A Comprehensive Catalogue of Stromal Cells in Human Lung Tumours at Single-Cell Resolution

Novel study reveals a highly complex tumour microenvironment that profoundly molds stromal cells

On 9 July 2018, Belgian researchers published in the Nature Medicine a comprehensive catalogue of stromal cells in human lung tumours and non-malignant lung tissue at single-cell resolution. In describing key molecular differences between stromal cells co-opted by tumours and those in matching non-malignant samples, their analyses confirm many important observations made previously either in vitro, in bulk or using animal models, and highlight key areas for further advances in stromal cell biology. By identifying novel cell subtypes and altered pathways, by highlighting the cellular sources of stromal signals and by cataloguing marker genes, this dataset has potentials to fuel advances in lung cancer diagnosis and therapy. It can serve to validate a priori hypotheses from independent data, but also highlight novel targets meriting functional validation.

The authors wrote in study background that tumours are characterised by extensive heterogeneity, but so far efforts in understanding this heterogeneity were largely limited to cancer cells. These revealed a remarkably complex and diverse landscape of cancer cells, with evidence for genetic diversification and clonal selection. However, the stromal cells associated with tumours, and the complex cellular system they build to form the tumour microenvironment, may themselves be as complex and heterogeneous as the cancer cell compartment. Particularly, an increasing number of studies suggest that stromal cells, such as macrophages, T cells and fibroblasts, are highly heterogeneous. The extent of this heterogeneity, how it is shaped by other cells in the tumour and vice versa also directly affects them, remains however poorly characterised, in part because of a historical lack of methods to study these cells in isolation.

The tumour microenvironment is increasingly recognised as a cancer therapy target. PD-1 or PD-L1 antibodies activate anti-tumour responses of cytotoxic T cells. In patients with advanced non-small-cell lung cancer (NSCLC), these treatments demonstrated response rates up to 45%, and some responses are remarkably durable. Nintedanib, when added to docetaxel, significantly extends median overall survival in previously-treated patients with NSCLC. Despite the paramount therapeutic importance, the in situ phenotype of stromal cells targeted remains elusive.

Cancer cells are embedded in the tumour microenvironment, a complex system of stromal cells. The study team presents a 52,698-cell catalogue of the tumour microenvironment transcriptome in human lung tumours at single-cell resolution, validated in independent samples where 40,250 additional cells were sequenced. By comparing with matching non-malignant lung samples, they revealed a highly complex tumour microenvironment that profoundly molds stromal cells.

The study team identified 52 stromal cell subtypes, including novel subpopulations in cell types hitherto considered to be homogeneous, as well as transcription factors underlying their heterogeneity. They discovered fibroblasts expressing different collagen sets, endothelial cells downregulating immune cell homing and genes coregulated with established immune checkpoint transcripts and correlating with T-cell activity.

By assessing marker genes for these cell subtypes in bulk RNA-sequencing data from 1,572 patients, the investigators illustrated how these correlate with survival, while immunohistochemistry for selected markers validates them as separate cellular entities in an independent series of lung tumours.

Some key observations emerged from this study. Firstly, the lung tumour microenvironment is more complex and heterogeneous. Secondly, most clusters were composed of cells originating from different patients, while independent analysis of 3 additional patients validated 45 of 52 cell subtypes. However, marker gene expression in TCGA indicated that some cell subtype abundances differed between lung squamous carcinoma or adenocarcinoma subtype, that they were influenced by tumour characteristics such as tumour stage and that they correlated with patient survival in squamous histology but not adenocarcinoma or vice versa. While most stromal subtypes were detected in several tumours, their abundances and functions could differ between tumours. Intriguing questions remain as to whether these stromal cell phenotypes also exist in tumours affecting other organs, and whether they recur in metastases of lung tumours to other organs. A third observation relates to the association of cell subtypes with patient survival: squamous tumours showed a negative correlation between many stromal cell markers and survival, independently of tumour stage. Whether this difference reflects a genuine effect of stromal cells on squamous cancer cell biology remains to be elucidated. The stromal cell number in tumour samples could also reflect invasive properties of cancer cells, with more invasive tumours containing more stromal cells. Nevertheless, observations that some marker genes correlated robustly with survival in various cohorts suggest exciting opportunities for use of these subtype-specific marker genes as biomarkers for prognosis but also for therapy response prediction. Finally, distinctive features of tumour stroma may represent vulnerabilities and provide exciting entry points for the design of novel therapies.

Single-cell RNA-sequencing (scRNA-seq) enables specific profiling of cell populations at the single-cell level. While conventional bulk RNA-sequencing methods process millions of cells, averaging out underlying differences, scRNA-seq can reveal changes that render each individual cell type unique. Previous scRNA-seq studies on glioblastoma, melanoma and oligodendroglioma focused largely on cancer cells, analyzing few stromal cells from tumours, and not from matching non-malignant tissue. By analyzing cells from lung tumours and matching non-malignant tissue at a much higher scale, the study team uncovered stromal cell heterogeneity and adaptation to the tumour. 

By providing a comprehensive catalogue of stromal cell types and by characterising their phenotype and co-optive behaviour, this resource provides deeper insights into lung cancer biology.

The study was supported by multiple grants. The authors declared no competing interests. 


Lambrechts D, Wauters E, Boeckx B, et al. Phenotype molding of stromal cells in the lung tumor microenvironment. Nature Medicine; Published online 9 July 2018. DOI: https://doi.org/10.1038/s41591-018-0096-5