From bench to bedside: Stem cell applications in gastric cancer therapy and their emerging clinical relevance

Abstract

Gastric cancer is one of the most aggressive malignancies worldwide, with high recurrence, metastasis, and resistance to conventional therapies. Increasing evidence points to gastric cancer stem cells (GCSCs) as critical drivers of tumor initiation, progression, and therapeutic failure. These cells are regulated by developmental signaling pathways, epigenetic networks, metabolic adaptations, and interactions with the tumor microenvironment, all of which sustain stemness and promote immune evasion. Emerging therapeutic strategies now target these vulnerabilities through pathway inhibition, RNA-based approaches, modulation of the tumor microenvironment, nanotechnology platforms, and immunotherapies, supported by preclinical and translational evidence. This review focuses on the clinical relevance of GCSCs, the molecular and cellular mechanisms underlying their maintenance, and the evolving therapeutic and precision medicine strategies aimed at overcoming stemness-driven resistance in gastric cancer. However, despite substantial preclinical progress, clinical translation remains limited by the lack of standardized biomarkers, validated endpoints and prospective trials, highlighting the urgent need to bridge this gap to achieve true stemness-informed precision therapy.

Key Words: Gastric cancer; Gastric cancer stem cells; Stemness; Tumor microenvironment; Therapy resistance; Biomarkers; Precision medicine

Core Tip: Gastric cancer stem cells (GCSCs) are key mediators of recurrence, metastasis, and therapy resistance. Understanding their regulation through signaling pathways, epigenetic mechanisms, and tumor microenvironmental support provides opportunities for novel therapeutic interventions. This review highlights the clinical relevance of GCSCs, current mechanistic insights, and emerging strategies such as RNA-based therapies, immunotherapy, and nanotechnology offering a precision medicine framework to improve outcomes in gastric cancer.

INTRODUCTION

Gastric cancer remains among the most lethal malignancies worldwide ranking fifth in incidence and third in cancer-related mortality. Despite advances in surgery, chemotherapy, targeted therapy and immunotherapy, recurrence, metastasis and therapy resistance continue to drive its persistently high fatality[1-5]. Increasing evidence implicated a rare subpopulation of tumor cells gastric cancer stem cells (GCSCs) as the main drivers of these outcomes. These cells possess self-renewal capacity, tumor-initiating potential, and multilineage differentiation and are typically identified by specific surface markers[4-7].

Mechanistically, GCSC maintenance is orchestrated by developmental and stress-adaptation programs such as Wnt/β-catenin, Hedgehog, Notch, transforming growth factor (TGF)-β/suppressor of mothers against decapentaplegic proteins (SMAD), epithelial-mesenchymal transition (EMT), hypoxia signaling and metabolic rewiring which collectively promote invasion, immune evasion, and treatment failure[5,8,9]. Tumor microenvironment (TME) plays an equally decisive role: Mesenchymal stem cells (MSCs) and cancer-associated fibroblasts (CAFs) remodel extracellular matrix, secrete cytokines [like interleukin (IL)-6, IL-8, TGF-β], and create an immunosuppressive milieu that reinforces stemness and chemoresistance[10-12]. Extracellular vesicles further mediate cancer stem cell (CSC)-non-CSC communication; amplifying metastatic and stem-like traits through paracrine signaling[13,14].

At the systems level, transcriptomic and single-cell studies have characterized stemness-based molecular subtypes of gastric cancer that differ in survival, immune landscape and therapy response[15-19]. High-stemness tumors show enhanced EMT and immune evasion features, linking stemness with immune-oncology and therapy-resistance.

Across molecular layers numerous regulators of GCSC biology have been identified, including transcription factors (E2F1, HMGA1, PRDM1/BLIMP1, AR, ATOH1), epigenetic and epitranscriptomic modulators (METTL14, WTAP, m6A machinery) noncoding RNAs, and membrane proteins[20-32]. Together these findings highlight the complex regulation of stemness, yet standardized markers and clinically actionable endpoints remain scarce, limiting translation into perspective trials[2,16].

Therapeutic innovation is rapidly advancing with strategies targeting core stemness pathways, metabolic dependencies (like glycolysis, ferroptosis resistance), tumor stroma interactions and immune landscape mechanisms. Approaches such as pathway blockade, transcription factor inhibition, anti-cluster of differentiation (CD) 33 chimeric antigen receptor T (CAR-T) therapy and programmed death-ligand 1 (PD-L1)/CTCF modulation are being complemented by novel delivery systems including MSC-based carriers, CSC-selective nanoparticles and locoregional nanocomplexes that co-target CSCs and bulk tumor cells[6,9,33-37]. In parallel, organoid systems and patient-derived models are transforming preclinical testing and aligning stemness biology with genotype, microenvironment, and drug sensitivity[38,39].

Against this backdrop, our article synthesizes current evidence on the clinical relevance of GCSCs and stemness-defined subtypes, the key molecular and cellular mechanisms driving their maintenance and emerging therapeutic and precision medicine strategies aimed at overcoming stemness-driven resistance and improving outcomes in gastric cancer[15-18,38].

METHODOLOGY

This narrative review was developed following the general framework of the preferred reporting items od systematic reviews and meta-analyses (PRISMA) guidelines, which were applied to ensure precision, transparency and methodological consistency in the literature selection process. PubMed, Cochrane Library, MEDLINE, Scopus, ClinicalTrials.gov, and Web of Science databases were searched by two of the authors (Prisacariu IA and Koumarelas KE) to identify studies reporting data on GCSCs from 2020 up to June 2025. The following medical subject headings and free-text terms, alone or combined with the logical operators “OR” and “AND” were used: Gastric cancer, stomach neoplasms, gastric carcinoma, cancer stem cells, stemness and gastric cancer stem cells. Old, duplicate and non-English studies were excluded.

After screening 517 records by title and abstract, 99 representative publications were included based on their relevance, scientific quality and contribution to the thematic synthesis. Inclusion criteria comprised peer-reviewed Englisch studies providing preclinical, clinical or translational data related to GCSCs, their molecular regulation or therapeutic targeting. High-quality reviews were also considered when they offered relevant mechanistic insight. Exclusion criteria included editorials, conference abstracts, commentaries, non-gastric cancer models, duplicate publications and studies lacking methodological clarity or accessible data. The screening and selection process were performed independently by two distinct authors (Prisacariu IA and Koumarelas KE) and any uncertainties regarding inclusion or exclusion criteria were resolved through discussion with the corresponding author to ensure accuracy and consistency. Although elements of the PRISMA framework were followed to enhance methodological precision, the present work constitutes a narrative review, aiming to integrate mechanistic, clinical and translational insights rather than to perform quantitative meta-analysis or risk-of bias assessment.

GCSCS CLINICAL RELEVANCE

GCSCs are a distinct subpopulation of tumor cells with the ability to self-renew, differentiate, and drive tumor regrowth after treatment. They are believed to play a central role in disease recurrence, metastasis, and therapy resistance. Clinically, GCSCs are identified using several markers, including CD44, CD133, ALDH1A1, EpCAM, SOX2, NANOG, OCT4, and LGR5, which are often expressed at higher levels in advanced disease and correlate with poor patient survival[13,40-43] (Table 1).

Table 1 Clinical relevance of gastric cancer stem cells key markers and their significance.

Gene/marker	Clinical significance
CD44	Common GCSC marker; associated with self-renewal, metastasis, and poor prognosis
CD133	Identifies chemoresistant GCSCs; linked with tumor initiation and relapse
ALDH1A1	Enzyme marker of stemness; correlates with advanced disease and survival
EpCAM	Cell adhesion molecule; supports tumor-initiating capacity
SOX2, NANOG, OCT4	Pluripotency transcription factors; sustain stem-like properties and correlate with aggressive phenotypes
LGR5	Stem cell receptor; marks progenitor-like GCSCs with poor clinical outcomes

CD: Cluster of differentiation; GCSC: Gastric cancer stem cell.

Importantly, these cells can also be detected in the bloodstream. Circulating GCSCs identified by combinations such as CD24⁺ CD44⁺EpCAM⁺ CD54^- were found in virtually all gastric cancer patients and were especially enriched in those with metastatic disease. In fact, one flow-cytometry panel achieved an area under the curve (AUC) of 0.911, with 83% sensitivity and 95% specificity for distinguishing gastric cancer patients from healthy individuals, highlighting the potential of liquid biopsies to monitor stemness in real time[7,44]. Parallel analyses of tumor tissue using bulk and single-cell transcriptomic “stemness indices” consistently showed that high-stemness tumors were more aggressive, enriched for EMT programs, and carried features of immune evasion, again translating into worse clinical outcomes[15,19,45-47].

Several tumor-intrinsic regulators of GCSCs have direct clinical implications. The DNA demethylase TET1 and the transcription factor FOXO4 (a negative regulator of Wnt/β-catenin signaling) both decline progressively from primary tumors to lymph node and distant metastases; their low expression predicts poor survival, with FOXO4 confirmed as an independent prognostic factor[48] (Table 2). Similarly, high SOX9 expression identifies progenitor-like tumor cells, and in patient cohorts, SOX9 positivity was linked with higher recurrence and resistance to chemotherapy[49]. Other oncogenic drivers such as E2F1, c-Myc, TGF-β-activated kinase 1 (TAK1), STMN1, AR, SYT11, DAZAP1, and CYB5R1 have been shown to increase stemness marker expression and independently predict worse outcomes. For example, patients with high TAK1 expression had a hazard ratio (HR) of 2.01 for overall survival and 1.79 for recurrence-free survival, while CYB5R1 overexpression predicted both shorter overall and progression-free survival[8,20,30,50-52]. On the other hand, tumor-suppressor genes such as BATF2, ATOH1, METTL14, nuclear receptor RORβ, and CDK5RAP3 are associated with reduced stemness and better survival. For example, patients with high ATOH1 expression had a 5-year survival of 62.3% compared to 44.3% in the low-expression group[5,24,53,54].

Table 2 Regulatory mechanisms of gastric cancer stem cells genes and pathways.

Gene/pathway	Role in stemness regulation
Wnt/β-catenin	Promotes self-renewal, EMT, and therapy resistance
Notch/Hedgehog/TGF-β	Maintain stemness programs, regulate cell fate and differentiation
FOXO4, TET1	Tumor suppressors; their loss enhances stemness and predicts poor prognosis
SOX9	Promotes symmetric division and expansion of GCSC pool; linked with recurrence
E2F1, PRDM1, AR, STMN1, c-Myc	Oncogenic drivers; upregulate stemness markers, proliferation, and resistance
ncRNAs (LINC00520, HCP5, circUBA2, miR-378a-3p, miR-15a-5p)	Regulate signaling pathways (Wnt, Hedgehog, Hippo) to promote or suppress stemness traits

ncRNA: Noncoding RNA; TGF: Transforming growth factor; EMT: Epithelial-mesenchymal transition; GCSC: Gastric cancer stem cell.

Epigenetic and noncoding RNA networks also play a clinically relevant role in GCSC regulation. Long noncoding RNAs (lncRNA) (LINC00520, HCP5, HNF1A-AS1) and circular RNAs (circUBA2, circFAM73A, circ 0051246) are frequently overexpressed in gastric tumors, where they sustain stemness programs and correlate with poor prognosis. Conversely, microRNAs such as miR-148/152, miR-378a-3p, miR-375, and miR-15a-5p act as tumor suppressors; their loss leads to upregulation of pathways like Hedgehog or ferroptosis resistance and poorer survival outcomes[55-63].

The TME strongly supports GCSCs. High expression of CAF or MSC signatures correlates with shorter survival and higher stemness indices. Mechanistically, CAFs remodel the extracellular matrix and secrete cytokines that sustain stem-like states, while MSCs enhance tumor sphere formation and drug resistance. For example, gastric cancer MSC-conditioned medium increased the half-maximal inhibitory concentration (IC₅₀) of 5-luorouracil (5-FU) and paclitaxel, reflecting increased chemoresistance[11,12,36,64-66]. Immune interactions further reinforce stemness: Cholesterol metabolism via SREBP2 reduces tumor stiffness and blunts natural killer (NK) cell activity, while loss of TAP1 compromises antigen presentation and promotes immune escape[2,3]. Even the microbiome can influence stemness, as shown by Helicobacter pylori Tipα, which activates Wnt/β-catenin signaling and induces CD44, OCT4, and NANOG expression[67].

Therapy itself can select for GCSCs. Cisplatin exposure increases the proportion of CD133⁺ cells in gastric cancer biopsies, and xenografts derived from CD44⁺ cells remain larger under cisplatin than CD44^- counterparts[6,59]. Patient-derived organoids resistant to oxaliplatin showed markedly higher fractions of CD133⁺ cells (14.5%-28.0%) compared with sensitive organoids (0.3%-3.9%), with IC₅₀ values 10-40 times higher (32-62 μmol/L vs 1.6-3.3 μmol/L), directly linking stemness to chemoresistance[68]. These findings explain why relapsed or refractory tumors often harbor higher proportions of stem-like cells and suggest that treatment itself may enrich for GCSCs.

Finally, composite risk models and multi-gene signatures have been developed to stratify patients by stemness. These include the GCSC-related score (GCScore) (HR approximately 9.15) and lncRNA-based signatures, which consistently outperform tumor node metastasis (TNM) staging in predicting survival and therapy response[16,17,69]. For example, low-stemness patients often respond better to programmed cell death protein 1 (PD-1)/PD-L1 checkpoint inhibitors, while high-stemness tumors may benefit from kinase inhibitors or ferroptosis-targeting approaches. Classical pathways such as Sonic Hedgehog (SHH) signaling also predict poor survival and are now being reconsidered as therapeutic targets[70].

In summary, GCSCs are clinically relevant because they are detectable in both tissue and blood, correlate with tumor aggressiveness, predict survival, and mediate resistance to chemotherapy and immunotherapy. Their clinical impact is reinforced by their strong links to TMEal support and therapy-induced selection. These findings position stemness-related markers and signatures as actionable tools for both biomarker-guided prognosis and development of targeted therapies in gastric cancer.

STEM CELL BASED THERAPEUTIC STRATEGIES

Therapeutic strategies for GCSCs have evolved from conceptual models to a growing body of translational approaches aimed at blocking stemness, reversing drug resistance, and enhancing immune clearance. Although direct stem cell therapies are not yet in clinical use for gastric cancer, multiple avenues are being investigated to therapeutically exploit GCSC biology with some strategies, particularly pathway inhibitors, repurposed drugs and immunotherapies, being closer to clinical application than others.

One promising approach involves targeting key transcriptional and signaling regulators of stemness. For example, the TET1/FOXO4 axis normally restrains Wnt/β-catenin signaling; when this axis is lost, GCSCs expand and undergo epithelial-to-mesenchymal transition. Pharmacological Wnt inhibitors such as ICG001 restored control in this setting, reducing CSC markers[48]. Similarly, deletion of SOX9, a stemness-associated transcription factor, impaired organoid formation and growth, suggesting that SOX9 is an actionable CSC vulnerability[49]. Other regulators such as TMEM206, PRDM1, CYB5R1, and STMN1 have been identified as mediators of stemness, where their knockdown reduced tumor sphere formation and increased chemosensitivity. Conversely, enhancing tumor suppressors such as RORβ, BATF2, CDK5RAP3, TRIM28, and METTL14 reduced CSC traits and tumorigenicity in animal models, supporting these molecules as potential therapeutic entry points[5,22,25,50,52-54,71,72].

Noncoding RNAs are also central regulators of GCSC biology and represent a rich pool of therapeutic candidates. Silencing oncogenic lncRNAs (LINC00520, HCP5), circular RNAs (circ 0051246, circFAM73A, circUBA2), or restoring tumor-suppressive microRNAs such as miR-148/152, miR-378a-3p, miR-144-3p, miR-375, and miR-15a-5p consistently reduced stemness traits, suppressed tumor growth, and in some cases triggered ferroptosis, a form of programmed cell death[27,28,55,56,58-62,73]. These findings highlight how RNA-based therapeutics could be harnessed to reprogram stem-like tumor cells and restore treatment sensitivity, although clinical translation will require overcoming challenges in delivery, stability and off target effects.

The TME, including stromal and immune components, is another critical therapeutic target. MSC-derived exosomes loaded with microRNAs showed strong anti-tumor effects in both cell culture and xenograft models, inducing apoptosis and suppressing tumor growth without the risks of immune rejection that accompany MSC transplantation[74]. Disrupting pathological MSC tumor interactions can also overcome resistance. For instance, inhibiting mitochondrial transfer from MSCs to cancer cells using ROCK1 inhibitors restored oxaliplatin sensitivity[65]. Similarly, exosomes enriched in Wnt5a were shown to reprogram stromal cells via YAP activation, but inhibition of this pathway could reverse their pro-tumorigenic effects[14]. Other TME-focused strategies include IL-6 blockade, which abrogated adipose-derived stem cell-induced stemness[64], and inhibition of CTCF-PD-L1 signaling, which reversed MSC-mediated chemoresistance[36]. Importantly, a study conducted by Zhu and Wang[3] showed that cholesterol metabolism inhibitors such as simvastatin restored NK cell cytotoxicity against GCSCs, offering a metabolic route to improve immunosurveillance. Because statins and IL-6 targeting agents are already used clinically in other settings, these approaches may be among the more readily translatable TME-directed strategies.

Beyond the TME, natural compounds and pharmacologic agents have demonstrated CSC-directed activity making them particularly attractive for repurposing. Rhaponticin selectively eliminated CD133⁺/CD166⁺ GCSCs, reduced PD-L1 expression, and promoted infiltration of CD4⁺ and CD8⁺ T cells into xenograft tumors[75]. Apatinib, a vascular endothelial growth factor receptor-2 inhibitor, suppressed stemness markers and decreased spheroid viability, while combination regimens such as apatinib + pyrotinib or imatinib overcame acquired resistance in human epidermal growth factor receptor 2 (HER2)-positive tumors[70,76]. Additional agents, including atorvastatin, chloroquine, loganetin, 4’-bromo-resveratrol (4-BR), and superparamagnetic iron oxide nanoparticles (Atranorin^@SPION nanoparticles) have suppressed CSC proliferation, promoted apoptosis, and in some cases activated alternative cell death pathways such as ferroptosis or necroptosis[77-82]. Collectively, these data suggest that rational combinations of tyrosine kinase inhibitors (TKIs), metabolic modulators and autophagy or ferroptosis regulators may be among the most clinically approachable strategies for GCSC targeting.

Immunotherapy and advanced delivery platforms are also being applied to CSCs. Anti-CD133 CAR-T cells effectively recognized and killed cisplatin-resistant CD133⁺ gastric cancer cells, producing high levels of cytokines such as IL-2 and interferon-γ[6]. Similarly, antibody drug conjugates targeting LGR5⁺ CSCs eradicated tumors in preclinical models[38]. Novel delivery systems include Nano-EN-IR^@Lip nanoparticles, which combined CSC inhibition with photothermal therapy, and MSC-loaded nanoparticles carrying paclitaxel, which provided tumor-specific drug delivery while allowing real-time imaging[37,83]. Engineered exosomes carrying therapeutic microRNAs, such as miR-424-3p, further exemplify the precision with which CSC-directed therapies can be designed[66]. Although these modalities are still at a preclinical stage, they illustrate how CSC-directed immunotherapy and nanotechnology could eventually be integrated with existing treatment regimens.

Finally, precision medicine approaches integrate stemness biology with therapy stratification. Patients with high-risk mesenchymal or CAF signatures were predicted to respond better to TKIs and checkpoint inhibitors, whereas low-risk groups were more likely to benefit from cytotoxic chemotherapy and anti-PD-1 immunotherapy[11,12,15,19,47]. Prognostic models such as the GCScore and identification of druggable stemness modules (e.g., PLK4, EZH2, PARP1, DAZAP1) provide actionable biomarkers to select therapies and monitor patient response[16,25,51,68]. These tools offer a framework in which GCSC-associated markers and pathways can be aligned with specific treatment modalities.

In summary, stem cell-based therapeutic strategies in gastric cancer encompass a wide range of approaches: Direct targeting of stemness regulators, RNA-based interventions, disruption of tumor-stroma interactions, pharmacologic CSC inhibitors, immune-based therapies, and precision delivery systems. Among these, repurposed agents with existing clinical use (such as apatinib, statins and cloroquine), pathway-directed inhibitors and immunotherapy- based combinations currently appear the most immediately translatable. By contrast, gene- and vesicle-based therapies and complex nanoplatforms, while high promising, remain at an earlier development stage. Collectively, these efforts highlight how GCSC biology is beginning to inform personalized treatment strategies, but they also underscore the need for rigorous toxicologic assessment, pharmacokinetic studies and multicenter clinical trials before these approaches can be routinely implemented to overcome drug resistance, reduce recurrence and improve survival.

PRECLINICAL EVIDENCE

Extensive preclinical research provides strong evidence that GCSCs play a central role in tumor initiation, progression, and resistance to therapy. Both in vitro and in vivo models have been widely used to dissect the molecular mechanisms of stemness and to evaluate novel therapeutic strategies.

In vitro assays consistently demonstrate that manipulating regulators of stemness alters cancer cell behavior. For instance, TET1 knockdown increased tumor sphere formation, epithelial mesenchymal transition markers, and invasion, while TET1 overexpression restored an epithelial phenotype[48]. Similarly, deletion of SOX9 in organoid and mouse models abrogated neoplasia, confirming its essential role in malignant stemness[49]. Several oncogenic drivers, including E2F1, TMEM206, AR, STMN1, SYT11, cyclophilin A (CyPA) and multiple circular RNAs and lncRNAs, promote proliferation, invasion, tumor sphere formation and drug resistance[84-86]. Silencing these regulators reduces stemness features and increases apoptosis. On the other hand, tumor-suppressive regulators such as BATF2, RORβ, CDK5RAP3, ATOH1, METTL14, FPR3, and miR-378a-3p consistently reduced GCSC traits, sensitized cells to chemotherapy, and impaired EMT when overexpressed[5,24,25,29,53,54,61].

In vivo studies confirmed these findings by demonstrating that GCSCs are potent tumor initiators. Xenograft models showed that knockdown of PRDM1, DAZAP1, SIX2, TAK1, CD44v6, WTAP, SYT11, CYB5R1, and circular RNAs significantly reduced tumor growth, stem cell frequency, and resistance to chemotherapy, while overexpression of these drivers enhanced tumorigenicity[8,22,26,30,87]. Limiting dilution assays across several models confirmed that far fewer CSC-like cells are needed to initiate tumors compared to non-stem populations, providing functional proof of their tumor-initiating ability[41,43,44,59]. Importantly, xenografts derived from miR-15a-5p, miR-144-3p, and miR-375 mimics were significantly smaller and less invasive, suggesting the therapeutic potential of restoring tumor-suppressive microRNAs[60,62,63].

Mechanistic studies have revealed that GCSCs use multiple survival pathways to resist treatment. MSC-to-GC mitochondrial transfer promoted oxaliplatin resistance but was reversed by ROCK1 inhibition[65]. MSC-conditioned media also induced glycolysis via HK2, enhancing tumor growth and metastasis, effects neutralized by hepatocyte growth factor blockade or nuclear factor kappa-B (NF-κB) inhibition[88]. Similarly, tumor-promoting cytokines secreted from MSCs activated NF-κB dependent signaling and sustained CSC features[10]. Autophagy emerged as a recurrent survival mechanism, with IL-17B/beclin-1, AQP5, and hypoxia-driven autophagy shown to sustain CSC maintenance and resistance[78,89,90]. Epitranscriptomic regulation also proved crucial: METTL3/PARP1, METTL14/ATF5, and WTAP/TGF-β modulated DNA repair, EMT, and therapy resistance[25,26,68]. In addition, ferroptosis resistance was linked to the OTUD5/GPX4, POLQ (DNA polymerase theta)/dihydroorotate dehydrogenase (DHODH), and ferroptosis-related lncRNA (lncFERO)/SCD1 pathways, providing rationales for ferroptosis-inducing therapies[33-35,91] (Table 3).

Table 3 Therapy resistance mediated by gastric cancer stem cells gene-level insights.

Gene/marker	Contribution to therapy resistance
CD133, CD44	Enriched after chemotherapy (cisplatin/oxaliplatin); mediate chemoresistance and tumor regrowth
GPX4, OTUD5, POLQ, DHODH	Protect GCSCs from ferroptosis; therapeutic vulnerability for drug development
PD-L1	Enhances immune evasion and CSC survival; predicts poor response to immunotherapy
TAP1 (low expression)	Impairs antigen presentation, promoting immune escape
AQP5, beclin-1, ULK1	Autophagy regulators that sustain stemness under stress and confer resistance
CYB5R1	Links redox balance to stemness and chemoresistance; high expression predicts poor survival

CD: Cluster of differentiation; DHODH: Dihydroorotate dehydrogenase; PD-L1: Programmed death-ligand 1; GCSC: Gastric cancer stem cell; CSC: Cancer stem cell.

Preclinical interventions tested in these models also demonstrated significant therapeutic promise. Apatinib suppressed SHH signaling, reducing tumor volume by more than 50% without systemic toxicity[70]. Simvastatin restored NK cell cytotoxicity against CSCs by inhibiting SREBP2, improving survival in mice[3]. Natural compounds such as rhaponticin, atorvastatin, loganetin, clofoctol (CFT), 4-BR, prostaglandin D2, and Atranorin^@SPION consistently reduced CSC proliferation, triggered apoptosis or ferroptosis, and diminished xenograft tumor sizes[75,77,92]. Immunotherapy approaches, including anti-CD133 CAR-T cells and exosome-delivered miR-424-3p, demonstrated potent tumor regression and suppression of EMT in xenograft models[6,66]. Cutting-edge nanotechnologies such as Nano-EN-IR^@Lip photothermal therapy, MSC-loaded paclitaxel nanoparticles, and cisplatin-loaded hydroxylated single-walled carbon nanotubes offered robust tumor suppression with minimal toxicity[37,83,93].

Finally, organoid and patient-derived models replicated the histology and mutational landscape of gastric tumors, confirming their translational utility for drug testing and biomarker discovery[38,39,94,95]. Systems-level and computational studies further demonstrated that high stemness signatures correlate with immune evasion, M2 macrophage infiltration, and worse prognosis, complementing experimental evidence[7,11,15-19,45-47,84].

However, despite these encouraging results, preclinical models represent inherent limitations that must be acknowledged. Most in vitro systems rely on established cell lines or tumor sphere assays that only partially recapitulate the cellular heterogeneity, immune contexture and stromal interactions of primary gastric tumors. Xenograft models often use immunodeficient mice, which precludes evaluation of immune-mediated mechanisms and may overestimate therapeutic efficacy. Organoid and patient-derived-xenograft (PDX) models improve physiological relevance but still lack vascular and immune components critical for therapy response. Furthermore, variability in model design, endpoint definitions and dosing regimens across studies complicated reproducibility and translation to human disease.

In summary, preclinical evidence from cellular, animal, and computational studies confirms that GCSCs drive tumorigenesis, metastasis, and therapeutic resistance in gastric cancer. These models have been indispensable in uncovering key pathways such as ferroptosis defense, autophagy and immune evasions. And in identifying potential therapeutic vulnerabilities. Yet their inherent limitations underline the need for integrated models that incorporate immune and stromal compartments and for early-phase clinical validation to determine which GCSC-directed strategies truly translate to patient benefit. Building on these experimental insights, recent translational research has begun to explore how GCSC-associated biomarkers and therapeutic targets can be integrated into clinical decision-making.

CLINICAL EVIDENCE

Although interventional clinical trials specifically targeting GCSCs remain limited, a growing body of patient-derived data, retrospective cohorts, and translational analyses provides robust evidence linking stemness to clinical outcomes, therapy response, and biomarker development. However, it is important to note that the strength of association varies across cohorts, and not all studies consistently produce these findings. Key stemness-associated markers and their prognostic or predictive implications are summarized in Table 4.

Table 4 Key stemness-associated markers and clinical implications in gastric cancer.

Marker/signature	Type	Clinical association	Prognostic/predictive implication	Ref.
TET1, FOXO4	Tumor suppressors	Progressive loss from primary to metastasis	Low expression poor survival; FOXO4 = independent adverse factor	Qi et al[48]
SOX9	Transcription factor	High in tumors; linked to recurrence (especially liver)	High SOX9 worse survival, chemo resistance; low SOX9 better response to adjuvant chemo	Chen et al[49]
E2F1, PRDM1, STMN1, TMEM206, SYT11, WTAP, c-Myc, TAK1	Oncogenic drivers	Associated with stemness programs	High expression poor survival; validated as independent adverse markers	Wang et al[8]; Fu et al[20]; Qin et al[22]; Liu and Da[26]; Kim et al[30]; Wei et al[50]; Zhang et al[71]; Yang et al[83]
RORβ, CDK5RAP3, BATF2, ATOH1, TRIM28, METTL14	Tumor suppressors	Restrain stemness/EMT pathways	High expression better prognosis and/or improved chemo sensitivity	Wen et al[5]; Zhong et al[24]; Zhang et al[25]; Cao et al[53]; Lin et al[54]; Ning et al[72]
LINC00520, HCP5, circUBA2, circFAM73A, circ 0051246	Oncogenic ncRNAs	Promote stemness, EMT, aggressive phenotype	High expression worse survival, adverse clinicopathologic features	Xia et al[27]; Liu et al[55]; Yu et al[56]; Deng et al[58]; Li et al[73]
miR-148/152, miR-378a-3p, miR-375, miR-144-3p	Tumor-suppressive microRNAs	Inhibit stemness pathways	Higher levels improved prognosis, reduced stemness/chemoresistance	Xia et al[27]; Xu et al[61]; Shen et al[62]; Lu et al[63]; Li et al[73]
CD44/miR-21-5p/TGF-β2 axis	Stemness/chemoresistance axis	Linked to strong chemoresistance	Activation poor treatment response	Nie et al[59]; Li et al[73]
IL-17B/IL-17RB	Cytokine axis	Elevated serum and tumor expression	Potential circulating marker of CSC activity; associated with adverse features	Bie et al[89]
MSC- and CAF-derived signatures	Stromal signatures	Reflect TME remodeling, immune suppression	High scores poor survival; nomograms outperform TNM for 1-, 3-, 5-year OS	Mak et al[11]; Shen et al[12]
cGCSCs (CD24+ CD44+ EpCAM+ CD54-)	Circulating CSCs	Detected in GC patients only	High diagnostic accuracy (AUC = 0.911); support liquid biopsy monitoring	Becerril-Rico et al[7]
Exosomal Wnt5a	Exosomal marker	Linked to lymph node metastasis	Indicates pro-metastatic, stemness-supporting signaling	Wang et al[14]
Exosomal lncFERO	Exosomal ncRNA	Associated with ferroptosis resistance	High levels poor prognosis; marker of chemotoxicity-induced stemness	Zhang et al[91]

CD: Cluster of differentiation; ncRNA: Noncoding RNA; TGF: Transforming growth factor; IL: Interleukin; MSC: Mesenchymal stem cells; CAF: Cancer-associated fibroblast; cGCSCs: Circulating gastric cancer stem cells; lncFERO: Ferroptosis-related long noncoding RNA; CSC: Cancer stem cell; EMT: Epithelial-mesenchymal transition; TME: Tumor microenvironment; GC: Gastric cancer; TNM: Tumor node metastasis; OS: Overall survival; AUC: Area under the curve.

Several studies highlight the prognostic value of stemness regulators. Tissue microarrays confirmed progressive loss of TET1 and FOXO4 in primary and metastatic gastric cancer, with Kaplan-Meier analyses showing that low expression predicted poor survival, and FOXO4 emerging as an independent prognostic factor[48]. In a large cohort (n = 244), high SOX9 expression was associated with significantly worse survival and increased recurrence, particularly in the liver, while SOX9-low patients showed improved outcomes following adjuvant chemotherapy[49]. Additional stemness drivers, such as E2F1, PRDM1, STMN1, TMEM206, SYT11, WTAP, c-Myc, and TAK1 correlate with poor survival across multiple cohorts and function as independent prognostic factors[8,20,30,50,71,83]. Conversely, tumor suppressors such as RORβ, CDK5RAP3, BATF2, ATOH1, TRIM28, and METTL14 were linked with favorable prognosis, and higher expression predicted better chemotherapy response and longer survival[5,24,25,53,54,72]. Yet for several markers, including c-Myc, WTAP and CD133, prognostic value has varied among studies, suggesting context-specific roles, subtype dependence or differences in detection methods.

Noncoding RNAs also demonstrated strong clinical relevance. High expression of LINC00520, HCP5, circUBA2, circFAM73A, and circ 0051246 correlated with poor overall survival and adverse clinicopathological features, whereas upregulation of tumor-suppressive miRNAs such as miR-148/152, miR-378a-3p, miR-375, and miR-144-3p was associated with significantly improved prognosis[27,55,61,63,73]. For instance, in 52 paired samples, high miR-148/152 expression predicted superior survival outcomes, while in another cohort, activation of the CD44/miR-21-5p/TGF-β2 axis correlated with strong chemoresistance (P < 0.0001)[59,73]. Elevated serum IL-17B levels in patients also correlated with tumor IL-17RB expression, suggesting potential as a circulating biomarker for CSC-related activity[89]. Despite these encouraging trends, noncoding RNAs signatures often suffer from limited cohort sizes and lack of external validation, and some markers have shown weaker association in independent datasets.

Clinical evidence also extends to the TME and circulating biomarkers. Stromal-based prognostic scores, such as MSC- and CAF-derived signatures, stratified patients into risk groups, with high scores predicting poor survival. Importantly, nomograms integrating these signatures with clinical variables outperformed TNM staging in predicting 1-, 3-, and 5-year survival[11,12]. Exosomal markers provide additional translational potential: Wnt5a-enriched serum exosomes were linked with lymph node metastasis[14], while elevated circulating exo-lncFERO predicted poor outcomes[91]. Circulating GCSCs, defined by the CD24⁺ CD44⁺EpCAM⁺ CD54^- phenotype, were detected in patients but not in healthy volunteers, with diagnostic accuracy of 90.3% (sensitivity 83%, specificity 95%, AUC of 0.911, P < 0.0001), underscoring their value for liquid biopsy applications[7]. However, circulating CSC assays are not yet standardized, and variations in gating strategies, antibody panels and processing workflows can impact reproducibility.

Patient-derived models corroborate these findings. Organoids replicated intratumoral heterogeneity and therapy responses, making them powerful tools for biomarker validation and drug testing[38,39,95]. In oxaliplatin-resistant tumors, single-cell RNA sequencing confirmed ECM remodeling, mitochondrial enrichment, and stromal tumor interactions as resistance mechanisms[65]. Similarly, cisplatin-treated biopsies showed enrichment of CD133⁺ GCSCs, providing direct evidence for chemotherapy-induced stemness[6].

Markers predictive of therapy response are also emerging. High ATOH1 and BATF2 expression correlated with significantly better responses to 5-FU based chemotherapy, confirmed by both xenograft and organoid models[24,53]. Risk models integrating CSC features consistently stratified patients into high- and low-risk groups, independent of TNM stage (multivariate Cox HR approximately 2.0, P < 0.001). Importantly, patients in low-stemness subgroups achieved higher response rates to anti–PD-1/PD-L1 immunotherapy (44.6% vs 21.8%, P < 0.001) and showed greater immune infiltration, validating biomarker-driven immunotherapy stratification[16,17,19,45]. Nonetheless, predictive accuracy caries across studies and some stemness signatures exhibit reduced performance in external cohorts, emphasizing the need for harmonized pipelines.

Other translational evidence identifies emerging therapeutic targets such as CyPA/CD147, FPR3-NFATc1-NOTCH3, OTUD5/GPX4, and PARP1, each correlating with poor survival and resistance in patient datasets[29,68,85,96]. Additional studies linked obesity-driven leptin signaling to aggressive gastric cancer biology[97] and showed that serum HER2, rather than tissue HER2, better predicted survival outcomes[4].

While direct early-phase clinical trials specifically targeting GCSCs are lacking, indirect clinical evidence is robust. For example, phase III trials of apatinib demonstrated overall survival benefits in third-line gastric cancer, with preclinical and case reports suggesting synergy when combined with pyrotinib[70,76]. Collectively, these findings emphasize that although CSC-directed therapies are still at the translational stage, validated biomarkers, organoid models, circulating stemness markers, and stromal risk scores already provide meaningful tools for patient stratification and therapeutic planning in gastric cancer. Yet, contradictory and null findings across certain datasets highlight the need for larger, multicenter validation studies before these biomarkers can be adopted broadly in clinical practice.

MECHANISTIC INSIGHTS

Multiple mechanistic pathways sustain GCSCs, integrating intrinsic transcriptional regulation, epigenetic control, metabolic rewiring, and TME interactions.

At the molecular level, regulators such as TET1 and FOXO4 modulate stemness programs and influence differentiation and invasive phenotypes[48]. Similarly, SOX9 alters the orientation of stem cell division fueling tumorigenesis[49]. Transcription factors including E2F1, AR, PRDM1, c-Myc, and HMGA1 enhance stemness by upregulating pluripotency factors (NANOG, OCT4, SOX2, BMI1) and driving EMT, proliferation, invasion, and chemoresistance[20-23,83]. Conversely, tumor suppressors such as RORβ, BATF2, CDK5RAP3, and ATOH1 restrain stemness by inhibiting Wnt/β-catenin or stabilizing PTEN, thereby restoring chemotherapy responsiveness[5,24,53,54].

Epigenetic and noncoding RNA networks also play crucial roles. LINC00520 stabilizes c-Myc messenger RNA, sustaining EMT and chemoresistance, while HCP5 activates signal transducer and activator of transcription 3 (STAT3) and Notch pathways via microRNA sponging[55,56]. Circular RNAs such as circ 0051246 relieve repression of YAP1 by sequestering miR-375, amplifying Hippo signaling, while circUBA2 upregulates IL-6/STAT3 signaling to maintain self-renewal[28,58]. MicroRNAs also regulate stemness: MiR-148/152 target ITGA5 to reduce clonogenicity, miR-378a-3p suppresses Hedgehog signaling via RAB31, and miR-15a-5p inhibits ONECUT2 to downregulate β-catenin, while miR-21-5p paradoxically sustains stemness by upregulating TGF-β2/SMAD signaling[59,61,62,73].

The TME exerts strong influence on CSC maintenance. MSC-related genes activate angiogenesis and EMT while suppressing immunity, correlating positively with M2 macrophages and negatively with CD8⁺ T cells[12]. CAF-driven ECM remodeling enhances stemness through cytokine secretion and immune evasion[11]. Exosomal communication plays a central role: CSC-derived exosomes carrying H19 activate YAP/CDX2 in neighboring cells, driving heterogeneity, while Wnt5a-rich exosomes stimulate YAP in bone marrow-derived MSCs to promote lymphangiogenesis[13,14]. Moreover, chemotherapy itself can remodel the TME: Oxaliplatin-resistant tumors stiffen ECM, activating ROCK1-mediated mitochondrial transfer from MSCs to cancer cells, restoring oxidative metabolism and survival[98].

Metabolic pathways further sustain GCSCs. Heat shock protein 90 (Hsp90) organizes glycolytic enzyme complexes (PKM2, ENO1) at lamellipodia, enhancing local adenosine triphosphate production for migration and EMT[9]. CYB5R1 links iron-sulfur cluster biogenesis to oxidative stress resistance and chemoresistance[52]. Autophagy provides a key survival mechanism: AQP5TRIM21-ULK1 signaling, IL-17B/beclin-1 ubiquitination, and TRPV4-dependent autophagy maintain CSC viability under stress, while chloroquine or ATG7 inhibition collapses these programs[31,78,89,90]. Ferroptosis is actively suppressed in GCSCs via the p53-OTUD5-GPX4 loop, POLQ-DHODH antioxidant axis, and USP7/LncFERO/SCD1 signaling, establishing ferroptosis defense as a hallmark of CSC biology[33-35,91].

Immune evasion mechanisms extend beyond PD-1/PD-L1. CSCs exploit cholesterol metabolism via SREBP2 to reduce stiffness and impair NK cytotoxicity, creating a “mechanical immune checkpoint”[3]. PD-L1 itself not only mediates immune escape but also enhances CSC survival and chemoresistance, regulated through phosphoinositide 3-kinase/protein kinase B/mammalian target of rapamycin and mitogen-activated protein kinase signaling[75]. Loss of regulatory T cells (Tregs) weakens major histocompatibility complex class I presentation, further enabling immune escape[2]. Stromal-immune crosstalk also reinforces stemness: Tregs induce IL-13 to STAT3 signaling, promoting SOX2, LGR5, and PROM1 expression[99], while high CAF or macrophage infiltration sustains EMT and stemness traits[16,47].

Together, these findings delineate GCSCs as integrators of oncogenic transcriptional programs, epigenetic circuits, metabolic adaptations, and TME interactions. The convergence of these mechanisms explains their central role in chemoresistance, immune evasion, relapse, and metastasis, while simultaneously offering multiple therapeutic vulnerabilities.

CLINICAL TRANSLATION

Translating insights from GCSC biology into clinical practice requires integrating biomarkers into patient stratification and developing therapies that target stemness alongside existing modalities. Multiple studies highlight candidate biomarkers with direct prognostic and therapeutic implications. Several stemness-associated markers, including FOXO4 and SOX9, have emerging value for risk stratification and treatment allocation, supporting their integration into biomarker-guided clinical framework[48,49]. Similarly, EPHB2 expression has been linked to reduced lymph node metastasis and may guide decisions about lymphadenectomy or adjuvant therapy in early-stage gastric cancer[40]. High MSC-risk and CAF-signatures predict poor outcomes and may direct high-risk patients to checkpoint inhibitors or stroma-targeting therapies[11,12].

Translational work also identifies actionable therapeutic targets. For example, E2F1, a driver of 5-FU resistance, emerges as a biomarker and vulnerability that could improve responses to adjuvant chemotherapy[20]. BATF2 expression stratifies patients likely to benefit from adjuvant chemotherapy, while TMEM206 and AR suggest opportunities for targeted therapy in resistant disease[23,53,71]. RORβ, WTAP, and METTL14 function as prognostic tools and may guide RNA-targeting therapies in resistant tumors[5,25,26]. At the noncoding RNA level, targeting LINC00520, HCP5, circ 0051246, circUBA2, and circFAM73A or restoring miR-375, miR-148/152, and miR-378a-3p has been proposed to enhance chemotherapy sensitivity, supported by clinical correlations linking their expression with survival and recurrence risk[27,28,55,61,73].

Emerging translational platforms include exosomes and nanomedicine. Compared with cell-based therapies, exosomes offer scalability, lower immunogenicity, and no tumorigenic risk, and may be integrated with checkpoint inhibitors to overcome CSC-driven resistance[74]. For example, exosomal Wnt5a correlates with lymph node metastasis and may serve as a biomarker, while verteporfin-mediated YAP inhibition reversed its effects[14]. MSC-derived exosomes engineered to deliver miR-424-3p reduced EMT and stemness, suggesting therapeutic feasibility[66]. Similarly, Nano-EN-IR^@Lip and Atranorin^@SPION formulations combine imaging, photothermal therapy, and CSC inhibition, offering theranostic potential[82,83].

Drug repurposing is another promising avenue. Statins restore NK cytotoxicity by disrupting cholesterol metabolism in GCSCs[3], while atorvastatin shows synergy with sorafenib[77]. Other repurposed candidates include chloroquine (targeting autophagy in chemoresistant tumors), 3,3’-diindolylmethane (DIM) + NF-κB inhibitors (to counteract stromal activation), and CFT (an antibacterial drug with anti-CSC effects)[10,78,81].

Finally, translational strategies emphasize integration into multimodal care. Biomarker-driven classifiers such as the stemness score, 9-gene signature, and lncRNA pair risk score predict overall survival (HR approximately 2.10, P < 0.001) and therapy response, outperforming TNM staging for immunotherapy stratification[15,17,19,45]. High-stemness tumors may benefit from TKIs or ferroptosis inducers[34,35], while low-stemness tumors respond better to PD-1/PD-L1 blockade and conventional chemotherapy. Importantly, early clinical signals suggest that combining apatinib with pyrotinib in HER2⁺ GC may synergistically suppress CSC-driven resistance, while apatinib alone may reduce relapse via CSC depletion[70,76].

Altogether, these findings illustrate how stemness research is transitioning toward clinical application. Biomarkers such as SOX9, FOXO4, EPHB2, and MSC/CAF signatures can guide prognosis and therapy selection, while exosome-based delivery, nanoplatforms, and repurposed drugs expand therapeutic opportunities. The integration of these approaches with surgery, chemotherapy, targeted therapy, and immunotherapy marks a critical step toward CSC-informed precision medicine in gastric cancer.

SAFETY AND REGULATORY CONCERNS

While stem cell-based therapies and CSC-targeting strategies hold promise, their clinical translation raises important safety and regulatory challenges. Direct manipulation of epigenetic or developmental pathways, such as TET1 activity or Wnt/β-catenin signaling, risks disrupting normal stem cell compartments and tissue homeostasis[40,48,67,72,80]. Similarly, transcription factors such as E2F1, PRDM1, SIX2, and SOX9 regulate both cancer and normal tissue differentiation; systemic inhibition may impair normal proliferation, making tumor-specific delivery essential[20,22,49,87].

Noncoding RNA-based therapeutics face hurdles in specific delivery, stability, and immune activation, with risks of off-target hybridization or interference in normal RNA networks[27,55,56,61,62]. Exosome-mediated therapies appear safer than whole-cell approaches, but challenges remain in standardizing isolation, purification, and production under good manufacturing practice conditions, alongside the risk that blocking exosomal axes (e.g., Wnt5a/YAP, YAP/CDX2) may interfere with physiological processes beyond cancer[13,14,66,74]. Similarly, MSC-related pathways such as TGF-β and EMT play vital roles in stromal regeneration, meaning interventions risk compromising normal tissue repair[11,12,64].

Several pharmacological agents raise specific safety concerns. Chloroquine is Food and Drug Administration-approved but carries well-documented ocular and cardiac toxicities, requiring careful monitoring in oncology use[78]. Statins, while widely prescribed in cardiology, may impact normal stem cell and immune functions at oncologic dosing[3,77]. Natural compounds such as DIM could paradoxically activate stromal MSCs, accelerating tumor progression, highlighting the importance of evaluating stromal effects in preclinical safety studies[10]. Similarly, Hsp90 inhibitors (e.g., TAS116) show ocular and hepatic toxicities in clinical testing, requiring dose optimization[9], while OTUD5-GPX4 ferroptosis targeting, though effective, risks redox toxicity and off-target ferroptosis in normal tissues[33-35,82,91].

Cell-surface targets such as CD44, CD133, AQP5, and LGR5 are not exclusively CSC-specific and are expressed in some normal stem or epithelial cells, raising concerns of on-target, off-tumor toxicity in CAR-T or antibody-based therapies[6,32,42,44,90]. Immunotherapy strategies, including PD-L1 blockade or FPR3-Notch modulation, carry risks of autoimmune toxicity or infections due to their physiological immune roles[2,36,99]. Likewise, CAF-targeting therapies must account for stromal heterogeneity and the risk of impairing normal tissue integrity[11,94].

Nanotechnology platforms, while promising, face regulatory scrutiny regarding biodistribution, clearance, and long-term biocompatibility. For example, hydroxylated single-walled carbon nanotubes carriers and IR780-based nanoliposomes showed no acute toxicity in mice but require rigorous good laboratory practice toxicology and immune response assessment before human application[83,93]. Likewise, induced pluripotent stem cells- and organoid-based models, while safe for preclinical testing, face regulatory barriers in scaling and quality control for therapeutic use[38,39].

Altogether, these findings emphasize that while many CSC-directed strategies demonstrate efficacy, their translation must balance tumor-specific targeting with preservation of normal stem and immune functions. Moving forward, clinical application will require selective delivery systems (e.g., nanoparticles, tumor-specific promoters), biomarker-driven patient selection, rigorous toxicology studies, and compliant manufacturing to ensure safety and regulatory approval.

PRECISION MEDICINE PERSPECTIVE

The growing recognition of GCSCs as key drivers of therapy resistance and relapse has positioned them at the center of precision medicine strategies. Several biomarkers have been proposed for patient stratification, aiming to guide therapy allocation and improve outcomes. For example, decreased FOXO4 expression has been associated with poor survival, while elevated SOX9 not only marks an aggressive phenotype but also predicts reduced responsiveness to chemotherapy. Together, these findings suggest that routine assessment of FOXO4 and SOX9 could help identify patients who might benefit from intensified or alternative regimens rather than standard protocols[48,49]. Similarly, the absence of EPHB2 has been linked with early tumor progression and increased lymph node metastasis, indicating its potential role in stratifying early gastric cancer patients for closer surveillance or adjuvant treatment[40].

Beyond protein markers, noncoding RNAs are emerging as powerful stratification tools. Elevated expression of LINC00520 or HCP5 identifies patients with higher relapse risk and resistance to conventional chemotherapy, whereas higher levels of miR-148/152 or miR-375 correlate with improved survival. This highlights the possibility of using RNA-based biomarkers not only for prognosis but also to guide CSC-targeted therapy approaches[55,56,73]. In parallel, circular RNAs such as circUBA2 and circ 0051246 have been linked to IL-6/STAT3 and Hippo signaling, respectively, and may serve as indicators of tumors particularly reliant on stemness pathways, suitable for therapies aimed at disrupting these circuits[28,58].

Exosome profiling represents another promising precision medicine avenue. For instance, serum exosomal Wnt5a has been correlated with lymph node metastasis and poor prognosis, suggesting its use as a liquid biopsy marker to stratify patients who require intensified therapy[14]. Likewise, exosomal lncRNAs such as lncFERO have been linked with ferroptosis resistance, pointing toward stratification strategies where high-lncFERO tumors could be prioritized for ferroptosis-inducing therapies[91]. In addition, engineering exosomes to deliver specific microRNAs, such as miR-424-3p, has shown preclinical success in reducing stemness and EMT, raising the possibility of combining biomarker testing with therapeutic exosome delivery for tailored treatment[66].

Risk-score models incorporating stemness signatures also play a central role in precision oncology. Stratification systems based on MSC- or CAF-derived signatures have been shown to predict prognosis and immunotherapy responsiveness, outperforming TNM staging in some analyses[11,12]. Similarly, composite scores such as the GCScore or stemness-risk models have been validated as independent predictors of overall survival and therapy response, with high-stemness tumors often more resistant to checkpoint inhibitors but potentially more sensitive to targeted kinase inhibitors or ferroptosis-based regimens[15,16,19,34]. These frameworks exemplify how integrating molecular profiling with stemness biology could refine therapeutic allocation.

DISCUSSION

GCSCs constitute a clinically meaningful subpopulation that underpins recurrence, metastasis, and treatment failure. In tissue, GCSCs are enriched for markers such as CD44, CD133, ALDH1A1, EpCAM, SOX2, NANOG, OCT4, and LGR5; higher expression of these markers tracks with advanced stage and inferior survival[13,40-43]. Stemness can also be profiled non-invasively: Circulating GCSCs e.g., CD24⁺ CD44⁺EpCAM⁺ CD54^- discriminate patients from healthy individuals with high accuracy (AUC = 0.911; 83% sensitivity; 95% specificity), supporting liquid biopsy-based monitoring[7,44]. Transcriptomic stemness indices from bulk and single-cell data consistently associate with EMT programs, immune evasion, and worse outcomes[15,19,45-47].

Tumor-intrinsic regulators (such as TET1, FOXO4, SOX9, TAK1, CYB5R1) stratify prognosis in multiple cohorts, whereas suppressors including BATF2, ATOH1, METTL14, RORβ, and CDK5RAP3 correlate with more favorable outcomes[5,8,20,24,30,48-54]. The TME including CAF and MSC signatures further amplifies stemness and chemoresistance, while therapy itself can select for GCSCs, as demonstrated by enrichment of CD133⁺/CD44⁺ populations following cisplatin exposure and increased stemness in drug-resistant organoids[6,11,12,36,59]. Importantly, multigene stemness scores (such as GCScore and lncRNA panels) outperform TNM staging for risk prediction and may guide therapy allocation. Low-stemness tumors respond more favorably to PD-1/PD-L1 blockade whereas high-stemness tumors may benefit from kinase inhibitors or ferroptosis-targeting strategies; canonical pathways such as SHH also remain relevant therapeutic targets[16,17,34,35,69,70]. Nevertheless, many of these associations are based on retrospective datasets or preclinical systems, and their robustness across diverse patient populations has yet to be confirmed.

Across the available evidence, most data remain preclinical or retrospective with limited prospective clinical validation. Many human studies are single-center, small, or lack harmonized endpoints, which restricts their statistical power and generalizability. Several findings rely on in silico data (single-cell RNA sequencing, bulk transcriptomics, risk models) where correlative associations may not reflect causal biology; functional validation in patient-relevant systems is often incomplete. Tumor heterogeneity spatial, temporal, and molecular is incompletely captured by tissue microarrays or small validation cohorts, and marker specificity for GCSCs (e.g., CD44/CD24/CD133) remains debated. Methodological variability across studies reliance on a narrow range of cell lines, and the absence of pharmacokinetic, biodistribution, and safety data for emerging modalities (e.g., exosomes, nanocarriers, ferroptosis inducers) further limit immediate translation. Current organoid and PDX platforms also lack immune, vascular, and neural compartments, reducing their ability to model immunotherapy response or stromal crosstalk. These limitations underscore the need for careful interpretation and emphasize that many stemness-based biomarkers are not yet ready for routine clinical application.

Future research should focus on translating the expanding knowledge of GCSCs into clinically actionable strategies. Large, prospective, multicenter validation studies are essential to confirm the prognostic and predictive value of markers such as FOXO4, SOX9, EPHB2, TAP1, and CAF/MSC-derived signatures and to establish standardized assays and cutoff. Incorporating patient-derived organoids and xenografts into co-clinical pipelines could enable real-time drug sensitivity testing and biomarker-guided therapy selection.

On the therapeutic side, combination approaches targeting multiple survival circuits, such as developmental signaling pathways (Wnt/β-catenin, Hippo/YAP, Notch), immune–stemness crosstalk (PD-L1, FPR3-NOTCH3, TAP1), metabolic programs (OTUD5-GPX4 or POLQ-DHODH, cholesterol-mediated NK suppression), and stress-adaptation mechanisms (autophagy, m⁶A epigenetic regulation) deserve further investigation. Novel delivery platforms including engineered exosomes, ligand-targeted nanoparticles, and RNA-based therapeutics should be developed alongside rigorous toxicology and manufacturing assessments. Although these approaches hold considerable promise, their clinical value will depend on rigorous multicenter validation and careful integration into existing treatment frameworks.

Ultimately, future directions in this field should integrate multi-omic profiling, organoid-based drug screening, and artificial intelligence-driven analytics to design patient-specific treatment strategies. By embedding GCSC biology into precision medicine frameworks and combining it with chemotherapy, targeted therapy, and immunotherapy, upcoming studies have the potential to transform outcomes for gastric cancer patients, making treatments not only more effective but also safer and more durable.

CONCLUSION

GCSCs are pivotal drivers of recurrence, metastasis, and therapy resistance, and their signatures are now measurable in both tissue and blood. Integrating stemness-based biomarkers with standard clinicopathologic factors can sharpen risk stratification and help match patients to immunotherapy, targeted agents, or cytotoxic regimens more effectively. Preclinical advances including inhibition of developmental pathways, stromal crosstalk, and metabolic defenses outline multiple, testable vulnerabilities. The next step is prospective validation: Harmonized assays, liquid-biopsy monitoring, and biomarker-guided trials that combine GCSC-targeted strategies with current standards of care. To ensure reproducibility and board clinical applicability, these efforts must be carries out through large, multicenter validation studies that confirm the robustness of proposed biomarkers across diverse populations and treatment setting.