Lung cancer stem cells in the tumor microenvironment: translational biomarkers and therapeutic strategies for immune evasion and treatment resistance Authors: Li Li Shao, Jing Jing Shao, Tian Ye Zhao, Er Dong Cai, Yaxuan Wang, Ru Jian Lu, Sai Hua Chen Published: 2026 ABSTRACT Highlights•LCSC–TIME crosstalk drives immune evasion and therapy resistance.•TAMs, MDSCs, Tregs, CAFs and hypoxia sustain LCSC stemness.•LCSCs reshape immunity through cytokines, checkpoints and antigen loss.•Targeting LCSCs and the TIME may improve durable immunotherapy responses.Keywords: Lung cancer stem cells, Tumor immune microenvironment, Translational biomarkers, Immune evasion, Immunotherapy resistance Introduction Lung cancer remains the most common and deadliest cancer worldwide, and its prevention and treatment still face major challenges [1,2]. Immune checkpoint inhibitors (ICIs), especially therapies targeting PD-1/PD-L1, have greatly improved the treatment of NSCLC, while chemoimmunotherapy has also changed the therapeutic landscape of extensive-stage SCLC. However, only a subset of patients achieve durable remission. Many tumors show primary resistance, mixed response, acquired resistance, or relapse after an initial response, indicating that immune escape is driven by tumor-intrinsic programs as well as microenvironmental selection pressures [3,4]. The cancer stem cell hypothesis provides an important framework for understanding this problem [5]. LCSCs represent a plastic tumor subpopulation with self-renewal ability, differentiation potential, and resistance to chemotherapy, radiotherapy, targeted therapy, and immunotherapy [[6], [7], [8]]. These cells do not function in isolation. They exist within the TME, which contains immune cells, stromal cells, extracellular matrix, hypoxic niches, and soluble mediators. Increasing evidence demonstrates strong bidirectional communication between LCSCs and the TIME [[9], [10], [11]]. Within this network, TAMs, MDSCs, Tregs, CAFs, and hypoxia-related programs sustain LCSC maintenance through paracrine signaling, metabolic reprogramming, epigenetic regulation, and direct cell-cell interactions. In turn, LCSCs secrete immunosuppressive cytokines, chemokines, and extracellular vesicles, recruit suppressive immune populations, impair cytotoxic T-cell and NK-cell function, and downregulate antigen-presentation machinery. Together, these interactions create an immune-privileged niche that promotes immune evasion and treatment resistance (Fig. 1). Compared with previous reviews that separately describe CSC biology or the lung cancer immune microenvironment, this review specifically integrates LCSC-TIME reciprocity with translational biomarker development and therapeutic intervention. We also distinguish evidence derived mainly from NSCLC, SCLC, or both, and separate established clinical strategies from experimental LCSC-directed approaches. This framework may help identify rational combination therapies capable of disrupting the self-reinforcing LCSC-TIME network. Bidirectional immune crosstalk between LCSCs and the immune tumor microenvironment. Immunosuppressive immune cells within the TIME, including M2 TAMs, MDSCs, Tregs, and dysfunctional NK-cell surveillance, support LCSC maintenance, immune escape, and therapy resistance. In turn, LCSCs remodel the immune TIME by releasing cytokines, chemokines, and exosomes, including CXCL12/CXCR4-related signals, CCL2, IL-6, IL-10, TGF-β, and GM-CSF. These mediators promote recruitment and expansion of MDSCs, Tregs, and M2 TAMs. LCSCs also express PD-L1, B7-H3, and CD47, which contribute to T-cell exhaustion, reduced macrophage phagocytosis, and innate immune escape. Reduced MHC-I and TAP1 expression further impair antigen presentation and CD8⁺ T-cell recognition. This reciprocal network promotes immune evasion, treatment resistance, and recurrence. Biological characteristics of lung cancer stem cells Core features of lung cancer stem cells LCSCs are major contributors to tumor heterogeneity. Their defining characteristics include self-renewal, multipotent differentiation, and resistance to chemotherapy and radiotherapy. These properties drive tumor progression, therapeutic resistance, and distant metastasis [12]. A key feature of LCSCs is asymmetric cell division. This process maintains the stem cell population while also supporting continuous tumor growth. During asymmetric division, one daughter cell retains stemness, whereas the other undergoes differentiation and contributes to tumor expansion [13]. Experimental studies support this concept. For example, CD133⁺ cells isolated from the A549 cell line show a sphere-forming efficiency that is 3.2-fold higher than that of CD133⁻ cells [14]. This finding reflects the strong self-renewal and proliferative capacity of LCSCs.LCSCs also possess multipotent differentiation ability. This property allows them to generate phenotypically diverse progeny and contributes to tumor heterogeneity. Studies have shown that CD133⁺ LCSCs can differentiate into epithelial-like cells expressing CK7 and mesenchymal-like cells expressing vimentin [15]. This phenotypic plasticity enables tumors to adapt to environmental stress and therapeutic pressure. As a result, tumors gain enhanced invasive and metastatic potential. Resistance to chemotherapy and radiotherapy is another critical feature of LCSCs. This resistance is a major cause of tumor recurrence. Several mechanisms contribute to this phenotype. First, LCSCs establish intrinsic defense systems. They upregulate ATP-binding cassette transporters, enhance DNA damage repair, and reduce the expression of apoptotic effectors such as caspase-3. Experimental studies show that the IC50 of cisplatin in CD133⁺ LCSCs is 2.8-fold higher than that in CD133⁻ cells [15]. Second, LCSCs can enter a quiescent G0 state. This state helps them avoid cell cycle–dependent damage. Under hypoxic conditions, the proportion of LCSCs in the G0 phase increases from 12% to 35% [13]. At the molecular level, multiple signaling pathways regulate the resistant phenotype of LCSCs. Persistent activation of the Wnt/β-catenin pathway plays a central role in cisplatin resistance. Inhibition of this pathway increases cisplatin sensitivity in CD133⁺ LCSCs by 4.5-fold [16]. LCSCs also undergo metabolic reprogramming. These cells preferentially rely on glycolysis for energy production. Their glucose uptake and lactate secretion rates are significantly higher than those of bulk tumor cells [14]. This metabolic shift supports rapid proliferation and contributes to the formation of an immunosuppressive microenvironment through extracellular acidification. Interactions with the TIME further strengthen therapeutic resistance. For example, IL-6 secreted by TAMs activates the STAT3 pathway in LCSCs. This activation upregulates MDR1 expression and increases paclitaxel resistance by 3.1-fold [17]. Other studies have shown that LCSCs upregulate PD-L1 expression after chemotherapy, increasing expression levels from 15% to 42% [18]. This change helps LCSCs evade immune surveillance. Overall, the core characteristics of LCSCs arise from the interaction between intrinsic genetic programs and extrinsic microenvironmental signals. These findings highlight the importance of developing dual-targeting strategies that simultaneously disrupt stemness and the supportive microenvironment. Together, these interconnected mechanisms form a dynamic regulatory network that drives LCSC stemness, plasticity, and therapeutic resistance (Fig. 2). Intrinsic features and resistance mechanisms of LCSCs. LCSCs possess core stem-cell properties, including self-renewal, asymmetric division, multipotent differentiation, and phenotypic plasticity, which support tumor heterogeneity and long-term persistence. Stemness-maintaining pathways, including Wnt/β-catenin, Notch, and Hedgehog signaling, sustain LCSC survival, self-renewal, and therapy resistance. LCSCs also display intrinsic resistance mechanisms, such as increased ABC transporter-mediated drug efflux, enhanced DNA damage repair, reduced apoptosis with low caspase-3 activity, and quiescence in the G0 state. Metabolic reprogramming, characterized by increased glucose uptake, glycolysis, and lactate production, further promotes survival under stress. Stemness-associated transcription factors, including SOX2, OCT4, and NANOG, stabilize the stem-like state. These intrinsic programs collectively contribute to chemotherapy resistance, radiotherapy resistance, recurrence, metastasis, and poor prognosis. Specific surface markers of lung cancer stem cells Specific surface markers are essential for enriching, identifying, and functionally studying LCSCs. The most widely used markers include CD133, CD44, ALDH1, and EpCAM [19]. CD133 is among the earliest LCSC markers and is frequently enriched in aggressive NSCLC tissues. Functionally, CD133+ cells show enhanced sphere formation, tumor initiation, and therapy resistance compared with CD133- cells [14,18,20]. CD44 and its variants, especially CD44v6, are closely associated with migration, invasion, EMT, and metastatic potential [20,21]. ALDH1 reflects intracellular detoxification capacity and has been linked to stemness and EGFR tyrosine kinase inhibitor resistance in EGFR-mutant NSCLC [22]. EpCAM, particularly in combination with CD133, can help identify highly tumorigenic epithelial LCSC populations [18,19]. However, these markers should not be interpreted as fixed or universal identifiers of LCSCs. Their expression is context-dependent and can be dynamically regulated by hypoxia, EMT, chemotherapy, targeted therapy, inflammatory cytokines, stromal signals, and sampling site. In addition, marker performance differs between NSCLC and SCLC and may vary across adenocarcinoma, squamous cell carcinoma, and neuroendocrine phenotypes. Therefore, single-marker strategies are insufficient for clinical LCSC definition. More robust approaches should combine marker panels, stemness-related transcriptional programs, functional assays such as sphere formation or tumor initiation, single-cell profiling, and spatial validation within intact tumor niches. This point is particularly important for translational biomarker development, because a clinically useful LCSC marker must identify a biologically functional population and provide prognostic, predictive, monitoring, or therapeutic value. Key signaling pathways driving the maintenance of LCSC stemness The stem-like properties of LCSCs are not controlled by a single factor. Instead, they depend on the coordinated activity of several intrinsic signaling pathways. Among these pathways, Wnt, Notch, and Hedgehog play the most important roles. These pathways regulate self-renewal, cell fate determination, therapy resistance, and tumor regeneration [23]. The Wnt/β-catenin pathway is widely recognized as a major regulator of LCSC stemness. Aberrant activation of this pathway strongly supports the maintenance of stem-like properties. In particular, the nuclear localization of β-catenin is closely associated with clonogenic potential. Experimental studies show that cells with nuclear β-catenin positivity display a sphere formation rate of 62%, whereas negative cells show a rate of only 18% [16]. Several upstream regulators also influence this pathway. For example, SLC34A2 promotes the self-renewal of CD166⁺ LCSCs by activating the Wnt/β-catenin signaling axis. Knockdown of SLC34A2 reduces sphere-forming ability by 75% [24]. These findings highlight the importance of Wnt signaling in maintaining LCSC stemness. The Notch pathway also plays a critical role in LCSCs. However, its effects are highly context-dependent. On one hand, activation of Notch1 is essential for preserving stemness. Cells with high Notch1 expression exhibit a 3.5-fold increase in tumor-initiating capacity compared with cells showing low expression [25]. On the other hand, different Notch receptor subtypes may exert distinct or even opposite effects. For instance, knockdown of Notch2 increases the proportion of CD133⁺ cells within the LCSC population by 2.1-fold [26]. This receptor-specific behavior suggests that individual Notch family members should be evaluated separately in future targeting strategies. The Hedgehog pathway also contributes strongly to LCSC maintenance and therapeutic resistance. This pathway mainly functions through Gli transcription factors. High Gli1 expression is closely associated with reduced sensitivity to chemotherapy and targeted therapies. Experimental data show that the IC₅₀ of cisplatin in Gli1⁺ cells is 3.2 times higher than that in Gli1⁻ cells [27]. Pharmacological inhibition of Hedgehog signaling can reverse drug resistance. For example, sulforaphane enhances sensitivity to EGFR tyrosine kinase inhibitors and increases drug responsiveness in resistant LCSCs by 4.8-fold [28]. Importantly, these signaling pathways do not function independently. Instead, they form an interconnected regulatory network with extensive crosstalk. For example, activation of Wnt signaling can induce Notch1 expression. Similarly, inhibition of the Hedgehog pathway can reduce nuclear β-catenin levels [23]. This interconnected network increases the stability and robustness of stemness maintenance. It also helps explain why therapies targeting a single pathway often show limited clinical efficacy. In contrast, combined inhibition of the Wnt and Hedgehog pathways reduces sphere formation by up to 90%, whereas single-pathway inhibition achieves only 40–50% suppression [27]. Overall, a deeper understanding of the interactions among these signaling pathways is essential. Multi-target therapeutic strategies may provide a more effective approach for eliminating LCSCs and overcoming therapeutic resistance. Translational classification of LCSC-related biomarkers To enhance the translational relevance of LCSC-related biomarkers, these molecules can be classified according to their potential clinical applications rather than by biological function alone. Identification markers are mainly used to enrich or characterize stem-like tumor cell populations in experimental and clinical specimens. Prognostic markers are associated with aggressive tumor behavior, metastasis, recurrence risk, or poor survival. Predictive markers may help estimate sensitivity or resistance to immunotherapy, targeted therapy, chemotherapy, or radiotherapy. Monitoring biomarkers, particularly those detectable through circulating tumor cells, extracellular vesicles, ctDNA, or other liquid biopsy approaches, may support longitudinal assessment of minimal residual disease, therapeutic resistance, and recurrence. In addition, several LCSC-related molecules function as therapeutic targets or actionable nodes within the LCSC–TIME regulatory network. Table 1 summarizes representative LCSC-associated biomarkers, their major biological functions, potential clinical relevance, therapeutic implications, and the predominant evidence context in NSCLC, SCLC, or both. Table 1Translational categories of representative LCSC-related biomarkers.Table 1 dummy alt textCategoryRepresentative biomarkersMain biological functionEvidence baseTranslational implicationLCSC identification / enrichmentCD133, CD44/CD44v6, ALDH1, EpCAM, CD166Self-renewal, tumor initiation, EMT, epithelial or mesenchymal plasticityMainly NSCLC; selected markers also reported in SCLC or mixed lung cancer modelsUseful for LCSC enrichment, patient stratification, and construction of combinational marker panels; single markers are insufficient.Prognostic / recurrence markersCD133, CD44v6, ALDH1, EpCAM+CD133+, stemness programsHigh tumorigenicity, metastatic potential, drug resistance, minimal residual diseaseStronger clinical evidence in NSCLC; limited but emerging evidence in SCLCMay identify patients at high risk of relapse, metastasis, or poor survival and support intensified monitoring.Immunotherapy-response markersPD-L1, B7-H3, MHC-I/TAP1/LMP2 loss, CXCL12/CXCR4, CD47T-cell exhaustion, immune checkpoint signaling, defective antigen presentation, suppressive-cell recruitment, innate immune evasionPD-L1 and CXCR4: mainly NSCLC; antigen-presentation rescue: strong SCLC evidence; CD47: preclinical lung cancer evidenceMay help explain primary/acquired ICI resistance and guide combinations with checkpoint blockade, epigenetic therapy, CXCR4 antagonists, or CD47 blockade.Therapeutic targetsPD-L1, CD47-SIRPα, CXCR4, IL-6/STAT3, Wnt/β-catenin, Notch, HedgehogMaintenance of immune privilege, stemness, macrophage polarization, and therapy resistanceApproved ICIs are clinical; most LCSC-directed targets remain preclinical or early translationalRational targets for combination therapy, but specificity, toxicity, pathway redundancy, and compensatory escape require careful evaluation.Monitoring / liquid-biopsy candidatesCTCs with stem-like markers, extracellular vesicles, circulating cytokines/chemokines, ctDNA-defined resistance contextDynamic tracking of residual disease, immune remodeling, and therapy-induced stemnessEmerging evidence; stronger clinical infrastructure in NSCLC than SCLCCould support serial assessment of LCSC burden and adaptation during immunotherapy or targeted therapy. Translational categories of representative LCSC-related biomarkers. LCSC-mediated remodeling of the immune microenvironment and mechanisms of immune evasion Recruitment and polarization of immunosuppressive cells driven by LCSC-derived cytokines LCSCs play a central role in shaping the tumor immune microenvironment. Rather than acting only as passive therapy-resistant cells, they actively construct a suppressive niche through cytokines, chemokines, immune checkpoints, extracellular vesicles, metabolic products, and antigen-presentation defects [[29], [30], [31]]. This reciprocal network can be viewed as a feed-forward circuit: LCSCs recruit and educate suppressive immune cells, and these immune and stromal cells in turn reinforce LCSC stemness, plasticity, and survival. Such crosstalk provides a mechanistic explanation for immune evasion, primary ICI resistance, incomplete response, persistence of minimal residual disease, and recurrence. Regulation of M2 polarization of tumor-associated macrophages by LCSCs LCSCs promote macrophage polarization toward the immunosuppressive M2 phenotype. They achieve this through soluble cytokines, membrane-associated molecules, and extracellular vesicles [32]. These interactions reshape immune balance within the tumor microenvironment. Macrophage colony-stimulating factor (M-CSF) is a key mediator in this process. The concentration of M-CSF in LCSC-conditioned medium reaches 80 pg/mL, which is much higher than the 15 pg/mL observed in non-stem cells. M-CSF activates the CSF1R pathway and promotes M2 polarization. As a result, CD206 expression increases by 3.5-fold [32]. LCSC-derived IL-10 further strengthens this suppressive phenotype. IL-10 at 50 pg/mL increases macrophage-derived IL-10 production by 4.2-fold [33]. This process forms a positive feedback loop that stabilizes M2 polarization. Membrane-associated molecules also contribute to macrophage regulation. The transmembrane glycoprotein MUC1 is highly expressed in LCSCs. MUC1 directly promotes M2 polarization through activation of NF-κB signaling. LCSCs with high MUC1 expression generate 75% M2 TAMs, whereas low-expression groups produce only 28% [34]. At the epigenetic level, LCSC-derived exosomes also influence macrophage behavior. These exosomes contain high levels of miR-1207-5p. After uptake by macrophages, miR-1207-5p suppresses PTEN and activates the PI3K/Akt pathway. This process increases Arg1 expression by 3.8-fold [35]. Importantly, interactions between LCSCs and TAMs are bidirectional. M2-polarized TAMs secrete IL-6, which helps maintain LCSC stemness. LCSCs exposed to conditioned medium from M2 TAMs show a 2.5-fold increase in CD133 expression [17]. Disrupting this interaction network may provide therapeutic benefit. For example, neutralization of M-CSF reduces the proportion of M2 TAMs by 55% in LCSC models [32]. Molecular pathways promoting expansion of MDSCs and Tregs LCSCs promote the parallel expansion of MDSCs and Tregs through coordinated inflammatory and immunoregulatory signaling pathways [36]. These interactions establish a dual-layered immunosuppressive barrier within the tumor microenvironment. In the myeloid compartment, IL-6 and GM-CSF are major regulators. IL-6 at 100 pg/mL expands MDSCs by 3.5-fold. At the same time, IL-6 activates STAT3 signaling and increases the expression of Arg1 and iNOS. These changes strengthen the suppressive effect of MDSCs on CD8⁺ T cells by 2.8-fold [36]. GM-CSF also contributes to MDSC expansion. At a concentration of 50 pg/mL, GM-CSF increases MDSC accumulation by 2.7-fold through activation of the PI3K/Akt pathway [36]. In the lymphoid compartment, LCSCs rely mainly on TGF-β and chemokine signaling to regulate Tregs. TGF-β at 20 ng/mL activates the Smad pathway and increases Foxp3 expression by 4.5-fold. This process expands Treg populations by 3.2-fold [37]. LCSCs also secrete CCL22. This chemokine further enhances Treg recruitment and increases their proportion by 3.8-fold [36]. Direct cell–cell interactions additionally contribute to immunosuppression. PD-L1 expressed on LCSCs binds to PD-1 on Tregs. This interaction enhances the suppressive activity of Tregs against CD8⁺ T cells by 2.5-fold [18]. Targeting these pathways may effectively reduce immunosuppression. For example, blockade of IL-6 signaling decreases MDSC expansion by 45% and reduces Treg expansion by 50% in LCSC models [36]. High expression of immune checkpoint molecules and T cell exhaustion LCSCs express high levels of immune checkpoint molecules. Among these molecules, PD-L1 and B7-H3 are particularly important because they transmit inhibitory signals and induce functional exhaustion of effector T cells [18,20]. PD-L1 expression has been observed on lung cancer stem-like cells, including cells from metastatic lymph-node aspirates, and may allow LCSC-enriched populations to suppress CD8+ T-cell proliferation and IFN-γ production through the PD-1/PD-L1 axis [18,38]. B7-H3 may further inhibit early T-cell activation and contribute to a non-inflamed or exhausted immune phenotype [20]. Clinically, this biology is relevant to ICI resistance in both NSCLC and SCLC, although the level of supporting evidence differs by subtype. In NSCLC, LCSC-enriched phenotypes may contribute to primary resistance by producing a poorly antigenic and suppressive niche before treatment. During therapy, immune pressure may select for stem-like clones with higher checkpoint expression, EMT programs, altered cytokine secretion, or reduced antigen presentation, thereby contributing to acquired resistance, mixed response, and residual disease. In SCLC, where immunotherapy benefit is often modest and antigen-presentation defects are common, stem-like or neuroendocrine-plastic states may support immune escape through MHC-I downregulation and epigenetic silencing [3,39]. Therefore, checkpoint expression should be interpreted together with LCSC state, antigen presentation, and immune-cell spatial context rather than as an isolated biomarker. Immune evasion through defective antigen presentation In addition to immunosuppressive signaling, LCSCs evade immune surveillance by impairing antigen presentation. They downregulate MHC-I molecules and antigen-processing proteins such as TAP1 and LMP2, thereby weakening recognition by CD8+ T cells. EMT and TGF-β signaling can further reduce MHC-I expression and limit the ability of tumor-resident memory T cells to recognize cancer stem-like cells [40]. This mechanism has direct therapeutic relevance. In NSCLC, defective antigen presentation may contribute to immune-cold niches and incomplete response to PD-1/PD-L1 blockade. In SCLC, epigenetic repression of antigen-presentation machinery has been implicated in resistance to PD-L1 blockade, and inhibition of lysine-specific demethylase 1 (LSD1) can restore MHC-I expression and improve immune recognition in preclinical models [39]. HDAC- or LSD1-targeting strategies may therefore be combined with checkpoint blockade to convert LCSC-enriched, antigen-poor tumor states into more immunogenic states. Overall, restoring antigen presentation represents a rational approach to disrupt LCSC-mediated immune escape, but clinical translation requires careful selection of patients, dosing schedules, and immune-context biomarkers. Reciprocal crosstalk: how the immune microenvironment sustains and induces LCSC stemness Reinforcement of LCSC stemness by TAM-derived IL-6/STAT3 signaling Tumor-associated macrophages (TAMs) are major inflammatory regulators in the lung cancer immune microenvironment. These cells secrete high levels of IL-6, which activates the IL-6R/STAT3 signaling pathway in LCSCs. Activation of this pathway induces STAT3 phosphorylation and nuclear translocation. As a result, stemness-related transcription factors such as Sox2 and Oct4 become upregulated. This signaling cascade enhances the self-renewal and sphere-forming capacity of LCSCs. At the same time, STAT3 activation increases the expression of the multidrug resistance gene MDR1. This process promotes resistance to chemotherapeutic agents such as cisplatin. Importantly, the interaction between TAMs and LCSCs is bidirectional. LCSCs secrete CCL2, which recruits TAMs and stimulates additional IL-6 production. This interaction creates a positive inflammatory feedback loop that continuously reinforces stemness properties. Disruption of this signaling loop can restore treatment sensitivity. For example, IL-6 neutralizing antibodies significantly improve drug responsiveness in LCSCs [17]. CAF-mediated construction of a protective niche for LCSCs Cancer-associated fibroblasts (CAFs) support LCSCs by creating a protective microenvironmental niche. CAFs achieve this effect through paracrine signaling and extracellular matrix (ECM) remodeling. CAFs secrete insulin-like growth factor 2 (IGF2) and CXCL12. These molecules activate the IGF1R/PI3K/Akt and CXCR4 pathways, respectively. Activation of these pathways promotes the expression of stemness-related genes and enhances self-renewal capacity in LCSCs. At the structural level, ECM components produced by CAFs reduce the penetration of chemotherapeutic drugs into tumor tissues. This barrier effect further increases therapeutic resistance in LCSCs. CAFs and LCSCs also establish reciprocal signaling interactions through IL-6 and TGF-β. These interactions stabilize stemness features and promote cell survival. Targeting CAFs may therefore provide therapeutic benefit. For example, PDGF-AA neutralizing antibodies disrupt the protective niche, weaken stemness characteristics, and restore drug sensitivity [41,42]. Hypoxia-induced HIF-1α activation and dedifferentiation toward LCSCs Hypoxia is a defining characteristic of the tumor microenvironment. Low-oxygen conditions promote the stabilization and activation of hypoxia-inducible factor-1α (HIF-1α). Activated HIF-1α drives the conversion of non-stem cancer cells into LCSC-like cells. HIF-1α directly activates the transcription of stemness-associated genes such as Sox2. In addition, HIF-1α induces epithelial–mesenchymal transition (EMT). This process is characterized by reduced E-cadherin expression and increased vimentin expression. These molecular changes enhance tumorigenic potential and increase the proportion of cells expressing LCSC markers such as CD133. As a result, hypoxia promotes both maintenance and expansion of the LCSC population. Experimental studies further support this role. Pharmacological inhibition of HIF-1α significantly reduces the conversion of non-stem cells into LCSC-like cells [39]. TGF-β–induced EMT and acquisition of stemness Transforming growth factor-β (TGF-β) plays a dual role in lung cancer progression. This cytokine promotes both EMT and the acquisition of stem-like properties. Therefore, TGF-β acts as an important molecular link between the immune microenvironment and LCSC regeneration. Exposure to high levels of TGF-β activates EMT programs in lung cancer cells. During this process, epithelial markers decrease, whereas mesenchymal markers increase. At the same time, stemness-related transcription factors become upregulated, and sphere-forming ability increases. These molecular and phenotypic changes enhance tumorigenicity and increase resistance to cisplatin. In addition, the proportion of CD133⁺ cells expand significantly. Blocking TGF-β signaling can reverse these effects. Neutralizing antibodies against TGF-β reduce stemness-related features and restore chemosensitivity [40]. Overall, these findings demonstrate that the tumor microenvironment dynamically maintains and induces LCSC stemness through multiple signaling pathways and niche interactions (Fig. 3). Microenvironmental induction and maintenance of LCSC stemness. The tumor microenvironment promotes LCSC maintenance and plasticity through multiple extrinsic signals. TAMs release IL-6, which activates IL-6R/JAK/STAT3 signaling and upregulates SOX2, OCT4, and MDR1, thereby enhancing stemness, self-renewal, and drug resistance. CAFs provide niche support through IGF2/IGF1R/PI3K/Akt and CXCL12/CXCR4 signaling, extracellular matrix remodeling, and reduced drug penetration. Hypoxia stabilizes HIF-1α and promotes dedifferentiation of non-stem lung cancer cells toward an LCSC-like state. TGF-β signaling induces EMT through SMAD2/3 activation, with reduced E-cadherin and increased vimentin, SOX2/NANOG, sphere formation, and cisplatin resistance. These microenvironmental cues collectively drive LCSC expansion, plasticity, therapy resistance, tumor initiation, and recurrence. Emerging therapeutic strategies targeting LCSC–TME crosstalk Therapeutic strategies discussed in this section differ substantially in clinical maturity. Approved or established lung cancer strategies include immune checkpoint blockade, chemoimmunotherapy, radiotherapy, chemotherapy, and oncogene-directed targeted therapy in molecularly selected NSCLC. In contrast, most direct LCSC-targeted approaches, including CD133/CD44-directed agents, CD47-SIRPα blockade in the LCSC context, CXCR4 antagonism for LCSC-TIME disruption, and Wnt/Notch/Hedgehog inhibition, remain preclinical or early translational. The most realistic near-term approach is therefore not single-agent elimination of LCSCs, but rational combination therapy that simultaneously targets tumor-intrinsic stemness, suppressive immune niches, and clinically actionable resistance mechanisms. Immunotherapies targeting LCSC-specific markers: monoclonal antibodies and ADCs Monoclonal antibodies and antibody-drug conjugates (ADCs) targeting LCSC-associated surface markers such as CD133, CD44, and ALDH1-related phenotypes have shown promising therapeutic potential in experimental models [20]. These therapies may reduce self-renewal and selectively deliver cytotoxic agents to marker-enriched populations. However, they are not yet established clinical strategies for lung cancer LCSC eradication. Important obstacles include marker heterogeneity, incomplete overlap between marker expression and functional stemness, on-target/off-tumor toxicity in normal epithelial or hematopoietic progenitors, and the possibility that marker-negative cells may reacquire stem-like states after treatment. Therefore, antibody- or ADC-based approaches should be developed with combinational marker selection, functional validation, and immune-context biomarkers. Targeting stemness signaling pathways (Wnt and Notch) in combination immunotherapy The self-renewal capacity of LCSCs depends heavily on continuous activation of stemness-related signaling pathways. Wnt, Notch, and Hedgehog inhibitors can suppress stem-like phenotypes and may enhance immune checkpoint blockade by increasing CD8+ T-cell infiltration or reducing tumor-mediated immune exclusion [23,25,27,43]. However, these pathways are also involved in tissue homeostasis and normal stem/progenitor-cell maintenance. Their therapeutic targeting may therefore cause toxicity, and pathway redundancy may permit compensatory activation of alternative stemness circuits. This limitation supports the use of biomarker-guided, intermittent, or combination regimens rather than broad and continuous pathway inhibition. Chemokine receptor antagonists disrupting LCSC–immune cell communication Chemokine signaling provides a direct link between LCSC biology and immune-cell trafficking. LCSC-derived CXCL12 activates CXCR4 and promotes recruitment of Tregs, MDSCs, and other suppressive populations into tumor tissues [29]. CXCR4 antagonists, such as plerixafor, can disrupt this axis, reduce immunosuppressive-cell recruitment, and potentially improve immune-mediated clearance of LCSCs. This strategy is especially relevant to metastatic NSCLC stem-like populations, where CXCR4 inhibition has been reported to counteract immunosuppressive properties [29]. Nevertheless, CXCR4-targeted therapy remains investigational in this specific context, and optimal combinations with anti-PD-1/PD-L1 therapy, chemotherapy, or stemness-pathway inhibitors require prospective validation. Targeting the CD47–SIRPα axis to enhance innate immune clearance LCSCs may evade innate immune surveillance by overexpressing CD47, a "do not eat me" signal that inhibits macrophage-mediated phagocytosis through SIRPα [44]. Blocking the CD47-SIRPα axis can restore macrophage phagocytic activity against lung cancer cells and cancer stem-like cells in preclinical models [44]. Combination with anti-PD-1/PD-L1 therapy or stemness-pathway inhibition may theoretically activate both innate and adaptive immunity. However, clinical translation faces major challenges, including anemia and other toxicities related to CD47 expression on normal cells, incomplete selectivity for LCSCs, and compensatory immune escape through alternative checkpoints or macrophage polarization programs. These limitations should be considered when designing LCSC-directed immunotherapeutic combinations. Mechanism-guided combinations and translational status A useful translational framework is to connect each resistance mechanism to a therapeutic strategy and its current level of evidence. For example, LCSC-derived CXCL12/CXCR4 signaling can be linked to CXCR4 antagonists; defective antigen presentation can be linked to epigenetic therapies such as HDAC or LSD1 inhibition; TAM-derived IL-6/STAT3 signaling can be linked to IL-6/STAT3 blockade; and CD47-mediated innate immune evasion can be linked to CD47-SIRPα inhibitors. Table 2 summarizes these relationships while distinguishing clinically established immunotherapy backbones from investigational LCSC-directed interventions. Table 2Bidirectional LCSC-TIME mechanisms, matched interventions, and translational status.Table 2 dummy alt textNetwork mechanismRepresentative mediatorsDominant evidence baseMatched therapeutic strategyStatus and key challengesImmune checkpoint-driven T-cell exhaustionPD-L1, B7-H3, PD-1+ exhausted CD8+ T cellsPD-L1: NSCLC and SCLC clinical relevance; B7-H3 mainly investigationalPD-1/PD-L1 blockade; B7-H3-directed agents in selected settingsICI backbones are established, but LCSC enrichment may drive primary/acquired resistance. Spatial and stemness biomarkers are needed.Suppressive-cell recruitmentCXCL12/CXCR4, CCL2, CCL22, TGF-βMainly NSCLC LCSC and metastatic modelsCXCR4 antagonists; CCL2/CCR2 or TGF-β pathway combinationsInvestigational for LCSC-TIME disruption; optimal timing and immune combinations remain unresolved.TAM-LCSC inflammatory feedbackTAM-derived IL-6/STAT3, M-CSF/CSF1R, CCL2NSCLC and lung cancer model evidenceIL-6/STAT3 blockade; CSF1R/M-CSF targeting; macrophage reprogrammingPreclinical/early translational; redundancy of macrophage programs and systemic cytokine toxicity are concerns.Defective antigen presentationMHC-I loss, TAP1/LMP2 reduction, EMT/TGF-β, LSD1-regulated repressionSCLC evidence for LSD1/MHC-I rescue; NSCLC evidence for EMT-related evasionHDAC or LSD1 inhibitors plus PD-1/PD-L1 blockadeStrong mechanistic rationale; requires biomarker selection and careful management of epigenetic-drug toxicity.Innate immune evasionCD47-SIRPα axis, M2 TAMsPreclinical lung cancer and LCSC evidenceCD47 blockade; combination with ICIs or macrophage modulatorsInvestigational; on-target/off-tumor toxicity and macrophage compensation must be addressed.Intrinsic stemness and plasticityWnt/β-catenin, Notch, Hedgehog, EMT, HIF-1αMainly preclinical NSCLC/LCSC models; selected pathways relevant to SCLC plasticityWnt/Notch/Hedgehog inhibitors; HIF-1α or EMT-directed combinationsNot established for LCSC eradication; normal stem-cell toxicity and pathway redundancy limit monotherapy. Bidirectional LCSC-TIME mechanisms, matched interventions, and translational status. Discussion Bidirectional communication between LCSCs and the TME is a major driver of tumor progression, immune evasion, and therapeutic resistance. The central message of this review is that LCSCs and the immune microenvironment form a reciprocal network rather than a one-way regulatory axis. LCSCs remodel the TIME by expressing immune checkpoints, secreting suppressive mediators, recruiting TAMs/MDSCs/Tregs, impairing antigen presentation, and promoting innate immune escape. Conversely, TAMs, CAFs, hypoxia, TGF-β, extracellular matrix remodeling, and inflammatory cytokines sustain stemness, EMT, quiescence, and drug resistance. This network can explain several clinically important phenomena, including primary resistance to ICIs, acquired resistance after immune selection, mixed response among lesions, persistence of minimal residual disease, and recurrence after apparently effective treatment. Future translational work should address three priorities. First, LCSC identification should move beyond single markers toward integrated panels that combine surface markers, transcriptional states, spatial niche information, and functional assays. Single-cell RNA sequencing, single-cell epigenomics, spatial transcriptomics, and spatial proteomics can define LCSC-like states in relation to exhausted T cells, TAMs, CAFs, and hypoxic niches. Second, preclinical models should better preserve patient-specific tumor ecology. Patient-derived organoids, immune-organoids, co-culture systems, and patient-derived xenografts can help test whether biomarker-defined LCSC-TIME circuits predict response to combination therapy. Third, circulating biomarkers should be incorporated into longitudinal monitoring. CTCs with stem-like features, extracellular vesicles, circulating cytokines, and ctDNA-defined resistance context may help track LCSC dynamics, residual disease, and therapy-induced immune remodeling. Despite these opportunities, several challenges remain: LCSC markers are dynamic and context-dependent; NSCLC and SCLC differ in lineage programs, immune contexture, and therapeutic vulnerabilities; and many LCSC-directed interventions may damage normal stem/progenitor cells or trigger compensatory immune escape. Therefore, the most promising strategy is biomarker-guided combination therapy that disrupts both tumor-intrinsic stemness and the protective immune-stromal niche while maintaining an acceptable therapeutic window. Funding information This research was financially supported by five projects, including the Nantong Municipal Health Commission Project (QN2024026); Natural Science Foundation of Nantong Municipal Science and Technology Bureau (MSZ2024121, MSZ2025014). Nantong University (Grant No. 2024I14), Nantong Youth Medical Expert Project (Grant No. 400142). CRediT authorship contribution statement Li Li Shao: Writing – original draft, Data curation, Conceptualization. Jing Jing Shao: Writing – original draft, Supervision, Software. Tian Ye Zhao: Writing – original draft, Methodology, Investigation. Er Dong Cai: Writing – original draft, Software, Conceptualization. Yaxuan Wang: Writing – review & editing, Data curation, Conceptualization. Ru Jian Lu: Writing – original draft, Project administration, Data curation. Sai Hua Chen: Writing – review & editing, Methodology, Funding acquisition. 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. Contributor Information Yaxuan Wang, Email: 17805050201@163.com. Ru Jian Lu, Email: lurujian168@sina.com. Sai Hua Chen, Email: jsshcjl@ntu.edu.cn. References