CSCs are a distinct subpopulation of cancer cells characterized by their ability to self-renew, differentiate into various cell types and drive tumorigenesis. CSCs are hypothesized to be responsible for tumor initiation, progression and resistance to conventional therapies (1). Normal stem cells in tissues (such as the gut lining or bone marrow) are strictly controlled to replace damaged cells and maintain organ function. By contrast, CSCs acquire genetic and epigenetic abnormalities that disrupt this regulation, allowing them to proliferate uncontrollably and spread to other organs (metastasis) (2). Studies have shown that CSCs can repopulate tumors even after therapy, making them a critical target for therapeutic intervention (14,15). For example, in gastric cancer, CSCs have been implicated in therapeutic resistance and tumor relapse, highlighting their role in maintaining tumor growth and heterogeneity (16).

The identification and characterization of CSCs rely on specific surface markers and functional assays. Surface markers are proteins on cell membranes that can act as 'barcodes' to identify specific cell types. In CSCs, markers such as CD44, CD133 and aldehyde dehydrogenase 1 (ALDH1; an enzyme involved in metabolism) are often upregulated and are correlated with aggressive tumor behavior. For example, CD44⁺ CSCs in colorectal cancer show enhanced ability to initiate tumors and resist drugs (17,18). CD44, a transmembrane glycoprotein, is a defining CSCs marker in multiple cancer types, including breast, colorectal and head and neck malignancies. CD44⁺ CSCs exhibit enhanced tumor-initiating capacity and resistance to conventional therapies (19). CD133 (prominin-1) is another widely studied marker, implicated in CSC tumorigenicity in glioblastoma, liver and pancreatic cancer. CD133⁺ cells exhibit increased sphere-forming ability and heightened tumorigenicity in xenograft models (20). ALDH1 is an intracellular enzyme frequently used to identify CSCs in breast, lung and prostate cancer. High ALDH1 activity is linked to enhanced self-renewal, chemoresistance and tumor initiation (20). However, the variability in marker expression across different cancer types underscores the need for context-dependent analysis of CSC markers.

While CD44 and CD133 remain widely used CSC markers, their expression varies significantly across digestive tumor subtypes. For instance, CD44⁺ cells dominate in colorectal cancer, whereas CD133⁺ populations are more prevalent in pancreatic CSCs (21). Notably, Tian et al (22) demonstrated heterogeneous CD44/CD133 expression in esophageal squamous cell carcinoma (ESCC), with only 40% of tumors showing co-expression, questioning their universal applicability. Single-cell sequencing further revealed that ALDH1⁺ gastric CSCs (GCSCs) comprise distinct subclones with divergent Wnt and Notch dependencies (23), underscoring the need for context-specific marker validation.

The maintenance of CSCs is regulated by complex molecular pathways and influenced by both genetic and epigenetic factors. Key signaling pathways include Wnt/β-catenin (24), Notch (25) and Hedgehog (26), which are critical for CSC self-renewal and differentiation. For instance, aberrant activation of the Wnt/β-catenin pathway leads to β-catenin accumulation and nuclear translocation, activating genes associated with proliferation, invasion and survival (27). This dysregulation is observed in colorectal and breast cancer, where nuclear β-catenin expression is correlated with increased tumor aggressiveness and recurrence (28,29). Epigenetic factors, such as DNA methylation and histone modification, also serve significant roles in maintaining CSCs properties (30). Studies have shown that epigenetic modifications can silence tumor suppressor genes and activate oncogenes, contributing to CSC maintenance and tumorigenesis (31,32). Additionally, genetic mutations and chromosomal instability can drive the acquisition of stem cell-like properties in cancer cells, further promoting tumor progression and therapeutic resistance (33).

The TME is the 'ecosystem' surrounding cancer cells, including both living cells (such as immune cells and fibroblasts) and non-living elements [such as proteins and nutrients in the extracellular matrix (ECM)]. This environment supports tumor growth by providing nutrients, promoting blood vessel formation and shielding cancer cells from the immune system (34). Cellular components include cancer-associated fibroblasts (CAFs), immune cells [such as macrophages, T cells and natural killer (NK) cells] and endothelial cells. Non-cellular components consist of the ECM, growth factors, cytokines and metabolites (35). These components collectively influence tumor progression, metastasis and therapeutic resistance (Fig. 1). For example, CAFs promote tumor growth by secreting growth factors and remodeling the ECM, while immune cells can either suppress or enhance tumor growth depending on their polarization state (36). The ECM provides structural support and facilitates cell migration and invasion, contributing to metastasis (37).

Among the TME components, cytokines play a pivotal role in shaping the behavior of CSCs. For instance, interleukin-6 (IL-6) and IL-8 have been shown to activate critical signaling pathways such as STAT3 in CSCs, thereby promoting their self-renewal and survival (38). Matrix metalloproteinases (MMPs), which are enzymes that degrade the ECM, also contribute to CSC maintenance by facilitating tumor invasion and creating niches conducive to CSC proliferation (39). The intricate interplay between these TME components and CSCs highlights the importance of considering the microenvironment in therapeutic strategies.

The TME plays a critical role in supporting the survival and proliferation of CSCs. Hypoxia, a common feature of the TME, enhances the properties of CSCs by activating hypoxia-inducible factors (HIFs), which promote self-renewal and resistance to apoptosis (25) (Fig. 1). Studies have shown that hypoxia increases the expression of CSCs markers such as CD44 and ALDH1, driving tumor progression and therapeutic resistance (25,40). Additionally, angiogenesis within the TME provides nutrients and oxygen to CSCs, further promoting their survival and proliferation. For instance, HMGB2, a non-histone chromatin-binding protein, has been implicated in promoting angiogenesis and enhancing the properties of CSCs in digestive tract tumors. High HMGB2 expression is correlated with shorter OS and disease-free survival (DFS), highlighting its role in supporting CSC behavior (10).

Recent studies have also highlighted the role of the ECM in CSC maintenance (1,41). The ECM, composed of collagen, fibronectin and laminin, interacts with CSCs through integrin receptors. This interaction activates downstream signaling pathways such as focal adhesion kinase (FAK) and PI3K/AKT, which are essential for CSC survival and chemoresistance (1). For example, fibronectin has been found to enhance the stemness of colorectal CSCs by activating the FAK/PI3K/AKT pathway, underscoring the role of the ECM in maintaining CSC properties within the TME (1).

CSCs employ various strategies to evade anti-tumor immunity, often interacting with immunosuppressive cells such as regulatory T cells (Tregs) and myeloid-derived suppressor cells (MDSCs) (Fig. 1). Tregs suppress anti-tumor immune responses by inhibiting cytotoxic T cells, while MDSCs promote immune tolerance through the production of immunosuppressive cytokines (42,43). CSCs can secrete factors that attract and activate these immunosuppressive cells, creating a favorable microenvironment for tumor growth (44). Furthermore, CSCs express immune checkpoint molecules such as programmed death-ligand 1 (PD-L1), which inhibit T cell activity and enhance immune evasion (45). For example, HMGB2 has been shown to promote immune escape in non-small cell lung cancer by upregulating PD-L1 expression, highlighting its role in immune evasion mechanisms (46).

Moreover, CSCs can modulate the TME to reduce the infiltration of cytotoxic T cells and NK cells. For instance, in colorectal cancer, the fatty acid desaturase 1 (FADS1)/dihydroxydodecanoic acid (DDA) axis is activated under hypoxic conditions in CSCs, impairing NK cell cytotoxicity (47). This mechanism not only facilitates immune evasion but also enhances the metastatic potential of CSCs. Additionally, CSCs can exploit metabolic pathways to alter the TME. Lactate, a byproduct of glycolysis, can drive metastasis of normoxic colorectal CSCs (CCSCs) via peroxisome proliferator-activated receptor γ coactivator 1-α (PGC-1α)-dependent oxidative phosphorylation (48). This metabolic adaptation not only supports the survival of CSCs but also creates a hostile environment for immune cells.

CSCs exhibit notable metabolic flexibility, enabling them to survive and thrive in the harsh TME. CSCs can switch between glycolysis and oxidative phosphorylation depending on the availability of nutrients and oxygen. For example, in hypoxic conditions, CSCs upregulate glycolysis-related enzymes and transporters to maintain energy production (25). This metabolic reprogramming is often accompanied by the activation of specific signaling pathways, such as the HIF-1α pathway, which further enhances the stemness and survival of CSCs.

Furthermore, CSCs can utilize alternative metabolic substrates to adapt to nutrient limitations. In colorectal cancer, vitamin D has been shown to promote ferroptosis in CCSCs by downregulating SLC7A11, a cystine transporter critical for redox balance (49). This suggests that CSCs can alter their metabolic pathways to resist oxidative stress and maintain cellular homeostasis. Additionally, CSCs can exploit the metabolic products of surrounding cells. For instance, in gastric cancer, prostaglandin D2 (PGD2)/prostaglandin D2 receptor 2 (PTGDR2) signaling inhibits autophagy via autophagy-related protein 4B (ATG4B) ubiquitination, reducing CSC stemness (50). This highlights the intricate interplay between metabolic adaptation and signaling pathways in CSCs.

Esophageal CSCs drive tumor initiation, progression and therapeutic resistance through diverse molecular mechanisms (Table I). This subsection synthesizes key findings on CSC-related pathways in esophageal cancer, critically evaluates their implications and highlights unresolved controversies.

CSCs in ESCC exhibit a quiescent state, enabling evasion of conventional therapies targeting rapidly dividing cells. Chen et al (51) demonstrated that ESCC CSCs downregulate DDR pathways, reducing apoptosis under genotoxic stress. This quiescence is mediated by suppressed checkpoint kinase 1/2 activation and impaired p53 signaling, allowing CSCs to accumulate mutations and survive chemotherapy. However, this study relied on in vitro sphere-forming assays and xenograft models, which may not fully recapitulate the TME in human patients. Subsequent studies, such as that by Zhao et al (52), corroborated these findings but emphasized heterogeneity in DDR pathways across CSC subpopulations, suggesting context-dependent survival mechanisms.

Multiple signaling cascades converge to sustain CSC self-renewal and chemoresistance. The Hippo/Yes-associated protein 1 (YAP1) pathway is a critical regulator in esophageal cancer. Song et al (53) revealed that YAP1 upregulates SRY-box transcription factor 9, enhancing CSC properties such as tumorigenicity and spheroid formation. Similarly, Xu et al (54) identified a STAT3/miRNA (miR)-181b/cylindromatosis axis that promotes CSC proliferation by modulating IL-6/STAT3 signaling. While these studies highlight pathway-specific roles, conflicting evidence exists regarding cross-talk between pathways. For instance, Liu et al (55) linked the HSP27/AKT/hexokinase 2 (HK2) axis to CSC metabolic reprogramming, showing that AKT activation sustains stemness via HK2-dependent glycolysis. This contrasts with the study by Kai et al (56), in which myosin heavy chain 9 was implicated in activating PI3K/AKT/mTOR to drive CSC oncogenesis, suggesting overlapping yet distinct metabolic dependencies.

Autocrine signaling mechanisms facilitate CSC invasion and metastasis in esophageal cancer. Wang et al (57) demonstrated that C-X-C motif chemokine ligand 12/C-X-C motif chemokine receptor 4 (CXCR4) axis activation in CSCs enhances MMP secretion, promoting ECM degradation and metastatic spread. Conversely, Yue et al (58) linked TGF-β1 to CSC migration via Smad-dependent EMT activation. Despite mechanistic clarity, these studies predominantly utilized monolayer cell cultures, neglecting the contribution of stromal cells in the TME. Recent work by Wei et al (59) addressed this gap, showing that quiescin sulfhydryl oxidase 1 (QSOX1) in CSCs upregulates PD-L1 to exclude CD8⁺ T cells, illustrating how cytokine networks synergize with immune evasion.

Dysregulation of non-coding RNAs and epigenetic modifiers underpins CSC plasticity in esophageal cancer. Guo et al (60) reported that miR-637 loss activates the Wiskott-Aldrich syndrome protein and SCAR homolog/IL-8 pathway, augmenting CSC stemness and metastasis. Similarly, Xun et al (61) identified miR-191-3p as a suppressor of regulator of G-protein signaling 1, which inhibits CXCR4/PI3K/AKT signaling. These findings align with genome-wide methylation analyses by Yu et al (62), which revealed hypermethylation of tumor suppressor genes (such as Cadherin 1) in CSCs. However, inconsistencies arise in biomarker specificity; for example, Gupta et al (63) found variable CD44/CD133 expression across ESCC subtypes, questioning the universality of these markers.

CSCs employ multifaceted strategies to resist treatment in esophageal cancer. Xu et al (64) showed that STAT3 silencing sensitizes CSCs to the HSP90 inhibitor, SNX-2112, while Liu et al (65) linked ferroptosis resistance to HSP27-glutathione peroxidase 4 (GPX4) upregulation. Notably, Wei et al (59) uncovered a novel immune evasion mechanism where QSOX1 elevates PD-L1, enabling CSCs to bypass cytotoxic T cell surveillance. These studies underscore the need for combinatorial therapies targeting both CSCs and immune cells, although clinical validation remains limited.

GCSCs are central to tumor initiation, metastasis and therapeutic resistance. The functional regulation of GCSCs involves intricate signaling networks, epigenetic reprogramming and interactions with the TME (Table II). In this subsection, key mechanisms governing GCSC biology are dissected, representative studies are critically evaluated and therapeutic implications are discussed.

The Wnt/β-catenin pathway is a cornerstone of GCSC self-renewal. Mao et al (66) demonstrated that β-catenin activation enhances GCSC proliferation, while salinomycin inhibits this pathway, reducing tumorigenicity in xenograft models. Similarly, Wu et al (34) identified miR-483-5p as a promoter of Wnt/β-catenin signaling, driving GCSC invasion and self-renewal via direct suppression of GSK3β. However, discrepancies exist in downstream effectors. Xu et al (67) reported that the bone marrow X kinase-Rho GTPase activating protein fusion proteins activate Janus kinase/STAT3, bypassing canonical Wnt signaling to sustain GCSC survival. These findings suggest pathway plasticity, where GCSCs utilize both canonical and non-canonical Wnt signaling to adapt to therapeutic pressures.

Epigenetic modifiers and miRNAs tightly regulate GCSC plasticity. Wu et al (68) revealed that the miR-19b/20a/92a cluster maintains GCSC self-renewal by targeting PTEN and activating AKT/mTOR signaling. By contrast, Han et al (69) showed that atonal transcription factor 1 (ATOH1) induces GCSC differentiation by suppressing Notch1, highlighting a tumor suppressive role of transcription factors. The detailed mechanism involves the direct interaction of ATOH1 with the Notch1 signaling pathway. Notch1 is a well-known promoter of stemness and tumorigenicity in various cancer types, including gastric cancer. By inhibiting Notch1, ATOH1 effectively reduces the self-renewal and proliferative capacity of cancer stem cells, thus exerting a tumor suppressive effect. Notably, Shen et al (70) identified one cut homeobox 2 as a miR-15a-5p target that amplifies GCSC stemness through SOX2 upregulation. Such studies underscore the dual role of epigenetic regulators in either sustaining or suppressing GCSC phenotypes, depending on cellular context.

Autophagy enables GCSCs to survive stress and resist therapy. Li et al (71) linked Notch-induced autophagy to chemoresistance, where inhibition of autophagy sensitizes GCSCs to 5-fluorouracil (5-FU). Xin et al (72) further demonstrated that methionine represses autophagy by methylating RAB37, enhancing GCSC survival under nutrient deprivation. Recent work by Togano et al (73) corroborated these findings, showing that autophagy inhibitors impair GCSC viability in hypoxic niches. Conversely, Zhang et al (50) reported that PGD2/PTGDR2 signaling suppresses GCSC stemness by promoting ATG4B ubiquitination and inhibiting autophagic flux. These contrasting roles of autophagy emphasize the need for context-specific therapeutic targeting.

GCSCs exploit microenvironmental cues to evade immune surveillance and foster metastasis. Yang et al (74) revealed that HIF-1α induces EMT in GCSCs via Snail upregulation, facilitating dissemination under hypoxia. Sun et al (75) implicated HER2 in GCSC invasion, where HER2 inhibition reduces tumorigenicity in patient-derived xenografts. Additionally, Seeneevassen et al (76) identified leukemia inhibitory factor as a Hippo pathway activator that suppresses GCSC tumorigenicity, suggesting a dual role for cytokines in niche regulation. Recent studies also highlight the influence of the microbiome. PGD2/PTGDR2 signaling, downregulated in gastric cancer tissues, restricts GCSC self-renewal by inhibiting STAT3 phosphorylation (50,77), aligning with findings from GCSC models.

GCSCs employ diverse mechanisms to resist conventional therapies. Wang et al (78) showed that docetaxel/cisplatin/5-FU regimens fail to eradicate ALDH1⁺ GCSCs due to upregulated ABC transporters. Zhang et al (79) uncovered exosomal long non-coding RNA FERO as a ferroptosis suppressor, linking chemotoxicity to GCSC stemness preservation. Novel strategies targeting metabolic vulnerabilities, such as inhibition of stearoyl-CoA desaturase 1 to induce ferroptosis via cholesterol-mTOR axis disruption (80), show promise. Similarly, Ni et al (81) demonstrated that Celastrus orbiculatus extract suppresses GCSCs through TGF-β/Smad pathway inhibition.

CCSCs are pivotal drivers of tumor initiation, progression, chemoresistance and metastasis. The unique biological properties and regulatory mechanisms of CCSCs have been extensively studied, revealing complex interactions between intrinsic signaling pathways, epigenetic modifications and the TME (Table III). In this subsection, the key mechanisms underlying CCSC functionality, supported by representative studies, are summarized and their implications for therapeutic targeting are discussed.

The Wnt/β-catenin pathway is a central regulator of CCSC self-renewal and chemoresistance. Chen et al (82) demonstrated that miR-199a/b upregulation in ALDH1⁺ CCSCs activates Wnt/β-catenin signaling, enhancing ATP-binding cassette subfamily G member 2 (ABCG2)-mediated drug efflux and cisplatin resistance. Similarly, Li et al (83) identified that lysine-specific demethylase 3 epigenetically activates Wnt/β-catenin by removing repressive H3K9 methylation marks, thereby promoting stemness and tumorigenicity. These findings align with Hua et al (84), who showed that Tribbles pseudokinase 3 stabilizes β-catenin/T-cell factor 4 complexes, amplifying stemness and EMT in CCSCs. However, discrepancies arise in the role of downstream effectors. While Yu et al (85) implicated Special AT-rich sequence-binding protein 2 as a Wnt-driven transcriptional coactivator, Zhu et al (86) highlighted SOX2-mediated β-catenin/Beclin1 crosstalk in chemoresistance, suggesting pathway plasticity across CCSC subpopulations.

CCSCs evade chemotherapy through dynamic post-translational modifications. Izumi et al (87) revealed that F-box and WD repeat domain-containing 7 (FBXW7), an E3 ubiquitin ligase, is downregulated in CCSCs, leading to c-Myc stabilization and enhanced survival under 5-FU treatment. Conversely, Honma et al (88) reported that FBXW7 upregulation degrades pro-survival proteins, sensitizing CCSCs to oxaliplatin. This paradox may reflect context-dependent roles of FBXW7 in different chemotherapeutic regimens. Epigenetically, Mukohyama et al (89) identified miR-221 as a driver of chemoresistance by targeting Quaking homolog, KH domain RNA-binding protein (QKI), a tumor suppressor that restrains CCSC proliferation. These studies underscore the need for personalized strategies targeting ubiquitination or miRNA networks.

CCSCs exhibit metabolic flexibility to survive hostile microenvironments. Liu et al (48) demonstrated that lactate, a byproduct of glycolysis, induces PGC-1α-dependent oxidative phosphorylation in normoxic CCSCs, facilitating liver metastasis. Conversely, Guo et al (49) found that vitamin D triggers ferroptosis in CCSCs by suppressing SLC7A11, a cystine transporter critical for redox balance. These contrasting mechanisms highlight the dual role of metabolism in CCSC survival and vulnerability. Additionally, hypoxia-driven immune evasion was explored by Geng et al (47), in which it was observed that CCSCs upregulated FADS1/DDA to impair NK cell cytotoxicity, a process exacerbated by hypoxic conditions.

The niche surrounding CCSCs notably influences their behavior. Liu et al (90) showed that CAFs secrete exosomes enriched with TGF-β1, which enhance CCSC radioresistance via Smad-dependent stemness pathways. Montalbán-Hernández et al (91) further revealed that CCSCs fuse with monocytes to form hybrid cells capable of immune evasion and metastatic dissemination. These findings align with those of Cavallucci et al (92), who identified Fusobacterium nucleatum as a pathobiont that directly activates pro-inflammatory and stemness pathways in CCSCs, emphasizing the role of the microbiome in CCSC regulation.

EMT is a hallmark of CCSC plasticity. Tamura et al (93) linked E-cadherin loss to Nanog upregulation, which drives CCSC proliferation. Similarly, Zou et al (94) demonstrated that CD44 knockdown suppresses EMT and invasiveness by inhibiting Snail and Twist family bHLH transcription factor 1. Immune evasion mechanisms were highlighted by Vishnubalaji et al (95), where ALDH1⁺ CCSCs were shown to activate MAPK and FAK pathways to resist oxidative stress and immune surveillance. These studies collectively suggest that targeting EMT regulators or immune checkpoints could disrupt CCSC dissemination.

CSC-associated markers serve as robust predictors of patient survival and treatment response. In esophageal cancer, CD44 and CD133 expression is associated with resistance to neoadjuvant chemotherapy and poor survival. Specifically, Agawa et al (96) demonstrated that CD44⁺/CD133⁺ CSCs in ESCC predict poor pathological response to cisplatin/5-FU regimens, with a 3-year survival rate of 28 vs. 72% for marker-negative patients. Similarly, Claudin 4 (CLDN4)-high ESCC cells exhibit stem-like properties and resistance to concurrent chemoradiotherapy, as shown by Lin et al (97), linking CLDN4 to reduced OS [hazard ratio (HR), 2.1; P=0.003]. Meta-analyses by Trevellin et al (98) further confirmed that CSC markers (such as ALDH1 and CD44) are consistently associated with shorter DFS across esophageal and gastric cancer.

In gastric cancer, ALDH1⁺ CSCs are linked to chemoresistance and recurrence. Nishikawa et al (99) reported that ALDH1⁺ cells survive 5-FU treatment due to upregulated ABCG2, with ALDH1 positivity correlating with a 2.3-fold increased risk of relapse. Gong et al (100) identified epithelial cell transforming 2 (ECT2) as a novel prognostic marker, where high ECT2 expression promotes stemness via Wnt/β-catenin signaling and predicts poor differentiation (HR, 1.9; P=0.01). Circulating GCSCs detected in peripheral blood are also correlated with advanced tumor stage and metastasis (P<0.001) (101).

For colorectal cancer, Catalano et al (102) revealed that thyroid hormone receptor activation reduces CSC viability by suppressing the Wnt and bone morphogenetic protein 4 pathways, improving survival in patients with low Nanog expression (HR, 0.6; P=0.04). Conversely, Prieur et al (103) demonstrated that anti-progastrin antibodies targeting Wnt-driven CSCs enhance chemosensitivity in KRAS-mutated colorectal cancer, with a 40% reduction in tumor sphere formation.

CSCs evade therapy through intrinsic and extrinsic mechanisms. In esophageal cancer, ABT-263 (a BCL-2 inhibitor) synergizes with chemotherapy by depleting CSCs via apoptosis induction, achieving a 60% reduction in tumor volume in xenografts (104). Nanog, a stemness marker, mediates resistance to cisplatin; iron chelators targeting Nanog reduce chemoresistance by 50% in vitro (105).

GCSCs employ metabolic adaptations and epigenetic reprogramming. Xu et al (106) showed that 5-FU enriches ALDH1⁺ CSCs through reactive oxygen species-mediated autophagy, while methionine restriction sensitizes CSCs by inhibiting RAB37 methylation (107). HNRNPA2B1, an RNA-binding protein, stabilizes c-Myc mRNA in CSCs, and its knockdown restores oxaliplatin sensitivity (P<0.01) (108).

In colorectal cancer, dual PI3K/mTOR inhibitors induce the differentiation of CD133⁺ CSCs, reducing tumorigenicity by 70% (109). CD44⁺/CD133⁺ CSCs are paradoxically sensitive to trifluridine (110), suggesting metabolic vulnerabilities. However, sphingosine kinase 1/HIF-1 axis inhibition (111) and aurora kinase A/YAP1) targeting (112) overcome microenvironment-driven resistance, suppressing metastasis by 45-60%.

Verteporfin, a YAP1/TAZ-TEAD inhibitor, suppresses GCSC tumorigenicity by disrupting transcriptional activity, leading to reduced spheroid formation and metastasis (121) (Fig. 2). This aligns with findings in colorectal cancer, where tankyrase inhibitors downregulate YAP1-associated c-KIT, impairing CCSC viability (122). However, YAP1 crosstalk with other pathways (such as PI3K/AKT) may necessitate dual targeting to prevent resistance.

Src kinase inhibition blocks GCSC proliferation and EMT by suppressing STAT3 and AKT phosphorylation (123) (Fig. 2). Similarly, in CRC, MEK inhibitors combined with CSC-targeting agents overcome resistance by depleting ALDH1⁺ CCSCs (124). These findings emphasize the potential of kinase inhibitors in disrupting CSC survival networks. The pan-BCL-2 inhibitor, navitoclax (ABT-263), synergizes with chemotherapy in gastroesophageal carcinoma, inducing apoptosis in CD44⁺ CSCs and reducing tumor recurrence (125). However, heterogeneity in BCL-2 family expression across CSC subpopulations may limit efficacy, as observed in colorectal cancer models (126).

Combining CSC-targeted agents with conventional therapies improves outcomes by addressing bulk tumors and residual CSCs (Fig. 2). For example, salinomycin-loaded carbon nanotubes selectively kill GCSCs while sparing normal cells, enhancing the efficacy of 5-FU (127). In colorectal cancer, polymeric micelles targeting CD44v6 deliver niclosamide, synergizing with oxaliplatin to reduce CCSC-driven metastasis (128). Similarly, mithramycin A represses ABCG2 in CRCSCs, sensitizing them to chemotherapy (129). In addition, aptamer-mediated survivin knockdown in CCSCs enhances 5-FU-induced apoptosis and reduces immune evasion (130). Furthermore, inhibition of cholesterol synthesis in GCSCs attenuates NK cell evasion, suggesting a role for metabolic-immune crosstalk in combinatorial regimens (131).

CSCs serve pivotal roles in the initiation, progression and therapeutic resistance of digestive tract tumors. The heterogeneity and plasticity of these cells pose significant challenges for effective treatment (132). Emerging therapeutic strategies targeting CSC-specific pathways, immune modulation and combination therapies offer promising avenues to overcome resistance and enhance patient outcomes (133). Future research should focus on integrating biomarker-driven approaches and innovative technologies to advance precision medicine in the field of digestive oncology.

Despite promising preclinical data, the clinical translation of CSC-targeted therapies remains challenging. For example, vismodegib, a Hedgehog inhibitor, failed to improve survival in gastric cancer trials due to compensatory YAP1 activation in residual CSCs (134,135). Similarly, Wnt inhibitors such as PRI-724 have shown limited efficacy against colorectal cancer, as CAF-derived IL-6 reactivates β-catenin via STAT3 (136). These failures highlight the need to target CSC-stroma crosstalk. Emerging strategies such as microbiota modulation (including F. nucleatum eradication) and metabolic-immune combinations (including vitamin D + anti-PD-1) may overcome these hurdles by simultaneously disrupting CSC niche support and immune evasion (137,138).

Identifying and targeting CSCs is technically challenging due to their heterogeneity and plasticity. Universal markers such as CD44 and CD133 show variable expression across tumor subtypes, necessitating context-specific validation (139). Furthermore, CSC subpopulations exhibit divergent responses to therapies. For instance, delta-like protein 1 inhibition suppresses Wnt-driven CCSCs but spares LGR5⁺ subsets (140), underscoring the need for multitargeted strategies. Single-cell sequencing and spatial transcriptomics may elucidate clonal dynamics, guiding personalized therapies. Furthermore, the TME shields CSCs via cytokine loops and stromal support. Blocking proprotein convertases disrupts GCSC-TME crosstalk, reducing invasiveness (141). Similarly, ruthenium-xanthoxylin complexes target HSP90 in CRCSCs, thereby overcoming stroma-mediated resistance (142). Advances in genomic and proteomic technologies have enabled tailored interventions against CSC heterogeneity in digestive tract tumors.

However, preclinical models have inherent limitations. Although animal models provide mechanistic insights, they often fail to recapitulate human tumor complexity and microenvironmental interactions, leading to clinical discrepancies (143). Additionally, tumor heterogeneity and CSC plasticity complicate therapy development, as CSCs can adapt to microenvironmental changes and therapeutic pressures. The TME, which is crucial for CSC support, is difficult to target without affecting normal tissue homeostasis (144). Furthermore, a lack of robust biomarkers for patient stratification and treatment response assessment hinders the development of effective CSC-targeted therapies.

Nanomaterials and CRISPR-based approaches offer precision in future treatment. For example, SPION-driven atranorin induces ferroptosis in GCSCs by modulating Xc-/GPX4 (6,145). Additionally, low-dose vitamin C promotes CCSC differentiation via β-catenin membrane retention, a strategy that is compatible with immune checkpoint inhibitors (146). Although preclinical data are promising, clinical trials remain limited. The active hexose correlated compound/epigallocatechin gallate combination reduces LGR5⁺ CCSCs in early-phase studies (147); however, its scalability and toxicity require further evaluation. Future research should prioritize biomarkers (such as ALDH1 and CD44) to stratify patients and optimize trial designs.

In addition, the application of artificial intelligence (AI) in predicting the vulnerabilities of CSCs has recently been demonstrated, providing new ideas and methods for the development of targeted therapies against CSCs (148). For example, deep learning models have been employed to analyze large-scale CSC genomic and transcriptomic data, aiming to identify potential therapeutic targets and drug candidates. These models can predict the drug sensitivity and resistance of CSCs, thereby offering a theoretical basis for drug repurposing (149). For example, studies have utilized deep learning algorithms to analyze the gene expression profiles of CSCs from various digestive tract tumors and to predict their responses to different drugs (150,151). The results have shown that some conventional drugs may have potential inhibitory effects on CSCs. Further experimental verification is ongoing to explore the feasibility of these drugs in clinical applications. This integration of AI with CSC research not only increases the efficiency of target discovery but also optimizes therapeutic strategies, offering new avenues for the individualized treatment of digestive tract tumors.

Finally, patient-derived organoids (PDOs) have garnered significant attention as promising tools for personalized therapy screening. PDOs are three-dimensional cellular models cultured from patient tumor tissues (152). PDOs retain the histological structure and genetic characteristics of the original tumor to some extent, making them powerful tools for evaluating drug efficacy and guiding clinical treatment. A previous study has shown that PDOs can effectively mimic the TME and CSC niche of digestive tract tumors (153). By screening various drugs using PDOs, researchers can identify therapeutic regimens that are effective against CSCs and the bulk tumor cells, providing a basis for personalized treatment. For example, researchers successfully established PDO models from patients with gastric or pancreatic ductal adenocarcinoma, and drug screening experiments revealed that certain drugs could specifically targeted CSCs in the PDOs, inhibiting tumor growth and metastasis (154,155). These findings demonstrate the potential of PDOs to help predict therapeutic responses and guide clinical treatment. The application of PDOs in personalized therapy screening offers a new approach for improving the treatment outcomes of digestive tract tumors.

In conclusion, although significant progress has been made in understanding the biology of CSCs and their role in digestive tract tumors, several challenges remain in translating this knowledge into clinical practice. Future research should focus on addressing the limitations of current CSC markers, reconciling discrepancies between preclinical and clinical findings and developing strategies to overcome CSC plasticity and microenvironment-driven resistance. By integrating innovative technologies and biomarker-driven approaches, the field of CSC-targeted therapies can be advanced and precision medicine in the field of digestive oncology can be improved.