Co-evolution of Subclonal Expansion and Metastatic Microenvironment in Lung Adenocarcinoma Captured by Spatial Lineage Tracing
Background
Cancer progression is both a process of selecting genetically distinct cancer cell populations and an ecological change marked by the accumulation of hypoxia, immune suppression, and fibrosis. However, single-cell RNA sequencing disrupts tissue architecture, erasing the original spatial positions of cells, while multi-region sequencing often samples limited areas, missing the continuous spatial structure of cell states and lineages. It remains poorly understood where subclones grow, how they alter surrounding cells, and where metastasis-competent cells emerge within the primary tumor.
To address these gaps, the researchers combined high-resolution spatial transcriptomics with lineage tracing that accumulates genetic barcodes over time. The model used was KP-Tracer mice, in which lung adenocarcinoma is induced by Kras activation and Trp53 deletion. Cre virus was administered to type 2 alveolar cells in 8–12-week-old mice to simultaneously initiate tumor development and CRISPR–Cas9 lineage recording. The design aimed to reconstruct not only the current transcriptional state of cancer cells but also their common ancestry and proliferation history within tissue coordinates.
Key Findings
The researchers used Slide-seq, which reads large tissue areas at subcellular resolution, and Slide-tags, which provide single-nucleus sensitivity. They analyzed over 100 tumors across 49 spatial transcriptomic arrays and applied a correction algorithm to recover missing lineage barcodes using spatial and genetic information from neighboring cancer cells. The median accuracy of restoring hidden barcodes in real data was 90%, and iterative correction recovered 4–58% (mean 31%) of missing information per dataset. The median accuracy of random predictions was 67%.
As tumors progressed, cancer cells maintaining an alveolar state were mainly located at the periphery, while those with epithelial-mesenchymal transition (EMT) and hypoxia gene programs were found in the interior. Around rapidly expanding high-fitness subclones, Arg1-positive immunosuppressive tumor-associated macrophages and Postn-positive myofibroblast-like cancer-associated fibroblasts were enriched. Fitness calculated from lineage trees showed a Pearson correlation coefficient of 0.4 with existing transcriptome-based fitness metrics, and hypoxia–EMT spatial communities were most strongly associated with high fitness. This supports the sequential interplay of subclone proliferation, oxygen depletion, immune and stromal reprogramming, and metastasis-promoting cell states. Spatial analysis also identified an unreported cancer cell state expressing Piezo2, Robo1, and Pecam1, maintaining Nkx2-1 but not expressing Vim.
In mice with widespread metastasis, the primary tumor was reconstructed in 3D by sectioning the lung every 200–500 micrometers. Metastatic lesions in mediastinal lymph nodes, ribs, and diaphragm were all lineage-linked to spatially restricted subclones within the primary tumor T2. The TGF-β program in metastatic lesions was nearly identical to that of the corresponding primary subclone (log2 fold change −0.14, P=1.0), but collagen-related expression was significantly increased (log2 fold change 3.81, P<10⁻⁵). COL3A1 protein staining further supported fibrosis in metastatic sites. Nature Genetics paper
Implications and Outlook
This study provides spatial evidence that aggressive cancer cell states are not fixed solely by internal mutations but are co-shaped by the hypoxic, immunosuppressive, and fibrotic environment created by subclone expansion. The finding that metastatic seeds are concentrated in specific local ecosystems rather than uniformly distributed throughout the primary tumor is also noteworthy. After metastasis, the existing TGF-β–EMT program is maintained, but collagen deposition is added, creating a new environment suitable for metastatic colonization.
However, the results are largely dependent on the Kras;Trp53-based mouse lung adenocarcinoma model. Thin tissue sections may not fully represent the 3D structure of tumors, and the diversity of lineage barcodes in spatial data is lower than in dissociated single-cell analysis. The method of correcting missing barcodes using neighboring cells may introduce errors in highly mobile cancer cells. It remains to be verified whether the same spatial structure is reproducible in human lung adenocarcinoma cohorts and treatment-naïve specimens, and whether subclone metastatic potential is indeed reduced by blocking hypoxia or fibrosis signals.
Nature Genetics, Published online: 02 September 2026; doi:10.1038/s41588-026-02739-zTumors are driven by dynamic interactions between cancer cells and their environment. The authors use spatial lineage tracing to map the spatiotemporal trajectories of a lung adenocarcinoma model and dissect how extrinsic signals drive progression.
Clinically, this opens the possibility of adding a spatial risk map to the current practice of classifying surgical tissues based solely on genomic mutations. Identifying regions in histological sections where hypoxia–EMT cancer cells, Arg1-positive macrophages, and Postn-positive fibroblasts are in close proximity could serve as a basis for selecting high-metastasis-risk subclones and determining the intensity of adjuvant therapy. Pharmaceutical companies could evaluate combination therapies targeting TGF-β, hypoxia response, macrophage suppression, or collagen formation using organoid co-culture systems that mimic these regions. However, the current achievement is at the stage of proposing candidate ecosystems rather than proving clinical efficacy of therapeutic targets. Patient specimen prognosis linkage and prospective drug trials are still needed.