Treatment-induced neuroendocrine prostate cancer (NEPC) represents an aggressive form of castration-resistant prostate cancer (CRPC) associated with lineage plasticity and therapeutic resistance. In this study, we investigated the role of the Hippo signaling axis in the transdifferentiation from androgen receptor-positive prostate cancer (ARPC) to NEPC. RNA sequencing analyses of CRPC metastases revealed coordinated alterations in Hippo pathway components, with decreased expression of YAP1, LATS2, and TEAD2 and increased expression of LATS1, TEAD1, and the RNA splicing regulator RBFOX2 in NEPC. These transcriptional alterations were consistently observed across multiple model systems and patient samples. Epigenetic analyses demonstrated that reduced expression of YAP1, TEAD2, and LATS2 was associated with increased DNA methylation, whereas elevated TEAD1 expression correlated with DNA hypomethylation in NEPC. NEPC selectively retained TEAD1 expression, including a spliced isoform not detected in ARPC. Proteomic interactome analyses revealed that TEAD1 associated with RNA splicing factors and DNA repair proteins. Functional studies showed that TEAD1 knockdown led to the reversion of gene programs associated with epithelial differentiation. These findings indicate that the conversion of ARPC to NEPC involves coordinated loss of AR, YAP1, and REST activity alongside sustained TEAD1 expression and altered RNA processing. Our data identify TEAD1 as a transcriptional regulator associated with the NEPC state and suggest a role for TEAD1-linked transcriptional and post-transcriptional mechanisms in prostate cancer lineage plasticity.

Circulating tumor DNA (ctDNA) profiling from liquid biopsies is increasingly adopted as a minimally invasive solution for clinical cancer diagnostic applications. Current methods for inferring gene expression from ctDNA require specialized assays or ultra-deep, targeted sequencing, which preclude transcriptome-wide profiling at single-gene resolution. Herein we jointly introduce Triton, a tool for comprehensive fragmentomic and nucleosome profiling of cell-free DNA (cfDNA), and Proteus, a multi-modal deep learning framework for predicting single gene expression, using standard depth (~30-120x) whole genome sequencing of cfDNA. By synthesizing fragmentation and inferred nucleosome positioning patterns in the promoter and gene body from Triton, Proteus reproduced expression profiles using pure ctDNA from patient-derived xenografts (PDX) with an accuracy similar to RNA-Seq technical replicates. Applying Proteus to cfDNA from four patient cohorts with matched tumor RNA-Seq, we show that the model accurately predicted the expression of specific prognostic and phenotype markers and therapeutic targets. As an analog to RNA-Seq, we further confirmed the immediate applicability of Proteus to existing tools through accurate prediction of gene pathway enrichment scores. Our results demonstrate the potential clinical utility of Triton and Proteus as non-invasive tools for precision oncology applications such as cancer monitoring and therapeutic guidance.

Abstract Background: Metastatic castration-resistant prostate cancer (mCRPC) is a lethal disease with a major unmet clinical need to identify therapeutics that extend survival and biomarkers that predict treatment responses. The prostate-specific membrane antigen (PSMA) radioligand, 177Lu-PSMA-617 (177Lu-PSMA) is documented to increase overall survival (OS), but responses are highly variable. Analysis of cell-free DNA (cfDNA) from liquid biopsies offers a non-invasive window to profile tumor genomics and phenotypes, and identify markers predictive of 177Lu-PSMA outcomes. Methods: We interrogated a real-world prospective cohort of 140 patients with mCRPC meeting clinical guidelines for 177Lu-PSMA treatment. We analyzed pre-treatment and on-treatment circulating tumor DNA (ctDNA) by whole genome sequencing to quantitate ctDNA abundance and assess tumor-derived genomic features such as copy number alterations, somatic mutations, and structural rearrangements. We also employed a novel computational tool termed Proteus that uses ctDNA fragmentomic profiles to determine the expression of individual genes as well as expression programs that comprise a spectrum of relevant tumor phenotypes including mCRPC lineage subtypes, proliferation scores, and hypoxia signatures that may plausibly influence responses to 177Lu-PSMA. Tumor phenotype characteristics were integrated with PSMA-PET and SPECT imaging findings and evaluated for associations with clinical outcomes. Results: Pre-treatment ctDNA fraction was strongly associated of OS (multivariate HR=2.19, P=0.002), independent of prostate-specific antigen (PSA) response, and correlated positively with PSMA- and FDG-PET total tumor volume (R=0.59 and R=0.69, respectively). Poor responders were characterized by higher baseline ctDNA fraction, genomic instability, and overall tumor mutational burden. Deleterious alterations in the DNA repair gene ATM were associated with improved OS, whereas MYC amplifications were associated with poorer survival. Tumor phenotype features derived from ctDNA including cell cycle proliferation score, neuroendocrine differentiation, and complex copy number variation on chromosome 8 independently associated with clinical outcomes. Conclusions: Our findings demonstrate that baseline ctDNA analysis provides prognostic information for mCRPC patients undergoing 177Lu-PSMA therapy. ctDNA fraction and quantity, alongside specific genomic alterations and tumor phenotype features represent promising, non-invasively acquired biomarkers to guide patient selection, improve therapeutic monitoring, and further dissect genomic mechanisms of resistance to 177Lu-PSMA. Citation Format: Arthur P. McDeed, Robert Patton, Roman Gulati, Lukas Owens, Patricia Galipeau, Pooja Chandra, Aditya Pawar, Weston Hanson, Ruth Dumpit, Amir Iravini, Alireza Ghodsi, Delphine Chen, Michael T. Schweizer, Ruben Raychaudhuri, Michael Haffner, Gavin Ha, Peter S. Nelson. Prostate cancer phenotypes determined from circulating tumor DNA associate with outcomes to 177Lu-PSMA-617 therapy [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(8_Suppl):Abstract nr LB006.

Metastatic prostate cancer is a clinically and molecularly heterogeneous disease. Under the selective pressure of androgen receptor (AR)-directed therapies, resistant phenotypes frequently emerge, posing significant diagnostic and therapeutic challenges. Neuroendocrine prostate cancer (NEPC) is a clinically important phenotype characterized by lineage plasticity, neuroendocrine features, visceral metastases and poor prognosis. Accurately diagnosing NEPC remains difficult due to its histologic and molecular complexity but has high clinical relevance. In this study, we developed a deep learning model that leverages interpretable cellular features to improve feature extraction from H&E-stained tissue sections (NEURAL-PC). By incorporating a multiple instance learning (MIL) framework, NEURAL-PC enables robust NEPC classification solely from H&E tumor images, achieving an area under the receiver operating characteristic curve (AUROC) of 0.921 in independent external validation. In addition to its diagnostic utility, NEURAL-PC provides prognostic information that enables further subclassification of advanced prostate cancer across diverse datasets supporting its strong prognostic value and generalizability. Broadly, our work highlights a hybrid approach that integrates features across different domains, offering a promising strategy for developing reliable deep learning tools in pathology. Built on this framework, NEURAL-PC represents an extensively validated diagnostic and prognostic model for advanced prostate cancer.

Metastatic castration-resistant prostate cancer (mCRPC) is an aggressive subtype of prostate cancer (PC) without curative treatments. Antibody-drug conjugates (ADCs) emerged as promising cancer therapeutics that selectively deliver cytotoxic agents (payloads) to the tumors. Although ADCs have been successfully applied to treat hematological and solid tumors, ADC monotherapy has not demonstrated durable responses in mCRPC, and mechanisms of PC resistance to ADCs have not been thoroughly investigated. Our study aimed to improve ADC efficacy using an integrated approach for a custom ADC design and multiplexing. To nominate rational combinations of ADC targets and payloads, we (a) examined protein coexpression of 3 clinically relevant surface antigens - B7-H3, PSMA, and STEAP1 - in human mCRPCs and (b) screened established ADC payloads and their combinations in mCRPC cell lines with different molecular backgrounds. Identified synergistic interactions between DNA-damaging payloads and the BCL-XL inhibitor A-1331852 as well as their coordinated induction of the intrinsic apoptosis pathway were evaluated in PC cell lines. Functional relevance between isolated p53 loss and PC responses to 3 genotoxic ADCs - B7-H3-seco-DUBA, PSMA-SG3249, and STEAP1-DXd - and their combinations with A-1331852 were established using genetic knockout models. Lastly, enhanced in vivo antitumor activity of vobramitamab duocarmazine by systemic A-1331852 was shown. Collectively, our findings provide rationale for development of ADC therapies combining genotoxic payloads with BCL-XL inhibitors for mCRPC.

The Androgen Receptor (AR) is a crucial master regulator for Prostate Cancer (PC) survival. Most metastatic PCs exhibit an AR-active phenotype- ARPC, but under treatment pressure, a subset of ARPC undergoes a lineage transition to neuroendocrine PC- NEPC or double negative PC (DNPC). LTL331 is an in vivo model that undergoes a treatment-induced lineage switch from ARPC to NEPC. We hypothesized that double negative (AR-/NE-) phenotype represents a de-differentiated intermediate stage from ARPC to NEPC trans-differentiation, where AR downstream gene function loss precedes the AR loss, potentially via reversible epigenetic mechanisms. We implemented both RNA FISH-IF imaging and sequencing approaches to characterize and explore the underlying mechanisms. We adapted the LTL331 ARPC PDX to grow as a 2D cell strain and determined that these cells, LTL331_DNPC_CL exhibit features of DNPC. Notably, LTL331_DNPC_CL expressed high levels of AR transcripts yet low AR protein. Similarly, AR regulated genes (e.g., KLK3, NKX3.1) exhibited transcription-translation discordance, suggesting post-transcriptional alterations in gene regulation. Supraphysiological levels of androgens, as well as proteasome inhibition restored AR protein levels but did not rescue AR-regulated gene expression. LTL331_DNPC_CL cells exhibit a DNPC phenotype initially, with late passage cells activating neuroendocrine and EMT related genes including CEACAM5, FOXF1, and HGF. These results indicate that the transition from an ARPC state to a NEPC state includes an intermediate phenotype with maintenance of AR expression but loss of canonical AR activity, followed by complete AR loss and gain of NE characteristics. Notably, early steps in the process are regulated via post-translational regulation with a degree of reversibility. The model provides insights into the transdifferentiation process and possible treatment window for lineage reversion. Ruihong Wang, Dapei Li, Sander Frank, Ruth Dumpit, Armand Bankhead, Ilsa Coleman, Peter Nelson. Identification of a potential intermediate cell state in the treatment-induced lineage transition from AR-driven prostate cancer to neuroendocrine prostate cancer [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Innovations in Prostate Cancer Research and Treatment; 2026 Jan 20-22; Philadelphia PA. Philadelphia (PA): AACR; Cancer Res 2026;86(2_Suppl):Abstract nr A073.

Differential regulation metrics and values for select features Sheet 1: Log2 fold-change, p-value, and q-value between ARPC and NEPC lines for NPS in the 47 phenotype defining gene bodies (two tailed Mann-Whitney U test, Benjamini-Hochberg adjusted). Sheet 2: Differentially expressed list of 514 transcription factor (TF). All statistical comparison and fold change estimation was done against ARPCs. Sheet 3: Log2 fold-change, p-value, and q-value between ARPC and NEPC lines for central mean coverage in 108 TFs overlapping RNA-Seq up/down regulated TFs (two tailed Mann-Whitney U test, Benjamini-Hochberg adjusted). Sheet 4: Paralogous transcription factors (TFs) for each of the 38 TFs with differential RNA expression between ARPC and NEPC and differential TFBS accessibility. Paralogous TFs that are also differentially expressed by RNAseq analysis of PDX tumors are shown in red text. Paralogs were obtained from ensembl biomart human genes GRCh38.p13 (http://uswest.ensembl.org/biomart/martview/64d8bd7fe9851a2501aece9b74b03631) Sheet 5: Mean values in ARPC, NEPC, and HD lines for central and window means for ARPC and NEPC specific open chromatin regions.

Unsupervised model predictions and patient/validation cohort sequencing metrics Sheet 1: DFCI Cohort I: Tumor phenotype by histology and estimates of tumor fraction, subtyping score, and inferred subtype calls using ctdPheno. Sheet 2: DFCI cohort I: Tumor phenotype by histology and estimates of tumor fraction, subtyping score, and inferred phenotypic subtype calls using different ATAC-seq site restricted analysis using ctdPheno. Sheet 3: DFCI cohort I: Tumor phenotype by histology and estimates of tumor fraction, subtyping score, and inferred phenotypic subtype calls using different ATAC-seq site restricted analysis using Keraon. Sheet 4: DFCI Cohort II: Tumor phenotype by histology, summary of clinical correlatives and estimates of tumor fraction, subtyping score, and inferred subtype calls using ctdPheno. Sheet 5: Histology, tumor fraction, subtype score, NEPC fraction, and subtype calls for WGS and ULP patient samples (UW cohort). Sheet 6: Complete sequencing metrics for DFCI cohort I. Sheet 7: Complete sequencing metrics for DFCI cohort II (ULP). Sheet 8: Complete sequencing metrics for DFCI cohort II (deep WGS) Sheet 9: Complete sequencing metrics and ichorCNA estimated tumor fractions for the UW cohort (ULP). Sheet 10: Complete sequencing metrics for the UW cohort (deep WGS) Sheet 11: Complete sequencing metrics for the healthy donor cohort. Sheet 12: Clinical data summary for UW cohort.

ABSTRACTGenomic loss of the transcriptional kinaseCDK12occurs in ∼6% of metastatic castration-resistant prostate cancers (mCRPC) and correlates with poor patient outcomes. Prior studies demonstrate that acute CDK12 loss confers a homologous recombination (HR) deficiency (HRd) phenotype via premature intronic polyadenylation (IPA) of key HR pathway genes, includingATM.However, mCRPC patients have not demonstrated benefit from therapies that exploit HRd such as inhibitors of polyADP ribose polymerase (PARP). Based on this discordance, we sought to test the hypothesis that an HRd phenotype is primarily a consequence of acuteCDK12loss and the effect is greatly diminished in prostate cancers adapted toCDK12loss. Analyses of whole genome sequences (WGS) and RNA sequences (RNAseq) of human mCRPCs determined that tumors with biallelicCDK12alterations (CDK12BAL) lack genomic scar signatures indicative of HRd, despite carrying bi-allelic loss and the appearance of the hallmark tandem-duplicator phenotype (TDP). Experiments confirmed that acute CDK12 inhibition resulted in aberrant polyadenylation and downregulation of long genes (includingBRCA1andBRCA2) but such effects were modest or absent in tumors adapted to chronicCDK12BAL. One key exception wasATM, which did retain transcript shortening and reduced protein expression in the adaptedCDK12BALmodels. However,CDK12BALcells demonstrated intact HR as measured by RAD51 foci formation following irradiation.CDK12BALcells showed a vulnerability to targeting of CDK13 by sgRNA or CDK12/13 inhibitors andin vivotreatment of prostate cancer xenograft lines showed that tumors withCDK12BALresponded to the CDK12/13 inhibitor SR4835, while CDK12-intact lines did not. Collectively, these studies show that aberrant polyadenylation and long HR gene downregulation is primarily a consequence of acute CDK12 deficiency, which is largely compensated for in cells that have adapted to CDK12 loss. These results provide an explanation for why PARPi monotherapy has thus far failed to consistently benefit patients with CDK12 alterations, though alternate therapies that target CDK13 or transcription are candidates for future research and testing.

PDX sequencing metrics Phenotype and sequencing metrics for all PDX LuCaP lines.

Supplementary Figures S1-S18 and Supplementary Table legends.

PTM peak data and phenotype 47 fragment variability Sheet 1: PDX sample representation in 3 histone PTM CUT&RUN nucleosome profiling assays (H3K4me1, H2K27ac and H3K27me3). Sheet 2: Log2 fold-change, p-value, and q-value between ARPC and NEPC lines for coefficient of variation in the 47 phenotype defining gene bodies (two tailed Mann-Whitney U test, Benjamini-Hochberg adjusted). Sheet 3: Log2 fold-change, p-value, and q-value between ARPC and NEPC lines for coefficient of variation in the 47 phenotype defining gene promoters (two tailed Mann-Whitney U test, Benjamini-Hochberg adjusted).

Feature-region combination AUCs and benchmarking Sheet 1: Log2 fold-change, p-value, and q-value between ARPC and NEPC lines for central mean coverage in all queried (338) TFs (two tailed Mann-Whitney U test, Benjamini-Hochberg adjusted). Sheet 2: 100-fold cross-validation AUCs for all region and feature combinations subset by the ‘AR10’ overlapping features (see Methods). Sheet 3: 100-fold cross-validation AUCs for all region and feature combinations subset by the ‘Phenotype 47-defining’ overlapping features. Sheet 4: 100-fold cross-validation AUCs for all region and feature combinations (global). Sheet 5: Predictions scores, tumor fraction, depth of coverage, and subtype for benchmarking admixtures. Sheet 6: AUCs for unsupervised prediction of admixture subtypes grouped by tumor fraction and depth. Sheet 7: NEPC:ARPC ratio, tumor fraction, and ARPC and NEPC fraction predictions for mixed phenotype admixtures using Keraon (see Methods).

A cornerstone of research to improve cancer outcomes involves studies of model systems to identify causal drivers of oncogenesis, understand mechanisms leading to metastases, and develop new therapeutics. Although most cancer types are represented by large cell line panels that reflect diverse neoplastic genotypes and phenotypes found in patients, prostate cancer is notable for a very limited repertoire of models that recapitulate the pathobiology of human disease. Of these, the lymph node carcinoma of the prostate (LNCaP) cell line has served as the major resource for basic and translational studies. Here, we delineated the molecular composition of LNCaP and multiple substrains through analyses of whole-genome sequences, transcriptomes, chromatin structure, androgen receptor (AR) cistromes, and functional studies. Our results determined that LNCaP exhibits substantial subclonal diversity, ongoing genomic instability, and phenotype plasticity. Several oncogenic features were consistently present across strains, but others were unexpectedly variable, such as ETV1 expression, Y chromosome loss, a reliance on WNT and glucocorticoid receptor activity, and distinct AR alterations maintaining AR pathway activation. These results document the inherent molecular heterogeneity and ongoing genomic instability that drive diverse prostate cancer phenotypes and provide a foundation for the accurate interpretation and reproduction of research findings.

Metastatic prostate cancer (mPC) is a clinically and molecularly heterogeneous disease. While there is increasing recognition of diverse tumor phenotypes across patients, less is known about the molecular and phenotypic heterogeneity present within an individual. In this study, we aimed to define the patterns, extent, and consequences of inter- and intratumoral heterogeneity in lethal prostate cancer. By combining and integrating in situ tissue-based and sequencing approaches, we analyzed over 630 tumor samples from 52 patients with mPC. Our efforts revealed phenotypic heterogeneity at the patient, metastasis, and cellular levels. We observed that intrapatient intertumoral molecular subtype heterogeneity was common in mPC and showed associations with genomic and clinical features. Additionally, cellular proliferation rates varied within a given patient across molecular subtypes and anatomic sites. Single-cell sequencing studies revealed features of morphologically and molecularly divergent tumor cell populations within a single metastatic site. These data provide a deeper insight into the complex patterns of tumoral heterogeneity in mPC with implications for clinical management and the future development of diagnostic and therapeutic approaches.

Abstract Study Purpose: Determine the role of the transcription factor TEAD1 in promoting the neuroendocrine phenotype in castration-resistant prostate cancer (CRPC). Experimental Procedures: We quantitated YAP-pathway associated transcripts, protein abundance, and splicing events in androgen receptor (AR) positive prostate adenocarcinoma (ARPC) and small cell or neuroendocrine prostate cancer (SCNPC) using RNAseq, scRNAseq, qPCR, immunohistochemistry (IHC), CpG methylation, and western blot analyses on patient samples, patient-derived xenograft (PDX) models, and cell lines. In addition, we tested the impact of the pharmacological inhibition and knockdown of proteins and genes associated with TEAD1 activity in ARPC and SCNPC cell lines on TEAD1 expression, cell number, and tumor cell differentiation. Results: Transcriptomic analyses revealed a decrease in YAP1, TAZ, LATS2, and TEAD2 and an increase in LATS1, TEAD1 and RBFOX2 transcript levels in SCNPCs compared to ARPCs in SU2C, rapid autopsy metastases, and PDX models. Similar results were observed in ARPC and SCNPC cancer cell lines, in the transplantable LTL331 prostate cancer transdifferentiation model, and scRNAseq of liver CRPC metastases from patients. The decrease in YAP1, TEAD2 and LATS2 expression correlates with increased methylation and the increase in TEAD1 correlates with decreased methylation in SCNPC PDX models, suggesting epigenetic control. The changes in YAP and YAP-associated proteins (most importantly TEAD1) at the epigenetic and transcriptional level were confirmed by IHC analysis comparing ARPC to SCNPC PDX models. Of the four TEAD proteins, TEAD1 expression was increased, with a concomitant decrease in TEAD2 and TEAD3, whereas low levels of TEAD4 transcript are maintained in the SCNPC phenotype. Notably, RBFOX2 was increased in the SCNPC datasets. RBFOX2 is a pre-RNA splicing regulator that promotes the inclusion of exon 6 in TEAD1 mRNAs associated with increased TEAD1 activity. The RBFOX2 spliced TEAD1 was observed in the SCNPC PDX models, while not detected in the ARPC PDX models. Conclusions: The conversion of ARPC to SCNPC involves the loss of transcriptional regulators: AR, YAP, and REST inactivation in SCNPC. It also involves the expression of ASCL1 or NEUROD1, well described master regulators that are implicated in the conversion to and maintenance of the SCNPC phenotype. Our work suggests TEAD1 is another transcriptional regulator in SCNPC that warrants further investigation. Citation Format: Lisha G. Brown, Ilsa M. Coleman, Tony L. Chu, Daniel W. Lin, John K. Lee, Erolcan Sayar, Lawrence D. True, Ruth Dumpit, Eva Corey, Peter S. Nelson, Michael C. Haffner, Colm Morrissey. Uncovering the role of TEADs in treatment induced small cell or neuroendocrine prostate cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2754.