Cancer’s Family Tree Gets a Map: Lineage Tracing Reveals How Tumors Grow and Spread

Tumors are not the uniform masses of cells that early cancer biology often imagined them to be. They are evolving ecosystems, in which genetically distinct subclones of cancer cells compete, cooperate and reshape the tissue around them as they expand. A new study published in Nature Genetics has now brought an unprecedented level of resolution to this dynamic picture, combining high-resolution spatial transcriptomics with evolving lineage-tracing technologies to map, in both space and time, how lung tumors grow, remodel their microenvironment and seed metastases. The work, led by Matthew G. Jones, Dawei Sun and colleagues across a large multi-institutional collaboration, offers one of the most comprehensive datasets yet assembled to connect a tumor’s genealogy to its geography.

The research team focused on a well-established mouse model of lung adenocarcinoma driven by mutations in the Kras and Trp53 genes, a system that closely recapitulates the progression of human non-small-cell lung cancer. In this model, known as KP-Tracer, cancer cells carry heritable molecular barcodes that accumulate edits as cells divide, allowing researchers to reconstruct family trees of tumor cells long after the fact. By reading out these barcodes alongside genome-wide gene expression, the team could infer not only which cells were related to one another but also where they sat within the tumor and what molecular programs they were running.

Technically, the platform integrates two complementary spatial assays. Slide-seq provides genome-wide expression measurements on dense arrays of bead-based spots at near-cellular resolution across entire tissue sections, while Slide-tags assigns spatial coordinates to individual cell nuclei, enabling single-cell profiling with positional information. The lineage barcodes embedded in the tumor cells could be captured in both assays, though with substantial dropout and missing data. To address this, the team developed computational methods, including spatial imputation strategies that borrow lineage information from neighboring spots, and benchmarked their phylogeny-reconstruction pipelines extensively on simulated data to ensure that the inferred evolutionary trees were robust to the noise inherent in spatial measurements.

With this integrated platform in hand, the researchers asked a fundamental question: where within a tumor does expansion actually happen, and what does the microenvironment look like in those regions? By combining the reconstructed phylogenies with spatial maps, an approach the authors describe as tumor phylogeography, they identified regions of recent subclonal expansion, essentially the growing edges of the tumor’s family tree. These expanding subclones were not randomly distributed. Instead, they were consistently associated with a distinctive microenvironmental signature: hypoxia, fibrosis and immunosuppression.

The association was striking. Areas harboring rapidly expanding subclones were enriched for low-oxygen conditions, marked by expression of hypoxia-response genes such as the glucose transporter GLUT1. They also contained dense deposits of extracellular matrix produced by activated fibroblasts and were populated by immunosuppressive immune cells, including Arg1-expressing tumor-associated macrophages. In other words, the most successful cancer clones were not simply the ones with the best intrinsic growth programs; they were the ones that had managed to engineer, or at least exploit, a microenvironment that suppressed immune attack and supplied the conditions for aggressive proliferation.

To disentangle cause from correlation, the team turned to controlled experiments. Using organoid co-culture systems, they exposed cancer cells to hypoxic conditions and to specific stromal cell partners, testing how these extrinsic factors influenced cancer cell state. The results supported a model in which hypoxia and intercellular signaling integrate to push cancer cells toward prometastatic, high-plasticity states, including epithelial-to-mesenchymal transition-like programs previously linked to metastatic competence. Spatially aware ligand-receptor analysis, performed with a purpose-built algorithm called LARIS, further revealed that the rewired interactions between macrophages, fibroblasts and cancer cells in expanding niches differed markedly from those in non-expanding regions, pinpointing candidate signaling pathways that sustain the aggressive state.

Perhaps the most consequential findings concern metastasis. By tracing lineage barcodes from primary tumors into metastatic lesions found in lymph nodes, the diaphragm and other sites, the researchers showed that metastases arise from spatially confined subclones within the primary tumor rather than from cells scattered broadly across it. The metastasis-seeding subclones occupied identifiable niches at the primary site, and their genealogical signatures could be detected across serial tissue sections, effectively allowing the team to watch the metastatic cascade unfold backward from the established lesion to its birthplace in the primary tumor.

Equally important, the study found that metastases do not merely inherit traits from their parent clones; they actively remodel the distant sites they colonize. Metastatic lesions, and even the pre-metastatic neighborhoods surrounding them, became fibrotic and collagen-rich, with elevated TGF-beta signaling. The team extended this observation to human disease by analyzing single-cell data from a pan-cancer atlas of human brain metastases and spatial transcriptomics datasets of human non-small-cell lung cancer, finding that collagen deposition, TGF-beta activity and hypoxia signatures were similarly elevated in human metastatic compartments. This convergence between the mouse model and human data strengthens the case that the mechanisms uncovered are not artifacts of the experimental system.

The implications for cancer medicine are substantial. If prometastatic cell states emerge specifically within hypoxic, fibrotic and immunosuppressive niches, then targeting the microenvironment, for example by alleviating hypoxia, modulating fibroblast activity or reprogramming suppressive macrophages, could potentially prevent the emergence of metastatic competence before it arises. The findings also suggest that sampling strategies in the clinic, which often rely on a single biopsy, may miss the spatially restricted subclones that matter most for a patient’s prognosis. Understanding where within a tumor the dangerous clones reside could inform how biopsies are taken and how risk is assessed.

The study also represents a methodological milestone for the field of spatial lineage tracing. The authors have released their processed data via Zenodo, deposited raw sequencing data under a public BioProject accession, and made their analysis code, including the Cassiopeia lineage-reconstruction framework and spatial analysis notebooks, freely available on GitHub under an open license. As these tools proliferate, the ability to read a tumor’s history directly from its architecture may become a standard part of the cancer biologist’s toolkit, transforming how researchers study not only lung cancer but the evolutionary dynamics of malignancies throughout the body.

Subject of Research: Spatiotemporal lineage tracing of lung adenocarcinoma to map tumor growth, microenvironmental remodeling and metastasis

Article Title: Spatiotemporal lineage tracing reveals the dynamic spatial architecture of tumor growth and metastasis

Article References: Jones, M. G., Sun, D., Min, K. H. J., Colgan, W. N., Wang, H., Török, T., Ribeiro, J., Xue, J., Cardoso, E. C., Rong, Y., Tian, L., Weir, J. A., Chen, V. Z., Koblan, L. W., Yost, K. E., Mathey-Andrews, N., D’Souza, E., Russell, A. J. C., Stickels, R. R., … Yang, D. (2026). Spatiotemporal lineage tracing reveals the dynamic spatial architecture of tumor growth and metastasis. Nature Genetics, 58(9), 2398-2410. https://doi.org/10.1038/s41588-026-02739-z

Image Credits: AI Generated

DOI: 10.1038/s41588-026-02739-z

Keywords: lineage tracing, spatial transcriptomics, tumor evolution, lung adenocarcinoma, tumor microenvironment, metastasis, hypoxia, fibrosis, immunosuppression, tumor phylogeography, Slide-seq, Slide-tags