Back

Neuroendocrine prostate cancer converges on a fetal pulmonary neuroendocrine-like program

Yue, W.; Kyprianou, N.; Tewari, A. K.; Padanilam, B. J.

2026-07-08 cancer biology
10.64898/2026.07.07.737089 bioRxiv
Show abstract

Advanced prostate cancers can relapse as neuroendocrine prostate cancer (NEPC), a treatment-resistant state that is independent of androgen receptor (AR) signaling and whose developmental identity remains unclear. We integrated single-cell, bulk, cistrome and spatial datasets across published cohorts and mapped malignant tumor states to human fetal atlases. In the discovery cohort (35,696 tumor cells, 20 patients), NEPC cells were malignant and shared copy-number architecture with adenocarcinoma (chromosome-level r = 0.64). NEPC most closely matched a fetal, lineage-restricted neuroendocrine program, best supported as pulmonary neuroendocrine-like among the alternatives tested (islet, chromaffin and sympathoblast), with the same direction across three cohorts; the pulmonary-versus-islet distinction was method-sensitive and remains provisional. Cross-sectional stage and pseudotime analyses were consistent with an ordered luminal-to-neuroendocrine sequence, with FOXA2/SOX2 changing early and ASCL1/NEUROD1/MYCN late across five datasets. A regulatory screen using five methods nominated a circuit centered on ASCL1 whose convergent enhancers overlapped fetal pulmonary neuroendocrine chromatin and showed FOXA1 binding in cell-line and xenograft models (87.0% versus 33.8%; P = 0.016). In one NEPC Visium section, the program localized to malignant NEPC territory defined by copy number, with an area under the receiver operating characteristic curve (AUROC) of 0.85. Based on three NEPC patients with cross-cohort replication of direction, these data support NEPC as an adenocarcinoma-related malignant state converging on a fetal pulmonary neuroendocrine-like reference program and nominate regulatory and surface targets linked to ASCL1 for functional study.

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.