Publication in Signal Transduction and Targeted Therapy

How immune-excluded lung tumors differ from infiltrated ones — and why it matters for immunotherapy

Explicyte collaborated with: GSK·Institut Bergonié·Institut Gustave Roussy·Paris Saclay University
First page of "Spatial transcriptomics identifies tumor-intrinsic and stromal programs associated with immune exclusion in NSCLC", Signal Transduction and Targeted Therapy, 2026
ThoracicBiomarker analysisBiomarker discoveryDiscoveryTrialsFFPE tissueBioinformaticsGeoMx DSPMultiplex IF/IHCRNAseq
JournalSignal Transduction and Targeted Therapy
DateOct 2026
Read full paper →

Checkpoint inhibitors help only a minority of patients with advanced lung cancer, and immune exclusion in NSCLC — CD8+ T cells stuck in the stroma, outside tumor nests — is one recurring reason. Across 261 ICB-treated patients, Explicyte and its partners combined GeoMx spatial transcriptomics, HTG EdgeSeq and multiplex immunofluorescence. Excluded tumors showed lower HLA class I and TAP1, peri-tumoral myofibroblastic CAFs, lower response rates and shorter progression-free survival.

A new study in Signal Transduction and Targeted Therapy, led by Prof. Antoine Italiano (Gustave Roussy, Université Paris-Saclay) with Explicyte CTO Jean-Philippe Guégan as first author, dissects the immune-excluded phenotype in advanced non-small cell lung cancer. The work draws on pretreatment core-needle biopsies from patients enrolled in the prospective BIP precision medicine study (NCT02534649). It was funded by GSK, whose scientists co-authored the paper. Explicyte co-designed the study and generated its core molecular and spatial data. This covered GeoMx DSP whole-transcriptome profiling with the TCR module across tumor, stroma and CD8-enriched compartments, plus five multiplex immunofluorescence panels covering lymphoid and myeloid subsets, CAF subtypes, T-cell exhaustion and HLA-ABC. It also covered HTG EdgeSeq whole-transcriptome profiling of 130 tumors. The team carried out the statistical, bioinformatic and single-cell spatial analyses that tie these layers to clinical outcomes.

The question

What distinguishes NSCLC tumors that keep CD8+ T cells stranded in the stroma from those they infiltrate, and how does that pattern relate to checkpoint-inhibitor outcomes?

Key steps

  1. 1

    Classify immune contexture in two cohorts

    CD8/pan-cytokeratin staining on biopsies from 261 ICB-treated patients sorted tumors into infiltrated, excluded or desert phenotypes, using a 2% CD8+ cell threshold. Pathologist and digital image-analysis scoring agreed closely. In the 144-patient discovery cohort, objective response reached 58% in infiltrated tumors vs 38% in excluded and 20% in desert ones. Median PFS was 12.8 vs 4.1 vs 2.8 months, and the association held after adjustment for ECOG status, PD-L1 and tertiary lymphoid structures. The pattern was consistent in the 117-patient internal validation cohort.

  2. 2

    Resolve tumor, stroma and T-cell compartments

    GeoMx DSP whole-transcriptome profiling covered 8 tumors (4 infiltrated, 4 excluded), with 9–12 regions each segmented into tumor, stroma and CD8-enriched compartments. In the tumor compartment, excluded cases expressed less HLA-A, HLA-B, HLA-F and TAP1. They also lost the link between IFN-γ signaling and antigen-presentation genes (r = 0.49 for HLA in infiltrated tumors vs r ≈ 0 in excluded), running 21.8% below the HLA level their IFN-γ signature predicted. HTG EdgeSeq profiling of 130 patients confirmed higher antigen-presentation activity in infiltrated tumors and TGF-β signaling in excluded ones.

  3. 3

    Measure HLA loss and T-cell proximity in situ

    Multiplex immunofluorescence on 152 tumors showed more CD8+ T cells, M1-like macrophages and HLA-ABC+ tumor cells in infiltrated cases. In excluded tumors, HLA-ABC– tumor cells sat farther from the nearest CD8+ T cell (median 54.4 vs 43.4 µm). Low tumor HLA-ABC in excluded tumors was associated with shorter PFS (HR 1.61, p = 0.049) after ECOG adjustment.

  4. 4

    Profile fibroblast subsets and T-cell states

    A six-marker CAF panel on 80 tumors found myofibroblastic CAFs (COL1A1+/αSMA+/TGF-β+) enriched in excluded tumors and positioned closer to tumor epithelium (p < 0.001), while total fibroblast density did not differ. In CD8-enriched regions, infiltrated tumors showed oxidative-phosphorylation and translation programs and higher TCR-probe diversity. Canonical exhaustion markers (PD-1, LAG-3, TIM-3, TIGIT) were similar across phenotypes.

Impact

Immune exclusion in NSCLC looks less like checkpoint-driven exhaustion and more like a combined problem of poor tumor visibility and a stromal barrier, and both features can be measured on routine biopsies.

261
ICB-treated NSCLC patients across discovery and validation cohorts
58%
objective response to ICB in infiltrated tumors vs 38% in excluded ones (discovery cohort)
12.8 mo
median PFS in infiltrated tumors vs 4.1 months in excluded tumors
21.8%
lower HLA class I expression in excluded tumors than their IFN-γ signature predicts

For checkpoint-inhibitor developers, spatial CD8+ T-cell localization emerges as a candidate stratification biomarker. It adds information beyond PD-L1 and TLS status and can be read on standard core-needle biopsies. The data point to combination strategies at two levels: restoring tumor antigen presentation, and reshaping myCAF-rich stroma, for example through TGF-β pathway modulation. They also argue for spatially resolved biomarker endpoints in trials enrolling immune-excluded patients.

Building a spatial biomarker strategy for checkpoint or stroma-targeting combinations in NSCLC? Let's talk.

Talk to us