How Lung Cancers Shed Their Identity to Outsmart KRAS Drugs

When the first KRAS inhibitors finally reached the clinic after decades of failure, they were hailed as a turning point for lung cancer. For patients whose tumors carry the KRAS G12C mutation, drugs such as sotorasib and adagrasib delivered real, measurable regressions. Yet the victories proved fragile. Responses were often incomplete, and tumors that initially shrank frequently found ways to grow again. A new study published in Nature Genetics by Nicolas Mathey-Andrews, Carrie L. Rodriguez, Tyler Jacks and colleagues at the Massachusetts Institute of Technology and Columbia University now reveals a striking dimension of that resistance: the identity of the tumor cell itself, not just its genetics, determines whether the cancer remains dependent on KRAS or learns to live without it.

The research team turned to genetically engineered mouse models of Kras-driven lung adenocarcinoma, the workhorse systems that have long defined lung cancer biology. These autochthonous models, in which tumors arise in their natural tissue environment after viral activation of oncogenic Kras, faithfully recapitulate the anatomy and evolution of human disease. When the investigators treated mice bearing Kras G12C tumors with sotorasib, the response was rapid and robust. Micro-computed tomography scans showed aeration returning to compressed lung tissue, and histological analysis confirmed widespread tumor regression. But the response was never complete. Even after two weeks of treatment, residual disease persisted, and it was this residual population that held the clues to what would come next.

Single-cell RNA sequencing of the residual tumors revealed something unexpected. Rather than resembling the aggressive adenocarcinoma cells that dominated before treatment, the surviving cells had shifted toward a transcriptional program resembling alveolar epithelial cells, the specialized gas-exchange cells of the lung. Markers such as HOPX and surfactant proteins became prominent. This finding aligns with a growing body of evidence that alveolar differentiation itself can blunt dependence on KRAS signaling, suggesting that the drug pushes tumors into a state where the oncogene matters less. The residual disease was not simply a smaller version of the original tumor; it was a fundamentally different cellular entity, poised at a lineage crossroads.

What happened under continued treatment was equally revealing. With sustained sotorasib administration, the researchers invariably observed genetic amplification of the Kras oncogene as the dominant resistance mechanism. Using advanced genomic tools, including whole-genome sequencing and computational reconstruction of amplification architecture, the team found that resistant tumors had accumulated extra copies of the mutant Kras allele, in some cases organized on circular extrachromosomal DNA, a notoriously dynamic form of oncogene amplification. Some tumors also amplified Myc or Mycn. This is a classic oncogene-centric escape: the cancer simply makes more of the drug target, overwhelming the inhibitor. In these tumors, MAPK pathway activity, measured by phosphorylated ERK staining, remained evident, confirming that KRAS signaling had been restored.

But oncogene amplification is only one route to resistance. The more provocative question the study addressed is whether tumors can escape KRAS dependence altogether by changing who they are. Histologic transformation, in which an adenocarcinoma converts to a squamous cell carcinoma or even small-cell lung cancer under the pressure of therapy, is a documented phenomenon in patients treated with EGFR and ALK inhibitors. The team used functional CRISPR approaches in their mouse models to deliberately promote squamous lineage transformation and ask whether this change in cellular identity could render tumors indifferent to KRAS inhibition.

The answer, established through a series of elegant experiments, was a clear yes. In alveolar organoid cultures derived from Kras-mutant, p53-deficient lung tissue, forced expression of the transcription factor ΔNp63 was sufficient to reprogram the adenocarcinoma cells into a squamous state. These reprogrammed organoids became strikingly insensitive to KRAS inhibitors, surviving drug concentrations that killed their untransformed counterparts. Critically, the squamous organoids did not show reactivation of KRAS or MAPK signaling when treated. Their resistance was not about reactivating the oncogene; it was about no longer needing it. The drug still hit its target, but the target no longer mattered to the cell’s survival.

In living animals, the team identified the genetic conditions that poise tumors for this fate. Loss of Nkx2-1, the master transcription factor that maintains lung adenocarcinoma identity, facilitated squamous transformation in autochthonous tumors. Ectopic expression of Sox2, the lineage-defining transcription factor of squamous lung cancer, cooperated with Nkx2-1 loss to drive adenosquamous tumors, mixed lesions containing both adenocarcinomatous and squamous components. When mice bearing these transformed tumors were treated with sotorasib for extended periods, the squamous compartments persisted and progressed while showing no evidence of KRAS or MAPK reactivation. Lineage transcription factors, the study concludes, can mediate genuine KRAS independence, a mechanism fundamentally distinct from the amplification-driven resistance seen in adenocarcinomas that retain their identity.

The mechanistic logic of this transformation is rooted in developmental biology. NKX2-1 acts as a guardian of alveolar and pulmonary fate, and its loss releases tumor cells from that commitment, allowing alternative lineage programs to take over. ΔNp63, the dominant-negative isoform of the p53 family member p63, is the molecular switch for stratified epithelial identity, essential for the proliferative potential of basal cells in squamous epithelia. SOX2, meanwhile, has been shown in prior work to be the determining oncogenic switch that promotes lung squamous carcinoma from multiple cells of origin. The new study ties these threads together into a coherent model: the balance between NKX2-1 and factors such as ΔNp63 and SOX2 governs whether a KRAS-mutant tumor remains addicted to its driver or can be reprogrammed into a state where the driver is dispensable.

The clinical implications are substantial. KRAS G12C inhibitors are now approved therapies, but clinical responses are incomplete and resistance is nearly universal, with documented mechanisms including new KRAS mutations, bypass signaling through EGFR and other receptors, and epithelial-to-mesenchymal transition. The mouse models in this study mirror that reality: initial responses were rapid but partial, and residual disease consistently evolved under drug pressure. If a subset of human tumors can achieve resistance through lineage plasticity rather than target reactivation, then combination strategies aimed solely at suppressing MAPK signaling will fail for those patients. Monitoring for lineage markers, such as NKX2-1 loss, ΔNp63 or SOX2 emergence, could identify patients whose tumors are drifting toward KRAS independence before overt progression, and biopsy of resistant lesions may need to look at histology as closely as genomics.

There is also a broader lesson for targeted therapy across oncology. The study echoes patterns seen in EGFR-mutant lung cancer, where squamous and small-cell transformation accompany resistance to osimertinib, often in tumors with RB1 and TP53 alterations. In those contexts too, the transformed tumors frequently lose dependence on the original driver. What this new work adds is a mechanistic, causally tested framework in mouse models: specific transcription factors are sufficient to confer drug resistance by rewriting cellular identity, and the resulting state is stable and heritable. Therapies that target the lineage machinery itself, or that prevent the plastic transition in the first place, may be needed alongside KRAS inhibitors to make responses durable. As KRAS drugs become a cornerstone of lung cancer treatment, this study makes clear that the battle against resistance will be fought not only over the oncogene, but over the very identity of the tumor cell.

Subject of Research: Lineage plasticity and transcription factor-driven squamous transformation as mechanisms of resistance to KRAS inhibitors in mouse models of non-small-cell lung cancer

Article Title: Lineage identity governs oncogene dependence in mouse NSCLC models of KRAS inhibitor resistance

Article References: Mathey-Andrews, N., Rodriguez, C. L., Shui, B., Patriotis, A. L., Rideout, W. M., III, Chen, V. Z., Murazzi, I., Ghazi, P., Cornwall-Brady, M. R., Liu, M., Heileman, M. G., Trakala, M., Concepcion-Crisol, C. P., Yang, D., & Jacks, T. (2026). Lineage identity governs oncogene dependence in mouse NSCLC models of KRAS inhibitor resistance. Nature Genetics. https://doi.org/10.1038/s41588-026-02768-8

Image Credits: AI Generated

DOI: 10.1038/s41588-026-02768-8

Keywords: KRAS inhibitors, lung adenocarcinoma, lineage plasticity, squamous transformation, NKX2-1, SOX2, DeltaNp63, drug resistance, genetically engineered mouse models, targeted therapy, MAPK signaling, histologic transformation