Neuroendocrine Differentiation in Prostate Cancer Requires ASCL1

Abstract

Most patients with prostate adenocarcinoma develop resistance to therapies targeting the androgen receptor (AR). Consequently, a portion of these patients develop AR-independent neuroendocrine (NE) prostate cancer (NEPC), a rapidly progressing cancer with limited therapies and poor survival outcomes. Current research to understand the progression to NEPC suggests a model of lineage plasticity whereby AR-dependent luminal-like tumors progress toward an AR-independent NEPC state. Genetic analysis of human NEPC identified frequent loss of RB1 and TP53, and the loss of both genes in experimental models mediates the transition to a NE lineage. Transcriptomics studies have shown that lineage transcription factors ASCL1 and NEUROD1 are present in NEPC. In this study, we modeled the progression of prostate adenocarcinoma to NEPC by establishing prostate organoids and subsequently generating subcutaneous allograft tumors from genetically engineered mouse models harboring Cre-induced loss of Rb1 and Trp53 with Myc overexpression (RPM). These tumors were heterogeneous and displayed adenocarcinoma, squamous, and NE features. ASCL1 and NEUROD1 were expressed within NE-defined regions, with ASCL1 being predominant. Genetic loss of Ascl1 in this model did not decrease tumor incidence, growth, or metastasis; however, there was a notable decrease in NE identity and an increase in basal-like identity. This study provides an in vivo model to study progression to NEPC and establishes the requirement for ASCL1 in driving NE differentiation in prostate cancer. Significance: Modeling lineage transitions in prostate cancer and testing dependencies of lineage transcription factors have therapeutic implications, given the emergence of treatment-resistant, aggressive forms of neuroendocrine prostate cancer. See related commentary by McQuillen and Brady, p. 3499.

Figure 1.. Prostate organoid allografts with Myc overexpression and Rb1 and Trp53 KO generate heterogeneous tumors with regions of NE differentiation.

(A) Diagram of the prostate organoid allograft procedure (created in BioRender.com). Prostates from RPM (Rb1^fl/flTrp53^fl/flMycT58A^LSL/LSL) or RPR2 (Rb1^fl/flTrp53^fl/flRbl2^fl/fl) mice were isolated, grown as organoids in vitro, and infected with Lentivirus expressing Cre-recombinase to delete the floxed alelles. Genotype confirmed organoids were injected subcutaneously into the flanks of male NSG mice to test tumor formation activity. (B) Heat map of bulk RNA-seq from four independently derived RPM in vitro organoids after Cre recombination. Expression of NE, luminal, and basal cell markers is shown and measured by counts per million (CPM). (C) Efficiency of generating tumors from each organoid (five organoids for each genotype with 8 to 28 mice per organoid; mean with SEM, unpaired t-test), and (D) tumor growth in mice with RPM or RPR2 allografts (mean with SEM; unpaired t-test). (E) Percentage of RPM and RPR2 tumors expressing NE markers (INSM1 or SYP) (36% for RPM, 6% for RPR2). (F) Violin plots representing the percentage tumor area positive for each cell identity marker in each RPM or RPR2 tumor (median; unpaired t-tests). (G) Representative images of IHC with cell-type specific markers in RPM and RPR2 allograft tumors. Note, two fields for each tumor are shown to illustrate the heterogeneity seen in these tumors. n = the number of mice enrolled (C) or the number of tumors analyzed (D,E). ns=p>0.05; *p≤0.05; **p≤0.01; White scale bar=70μm. Supporting RNA-seq data is in Supp Table S1.

Figure 2.. RPM prostate organoid allografts contain NE regions expressing ASCL1 and NEUROD1.

(A) H&E and IHC for INSM1, ASCL1, NEUROD1 and AR of an RPM tumor containing NE marker expression (top) versus a tumor not expressing NE markers (bottom). (B) Violin plot of the quantification of IHC for ASCL1 or NEUROD1+ cells in 10 NE containing RPM tumors (median shown). (C) Schematic of the Ascl1^TdTom allele utilized to mark ASCL1+ cells (created in BioRender.com). The Ascl1 coding region is intact but a tandem TdTom encoding sequence separated by P2A sequences was knocked in (Lin et al. 2017). Prostate organoids from RPM with one allele of Ascl1^TdTom were made and used for allograft tumor generation. (D) IF images of a RPM^TdTom allograft tumor. Blue is DAPI. Essentially complete colocalization of ASCL1 and TdTom supports the use of TdTom as a proxy for ASCL1. (E) IF colocalization of INSM1 with either TdTom (ASCL1) or NEUROD1 (white arrows), or both factors (yellow arrow). Some INSM1 cells do not express either transcription factor (orange arrows). (F) Percentage of INSM1+ cells co-expressing TdTom (ASCL1), NEUROD1, both or neither from IF in (E) (mean with SEM). (G) IHC for ASCL1 and CK8 of an RPM tumor in an area with luminal cell histology. (H) Colocalization of TdTom (ASCL1) with other cell identity markers in a RPM^TdTom tumor. ASCL1 is co-expressed with CK8 (yellow arrows) but not P63 or VIM. ASCL1 is expressed in a subset of EPCAM and SOX2+ cells. Black scale bar=50μm. White scale bar=20μm

Figure 3.. ASCL1+ cells can transition to NEUROD1+ cells in RPM tumors.

(A) Diagram of the procedure to establish allograft tumors from ASCL1- or ASCL1+ RPM tumor cells (created in BioRender.com). (B) Tumor weights from allografts seeded with ASCL1- (TdTom-) or ASCL1+ (TdTom+) cells were comparable (mean with SEM; unpaired t-test). (C) Percent area of ASCL1 or NEUROD1 by IHC in tumors arising from TdTom- or TdTom+ cells (mean with SEM; unpaired t-tests). (D) H&E and IHC for INSM1, SYP, ASCL1, NEUROD1, and AR from allograft tumors arising from TdTom- (two different fields of view) or TdTom+ cells. (E) Violin plot representing the percentage area of each cell identity marker in tumors from TdTom- or TdTom+ cells (median; unpaired t-test). (F) Percentage of tumors expressing NEUROD1 before sorting and after sorting into TdTom- and TdTom+. (G) Percentage of INSM1+ cells co-expressing TdTom (ASCL1), NEUROD1, both, or neither in tumors from TdTom- or TdTom+ cells (mean with SEM; unpaired t-tests). (H) Colocalization of TdTom (ASCL1) with NEUROD1 (white arrows). Some INSM1 cells do not express either transcription factor (orange arrows). (I) H&E of tumors from TdTom- or TdTom+ cells. ns=p>0.05. Black scale bar=50μm. White scale bar=20μm.

Figure 4.. ASCL1 loss prevents efficient NE marker expression in RPM allograft tumors.

(A) Efficiency of generating tumors, and (B) tumor growth in mice with RPM or RPMA organoid allografts harvested between 7–9 weeks after injection. Data include tumors from 5 independently derived RPM (same as shown in Fig. 1C, D for comparison) and 2 RPMA organoids (mean with SEM; unpaired t-test). (C) IHC with cell-type specific markers in an RPMA tumor expressing NE markers (INSM1) (left) and a typical RPMA tumor with no NE marker detected (right). (D) Violin plot representing the percentage area of each cell identity marker in each RPM or RPMA tumor (median; unpaired t-test). (E) Percentage of RPM and RPMA tumors expressing CK8, P63, and NE markers (INSM1 or SYP). (F) Heat map of bulk RNA-seq from 2 independently derived RPM (from Fig. 1B for comparison) or RPMA in vitro organoids after Cre recombination. Expression of NE, luminal, and basal cell markers is shown and measured by counts per million (CPM). (G) Representative images of IHC with ASCL1, NEUROD1, and POU2F3 in an INSM1+ region as shown in 4C of an RPMA tumor and positive controls for the staining. n = the number of mice enrolled (A) or the number of tumors analyzed (B,E). ns=p>0.05. Black scale bar=50μm. Supporting RNA-seq data is in Supp Table S1.

Figure 5.. snRNA-seq confirms the presence of luminal, basal, and NE features in the RPM allograft tumors.

(A) Combined UMAP from snRNA-seq of RPM tumors (n=3) and RPMA tumors (n=3). Cluster 4 was removed due to high mitochondrial gene content. Cluster identity was assigned based on markers as shown in (B). (B) Dot plot with expression levels of differentially expressed genes in each cluster color coded for cell type classification. (C) Reclustering of tumor cell populations from (A). (D) Dot plot with expression levels of differentially expressed genes in each cluster color coded for cell type classification from the reclustered tumor populations (C). (E) UMAP showing the distribution of specific luminal, basal, and NE markers as indicated. For cluster markers see Supplementary Tables S2 and S3.

Figure 6.. RPM tumors with Ascl1 deleted are enriched with basal markers and depleted of neuroendocrine markers.

(A) Dot plot showing expression levels of different cell identity markers between RPM and RPMA tumors from the reclustered data in Fig. 5C. RPM, blue; RPMA, red. (B) UMAP separating the RPM (n=3) and RPMA (n=3) nuclei. (C) Proportion of nuclei in each cluster in RPM versus RPMA tumors showing underrepresentation of RPMA in NE defined clusters and overrepresentation in clusters with basal cell-like identity. (D) UMAPs showing Ascl1 and Ar expression, (E) basal cell signature enrichment and NE cell signature depletion using the indicated markers comparing RPM and RPMA genotypes. (F) UMAPs showing the differences in the distribution of CRPC (NE) vs CRPC (Adeno) gene signatures (from Beltran et al. 2016) between RPM and RPMA tumors. For cluster markers separated by RPM versus RPMA see Supplementary Tables S4 and S5. For differentially expressed genes between RPM and RPMA by cluster see Supplementary Table S6.