Following cancer through the body: The heterogeneity of mammalian tumors has been well documented, but it remains unknown how differences between individual cells lead to metastasis and spread throughout the body. Quinn et al. created a Cas9-based lineage tracer and used single-cell sequencing to generate phylogenies and follow the movement of metastatic human cancer cells implanted in the lung of a mouse xenograph model. Using this model, they found that within the same cell line, cancer cells exhibited diverse metastatic phenotypes. These subclones exhibited differential gene expression profiles, some of which were previously associated with metastasis. Science , this issue p. eabc1944
Charting the spatiotemporal dynamics of cell fate determination in development and disease is a long-standing objective in biology. Here, we present the design, development, and extensive validation of PEtracer, a prime editing (PE)–based, evolving lineage tracing technology compatible with both single-cell sequencing and multimodal imaging methodologies, created to jointly profile cell state and lineage in dissociated cells or while preserving cellular context in tissues with high spatial resolution. Using PEtracer coupled with MERFISH spatial transcriptomic profiling in a syngeneic mouse model of tumor metastasis, we reconstructed the growth of individually seeded tumors in vivo and uncovered distinct modules of cell-intrinsic and -extrinsic factors that coordinate tumor growth. More generally, PEtracer enables systematic characterization of cell state and lineage relationships in intact tissues over biologically relevant temporal and spatial scales.
Oncogenes amplified on extrachromosomal DNA (ecDNA) contribute to treatment resistance and poor survival across cancers. Currently, the spatiotemporal evolution of ecDNA remains poorly understood. In this study, we integrate computational modeling with samples from 94 treatment-naive human glioblastomas (GBM) to investigate the spatiotemporal evolution of ecDNA. We observe oncogene-specific patterns of ecDNA spatial heterogeneity, emerging from random ecDNA segregation and differing fitness advantages. Unlike PDGFRA-ecDNAs, EGFR-ecDNAs often accumulate prior to clonal expansions, conferring strong fitness advantages and reaching high abundances. In corroboration, we observe pretumor ecDNA accumulation in vivo in genetically engineered mouse neural stem cells. Variant and wild-type EGFR-ecDNAs often coexist in GBM. Those variant EGFR-ecDNAs, most commonly EGFRvIII-ecDNA, always derive from preexisting wild-type EGFR-ecDNAs, occur early, and reach high abundance. Our results suggest that the ecDNA oncogenic makeup determines unique evolutionary trajectories. New concepts such as ecDNA clonality and heteroplasmy require a refined evolutionary interpretation of genomic data in a large subset of GBMs. Significance: We study spatial patterns of ecDNA-amplified oncogenes and their evolutionary properties in human GBM, revealing an ecDNA landscape and ecDNA oncogene–specific evolutionary histories. ecDNA accumulation can precede clonal expansion, facilitating the emergence of EGFR oncogenic variants, reframing our interpretation of genomic data in a large subset of GBMs. See related commentary by Korsah et al., p. 1979
Oncogene amplification on extrachromosomal DNA (ecDNA) is a pervasive driver event in cancer, yet our understanding of how ecDNA forms is limited. In this study, we couple a CRISPR-based method for ecDNA induction with extensive characterization of newly formed ecDNAs to examine their biogenesis. We find that DNA circularization is efficient, irrespective of 3D genome context, with the formation of 800 kb, 1 Mb, and 1.8 Mb ecDNAs reaching or exceeding 15%. We show nonhomologous end joining and microhomology-mediated end joining both contribute to ecDNA formation, whereas inhibition of DNA-PK catalytic subunit and ATM have opposing impacts on ecDNA formation. ecDNA and the corresponding chromosomal excision scar can form at significantly different rates and respond differently to DNA-PK catalytic subunit and ATM inhibition. Taken together, our results support a model of ecDNA formation in which double-strand break ends dissociate from their legitimate ligation partners prior to joining of illegitimate ends to form the ecDNA and excision scar. Significance: Our study harnesses a CRISPR-based method to examine ecDNA biogenesis, uncovering efficient circularization between double-strand breaks. ecDNAs and their corresponding chromosomal scars can form via nonhomologous end joining or microhomology-mediated end joining, but the ecDNA and scar formation processes are distinct. Based on our findings, we establish a mechanistic model of excisional ecDNA formation.
Foundation models have transformed natural language processing and computer vision, yet their potential in single-cell biology—particularly for complex diseases such as cancer— remains underexplored. We present Tahoe-x1 (Tx1), a family of perturbation-trained single-cell foundation models with up to 3 billion parameters. Tx1 is pretrained on large-scale single-cell transcriptomic datasets, including the Tahoe-100M perturbation compendium, and fine-tuned for cancer-relevant tasks. Through architectural optimizations, data loader refinements, and efficient training strategies, Tx1 achieves 3–30× higher compute efficiency than prior implementations of cell-state models. Tx1 jointly learns representations of genes, cells, and compounds using a masked-expression generative objective that incorporates a drug token, enabling flexible adaptation to diverse downstream applications. We evaluate Tx1 across four key disease-relevant benchmarks: (1) prediction of overall and context-specific gene essentiality, (2) identification of genes contributing to the hallmarks of cancer, (3) cell-type classification, and (4) prediction of perturbation responses in held-out cellular contexts. Tx1 achieves state-of-the-art performance across all tasks. We release pretrained checkpoints, training code, and evaluation workflows to accelerate the development of perturbation-trained single-cell foundation models for applications in precision oncology and beyond.
A bstract: Novel experimental assays now simultaneously measure lineage relationships and transcriptomic states from single cells, thanks to CRISPR/Cas9-based genome engineering. These multimodal measurements allow researchers not only to build comprehensive phylogenetic models relating all cells but also infer transcriptomic determinants of consequential subclonal behavior. The gene expression data, however, is limited to cells that are currently present (“leaves” of the phylogeny). As a consequence, researchers cannot form hypotheses about unobserved, or “ancestral”, states that gave rise to the observed population. To address this, we introduce TreeVAE: a probabilistic framework for estimating ancestral transcriptional states. TreeVAE uses a variational autoencoder (VAE) to model the observed transcriptomic data while accounting for the phylogenetic relationships between cells. Using simulations, we demonstrate that TreeVAE outperforms benchmarks in reconstructing ancestral states on several metrics. TreeVAE also provides a measure of uncertainty, which we demonstrate to correlate well with its prediction accuracy. This estimate therefore potentially provides a data-driven way to estimate how far back in the ancestor chain predictions could be made. Finally, using real data from lung cancer metastasis, we show that accounting for phylogenetic relationship between cells improves goodness of fit. Together, TreeVAE provides a principled framework for reconstructing unobserved cellular states from single cell lineage tracing data.