plausible that GC proliferation similarly depends on such mechanisms. For example, a novel 127-amino-acid micropeptide encoded by circSpdyA binds to the ER catalytic domain of FA synthase, activating de novo FA synthesis, enhancing FAO and upregulating CPT1A activity. This promotes the uptake of accumulated FAs to supply membrane biosynthesis and support tumor cell proliferation and metastasis in breast cancer (117). Another micropeptide, E2F1-99aa, is encoded by circE2F1, which contains an IRES and a 297-nucleotide open reading frame. E2F1-99aa mimics the DP-binding domain of E2F1, interacts with SPIB and competitively inhibits E2F1-SPIB binding. This intervention downregulates nucleotide biosynthetic gene expression, reduces purine and pyrimidine synthesis and consequently suppresses neuroblastoma progression (118). Future research should focus on identifying GC-specific circRNA-encoded micropeptides and validating whether their functions align with these conserved metabolic pathways.

m⁶A modification of circRNAs is regulated by methyltransferases, demethylases and reader proteins, primarily occurring at the N⁶ position of adenine residues (119). The covalently closed circular structure of circRNAs restricts the spatial distribution of modification sites, facilitating the formation of an 'aggregation effect' of m⁶A modifications. This effect enhances the recruitment efficiency of binding proteins and results in a more concentrated distribution of modification sites within the exonic sequences of circularized regions (120). Notably, m⁶A modification of circRNAs primarily reinforces and expands their non-coding functional repertoire; it enables the subcellular trafficking and localization of circRNAs between the nucleus and cytoplasm and regulates processes such as RNA processing, degradation, nuclear export and translation (121). By modulating the abundance and functional duration of circRNAs, m⁶A modification serves as one of the critical pathways mediating circRNA degradation (122), thereby further interfering with the metabolic homeostasis of tumor cells and contributing to tumor progression (123). However, in rare, specialized contexts, m⁶A modification can endow circRNAs with coding potential, whereby the m⁶A reader protein YTHDF3 recognizes m⁶A-modified circRNAs, recruits the translation initiation factor eIF4G2, and mediates ribosome assembly on circRNAs. By providing an IRES, this process drives the translation of circRNAs into short peptides (124). Notably, a single m⁶A modification is sufficient to initiate circRNA translation (125). In such cases, certain circRNA molecules temporarily acquire coding capacity upon m⁶A modification, which contributes to the complex regulatory networks underlying tumor initiation and progression.

Abnormal glycolytic flux is a key hallmark of bioenergetic metabolism in GC. circRNAs can regulate glycolytic pathways through epigenetic remodeling, while dynamic m⁶A epitranscriptomic modifications modulate the expression of glycolytic enzymes via methyltransferase/demethylase-mediated RNA methylation cycles. These interconnected mechanisms collectively maintain the metabolic plasticity of GC cells (126). FTO, an RNA demethylase with specific m⁶A demethylation activity (127), binds to circFAM192A at specific sites and removes its m⁶A modifications to prevent degradation. Subsequently, circFAM192A interacts with the membrane-localized leucine transporter SLC7A5 to enhance its stability, increasing the membrane localization of SLC7A5 for leucine transport. Elevated leucine uptake thereby promotes GC cell proliferation (128). FTO can also specifically remove the m⁶A modification of circ_0112136 to maintain its stability, thereby activating the PI3K/AKT/mTOR pathway. Notably, mTORC1 is a critical signaling molecule that activates the rate-limiting enzyme of pyrimidine synthesis (129,130). It should be emphasized that whether the FTO/circ_0112136 axis regulates pyrimidine biosynthesis, carbamoyl-phosphate synthetase 2, aspartate transcarbamylase and dihydroorotase (CAD) activity or nucleotide levels in GC cells via the mTORC1 signaling pathway warrants further investigation. This speculation is derived solely from the established function of mTORC1 in nucleotide metabolism.

Although the most abundant epitranscriptomic modification of RNA, direct evidence for m⁶A-modified circRNAs regulating metabolic remodeling pathways in GC remains limited. However, findings from other tumor types provide valuable insights. For example, in hepatocellular carcinoma, m⁶A-modified circRAPGEF1 competitively binds to IGF2BP3 via its KH domain, leading to the degradation of certain enzymes in aspartate metabolism. Aspartate activates CAD, the rate-limiting enzyme in de novo pyrimidine synthesis, thereby enhancing tumor cell stemness (131). The 'm⁶A-circRNA-metabolic reprogramming' axis represents a conserved regulatory module across tumors, which strengthens the credibility of a similar mechanism operating in GC. Table I systematically summarizes the molecular mechanisms of circRNAs in the metabolic reprogramming of GC.

Plasma and serum are clinically accessible body fluid samples, and research on circRNA biomarkers in these fluids is relatively advanced. On the one hand, plasma circRNAs can be used for early detection and to distinguish between healthy individuals and patients with GC. For example, the combined detection of circulating hsa_circ_0001017 and hsa_circ_0061276 is closely associated with major clinicopathological factors of GC, showing a sensitivity and specificity of 84.7 and 96.6%, respectively, which demonstrates notable clinical value for GC diagnosis and prognosis assessment (140). On the other hand, plasma circRNAs can differentiate between different clinical stages of patients with GC. For instance, the expression level of plasma exosomal circLPAR1 in patients with GC is notably decreased in the advanced stages (stages IIb and III), while it is elevated in the early and intermediate stages (stages I and IIa) or after surgery. Moreover, the diagnostic performance of plasma exosomal circLPAR1 is superior to that of CEA and CA19-9, with notably improved diagnostic efficiency, demonstrating good sensitivity and specificity, making it a potential biomarker for GC diagnosis and staging (141). Additionally, studies have isolated exosomes from the plasma of patients with GC and healthy individuals and used high-throughput RNA-seq to analyze circRNA expression in circulating exosomes. The authors found that hsa_circ_0093425, hsa_circ_0007247 and hsa_circ_0029780 were significantly upregulated, while hsa_circ_0139983, hsa_circ_0001989, hsa_circ_0011115 and hsa_circ_0019618 were significantly downregulated in patients with GC, indicating that plasma exosomal circRNAs exhibit differential expression in GC (142). In addition, plasma-derived hsa_circ_0000745 (143), hsa_circ_0000190 (144), hsa_circ_0000181 (145), hsa_circ_0003195 (146), circPTPN22 (147), hsa_circ_0005927 (148) and hsa_circ_0065149 (149), as well as serum-derived hsa_circ_0001649 (150), hsa_circ_0000231 (151), circHAS2 (152), hsa_circ_0000702 (153), hsa_circ_0050547 (154) and hsa_circ_000200 (155), also play important roles in the screening, diagnosis and prognosis of GC.

Gastric juice is directly secreted by gastric cells and more accurately reflects local gastric lesions compared with serum and plasma; it is not cleared by the liver and, in theory, may lack ribonucleoprotein and lipoprotein complexes, thereby reducing the risk of ncRNA degradation by nucleases, proteases and the acidic environment of gastric juice. This makes gastric juice a potentially ideal source of biomarkers for the dynamic monitoring of GC screening, diagnosis, treatment efficacy and recurrence (156). Shao et al (157) studied circRNA expression changes in gastric juice and confirmed that hsa_circ_0014717 can stably exist in human gastric juice. By comparing healthy individuals, patients with gastric ulcer, patients with chronic atrophic gastritis and patients with GC, they found that hsa_circ_0014717 expression was significantly downregulated in the chronic atrophic gastritis group, but increased in the GC group, a trend similar to that observed for hsa_circ_0005927 in gastric juice (148). This suggests a possible link to tumor cell secretion mechanisms and changes in the TME. Additionally, hsa_circ_0065149 was shown to stably exist in gastric juice, but its expression did not notably differ among the four groups. Furthermore, the expression trend of hsa_circ_0065149 in gastric juice was not consistent with that in GC tissues or plasma exosomes, suggesting that this may be related to small sample size or selective changes in circRNA sources (149). Therefore, gastric juice circRNA holds unique value for early evaluation of GC treatment efficacy and is especially suitable for patients unable to tolerate repeated endoscopic examinations.