Prostate cancer’s shift to a lethal form traced step by step in single cells

When prostate tumors stop responding to hormone therapy, some transform into neuroendocrine prostate cancer, an aggressive and hard-to-treat form of the disease. How this transformation unfolds inside a tumor has been difficult to pin down, because clinical samples usually capture only a single snapshot of either the original adenocarcinoma or the evolved neuroendocrine state. A new study in Genome Medicine now offers a time-resolved view of the process, using a patient-derived model followed across the entire course of treatment-induced change.

Researchers led by Colin C. Collins and Yuzhuo Wang, with co-first authors Funda Sarı, Hee Chul Chung and Dong Lin, performed longitudinal single-cell RNA sequencing of the LTL331 patient-derived xenograft. The model spans eight timepoints, from pre-castration prostate adenocarcinoma through relapsed neuroendocrine prostate cancer after androgen deprivation, covering 51,726 cells in total. The team matched this transcriptomic timeline with whole-exome sequencing, integrated inferred single-cell copy-number profiles, applied trajectory inference, and analyzed regulon and transcription-factor activity, drawing on chromatin-accessibility data from a clinical metastatic castration-resistant prostate cancer atlas.

The central finding is that neuroendocrine transdifferentiation proceeded through transcriptional reprogramming of a single clone rather than through clonal outgrowth. The adenocarcinoma and neuroendocrine states shared the same copy-number architecture and a truncal TMPRSS2::ERG rearrangement, and the androgen receptor locus remained copy-number stable even as androgen receptor transcription was lost. In other words, the genetic substrate did not change in a way that would explain the shift; instead, the cells’ transcriptional identity was rewritten.

Every trajectory the team reconstructed from adenocarcinoma to neuroendocrine cancer funneled through one transient, obligate mesenchymal gateway. This state accounted for roughly 0.25 percent of cells and peaked between 12 and 16 weeks after castration. From this gateway, the lineage re-engaged an early neural-crest program in developmental order: the neural-plate-border regulator MSX1 peaked together with the epithelial-to-mesenchymal delamination program, and the border specifier PAX7 persisted into the early neuroendocrine state.

After passing through the gateway, the lineage bifurcated with a strong bias toward a tuft-negative, ASCL1-negative fate, with a probability of about 0.94 versus about 0.06 for the alternative, a paired comparison yielding a p value of 4.8 × 10−23, and POU2F3 activity below one percent. This route diverges from the canonical ASCL1 and PHOX2B pathway and from the ASCL1-versus-tuft split described in other neuroendocrine prostate cancer models, suggesting that the developmental route taken in this model is distinctive.

MSX1 emerged as a candidate priming regulator of the transition. The evidence comes from transcription-factor-activity analysis together with chromatin footprints specific to androgen receptor-negative, neuroendocrine-positive cells in patient neuroendocrine prostate cancer. The authors emphasize that this is candidate-level support, with the gateway and its regulators framed as potential targets for intercepting the transition rather than validated therapeutic points.

The study also tested whether the lineage states observed in the model could be seen in patients. Adenocarcinoma and ASCL1-positive neuroendocrine programs replicated in two clinical single-cell cohorts using an embedding-free, label-level comparison, with area under the receiver operating characteristic curve values of 0.82 to 0.83. Neuroendocrine identity tracked the clinical neuroendocrine-differentiation annotation, with odds ratios of 134.6 and 31.4 in the two cohorts.

The authors conclude that, in the LTL331 patient-derived model, neuroendocrine transdifferentiation is a directional, stepwise transition rather than a stochastic loss of identity. They state that, to their knowledge, this is the first time-resolved evidence that prostate neuroendocrine transdifferentiation re-engages the early neural-crest program in developmental order before committing to a predominantly ASCL1-negative fate. The transient mesenchymal gateway is described as a potential opportunity for intervention, and its regulators as candidate targets.

Several caveats frame these conclusions. The longitudinal analysis rests on a single patient-derived xenograft model, and clinical replication was limited to label-level comparisons of lineage states in two patient cohorts rather than direct tracking of the transition in patients. The gateway’s small size, about a quarter of one percent of cells, reflects its transient nature but also means the model’s timing and proportions may not map directly onto human disease. The authors note that clinical single-cell studies to date primarily capture established adenocarcinoma or neuroendocrine cancer at single timepoints, which is precisely why the order and direction of the transition have remained only partially defined. The work, published open access on 28 October 2026, is presented as an early-release version subject to further edits before the final version of record.

Subject of Research: Time-resolved single-cell analysis of treatment-induced neuroendocrine transdifferentiation in prostate cancer

Article Title: Longitudinal single-cell RNA sequencing of a neuroendocrine transdifferentiation model reveals transcriptional reprogramming in treatment-induced neuroendocrine prostate cancer

Article References: Sar, F., Chung, H. C., Lin, D., Morova, T., Haegert, A., Volik, S., Bell, R., LeBihan, S., Ozturan, D., Xue, H., Dong, X., Wu, R., Li, J., Kung, H.-J., Gleave, M. E., Lack, N. A., Lin, Y.-Y., Wang, Y., & Collins, C. C. (2026). Longitudinal single-cell RNA sequencing of a neuroendocrine transdifferentiation model reveals transcriptional reprogramming in treatment-induced neuroendocrine prostate cancer. Genome Medicine. https://doi.org/10.1186/s13073-026-01806-8

Image Credits: AI Generated

DOI: 10.1186/s13073-026-01806-8

Keywords: neuroendocrine prostate cancer, single-cell RNA sequencing, patient-derived xenograft, transcriptional reprogramming, neural-crest program, MSX1, androgen deprivation, mesenchymal gateway, ASCL1, trajectory inference, TMPRSS2::ERG, Genome Medicine