Understanding immune checkpoint inhibitor efficacy through spatial decoding of the lung cancer tumor immune microenvironment

Immune checkpoint inhibitors (ICIs) have improved patient outcomes substantially in non–small cell lung cancer (NSCLC). Despite considerable effort, our understanding of the features that predict for immunotherapy response and resistance in patients remains incomplete. In this issue of the JCI, Isomoto and colleagues utilized a multiplex IHC platform to profile the spatial organization of the lung cancer tumor immune microenvironment, enabling the identification of spatial immune features that correlate with immunotherapy efficacy. This study enhances our knowledge of the spatial organization of features impacting ICI efficacy by identifying a three-variable spatial composite — including CD73 upregulation in EGFR-mutant NSCLC — that substantially outperforms PD-L1 expression in predicting immunotherapy efficacy. Moreover, it establishes spatial proteomic profiling as a platform for generating therapeutic hypotheses that are actionable and mechanistic in NSCLC.

Lung cancer remains the leading cause of cancer death worldwide (1). The most common form of lung cancer is non–small cell lung cancer (NSCLC), which represents approximately 80% of all new cases of this disease (1). While immune checkpoint inhibitors (ICIs) form the backbone of treatment for most cases of advanced NSCLC, the features of NSCLC tumors that govern the efficacy of these cancer immunotherapies remain incompletely defined (2).

Early studies identified PD-L1 expression and tumor mutation burden as nonoverlapping predictors of therapeutic response to PD-1/PD-L1–based ICIs in NSCLC (2). On the other hand, genomic alterations in STK11 and KEAP1 confer ICI resistance, particularly in the context of a concurrent KRAS mutation (3). Beyond immunohistochemical and genomic characterization, subsequent work has suggested that spatial organization of immune cells within the NSCLC tumor microenvironment (TME) — including the presence of stem-immunity hubs (4) and TCF1⁺ stem-like CD8⁺ T cells within tertiary lymphoid structures (TLSs) (5) — may also be associated with immunotherapy response.

In this issue of the JCI, Isomoto and colleagues used a bespoke multiplexed IHC platform combined with computational tissue segmentation to perform spatial profiling of the tumor immune microenvironment in 103 patients with metastatic NSCLC, 81 of whom received ICI therapy (6). Based on analysis of pretreatment samples, the study identified several spatial immune features correlated with immunotherapy efficacy. The proximity of CD8⁺ tumor-infiltrating lymphocytes (TILs) to tumor nests and markers of tissue residence and proliferation in TILs were associated with ICI efficacy, while CD206⁺ M2-like tumor-associated macrophages (TAMs) and fibroblast activation protein⁺ (FAP⁺) cancer-associated fibroblasts (CAFs) correlated with poorer outcomes after ICI treatment (Figure 1).

Figure 1

Spatial proteomics enables scientific discovery in the treatment of NSCLC. Isomoto et al. (6) identified spatial features of the NSCLC tumor immune microenvironment that are correlated with ICI efficacy (top) and ICI resistance (bottom), laying the groundwork for spatial proteomics to be utilized as a guide for scientific discovery and personalized cancer therapy.

Critically, CD8⁺ TIL density in the intratumoral stromal area — the compartment typically captured by standard immunohistochemical TIL quantification and artificial intelligence–based H&E scoring — did not independently predict ICI efficacy in multivariable analysis. This finding provides a mechanistic, spatial explanation for the inconsistent clinical utility of bulk TIL assessment.

Spatial validation for Isomoto et al.’s finding is provided by a recent multiplexed immunofluorescence and deep learning study of NSCLC biopsies by Monkman et al. (7), which performed with high accuracy for predicting ICI progression-free survival >24 months. In that study, Monkman et al. used cell–cell spatial interaction features to confirm that tumor nest proximity relationships, not aggregate TIL density, constituted the critical predictive unit (7).

The finding that CD8⁺ TILs localized to the tumor nest predict ICI response more accurately than those in the aggregate stromal compartment raises a fundamental question: are these spatially privileged T cells recognizing tumor antigens? T cells engaged with cognate antigen should reinvigorate more robustly after ICIs than bystanders that colocalize without antigen recognition. Identifying which tumor antigens drive TIL reactivity has proved challenging because solid tumors harbor abundant bystander T cells that are reactive to common viral antigens rather than tumor-derived antigens (8, 9). Beyond classical neoantigens, cancer cells, including NSCLC cells, express noncanonical antigens: peptides derived from noncoding RNAs, UTRs, and alternative reading frames that are presented on MHC molecules and can be targeted by HLA-restricted T cells (10–12). Emerging spatial single-cell T cell receptor (TCR) sequencing technologies (13, 14) combined with evolving TCR antigen discovery platforms (15) may be used to determine whether nest-localized CD8⁺ TILs are enriched for tumor antigen reactivity and whether stromal TILs are bystanders reactive to common viral antigen.

Isomoto et al. found that TILs with markers of tissue residency within tumor nests were also associated with ICI efficacy (6), a finding supported by prior datasets in solid tumors (2, 16). While tumor-reactive TILs and bystander T cells may both express tissue residency markers, CD39 and CD103 remain imprecise surrogates (17, 18). Critically, exhausted T cells may proliferate and mediate antitumor immunity in response to ICIs, whereas tissue-resident memory (Trm) T cells may represent bystanders lacking tumor antigen reactivity; distinct ontogenies and divergent ICI responses are now well documented for these T cell subtypes (17, 18). We therefore hypothesize that CD39⁺CD103⁺ expression marks tumor-reactive tissue-resident exhausted T cells within NSCLC tumor nests that retain capacity for reinvigoration by ICIs.

Notably, expression of coinhibitory receptors, including LAG-3, TIM-3, and TIGIT, showed positive correlation with ICI response in this cohort, consistent with their role as markers of antigen-driven T cell activation in the pretreatment setting rather than terminal effector paralysis. This finding has direct implications for the interpretation of ongoing trials combining ICIs with anti–LAG-3, anti–TIM-3, or anti-TIGIT agents.

High Ki-67 positivity within Trm-like TILs likely distinguishes the TCF1⁺ progenitor-exhausted subset from TCF1^– terminally exhausted TILs: Ki-67⁺ and TCF1⁺ progenitor-exhausted TILs retain proliferative self-renewal capacity and can be selectively reinvigorated by PD-1/PDL-1 blockade (19), whereas the TCF1^– subset remains largely refractory to ICIs. These progenitor-exhausted cells are found within stem-immunity hubs and TLSs in NSCLC (4, 5), suggesting that Ki-67⁺ Trm-like TILs in tumor nests may represent effector progeny of TLS-resident precursors that have migrated upon antigen encounter.

CAFs have been reported to have diverse functions in promoting NSCLC, including mediating immunosuppression and therapy resistance (20, 21). High numbers of FAP⁺ CAFs in intratumoral stromal areas were associated with inferior ICI outcomes in the Isomoto cohort. Imaging mass cytometry and pan-cancer scRNA-seq studies have identified multiple CAF subtypes with heterogeneous survival and immunosuppressive associations (20, 21). As spatial technologies enable further CAF subtype refinement, determining which populations most potently suppress ICI efficacy in NSCLC will be essential for developing CAF-targeted therapies.

Integrating all three spatial features — high density of Ki-67⁺CD39⁺CD103⁺ Trm-like TILs in tumor nests, low CD206⁺ M2-TAM burden, and absence of FAP⁺ CAFs — identified a favorable TME (fTME) subgroup comprising 11 of 81 ICI-treated patients, whose median progression-free survival was not reached versus 5.6 months in the remainder of the cohort (HR = 0.13) (6). This effect size markedly exceeded the PD-L1–based tumor proportion score (TPS; reflecting the percentage of tumor cells that express the protein PD-L1), a standardized clinical measurement for predicting immunotherapy response. Direct comparison within this cohort also demonstrated that CD206 outperformed CD163 as a predictor of poor ICI outcome, providing marker selection guidance for future studies incorporating immunosuppressive macrophage phenotyping into NSCLC biomarker panels. The fTME composite is independently corroborated by pan-cancer analyses: a five–latent factor decomposition of ICI response across solid tumor types identified effective T cell infiltration and TGF-β microenvironment activity as universal axes of ICI resistance (22), and the LORIS composite scoring framework demonstrated stable ICI outcome prediction across 2,881 patients from 18 tumor types (23).

ICI therapy is currently not recommended for patients with EGFR-mutant lung adenocarcinoma. Isomoto et al.’s finding that NT5E-encoded CD73 (ecto-5′-nucleotidase) protein expression was highly enriched in EGFR-mutant NSCLC — versus EGFR WT — thus identifies both a mechanism for ICI resistance in EGFR-mutant lung adenocarcinoma and a therapeutic target of importance. The proposed mechanistic circuit is as follows: oncogenic EGFR activation drives autophagy and extracellular ATP release, which CD73 hydrolyzes to immunosuppressive adenosine, in turn promoting M2-TAM accumulation, aberrant angiogenesis, and TGF-β upregulation that collectively suppress TIL function. This circuit nominates CD73 blockade, which is currently under clinical investigation in multiple solid tumor types (24), as a potentially synergistic combination partner with ICI in EGFR-mutant NSCLC, complementary to the VEGF inhibition strategy (25).

Transcriptomic profiling in Isomoto et al. (6) revealed enrichment of angiogenesis gene expression signatures in EGFR-mutant NSCLC, consistent with recognition of CD206⁺ M2-TAMs and FAP⁺ CAFs as VEGF sources in the TME. The HARMONi-A trial demonstrated that ivonescimab — a bispecific antibody cotargeting PD-1 and VEGF — significantly improved progression-free survival versus chemotherapy alone in previously treated EGFR-mutant NSCLC (26), validating VEGF inhibition as a strategy for rendering the EGFR-mutant TME more responsive to ICI. These findings collectively suggest that modulating the functional outputs of immunosuppressive TME populations, rather than depleting them directly, may represent a tractable approach in oncogene-driven NSCLC.

In summary, Isomoto and colleagues present a landmark spatial proteomic characterization of the NSCLC tumor immune microenvironment. We view this platform as a research-grade proof of concept: its discriminatory power is compelling, but prospective multi-institutional validation, reagent and computational pipeline standardization, and reduction to a deployable clinical surrogate are required before routine implementation. The dual mechanistic circuit in the EGFR-mutant TME — linking EGFR to CD73 and to excess adenosine, and also linking EGFR to VEGF — opens complementary immunotherapeutic strategies, with CD73 blockade as a rational ICI partner that is potentially synergistic with the VEGF inhibition approach validated by HARMONi-A.

Looking forward, we envision this spatial proteomic platform as the foundation of a precision medicine workflow for individual patients with NSCLC. Integration with oncogenotype, RNA-seq, scRNA-seq, and epigenomic data will reveal whether specific tumor molecular states drive favorable or unfavorable TME configurations and whether oncogene-targeted therapies can beneficially shift them (Figure 1). When spatial profiling reveals an unfavorable TME, the identified suppressive mechanisms, exemplified here by CD73-driven adenosine and VEGF-mediated immunosuppression, serve as mechanistic road maps for individualized combination immunotherapy design. Extension to the tumor cell surfaceome will enable rational antibody-drug conjugate and bispecific T cell engager deployment. Moreover, as an extension of the TCR and antigen discovery framework discussed above, patient-specific neoantigen and T cell clone identification will support peripheral blood monitoring, neoantigen vaccination, and CAR T development for patients with NSCLC.

JDM receives licensing royalties from the NIH and the University of Texas Southwestern for distribution of human tumor cell lines. TZ’s spouse has equity interest in Novartis.

This work is the result of NIH funding, in whole or in part, and is subject to the NIH Public Access Policy. Through acceptance of this federal funding, the NIH has been given a right to make the work publicly available in PubMed Central.

• National Cancer Institute K08 CA262169 (to TZ).
• Cancer Prevention and Research Institute of Texas grant RR250063 (to TZ).
• Lung Cancer SPORE P50 CA070907 (to JDM).
• Simmons Comprehensive Cancer Center grant P30 CA142543.

We acknowledge Huocong Huang for helpful discussions during the preparation of this work.

Address correspondence to: Tao Zou, Hamon Center for Therapeutic Oncology Research, University of Texas Southwestern Medical Center, 6000 Harry Hines Blvd., Dallas, Texas 75390-8593, USA. Email: tao.zou@utsouthwestern.edu.

Copyright: © 2026, Zou et al. This is an open access article published under the terms of the Creative Commons Attribution 4.0 International License.

Reference information: J Clin Invest. 2026;136(10):e206316. https://doi.org/10.1172/JCI206316.

See the related article at .

1. World Health Organization. Lung Cancer Fact Sheet. https://www.who.int/news-room/fact-sheets/detail/lung-cancer Updated June 26, 2023.
2. Wang SL, Chan TA. Navigating established and emerging biomarkers for immune checkpoint inhibitor therapy. Cancer Cell. 2025;43(4):641–664.
   View this article via: CrossRef PubMed Google Scholar
3. Skoulidis F, et al. STK11/LKB1 mutations and PD-1 inhibitor resistance in KRAS-mutant lung adenocarcinoma. Cancer Discov. 2018;8(7):822–835.
   View this article via: CrossRef PubMed Google Scholar
4. Chen JH, et al. Human lung cancer harbors spatially organized stem-immunity hubs associated with response to immunotherapy. Nat Immunol. 2024;25(4):644–658.
   View this article via: CrossRef PubMed Google Scholar
5. Im SJ, et al. Characteristics and anatomic location of PD-1⁺TCF1⁺ stem-like CD8 T cells in chronic viral infection and cancer. Proc Natl Acad Sci U S A. 2023;120(41):e2221985120.
   View this article via: CrossRef PubMed Google Scholar
6. Isomoto K, et al. Spatial single-cell proteotyping reveals immunotherapy-resistant features within the complex tumor microenvironment of metastatic NSCLC. J Clin Invest. 2026;136(10):e195021.
   View this article via: JCI PubMed CrossRef Google Scholar
7. Monkman J, et al. Metabolic characterization of tumor-immune interactions by multiplexed immunofluorescence reveals spatial mechanisms of immunotherapy response in non-small cell lung carcinoma (NSCLC). Nat Commun. 2026;17(1):837.
   View this article via: CrossRef PubMed Google Scholar
8. Rosato PC, et al. Virus-specific memory T cells populate tumors and can be repurposed for tumor immunotherapy. Nat Commun. 2019;10(1):567.
   View this article via: CrossRef PubMed Google Scholar
9. Simoni Y, et al. Bystander CD8⁺ T cells are abundant and phenotypically distinct in human tumour infiltrates. Nature. 2018;557(7706):575–579.
   View this article via: CrossRef PubMed Google Scholar
10. Laumont CM, et al. Noncoding regions are the main source of targetable tumor-specific antigens. Sci Transl Med. 2018;10(470):eaau5516.
    View this article via: CrossRef PubMed Google Scholar
11. Ouspenskaia T, et al. Unannotated proteins expand the MHC-I-restricted immunopeptidome in cancer. Nat Biotechnol. 2022;40(2):209–217.
    View this article via: CrossRef PubMed Google Scholar
12. Ely ZA, et al. Pancreatic cancer-restricted cryptic antigens are targets for T cell recognition. Science. 2025;388(6747):eadk3487.
    View this article via: CrossRef PubMed Google Scholar
13. Liu S, et al. Spatial maps of T cell receptors and transcriptomes reveal distinct immune niches and interactions in the adaptive immune response. Immunity. 2022;55(10):1940–1952.
    View this article via: CrossRef PubMed Google Scholar
14. Engblom C, et al. Spatial transcriptomics of B cell and T cell receptors reveals lymphocyte clonal dynamics. Science. 2023;382(6675):eadf8486.
    View this article via: CrossRef PubMed Google Scholar
15. Kasbe M, et al. De novo identification of the specificities of recurrent human T cell receptors [preprint]. https://doi.org/10.1101/2025.08.03.668342 Posted on bioRxiv August 4, 2025.
16. Caushi JX, et al. Transcriptional programs of neoantigen-specific TIL in anti-PD-1-treated lung cancers. Nature. 2021;596(7870):126–132.
    View this article via: CrossRef PubMed Google Scholar
17. Park SL, et al. Tissue-resident exhausted and memory CD8⁺ T cells have distinct ontogeny, function and role in disease. Nat Immunol. 2026;27(1):110–125.
    View this article via: CrossRef PubMed Google Scholar
18. Burn TN, et al. Antigen reactivity defines tissue-resident memory and exhausted T cells in tumors. Nat Immunol. 2026;27(1):98–109.
    View this article via: CrossRef PubMed Google Scholar
19. McLane LM, et al. CD8 T cell exhaustion during chronic viral infection and cancer. Annu Rev Immunol. 2019;37:457–495.
    View this article via: CrossRef PubMed Google Scholar
20. Cords L, et al. Cancer-associated fibroblast phenotypes are associated with patient outcome in non-small cell lung cancer. Cancer Cell. 2024;42(3):396–412.
    View this article via: CrossRef PubMed Google Scholar
21. Chen X, et al. Single-cell resolution spatial analysis of antigen-presenting cancer-associated fibroblast niches. Cancer Cell. 2025;43(12):2224–2240.
    View this article via: CrossRef PubMed Google Scholar
22. Usset J, et al. Five latent factors underlie response to immunotherapy. Nat Genet. 2024;56(10):2112–2120.
    View this article via: CrossRef PubMed Google Scholar
23. Chang TG, et al. LORIS robustly predicts patient outcomes with immune checkpoint blockade therapy using common clinical, pathologic and genomic features. Nat Cancer. 2024;5(8):1158–1175.
    View this article via: CrossRef PubMed Google Scholar
24. Augustin RC, et al. Next steps for clinical translation of adenosine pathway inhibition in cancer immunotherapy. J Immunother Cancer. 2022;10(2):e004089.
    View this article via: CrossRef PubMed Google Scholar
25. Le X, et al. Characterization of the immune landscape of EGFR-mutant NSCLC Identifies CD73/adenosine pathway as a potential therapeutic target. J Thorac Oncol. 2021;16(4):583–600.
    View this article via: CrossRef PubMed Google Scholar
26. Harmoni-A Study Investigators, et al. Ivonescimab plus chemotherapy in non-small cell lung cancer with EGFR variant: a randomized clinical trial. JAMA. 2024;332(7):561–570.
    View this article via: CrossRef PubMed Google Scholar