Increased Serine and One Carbon Pathway Metabolism by PKCλ/ι Deficiency Promotes Neuroendocrine Prostate Cancer

SUMMARY

Increasingly effective therapies targeting the androgen receptor have paradoxically promoted the incidence of neuroendocrine prostate cancer (NEPC), the most lethal subtype of castration-resistant prostate cancer (PCa), for which there is no effective therapy. Here we report that protein kinase C (PKC)λ/ι is downregulated in de novo and during therapy-induced NEPC, which results in the upregulation of serine biosynthesis through an mTORC1/ATF4-driven pathway. This metabolic reprogramming increases intracellular SAM levels to support cell proliferation and epigenetic changes that favor the development of NEPC characteristics. Altogether, we have uncovered a metabolic vulnerability triggered by PKCλ/ι deficiency in NEPC, which offers potentially actionable targets to prevent therapy resistance in PCa.

Graphical Abstract

IN BRIEF

Reina-Campos et al. find reduced PKCλ/ι in de novo and therapy-induced neuroendocrine prostate cancer (NEPC). PKCλ/ι loss upregulates mTORC1/ATF4 to promote serine biosynthesis, resulting in increased S-adenosyl methionine that supports cell proliferation and epigenetic changes favoring the development of NEPC.

INTRODUCTION

Acquired resistance to targeted therapies in cancer is a rising unmet clinical need. Therapeutic agents exert a natural selection and promote the acquisition of new tumor traits. Cellular plasticity endows cancer cells with the ability to overcome therapy by switching their cellular state. At the same time, metabolic dependencies arise in response to alterations in oncogenic growth signaling and might constitute an Achilles heel that can be exploited therapeutically. Therapy resistance is a major clinical problem in the treatment of prostate cancer (PCa) (Mu et al., 2017), in which although androgen deprivation therapy (ADT) has proven effective for its early management (Watson et al., 2015), tumors become resistant to ADT leading to a lethal outcome known as castration resistant prostate cancer (CRPC). As a result, increasingly potent second-generation AR-targeted therapies, such as enzalutamide and abiraterone, have been successfully implemented as a second line of therapy (de Bono et al., 2011; Scher et al., 2012). However, response to these agents is also limited and nearly all treated individuals will ultimately progress and develop resistance. One major form of tumor relapse involves the reactivation of the AR axis through several different mechanisms (Watson et al., 2015). Alternatively, disease progression may arise in the absence of a functional AR, as an extremely aggressive, highly proliferative, and metastatic PCa variant, termed neuroendocrine prostate cancer (NEPC) (Aggarwal et al., 2018; Beltran et al., 2016). More potent and sustained AR targeting has driven an increased incidence of NEPC, which has histological features of small cell carcinoma and neuroendocrine differentiation. Thus, understanding the molecular mechanisms that tumor cells exploit to undergo NEPC differentiation is a pressing unmet clinical need.

Despite the higher incidence of certain genomic alterations, such as RB1 and TP53 loss and amplification of MYCN and AURKA in NEPC compared to prostate adenocarcinoma, NEPC consists of a heterogeneous and complex spectrum of tumors. Lineage plasticity manifests with a markedly different epigenetic profile, low AR signaling, and acquired expression of neuroendocrine, neuronal, developmental, and stem cell markers (Rickman et al., 2017). Due to the heterogeneity of NEPC and because it is still unclear how these different players cooperate to control NEPC differentiation, it is conceivable that targeting individual signaling molecules might not lead to effective therapies. On the other hand, the identification of metabolic dependencies and other non-oncogenic addictions might reveal vulnerabilities that can open potentially therapeutic avenues for the treatment of this type of highly aggressive and treatment-resistant cancer. We currently lack the mechanistic insight as to which metabolic pathways play a role in cellular plasticity, adaptation, treatment resistance, and survival in NEPC.

Our initial unpublished bioinformatics analyses of human PCa data sets suggested that reduced levels of the atypical protein kinase C, PKCλ/ι (encoded by PRKCI), could be a determinant event in the mechanisms leading to NEPC differentiation. PKCλ/ι, together with PKCζ, constitute the atypical PKC (aPKC) family (Diaz-Meco and Moscat, 2012). Mounting evidence strongly suggests that these kinases play decisive roles in tumorigenesis (Diaz-Meco and Moscat, 2012). PKCζ has been demonstrated to play a tumor suppressor role in colorectal tumors through metabolic reprogramming (Ma et al., 2013), and in other types of cancer as well (Galvez et al., 2009; Kim et al., 2013). Consistently, a recent analysis of cancer-associated mutations in all PKC isoforms from different types of neoplasias revealed that these kinases are generally inactivated in cancer, supporting their potential role as tumor suppressors (Antal et al., 2015). However, the role of PKCλ/ι has been proposed to be an exception to this general paradigm, and its function has been suggested to be cancer-specific. Thus, although PKCλ/ι loss of function mutations have been identified (Antal et al., 2015), PRKCI is located in the 3q amplicon, which supports its potential oncogenic role in cancers with this genomic amplification, such as lung squamous cell carcinoma (Justilien et al., 2014). In contrast, recently published data demonstrate that PKCλ/ι, similarly to PKCζ, is a tumor suppressor in other cancers, like in intestine in which PKCλ/ι levels are reduced and its genetic inactivation results in enhanced tumorigenesis (Nakanishi et al., 2018; Nakanishi et al., 2016). This suggests that PKCλ/ι can be pro-oncogenic or tumor suppressive, depending on the cancer context. However, its potential role in other cancers remains unexplored.

Here we have rigorously tested the hypothesis that loss of PKCλ/ι plays a key role in driving NEPC progression to unveil potential metabolic non-oncogenic vulnerabilities that could be susceptible of therapeutic intervention during NEPC progression.

RESULTS

PKCλ/ι expression is downregulated in NEPC

Bioinformatics analysis of 60 PCa datasets (Oncomine) revealed that while PRKCI expression was upregulated in primary tumors compared to normal tissue, it was downregulated in metastases (Figure 1A). In agreement with the downregulation of PKCλ/ι in aggressive disease, a molecular concept map analysis using signatures of PRKCI-correlated or PRKCI-anticorrelated genes derived from the MSKCC 2010 PCa dataset (Taylor et al., 2010), demonstrated a negative association between PRKCI-correlated genes and metastasis, advanced cancer stage, and poor clinical outcome (Figure 1B). Patients with low levels of PRKCI expression had much lower relapse-free survival than patients with high PRKCI expression (Figure 1C). These results suggest that PKCλ/ι might be playing an unanticipated role as a tumor suppressor in late-stage PCa. To assess whether the loss of PKCλ/ι could be attributed to a defined molecular subset, we queried a cohort of metastatic CRPC tumors (Robinson et al., 2015). Patients with low PKCλ/ι expression were significantly exclusive from the group with activating alterations in the AR gene (Figure S1A). Correlation analysis showed that low PKCλ/ι expression was associated with diminished AR activity (Figure S1B). When evaluating datasets of NEPC, we found that PRKCI expression was significantly downregulated in NEPC compared to prostate adenocarcinoma CRPC samples (Figures 1D and 1E). No genomic amplifications or deletions of PRKCI were found in NEPC samples (Figure S1C). Molecular classification of primary and metastatic PCa samples (Taylor et al., 2010) revealed that PKCλ/ι downregulation was associated with the NEPC phenotype, independently of being primary or metastatic (Figure 1E). PKCλ/ι expression was found highly correlated with genes that are differentially expressed in NEPC (Figure 1F). These included the negative correlation with NEPC markers such as SYP, ENO2, NCAM1, and CHGA, and the positive correlation with the AR downstream target KLK3 (PSA), as well as the NEPC repressor REST (Figure 1G).

Figure 1. PKCλ/ι Expression is Downregulated in NEPC.

(A) PRKCI mRNA levels in PCa datasets separated by Cancer vs. Normal (upper graph) (Tomlins^$ denotes Benign Prostatic Hyperplasia Epithelia vs. Normal) and Metastasis vs. Primary cancer (lower graph) (Oncomine).

(B) Odds Ratio of the overlap between the gene signatures “PRKCI correlated” and “PRKCI anticorrelated”, and clinical subgroups generated from the specified PCa datasets. Taylor: recurrence at 3 years (^$) and 5 years (^$ $). Taylor 3^: recurrence at 1 (⁺), 3 (⁺⁺) and 5 years (⁺⁺⁺). (Oncomine). Cancer (C), Normal (N).

(C) Recurrence free survival (RFS) of patients stratified by PRKCI mRNA expression in a published microarray gene expression dataset (GSE21034). Log-rank (Mantel-Cox) test.

(D) PRKCI mRNA levels in Adenocarcinoma (Adeno) and NEPC samples.

(E) PRKCI mRNA levels in GSE21034 samples classified. Adeno: Adenocarcinoma; MET: metastasis.

(F) GSEA of NEPC signatures (Beltran et al 2011) in PRKCI correlated genes in GSE21034.

(G) Pearson correlation analysis of PRKCI and NEPC markers in GSE21034.

(H) Representative staining of PKCλ/ι and NCAM1, quantification (normalized values presented as mean fluorescence intensity (MFI)), and H&E, in a PCa cohort containing Adenocarcinoma (Adeno; n = 7) and NEPC (n = 14). Scale bars, 100 μm.

(I) Western blot of PKCλ/ι and Actin loading control in the indicated cell lines (n = 3).

(J) FACS analysis of CD44 or NCAM1 in LNCaP or LNCaP-ER cells (n = 3).

(K) qPCR analysis of indicated genes in sorted CD44^high or NCAM1^high cells in arbitrary units (AU) and western blot of NCAM1 and PKCλ/ι in LNCaP or LNCaP-ER cells (n = 2).

(L) Staining of PKCλ/ι and NCAM1 and quantification (normalized values) in an Adenocarcinoma (Adeno) with NEPC differentiation sample. Scale bars, 100 μm.

(M) Gene expression levels of PRKCI, NCAM1 and KLK3 in a published RNA-seq gene expression dataset (GSE70380). V1_met: rib metastasis obtained during first visit, before treatment. V2_met: metastasis obtained during second visit after enzalutamide (enza) treatment for 12 weeks.

Mean ± SEM (H, L, K). Two-tailed t-test (D, E, H, L, K). In (D) and (E), box and whiskers graphs indicate the median and the 25th and 75th percentiles, with minimum and maximum values at the extremes of the whiskers. *p < 0.05, **p < 0.01, ***p < 0.001. See also Figure S1.

Reduced protein levels of PKCλ/ι were observed in a cohort of hormone naive NEPC with a de novo presentation, as compared to primary adenocarcinomas with high Gleason score (Figure 1H). PKCλ/ι expression was severely decreased in NCAM1-positive prostate tumor areas, as well as in liver and lung NEPC metastases in TRAMP (transgenic adenocarcinoma mouse prostate) mice, a mouse model of NEPC (Figure S1D). Analysis of gene expression data from the NEPC mouse model of prostate-specific simultaneous deletion of Pten and Rb1, or Pten, Rb1, and Trp53 (Ku et al., 2017) showed reduced Prkci expression in NEPC tumors (Figure S1E). Consistently, PKCλ/ι protein levels were decreased in two different human NEPC cells lines (NE-1–3 and H660) (Figure 1I). These results suggest that NEPC tumors display low levels of PKCλ/ι expression.

To test whether PKCλ/ι loss could occur in therapy-related NEPC differentiation, we generated enzalutamide resistant LNCaP (LNCaP-ER) cells. LNCaP-ER cells had increased levels of NEPC markers CD44 and NCAM1 (Figure 1J). qPCR analysis of FACS-sorted CD44^high and NCAM1^high cell populations from LNCaP-ER cultures showed additional hallmarks of NEPC concomitant with reduced PKCλ/ι mRNA and protein levels (Figure 1K). Analysis of a tumor sample from a treated patient with high-burden disease, containing mixed prostate adenocarcinoma with extensive areas of NEPC differentiation, demonstrated a significant loss of PKCλ/ι expression in the NEPC areas as compared to the adenocarcinoma foci (Figure 1L). Interrogation of a dataset from a patient progressing on ADT comparing before and after enzalutamide treatment (Kohli et al., 2015) showed lower PRKCI mRNA levels in metastasis after treatment, which correlated with markers of NEPC differentiation and less AR activity (Figure 1M). These data support the notion that PKCλ/ι is reduced both in de novo NEPC and during treatment-related NEPC differentiation.

Prostate-Specific Deletion of PTEN and PKCλ/ι Promotes Basal/NEPC Tumorigenesis In Vivo

Next, we crossed the Pten^f/f-PbCre4⁺ mouse line (PTEN KO) in which Pten is specifically deleted in the prostate epithelium, with Prkci^f/f mice to generate Pten^f/f-Prkci^f/f-PbCre4⁺ mice (referred as DKO). DKO mice had a much shorter lifespan than PTEN KO mice with a median overall survival of 180 days (Figure 2A). DKO prostates had enlarged dorsolateral and ventral prostate lobes, while age-matched PTEN KO mice only showed mild prostate enlargement (Figure 2B). Histopathology of prostate sections revealed that tumor lesions from the DKO mice included more advanced prostate intraepithelial neoplasias (PIN) and increased incidence of carcinoma, with comedo-type necrosis and areas of invasion into the adjacent stroma (Figure 2C). DKO prostates showed increased stromal reaction and fibrosis, as determined by Masson’s trichrome and αSMA staining (Figures 2D and S2A). DKO tumors displayed a marked increase in Ki67⁺ cells per gland, with proliferative cells mostly accumulated in the outer regions where basal cells are usually located (Figures 2E and 2F). DKO prostates also showed higher abundance of TP63⁺ cells that overlapped with focal areas of high Ki67 staining (Figures 2E and 2F). Approximately two thirds of the examined DKO prostates presented prostatic urothelial carcinomas (Figures 2C and 2G). These results suggested that PKCλ/ι loss in a PTEN KO background promoted an increase in highly proliferative basal-like cells. Consistently, DKO PIN showed increased cytokeratin 5 (CK5), decreased cytokeratin 8 (CK8) and lower nuclear AR staining (Figures 2H and S2B), which is in keeping with the emergence of basal cell identity and the loss of luminal features. More advanced tumor areas, including prostatic urothelial carcinomas, showed marked CK5 upregulation with high Ki67⁺ and TP63 staining, loss of nuclear AR, and areas positive for the NE marker chromogranin A (CHGA; Figure 2I). FACS analysis showed an increase in NCAM1 staining in DKO basal cells (Figure 2J).

Figure 2. Prostate-Specific Deletion of PTEN and PKCλ/ι Promotes Basal/NEPC Tumorigenesis In Vivo.

(A) Overall survival of PTEN KO (n = 9) and DKO mice (n = 12). Log-Rank (Mantel-Cox) test.

(B) Photographs of PTEN KO and DKO genitourinary organs. Scale bar, 5 mm.

(C) Frequency of prostatic lesions and representative H&E of PTEN KO (n = 8) and DKO (n = 9) prostates. Invasive: Area of invasion into the stroma. Necrosis: area of comedo-type necrosis. mPIN1/2: low grade mouse intraepithelial neoplasia; mPIN3/4: high grade mouse intraepithelial neoplasia. Scale bars, 100 μm. Chi-square test.

(D) Masson’s Trichrome staining of PTEN KO (n = 5) and DKO prostates (n = 7). Scale bar, 100 μm.

(E) Ki67 and TP63 IHC of PTEN KO (n = 5) and DKO (n = 5) prostates. Scale bars, 200 μm.

(F) Quantification of Ki67 and TP63 IHC from panel E.

(G) Representative H&E of DKO prostate. Yellow line, prostatic urothelial carcinoma. Scale bars, 200 μm

(H) CK5, CK8 and AR IHC of PTEN KO and DKO (n = 4) prostates. Arrows denote nuclei with lower AR staining. Scale bars, 30 μm.

(I) CK5, CHGA, TP63, AR and Ki67 IHC of DKO prostate with magnification of prostatic urothelial carcinoma. Scale bar, 50 μm.

(J) FACS analysis of NCAM1 of PTEN KO and DKO (n = 3) prostate basal cells.

(K) Organoid growth of PTEN KO and DKO prostate-derived organoids (n = 3). Two-way ANOVA.

(L) qPCR of indicated genes in PTEN KO and DKO-derived prostate organoids (n = 3–5).

(M) Average organoid size of prostate cancer-derived organoids from PTEN KO and DKO mice treated with enzalutamide (enza, 10 μM) for 4 days (n = 3).

Mean ± SEM (F, J, K-M). Two-tailed t-test (F, J, L, M). *p < 0.05, **p < 0.01, ***p < 0.001. See also Figure S2.

DKO-derived prostate organoids had increased growth (Figure 2K) and higher expression of NEPC and basal markers (Figure 2L). Similar results were obtained in PTEN KO organoids upon lentiviral shRNA of PKCλ/ι (Figure S2C). DKO-derived prostate organoids were resistant to enzalutamide treatment as compared to PTEN KO organoids that showed a significant reduction in size upon treatment (Figure 2M). These results demonstrate that the concomitant deletion of PTEN and PKCλ/ι in the mouse prostate epithelium accelerates PCa progression and leads to the emergence of highly proliferative lesions with loss of luminal cell identity and gain of basal and NEPC features.

Loss of PKCλ/ι is Sufficient to Promote NEPC Differentiation at a Cellular Level

Lentiviral-mediated shRNA of PKCλ/ι in LNCaP cells was sufficient to induce NEPC and basal cell markers concomitant with the downregulation of AR and of AR targets (Figures 3A and 3B). Similar gene expression changes were observed upon CRISPR/Cas9-mediated inactivation of PKCλ/ι (sgPKCλ/ι) in C42B cells (Figures 3C and 3D). Re-expression of PKCλ/ι into sgPKCλ/ι cells (sgPKCλ/ι-R) restored these markers to the levels of control cells (sgC) (Figures 3C and 3D). We next stably knocked down (KD) PKCλ/ι in the immortalized non-transformed prostate epithelial cell line (PrEC). GSEA of genes differentially expressed in shPKCλ/ι cells showed similar transcriptomic profiles to human NEPC (Beltran et al., 2011) (Figures 3E and S3A). NEPC is also characterized by a remarkable increase in cellular proliferation, which is mediated by the activation of E2F (Figure S3B, left panel). Interestingly, PRKCI expression was negatively correlated with the hallmark “E2F Targets”, and positively correlated with “Androgen Response” gene set (Figure S3B, right panel). LNCaP shPKCλ/ι cells showed increased in vitro proliferation in androgen-deficient conditions and were able to grow even in the presence of enzalutamide treatment whereas shNT LNCaP cells were not (Figure 3F). LNCaP shPKCλ/ι cells showed increased expression of cell cycle regulatory genes and increased colony formation activity (Figures 3G and S3C). KD of PKCλ/ι in PC3 and DU145, two androgen-independent cell lines, enhanced proliferation concomitant with higher expression of NEPC markers (Figures S3D–S3G). Conversely, overexpression of PKCλ/ι in PC3 or DU145 strongly decreased cell proliferation, colony formation and SYP expression (Figures 3H and S3H–S3J). sgPKCλ/ι C42B cells showed higher proliferation under androgen-deficient condition and had increased basal levels of E2F1 (Figures 3I and 3J). sgPKCλ/ι cells displayed also a higher proliferative capacity in vivo in tumor xenografts than sgC cells even in the presence of enzalutamide treatment (Figures 3K and S3K). Tumor growth was rescued upon re-expression of PKCλ/ι in sgPKCλ/ι cells (Figure 3K). These results demonstrate that the loss of PKCλ/ι is sufficient to promote NEPC differentiation in a cell-autonomous manner in vitro and in vivo.

Figure 3. Loss of PKCλ/ι is Sufficient to Promote NEPC Differentiation at a Cellular Level.

(A) Western blot of indicated proteins in shNT and shPKCλ/ι LNCaP cells.

(B) qPCR of indicated NEPC-, Basal-, and AR-related genes in shNT and shPKCλ/ι LNCaP cells (n = 3–4).

(C) Western blot of indicated proteins in sgC, sgPKCλ/ι and sgPKCλ/ι-R C42B cells (n = 3).

(D) qPCR of indicated genes in sgC, sgPKCλ/ι and sgPKCλ/ι-R C42B cells (n = 3).

(E) GSEA of NEPC signatures (Beltran et al 2011) in microarray data of shNT (NT) and shPKCλ/ι PrEC cells.

(F) Cell proliferation of shNT and shPKCλ/ι LNCaP cells (n = 3) under androgen deprivation (ADT) with or without enzalutamide (enza, 10 μM). Western blot of PKCλ/ι.

(G) qPCR of indicated genes in shNT and shPKCλ/ι LNCaP cells (n = 3–6).

(H) Cell proliferation in ADT (n = 3) and western blot of PC3 cells expressing FLAG or FLAG-PKCλ/ι.

(I) Cell proliferation of sgC and sgPKCλ/ι C42B cells in ADT (n = 3).

(J) qPCR analysis of E2F1 mRNA levels in sgC and sgPKCλ/ι C42B cells (n = 4).

(K) Tumor growth of xenografts of sgC, sgPKCλ/ι and sgPKCλ/ι-R C42B cells (n = 4–12).

Mean ± SEM (B, D, F, G, H-K). Two-tailed t-test (B, D, G, J). Two-way ANOVA (F, H, I, K). *p < 0.05, **p < 0.01, ***p < 0.001. See also Figure S3.

PKCλ/ι Regulates mTORC1 Activity through LAMTOR2 Phosphorylation

Next, we carried out transcriptomic profiling of shNT and shPKCλ/ι PrEC cells as well as of sgC, sgPKCλ/ι, and sgPKCλ/ι-R C42B cells. GSEA revealed “mTORC1 Signaling”, “MYC targets” and cell-cycle related signatures to be upregulated in PKCλ/ι deficiency conditions, which was reverted in sgPKCλ/ι-R cells (Figures 4A and 4B). Investigate Gene Set analysis (Broad Institute) with genes that were differentially expressed in sgPKCλ/ι as compared to sgC and rescued in sgPKCλ/ι-R identified gene sets corresponding to mTORC1 and AR signaling as the most upregulated and downregulated, respectively (Figures S4A and S4B). mTORC1 was activated in sgPKCλ/ι cells as determined by western-blot of three downstream effectors, p4EBP1, pS6K and cMYC (Figure 4C). The kinase activity of PKCλ/ι was required to inhibit mTORC1 since the WT form of PKCλ/ι, but not its kinase inactive mutant (K274W), was able to rescue mTORC1 activation in sgPKCλ/ι cells (Figure 4C). Increased activation of mTORC1 was also detected in PIN lesions of DKO mice, as determined by pS6 (Ser240/244) staining (Figures 4D and 4E). mTORC1 inhibition by rapamycin treatment in sgPKCλ/ι cells restored the levels of NEPC and basal markers to control levels (Figures 4F and 4G). GSEA revealed that inhibition of mTORC1 with Torin1 in LNCaP cells that had been starved of androgens for 2 days significantly correlated with a decrease in genes overexpressed in NEPC (Figure 4H). These results demonstrate that the activation of mTORC1 upon PKCλ/ι deficiency is central to NEPC differentiation.

Figure 4. PKCλ/ι Regulates mTORC1 Activity through LAMTOR2 Phosphorylation.

(A) Top 5 GSEA results of shPKCλ/ι vs. shNT comparison of PrEC cells using compilation H (MSigDb).

(B) Top 5 GSEA results of sgPKCλ/ι vs. sgC and sgPKCλ/ι vs. sgPKCλ/ι-R comparisons of C42B cells using compilation H (MSigDb).

(C) Western blot of indicated proteins in sgC, sgPKCλ/ι, sgPKCλ/ι-R WT and sgPKCλ/ι-R K274W C42B cells.

(D) pS6 (Ser 235/236) IF of PTEN KO and DKO prostates (n = 4). Scale bar, 50 μM.

(E) Quantification of pS6 in (D).

(F) Western Blot of indicated proteins in sgC and sgPKCλ/ι C42B cells treated with 50 nM rapamycin (rapa).

(G) qPCR of indicated genes in sgC and sgPKCλ/ι C42B cells treated with vehicle (Veh) or 50 nM rapamycin (n = 5).

(H) GSEA of “mTORC1_Signaling” and “NEPC_UP” gene sets in the comparison of Torin1-treated LNCaP vs. vehicle, GSE93603.

(I) Volcano plot of biotinylated proteins in PKCλ/ι-BioID2 vs. Empty-BioID2 LNCaP cells (n = 5). Uniques: Proteins identified only in PKCλ/ι-BioID2 cells, not present in Empty-BioID2 cells.

(J) Double staining of PKCλ/ι and LAMTOR2 in LNCaP cells. Yellow dotted line marks the nuclei. Scale bar, 25 μM.

(K) Co-IP of PKCλ/ι and LAMTOR2 in 293FT cells transfected with FLAG-LAMTOR2.

(L) In vitro phosphorylation of recombinant LAMTOR2 by recombinant PKCλ/ι with ATPγS followed by PNBM alkylation and immunoblotting for the indicated proteins.

(M) In vitro phosphorylation of FLAG-tagged WT and S30A LAMTOR2 immunoprecipitates by recombinant PKCλ/ι with ATPγS followed by PNBM alkylation and immunoblotting for the indicated proteins.

(N) Western blot of indicated proteins in C42B cells transfected with WT or S30A FLAG-LAMTOR2.

(O) Immunostaining for the indicated proteins and radial distribution profiling of mean intensities from the nucleus center in sgC and sgPKCλ/ι. Yellow dotted line marks the nuclei. Scale bars, 10 μM. Results are shown as mean from n = 6 cells per condition. Unadjusted multiple comparisons by t-test over 750 data points, significance of at least p < 0.05 indicated by black line.

(P) Western Blot of indicated proteins in sgC and sgPKCλ/ι treated with Ciliobrevin D (40 μM) for 8 hr.

(Q) LAMP2 staining and radial distribution profiling of mean intensities from the nucleus center (n = 6 cells, each condition) in LNCaP cells transfected with LAMTOR2 WT or LAMTOR2 S30A. Scale bars, 10 μM. Yellow dotted line marks the nuclei. Scale bars, 10 μM. Results are shown as mean from n = 6 cells per condition. Unadjusted multiple comparisons by t-test over 750 data points, significance of at least p < 0.05 indicated by black line

(R) Co-IP of mTOR and LAMTOR2 in sgC and sgPKCλ/ι C42B cells transfected with FLAG-LAMTOR2.

(S) Co-IP of mTOR and LAMTOR2 in 293FT cells transfected with FLAG-LAMTOR2 WT or S30A.

(T) Model: the loss of PKCλ/ι promotes the perinuclear aggregation of lysosomes, which favors the interaction of mTOR with its co-activator Rheb.

Mean ± SEM (E, G). two-tailed t-test (E, G). *p < 0.05, **p < 0.01, ***p < 0.001. q value is p value with adjustment for multiple hypotheses. See also Figure S4.

To dissect the regulatory mechanism between PKCλ/ι and mTORC1, we identified interactors of PKCλ/ι using the proximity-dependent biotin method (BioID2) in LNCaP cells (Figure S4C). p62/SQSTM1, a bona fide PKCλ/ι interactor, was identified in this screening, which validated the approach (Figure 4I). Among the proteins found exclusively in PKCλ/ιBioID2 samples, LAMTOR2 stood out as a potential link between PKCλ/ι and mTORC1 (Figure 4I). LAMTOR2 is part of the Ragulator, which is a pentameric scaffold complex that regulates mTORC1 at the lysosome membrane (Sancak et al., 2010). Other components of the endosomal/lysosomal system, including LAMP2, KIF2C/B and V-ATPases were also found as PKCλ/ι interactors (Figure 4I). LAMTOR2 colocalized and co-immunoprecipitated with endogenous PKCλ/ι (Figures 4J and 4K). PKCλ/ι was able to directly phosphorylate LAMTOR2 in an in vitro kinase assay (Figure 4L). Serine 30 in LAMTOR2 was predicted as the unique phosphorylation site by PKCλ/ι using Scansite 4.0 and is highly conserved across species (Figure S4D). Mutation of LAMTOR2 S30 site to alanine significantly reduced PKCλ/ι phosphorylation (Figure 4M). Analysis of the crystal structure of the Ragulator complex revealed that LAMTOR2 associates with the RagA/B and RagC/D GTPases, which are important to activate mTORC1. Serine 30 is located at the interface of LAMTOR2 with the Rag GTPases (Figure S4E), which could impact complex formation, recruitment to lysosomes and mTORC1 activation. In fact, we detected enriched expression of LAMTOR2, and the other components of the Ragulator:Rag complex, in lysosomal membranes of sgPKCλ/ι cells (Figure S4F). Expression of the LAMTOR2 S30A mutant in C42B cells was sufficient to increase mTORC1 activity and NEPC markers (Figure 4N). PKCλ/ι deficient cells revealed a strikingly different lysosomal distribution compared to sgC cells, which included perinuclear aggregation of lysosomes, and increased amount of mTOR and RagC proximal to the nucleus, where Rheb, a key activator of mTORC1 is enriched (Figure 4O). Treatment with Ciliobrevin D (CilioD), a dynein inhibitor that blocks the retrograde transport system, reduced mTORC1 activity of sgPKCλ/ι cells to control levels (Figure 4P). The expression of LAMTOR2 S30A mutant mimicked the lysosomal phenotype of sgPKCλ/ι cells and promoted the perinuclear aggregation of lysosomes (Figure 4Q). Consistent with increased mTORC1 activation, the interaction of mTOR with LAMTOR2 was stronger in sgPKCλ/ι cells than in sgC cells (Figure 4R). Likewise, the LAMTOR2 S30A mutant displayed enhanced affinity for mTOR than LAMTOR2 WT (Figure 4S). These data support a model whereby PKCλ/ι inhibits mTORC1 activation by forcing the dispersion of lysosomes and impairing its interaction with LAMTOR2 through phosphorylation of the S30 (Figure 4T).

Loss of PKCλ/ι Increases Serine Metabolism through the mTORC1/ATF4/PHGDH Axis

Next, we carried out pathway analysis of gene expression data generated in our C42B cell line model. This analysis revealed ATF4 as the main activated upstream regulator of the transcriptional changes observed in PKCλ/ι-deficient cells (Figure 5A). Consequently, many targets of this transcription factor, including ATF4 itself, appeared upregulated in sgPKCλ/ι cells (Figure S5A). Increased ATF4 was confirmed by immunoblotting (Figure 5B). siRNA-mediated KD of PKCλ/ι in LNCaP cells promoted upregulation of ATF4 (Figure S5B), and tumor lesions in the DKO mice showed increased nuclear ATF4 staining than PTEN KO tumors (Figure 5C). Upregulation of ATF4 in PKCλ/ι-deficient cells was reduced by inhibition of mTORC1 with rapamycin or siRNA-mediated KD of Raptor, while no changes in any of the three branches of the unfolded protein response (UPR) were observed (Figures 5B and S5C–S5E). ATF4 KD by shRNA in sgPKCλ/ι cells resulted in reduced levels of NEPC and basal markers (Figures 5D and 5E), as well as impaired cell proliferation (Figure 5F). To delineate which subset of ATF4 targets were upregulated upon PKCλ/ι loss, we compared the ATF4 cistrome with genes upregulated in sgPKCλ/ι cells and performed unbiased pathway analysis. Serine and glycine biosynthesis appeared as the most significantly upregulated signature (Figure S5F). GSEA revealed significant enrichment in sgPKCλ/ι of a recently generated serine, glycine, one-carbon pathway (SGOCP) gene signature (Mehrmohamadi et al., 2014), as compared to sgC and sgPKCλ/ι-R cells (Figure 5G). This included key genes of the pathway such as PHGDH, MTHFD2 and PSAT1 (Figure 5G). Several genes of the SGOCP were found to be upregulated in LNCaP cells transfected with an siRNA PKCλ/ι (Figure S5G), as well as in shPKCλ/ι PrEC cells (Figure S5H). ATF4, PHGDH and NEPC markers were rescued by the WT form of PKCλ/ι, but not by its kinase-dead mutant (Figure 5H). Both shRNA-mediated KD of ATF4 and rapamycin treatment inhibited the expression of the SGOCP genes (Figures 5I and 5J). Ectopic expression of ATF4 in LNCaP cells promoted the induction of SGOCP genes, concomitant with an increase in NEPC and basal markers (Figures 5K and 5L). E2F1 upregulation was reduced upon mTORC1 and cMYC inhibition but was not altered by changes in ATF4 levels (Figures 5I–5L and S5I), suggesting PKCλ/ι-mediated regulation of E2F1 is ATF4-independent. shRNA-mediated KD of PHGDH reduced NEPC and basal markers in sgPKCλ/ι cells (Figures 5M and 5N), and partially inhibited their growth advantage (Figure 5O).

Figure 5. Loss of PKCλ/ι Increases Serine Metabolism through the mTORC1/ATF4/PHGDH axis.

(A) Upstream regulator analysis by Ingenuity Pathway Analysis (IPA) of PKCλ/ι-dependent genes in C42B cells.

(B) Western blot analysis of indicated proteins in sgC and sgPKCλ/ι C42B cells treated with 50 nM rapamycin (rapa).

(C) ATF4 staining in PTEN KO and DKO prostates and quantification of nuclear ATF4⁺ cells (n = 5–7). Scale bar, 25 μM.

(D) Western blot analysis of indicated proteins in sgPKCλ/ι and sgC stably transfected with shNT or shATF4.

(E) qPCR of indicated genes in sgPKCλ/ι and sgC C42B cells stably transfected with shNT or shATF4 (n = 3).

(F) Cell proliferation of sgPKCλ/ι and sgC C42B cells stably transfected with shNT or shATF4 (n = 3).

(G) GSEA of the gene set “SGOCP” in the comparisons sgPKCλ/ι vs. sgC C42B cells (left) and sgPKCλ/ι-R vs. sgPKCλ/ι (right) and heatmap of PHGDH, PSAT1 and MTHFD2 expression with Log2FC values for the sgPKCλ/ι vs. sgC comparison.

(H) Western blot of indicated proteins in sgC, sgPKCλ/ι, sgPKCλ/ι-R WT and sgPKCλ/ι-R KiD C42B cells.

(I) qPCR of indicated genes in sgC and sgPKCλ/ι C42B cells stably transfected with shNT or shATF4 (n = 3).

(J) qPCR of indicated genes in sgC and sgPKCλ/ι C42B cells treated with 50 nM rapamycin (n = 3).

(K) Western blot analysis of indicated proteins in control and ATF4-transfected LNCaP cells.

(L) qPCR of indicated genes in control and ATF4-transfected LNCaP cells (n = 3).

(M) Western blot analysis of indicated proteins in sgPKCλ/ι and sgC C42B cells stably transfected with shNT or shPHGDH.

(N) qPCR of indicated genes in sgC and sgPKCλ/ι C42B cells stably transfected with shNT or shPHGDH (n = 3).

(O) Cell proliferation of sgC and sgPKCλ/ι C42B cells stably transfected with shNT or shPHGDH (n = 3).

(P) Fraction of labeled [U-¹³C₆]Glucose-derived intracellular serine and glycine (n = 3). Statistical significance for sgC vs. sgPKCλ/ι comparison. Mean ± SD.

(Q) Isotopologue distribution (M0 to M3 according to labeled carbons) of [U-¹³C₆]Glucose-derived intracellular serine and glycine in sgC (C), sgPKCλ/ι (λ/ι), sgPKCλ/ι-R WT (R) C42B cells (n = 3).

(R) Fraction of labeled [α-¹⁵N]Glutamine-derived intracellular glutamate (Glu), serine (Ser) and glycine (Gly) at 24 hr in sgC, sgPKCλ/ι, and sgPKCλ/ι-R C42B cells (n = 3).

Mean ± SEM (C, E, F, I, J, L, N, O, Q, R). Two-tailed t-test (C, E, I, J, L, N, R). *p < 0.05, **p < 0.01, ***p < 0.001. Two-way ANOVA (F,O,P). See also Figure S5.

Next, we investigated whether transcriptional changes driven by the ATF4/SGOCP axis translated into a metabolic phenotype in sgPKCλ/ι cells. Analysis of metabolite abundances revealed that sgPKCλ/ι cells displayed increased intracellular levels of glycine, but not serine (Figure S5J). Since serine conversion to glycine serves to donate 1C units to the One Carbon Pool, these results suggest an accumulation of glycine downstream of increased serine synthesis. sgPKCλ/ι cells had lower lactate/glucose ratios than control cells, despite increased cellular proliferation (Figure S5K), which indicated a difference in glucose utilization, likely by the diversion of carbons to the SGOCP. Tracing analysis of uniformly labeled [U-¹³C₆]Glucose showed serine and glycine as the top two metabolites with increased labeling in sgPKCλ/ι cells (Figures 5P, 5Q and S5L). sgPKCλ/ι cells displayed higher glycine efflux despite increased glycine accumulation, while no changes in serine efflux were observed (Figure S5M). Tracing of [α-¹⁵N]Glutamine, which donates 1N via glutamate during the step catalyzed by PSAT1, showed increased labeling in serine and glycine (Figure 5R). These results demonstrate that the loss of PKCλ/ι activates mTORC1 to drive an ATF4-dependent gene transcription program that promotes NEPC differentiation and results in the metabolic reprogramming of PCa cells to increase the flux through the SGOCP.

Human Relevance of the mTORC1/ATF4/PHGDH Axis in NEPC

Comparative bioinformatics analysis of NEPC vs. prostate adenocarcinoma human samples by GSEA showed a significant enrichment in the “mTORC1 Signaling” signature in NEPC (Figure 6A). Gene expression analysis of datasets of enzalutamide resistant C42B cells (Figure 6B) demonstrated the upregulation of SGOCP enzymes concomitant with downregulation of PKCλ/ι in the resistant samples (Figures 6B). Similar alterations were seen following enzalutamide treatment in an enzalutamide resistant human metastasis profiled before and after therapy (Figure 6C). Interrogation of a well-characterized collection of tumor-derived organoids of advanced PCa (Gao et al., 2014) revealed downregulation of PKCλ/ι concomitant to an increase in NEPC markers and SGOCP enzymes in an organoid line derived from a treatment-induced case of NEPC (PCa4) (Figure 6D). Further bioinformatics interrogation of a human NEPC dataset showed increased mRNA expression of SGOCP enzymes as compared to adenocarcinoma samples (Figure 6E). Immunohistochemistry analysis of adenocarcinoma and NEPC tissues from a cohort of 19 patients (Figures S6A and S6B), showed higher mTORC1 activity, as determined by p4EBP1 (Thr37/46) staining, concomitant with increased expression of PHGDH, in NEPC as compared to adenocarcinoma (Figures 6F, 6G, S6A and S6B). Notably, areas of NEPC showed much stronger nuclear ATF4 staining than adenocarcinoma lesions (Figures 6H, S6A and S6B). These results demonstrate the clinical relevance of the hyperactivation of the mTORC1/ATF4/PHGDH axis, which occurs in de novo NEPC as well as in NEPC lesions that arise after treatment.

Figure 6. Human Relevance of the mTOR/ATF4/PHGDH Axis in NEPC.

(A) GSEA of “mTORC1 Signaling” Hallmark gene set in NEPC vs. Adeno (Adenocarcinoma) in two published human PCa datasets of RNA-seq (Beltran 2011) and microarray (GSE32967) gene expression.

(B) Expression values of PRKCI, ATF4, ASNS and PHGDH in enzalutamide sensitive (Sen) and resistant (Res) C42B cells from a published microarray gene expression dataset (GSE64143).

(C) Expression values of the indicated genes from a published RNA-seq gene expression dataset (GSE70380). V1_met: rib metastasis obtained during first visit, before treatment. V2_met: metastasis obtained during second visit after enzalutamide (enza) treatment for 12 weeks.

(D) Heatmap of PRKCI, NE markers and SGOCP gene expression in a published RNA-seq dataset of patient derived prostate cancer organoid lines.

(E) Expression for the indicated genes in Adenocarcinoma (Adeno) and NEPC samples in a published RNA-seq gene expression dataset (Beltran 2011). Box and whiskers graphs indicate the median and the 25th and 75th percentiles, with minimum and maximum values at the extremes of the whiskers.

(F) IHC staining of p4EBP1 (Thr37/46) and H&E, with quantification (normalized values) in Adenocarcinoma (Adeno, n = 6) and NEPC (n = 10). Scale bars, 100 μm.

(G) p4EBP1 (Thr37/46) and PHGDH IHC, and H&E staining, in a primary NEPC lesion with adenocarcinoma (yellow-dashed line) and an adjacent NEPC lesion (black-dashed line) with quantification of PHGDH staining (normalized values) in Adenocarcinoma (n = 6) and NEPC (n = 7). Scale bars, 100 μm.

(H) Double staining of SYP and ATF4 with quantification of nuclear ATF4 in Adenocarcinoma (Adeno, n = 6) and NEPC (n = 13). White scale bars, 50 μm. Black scale bar, 100 μm. Quantification of ATF4 intensity.

Mean ± SEM (F, G, H). Two-sided t-test (E, F, G, H). *p < 0.05, **p < 0.01, ***p < 0.001. See also Figure S6.

Loss of PKCλ/ι Increases DNA Methylation to Promote NEPC Differentiation

Serine metabolism fuels the methionine salvage pathway to produce SAM for DNA methylation (Yang and Vousden, 2016). Since epigenetic regulation has been recently recognized as a prognostic value to distinguish indolent from aggressive forms of prostate cancer (Bhasin et al., 2015; Rickman et al., 2017), we tested the hypothesis that increased SGOCP could affect the lineage specificity of NEPC through epigenetic modulation. First, we assessed whether increased SGOCP in PKCλ/ι deficient cells could promote incorporation of SGOCP-derived carbon units into genomic DNA. Indeed, sgPKCλ/ι cells had increased methyl-cytosine labeling from [methyl-¹³C]Methionine (Figure S7A), a measure of 1C units donated from SAM pools. 5mC dot-blot analysis of whole genomic DNA extracts showed increased DNA methylation in sgPKCλ/ι cells, which was confirmed by MS data of total MeCyt to Cytosine ratios (Figures 7A and S7B). The intracellular levels of SAM were upregulated in sgPKCλ/ι cells, which were rescued by the KD of PHGDH that also inhibited global genomic methylation levels (Figures 7A and 7B). Similarly, KD of ATF4 or inhibition of mTORC1 with rapamycin reduced SAM to control levels in sgPKCλ/ι cells (Figures S7C and S7D).

Figure 7. Loss of PKCλ/ι Increases DNA Methylation to Promote NEPC Differentiation.

(A) Dot-blot analysis of total genomic DNA methylation levels of sgC shNT, sgPKCλ/ι shNT and sgPKCλ/ι shPHGDH C42B cells. Methyl blue (MB) staining for total genomic DNA loading.

(B) Intracellular SAM levels in sgPKCλ/ι and sgC C42B cells with stable shPHGDH or shNT (n = 3), and western blot of indicated proteins.

(C) Violin plot of pileup values (amount of reads at peak summit) for mapped regions of the Medip-seq in sgC and sgPKCλ/ι C42B cells, black lines show magnification of the 0 to 100 score region with median pileup value in red.

(D) Venn diagram of overlap between differentially methylated regions (DMR) in sgC and sgPKCλ/ι C42B cells.

(E) Venn diagram of overlap between differentially methylated regions (DMR) in sgPKCλ/ι cells and differentially expressed genes (DEG) in sgPKCλ/ι vs. sgC cells.

(F) Medip-seq and qPCR of ADAMTS1 and CDKN1A in sgC and sgPKCλ/ι C42B cells stably transfected with shPHGDH or shNT (n = 3).

(G) Venn diagram of gene overlap between DMR sgPKCλ/ι and Hypermethylated CpG regions in NEPC and between DMR sgPKCλ/ι and DMR in High Grade PCa.

(H) Western blot of indicated proteins in sgC and sgPKCλ/ι C42B cells treated with 5 μM aza.

(I) qPCR of indicated genes in sgC and sgPKCλ/ι C42B cells treated with 5 μM aza for 4 days (n = 3).

(J) Western blot of indicated proteins in sgC and sgPKCλ/ι C42B cells treated with 2 mM cyclo for 6 days.

(K) qPCR of indicated genes in sgC and sgPKCλ/ι C42B cells treated with 2 mM cyclo for 6 days.

(L) Cell proliferation of sgC and sgPKCλ/ι C42B cells with cyclo or aza (n = 3).

(M) Average size of PTEN KO and DKO prostate-derived organoids treated with 5 μM aza for 4 days (n = 3).

(N) qPCR of Ncam1 in PTEN KO and DKO prostate-derived organoids treated as in (M; n = 3).

(O) Tumor growth of xenografts of sgC shNT (n = 6), sgPKCλ/ι shNT (n = 5) and sgPKCλ/ι shPHGDH (n = 6) C42B cells treated with Veh, and sgPKCλ/ι shNT treated with aza (n = 4) starting at the time indicated by the arrow.

(P) Staining and quantification of 5mC and NCAM1 in sgC and sgPKCλ/ι C42B xenograft tumors treated with veh or aza (n = 3) and H&E staining. Scale bars, 50 μm.

(Q) Diagram of proposed mechanism.

Mean ± SEM (B, F, I, K-P). Two-tailed T-test (B, F, I, K, M, N, P). *p < 0.05, **p < 0.01, ***p < 0.001. Two-way ANOVA (L, O). Hypergeometric test (E, G). See also Figure S7.

We next characterized the methylome of sgPKCλ/ι and sgC cells by Methylated DNA Immunoprecipitation followed by sequencing (Medip-seq). PKCλ/ι loss promoted an increase in genome wide methylation, as detected by enhanced median region methylation scores (Figure 7C), which included differential methylation in regions within and around gene bodies (Figure S7E). Differentially methylated regions (DMR) in sgPKCλ/ι cells (DMR sgPKCλ/ι) had a significant overlap with their differentially expressed genes (Figures 7D and 7E). These results demonstrate that the alterations in DNA methylation in sgPKCλ/ι cells are correlated with transcriptional changes. We next performed transcription factor enrichment analysis of clustered regions based on methylation profiles around genes differentially expressed between sgPKCλ/ι and sgC cells. Upregulated genes clustered by methylation profiles showed an enrichment of binding motifs for Pax5, Nr5a2 and Oct4-Sox2-Tcf-Nanog transcription factors (Figure S7F), which have been previously involved in embryonic morphogenesis, neural development and maintenance of stem cell traits. On the other hand, downregulated genes clustered by methylation profiles included a subset that showed an enrichment for Androgen Response Element (ARE) (Figure S7G), which suggested that DNA methylation could contribute to the loss of AR activity. In fact, the KD of PHGDH rescued the expression of hypermethylated and downregulated genes in sgPKCλ/ι cells, including the AR-dependent gene ADAMTS1 and the negative regulator of cell cycle CDKN1A (p21) (Figure 7F). These observations are of clinical relevance since differentially methylated regions (DMR) in sgPKCλ/ι cells contained a statistically significant overlap with CpG areas found hypermethylated in NEPC patients, as well as those found hypermethylated in highly aggressive PCa (Beltran et al., 2016; Bhasin et al., 2015) (Figure 7G). Gene Ontology (GO) pathway analysis of DMR sgPKCλ/ι-associated genes, hypermethylated NEPC CpGs, and DMR of highly aggressive PCa, showed a clear enrichment in development-related gene signatures, including “embryonic morphogenesis”, “pattern specification”, and “cell morphogenesis involved in neuron differentiation” processes (Figure S7H). These data indicate that the DNA methylation changes observed upon PKCλ/ι deficiency impact the expression of genes central to the control of cell-type specification, which are similar to those observed during NEPC differentiation.

DNMT Inhibition Blocks NEPC Differentiation and Tumor Growth

To investigate the therapeutic potential of the inhibition of DNA methyltransferase (DNMT) activity in NEPC, sgPKCλ/ι cells were treated with decitabine (aza), a pharmacological inhibitor of DNMT, or with cycloleucine (cyclo), which blocks the last enzyme in the production of SAM (MAT), and therefore the metabolic flux that supplies methyl donors for the DNMT reaction (Figure S7I). Aza treatment severely reduced NEPC and basal markers in sgPKCλ/ι cells to levels like those in sgC cells (Figures 7H and 7I). Similarly, cyclo reduced the expression of basal and NEPC markers in sgPKCλ/ι cells (Figures 7J and 7K). Both drugs had a potent anti-proliferative effect on sgPKCλ/ι C42B cells (Figure 7L). Aza treatment was effective in blocking the proliferative advantage of DKO organoids and reduced the expression of Ncam1 to levels of PTEN KO organoids (Figures 7M and 7N). Aza treatment of sgPKCλ/ι C42B-derived xenografts inhibited both the growth of already stablished tumors and the levels of 5mC and NCAM1 to those of vehicle-treated samples (Figures 7O and 7P). Importantly, KD of PHGDH also completely inhibited the growth advantage of sgPKCλ/ι tumors (Figure 7O). These results support a model whereby the metabolic reprogramming orchestrated by PKCλ/ι deficiency through the mTORC1/ATF4/PHGDH axis creates a vulnerability in NEPC that can be exploited therapeutically by targeting the SGOCP and DNA methylation.

DISCUSSION

Here we report that PKCλ/ι is a critical suppressor of NEPC, which has broader implications in cancer since lineage plasticity has emerged as a mechanism of drug resistance, not only in PCa, but also in other types of neoplasia (Niederst et al., 2015). Similar to the simultaneous loss of Trp53, and Rb1 with Pten (Ku et al., 2017), PKCλ/ι deficiency regulates lineage plasticity to facilitate the transition from a luminal to basal phenotype. Luminal prostate epithelial cells depend on AR for survival, but basal cells do not, which allows tumor basal cells to escape ADT and proliferate in the absence of AR, which is a common and fundamental feature of NEPC (Rickman et al., 2017). Importantly, PKCλ/ι-deficient prostate epithelial cells showed a sustained inhibition of AR both in vitro and in vivo, which makes them resistant to enzalutamide. Therefore, our results support a model whereby luminal cells that initiate the lineage switch to a basal/neuroendocrine phenotype upon PKCλ/ι loss, acquire survival and proliferative advantages that allow them to adapt and escape therapy.

At the molecular and mechanistic levels, our study provides insights into the signaling cascades that govern NEPC and identifies metabolic and synthetic vulnerabilities that occur upon PKCλ/ι loss to target this lethal type of PCa. PKCλ/ι deficiency in prostate epithelial cells rewires the tumor metabolism towards the serine biosynthetic pathway through an mTORC1-dependent upregulation of ATF4 (Figure 7Q). Of note, PKCλ/ι represses basal mTORC1 activity by directly phosphorylating LAMTOR2, a central component of the Ragulator complex that controls mTORC1 localization to the lysosomes and, therefore, its activity (Bar-Peled et al., 2012). LAMTOR2 has also been shown to regulate lysosomal positioning (Sancak et al., 2010). Our data demonstrate that the lack of LAMTOR2 phosphorylation, as consequence PKCλ/ι deficiency, results in the perinuclear redistribution of lysosomes, which serves to favor the proximity of mTOR to Rheb and the Rag GTPases, its obligate activators. These observations agree with the lysosomal phenotype previously reported in LAMTOR2 KO cells (Teis et al., 2006), and is reminiscent of recent work linking lysosomal redistribution with mTORC1 activation in the regulation by acid of the circadian clock (Walton et al., 2018). Therefore, one of the important roles of PKCλ/ι as a tumor suppressor is to act as a repressor of mTORC1 by promoting the dispersion of lysosomes, which is essential for the acquisition of the NEPC phenotype in PCa.

PCa has a unique metabolism that, unlike most solid tumors, it is not highly glycolytic (at least at early stages) and prostate cancer cells rely mostly on glutamine and lipids. However, whether metabolic differences exist between adenocarcinoma and NEPC was not known. In fact, the hyperproliferative state of NEPC suggests that a different set of metabolic pathways would be needed to fulfill its energetic and anabolic needs. Our data demonstrate that PKCλ/ι-controlled metabolic reprograming offers competitive advantages during NEPC differentiation through the regulation of SGOCP, a critical metabolic node for proliferation and the epigenetic control of gene expression (Yang and Vousden, 2016). This metabolic rewiring results in the increased expression of key metabolic enzymes, including PHGDH, which is the first and limiting step in SGOCP, and is upregulated in several aggressive tumors (Yang and Vousden, 2016; Ma et al., 2013), which creates a metabolic vulnerability in high PHGDH expressing tumors like NEPC.

The central role of SGOCP in cell growth and transformation is due to its pleiotropic function in the control of key metabolic pathways including the synthesis of nucleotides needed for cell growth, and that of NAD(P)H and glutathione to maintain an adequate redox balance in cancer cells (Yang and Vousden, 2016). However, another important function of SGOCP is the biosynthesis of SAM, which promotes epigenetic changes conducive to NEPC differentiation. Recent evidences demonstrated that NEPC tumors showed a distinct DNA methylation profile, which segregates them from adenocarcinomas (Beltran et al., 2016). Importantly, the epigenetic differences induced by the loss of PKCλ/ι overlap with CpG areas found hypermethylated in highly aggressive PCa and NEPC patients (Beltran et al., 2016; Bhasin et al., 2015). This demonstrate that PKCλ/ι is a critical tumor suppressor during NEPC by impacting the epigenome through the SGOCP. In fact, the epigenetic enzymes EZH2, DEK and DNMTs are commonly found upregulated in NEPC, and inhibition of EZH2 has been shown to revert key NEPC features and prevent growth in combination with AR-directed therapies (Beltran et al., 2016; Ku et al., 2017). Therefore, our results could have potential therapeutic implications since the inhibition of DNA methylation with the FDA-approved drug decitabine blocks proliferation and reverts basal and NEPC markers. This strongly suggest that targeting DNMTs could offer a strategy for treatment of NEPC, potentially in combination with enzalutamide. Our results reported here establish that the DNA methylation changes observed upon PKCλ/ι deficiency are central to the control of cell specification and identity, which supports the relevance of the PKCλ/ι-mTORC1-ATF4-SGOCP cascade in NEPC and provides additional rationale to therapeutically exploit these pathways.

STAR★METHODS

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Maria T Diaz-Meco (mdmeco@sbpdiscovery.org)

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Mice

C57BL/6 background Pten^f/f-PbCre4⁺ mice were generated by breeding Pten^f/f-PbCre4⁻ females with Pten^f/wt-PbCre4⁺ males. C57BL/6 background Pten^f/f-Prkci^f/f-PbCre4⁺ mice were obtained from crossing Pten^f/f-Prkci^f/f-PbCre4⁻ females with Pten^f/f-Prkci^f/f-PbCre4⁺ males. Prkci^f/f mice have been previously described in (Nakanishi et al., 2016). For the xenograft experiments, NSG mice of 2 months of age were used. NSG mice were purchased from the animal core at SBP Medical Discovery Institute. All mice were born and maintained under pathogen-free conditions. Animal handling and experimental procedures conformed to institutional guidelines and were approved by the Institutional Animal Care and Use Committee at SBP Medical Discovery Institute. All genotyping was done by PCR. The WD (D12079B; OpenSource Diets) was available ad libitum. Experimental mice were all males. NSG mice were randomized between litters to prevent a bias towards age. Other animal subjects were not randomly assigned in animal experiments. Sex- and age-matched animals were allocated from each genotype into experimental groups. An identification code was assigned to each animal and the investigators were not blinded to group allocation at the time of data collection and analysis. For prostate organoid isolation, dorsolateral and ventral lobes for each mouse genotype were used and established as described under prostate organoid culture. For mouse xenografts, sgC or sgPKCλ/ι C42B cells were trypsinized, washed two times in cold PBS, aliquoted to 250K per dose in a 50 μl volume and injected subcutaneously into both flanks of male NSG mice. Measurements were done weekly using a caliper. Daily gavage treatment with 10mg/kg enzalutamide or vehicle (using a formulation of 1% carboxymethyl cellulose, 0.1% Tween-80, 5% DMSO) was initiated on the day of injection. Treatment by IP injection of 1mg/kg aza in PBS or vehicle was initiated when tumors reached an average size of 250 mm³ and administered three consecutive days weekly.

Human Samples

FFPE tissue samples from male PCa patients were obtained from Scripps Clinic and Complejo Hospitalario Universitario de Albacete (CHUA) (Albacete, Spain). Written informed consent was obtained from all patients with the protocol approved by the Ethics Committee of CHUA and Scripps Green Hospital in accordance with the ethical guidelines for epidemiological research, as well as the principles expressed in the Declaration of Helsinki. Samples were de-identified in terms of all the covariate relevant details such as age, sex, and past history, and were sent to SBP Medical Discovery Institute and used for histological analyses. Informed consent was obtained from all participants. The study was approved by the IRB Committee of SBP Medical Discovery Institute.

Cell lines

HEK293T (sex: female), C42B (sex: male), LNCaP-FGC (sex: male), PC3 (sex: male), DU145 (sex: male) and Phoenix-GP (sex: female) were obtained directly from ATCC. PrEC cell line (sex: male) was purchased from LONZA. All cells were negative for mycoplasm.

METHOD DETAILS

Cell culture experiments

LNCaP and C42B cells were cultured in Rosewell Park Memorial Instiute 1640 (RPMI 1640, GIBCO). PC3, HEK293T, and DU145 were cultured in Dulbecco’s Modified Eagles Medium (DMEM). NE-1–3 and H-660 were cultured in RPMI1640 without phenol red (GIBCO). All base mediums were supplemented with 10% fetal bovine serum (FBS), 2 mM glutamine, and 100 U/mL penicillin and 100 μg/mL streptomycin, in an atmosphere of 95% air and 5% CO2, except for NE-1–3 and H-660 that 5% Charcoal Stripped Serum (CSS) was used. Lentiviruses and retroviruses were prepared and used as previously described. The following treatments were applied as follows; rapamycin was used at 100 nM, cycloleucine was used at 2 mM with daily media change, decitabine was used at 5 μM, and DHT was used at 10 nM. Androgen Deprivation Therapy conditions consisted on RPMI1640 media without phenol red (GIBCO), 10% dialyzed FBS, Glutamax and 100 U/mL penicillin 100 and 100 μg/mL streptomycin.

Isolation and Culture of Prostate Epithelial Cells

Murine prostates were isolated from either 8-week-old PTENfl/fl-PBcre or age-matched ~22-week-old PTENf/f-PBCre and PTENf/f-PKCλ/ιf/f-PBCre male mice and were placed in 5 mg/mL collagenase type II in ADMEM/F12 and digested for 1 to 2 hr at 37 °C. Glandular structures were washed with ADMEM/F12 and centrifuged at 300 G. Subsequently, structures were digested in 5 mL TrypLE with the addition of Y-27632 to 10 μM for 15 min at 37 °C. Trypsinized cells were washed and seeded in growth factor-reduced Matrigel. Murine prostate epithelial cells were cultured in ADMEM/F12 supplemented with B27, 10 mM HEPES, Glutamax, and penicillin/streptomycin and containing the following growth factors: EGF 50 ng/mL, Rspondin1-conditioned medium, 100 ng/mL recombinant Noggin, TGF-β/Alk inhibitor A83–01 and 0.1–1nM DHT. Murine prostate organoids were passaged by trypsinization with TrypLE for 5 min at 37 °C. Lentiviral infections were performed using pLKO .1-puro targeting mouse PKCλ/ι or control scramble. In short, 250,000 single cells were infected with 1:1000 dilution from stock concentrated virus. Infection was done during centrifugation for 1 hr at 600 G at RT. Cells were subsequently placed at 37 °C, 5% CO ₂ for 3 hr to recover. Cells were plated in Matrigel and, 24 hr post seeding, 1 μg/mL puromycin was applied for 7 days to ensure only infected cells remained.

FACS and FACS-sort analysis

Cell lines in culture were washed in PBS once and collected by scraping them in FACS buffer (1X PBS with 2% FBS, 25 mM HEPES pH 7.0 and 2 mM EDTA), pelleted by centrifugation, resuspended in appropriate amount of FACS buffer, stained in the dark for 30 minutes and analyzed with the addition of propidium iodide right before FACS analysis using FACS ARIA at the SBPMDI FACS core. FACS analysis of prostate-derived cell preparations was performed as follows; prostate cell preparations were obtained as described under “Isolation and Culture of Prostate Epithelial Cells”. Cells were incubated with anti-CD16/CD32 antibody (2.4G2) diluted in FACS buffer before staining with labeled antibodies: anti-CD49f (GoH3) was purchased from BD Horizon, anti-CD31 (390), anti-CD45 (30-F11), anti-TER-119 (TER-119) and anti-CD140a (APA5) from eBioscience, anti-Sca-1 (E13–161.7) and anti-CD56 (HCD56) from BioLegend. Then, cells were washed and subjected to FACS analysis. Data were acquired with a LSRFortessa 14-colors cytometer (BD Biosciences) at the SBPMDI FACS core and analyzed with FlowJo software (Tree Star Inc.).

Cell lysis, immunoprecipitation and western blotting

Cells for protein analysis were lysed in RIPA buffer (20 mM Tris-HCl, 37 mM NaCl, 2 mM EDTA, 1% Triton-X, 10% glycerol, 0.1% SDS, and 0.5% sodium deoxycholate, with phosphatase and protease inhibitors). For immunoprecipations, cells were lysed in σ3 buffer (100 mM NaCl, 25 mM Tris, 1% Triton-X, 10% glycerol, with phosphatase and protease inhibitors) and immunoprecipitated with 25 μl of 50% slurry of protein Glutathione-Sepharose 4B beads (Bioworld). Immunoprecipitates were washed three times with lysis buffer, once with high salt (500 mM NaCl), and once more with lysis buffer. Cell extracts and immunoprecipitated proteins were denatured, subjected to SDS-PAGE, transferred to PVDF membranes (GE Healthcare), and immunoblotted with the specific antibodies as listed in STAR methods Reagents table.

Immunostaining quantification

All samples for each experiment were processed equally and simultaneously to reduce batch effects. Image acquisition was performed using the same settings for all samples. PKCλ/ι expression levels were assessed in regions that were simultaneously stained for SYP/NCAM1 to define Adenocarcinoma (Adeno) or NEPC lesions in patient samples previously characterized by a clinical pathologist. Regions of interest were selected by observing SYP/NCAM1 positive or negative tumor areas and quantified with ImageJ. IHC staining was quantified using the Color deconvolution plugin on ImageJ on H DAB mode to obtain mean grey values of DAB channel. 5mC staining was quantified by selecting ROI using the DAPI channel with auto threshold and Analyze particles with 20–500 size and circularity 0.2–1.0 with masks. ROI were overlapped with 5mC channel to measure mean intensity values for each nucleus and find average intensity values per sample. % of positive nuclei staining was calculated as colocalization fraction between DAPI channel and channel of interest using JaCOP plugin in ImageJ or by manual counting of selected areas. Relative intensity of staining from nuclei center was performed by manually cropping individual cells on the channel of interest, outlining the nucleus and running Radial Profile plugin on ImageJ with a 1000-pixel radius. Intensity values were normalized to the sum of all values of the channel for each cell to find relative intensity distribution. All tissue slides were imaged with Zeiss LSM 710 NLO Confocal Microscope. Cell culture slides were imaged in a Nikon Super-Res Confocal Microscope with a high-resolution Galvano scanner.

Sphere formation assay

To determine the ability of cells to grow in soft agar, 2 × 104 cells were suspended in 0.3%agar in DMEM plus 10% FCS and overlaid on 0.5% agar in the same medium. Cells were re-fed with 10% FCS-containing medium every 5 days.

Clonogenic Assay

Single cell suspensions were prepared by trypsinization. Cells were washed with PBS and incubated with a 0.05% trypsin / EDTA solution for 5–10 minutes. Trypsin was quenched with 3 volumes of Dulbecco’s modified eagle medium containing 10% fetal bovine serum. The number of cells in each sample were counted using a hemocytometer and diluted such 1000 cells from each cell line/construct were seeded into petri dishes (three replicates of each in 35 mm wells). Cells were incubated for 2 weeks without media change. Crystal violet staining was performed by removing cell media, washing one time with PBS, fixing in 10% neutral buffered formalin for 20 minutes and staining in 0.01% (w/v) crystal violet in distilled water for 60 minutes.

Cell membranes fractionation

Confluent C42B sgC and sgPKCλ/ι cells, plated in 2 × 10 cm dishes, were rinsed once in cold PBS, then scraped, spun down and resuspended in 750 μl of fractionation buffer: 140 mM KCl, 1 mM EGTA, 2.5 mM MgCl₂, 50 mM sucrose, 20 mM HEPES, pH 7.4, supplemented with protease inhibitors. Cells were mechanically broken by spraying four to five times through a 23G needle attached to a 1 mL syringe, then spun down at 2,000g for 10 min, yielding a post nuclear supernatant (PNS). The PNS was further spun at maximum speed for 15 min in a tabletop refrigerated centrifuge, thus separating the cytosol (the supernatant) from the light organellar fraction (the pellet). The light organelle fraction was resuspended in 50/50 Laemmli/fractionation buffer, bringing it to an equal volume to the Laemmli-supplemented cytosol. Equal volumes of each fraction were gel-loaded and subjected to immunoblotting.

SAM quantification assay

Quantification of intracellular SAM levels was performed according to manufacturer’s instructions (Bridge-It S-Adenosyl Methionine (SAM) Fluorescence Assay, 96-well microplate format, Mediomics). In brief, 5×10⁴ cells were pelleted in PBS and resuspended in 30 μl of Buffer CM and incubated at 24 °C for 1hr, with occasional vortex. Samples were cleared by centrifugation at 4 °C 10,000 × g for 5 minutes and the supernatants were analyzed and quantified together with the SAM standard curve.

Gene-expression analyses

Total RNA was extracted by using TRIzol reagent (Invitrogen), and purified by using RNA miniprep kit (OmnigenX, OR3220–2) following the manufacturer’s protocol. After quantification using a Nanodrop 1000 spectrophotometer (Thermo Scientific), 1 g of RNA was reverse-transcribed using random primers and MultiScribe Reverse Transcriptase (Applied Biosystems). Gene expression was analyzed using the CFX96 Real Time PCR Detection System with SYBR Green Master Mix (BioRad). Primer sequences are listed in Table S1. The amplification parameters were set at 95 °C for 30 s, 58 °C for 30 s, and 72 °C for 30 s (40 cycles total). Gene expression values for each sample were normalized to the 18S rRNA or b-actin indistinctively. RNA-seq studies were performed in the Genomics Core at SBP Medical Discovery Institute. Briefly, total RNA was extracted from sgC, sgPKCλ/ι, and sgPKCλ/ι-R. PolyA RNA was isolated using the NEBNext® Poly(A) mRNA Magnetic Isolation Module and barcoded libraries were made using the NEBNext® Ultra™ Directional RNA Library Prep Kit for Illumina® (NEB, Ipswich MA). Libraries were pooled and single end sequenced (1X75) on the Illumina NextSeq 500 using the High output V2 kit (Illumina Inc., San Diego CA). Sequencing Fastq files were uploaded to BaseSpace and processed with RNAexpress App (Illumina) to obtain raw reads counts for each gene.

Bioinformatics analysis of gene expression

Microarray studies were performed in the Genomics and Microarray Laboratory at the Department of Environmental Health, University of Cincinnati Medical Center. Briefly, total RNA was extracted from 3 independent plates of PrEC siC and siPKCλ/ι and hybridized on Affymetrix mouse ST 1.0 microarrays. Scanning of the images and the first pass processing of probe-level fluorescence intensities was performed using the Microarray Suite 5.0 software (MAS 5.0; Affymterix, Santa Clara, CA). Gene expression tables (.gct) used as input for GSEA were created by processing CEL files with ExpressionFileCreator module hosted at GenePattern (https://genepattern.broadinstitute.org) with RMA method, using quantile normalization and background correction. Gene matrix files generated with RNAexpress app (Illumina) were modified to comply with .gct format and used as input file for Gene set enrichment analysis. GSEA was performed using GSEA v2.0.14 software (http://www.broadinstitute.org/gsea/index.jsp) with 5000 gene-set permutations using the gene-ranking metric T-test with Hallmark, C2, C3, C5 and C6 MSigDb collections as specified in each case. Heat-map representation of gene expression was generated using Morpheus (Broad Institute). Differentially expressed genes using Deseq2 with q-value<0.05 were considered to be significantly differentially expressed and uploaded to IPA (QIAGEN) to be analyzed using “Core Pathway Analysis”. Mutual exclusivity analysis of molecular alterations and PKCλ/ι expression in human patients (Robinson 2015 dataset) was performed using cBioPortal (http://www.cbioportal.org/) using samples with transcriptomic data (n = 118). Potential ATF4 targets in PCa (n = 3240 genes) were obtained from a comprehensive resource of predicted transcription factor (TF) targets and enhancer profiles in cancer, which integrates analysis of TCGA expression profiles and public ChIP-seq profiles (cistrome.org). 114 potential ATF4 targets were upregulated in sgPKCλ/ι cells compared to sgC. The list of 114 genes was used to perform unbiased pathway analysis on IPA.

Copy Number status with ploidy correction

Genomic status of select genes with ploidy correction was performed based on the CNA log2 scores as part of the NEPC cohort published in Beltran et al, Nat Med 2016 (http://download.cbioportal.org/nepc_wcm_2016.tar.gz). Ploidy values were obtained from the clinical data table provided as calculated by CLONET. For each log2 score, the score was rescaled via the following formula:

$$\mathit{u}\mathit{p}\mathit{d}\mathit{a}\mathit{t}\mathit{e}{\mathit{d}}_{\mathit{l}\mathit{o}\mathit{g}{2}_{score}}=\mathit{l}\mathit{o}{\mathit{g}}_{\mathit{C}\mathit{L}\mathit{O}\mathit{N}\mathit{E}{\mathit{T}}_{PLOID{Y}_{ESTIMATE}}}(\mathit{o}\mathit{r}\mathit{i}\mathit{g}\mathit{i}\mathit{n}\mathit{a}{\mathit{l}}_{\mathit{l}\mathit{o}\mathit{g}{2}_{score}})$$

Unsupervised clustering of patient samples

MSKCC PCa 2010 was directly downloaded from MSKCC. K-means (n=2) clustering was performed using gene expression values for the genes included in the previously published Integrated NEPC Signature (Beltran et al., 2016).

Medip-seq

Methylated DNA Immunoprecipitation was performed as follows. Total genomic DNA from samples was extracted using QIAamp DNA mini kit (QIAGEN) and eluted in DDW. 20 μg of total genomic DNA was diluted in 50 μl TE and sonicated using Covaris S2 with the following parameters (Duty 5%, Cycles/burst = 200, time=60s, 4 cycles using a microTUBE AFA Fiber Snap-Cap). Sonicated DNA was cleaned with Qiagen PCR clean UP (QIAGEN) and eluted in 30 μl of EB. 2 μg of sonicated genomic DNA was used to prepare libraries using NEBNext Ultra II DNA Library Prep kit (Illumina) following manufacturer’s instructions. Size-selected, end-repaired and adapter-ligated fragments (libraries) were denatured 10 min at 95C and immediately transferred to ice. 500 ng of library DNA were diluted in 500 μl of ice-cold IP buffer (20mM Tris-Cl, pH 8, 2mM EDTA, 1% Triton-X, 150 mM NaCl) with the addition of 1 μg of anti-5mC antibody (33D3, Diagenode) and incubated at 4 °C overnight. Same amount of library was kept aside for input sequencing and incorporated later to the Elution steps. IP pull down was performed with appropriate amount of Protein A Sepharose (SantaCruz) for 2 hr at 4 °C on orbital rotation. After Pull-down, Protein-A Sepharose beads were washed twice with IP Buffer, once with High Salt buffer (IP Buffer with 300 mM NaCl), twice with TE buffer and finally resuspended in 400 μl of Elution Buffer (25 mM Tris-Cl; pH 8.0, 10 mM EDTA, 0.5% SDS) with the addition of 10 μl of Proteinase K and samples were heated at 55C for 1 hr. After Elution, DNA was purified using PCR cleanup kit (QIAGEN) and eluted in 18 μl of EB. Library quality of IP 5mC outputs and inputs was analyzed with Bioanalyzer. DNA sequencing was done in a NextSeq 500 (Illumina).

Medip-seq data processing and analysis

Files were uploaded to usegalaxy.org server. After passing FASTQC analysis, raw FASTQ sequence quality files were converted with FASTQ Groomer and aligned to human genome (hg19) with BWA for Illumina with default parameters. Aligned SAM files were converted to BAM and sorted by chromosome position. Alignment statisitics was calculated using FlagStat. Peak Calling was performed using MACS2 peakcalling with the following parameters: Effective genome size=2451960000, Band width for picking regions to compute fragment size=300, Set lower mfold bound=5, Set upper mfold bound=50, Peak detection based on=q-value, Minimum FDR (q-value) cutoff for peak detection=0.01, Build Model=create_model, When set, scale the small sample up to the bigger sample=False, Use fixed background lambda as local lambda for every peak region=False, When set use a custom scaling ratio of ChIP/control (e.g. calculated using NCIS) for linear scaling=1.0, The small nearby region in basepairs to calculate dynamic lambda=1000, The large nearby region in basepairs to calculate dynamic lambda=10000, Composite broad regions=nobroad, How many duplicate tags at the exact same location are allowed=1. Bedgraph files from MACS2 output were converted to BigWig for Peak visualization. DMR were calculated using MACS2 bdgdiff using 5mC outputs and inputs for each sample with correction for sequencing depth and –g 60. Genomic Regions were annotated using EpiExplorer and NextBio using PileUp values as peak scores. Analysis of Overlap between bed files was performed using NextBio and the function Intersect at The Genomic Hyperbrowser as described in EpiExplorer documentation (epiexplorer.mpi-inf.mpg.de). Methylation profile analysis was performed using ComputeMatrix and PlotHeatmap packages within Deeptools with default options. Genomic coordinates for human Refseq genes were downloaded from UCSC Table browser. Transcription factor enrichment analysis was performed with HOMER.

Isotopic Labeling

sgC and sgPKCλ/ι C42B cells were cultured in RPMI 1640 medium including 10 mM [U-¹³C₆] glucose or 2 mM [α-¹⁵N]Glutamine, and 10% (v/v) dialyzed FBS for 24 and 48 hr prior to metabolite extraction. Labeling (corrected for natural abundance using in-house software) is depicted as isotopologue distributions or as labeled fraction of metabolites. Metabolite abundances were normalized to cell counts. Serine and glycine secretion rates are depicted as level of labeled secreted metabolite relative to integral of viable cells (IVC).

Gas Chromatography-Mass Spectrometry (GC-MS) Sample Preparation and Analysis

Cells were washed with saline solution and quenched with 0.5 mL −20 °C methanol. After adding 0.2 ml 4°C cold water, cells were collected in tube s containing 0.5 mL −20 °C chloroform. The extracts were vortexed for 10 min at 4°C and centri fuged at 16,000×g for 5 min at 4 °C. The upper aqueous phase was evaporated under vacuum at −4 °C. To determine extracellular metabolite levels, medium was centrifuged for 5 min at 300xg and 10 μl of supernatant was extracted using water/methanol/chloroform as described above. Derivatization for polar metabolites was performed using a Gerstel MPS with 15μl of 2% (w/v) methoxyamine hydrochloride (Thermo Scientific) in pyridine (incubated for 60 min at 45 °C) and 15 μl N-tertbutyldimethylsilyl-N-methyltrifluoroacetamide (MTBSTFA) with 1% tertbutyldimethylchlorosilane (Regis Technologies) (incubated further for 30 min at 45 °C). Polar derivatives were analyzed by GC-MS using a DB-35MS column (30 m × 0.25 i.d. × 0.25 μm) installed in an Agilent 7890A gas chromatograph (GC) interfaced with an Agilent 5975C mass spectrometer (MS) operating under electron impact ionization at 70 eV. The MS source was held at 230°C and the quadrupole at 150 °C and heli um was used as carrier gas. For MTBSTFA derivatized samples, the GC oven was held at 100 °C for 1 min, increased to 300 °C at 10 °C/min, and held at 325 °C for 3 min.

Analysis of methyl cytosine

Cells were grown in amino acid-free RPMI 1640 supplemented with all ¹²C amino acids and [methyl-¹³C]Methionine at standard RPMI with 10% FBS and 2 mM glutamine concentrations for 4 hr. Cells were rinsed with PBS, pellets collected by scraping and centrifugation, and then snap frozen in liquid nitrogen. Cellular biomass was collected and hydrolyzed for GC/MS analysis. Pellets were extracted by addition of 500 μL −80 °C methanol, 200 μL water, and 500 μL chloroform with subsequent pellet disruption in Restch Mixer Mill for 30 s at 30 Hz, vortexing for 10 s, and centrifugation at 20000 rpm for 10 min at 4 °C. Interfacial layer (containing nucleotide polymers) was isolated by discarding upper aqueous phase and lower organic phase from extraction and washing twice by addition of 500 μL methanol, vortexing, and discarding supernatant after centrifugation at 20000 rpm for 10 min at 4°C. Interfacial layer was dried ambiently overnight in fume hood. Nucleobases were isolated by acid hydrolysis for 2 hr at 80 °C with 2 M HCl and subsequently dried under ambien t air flow. Dried hydrolysate was resuspended in 500 μL 90% methanol and 50 μL redried for GC/MS analysis. Samples were derivatized with MTBSTFA and measured as described above. GC oven was held at 100 °C for 1 min after injection, increased to 255 °C at 3.5 ° C/min, and finally increased to 320 °C at 15 °C/min and held for 3 min.

BioID2-based Screening

LNCaP cells stably expressing myc-BioID2 or myc-BioID2-PKCλ/ι were grown for 48 hr in the presence of 50 μM biotin. Cells were lysed via sonication in 8 M urea, 50 μM ammonium bicarbonate, and extracted proteins were centrifuged at 14,000 × g to remove cellular debris and quantified by BCA assay (Thermo Scientific) according to the manufacturer’s specifications. A total of 700 μg of protein extract from each sample was used for affinity purification of biotinylated proteins. First, cysteine disulfide bonds were reduced with 5 mM tris(2-carboxyethyl)phosphine (TCEP) at 30 °C for 60 min f ollowed by cysteine alkylation with 15 mM iodoacetamide (IAA) in the dark at room temperature for 30 min. Affinity purification was carried out in a Bravo AssayMap platform (Agilent) using AssayMap streptavidin cartridges (Agilent). Briefly, cartridges were first primed with 50 mM ammonium bicarbonate, and then proteins were slowly loaded onto the streptavidin cartridge. Background contamination was removed by extensively washing the cartridges with 8 M urea, 50 mM ammonium bicarbonate. Finally, cartridges were washed with Rapid digestion buffer (Promega, Rapid digestion buffer kit) and proteins were subjected to on-cartridge digestion with mass spec grade Trypsin/Lys-C Rapid digestion enzyme (Promega, Madison, WI) at 70 °C for 2 hr. Digested peptides were then desalted in the Bravo platform using AssayMap C18 cartridges and the organic solvent was removed in a SpeedVac concentrator prior to LC-MS/MS analysis. Digested samples were analyzed on a Thermo Fisher Orbitrap Lumos mass spectrometer equipped with an Easy nLC 1200 ultra-high pressure liquid chromatography system. Samples were injected on an in-house packed C18 reverse phase column (25 cm × 75 um packed with 1.7 μm, 130 Å pore size Bridged Ethylene Hybrid particles (Waters)). Peptides were separated by an organic gradient from 5% to 30% ACN in 0.1% formic acid over 130 minutes at a flow rate of 300 nL/min. The MS continuously acquired spectra in a data-dependent manner throughout the gradient, acquiring a full scan in the Orbitrap (at 120,000 resolution with an AGC target of 1e6 and a maximum injection time of 100 ms) followed by as many MS/MS scans as could be acquired on the most abundant ions in 3 s in the Orbitrap (at 15,000 resolution, HCD collision energy of 30%, AGC target of 1e5, a maximum injection time of 22 ms, and an isolation width of 1.3 m/z). Singly and unassigned charge states were rejected. Dynamic exclusion was enabled with a repeat count of 1, an exclusion duration of 20 s, and an exclusion mass width of +/− 10 ppm. Raw mass spectrometry data were processed using the MaxQuant software package and Andromeda search engine (version 1.5.5.1). Peptides were generated from a tryptic digestion of up to two missed cleavages and were searched against the Uniprot human protein database (downloaded on August 13, 2015). Variable modifications were allowed for N-terminal protein acetylation, methionine oxidation, and lysine acetylation. A static modification was indicated for carbamidomethyl cysteine. All other settings were left using MaxQuant default settings. Peptide and protein identifications were filtered to a 1% false discovery rate (FDR). Statistical analysis was performed using the MSstats software package.

Analysis of Glucose and Lactate in culture media

Glucose and lactate concentrations were determined in 100 μl samples of medium using a YSI 2950 enzyme analyzer. Amounts in medium from cells were compared with amounts determined in medium incubated in parallel without cells and normalized to the total of cells for each time-point.

Protein Structure Analysis

The crystal structure of the human LAMTOR-RagA CTD-RagC CTD complex was downloaded from PDB (6EHR). Molecular graphics and analyses were performed with UCSF Chimera.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analyses for figures were performed using GraphPad Prism software (San Diego, CA). Data are presented as the mean ± SD unless otherwise specified. For qPCR experiments, Gaussian distribution was assumed and a Student’s t-test (two-tailed unpaired) was used to determine statistical significance. For human and mouse studies, significant differences between groups were determined using a Student’s t-test (two-tailed unpaired) when the data met the normal distribution tested by D’Agostino test. For statistical analysis of metabolite abundances and fraction labeling, significant differences were determined when both p values of sgPKCλ/ι vs. sgC and sgPKCλ/ι vs. sgPKCλ/ι-R comparisons were below p value<0.05 using Tukey’s multiple comparison test. For statistical analysis of radial profile distribution values, significance was determined by unadjusted multiple comparisons by t-test over 750 data points. All experiments were performed at least three independent times, unless otherwise noted. Gene expression correlation analyses were performed using Pearson’s correlation coefficients, with a two-tailed test. A log-rank (Mantel-Cox) test was used to evaluate statistical significance for the Kaplan-Meier survival plots. The significance level for statistical testing was set at p < 0.05. For animal xenograft studies, we used G*Power Data analysis to ensure adequate power to detect a pre-specified effect size. Specifically, based on preliminary qPCR data of E2F1 gene expression in sgC and sgPKCi C42B cells, as a surrogate for cell proliferation potential of these cells, we calculated sample size assuming power of 0.80 and 5% Type I error rate (alpha=0.05) with sampling ratio=1. No statistical method was used to predetermine any other sample sizes.

DATA AND SOFTWARE AVAILABILITY

The RNA-seq, Microarray, and Medip-seq data reported in this article have been deposited in NCBI GEO under the accession number GSE109763. Unprocessed original data have been deposited to Mendeley Data Repository and are available at http://dx.doi.org/10.17632/v4k68jpr9m.1

SIGNIFICANCE.

Acquired resistance to cancer treatment is a major setback to current therapies. In prostate cancer (PCa), this resistance is increasingly seen as a phenotypic transdifferentiation to a deadly form termed small cell/neuroendocrine prostate cancer (NEPC), for which there are no effective therapies available. Here, we have identified protein kinase C (PKC)λ/ι as a tumor suppressor in advanced PCa, whose loss promotes a metabolic reprogramming that prostate cancer cells exploit to sustain their increased proliferation and epigenetic needs, thus favoring cancer cell plasticity and NEPC differentiation. Importantly, our findings offer evidence of this metabolic axis being upregulated in human NEPC and provide a rationale to exploit this metabolic vulnerability in the clinic.

HIGHLIGHTS.

1. Loss of PKCλ/ι promotes basal and NEPC features in vivo

2. PKCλ/ι represses mTORC1 activation through LAMTOR2 phosphorylation

3. Loss of PKCλ/ι increases the SGOCP through mTORC1/ATF4 to fuel DNA methylation

4. The mTORC1/ATF4/PHGDH axis is a synthetic vulnerability of NEPC