Targeting tumor dormancy: the next frontier in gastrointestinal stromal tumor therapy.

Introduction
Gastrointestinal stromal tumors (GISTs), the most prevalent mesenchymal neoplasms of the gastrointestinal tract, originate from precursors of the interstitial cells of Cajal (ICC). Their incidence exhibits a rising trend [1,2]. The anatomical distribution varies significantly: the stomach is the most common site (∼60%), followed by the small intestine—including the jejunum and ileum (∼30%); less frequent sites include the duodenum (∼5%), rectum (∼3%), colon (∼1%), and esophagus (<1%), with rare cases arising in the omentum/mesentery or presenting as disseminated tumors without a known primary (Fig. 1) [3]. Autopsy studies reveal microGISTs (diameter <1 cm) in up to 20% of individuals [4,5]. Although typically indolent, this suggests that the true incidence of GIST far exceeds the clinical detection rate. Surgery remains the primary therapy for localized GISTs. However, approximately 30% of patients are ineligible due to tumor location or metastasis, and postoperative recurrence rates approach 40% [6]. As the most common soft tissue sarcoma (constituting ∼20%) [7], GISTs display inherent resistance to conventional chemotherapy. The discovery of activating c-KIT mutations in 1998 [8] and the successful treatment of the first metastatic GIST patient with imatinib (IM) in 2000 [9] ushered in a new era of targeted therapy with clinical trials confirming that IM prolongs median survival in advanced GIST from 18 months to over 5 years [[10], [11], [12], [13]].
Despite the transformative impact of imatinib, durable eradication of GIST remains uncommon. This limitation is partly attributable to the marked molecular heterogeneity of GIST, which shapes therapeutic sensitivity and resistance. KIT mutations are identified in approximately 60–70% of cases, PDGFRA mutations in 10–15%, and the remaining tumors are wild-type, often harboring alterations in SDH, RAS pathway components, NF1, or other genes (Fig. 1). Consistent with this heterogeneity, responses to IM vary substantially across molecular subtypes: a subset of patients exhibit primary resistance, exemplified by PDGFRA D842V-mutant tumors, whereas many initially responsive tumors eventually acquire secondary resistance through additional KIT or PDGFRA mutations [14]. However, clinical evidence suggests that resistance alone does not fully explain disease persistence. In the PERSIST-5 trial, recurrence during imatinib exposure was rare in patients with imatinib-sensitive disease, whereas most recurrences occurred after treatment discontinuation [15]. Similarly, the IMADGIST study showed that extending adjuvant imatinib to 6 years further reduced relapse risk compared with 3 years [16]. In advanced disease, the BFR14 phase III trial demonstrated that interruption of IM, even after prolonged disease control, rapidly led to re-progression in most patients [17]. Collectively, these findings suggest that, in many patients, residual GIST cells can survive in a clinically occult state during TKI therapy and subsequently re-expand once therapeutic pressure is withdrawn [[15], [16], [17], [18]].
This clinical behavior closely aligns with the concept of tumor dormancy. Tumor dormancy, a critical pathway to resistance, was first conceptualized by Willis in 1934 to explain the delayed metastasis that occurs after primary tumor resection [19]. Recent research delineates two distinct modes: Mass dormancy, where tumor bulk is maintained in dynamic equilibrium via angiogenesis suppression and immune surveillance; and Cellular dormancy, where individual tumor cells enter G0/G1 arrest, governed by multiple pathways [[20], [21], [22]]. In parallel, cancer stem cells (CSCs) are increasingly linked to dormancy, providing a cellular basis for therapy resistance and relapse. In GIST, dormancy is not merely a theoretical framework but a clinically meaningful explanation for several recurrent patterns: late postoperative relapse, recurrence clustering after cessation of adjuvant imatinib, incomplete eradication despite prolonged disease control, and eventual progression driven by persistent drug-tolerant cell populations. Recent studies have begun to characterize dormancy and stem-like/CSC-like biology in GIST, although definitive evidence that imatinib-residual cells are bona fide CSCs remains lacking [[23], [24], [25], [26]].
This review systematically synthesizes the core mechanisms of tumor dormancy and their specific regulatory pathways in GIST, integrating recent advances in GIST stem-like/CSC-like persistent-cell research. Building upon this mechanistic foundation, we propose novel therapeutic strategies targeting dormancy to overcome IM resistance.

Molecular heterogeneity and mutational evolution in GIST
Under the umbrella term of GIST, multiple molecularly defined subtypes are currently recognized. While most GISTs harbor activating KIT mutations, and a smaller proportion are driven by PDGFRA alterations, KIT/PDGFRA-wild-type GIST represents a biologically heterogeneous group that includes SDH-deficient GIST, NF1-associated GIST, RAS/BRAF-pathway–altered tumors, and rare cases with other actionable events such as gene fusions. These subtypes differ in pathogenesis, signaling dependencies, clinicopathological features, and therapeutic vulnerabilities [27,28]. Therefore, GIST should not be regarded as a single uniform KIT-driven disease, and the biological basis of tumor persistence may likewise vary across distinct molecular subgroups [27].
In advanced GIST, secondary resistance mutations constitute one of the best-established mechanisms of acquired resistance to tyrosine kinase inhibitors [29,30]. These mutations usually emerge in the kinase domain of KIT, and less frequently PDGFRA, under the selective pressure of treatment [29,31]. Mechanistically, secondary mutations may involve the ATP-binding pocket or the activation loop, thereby altering kinase conformation and reducing drug binding [30,31]. Clinically, such mutational heterogeneity is highly relevant because it contributes to inter- and intralesional polyclonal resistance and influences the activity of later-line TKIs. Accordingly, secondary mutations are central to the contemporary understanding of disease progression in advanced GIST [[31], [32], [33]].
Nevertheless, a mutation-centered model alone cannot fully explain all clinically observed patterns of GIST persistence. Some patients maintain low-volume residual disease for prolonged periods during therapy, whereas others experience rapid tumor regrowth after treatment discontinuation, suggesting that non-genetic persistence states may coexist with genetic resistance [34]. In addition, under sequential therapeutic pressure, more complex later-line mutational evolution, including compound or tertiary-like resistance alterations, may further increase biological heterogeneity, although this area remains less systematically characterized than classical secondary mutations in GIST [33,34]. These limitations highlight an important conceptual gap: while secondary and later-line mutations explain acquired drug resistance, they do not fully explain how residual tumor cells survive in a clinically occult state. In this context, tumor dormancy provides a complementary framework for understanding persistent disease, delayed relapse, and incomplete tumor eradication in GIST.

Overview of tumor dormancy
Tumor dormancy represents a critical biological phenomenon where neoplasms enter a state of growth arrest under specific conditions, closely associated with treatment resistance, recurrence, and metastasis. It operates through two synergistic modes—mass dormancy and cellular dormancy—governed by angiogenesis regulation, immune surveillance, cell-autonomous adaptations, and extracellular matrix (ECM) remodeling. The following sections detail the three core mechanisms: angiogenic dormancy, immune-mediated dormancy, and cellular dormancy (Fig. 2), with ECM remodeling and mechanotransduction discussed in a dedicated subsequent section.

Angiogenic dormancy
Angiogenic dormancy is a key state where solid tumors are restrained in a proliferation-death equilibrium due to insufficient vascularization. Solid tumors initially exist as avascular cell aggregates, their growth strictly dependent on the formation of new vascular networks [35,36]. When tumors fail to recruit adequate vasculature, internal cells undergo apoptosis from hypoxia and nutrient deprivation, ultimately maintaining size homeostasis – a phenomenon more pronounced in larger tumors [37]. Tumor angiogenesis proceeds through sprouting angiogenesis, intussusceptive angiogenesis, and vasculogenesis (Fig. 2) [[38], [39], [40], [41]]; dysregulation at any stage can perpetuate dormancy [42].
Maintenance of this dormant state is bidirectionally regulated by pro- and anti-angiogenic factors. In several non-GIST tumor models, endogenous inhibitors such as thrombospondin-1 (TSP-1) help preserve vascular quiescence, whereas pro-angiogenic mediators including VEGF and epoxyeicosatrienoic acids (EETs) can promote dormancy escape [43,44]. Physical properties of the tumor microenvironment (TME) also modulate angiogenic dormancy. Extracellular matrix stiffness regulates endothelial tip cell formation via the p-PXN-Rac1-YAP signaling axis [45]. Immune cell-endothelial cell (EC) interactions provide novel insights: tumor-associated macrophages (TAMs) secrete WNT7b, activating β-catenin signaling in ECs, elevating VEGFA expression, and amplifying pro-angiogenic networks; conversely, Tregs secrete TGF-β1, inducing EC expression of p21 and p16, accelerating EC senescence, and actively maintaining vascular dormancy [46]. The clinical relevance of angiogenic dormancy is validated by metronomic chemotherapy, which sustains dormancy by upregulating TSP-1 and inhibiting VEGF and pro-angiogenic cells [47,48]. However, whether these specific mechanisms directly govern angiogenic dormancy in GIST remains unknown. In GIST, currently available evidence more directly supports a role for angiogenic activity in tumor progression rather than a formally demonstrated angiogenic dormancy program. VEGF expression has been reported to correlate with microvessel density and adverse biological behavior in GIST, and high intratumoral VEGF expression in advanced GIST independently predicts early treatment failure and poor survival during imatinib therapy [49,50]. These findings suggest that angiogenic signaling may influence the transition between tumor restraint and progression in GIST, but the existence of a bona fide angiogenic dormancy state in GIST still requires direct mechanistic validation.

Immune-mediated dormancy
The immune system's role in suppressing tumor progression through surveillance is well-established; its critical function in sustaining tumor dormancy is increasingly recognized [20,51,52]. Direct evidence comes from organ transplantation: dormant donor-derived cancer cells can reactivate upon recipient immunosuppression, as exemplified by melanoma transmission via renal transplantation [[53], [54], [55], [56]]. Recent research further reveals that diverse immune cells within the TME regulate dormancy homeostasis through complex networks. CD8⁺ T cells are essential for maintaining disseminated tumor cell (DTC) dormancy; their depletion accelerates lung metastasis in melanoma models [57], with analogous observations in lymphoma [58] and breast cancer [59,60]. CD4⁺ T cells can induce anti-angiogenic factors (e.g., CXCL9, CXCL10) and suppress αvβ3 integrin expression, blocking tumor angiogenesis and cooperatively sustaining angiogenic dormancy [61]. Notably, dormant sarcoma tissues exhibit a significantly elevated CD8⁺/CD4⁺ T cell ratio [62,63], suggesting a dominant role for CD8⁺ T cells. Recent findings indicate that combining cisplatin and temozolomide induces "hypermutation" in tumor cells, generating numerous neoantigens that markedly enhance CD8⁺ T cell recognition and elimination, potentially offering a novel strategy against poorly immunogenic dormant cells [64].
Natural killer (NK) cells play a selective role in organ-specific dormancy. Within the liver microenvironment, NK cells maintain triple-negative breast cancer cell dormancy via IFN-γ signaling, their abundance correlating positively with dormancy duration [65]. TAMs exhibit functional heterogeneity: M1-like TAMs promote dormancy maintenance by recruiting CD8⁺ T cells, whereas M2-like TAMs secrete immunosuppressive factors (e.g., IL-10/TGF-β), subverting immune surveillance and facilitating dormancy escape [22,66]. Emerging research elucidates a key role for neutrophils in chemotherapy-induced dormancy escape: He et al. demonstrated that chemotherapeutics like doxorubicin induce fibroblast senescence in lung tissue, releasing senescence-associated secretory phenotype (SASP) factors that stimulate neutrophil extracellular trap (NET) formation. NETs, composed of DNA-histone fibers, remodel the ECM, activating integrin-FAK signaling and ultimately awakening dormant DTCs [67].
In GIST, immune regulation is likely to be relevant to tumor persistence, but direct proof that immune surveillance maintains a dormant residual-cell state is still limited. What is better established is that the GIST immune microenvironment is biologically active and closely linked to treatment response. Imatinib has been shown to potentiate antitumor T-cell responses in GIST by inhibiting indoleamine 2,3-dioxygenase (IDO), thereby activating intratumoral CD8+ T cells and inducing apoptosis of regulatory T cells [68]. More recent single-cell analyses of advanced GIST further revealed that imatinib-resistant lesions are enriched in immunosuppressive Treg-associated interactions and IDO1+ dendritic-cell populations, supporting the view that immune remodeling accompanies disease persistence and resistance in GIST [69]. These data do not yet establish immune-mediated dormancy in the strict mechanistic sense, but they strongly suggest that immune surveillance and immune escape may influence whether residual GIST cells remain clinically restrained or progress [[68], [69], [70]].

Cellular dormancy
Cellular dormancy represents the classical mode of tumor dormancy, with mechanisms more complex than mass dormancy. Maintenance of the dormant state typically relies on intrinsic alterations within the tumor cell itself, coupled with extrinsic regulation by the surrounding microenvironment [47,[71], [72], [73]].
Intrinsic regulation of cellular dormancy encompasses genetic alterations, autophagy, intracellular signaling, epigenetic modifications, and metabolic reprogramming [20,71,74]. The balance of key signaling pathways is crucial: ERK primarily promotes proliferation, while p38 often induces growth arrest. Consequently, a low ERK/p38 signaling ratio (ERKlow/p38high) is widely recognized as a core mechanism for maintaining tumor cell quiescence [[75], [76], [77]]. The dormant state of DTCs represents a recognized reservoir for breast cancer recurrence. Studies show that the Fbxw7 gene is indispensable for maintaining dormancy in breast cancer DTCs; its ablation disrupts quiescence and enhances DTC sensitivity to paclitaxel [78]. Notably, low Fbxw7 expression correlates significantly with malignant progression in GIST and is an important independent prognostic factor for 10-year recurrence-free survival [79], suggesting a pivotal role for Fbxw7 levels in GIST dormancy/awakening regulation. Autophagy, by degrading damaged organelles, misfolded proteins, and cytoplasmic components, provides energy for stressed cells to sustain survival [80]. In breast cancer, autophagy promotes recurrence by supporting the survival of dormant tumor cells [81,82]. Autophagy-related gene 7 (ATG7) is essential for autophagy activation; inhibiting autophagy in dormant cells induces mitochondrial dysfunction and reactive oxygen species (ROS) accumulation, ultimately triggering apoptosis [82]. Furthermore, dormant tumor cells adapt to energy demands via metabolic reprogramming; detailed mechanisms can be found in other excellent reviews [71,83].
Extrinsic regulation of cellular dormancy is primarily mediated by diverse components within the TME. Beyond immune cells, other non-malignant cells and ECM constituents are deeply involved. Hypoxia is a fundamental environmental factor inducing tumor cell dormancy [[84], [85], [86], [87]]. For instance, in salivary adenoid cystic carcinoma (SACC), the miR-922/DEC2 axis is essential for hypoxia-induced cellular dormancy, a process linked to lipid metabolic reprogramming [84]. Beyond immune cells, stromal populations such as cancer-associated fibroblasts (CAFs) and ECM components are increasingly recognized as critical regulators of tumor cell quiescence [[88], [89], [90]]. CAFs typically exhibit pro-tumorigenic properties, whereas normal fibroblasts may display tumor-suppressive capabilities. In a melanoma mouse model, dermal fibroblasts suppressed tumor growth by downregulating cyclin D1 and upregulating the cyclin-dependent kinase inhibitor p16 [91]. In GIST, emerging evidence indicates that fibroblast populations within the tumor stroma actively participate in tumor progression and therapeutic resistance. Recent studies have shown that CAFs derived from GIST tissues can secrete platelet-derived growth factor C (PDGFC), which activates PDGFRA signaling in tumor cells and promotes tumor growth and metastatic potential. These findings suggest that fibroblast-mediated paracrine signaling may contribute to shaping the biological behavior of GIST within the stromal niche [92]. Furthermore, CAF-derived cytokines such as TGF-β have been implicated in metabolic reprogramming and drug resistance, highlighting a potential role for stromal fibroblasts in regulating the survival of therapy-tolerant or dormant tumor cells [93]. Although direct mechanistic studies of CAF-regulated dormancy in GIST remain limited, insights from other tumor types suggest potential regulatory pathways. For example, CAF-derived chemokines such as CXCL1 have been reported to disrupt DEC2-mediated dormancy programs in oral squamous cell carcinoma cells, while fibroblast-mediated ECM remodeling can modulate integrin signaling pathways that influence tumor cell quiescence [[94], [95], [96]]. These cross-tumor observations provide a conceptual framework for understanding how stromal fibroblasts may regulate dormancy-associated phenotypes in GIST.
Notably, cellular dormancy is not a binary state but rather a continuum of growth-arrested phenotypes—including deep quiescence, therapy-tolerant persisters, and senescence-like programs—that differ in reversibility and therapeutic vulnerability [97]. For GIST, this nuance is clinically relevant: IM-induced residual cells may occupy multiple arrested states (e.g., autophagy-dependent quiescent cells versus inflammation-prone persisters), and single-cell profiling approaches will be essential to map these hierarchies and identify actionable targets for dormancy-directed therapy.

Microenvironmental axis in tumor dormancy: the role of ECM remodeling and mechanotransduction
Beyond angiogenic and immune dormancy, the ECM and tissue mechanics constitute a "microenvironmental axis" integrating inflammation, stromal remodeling, and cell-intrinsic signaling to modulate dormancy [98,99]. ECM remodeling, particularly in response to chronic inflammation, plays a critical role in regulating tumor cell fate. For example, recent NET-focused reviews synthesize evidence that NET-associated proteases can remodel laminin and trigger integrin-dependent signaling programs that promote dormant-cell reactivation and metastatic outgrowth [100,101]. Importantly, mechanical forces themselves are bidirectionally linked to dormancy. Under low-stress ECM conditions, integrin β1/β3 activation promotes stem-like traits and drug tolerance, as well as mitochondrial fission, elevated ROS, and NRF2-dependent antioxidant responses—all characteristics associated with therapeutic escape and quiescence. Conversely, excessive mechanical forces in a fibrotic ECM can induce dormancy in stem-like cancer cells via DDR2/STAT1/p27 signaling, with dormancy being reversible upon removal of mechanical constraints [102]. Given that GIST lesions are often found in the peristaltic, innervated, and fibrosis-prone gastrointestinal wall, with frequent recurrence in sites such as the liver or peritoneum (which exhibit distinct ECM composition and inflammatory trajectories), incorporating mechanobiology into GIST dormancy models is crucial to understand how these factors may drive late relapse.
Organ-specific ECM cues encode dormant programs through niche-dependent receptor–matrix interactions. Perivascular endothelial Wnt signals can epigenetically enforce dormancy [103], while niche-specific ligands such as laminin-211 sequester YAP via dystroglycan receptors [104]. Conversely, fibrotic remodeling—such as collagen I deposition—reactivates dormant cells through integrin β1/SRC/ERK signaling, and integrin cooperation, along with ROCK-mediated tensional regulation, maintains dormancy by shaping ECM adhesiveness [105,106]. These findings suggest that the GIST field might also benefit from exploring a similar ECM-centered perspective. Mapping integrin repertoires, YAP/TAZ activity, collagen and fibronectin dynamics, and NET-driven matrix cleavage may help reveal actionable vulnerabilities (e.g., FAK inhibition, YAP modulation, or anti-fibrotic and anti-NET strategies) that could ultimately inform efforts to prevent dormancy escape and relapse.

Dormancy regulation mechanisms in GIST
At present, direct mechanistic evidence for dormancy in GIST is derived predominantly from KIT-driven and imatinib-exposed models, whereas comparable studies in SDH-deficient, NF1-related, and other rare molecular subtypes remain limited. This likely reflects the rarity of these subgroups and the lack of representative experimental systems, rather than excluding a role for dormancy biology in these settings.

Cell cycle signaling pathways
In 2008, a landmark Cancer Research study first elucidated that imatinib induces GIST cell dormancy or quiescence via the APC/C(Cdh1)-SKP2-p27Kip1 signaling axis [23]. In this pathway, CDH1 promotes SKP2 degradation [[107], [108], [109]], thereby stabilizing the cyclin-dependent kinase inhibitor p27Kip1 [[110], [111], [112], [113], [114], [115], [116], [117]] and targeting cell cycle-promoting proteins such as Cyclin A, collectively maintaining G0/G1 arrest [23].
Investigating IM's mechanism, researchers found that treating GIST 882 cells with IM significantly downregulated SKP2 and Cyclin A protein levels while upregulating p27Kip1, indicating exit from the cell cycle into quiescence. Mechanistically, IM triggers CDH1 nuclear translocation via inhibition of the KIT/PDGFRA downstream PI3K/AKT pathway (not MAPK), thereby activating the APC/C(Cdh1) complex. This was confirmed pharmacologically: the PI3K inhibitor LY294002 replicated these effects, whereas the MEK1/2 inhibitor U0126 did not [23]. In 2013, the same team further identified the DREAM complex as central to IM-induced quiescence(Fig. 3) [118]. The DREAM complex (p130/E2F4/5/DP/MuvB core) drives and maintains G0 phase quiescence by binding and repressing promoters of cell cycle genes (e.g., CCNA2, CDC25A). Conversely, during proliferation, the MuvB core forms the MMB complex with B-MYB/FOXM1 to regulate S and G2/M phase gene expression [24,[119], [120], [121]]. IM was shown to induce p130 upregulation, enhanced p130/E2F4/LIN37 complex assembly, and LIN52-Ser28 phosphorylation in vitro and in vivo. Notably, these changes were fully reversible upon drug withdrawal, indicating dynamic regulation of IM-induced quiescence by the DREAM complex. Targeting DREAM assembly significantly sensitized cells to IM-induced apoptosis: siRNA knockdown of LIN52 or the DREAM-regulating kinase DYRK1A (mediating LIN52-Ser28 phosphorylation) markedly increased apoptosis rates following IM treatment. This discovery offers a potential novel target for overcoming quiescent cell-mediated resistance in GIST therapy.

Metabolic reprogramming
A complex, bidirectional relationship exists between metabolism and the cell cycle. As different cell cycle phases entail distinct biological functions, cells dynamically adjust their metabolic state to meet biosynthetic demands for proliferation. For instance, metabolic oscillations are driven by the SKP2-IDH1/2 pathway: G1 phase relies on mitochondrial TCA cycle, while S phase favors glycolysis; in tumors, SKP2 overexpression destabilizes IDH1, sustaining hyperglycolysis and driving aberrant cell cycle progression [122]. Conversely, metabolism can directly influence the cell cycle by regulating cyclins and other components [123].
Unlike proliferating cells, those in G0 "quiescence" exhibit significantly reduced biosynthetic demands. In IM-treated KITV558del/+ genetically engineered mouse models (developing singular intestinal GISTs [124]) and GIST cells, IM suppresses glucose uptake, lactate production, and glycolytic capacity by inhibiting KIT signaling (Fig. 3). Furthermore, IM treatment coordinately downregulates glycolytic pathway enzymes while upregulating mitochondrial enzymes associated with the TCA cycle (e.g., COX I/II/IV, TFAM). Significantly, this metabolic switch aligns with the reliance of dormant cells on TRAP1 protein to maintain OXPHOS: TRAP1 stabilizes mitochondrial complex II, enhancing respiratory capacity to support survival under stress [125]. Despite reduced glucose uptake, mitochondrial respiratory capacity increased: IM treatment elevated mitochondrial spare respiratory capacity (SRC) by 50-63%, enhanced glutamine and fatty acid oxidation capacity, and increased overall dependence on fatty acid-derived energy. Additionally, IM-induced dormant GIST cells showed reduced reactive oxygen species (ROS) levels, previously reported to positively correlate with GIST proliferation and invasion [126]. Targeting dormancy reversal enhanced IM efficacy: the mitochondrial OXPHOS inhibitor VLX600 alone suppressed GIST cell respiration. When combined with IM, VLX600 activated AKT signaling, upregulating GLUT1 expression and forcing cells back to a glycolytic phenotype. Simultaneously, VLX600 reversed the IM-induced increase in p27Kip1 levels, blocking APC-p27Kip1/DREAM complex-mediated cell cycle arrest and driving quiescent cells back into the proliferation cycle, thereby potentiating IM-induced apoptosis.

Autophagy
As noted, autophagy is a core mechanism sustaining tumor cell dormancy. Recent studies highlight its critical role in IM resistance in GIST. Ni et al. first elucidated that neural infiltration within the TME activates autophagic flux via the GDNF-GFRA1 axis, driving GIST cells into dormancy and mediating resistance [25] (Fig. 3). This discovery builds upon Chi et al.'s seminal work demonstrating that only ICC surrounding neurons possess tumorigenic potential, revealing the neural microenvironment's regulatory basis for GIST development [127].
Under imatinib pressure, completion of autophagic flux depends on neurogenic signaling: GDNF secreted by infiltrating nerves binds GDNF family receptor α1 (GFRA1) on tumor cells, enhancing lysosomal function via activation of the MCOLN1/Ca²⁺/TFEB pathway. Specifically, the GDNF-GFRA1 complex recruits the lysosomal calcium channel MCOLN1, triggering Ca²⁺ release and promoting nuclear translocation of the transcription factor TFEB, which subsequently upregulates lysosomal biogenesis genes (e.g., CTSB/CTSD), maintaining lysosomal acidification and cathepsin activity. Intact autophagic flux (evidenced by LC3-II degradation and p62 clearance) enables cells to maintain energy homeostasis by clearing damaged organelles while accelerating turnover of inactive KIT protein, establishing a Ki67-negative/TUNEL-negative anti-apoptotic/low-proliferation dormant phenotype. Clinically, high GFRA1 expression independently predicted increased recurrence risk, suggesting that targeting the GDNF-GFRA1-MCOLN1 axis could synergize with TKIs to eliminate dormant cells [25].

Cancer stem cells or stem-like cells: the root of GIST resistance and dormancy?
Emerging research increasingly reveals an intrinsic link between tumor dormancy and cancer stem cells. CSCs, characterized by unique self-renewal, differentiation potential, and treatment resistance, are considered the "seed cells" for tumor recurrence and metastasis. These cells can evade conventional therapy by entering quiescence, subsequently reactivating proliferation upon specific microenvironmental cues, ultimately leading to clinical relapse. Single-cell RNA sequencing has advanced this field: a 2024 Science study identified quiescent CSC subpopulations in cutaneous squamous cell carcinoma exhibiting dynamic plasticity–proliferation balance, directly regulating drug resistance and dormancy [128]. The conceptual underpinnings of the CSC hypothesis trace back to the mid-19th century, when Rudolf Virchow proposed the "embryonal rest” theory. This model posited that tumors originate from residual embryonic-like cells, establishing an early theoretical framework for CSC biology. By the 1960s, experimental evidence began to illuminate tumor cell heterogeneity. Bergsagel and Valeriote, utilizing a spleen colony transplantation assay, provided seminal evidence in multiple myeloma: only a minor subset of cells possessing clonogenic potential was capable of driving tumor growth and propagation [129]. Crucially, while this work identified the critical functional subpopulation, the term "cancer stem cell" was not employed; these cells were designated "clonogenic cells." The formal conceptualization and functional validation of CSC emerged decades later with John Dick's landmark 1997 study in acute myeloid leukemia, which definitively established the CSC paradigm [130]. Although a unified definition remains elusive, the widely accepted functional criteria are: a cell subpopulation capable of self-renewal, driving tumor heterogeneity, and serially propagating tumors in animal models [131].

Properties of cancer stem cells
In normal tissues (e.g., intestinal epithelium, hematopoietic system), tissue-specific stem cells maintain homeostasis through self-renewal and differentiate into diverse, short-lived functional cells [132]. CSCs exhibit more complex biological properties. As mentioned above, the CSC concept was first established in acute myeloid leukemia (AML), marked by the CD34⁺CD38⁻ population [130]. Subsequent studies confirmed stem-like cell populations exist in most hematologic malignancies and solid tumors, including breast [133,134], prostate [135,136], colon [137], ovarian [138], pancreatic [139], and lung cancers [140]. However, CSCs from different tissues display significant heterogeneity. Current CSC identification relies on specific surface markers (e.g., CD44⁺CD24⁻ in breast cancer [133,134]) (Table 1) and functional assays, including in vitro sphere formation, limiting dilution assay, and serial transplantation in animal models – the latter widely regarded as the gold standard for assessing in vivo tumorigenicity [141]. Beyond self-renewal, multipotency, and tumorigenicity, CSCs exhibit two therapeutically critical features: resistance to standard and emerging therapies, constituting a primary cause of treatment failure [141] and metabolic reprogramming - including altered methionine [142,143], glutamine [144], and proline metabolism [145,146] - enabling dynamic adaptation to microenvironmental stress. This metabolic plasticity allows CSCs to dynamically respond to microenvironmental stress, constituting a core mechanism of resistance [141].

Regulatory Mechanisms of CSC Dormancy
CSCs employ diverse intrinsic mechanisms to enter and maintain dormancy (Fig. 4). Imbalanced key signaling pathways play a central role: a low ERK/p38 signaling ratio (ERKlow/p38high) promotes quiescence by suppressing proliferation-associated transcription factors (e.g., c-Fos) [77]; developmental pathways like Notch and Wnt are reactivated, not only maintaining the CSC pool but also finely tuning the balance between dormancy and proliferation via regulation of CDKIs like p21 and p27 [147]. In colorectal cancer models, dormant LGR5⁺ CSCs co-express high p27 levels and maintain quiescence via collagen COL17A1-mediated cell-matrix adhesion; COL17A1 knockout forces these cells out of dormancy [148]. Notably, dormant CSCs typically upregulate glycolytic genes [149], contrasting the OXPHOS shift in IM-induced GIST dormancy and highlighting the context-dependent metabolic flexibility of CSCs.
The tumor microenvironment is indispensable for shaping CSC dormancy (Fig. 4), involving complex multicellular interaction networks: Immune cells, such as specific plasma cell subsets, can maintain CSC self-renewal potential and dormancy in glioblastoma by secreting immunoglobulin G (IgG) that binds Fcγ receptor IIA (FcγRIIA) on CSCs, activating downstream AKT-mTOR signaling [150]. CAFs maintain CSC dormancy by establishing spatially specialized metabolic-signaling niches. In hepatocellular carcinoma, specific CAF subsets within fibrotic rings (FR⁺) secrete high levels of hepatocyte growth factor (HGF), activating Wnt/β-catenin signaling and recruiting EpCAM⁺ CSCs, forming pro-metastatic "wound-healing hubs" [151]. ECM components and their mediated signaling are also crucial. For example, in hepatocellular carcinoma, semaphorin 3C (Sema3C) secreted by CSCs enhances hepatic stellate cell (HSC) contractility, leading to significant ECM collagen deposition and stiffening. The stiffened ECM, via the integrin β1-PI3K/AKT-p27 axis, forces CSCs into G0/G1 arrest. The stiff ECM further feeds back to stimulate CAF secretion of TGF-β1, which enhances Sema3C expression via AP1/c-Jun transcription factor binding to the Sema3C promoter, forming a self-reinforcing "CSCs-Sema3C-ECM stiffening-dormancy maintenance" loop [152].

GIST: bona fide CSCs or CSC-like therapy-persistent cells?
The purpose of discussing CSC biology in GIST is not to claim that imatinib-residual cells have already been definitively proven to be bona fide CSCs, but rather to propose a biologically plausible model, because intrinsic drug tolerance is a core property of CSC biology.
Although CD133 and CD44 are widely recognized CSC markers in many solid tumors, their ubiquitous overexpression in GIST may not specifically delineate a CSC subpopulation [153]. Intriguingly, the CD133- population often exhibits greater proliferative and invasive capacity. Rather than representing a simple inversion of the CSC hierarchy, this finding may indicate that canonical epithelial CSC markers are of limited applicability in GIST and may also reflect epigenetic regulation of CD133 expression, including promoter methylation in more aggressive tumors. Therefore, functional characteristics such as tumor-propagating capacity, therapeutic persistence, and quiescence-associated behavior may provide a more reliable basis than single surface markers for identifying stem-like dormant cell populations in GIST. Nonetheless, CD133 expression levels remain an independent prognostic factor for disease-free survival, significantly correlating with tumor size and location [154].
The strongest functional evidence instead points to KITlow populations. In mice, Bardsley et al. showed that KitlowCd44+Cd34+ ICC progenitors were clonogenic, self-renewing, capable of differentiating into mature ICC, and able to generate GIST-like tumors after spontaneous transformation; importantly, these cells remained resistant to KIT/PDGFRA inhibition, including imatinib, because their survival was relatively KIT-independent, and salinomycin suppressed their proliferation while increasing imatinib sensitivity [26]. In human GIST, Banerjee et al. identified a CD34+KITlow subpopulation with intrinsic imatinib resistance, increased OCT4/NANOG expression, self-renewal capacity, and differentiation into more imatinib-sensitive KIThigh progeny [155]. TKI exposure further enriched this compartment, whereas transcriptomic analyses linked low-KIT states to stem-cell programs together with Gas6/AXL and NF-κB activation; pharmacologic inhibition of AXL or NF-κB enhanced killing of these cells [155]. These data support KITlow cells as a prominent candidate reservoir of primary TKI resistance in GIST.
However, an important conceptual gap remains. The currently available studies do not prove that the residual cells persisting in patients during imatinib therapy are bona fide CSCs. Direct serial tumor-propagation or lineage-tracing evidence using imatinib-residual human GIST cells is still lacking. Rather, existing data support a spectrum of therapy-persistent states that may include KITlow CSC-like cells, quiescent dormant cells, and drug-tolerant persister cells [97]. Consistent with this broader framework, imatinib-induced BCL6 promotes tolerance by recruiting SIRT1 to the TP53 promoter and repressing TP53-dependent apoptosis, whereas a recent study demonstrated reversible YAP activation in imatinib-generated GIST drug-tolerant persister cells and showed that YAP inhibition suppressed these cells and delayed tumor regrowth after treatment withdrawal [156,157]. Therefore, in GIST, it is currently more accurate to refer to imatinib-residual cells as CSC-like or stem-like therapy-persistent cells rather than definitively established CSCs.

Novel therapeutic strategies for GIST: targeting tumor dormancy and stem-like persistent cells
Based on the foregoing, we systematically propose and discuss three promising novel therapeutic strategies: (1) Reactivation ("Awakening") of dormant tumor cells to re-enter the cell cycle and restore sensitivity to TKIs like IM; (2) Active Maintenance ("Sleep") of cancer cells in a persistent dormant state, converting them into clinically "harmless" quiescent lesions; (3) Specific targeting and eradication of stem-like/CSC-like persistent cell populations that may contribute to tumorigenesis, resistance, and relapse (Fig. 5). These strategies aim to fundamentally address the challenge of residual disease persistence with conventional targeted therapy, offering new directions for improving long-term survival in GIST patients. And several clinical studies targeting tumor dormancy and cancer stem cells have been conducted, with details summarized in Tables 2 and 3.

Activating dormant cells ("Awakening" strategy)
Forcing dormant cancer cells back into the cell cycle ("awakening") is a strategy to enhance the efficacy of antiproliferative agents [158]. Its core lies in targeting key pathways regulating cellular dormancy. For example, acute myeloid leukemia stem cells residing in the bone marrow endosteal niche are quiescent and chemoresistant; activating them into the cycle using granulocyte colony-stimulating factor (G-CSF) significantly enhances chemotherapy-induced apoptosis and clearance [159]. Further studies showed that combining cytarabine with the checkpoint kinase 1 inhibitor GDC-0575 more effectively kills leukemia cells after G-CSF reactivation [160]. Similarly, while imatinib is highly effective in chronic myeloid leukemia (CML), it fails to eradicate all leukemia cells, leading to relapse upon discontinuation. Research demonstrates that Fbxw7 induces quiescence in leukemia-initiating cells (LICs) by reducing c-Myc levels; targeting Fbxw7 inhibition significantly enhances LIC sensitivity to IM, offering a novel approach for CML therapy [161].
In GIST, metabolic modulation offers a promising awakening approach. As detailed above, VLX600 combined with IM reverses OXPHOS-dependent dormancy and restores IM sensitivity. Analogous strategies have shown efficacy in other settings: 2-deoxy-d-glucose, (2-DG) enhanced gemcitabine cytotoxicity against resistant pancreatic cancer CSCs [162]. OXPHOS complex inhibitors (e.g., oligomycin, rotenone) are also under investigation. Analogous to GIST, ovarian cancer cells resistant to cisplatin exhibit hyperactivated OXPHOS; treatment with metformin or oligomycin reverses this resistance [163]. Similarly, the ATP synthase inhibitor Oligomycin A showed efficacy in trastuzumab-resistant HER2⁺ breast cancer models [164]. However, caution is warranted: the "awakening" strategy carries inherent risk. Without effective subsequent antiproliferative therapy or if the original drug lacks sufficient potency, activating dormant cells could accelerate disease progression.

Maintaining Tumor Dormancy ("Sleep" Strategy)
Given the potential for awakening to trigger relapse, an alternative strategy aims to perpetually maintain cancer cells in a dormant, "harmless" state – the "sleep strategy." This primarily involves inhibiting proliferative signals and targeting cell cycle pathways, with some progress in breast cancer. For instance, standard adjuvant therapy for estrogen receptor-positive (ER⁺) breast cancer (tamoxifen or letrozole) inhibits the proliferation of dormant cells via anti-estrogen effects, significantly improving survival [165,166]. Pivotal phase III randomized trials confirmed that adding CDK4/6 inhibitors (e.g., palbociclib, FDA-approved in 2015 for ER⁺ advanced breast cancer [167]) to hormonal therapy reduces disease progression risk by approximately 50% [168]. By contrast, the clinical translation of direct cell-cycle targeting in GIST has so far been limited: in a biomarker-driven phase II study enrolling patients with advanced CDKN2A-deleted GIST refractory to imatinib and sunitinib, palbociclib monotherapy showed only limited activity, suggesting that cell-cycle blockade alone may be insufficient for effective disease control in molecularly advanced GIST [169]. Furthermore, Azacytidine combined with Retinoic Acid can induce tumor dormancy and suppress metastasis in various epithelial cancers, offering a potential strategy for controlling advanced metastatic disease [170]. Similarly, applying the "sleep strategy" to GIST faces a fundamental challenge: dormant cancer cells, if not eradicated, retain the potential for reactivation under specific conditions, leading to disease progression.

Targeting CSC-associated pathways and CSC-like persistent cells
CSCs, due to their low immunogenicity and heterogeneous immunomodulatory capabilities, are considered key factors in tumor immune evasion and treatment resistance [171,172]. In GIST, however, bona fide CSCs have not yet been definitively established. Therefore, CSC-directed therapeutic concepts should be viewed primarily as a useful framework for targeting KITlow stem-like cells and other therapy-persistent residual states. Rather than assuming a fully validated GIST CSC hierarchy, it is more appropriate to focus on CSC-associated vulnerabilities, including stemness-related surface phenotypes, non-genetic persistence programs, and survival pathways enriched in CSC-like residual cells.
In GIST specifically, the mechanistic insights discussed above have also begun to reveal actionable therapeutic vulnerabilities in CSC-like or therapy-persistent residual cell states. In human GIST, CD34+KITlow cells with intrinsic imatinib resistance were shown to be susceptible to bemcentinib and bardoxolone, either alone or in combination with imatinib [155]. Consistent with the previously described apoptosis-escape program, the BCL6 inhibitor BI-3802 restored imatinib sensitivity and showed synergistic effects with imatinib in both imatinib-responsive and imatinib-resistant GIST models [156]. Likewise, in line with the emerging role of YAP-dependent persister survival, verteporfin was reported to preferentially suppress imatinib-generated drug-tolerant persister cells and delay tumor regrowth after treatment withdrawal [157]. Earlier murine studies also showed that salinomycin could inhibit the proliferation of KitlowCd44+Cd34+ ICC stem/progenitor cells and enhance their sensitivity to imatinib [26]. Collectively, these findings support the feasibility of targeting CSC-like or stem-like residual states in GIST through state-directed therapeutic vulnerabilities.
Beyond these GIST-specific state-directed approaches, broader therapeutic strategies developed to target CSC-associated features in other malignancies may also provide useful conceptual and translational guidance for GIST. Given that CSCs from different tumors express specific surface markers, targeting these markers is an attractive strategy for their selective elimination [173]. Chimeric antigen receptor T-cell (CAR-T) therapy has pioneered a new chapter in tumor immunotherapy [174]. This approach engineers patient-derived T cells to express CARs targeting specific tumor antigens; after ex vivo expansion, the modified cells are reinfused to exert therapeutic effects. CAR-T therapies targeting CSC-associated antigens have shown preliminary efficacy in early-phase studies in AML [175] and glioblastoma [176], supporting the broader feasibility of immunologic strategies directed against stemness-enriched tumor cell subsets.
Another key strategy involves inducing CSC differentiation, causing loss of stemness and enhancing drug susceptibility. All-trans retinoic acid (ATRA) achieved landmark success in this domain (particularly in leukemia), inspiring its investigation in solid tumors. To date, ATRA has shown differentiation-inducing efficacy in non-small cell lung cancer [177], breast cancer [178], and glioma [179], among others. Furthermore, novel differentiation inducers targeting mutant isocitrate dehydrogenase (IDH1/IDH2) have shown promising clinical results in AML and gained approval [180]. These successes in other malignancies provide important insights for developing differentiation-based strategies targeting stem-like or CSC-like persistent cells in GIST.

Conclusion and perspectives
Tumor dormancy is not merely an "invisible driver" of GIST resistance and relapse but also the "final battleground" that must be confronted in the era of precision therapy. This review systematically integrates the latest evidence, uniquely positioning GIST imatinib resistance and tumor dormancy within a unified framework. Importantly, although secondary and later-line resistance mutations explain a substantial proportion of TKI failure in advanced GIST, they do not fully account for minimal residual disease, prolonged clinically occult persistence, or relapse following treatment interruption. For this reason, mutation-based resistance and dormancy-based persistence should be viewed as complementary rather than competing models. Integrating these two frameworks may provide a more complete explanation of GIST relapse biology and may open therapeutic opportunities beyond genotype-guided TKI sequencing alone. Furthermore, we propose three potential therapeutic modalities: "Awaken – Sleep – Eradicate": (1) Reawakening dormant cells into the cell cycle via metabolic-epigenetic-immune interventions to restore TKIs sensitivity; (2) Enforcing Permanent Quiescence ("Sleep") using microenvironment engineering or pharmacologic induction to achieve functional cure; (3) Eradicating the source of relapse by targeting CSC-associated surface markers or stemness/persistence-associated signaling axes with CAR-T, bispecific antibodies, or differentiation inducers. Current technologies – single-cell multi-omics, organoid-immune co-cultures, and in situ lineage tracing – now enable real-time monitoring of dormancy and awakening. We anticipate that future GIST therapy may shift from a "continuous dosing-passive monitoring" paradigm to an "active intervention based on dormancy status" paradigm, truly enabling the transition from "living with disease" to "tumor-free survival."
Data availability statement: Data sharing does not apply to this article because no new data were generated or analyzed in this study.

Funding
We would like to thank the support of the Funding for Scientific Research and Innovation Team of The First Affiliated Hospital of Zhengzhou University (QNCXTD2023022) and the Joint Fund of Henan Provincial Research and Development Program for Science and Technology (242301420008).

CRediT authorship contribution statement
Sihan Wu: Writing – original draft. Hao Liu: Writing – original draft. Yuhan Yin: Writing – review & editing. Jiehan Li: Writing – review & editing. Zhen Zhang: Writing – review & editing. Wei Li: Writing – review & editing, Conceptualization. Yang Fu: Writing – review & editing, Conceptualization.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.