Spatially mapping the tumour immune microenvironments of non-small cell lung cancer

Abstract

Lung cancer is the leading cause of cancer-related deaths. An enhanced understanding of the immune microenvironments within these tumours may foster more precise and efficient treatment, particularly for immune-targeted therapies. The spatial architectural differences between lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC) are relatively unexplored. Here, we applied imaging mass cytometry to a balanced cohort of LUAD and LUSC patients, matched for clinical factors such as age, sex, and smoking history, to analyze 204 histopathology images of tumours from 102 individuals with non-small cell lung cancer (NSCLC). By analyzing interactions and broader cellular networks, we interrogate the tumour microenvironment to understand how immune cells are spatially organized in clinically matched adenocarcinoma and squamous cell carcinoma subsets. This spatial analysis revealed distinct patterns of immune cell aggregation, particularly among macrophage populations, that correlated with patient prognosis differentially in adenocarcinoma and squamous cell carcinoma, suggesting potential new strategies for therapeutic intervention. Our findings underscore the importance of analyzing NSCLC histological subtypes separately when investigating the spatial immune landscape, as microenvironmental characteristics and cellular interactions differed by subtype. Recognizing these distinctions is essential for designing precision therapies tailored to each subtype's unique immune architecture, ultimately enhancing patient outcomes.

Fig. 1. Single-cell spatial resolution of NSCLC using imaging mass cytometry.

A Schematic depicting the acquisition and analysis of IMC multiplexed images from 102 LUAD and 102 LUSC cores. Created in BioRender. Quail, D. (2025) https://BioRender.com/k14v431. B Distribution of cell types in the tumour microenvironment and corresponding clinical variables (predominant histological pattern, stage, sex, age, body mass index (BMI), smoking status, pack years, and progression status) in LUAD (n = 102) and LUSC (n = 102). Defined cell types are in order of prevalence for each histological subtype, sorted by CD4+ T cell frequency for LUAD and neutrophil frequency for LUSC. C Representative images of multiplex antibody staining and paired segmented images for LUAD and LUSC. Scale bars = 100 μm. Legend for cell types can be seen in B. D Prevalence of total immune, myeloid, and lymphoid cells as % total cells across LUAD (n = 102) and LUSC (n = 102) cores. Two-tailed unpaired multiple comparison adjusted Student’s t tests; mean ± s.e.m. E Lymphoid infiltrate as % of total cells in male (56 cores; 28 patients) and female (46 cores; 23 patients) LUAD patients. Two-tailed unpaired Student’s t test; mean ± s.e.m. F Lymphoid infiltrate as % of total cells in younger (under 75 years, 68 cores; 34 pts) and older (over 75 years, 34 cores; 17 pts) LUAD patients. Two-tailed unpaired Student’s t test; mean ± s.e.m. G Comparison of cell frequencies between LUAD (n = 102) and LUSC (n = 102) cores for stromal cells, neutrophils and macrophages, other myeloid and lymphoid cells. Two-tailed unpaired multiple comparison adjusted Student’s t tests; mean ± s.e.m. H Proportion of each immune cell as % of total immune cells in LUAD (n = 102) cores from most to least prevalent; mean ± s.e.m. I Proportion of each immune cell as % total immune cells in LUSC (n = 102) cores from most to least prevalent; mean ± s.e.m. J Proportion of Ki67+ cancer cells across LUAD (n = 102) and LUSC (n = 102) cores. Two-tailed unpaired Student’s t test; mean ± s.e.m. P value = 0.00000271. K Proportion of pERK+ CD4+ and CD8+ T cells across LUAD (n = 102 and n = 102) and LUSC (n = 99 and n = 101) cores. Two-tailed unpaired Student’s t test; mean ± s.e.m. pERK+ CD4 + : P value = 0.0000248. pERK+ CD8 + : P value = 0.0000177. L Proportion of CD40 + B cells across LUAD (n = 101) and LUSC (n = 93) cores. Two-tailed unpaired Student’s t test; mean ± s.e.m. M Proportion of Arg1+ macrophages across LUAD (n = 102) and LUSC (n = 102) cores. Two-tailed unpaired Student’s t test; mean ± s.e.m. P value = 0.00000733. N Proportion of alpha-cleaved H3+ neutrophils across LUAD (n = 102) and LUSC (n = 102) cores. Two-tailed unpaired Student’s t test; mean ± s.e.m. P value = 0.000000000318. LUAD lung adenocarcinoma, LUSC lung squamous cell carcinoma, Cl Monocyte classical monocyte, Non-Cl Monocyte non-classical monocyte, Int Monocyte intermediate monocyte, NK cell Natural Killer cell, Treg regulatory T cell, CyTOF cytometry by time of flight, Arg1 Arginase 1.

Fig. 2. Analysis of tumour microenvironments in LUSC and LUAD highlights dynamic pairwise cell–cell interaction patterns and cellular neighbourhoods.

A Heat map of pairwise cell–cell interactions (red) or avoidances (blue) across LUAD (upper half of square, n = 102 images) and LUSC cores (lower half of square, n = 102 images) using 10,000 permutations. NK cells and dendritic cells were omitted from the graph given the lack of significant interactions. Black dotted boxes indicate interactions that are referenced in-text. B Plots depicting significant differences in interaction tendencies for neutrophils (left) and CD163- macrophages (right) between LUAD and LUSC. The size of the outer circles corresponds to the P-value comparing LUAD and LUSC interaction scores. Two-tailed unpaired Student’s t test. C Representative segmented images depicting neutrophil-CD4+ T cell and neutrophil-cancer interaction tendencies in LUAD (left) and LUSC (right), respectively. Scale bars = 100 μm. D Heatmap of cell types distributed across 10 neighbourhoods discovered in LUAD (n = 102) and LUSC (n = 102) (10 nearest neighbours, 10 neighbourhoods). E Average distribution of each CN in LUAD (n = 102) and LUSC (n = 102) (10 nearest neighbours,10 neighbourhoods). F Cellular composition of CN1, CN7, and CN8 discovered in LUAD and LUSC (10 nearest neighbours,10 neighbourhoods). G Number of cells belonging to each CN across LUAD (n = 102) and LUSC (n = 102) (10 nearest neighbours, 10 neighbourhoods). Two-tailed unpaired multiple comparison adjusted Student’s t tests; mean ± s.e.m. P value CN1 = 0.0000681, CN3 = 0.000000926, CN7 = 0.000000682, CN8 = 0.0000441. H Kaplan–Meier analysis of overall survival in CN3 high (z-score ≥0, n = 20) and low (z-score <0, n = 31) LUAD patients (10 nearest neighbours, 10 neighbourhoods). Log-rank Mantel–Cox test. I Cellular composition of CN3 discovered in LUAD and LUSC (10 nearest neighbours,10 neighbourhoods). LUAD lung adenocarcinoma, LUSC lung squamous cell carcinoma, Cl Monocyte classical monocyte, Non-Cl Monocyte non-classical monocyte, Int Monocyte intermediate monocyte, NK cell Natural Killer cell, Treg regulatory T cell.

Fig. 3. A unique role for macrophages in NSCLC.

A Heatmap of cell types distributed across 30 neighbourhoods discovered in LUAD (n = 102) and LUSC (n = 102) (10 nearest neighbours, 30 neighbourhoods). B Distribution of cell types in CN15, CN23, and CN26 discovered in LUAD and LUSC (10 nearest neighbours, 30 neighbourhoods). Legend for other cell types can be seen in J. C Kaplan–Meier analysis of overall survival in CN15 high (z-score ≥0, n = 19) and CN15 low (z-score <0, n = 32) LUAD patients. Log-rank Mantel–Cox test. D Kaplan–Meier analysis of overall survival in CN23 high (z-score ≥0, n = 22) and CN23 low (z-score <0, n = 29) LUAD patients. Log-rank Mantel–Cox test. E Kaplan–Meier analysis of overall survival in CN26 high (z-score ≥0, n = 11) and CN26 low (z-score <0, n = 40) LUAD patients. Log-rank Mantel–Cox test. F Heatmap of cell types distributed across 30 neighbourhoods discovered in LUAD (n = 102) and LUSC (n = 102) (3 nearest neighbours, 30 neighbourhoods). G Distribution of CD163+ macrophages across cellular neighbourhoods discovered in LUAD and LUSC (3 nearest neighbours, 30 neighbourhoods). Neighbourhoods with no CD163+ macrophages excluded from graph. H Kaplan–Meier analysis of overall survival in CN26 high (z-score ≥0, n = 16) and CN26 low (z-score <0, n = 35) LUSC patients. Log-rank Mantel–Cox test. I Kaplan–Meier analysis of disease-free survival in CN26 high (z-score ≥0, n = 9) and CN26 low (z-score <0, n = 42) LUAD patients. Log-rank Mantel–Cox test. J Cellular composition of CN26 and CN19 discovered in LUAD and LUSC (3 nearest neighbours, 30 neighbourhoods). LUAD lung adenocarcinoma, LUSC lung squamous cell carcinoma, Cl Monocyte classical monocyte, Non-Cl Monocyte non-classical monocyte, Int Monocyte intermediate monocyte, NK cell Natural Killer cell, Treg regulatory T cell.

Fig. 4. Impact of cellular architecture on survival outcomes in LUAD and LUSC.

A PCA based on the frequency of cells in LUAD (n = 102) and LUSC (n = 102). B PCA based on raw image masks in LUAD (n = 102) and LUSC (n = 102). C PCA based on a multi-layer perceptron (MLP) used to capture unique features from raw images in LUAD (n = 102) and LUSC (n = 102). D Heatmap of cell types distributed across 10 neighbourhoods discovered in LUAD (n = 102) (10 nearest neighbours,10 neighbourhoods). E Heatmap of cell types distributed across 10 neighbourhoods discovered in LUSC (n = 102) (10 nearest neighbours, 10 neighbourhoods). F Proportion of cell types in cellular neighbourhoods identified as vascular niche in LUAD and LUSC. G Kaplan–Meier analysis of overall and disease-free survival in LUAD patients with high (z-score ≥0, n = 15) and low (z-score <0, n = 36) proportion of Ki67+ endothelial cells of total endothelial cells. Log-rank Mantel–Cox test. H Kaplan–Meier analysis of overall and disease-free survival in LUSC patients with high (z-score ≥0, n = 19) and low (z-score <0, n = 32) proportion of Ki67+ endothelial cells of total endothelial cells. Log-rank Mantel–Cox test. I Frequency of Ki67+ endothelial cells across LUAD (n = 102) and LUSC (n = 102) cores. Two-tailed unpaired Student’s t test; mean ± s.e.m. J Proportion of Ki67+ endothelial cells interacting with neutrophils in LUAD (n = 99) and LUSC (n = 86). Excludes cores without endothelial cell and neutrophil interactions. Two-tailed unpaired Student’s t test; mean ± s.e.m. LUAD lung adenocarcinoma, LUSC lung squamous cell carcinoma, Cl Monocyte classical monocyte, Non-Cl Monocyte non-classical monocyte, Int Monocyte intermediate monocyte, NK cell Natural Killer cell, Treg regulatory T cell.