Tumour and microenvironment crosstalk in NSCLC progression and response to therapy

Abstract

The treatment landscape of non-small-cell lung cancer (NSCLC) is evolving rapidly, driven by advances in the development of targeted agents and immunotherapies. Despite this progress, some patients have suboptimal responses to treatment, highlighting the need for new therapeutic strategies. In the past decade, the important role of the tumour microenvironment (TME) in NSCLC progression, metastatic dissemination and response to treatment has become increasingly evident. Understanding the complexity of the TME and its interactions with NSCLC can propel efforts to improve current treatment modalities, overcome resistance and develop new treatments, which will ultimately improve the outcomes of patients. In this Review, we provide a comprehensive view of the NSCLC TME, examining its components and highlighting distinct archetypes characterized by spatial niches within and surrounding tumour nests, which form complex neighbourhoods. Next, we explore the interactions within these components, focusing on how inflammation and immunosuppression shape the dynamics of the NSCLC TME. We also address the emerging influences of patient-related factors, such as ageing, sex and health disparities, on the NSCLC–TME crosstalk. Finally, we discuss how various therapeutic strategies interact with and are influenced by the TME in NSCLC. Overall, we emphasize the interconnectedness of these elements and how they influence therapeutic outcomes and tumour progression.

Introduction

Non-small-cell lung cancer (NSCLC), with the most common subtypes being lung adenocarcinoma (LUAD) and lung squamous cell carcinoma, remains the leading cause of cancer-related mortality worldwide despite advances in diagnosis and therapy¹. Immune-checkpoint inhibitors (ICIs) and targeted therapies have revolutionized the treatment paradigms for patients with advanced-stage or metastatic, locally advanced unresectable and early stage resectable NSCLC². Unfortunately, however, the majority of patients do not benefit and/or do not obtain a durable benefit from these therapeutic strategies for reasons that are not well understood³. The important role of the tumour microenvironment (TME) in cancer initiation, progression, metastatic dissemination and response to therapy is now well established⁴. The interplay between cancer cells and the surrounding stromal, immune and endothelial components creates a dynamic ecosystem that influences disease trajectory and therapeutic outcomes⁵.

An important yet underexplored aspect of the NSCLC TME is its spatial heterogeneity, that is, the non-uniform distribution and organization of cellular and molecular components within and surrounding the tumour mass⁶. This spatial complexity is not merely a by-product of tumour growth but actively contributes to resistance to treatment and disease progression⁶. Conventional analytical techniques often overlook these spatial patterns, providing a homogenized view that fails to capture localized interactions and microniches that can drive distinct biological behaviours⁷.

Over the past decade, advances in spatial transcriptomics and high-resolution imaging technologies have opened new avenues to dissect the TME with unprecedented detail^7–11. These technologies enable the simultaneous mapping of gene expression and cellular localization, unveiling how spatial context influences cellular function and intercellular communication. In NSCLC, spatial heterogeneity has been linked to variations in immune cell infiltration, metabolic gradients and the composition of the stromal component, all of which can affect responses to treatments including immunotherapies and targeted agents^12–17.

Understanding the spatial organization of the NSCLC TME is not merely an academic exercise but a clinical imperative. Spatial niches within tumours can harbour subpopulations of cancer cells that evade immune detection or are resistant to therapy, acting as reservoirs for disease relapse¹⁸. Moreover, spatial interactions can modulate signalling pathways in ways that differ from those predicted by examining cells in isolation, sometimes leading to an unexpected lack of response to therapy¹⁹. Therefore, integrating spatial context into our understanding of the TME in NSCLC can guide the development of personalized and effective therapeutic strategies while also facilitating the discovery of new biomarkers.

Health disparities add yet another dimension to the complexity of the NSCLC TME. Socioeconomic factors, access to health care, environmental exposures and genetic ancestry can all influence tumour biology and cancer–TME interactions²⁰. For example, differences in immune profiles and metabolic states have been observed across populations, potentially affecting responses to immunotherapy and other treatment modalities²¹. Addressing these disparities requires a nuanced understanding of how extrinsic factors shape the TME in NSCLC and contribute to disease outcomes.

In this Review, we aim to shift the focus from a tumour-centric view to a holistic perspective that encompasses the spatial dynamics of the NSCLC TME. We explore the different TME components and their interactions, and discuss how spatial heterogeneity contributes to therapeutic resistance. We also address the emerging influences of patient-related factors such as ageing, sex and health disparities on TME characteristics and tumour behaviour. We synthesize how the TME, shaped by spatial microenvironmental dynamics, systemic and patient-specific factors, drives differential responses to therapy in patients with NSCLC. Understanding this complexity could pave the way for new interventions that address the current limitations of treatments for these patients and improve their outcomes.

Spatial niches and cellular interactions

The TME of NSCLC is a complex and dynamic ecosystem that influences tumour progression, metastatic dissemination and response to therapy. This TME comprises a diverse array of cellular components, such as immune cells, cancer-associated fibroblasts (CAFs), endothelial cells and pericytes, and non-cellular components, such as extracellular matrix (ECM) elements and various signalling molecules^16,22–24 (Table 1). In the past few years, advances in spatial transcriptomics and high-resolution imaging have unveiled the spatial organization of the NSCLC TME, revealing the existence of distinct niches characterized by unique cellular compositions, molecular signatures and functional states^12–15,25.

Table 1 |.

Major components of the non-small-cell lung cancer tumour microenvironment

Cell type	Major role	Function in the TME
Myeloid
TAMs44–46	Plastic role	Antitumorigenic subtypes include CXCL9+ TAMs, which promote T cell infiltration and improve responses to ICIs Protumorigenic subtypes include SPP1+, which promote immune exclusion; COL1A1+, which activate fibrosis and immune evasion; and TREM2+, which inhibit NK cell infiltration and activity and promote expansion of exhausted CD8+ T cells
Senescent TAMs76	Protumorigenic	Promote an immunosuppressive environment Promote enrichment of Treg cells
Alveolar macrophages23,41	Plastic role	Present in alveolar spaces Can perform phagocytosis and mediate inflammation or tolerance
Interstitial macrophages23	Plastic role	Reside in interstitial regions Regulate immune responses contextually
TANs13,15,58	Plastic role	The N1 subtype enhances antitumour immunity The N2 subtype promotes angiogenesis, EMT and immunosuppression
cDC1 (refs. 38,39)	Antitumorigenic	Present antigens to CD8+ T cells, supporting cytotoxic immune responses
cDC2 (refs. 38,286)	Plastic role	Present antigens to CD4+ T cells Can support various TH subsets, influencing treatment outcomes
cDC3 (refs. 38,286)	Plastic role	Modulate immune responses Exhibit immunosuppressive features and reduced T cell activation in the TME
Regulatory DCs39	Protumorigenic	Suppress immune responses Contribute to resistance to ICIs
Plasmacytoid DCs38,51	Plastic role	Secrete type I interferons Generally enhance immune surveillance and antitumour responses, but can be associated with a tolerogenic phenotype
Myeloid-derived suppressor cells287,288	Protumorigenic	Suppress T cell activation Promote tumour immune evasion
Lymphoid
CD8+ Teff cells16,27	Antitumorigenic	Direct cytotoxic function
CD8+ exhausted T cells27,29,30,289	Protumorigenic	Dysfunctional state: limited cytotoxicity and high expression of immune checkpoints
TH1 cells4,290	Antitumorigenic	Support CD8+ T cell activity Coordinate effective antitumour immune responses
TH2 cells290	Protumorigenic	Support protumorigenic macrophages Enhance humoral responses that favour immune evasion and tumour progression
TH17 cells22	Plastic role	Produce IL-17 Can recruit cytotoxic cells (antitumour function) or, conversely, foster angiogenesis and tumour progression
Treg cells27,83	Protumorigenic	Suppress antitumour immunity Promote immune evasion and progression
Follicular helper T cells30,37	Antitumorigenic	Promote B cell activity and antibody production Support TLS formation, often linked to a better prognosis
NK cells16,291,292	Antitumorigenic	Direct tumour cell killing Can be functionally compromised in hypoxic niches
Other ILCs293–296	Plastic role	ILC2 subtype is protumorigenic, promoting fibrosis and supporting protumorigenic TAMs ILC3 subtype can be antitumour (via IL-17-mediated activation) or protumorigenic (promoting angiogenesis)
Memory B cells28,32	Plastic role	Can enhance antigen-specific immunity or, in some cases, support tumour-suppressive responses
Plasma cells16,32	Plastic role	IgG+ plasma cells often enhance antitumour immunity IgA+ plasma cells can suppress antitumour immunity
Regulatory B cells297–299	Protumorigenic	Produce IL-10 and TGFβ Suppress activity of Teff cells and promote that of Treg cells
Naive B cells16,32	Plastic role	Precursors to differentiated B cells Generally, less active in antitumour responses
Germinal centre B cells36,300	Antitumorigenic	Aid in antigen-specific immune responses and TLS formation Correlated with improved outcomes
Exhausted B cells301	Protumorigenic	Reduced functionality Can contribute to immunosuppression
Stromal and vascular
Myofibroblastic CAFs53	Protumorigenic	Form a fibrotic barrier Limit immune cell infiltration
Hypoxic CAFs53	Protumorigenic	Associated with hypoxic tumour regions Promote angiogenesis and immunosuppression
Inflammatory CAFs53	Antitumorigenic	Support immune cell infiltration Correlate with improved survival and an inflammatory TME
Desmoplastic CAFs53	Protumorigenic	Promote stromal remodelling Can facilitate tumour progression
Pericytes64,91	Protumorigenic	Stabilize tumour vasculature Often facilitate sustained growth and immune evasion
Vascular endothelial cells64	Protumorigenic	Support angiogenesis Regulate immune cell trafficking into the TME, often favouring tumour growth
Lymphatic endothelial cells64	Protumorigenic	Facilitate lymphatic drainage and metastasis Generally support tumour spread
High endothelial venules64,302	Antitumorigenic	Permit lymphocyte infiltration Associated with better prognosis and antitumour responses

CAF, cancer-associated fibroblast; cDC, conventional dendritic cell; DC, dendritic cell; EMT, epithelial–mesenchymal transition; ICI, immune-checkpoint inhibitor; ILC, innate lymphoid cell; NK, natural killer; TAM, tumour-associated macrophage; TAN, tumour-associated neutrophil; T_eff, effector T; T_H, helper T; TLS, tertiary lymphoid structures; TME, tumour microenvironment; T_reg, regulatory T.

Conceptually, the NSCLC TME can be classified into two principal types of spatial niches defined by immune cell density and activity: immune rich and immune poor²⁶. Each type presents specific challenges and opportunities for intervention, and their crosstalk adds layers of complexity to NSCLC pathology.

Immune-rich niches

Immune-rich spatial niches are characterized by a high density of immune cells actively engaging in antitumour responses. These areas, often located at the invasive margins or within stromal regions adjacent to tumour nests, harbour a diverse array of immune cells such as T cells, B cells, natural killer (NK) cells, dendritic cells (DCs) and tumour-associated macrophages (TAMs)^22,23. The spatial proximity of these cells facilitates robust interactions and effective antitumour responses (Fig. 1).

Fig. 1 |. Spatial dichotomy of immune-poor and immune-rich niches in the NSCLC TME.

a, The tumour microenvironment (TME) in non-small-cell lung cancer (NSCLC) is characterized by a high degree of spatial heterogeneity, which is depicted in this figure highlighting immune-rich and immune-poor niches. Immune-rich niches foster robust immune activity, with closer spatial interactions between CD8⁺ cytotoxic T cells, B cells, dendritic cells (DCs) and tumour-associated macrophages (TAMs), and cancer cells. This proximity enables direct cytotoxic interactions and facilitates antitumour immunity, particularly within tertiary lymphoid structures, in which B cells and plasma cells enhance immune priming and adaptive immune responses. Immune-poor niches are characterized by the presence of dense extracellular matrix (ECM) networks, which physically restrict immune infiltration. Cancer-associated fibroblasts (CAFs) secrete immunosuppressive cytokines and remodel the ECM to create a fibrotic, immune-excluded microenvironment, reducing direct immune–cancer cell interactions. Myeloid-derived suppressor cells (MDSCs), tumour-associated neutrophils (TANs) and regulatory T (T_reg) cells further contribute to immunosuppressive signalling, leading to reduced immune activation and a weakened antitumour response. Cellular interactions in immune-rich (part b) and immune-poor niches (part c) are also shown. In immune-rich niches, immune cells are tightly clustered with cancer cells, promoting immune-mediated tumour control. By contrast, immune-poor regions are dominated by stromal barriers and suppressive cell populations, which create spatial separation between immune and cancer cells, impeding immune infiltration and tumour recognition.

At the forefront are effector CD8⁺ cytotoxic T lymphocytes (CTLs) that are able to directly kill cancer cells through recognition of tumour-associated antigens presented by MHC class I molecules. The presence of these cells is generally associated with improved patient outcomes and serves as a favourable prognostic marker^15,27,28. CD4⁺ helper T (T_H) cells, particularly the T_H1 subpopulation, support the function of CTLs by secreting essential cytokines, such as IL-2 and IFNγ^29,30. Immunity hubs enriched in T cell-attracting chemokines and an abundance of T cells have been identified in pretreatment biopsy tissue samples from patients who subsequently had a response to ICIs (or responders). Within these hubs, researchers identified a distinct niche referred to as the stem-immunity hub, which is enriched in stem-like TCF7⁺PD1⁺CD8⁺ T cells, activated CCR7⁺LAMP3⁺ DCs, CCL19⁺ CAFs and CXCL10⁺ TAMs preferentially engaging TCF7⁻CD8⁺ T cells, underscoring a spatially organized, chemokine-driven network crucial for effective PD-1–PD-L1 blockade¹⁴. Furthermore, single-cell RNA sequencing and spatial transcriptomic analyses of pretreatment NSCLC biopsy samples has revealed that responders to chemotherapy–ICI combinations had higher levels of expression of interferon-stimulated genes, including CXCL9, CXCL10 and CXCL11, relative to nonresponders³¹. Post-treatment spatial transcriptomic profiling has provided further insight into this chemokine–T cell interplay by locating CXCL9 in the tumour–stromal interface, where it co-localizes with CTLs, thereby establishing a positive feedback loop involving IFNγ secretion and T cell recruitment³¹.

B cells and plasma cells contribute to humoral immunity by participating in antigen presentation and antibody production in the NSCLC TME^23,32,33. Notably, the presence of B cell markers, specific B cell subsets and plasma cells correlates with an improved prognosis and superior efficacy of ICIs in patients with NSCLC, suggesting that humoral immunity has a major role in antitumour responses^28,34,35. The formation of tertiary lymphoid structures (TLS), which are organized aggregates resembling secondary lymphoid organs, is a signature of immune-rich niches. TLS facilitate efficient antigen presentation and lymphocyte activation, and their presence has been associated with superior clinical outcomes, including prolonged overall survival and improved responses to ICIs^28,36,37. Spatial profiling analysis has delineated distinct TLS states ranging from nascent ‘lymphoid aggregates’ to more-advanced ‘activated’, ‘declining’ and ‘late’ TLS within the NSCLC TME³¹. Complete responders to chemotherapy–ICI combinations frequently exhibited late TLS characterized by T cell exhaustion markers (such as IFNG or HAVCR2, encoding IFNγ and TIM3, respectively), implying that a robust CTL response might be more conducive to therapeutic efficacy than continuous germinal centre activity³¹. By contrast, nonresponders frequently had activated TLS with robust germinal centre signatures yet concomitant hypoxia, which can inhibit the maturation of these structures³¹.

DCs, particularly conventional type 1 DCs, have a pivotal role in antigen presentation and activation of naïve T cells. Their abundance enhances the initiation and maintenance of effective immune responses against cancer cells³⁸. Conversely, the presence of immunoregulatory PD-L1⁺ DCs might contribute to resistance to ICIs^39,40.

TAMs are among the most abundant immune cells in the TME and have remarkable levels of heterogeneity and adaptability, which substantially influence tumour progression and responses to treatment^41–43. Their activation exists as a continuum, dynamically shaped by TME cues, beyond simple binary M1–M2 polarization. CXCL9⁺ macrophages, driven by IFNγ signalling, orchestrate effector T (T_eff) cell recruitment via CXCR3 ligands, enhancing immune infiltration and response to ICIs⁴⁴. The presence of these cells correlates with an immunologically hot TME and improved outcomes in patients with NSCLC. By contrast, TAMs expressing SPP1 (which encodes secreted phosphoprotein 1) promote ECM remodelling and immune exclusion, with a low CXCL9:SPP1 ratio linked with resistance to ICIs and poor prognosis across several solid tumour types including NSCLC^31,45. TAMs expressing COL1A1 (encoding collagen α1 chain) reinforce a fibrotic, ECM-rich TME that physically and immunologically restricts T cell infiltration, thus dampening the efficacy of ICIs⁴⁴. Similarly, TREM2⁺ macrophages, more abundant in lipid-rich and hypoxic niches, sustain an immunosuppressive phenotype by curbing NK cell activity and expanding exhausted CD8⁺ T cell populations and fostering an immunologically cold TME⁴⁶.

NK cells are key components of the innate immune system with the ability to recognize and kill cancer cells without previous sensitization. In the NSCLC TME, NK cells are often depleted or functionally impaired owing to factors such as hypoxia and the expression of inhibitory ligands by cancer cells^16,46,47.

The adaptive coordination within immune-rich niches presents therapeutic opportunities. ICIs, such as antibodies targeting PD-1 or PD-L1 (PD-(L)1) or CTLA4, have demonstrated increased efficacy in patients with immune-rich TMEs owing to the abundance of their cellular targets^30,48. Combining ICIs with agents that stimulate immune cell function or promote TLS development could further enhance clinical outcomes. For example, analysis of NSCLCs from patients who received combinations comprising of the anti-PD-1 nivolumab, the anti-CTLA4 antibody ipilimumab and standard-of-care chemotherapy revealed activation of chemokine signalling pathways, such as the CXCL13–CXCR5 axis, thereby facilitating lymphocyte recruitment and TLS maturation²⁸.

Understanding the mechanisms governing the formation and maintenance of immune-rich niches could unlock new avenues for enhancing antitumour immunity. The potential to manipulate TLS as biomarkers for patient stratification and treatment optimization constitutes an exciting frontier in NSCLC research³⁷.

Immune-poor niches

Immune-poor niches in the NSCLC TME are often characterized by structural and biochemical barriers that impede immune cell infiltration. This barrier is reinforced by a high level of stromal complexity, which can be measured quantitatively as a high fractal dimension, highlighting the physical structure that actively excludes lymphocytes from tumour sites⁴⁹ (Fig. 1). In these niches, tumours evade immune detection through both intrinsic and extrinsic mechanisms attributed to low antigenicity, and the secretion of immunosuppressive factors⁴⁹. Cancer cells can have low immunogenicity owing to downregulation of components of the antigen-presentation machinery, such as MHC molecules, or a low neoantigen load resulting from limited expression of immunogenic mutations⁵⁰. Additionally, they can secrete immunosuppressive factors such as VEGF, IL-10 and indoleamine 2,3-dioxygenase 1, creating a milieu hostile for immune cell infiltration and function⁵¹. Systemic factors, including chronic inflammation and elevated levels of immunosuppressive myeloid cells, reinforce the immune-desert phenotype⁵¹.

CAFs have a central role in immune-poor niches⁴⁹. Certain subtypes, such as myofibroblastic CAFs and desmoplastic CAFs, contribute to ECM deposition and remodelling, leading to increased tissue stiffness (which effectively limits immune cell infiltration) and altered tumour architecture^52,53. CAFs can also secrete a plethora of growth factors, cytokines and chemokines, including TGFβ, which promotes fibrosis and induces immunosuppression by inhibiting T cell proliferation and function⁵⁴. The dense ECM and abnormal vasculature resulting from CAF activity creates a hostile environment for immune cells, effectively excluding them from the tumour core. Spatial transcriptomic analyses of NSCLC biopsy samples from patients following treatment with chemotherapy–ICI combinations revealed an enrichment in COL11A1⁺GREM1⁺ CAFs and SPP1⁺ TAMs adjacent to cancer cells, and their increased abundance correlated with a higher tumour burden and poor treatment outcomes. The COL11A1–SPP1 axis has been hypothesized to physically obstruct infiltration of CTLs, thereby undermining effective antitumour immunity³¹. These findings support the notion that immune desertification does not merely involve the passive absence of immune cells but is an exclusionary process driven by the spatial and structural configuration of stromal cells in the TME.

The abnormal vasculature of immune-poor niches, influenced by signals from CAFs and cancer cells, is characterized by disorganized and leaky blood vessels. This dysfunction leads to inefficient oxygen delivery and hypoxia, thus exacerbating immunosuppression by inducing the expression of hypoxia-inducible factors (HIFs) and upregulation of immune checkpoints, such as PD-L1, on cancer and stromal cells⁵⁵. Hypoxia also promotes the recruitment of immunosuppressive cells such as myeloid-derived suppressor cells (MDSCs) and polarizes macrophages towards a tumour-promoting phenotype¹³.

Neutrophils within immune-poor niches have substantial plasticity, which enables transitions between protumorigenic and antitumorigenic states^38,56,57. Tumour-associated neutrophils (TANs) can promote angiogenesis, tumour growth and metastasis through the secretion of factors such as VEGF⁵⁸. The polarization of TANs is influenced by pathways such as TGFβ–Smad3 signalling, with the activation of Smad3 promoting a protumorigenic phenotype⁵⁹.

Addressing the challenges posed by immune-poor niches requires strategies that enable remodelling of the NSCLC TME to facilitate immune cell infiltration. Agents that stimulate innate immunity, such as Toll-like receptor agonists or oncolytic viruses, have been shown in preclinical models and early phase clinical trials to induce local inflammation and attract immune cells to the tumour site^60–62. Epigenetic modifiers can upregulate the expression of tumour antigens and MHC molecules, enhancing tumour visibility to the immune system, as demonstrated in preclinical models and early phase clinical studies⁶³. Although antiangiogenic agents, such as the anti-VEGF antibody bevacizumab, have shown limited ability to improve the clinical outcomes of patients with NSCLC when used as monotherapy, their efficacy might improve when administered in combination with other therapeutic classes or with optimized timing strategies^64–66. Targeting CAFs to disrupt the fibrotic stroma using agents such as the anti-inflammatory and antifibrotic agent pirfenidone might reduce physical barriers and modulate the immunosuppressive milieu⁶⁷. Given the complexity and heterogeneity of CAFs, whether we can selectively target protumorigenic CAF subtypes without disrupting those that might have tumour-suppressive functions is an important consideration. Re-educating CAFs to adopt an immune-supportive phenotype could have implications for overcoming therapeutic resistance.

Combining TME-remodelling approaches with ICIs might help to convert immune-desert niches into immune-infiltrated ones, potentially rendering tumours susceptible to immunotherapy. Nonetheless, whether immune-desert niches are in an irreversible state or whether they can be reprogrammed into immune-responsive environments remains to be determined.

Spatial topography of the NSCLC TME

Advances in spatial profiling have enabled the granular delineation of functionally discrete microenvironments based on aspects beyond the immune-based classification discussed, such as tumour topography and histo-architectural patterning⁶⁸. Spatial profiling approaches now capture intratumoural heterogeneity not merely as a compositional variation, but also as dynamic molecular and cellular gradients that propagate radially from the tumour core to the invasive margin, encompassing the peritumoural stroma, lymphoid aggregates and TLS^6,26. Such refined topographical mapping can elucidate how gradients of hypoxia, nutrient availability and immune infiltration coalesce into distinct microecological niches, each exerting selective pressures on tumour evolution and response to therapy⁶⁸. LUAD provides a compelling example: the heterogeneous histological subtypes (lepidic, acinar, papillary, micropapillary and solid) reflect intrinsic tumour organization and morphology features that, in turn, could possibly frame the spatial distribution and phenotypic polarization of immune and stromal elements^13,69–72. A study described a correlation between the presence of micropapillary and solid patterns in lymph node metastases, and inferior prognosis⁷³, which might stem not solely from their intrinsic molecular underpinnings or driver mutation landscapes, but also from how these architectural variants foster immunosuppressive microniches or limit the formation of robust immune effector archetypes¹³.

Ultimately, incorporating a topographical lens enables researchers to move beyond binary categorization (immune rich versus immune desert), forging a path towards predictive atlases that map the interplay of tumour architecture, microenvironmental composition and functional states. Such atlases can guide the rational design of combination therapies, inform patient stratification and perhaps help to identify underappreciated therapeutic targets defined not only by molecular subtype but also by the spatial organization of the TME.

Interplay of spatial niches and TME factors

Understanding the characteristics of individual spatial niches is paramount. Nevertheless, the interplay among these niches and the external factors that modulate them are the aspects that ultimately dictate tumour progression and response to therapy.

Interconnectedness of spatial niches

The spatial niches within the NSCLC TME coexist and profoundly influence each other, forming a highly interactive ecosystem (Fig. 2). Hypoxic conditions within the tumour core can activate CAFs, leading to increased production of ECM and the development of fibrotic niches. This fibrosis exacerbates hypoxia by impeding efficient blood flow, creating a vicious cycle that promotes tumour survival and aggressiveness⁷⁴. The dense ECM and abnormal vasculature act as physical barriers, preventing immune cells generated and recruited in immune-rich niches, such as TLS, from infiltrating the tumour core and exerting their antitumour effects⁷⁵.

Fig. 2 |. Interconnected cellular and molecular networks shaping the NSCLC TME.

The dynamic crosstalk between cancer cells, immune effectors, stromal components and systemic factors collectively influences cancer progression and response to therapy in patients with non-small-cell lung cancer (NSCLC). Higher neoantigen burdens and the formation of tertiary lymphoid structures (TLS) support robust immune infiltration, promoting closer proximity between cytotoxic lymphocytes and malignant cells and enhancing tumour control. Counterbalancing these effects are regulatory T (T_reg) cells, tumour-associated macrophages (TAMs), regulatory dendritic cells (DCs) and tumour-associated neutrophils (TANs), which together dampen immune activation. Concurrently, hypoxia-induced activation of cancer-associated fibroblasts (CAFs) and extracellular matrix (ECM) deposition create fibrotic barriers that impede immune cell access and perpetuate immune desertification in tumours. Chronic IFNγ signalling, driven by CD8⁺ T cells and natural killer (NK) cells, contributes to immune pressure but also promotes T cell exhaustion. Additionally, TGFβ–Smad3 signalling, which can be activated by radiotherapy and hypoxia, fosters stromal remodelling and immunosuppression. Understanding these multilayered, bidirectional interactions is key to developing strategies that enhance immune infiltration, disrupt immunosuppressive networks and ultimately improve the outcomes of patients with NSCLC. EMT, epithelial–mesenchymal transition; infCAF, inflammatory CAF; MHC I, MHC class I; SMA-CAF, smooth muscle actin-expressing CAF; TCR, T cell receptor; T_FH, T follicular helper cell; TKI, tyrosine-kinase inhibitor; TME, tumour microenvironment.

Hypoxia within the TME introduces another layer of complexity. Hypoxic conditions can induce a deep senescent state in cancer cells, characterized by a senescence-associated secretory phenotype that promotes immunosuppression^76,77. HIF1α upregulation is associated with resistance to EGFR tyrosine-kinase inhibitors (TKIs) in preclinical models of NSCLC⁵⁵. Targeting pathways that regulate hypoxia, including those modulated by TGFβ, in combination with ICIs could help to overcome the resistance to these agents in hypoxic TMEs. In a mouse model of breast cancer, the combination of an anti-TGFβ antibody with ICIs led to such changes in the TME by promoting lymphocyte infiltration and antitumour immunity⁷⁸.

Immunosuppressive cells recruited to hypoxic and fibrotic areas can migrate to immune-rich areas, further dampening immune responses. Regulatory T (T_reg) cells are among the key mediators of immunosuppression within the TME. They overexpress molecules such as PR domain zinc finger protein 1 (commonly referred to as BLIMP1), IL-10 and IL-35, inducing CD8⁺ T cell exhaustion and diminishing their effector function⁷⁹. The proximity of T_reg cells to T_eff cells and B cells can negate the potential survival benefits conferred by the presence of the latter immune cells alone^{13,29,37,80,81}. Conversely, an enrichment in the environments in which T_eff cells and B cells interact without the inhibitory influence of T_reg cells is associated with enhanced antitumour immunity and superior clinical responses to ICIs in patients with sarcoma⁸². In a mouse model of LUAD, depletion of T_reg cells resulted in reprogramming of the TME, altering the expression profiles of fibroblasts, endothelial cells and TAMs, and enhancing VEGF signalling⁸³. This reprogramming led to increased vascularization and immunomodulation of innate and adaptive immune cells.

Environmental factors, such as tobacco smoke, further reinforce these immunosuppressive dynamics, exacerbating the spatial and functional heterogeneity of the NSCLC TME. Smoking-related tumours have higher PDL1 levels than non-smoking-related tumours^84,85, and an increased tumour mutational burden (TMB) and neoantigen burden, fostering chronic antigenic stimulation and immune editing^86,87. Persistent tobacco smoking drives inflammation, further upregulating PDL1 through activation of aryl hydrocarbon receptor–mTOR signalling, reinforcing immune evasion⁸⁸. Therefore, smoking-related alterations generate an inflamed yet immunosuppressive milieu, in which CTLs coexist with a dense infiltrate of T_reg cells and exhausted T cells, thus blunting antitumour immunity⁸⁹. Despite high levels of CD8⁺ T cell infiltration, the effector functions are suppressed by the proximity of T_reg cells and other immunosuppressive factors, mirroring the spatial constraints observed in fibrotic and hypoxic niches. This paradox contributes to a scenario in which individuals who smoke generally derive greater clinical benefit from ICIs albeit with counterbalancing resistance mechanisms that necessitate combination strategies to optimize responses^86,87. In EGFR-mutant LUAD, tobacco smoking further modifies the TME, shifting it towards immune exclusion and stromal activation. Unlike other smoking-driven NSCLCs, these tumours have a lower TMB, reduced immune cell infiltration and abundance of fibroblast-enriched niches, leading to enhanced resistance to both ICIs and EGFR TKIs^89,90.

In addition to immune modulation, stromal cell interactions further reinforce the complexity of the NSCLC TME. Interactions between CAFs, pericytes and endothelial cells, driven by activation of NOTCH signalling (and particularly NOTCH3), further illustrate the interconnectedness of the TME^64,91 (Fig. 2). NOTCH3-dependent signalling promotes tumour invasion, collagen production and expression of a TGFβ-related signature associated with a poor prognosis in patients with NSCLC^64,91. Disrupting these signalling pathways might impede tumour progression and enhance therapeutic responses, highlighting the potential of targeting cell–cell interactions within the TME.

These examples underscore that the spatial niches within the TME are not isolated compartments but dynamically interacting domains. Recognizing these interactions is essential for developing strategies that can effectively remodel the TME and ultimately result in improved clinical outcomes.

Influence of systemic and microenvironmental factors on niche interactions

Systemic factors originating outside of the NSCLC TME, including chronic inflammation, immunosuppressive signals and systemic cytokine levels, along with microenvironmental conditions such as hypoxia, have a major influence on the TME and modulate interactions among its spatial niches (Fig. 2). Inflammation, driven by certain cytokines and chemokines, can paradoxically foster an environment conducive to disease progression within tumour-supportive niches. For example, persistent IFNγ signalling within the TME is associated with immune dysfunction and acquired resistance to ICIs^92–94. Although IFNγ signalling is key for antitumour immunity, chronic activation of this pathway can lead to T cell exhaustion and upregulation of inhibitory pathways, resulting in a sustained but incomplete antitumour response to ICIs, as has been described in a mouse model of melanoma⁹⁵. Modulating IFNγ-related responses could help to overcome resistance mechanisms and restore effective immune surveillance.

Systemic factors originating outside tumour niches also influence the TME. Elevated activation of IL-4-dependent signalling in bone marrow myeloid progenitors leads to the generation of immunosuppressive myeloid cells, contributing to a poor prognosis in patients with NSCLC⁹⁶. In a first-in-human trial involving 6 patients with NSCLC, addition of the anti-IL-4Rα antibody dupilumab to anti-PD-(L)1 antibodies showed preliminary efficacy, and analysis of paired biopsy samples showed evidence of reduced circulating immunosuppressive monocytes and enhanced infiltration of CD8⁺ T cells⁹⁶.

The complement system adds yet another dimension to the interplay within the TME, owing to its dual role in tumour immunity. Inactivation of the complement cascade within the TME, which is mediated by upregulation of complement regulatory proteins CD55 and CD59, decreases CD8⁺ T cell cytotoxicity and facilitates tumour immune evasion⁹⁷. In a mouse model of lung cancer, combining antibodies targeting CD55 or CD59 with those targeting PD-1 had a synergistic antitumour effect, suggesting a new combinatorial therapeutic approach⁹⁷. However, certain complement components, such as C5a, can promote tumour progression by enhancing immunosuppression through the recruitment of MDSCs and inhibition of CD8⁺ T cell function⁹⁸. Therefore, therapeutic strategies leveraging the complement system must carefully consider this complexity to avoid unintended protumorigenic effects.

Patient-specific determinants

In the pursuit of personalized medicine and more-effective therapeutic strategies for patients with NSCLC, consideration of all influences that shape the TME is essential. An emergent body of evidence is showing that patient-related factors, such as sex, ageing, obesity and health disparities, introduce additional layers of complexity (Fig. 3).

Fig. 3 |. Patient-specific determinants of the NSCLC TME.

Age-related changes in the tumour microenvironment (TME) include an increase in immunosuppressive myeloid populations, reduced CD8⁺ T cell function and upregulation of inhibitory receptors, leading to decreased efficacy of immune checkpoint inhibitors (ICIs). Sex differences also affect responses to therapy, with males having androgen-driven CD8⁺ T cell exhaustion, while females show higher infiltration of immune cells but increased dominance of regulatory T (T_reg) cells. Obesity induces chronic inflammation, promoting T cell exhaustion but also increasing sensitivity to ICIs. Diet modulates immunity both directly through inflammatory pathways and indirectly via the microbiota, which can either promote or suppress antitumour immunity. Ethnic and health disparities, including socioeconomic status and access to therapies, further affect the TME composition and treatment efficacy. NSCLC, non-small-cell lung cancer.

Sex-based differences

Sex-specific hormones and genetic differences contribute to the distinct biology of NSCLC and variations in immune responses. In a mouse model of urothelial carcinoma, androgen signalling was associated with an increased presence of progenitor exhausted CD8⁺ T cells, leading to inferior control over tumour growth in males compared with females⁹⁹. In samples derived from treatment-naïve patients with NSCLC, exhausted CD8⁺ tumour-infiltrating lymphocytes (TILs) were more prevalent in males, correlating with decreased efficacy of ICIs⁹⁹. In preclinical models, androgen-depriving agents and androgen receptor antagonists have shown promise in enhancing the activity of ICIs across various cancer types, potentially owing to increased expression of IFNγ^99,100. In females, oestrogen and oestrogen receptor signalling can promote tumour progression and influence response to therapy¹⁰¹. In a mouse model of melanoma, oestrogen signalling led to TAM polarization towards an immunosuppressive phenotype, leading to CD8⁺ T cell dysfunction and exhaustion¹⁰². This phenotype was associated with resistance to ICIs, which could be overcome by combining ICIs with the selective oestrogen receptor degrader fulvestrant. Analyses of samples derived from patients with NSCLC revealed increased infiltration of immune cells, including naïve CD4⁺ and CD8⁺ T cells, NK cells, antigen-presenting cells and M1 TAMs, in females relative to males^13,103. This increased presence of immune cells might contribute to superior overall survival in females, although females also have higher T_reg:CD8⁺ T cell ratios and greater expression of immune cell exhaustion markers than males, indicating a more suppressed immune phenotype despite increased immune infiltration^104–107. These findings suggest that although the immune response in females is more robust, it is often counterbalanced by increased immunosuppression within the TME.

The sex-specific prevalences of certain oncogenic mutations further influence the TME and response to therapy. For example, among patients with LUAD, EGFR mutations are more common in females who have never smoked than in other demographic groups, with reported rates as high as 40–60% in Asian populations and 10–15% in non-Asian cohorts¹⁰⁸, and are associated with an immune-depleted TME¹⁰⁹. KRAS mutations — particularly KRAS^G12C, the most common KRAS mutation in LUAD, which accounts for ~15% of LUAD and ~40% of KRAS-mutant tumours — are also more prevalent in females^110–112. In a mouse model of KRAS-mutant LUAD, knockout of STAT3 resulted in sex-specific effects on tumour progression¹¹³. Female mice had enhanced antitumour immune responses and a reduced tumour burden, whereas males showed increased tumour growth owing to a tumour-promoting immune phenotype characterized by activation of NF-κB signalling, a CXCL1-mediated neutrophil response and IL-6 induction¹¹³. Antibody-mediated IL-6 blockade and neutrophil depletion in males mitigated these tumour-promoting effects, whereas in females, oestrogen-dependent signalling mediated the antitumour effect of STAT3 knockout¹¹³. In female mice, tamoxifen reversed this benefit, highlighting the interplay between hormone signalling and immune responses¹¹³.

Sex-based differences also extend to treatment efficacy and adverse events. Meta-analyses of data from randomized controlled trials have revealed that males with advanced-stage NSCLC derive more benefit from single-agent ICIs, such as anti-PD-1 or anti-CTLA4 antibodies¹⁰⁶ but less benefit from combinations of ICIs or ICI–chemotherapy¹⁰⁵ than females. Additionally, females are more likely to have immune-related adverse events, which can affect treatment adherence and quality of life¹¹⁴. These findings underscore the importance of considering sex as a biological variable in treatment planning. Moreover, the hormone therapies used in gender-affirming treatments can modulate immune responses and potentially influence the TME¹¹⁵. Testosterone treatment has been associated with suppression of type I interferon responses, particularly in plasmacytoid DCs and monocytes, whereas oestradiol potentiates type interferon I responses¹¹⁵. Testosterone also enhances TNF responses, leading to increased secretion of IL-6 and IL-15, and expression of NF-κB, which can modulate NK cell activity and potentially contribute to T cell exhaustion¹¹⁵.

Ageing

Ageing leads to substantial changes to the TME, affecting immune cell composition, function and overall tumour behaviour. Analysis of samples derived from patients with LUAD revealed that those aged >70 years generally have a more immunosuppressive TME than patients under that age¹¹⁶. LUAD biopsy samples from older patients show increased expression of immunosuppressive proteins, such as PD-L2 and TIM3, and higher levels of immunosuppressive TAMs and fewer TANs¹¹⁶. This TME correlates with reduced T cell function and overall immunity, leading to decreased overall survival in older patients with early stage disease, but not in those with advanced-stage disease¹¹⁶. Ageing is also associated with reduced infiltration of CD8⁺ T cells and inadequate activation of these cells within tumours, in a phenotype that differs from conventional T cell exhaustion^13,14,117. This phenotype is characterized by a marked reduction of subpopulations of exhausted T cells, with a shift towards less differentiated states such as naïve-like and memory-like T cells¹¹⁷. Additionally, diminished crosstalk among immune cells, including NK cells, DCs and T cells, leads to impaired immune responses^13,117. The ageing TME has an enhanced tumour-induced myelopoietic response, with increased presence of myeloid progenitor-like cells and elevated production of IL-1α¹¹⁸, which is known to have tumour-promoting roles, associated with inferior survival and cancer recurrence¹¹⁹.

Furthermore, ageing is linked to increased levels of certain metabolites such as methylmalonic acid, which can affect tumour progression¹²⁰. In a mouse model of NSCLC, elevated levels of methylmalonic acid promoted cancer aggressiveness and metastatic potential by inducing the expression of epithelial-to-mesenchymal transition markers such as fibronectin and vimentin, and reducing that of E-cadherin¹²⁰. Methylmalonic acid also enhanced activation of TGFβ signalling and expression of SOX4, both associated with poor prognosis and tumour progression¹²⁰.

Older patients tend to derive less benefit from standard-of-care therapies, including chemotherapy and ICIs. Retrospective studies have shown that ICIs are less efficacious in older patients with NSCLC, who have a shorter duration of overall survival compared with younger patients despite a similar duration of progression-free survival (PFS)^116,121. In one such study, patients aged >75 years but not those aged ≤55 years derived minimal benefit from ICIs, emphasizing the need to consider patient age in therapeutic decision-making¹²¹.

Despite the generally immunosuppressive and protumorigenic features of the ageing TME, the incidence of NSCLC decreases after the age of 80 years¹. This paradox raises intriguing questions about whether ageing-related changes might limit tumour initiation at extreme ages. For example, one study suggests that functional iron insufficiency, mediated by the nuclear protein 1–lipocalin-2 axis, could suppress alveolar stem cell renewal and tumorigenesis, potentially pointing to epigenetic mechanisms that shape the TME of older individuals¹²². Such mechanisms could constitute an underexplored intersection between cellular senescence and microenvironmental remodelling, with implications for therapeutic strategies for older patients.

Obesity, microbiota and diet

Obesity has emerged as a key modifier of the NSCLC TME and therapeutic outcomes. The ‘obesity paradox’, a phenomenon whereby a higher body mass index (BMI) is associated with longer overall survival in patients with cancer, has been observed in NSCLC, particularly in patients receiving ICIs^123,124. Obesity typically induces chronic inflammation and T cell exhaustion, resulting in immunosuppression, although evidence suggests that these effects might be counterbalanced by metabolic and immune reprogramming in the presence of ICIs¹²⁴.

Factors that are overexpressed by adipose tissue¹²⁵, such as leptin, IL-6, and TNF, promote T cell exhaustion and tumour progression but might also sensitize the NSCLC TME to ICIs by elevating PD-1 expression on T cells, creating a dependency on pathways targeted by these therapeutic agents¹²⁴. Leptin signalling, which contributes to T cell exhaustion and increased PD-1 expression, seems to have a key role. Consequently, PD-1 blockade in patients with obesity has been shown to reverse T cell dysfunction, leading to improved PFS and overall survival compared with those without obesity¹²⁴. Notably, additional benefits have been observed with the use of the glucose-lowering agent metformin, particularly in patients with obesity treated with ICIs^123,126. In a preclinical study, metformin enhanced the efficacy of ICIs by reprogramming the immune TME, reducing activity of T_reg cells and improving T_eff cell function¹²³.

The microbiome has emerged as a modifiable factor influencing the NSCLC TME and response to ICIs. Preclinical and clinical studies have shown that gut and lung microbiota can modulate systemic immune responses, directly affecting the efficacy of ICIs^127,128. Imbalances in microbial populations and the presence of pro-inflammatory bacteria, such as members of the Alistipes genus, can foster immunosuppression, promote tumour progression and drive resistance to therapy. These microbial shifts also alter the profiles of cytokines in the TME (for example, IL-6), further exacerbating resistance mechanisms¹²⁹. Conversely, certain bacterial genera, such as Akkermansia, Ruminococcus and Bifidobacterium, have been associated with enhanced antitumour immunity and improved responses to anti-PD-(L)1 antibodies^130–133. Nonetheless, the influence of specific microbial species on the TME and response to ICIs is highly context-dependent, varying across patients based on host immunity, tumour biology and broader microenvironmental factors¹³³.

The microbiome has potential as a non-invasive biomarker for predicting therapeutic outcomes^134,135. Faecal microbiota profiling can help to identify patients who are more likely to benefit from ICIs¹³⁰. This observation has spurred the development of microbiome-targeted interventions, including administration of probiotics and/or prebiotics, dietary modifications and faecal microbiota transplantation to modulate TME dynamics, enhance T cell function and reverse immunosuppression^136,137.

Diet has a pivotal role both in managing obesity and shaping microbiome composition, positioning it as a central modulator of the NSCLC TME. Diet-induced microbiome shifts can affect immune cell function, inflammatory responses and metabolic pathways within the TME, potentially altering the balance between immune activation and suppression^138,139. Together, these factors provide new avenues for designing personalized strategies aimed at overcoming TME-driven resistance and improving patient outcomes.

Health disparities

Health disparities, including ethnic and socioeconomic factors, contribute to variations in NSCLC incidence and prognosis, TME composition and patient outcomes in response to therapy. As discussed for sex disparities, certain oncogenic mutations have a differential prevalence among ethnic groups^140–142. KRAS^G12C mutations are more common in white (14%) and Black (11%) populations compared with Asian (2%) patients with LUAD¹¹¹. Activating mutations in EGFR have been described in ~50% of Asian patients and 10–15% of non-Hispanic white individuals^143–145. Black patients with NSCLC often present at later stages compared with non-Hispanic white patients¹⁴⁶, and more often tend to have immunologically cold tumours with lower levels of immune cell infiltration, particularly in individuals who smoke¹⁴⁷. Compared with non-Hispanic white patients, the NSCLC TME from Black patients has higher levels of resting DCs and γδ T cells but fewer CD8⁺ T cells and monocytes¹⁴⁶. Conversely, immune activation, including infiltration of CD4⁺ and CD8⁺ T cells, monocytes and CD31⁺ endothelial cells, and activity of nicotine degradation signalling pathways are higher in the TME of non-Hispanic white patients compared with Black patients, especially at later disease stages, contributing to an immunologically hot TME¹⁴⁷.

Although some studies have described correlations between ethnic disparities in TME composition and overall survival, others report similar outcomes across different ethnic groups that are receiving ICIs^{111,146–148}. Factors such as Eastern Cooperative Oncology Group performance status, BMI and PD-L1 expression seem to be more predictive of PFS than ethnicity alone¹⁴⁸. Socioeconomic status and access to health-care resources also substantially affect patient outcomes and should be considered in the context of health disparities¹⁴⁹.

Therapeutic targeting of the TME in NSCLC

Integrating disease-specific factors such as TME characteristics, molecular biomarkers and patient-specific determinants is essential for advancing precision oncology strategies that aim to tailor treatments for patients with NSCLC to improve treatment effectiveness. Current and emerging strategies are being used to target the TME, address resistance mechanisms and emphasize the importance of patient selection and optimal biomarker use (Fig. 4).

Fig. 4 |. Therapeutic strategies modulating the NSCLC TME.

Various approaches can be used to modulate the tumour microenvironment (TME) in patients with non-small-cell lung cancer (NSCLC), enhancing immune responses and overcoming resistance. Immune checkpoint inhibitors (ICIs) target PD-1, PD-L1, CTLA4, VISTA, TIM3, TIGIT and CD73–NKG2A, reinvigorating T cell function and countering immunosuppressive elements of the TME such as regulatory T (T_reg) cells and myeloid-derived suppressor cells (MDSCs). Bispecific antibodies, such as PD-1 × VEGF, and anti-VEGF antibodies inhibit tumour angiogenesis. Chemotherapy and radiotherapy promote antigen release and infiltration of cytotoxic T lymphocytes (CTLs). Targeted therapies, including EGFR and ALK tyrosine-kinase inhibitors (TKIs), and KRAS inhibitors, modulate cytokine profiles and immune cell dynamics in the TME, and could synergize with ICIs. Adoptive cell therapy involves using chimeric antigen receptor (CAR) T cells, T cell receptor (TCR)-engineered T cells or tumour-infiltrating lymphocytes (TILs) to counteract TME-induced immunosuppression. Oncolytic viruses selectively lyse cancer cells, releasing tumour antigens that can promote immune cell infiltration, whereas cancer vaccines prime T cell responses against tumour neoantigens, reshaping immunity in the TME. ADC, antibody–drug conjugates; DC, dendritic cell; MHC I, MHC class I.

Strategies using ICIs

ICIs have revolutionized the treatment of patients with NSCLC by targeting immune-inhibitory pathways such as those downstream of PD-1, PD-L1 and CTLA4, thereby reactivating antitumour immune responses within the TME^150–153. These approaches were initially deemed successful in patients with advanced-stage disease, although their use has been progressively extended to the locally advanced unresectable and early stage resectable settings, reflecting a paradigm shift in treatment strategies¹⁵⁰. The efficacy of ICIs is closely linked to the immunological landscape of the TME. Tumours with an inflamed or immunologically hot TME, characterized by high infiltration of T_eff cells and elevated PD-L1 expression, are more responsive to ICIs. By contrast, those with an immunosuppressive or immunologically cold TME, dominated by T_reg cells, MDSCs and TAMs, often have resistance to ICIs^28,151,152.

Combining chemotherapy with ICIs enhances the immunogenicity of the TME (Fig. 4). Chemotherapeutic agents can induce immunogenic cell death, promote antigen presentation and reduce immunosuppressive cell populations, thereby converting an immunologically cold TME into a hot one¹⁵⁴. This synergistic effect has been observed across various stages of NSCLC^155–158. However, certain cytotoxic regimens, particularly at high doses, can exert immunosuppressive effects by inducing myelosuppression, depleting effector lymphocyte subsets and impairing antigen presentation pathways¹⁵⁹. Thus, the net effect of chemotherapy on the TME depends on the balance between its immunostimulatory and immunosuppressive effects, highlighting the need for strategic selection of agents, doses and schedules to maximize synergy with ICIs¹⁶⁰. The addition of ICIs to neoadjuvant chemotherapy can lead to increased infiltration of effector memory CD8⁺ T cells and B cells, and the formation of TLS within the TME, correlating with improved pathological responses and long-term survival benefits^161–163. In locally advanced unresectable NSCLC, consolidation treatment with ICIs after chemoradiotherapy can enhance antigen presentation and immune cell infiltration, resulting in prolonged PFS and overall survival^164,165. However, not all patients benefit equally from this approach, highlighting the influence of TME heterogeneity on therapeutic outcomes¹⁶⁶.

Radiotherapy is integral to the management of patients with early stage unresectable NSCLC and substantially affects the TME^167,168. Exposure to radiation can enhance release of tumour antigens, promote activation of DCs and increase infiltration of CTLs into the tumour, augmenting the antitumour immune response^169,170 (Fig. 4). The abscopal effect, whereby localized delivery of radiotherapy leads to systemic antitumour effects, is thought to result from immune activation within the TME^171,172. Its mechanistic basis, however, remains incompletely understood, and its occurrence in clinical practice remains rare and inconsistent in solid tumours¹⁷³. Conversely, radiotherapy can induce immunosuppressive pathways in the TME through upregulation of immunosuppressive cytokines (such as TGFβ) and activation of CAFs, leading to a remodelling of the ECM that hinders immune cell infiltration¹⁷⁴. Combining radiotherapy with ICIs has shown synergistic effects, enhancing antitumour immunity by modulating the TME towards immune activation^168,175. By contrast, radiotherapy alone or with chemotherapy can perpetuate immunosuppressive TME states through recruitment of myeloid cells or hypoxia, which could limit the efficacy of ICIs if treatment timing or dosing fails to align with immune priming^176,177. Emerging studies are now exploring how radiotherapy dose fractionation and sequencing with ICIs can further refine modulation of the TME to potentially mitigate immunosuppressive rebound effects.

Targeted therapies

Oncogenic driver alterations, such as mutations in EGFR and KRAS, not only promote tumour proliferation but can also shape the TME^178–180 (Fig. 4). EGFR mutations are associated with an immunosuppressive TME characterized by increased expression of PD-L1 and infiltration of T_reg cells^55,141,180. EGFR TKIs effectively inhibit tumour growth but can alter the TME by affecting cytokine profiles and immune cell populations^{55,109,141,180–182}. Combinations of EGFR TKIs with ICIs has yielded limited improvements in efficacy and increased toxicity, possibly owing to complex interactions between EGFR-dependent signalling and immune regulation within the TME^141,180,182. Careful management is required to mitigate overlapping toxicities and improve efficacy.

KRAS mutations are prevalent in LUAD and have historically been challenging to target therapeutically. In the past few years, selective KRAS^G12C inhibitors have demonstrated some clinical efficacy and the ability to modulate the TME by enhancing immune cell infiltration and cytokine production^183,184. In mouse models of pancreatic and lung cancer, exposure to these inhibitors was associated with increased infiltration of CD8⁺ T cells and NK cells, as well as elevated secretion of pro-inflammatory cytokines within the TME¹⁸³. Resistance to KRAS^G12C inhibitors often involves reactivation of downstream signalling pathways and adaptation of the TME to promote immune evasion¹⁸⁵. Combination therapies targeting multiple TME pathways, including those involving ICIs, are being explored in an attempt to overcome resistance and enhance efficacy¹⁸⁶. Emerging therapies are under investigation to target mutations other than KRAS^G12C, such as new inhibitors targeting other KRAS alleles and pan-KRAS inhibitors. These agents hold potential for modulating the TME and improving therapeutic responses, emphasizing the need to consider the variability of KRAS mutations when planning treatment strategies^187,188.

Overcoming resistance to ICIs and targeted therapies

Resistance to ICIs and targeted therapies is often mediated by adaptations of the TME to enable immune evasion^189–191. Mutations in tumour suppressor genes such as STK11 and KEAP1 are linked to an immune-suppressive TME and poor responses to ICI^190,192,193. These mutations can decrease the expression of immunostimulatory cytokines and increase the recruitment of immunosuppressive cells, contributing to an immunologically cold TME¹⁹⁰. In a mouse model of NSCLC, dual immune-checkpoint blockade (targeting PD-(L)1 and CTLA4 pathways) modulated the TME to overcome resistance¹⁹² through enhanced T cell activation and infiltration within the TME. In patients with ICI-resistant NSCLC harbouring STK11 and/or KEAP1 mutations, treatment with the anti-PD-L1 antibody durvalumab plus the anti-CTLA4 antibody tremelimumab resulted in superior overall survival relative to durvalumab–chemotherapy or chemotherapy alone¹⁹².

Modulation of metabolic reprogramming within the TME offers a compelling avenue for overcoming resistance to ICIs. Cancer and stromal cells within the TME produce immunosuppressive metabolites, such as adenosine and lactic acid, which inhibit T cell function, enhance TAM polarization towards immunosuppressive phenotypes and promote resistance to therapy^194–196. Targeting these pathways with small-molecule inhibitors, such as antagonists of the adenosine signalling pathway, can restore immune cell activity and have synergistic effects with ICIs, as demonstrated primarily in preclinical models^197–199.

Antiangiogenic therapy can induce durable tumour regression by restricting the tumour vasculature and promoting vascular normalization within the TME. Vascular normalization upregulates the expression of leukocyte adhesion molecules (such as ICAM1 and VCAM1) and improves perfusion and oxygenation, thereby increasing T cell infiltration and converting the TME from immunosuppressive to immunosupportive^200–202. As a result, delivering ICIs during this window of vascular normalization could improve clinical outcomes^66,203.

Emerging therapeutic approaches

Bispecific antibodies.

By simultaneously targeting multiple immune-evasion pathways, bispecific antibodies can reshape the NSCLC TME (Fig. 4). The PD-1 × VEGF bispecific ivonescimab disrupts TME-driven resistance to ICIs by modulating angiogenesis and immunosuppression, enhancing immune infiltration and reducing tumour-induced vascular barriers in PD-L1-positive, EGFR-mutant NSCLC^204–206. The PD-1 × IL-2α bispecific IBI363 fosters T cell activation and expansion within the TME, counteracting immune exhaustion and reinvigorating antitumour immunity in immunotherapy-refractory NSCLC²⁰⁷. Similarly, PD-1 × CTLA4 bispecifics such as volrustomig and cadonilimab enhance immune cell infiltration and priming in the TME, even in PD-L1-low tumours, by broadening T cell recruitment and activation. Ongoing trials involving patients with NSCLC include the phase III eVOLVE-Lung02 (ref. 208) of chemotherapy plus either volrustomig or pembrolizumab, and those testing cadonilimab in combination with antiangiogenic agents or chemotherapy^209,210.

Antibody–drug conjugates.

Antibody–drug conjugates (ADCs) are a promising class of therapeutic agents that combine targeted delivery of cytotoxic agents with immunostimulatory effects²¹¹ (Fig. 4). In NSCLC, ADCs targeting HER2, such as trastuzumab deruxtecan²¹², and TROP2, such as datopotamab deruxtecan²¹³, have shown the potential to not only eliminate cancer cells but also enhance antigen release and immune activation within the TME. In the phase II DESTINY-Lung01 trial, trastuzumab deruxtecan demonstrated antitumour activity in patients with HER2-overexpressing metastatic NSCLC, with objective response rates (ORRs) of up to 34%²¹². In the phase II TROPION-Lung05 study, heavily pretreated patients with actionable genomic alterations receiving datopotamab deruxtecan had an ORR of 36%, and even 44% in those with EGFR mutations²¹³.

Novel approaches targeting immune checkpoints.

Targeting immune checkpoints other than PD-(L)1 and CTLA4 is emerging as a strategy to block complementary immune evasion mechanisms, particularly in PD-L1-negative tumours. Antibodies targeting immune checkpoints such as TIGIT, VISTA and LAG3 (alone or combined with anti-PD-L1 antibodies), and bispecific antibodies targeting two immune checkpoints are being tested to enhance antitumour responses, particularly in immunologically cold or excluded tumours^214–224. Nonetheless, the phase III SKYSCRAPER-06 trial²²⁵, testing combination of the anti-PD-L1 antibody atezolizumab and the anti-TIGIT antibody tiragolumab in patients with metastatic NSCLC, failed to meet its primary end points of PFS and overall survival improvement, highlighting the complexity of optimizing immune checkpoint modulation within the unique TME of NSCLC²²⁶.

Combinations of standard-of-care ICIs and those targeting other immune checkpoints, such as the anti-CD73 antibody oleclumab and the anti-NKG2A antibody monalizumab, are being tested in clinical trials owing to their potential to disrupt immune-suppressive pathways^227,228. These agents have demonstrated promising activity in combination with ICIs, enhancing CD8⁺ T cell infiltration and activation while reducing the immunosuppressive influence of adenosine metabolism and NK cell dysfunction^227,228.

Adoptive cell therapies.

Adoptive cell therapies (ACTs), including chimeric antigen receptor T cells and TILs, constitute another promising approach in TME modulation²²⁹, with several early phase clinical trials currently underway in NSCLC^230–233 (Fig. 4). Although challenges including TME-induced immunosuppression, poor cell trafficking and limited persistence have generally hindered the efficacy of ACT in patients with solid tumours, innovations in engineering and delivery methods are being developed to overcome these barriers in NSCLC²²⁹.

Nucleic acid-based therapeutic approaches.

Personalized cancer vaccines targeting patient-specific tumour neoantigens have emerged as a promising strategy for the treatment of patients with solid tumours^234–236 (Fig. 4). In those with NSCLC, personalized mRNA vaccines combined with pembrolizumab have demonstrated the ability to broaden T cell responses and reshape the TME^237,238. Oncolytic viruses, such as the oncolytic herpesvirus talimogene laherparepvec, in combination with ICIs enhances antigen release, T cell priming and infiltration within the TME, improving therapeutic efficacy^239–242.

Cytokine-based therapies.

Clinical trials are testing combinations of cytokines with ICIs to stimulate T cell and NK cell function, thereby improving antitumour immune responses in patients with NSCLC^243–247. Although preclinical studies^248,249 and the CANTOS trial²⁵⁰ suggested that exposure to the anti-IL-1β antibody canakimumab reduces the incidence of lung cancer, the therapeutic benefit of this approach in patients with NSCLC remains uncertain. Indeed, the CANOPY trials^251,252 demonstrated a lack of efficacy from combinations of canakimumab with chemotherapy or ICIs^251,252. These results highlight the need for further studies to identify optimal patient subsets, timing or combination strategies to fully harness the therapeutic potential of IL-1β blockade in NSCLC.

Strategic timing, sequencing and personalization of treatment

The potential of adjuvant and neoadjuvant ICI-based approaches to improve long-term outcomes in patients with NSCLC by enhancing immune-mediated control of residual disease is being increasingly recognized^{150,253–255}. The success of these approaches underscores the importance of proper timing and sequencing of ICIs as well as their synergy with other treatment modalities²⁵⁶. Incorporating patient-specific factors, such as molecular drivers and immune profiles, will further refine these strategies, ensuring that treatment interventions are optimally aligned with the unique biology and clinical presentation of disease for each patient. Personalized approaches to deliver TME-targeted therapy will rely on the identification of biomarkers that predict response to treatment. Established markers, such as PD-L1 expression and TMB, provide valuable insights but have limited predictive power when used in isolation²⁵⁷. Emerging techniques, including spatial transcriptomics and multiplex imaging, are enabling a more-granular characterization of the NSCLC TME. Yet, despite these successes, the application of such advanced techniques in routine clinical practice remains constrained by costs, technical complexity and the lack of standardized protocols. Interpreting multiomics data requires specialized expertise, and results often lack the consistency needed for widespread clinical adoption. Moreover, the development of robust, clinically actionable biomarkers remains a key hurdle, especially for predicting long-term responses and resistance mechanisms.

Conclusions

The spatial heterogeneity of the TME in NSCLC presents both challenges and untapped opportunities for advancing cancer therapy. Herein we have described the characteristics of different spatial niches and propose that a paradigm shift is needed to fully exploit the therapeutic potential of the TME. This shift involves reimagining the NSCLC TME not merely as a backdrop for tumour growth but as a dynamic and malleable ecosystem that can be strategically manipulated to improve patient outcomes.

Spatial transcriptomics and high-resolution imaging have unveiled the importance of spatial context in the NSCLC TME. However, temporal dynamics in the TME are equally relevant. NSCLCs are not static entities; their TMEs evolve over time owing to genetic drift, selective pressure from therapies and interactions with the host immune system^256,258 (Fig. 5). Longitudinal studies capturing the temporal evolution of the NSCLC TME could identify patterns of niche transition, such as the conversion of immune-rich niches into immune-poor ones following resistance to ICIs. Although repeat tissue biopsies remain the gold standard for analysing TME evolution, their invasiveness limits the practicality of this as an approach for serial monitoring (Fig. 5). Liquid biopsies offer a less-invasive alternative, with circulating tumour DNA and circulating tumour cells enabling real-time tracking of resistance mutations and shifts in the immune landscape^259–262. To complement molecular analyses, advanced imaging techniques — including PET–CT with new radiotracers and functional MRI — provide spatiotemporal mapping of stromal remodelling and immune infiltration in vivo^263–267. Proteomic profiling using mass spectrometry enables real-time monitoring of relevant TME biomarkers^268–270, while metabolomic analyses capture TME shifts that might influence the timing of therapy^271–273. Additionally, exosomal cargos, containing TME-derived RNA, PD-L1 and regulatory microRNAs, could be a non-invasive biomarker to monitor TME dynamics^274–277.

Fig. 5 |. Longitudinal mapping of TME dynamics in NSCLC and clinical applications.

Longitudinal collection of clinical specimens can be integrated to map the temporal dynamics of the tumour microenvironment (TME) in patients with non-small-cell lung cancer (NSCLC), enabling personalized therapeutic strategies. Specimens collected at baseline, during and after treatment through tissue or liquid biopsy, biofluid analysis and imaging techniques can provide a multiomic view of the evolving TME. Tissue biopsy samples can be used for genomic, spatial transcriptomic, epigenomic, proteomic and metabolomic analyses. Circulating material, including circulating tumour DNA (ctDNA), circulating tumour cells (CTCs) and exosomal cargos, provide non-invasive snapshots of tumour evolution. Biofluid analyses assess cytokine, metabolite and immune cell profiles, and the microbiome, whereas imaging techniques offer spatial and functional insights. These integrated datasets enable comprehensive temporal mapping of the TME, capturing immune landscape dynamics, stromal and spatial remodelling, metabolic reprogramming and the emergence of resistance mechanisms. In turn, this information could be used in the clinic for personalized treatment planning, dynamic treatment adaptation, biomarker-driven patient stratification, the development of resistance mitigation strategies and improved clinical trial design. Ultimately, this approach facilitates real-time therapeutic response monitoring, enabling precision oncology strategies tailored to individual patients. CAF, cancer-associated fibroblast; DC, dendritic cell; ECM, extracellular matrix; NK, natural killer; RBC, red blood cell; TAM, tumour-associated macrophage.

Beyond temporal dynamics within the lung, NSCLC metastases to sites such as the liver, bone, brain and adrenal glands encounter distinct organ-specific TMEs that reshape interactions with immune cells and stromal components. For example, brain metastases often develop immunosuppressive microenvironments that are characterized by reduced TILs and enrichment of TANs and protumorigenic TAMs, ultimately altering response to ICIs^278–281. These site-specific adaptations highlight the need for organ-tailored therapeutic strategies, extending the personalization challenge beyond the primary lung tumour and highlighting the complexity of NSCLC management.

The heterogeneity of the TME across patients with NSCLC underscores the need for personalized therapeutic strategies. Integrating spatial profiling into clinical practice could enable the identification of dominant niches within the tumour of a patient. Therapies could then be tailored to target these specific niches, and combination and sequential approaches could be better refined to offer superior efficacy. For example, tumours predominantly characterized by immune-desert niches might benefit more from treatments that stimulate innate immunity, such as oncolytic viruses, rather than from ICIs alone²³⁹.

The complexity of the TME in NSCLC necessitates the integration of multiomics data (genomic, transcriptomic, proteomic and metabolomic) with spatial and temporal information. Computational modelling and machine learning algorithms can identify key regulatory networks and predict how the TME might respond to various interventions. This systems biology approach could help to prioritize therapeutic targets and combinations, reducing reliance on trial-anderror methods (Fig. 5). Artificial intelligence and machine learning algorithms are increasingly being leveraged to integrate spatial, temporal and multiomic data, improving biomarker discovery, predictive modelling and clinical trial design^282,283. These computational tools hold the potential to optimize treatment sequencing and refine patient stratification, contributing to advances in precision oncology.

The novel strategies presented are promising, although their translational application faces several challenges. The heterogeneity of the TME between and within patients with NSCLC complicates the development of universally effective treatments²⁵⁸. In addition to peritumoural and intratumoural heterogeneity, patient-associated sociobiological factors (including sex, ageing and access to health care) and differential exposure to microenvironmental factors (such as diet or the microbiome) add another layer of complexity to the NSCLC TME^284,285. In this context, the identification of biomarkers for patient stratification becomes a priority. Additionally, manipulating the NSCLC TME carries the risk of unintended consequences, such as exacerbation of immune-related toxicities or pre-existing immune disorders, or triggering resistance mechanisms. Rigorous preclinical models that accurately recapitulate the human TME in NSCLC are needed to evaluate the safety and efficacy of the proposed interventions.

To capitalize on these insights, multidisciplinary collaboration is essential. Oncologists, surgeons, pulmonologists, pathologists and scientists including but not limited to immunologists, computational biologists and systems biologists must work together to develop integrated models of the TME. Investment in longitudinal studies and advanced imaging technologies will improve our understanding of TME dynamics. Furthermore, next-generation clinical trials could be designed to test not only the efficacy of therapeutic agents but also their effect on the spatial and temporal characteristics of the TME. Embracing this holistic view of the TME will be instrumental to advance the management of patients with NSCLC treatment and improving their outcomes in the era of precision oncology.

Key points.

• In patients with non-small-cell lung cancer (NSCLC), the tumour microenvironment (TME) comprises diverse immune, stromal and endothelial cells whose spatial and functional interactions drive tumour progression, metastatic dissemination and response to treatment.

• Spatial heterogeneity of the NSCLC TME, including immune-rich and immune-poor niches, shapes therapeutic outcomes by creating distinct local conditions that either foster effective antitumour immunity or facilitate immune evasion.

• Complex crosstalk among cancer cells, fibroblasts, macrophages and other immune components can either promote or suppress antitumour immune responses, influencing resistance to targeted therapies and immunotherapies.

• Patient-specific determinants, such as ageing, sex, lifestyle factors, comorbidities and health disparities, further modulate the TME, adding layers of complexity to individual responses to treatment.

• Emerging technologies, such as spatial transcriptomics and multiplex imaging, enable unprecedented insights into TME organization and evolution, guiding strategies to overcome therapeutic resistance.

• Tailored approaches that integrate TME analysis, optimal treatment sequencing and combination regimens can potentially help to overcome acquired resistance and yield durable disease control in patients with NSCLC.