Single-cell atlas reveals how hidden clones drive pancreatic cancer’s deadly evolution

Pancreatic ductal adenocarcinoma, the most common and most lethal form of pancreatic cancer, has long frustrated oncologists with its ability to resist chemotherapy, evade the immune system and spread with ruthless efficiency. Now, one of the most comprehensive single-cell studies of the disease ever conducted offers a detailed explanation for that behavior. A team led by researchers at NYU Grossman School of Medicine and UC San Diego Health has mapped, cell by cell, the genetic and spatial architecture of human pancreatic tumors, revealing that individual tumors are not uniform masses of malignant tissue but contested territories ruled by competing genetically distinct clones. The work, published in Nature Genetics, suggests that the most aggressive behaviors of pancreatic cancer emerge not from the tumor as a whole but from specific evolutionary branches within it.

The scale of the analysis sets it apart. The researchers profiled 62 primary pancreatic tumors using single-cell RNA sequencing, capturing the gene-expression fingerprints of 443,451 individual cells. For 11 of those tumors, they added spatial transcriptomics, a technique that preserves the physical location of gene activity across 53,236 tissue spots, allowing the team to see not just which cells existed but where they sat relative to one another. Crucially, the cohort also included matched samples from six tumors collected before and after chemotherapy, providing a rare before-and-after view of how treatment reshapes the cellular landscape of the disease.

Because single-cell RNA sequencing does not directly read DNA, the team inferred the copy-number alterations, gains and losses of chromosomal segments, from expression data, then validated the approach against whole-genome sequencing and confirmed key findings with fluorescence in situ hybridization. Using these inferred genetic signatures as barcodes, they grouped malignant cells into clones and reconstructed evolutionary relationships within each tumor. The result was unambiguous: multiple genetically distinct neoplastic clones coexist within individual tumors, each carrying its own complement of chromosomal alterations and, remarkably, its own transcriptional personality.

That clone-specific personality extends beyond the cancer cells themselves. The analysis revealed that different clones deploy different ligand–receptor programs, the molecular handshakes through which cells communicate with their surroundings, and that these communication programs are modulated by genomic dosage. When a clone carried an amplification of a ligand or receptor gene, the corresponding interactions with neighboring cells were enriched; when the gene was deleted, the interactions were suppressed. In other words, the raw genetic equipment a clone inherits determines the conversations it can hold with the tumor microenvironment. Spatial transcriptomics and multiplex immunofluorescence independently confirmed that these genetically encoded differences are visible in intact tissue, with one validated example being the NECTIN2–TIGIT axis, an immunosuppressive interaction between tumor cells and T cells that was strengthened in clones carrying NECTIN2 copy-number gains.

Perhaps the most striking finding concerns the transition between the two major molecular subtypes of pancreatic cancer, known as classical and basal. Tumors dominated by the classical program tend to respond somewhat better to therapy, while the basal program is associated with aggression and resistance. By integrating the clone-level genetic data with phylogenetic reconstruction, the researchers found that focal amplification of the 8q24 region, which contains the potent oncogene MYC, was recurrently associated with classical-to-basal transitions occurring within tumor lineages at a frequency above chance. These were not separate tumors with different subtypes; they were evolutionary events unfolding inside a single patient’s cancer, in which one branch of the tumor’s family tree switched identity.

The consequences of that switch were equally significant. The clones that underwent the classical-to-basal transition showed a coordinated depletion of immune-response programs, suggesting that as they shifted identity, they also remodeled the local immune environment in ways that shielded them from attack. To test whether MYC was a driver or merely a passenger, the team performed functional perturbation experiments in murine models and primary human tumor-derived systems. When they knocked down MYC, the cells reactivated classical transcriptional programs and restored immune-response pathways. The evidence supports a model in which MYC actively maintains the basal, immune-depleted state, positioning it as a linchpin connecting genome evolution, cell identity and immune evasion.

Spatial analysis added a third dimension to the story. The team identified dispersed clones, malignant populations that spread widely across the tumor rather than forming compact territories, that were enriched for epithelial-to-mesenchymal transition activity, the developmental program that endows epithelial cells with migratory, invasive properties, and for immune-suppressive signaling. These dispersed clones correlated with metastatic potential and were found to colonize lymph node niches, the anatomical way stations through which pancreatic cancer typically begins its deadly spread. The dispersed clones had also shifted toward a basal phenotype, tying together the threads of genetic evolution, transcriptional plasticity and anatomical progression into a single narrative of how the disease advances.

The chemotherapy-matched samples offered a sobering evolutionary perspective. Comparing clones before and after treatment, the researchers documented clones that were eradicated, clones that persisted in expanded, contracted or stable form, and clones that emerged de novo after therapy. Retained clones showed distinct patterns of interaction with lymphatic and stromal cells compared with eradicated ones, and the interactions between specific stromal populations, including inflammatory and myofibroblastic cancer-associated fibroblasts and endothelial cells, differed systematically across the eradicated, pre-treatment retained and post-treatment retained clone groups. Chemotherapy, in this view, is not simply a cytotoxic assault but an evolutionary filter that selects for clones best equipped to survive in a remodeled microenvironment.

The technical achievement underlying these findings deserves emphasis. Distinguishing malignant from normal epithelial cells, assigning cells to clones, deconvolving spatial spots into their constituent cell types and reconstructing phylogenies from noisy single-cell data required an integrated computational pipeline drawing on tools for copy-number inference, cell–cell communication analysis and pathway enrichment. The team validated their copy-number calls against whole-genome sequencing and confirmed spatial relationships with orthogonal imaging, an unusually rigorous level of cross-checking for a single-cell study. The processed data have been deposited in public repositories, and the raw data are available through the database of Genotypes and Phenotypes under controlled access, making the resource available to the wider research community.

For a disease with a five-year survival rate that remains among the lowest of any major cancer, the study’s implications are potentially significant. If the most dangerous properties of pancreatic cancer, immune evasion, subtype switching and metastatic seeding, are properties of specific clones rather than of the tumor as a whole, then therapies could be designed to target the mechanisms those clones exploit. The demonstration that MYC knockdown reverses the basal, immune-depleted state points toward strategies aimed at clonal evolution itself, while the clone-specific ligand–receptor programs suggest that blocking particular tumor–microenvironment interactions could disarm the most aggressive branches of a tumor’s family tree. The authors caution that these are avenues for future investigation rather than ready-made treatments, but the study provides something pancreatic cancer research has lacked: a detailed, spatially resolved map of the enemy’s internal politics, and a demonstration that within that map lie the vulnerabilities that evolution has left exposed.

Subject of Research: Clonal heterogeneity and tumor progression in human pancreatic ductal adenocarcinoma

Article Title: Clonal heterogeneity in human pancreatic ductal adenocarcinoma and its impact on tumor progression

Article References: Clonal heterogeneity in human pancreatic ductal adenocarcinoma and its impact on tumor progression. (n.d.). https://doi.org/10.1038/s41588-026-02779-5

Image Credits: AI Generated

DOI: 10.1038/s41588-026-02779-5

Keywords: pancreatic cancer, single-cell RNA sequencing, spatial transcriptomics, clonal heterogeneity, MYC, tumor microenvironment, epithelial-to-mesenchymal transition, chemotherapy resistance, tumor evolution, immune evasion, copy-number variation, metastasis