O-GlcNAc transferase (OGT) is a nutrient-sensitive glycosyltransferase that is overexpressed in prostate cancer, the most common cancer in males. We recently developed a specific and potent inhibitor targeting this enzyme, and here, we report a synthetic lethality screen using this compound. Our screen identified pan-cyclin-dependent kinase (CDK) inhibitor AT7519 as lethal in combination with OGT inhibition. Follow-up chemical and genetic approaches identified CDK9 as the major target for synthetic lethality with OGT inhibition in prostate cancer cells. OGT expression is regulated through retention of the fourth intron in the gene and CDK9 inhibition blunted this regulatory mechanism. CDK9 phosphorylates carboxy-terminal domain (CTD) of RNA Polymerase II to promote transcription elongation. We show that OGT inhibition augments effects of CDK9 inhibitors on CTD phosphorylation and general transcription. Finally, the combined inhibition of both OGT and CDK9 blocked growth of organoids derived from patients with metastatic prostate cancer, but had minimal effects on normal prostate spheroids. We report a novel synthetic lethal interaction between inhibitors of OGT and CDK9 that specifically kills prostate cancer cells, but not normal cells. Our study highlights the potential of combining OGT inhibitors with other treatments to exploit cancer-specific vulnerabilities.
The primary contribution of OGT to cell proliferation is unknown, and in this study, we used a compound screen to indicate that OGT and CDK9 collaborate to sustain a cancer cell–specific pro-proliferative program. A better understanding of how OGT and CDK9 cross-talk will refine our understanding of this novel synthetic lethal interaction.
O-GlcNAc transferase (OGT) is a nutrient-sensitive glycosyltransferase that modifies serine and threonine residues of nuclear and cytoplasmic proteins via single sugar conjugation to regulate their functions (1–3). Analogous to phosphorylation, this sugar mark can be removed by an enzyme termed O-GlcNAcase (OGA). However, unlike phosphorylation, which is regulated by hundreds of kinases, OGT is solely responsible for all nucleocytoplasmic glycosylation. As such, this enzyme is a unique sensor and regulator of cell function.
OGT and OGT's catalytic product, O-GlcNAcylation, are increased in many cancers, including prostate cancer, which is the most common cancer in males in the United States (4). Both OGT and O-GlcNAc levels correlate with aggressive disease (5, 6), and OGT knockout results in a complete loss of proliferation (7–9). Development of specific OGT inhibitors is essential for probing the functions of this unique enzyme.
We recently developed a family of OGT small-molecule inhibitors (OSMI-2 and OSMI-4) that bind in the OGT's active site with nanomolar dissociation constants (10). OSMI-2 and OSMI-4 reduce O-GlcNAc levels in multiple cell lines, but cells also rapidly respond by increasing OGT expression to compensate. Cells that are able to compensate for inhibition have limited overall transcriptomic and proteomic changes, providing evidence that these inhibitors have minimal off-target effects (10, 11). There are modest growth-suppressive effects on cells; however, we wondered whether the adaptations needed to respond to OGT inhibition may sensitize cancer cells to disruption of other cellular pathways.
In this study, we screened 5,000 bioactive compounds for synergism with OSMI-2 and identified the compound AT7519 as synthetically lethal with OGT inhibition in prostate cancer cells. AT7519 is a pan-cyclin-dependent kinase (CDK) inhibitor, and here, we show that reducing CDK9 activity is sufficient for synthetic lethality. The OGT inhibitor potentiated the effects of several structurally unrelated, clinically relevant CDK9 inhibitors on RNA Polymerase II (RNA Pol II) carboxy-terminal domain (CTD) phosphorylation, transcription, and antiproliferative effects on prostate cancer organoids. To conclude, prostate cancer cells become dependent on high OGT activity when CDK9 is inhibited, and our study proposes that OGT can be targeted to sensitize cancer cells to other treatments.
Materials and Methods
Cell culture and statistical analysis
LNCaP, C4-2, 22RV1, PC3, HCT116, MDAMB231, and RWPE-1 cells were obtained from ATCC and PNT2 cells were obtained from Sigma, and all cell lines were maintained as recommended by the provider. Cells were authenticated using short tandem repeat profiling performed by ATCC. Mycoplasma testing was performed regularly, using the colorimetric MycoAlert Mycoplasma Detection Kit (Lonza). LNCaP-95 cells and MSK-PCA3 (12) model were kindly provided by Drs. M. Brown (Dana-Farber Cancer Institute, Boston, MA) and Yu Chen (Memorial Sloan Kettering Cancer Center, New York, NY), respectively. LNCaP-95 cells were maintained in 10% charcoal-stripped serum in phenol red-free RPMI. MSK-PCA3 cells were maintained as reported previously (13). Organoid experiments were performed in 12-well plates and all the experiments represent at least two biological replicates. Organoids were allowed to form in Matrigel domes for 3–10 days prior to treatment. After this, media were changed every 4–7 days, coinciding with imaging using a Zeiss AxioObserver inverted widefield microscope (Plan-Apochromat 20×/0.8). The growth of the cells in Matrigel was measured manually and is presented as change in the organoid size relative to the start of the treatments. For all statistical analysis presented in the article, P values are from two-sided analysis unless otherwise specified. All the main figure experiments were performed in LNCaP cell line, unless otherwise noted.
Compounds and assays
OSMI-2 and OSMI-4 were synthesized in-house (10). AT7519, LDC000067, palbociclib, RO3306, and PHA848125 were purchased from Selleckchem. Staurosporine was purchased from Abcam. Dinaciclib was from Axon Medchem. NVP2 and SB1317 were obtained from MedChemExpress. ATP levels in cells were assessed using the CellTiter-Glo Luminescent Cell Viability Assay (Promega). Colony-forming assays were performed using the CytoSelect 384-well Cell Transformation Assay (BioCat) according to the manufacturer's instructions. Growth rate and cell death activation were evaluated using the Incucyte instrument according to the manufacturer's instructions. For detection of cell death activation, we used IncuCyte Caspase-3/7 Green Reagent for apoptosis (Essen Biosciences). Cell death activation for Fig. 2E; Supplementary Fig. S3 was performed using the ApoTox-Glo Triplex Assay (Promega) according to the manufacturer's instructions (the signal from caspase activity was normalized to signal from viable cells). Cycloheximide was used as a positive control to induce cell death. Knockdown experiments were performed using RNAiMax Reagent (Sigma). OGT targeting siRNAs were from Thermo Fisher Scientific: siOGT_1 s16094 and siOGT_2 s16095, and CDK9 targeting siRNA was from Qiagen (SI00024423).
Protein and mRNA profiling
Samples for Western blotting were prepared as reported previously (5). Antibodies used are as follows: from Cell Signaling Technology, Cl-PARP (9541), p-S2/S5-RNA Pol II (4735), p-S2-RNA Pol II (13499), CDK9 (2316), and OGT (24083). OGA (HPA036141) antibody was obtained from Sigma. Actin (ab49900) and RL2 (ab2739) antibodies were from Abcam.
For reverse-transcription quantitative polymerase chain reaction (RT-qPCR) and RNA-sequencing (RNA-seq), RNA isolation was performed using the illustra Mini Spin Kit (GE Healthcare) according to the manufacturer's instructions. cDNA was synthesized using the qScript cDNA Synthesis Kit (Quantabio). For spike-in RNA normalization, luciferase mRNA was obtained from Promega (L4561).
LNCaP cells were plated into a 384-well plate (1,000 cells/well). The next day, cells were treated with DMSO or 40 μmol/L OSMI-2 and compounds from the compound library were separately added to each well. Most of the compounds were screened using a single dose, as predefined by the ICCB-Longwood Screening Facility (Boston, MA). Compounds with well-defined target were often screened in multiple doses, as predetermined by the ICCB Facility. After 3 days, cells were simultaneously fixed and stained with 4% formaldehyde and Hoechst, respectively. Total fluorescence was recorded with Acumen Laser Scanning Cytometer. The quality of the screen was assessed by calculating the Z'-factor (14) for each plate and was consistently between 0.5 and 1. All treatments were first normalized to DMSO- or OSMI-2–only treatment and are presented as percent of either.
RNA-seq data analysis
Libraries for sequencing were prepared by the Norwegian High Throughput Sequencing Centre using Strand-specific TruSeq RNA-seq library prep kit. Two biological replicates of each condition were used for the analysis. Samples were multiplexed and paired-end sequenced in four lanes with 150-bp read length. Sequencing reads were aligned to Hg19 (GRCh37) using STAR (15). Raw counts of reads were mapped to genes using HTSeq counts (http://htseq.readthedocs.io/en/master/count.html) and differential expression analysis was performed on duplicate samples using DESEQ2 (16). Genes with a P value of 0.01 were taken forward for fold change analysis. RNA-seq BAM files were normalized and converted to BigWig for visualization on IGV 2.0, experimental replicate data tracks were combined, and group autoscale was used to enable comparable visualization between samples. RNA-seq data has been deposited to the Gene Expression Omnibus (GSE116778).
Comparison of change in mRNA abundance versus mRNA half-life
Log2 fold change and Benjamin–Hochberg adjusted P values obtained from DESeq2 fold change analysis were used to generate volcano plots, and plotted using matplotlib, colored according to half-life based upon data from Schwanhäusser and colleagues (17) mouse homologs. Only genes with homologs and half-life data were included (4,212 in total, with 137 genes with mRNA half-life below 4 hours).
Discovery of synthetic lethality between OSMI-2 and a pan-CDK inhibitor AT7519
Both OGT and O-GlcNAc levels are increased in aggressive prostate cancer (5, 6), but it is not known whether OGT coordinates with specific cellular processes to drive cancer progression. In LNCaP prostate cancer cells, the recently developed OGT inhibitor, OSMI-2 (10), dose dependently decreased total O-GlcNAc and reduced ATP levels by 50%, but only modestly affected the proliferation rate (Fig. 1A–C). OSMI-2 treatment also resulted in a prominent increase in OGT protein levels, a response previously reported for OGT inhibition (5, 10, 18). This response represents one of the mechanisms to restore O-GlcNAc homeostasis (18). The reduction in ATP content without a corresponding decrease in cell number implied metabolic adaptations in inhibited cells that sustain proliferation.
To test whether OGT inhibition sensitizes prostate cancer cells to inhibition of other pathways, we screened 5,000 biologically active compounds in LNCaP cells with and without cotreatment of OGT inhibitor, OSMI-2. The screen was based on cell number as detected via DNA stain. Our strongest hit was AT7519, a pan-CDK inhibitor (Fig. 1D and E; Supplementary Fig. S1A).
We confirmed that OGT inhibition sensitizes LNCaP cells to AT7519 using multiple assays. Live-cell imaging showed that treatment with OSMI-2 and AT7519 blocked the proliferation of LNCaP cells (Fig. 1F; Supplementary Fig. S1B). Soft agar colony formation assays demonstrated that OSMI-2 combined with AT7519 abolished the ability of LNCaP cells to form colonies (Fig. 2A and B). Knockdown of OGT and treatment with AT7519 significantly further decreased ATP levels when compared with any single treatment (Supplementary Fig. S1C). We confirmed the antiproliferative effects of the OGT inhibitor–AT7519 combination in another prostate cancer cell line (PC3), in triple-negative breast cancer cell line (MDAMB231), and in the colon cancer cell line (HCT116; Supplementary Fig. S2A, S2B and S2C). These data show that synthetic lethality between OGT inhibition and AT7519 is observed in multiple cancer models.
To better understand the basis for the combined antiproliferative effects of OSMI-2 and AT7519, we evaluated targets that are affected by AT7519 (19, 20). OSMI-2 treatment significantly enhanced the AT7519-induced decrease in the phosphorylation of protein phosphatase 1 (PP1α), nucleophosmin 1 (NPM1), and RNA Pol II's CTD), which are targets of CDK1, CDK2, and CDK9, respectively (Fig. 2C). These effects implied that cell cycle is affected. Combination of AT7519 with OSMI-2 resulted in the accumulation of cells in G2–M-phase and decreased the S-phase population by more than 50% at 24 hours (Fig. 2D). In addition, the sub-G1-phase population of cells increased for the combination treatment already at 24 hours, and was increased by 5-fold after 48 hours, indicating activation of the cell death response. To confirm cell death activation, we evaluated activation of caspases 3/7 in prostate cancer cells (LNCaP and PC3) and in normal prostate cells (RWPE-1 and PNT2). Strikingly, OSMI-2 potentiated AT7519-induced cell death activation up to 8-fold in prostate cancer cells, but the combination did not induce cell death in normal cells (Fig. 2E; Supplementary Fig. S3). We further confirmed cell death induction by showing a 2-fold increase in PARP cleavage in response to OSMI-2–AT7519 combination (Fig. 2C).
Taken together, we have identified the CDK inhibitor, AT7519, and the OGT inhibitor, OSMI-2, as an antiproliferative combination against cancer cells. Interestingly, OSMI-2 significantly enhanced the effects of AT7519 on its known targets. The observed antiproliferative effects are explained, at least in part, by activation of the cell death response in cancer cells, while normal prostate cells do not undergo apoptosis under the same treatment. To identify the processes that lead to cell death activation, we must focus on events prior to caspase activation.
Inhibition of RNA Pol II CTD kinase CDK9 is lethal in combination with OGT inhibitors
We used transcriptional profiling to probe the basis for synthetic lethality between OGT inhibition and AT7519. Combination treatment with OSMI-2 and AT7519 induced defects in proliferation already at 12 hours after the treatment and cell death was prominent after 24 hours (Figs. 1F and 2C). Importantly, OSMI-2 decreased O-GlcNAcylation at 4 hours (Supplementary Fig. S4A), thereby providing us with a window of opportunity before decreased proliferation and activation of apoptosis, in which we could identify the key events leading to toxicity in cancer cells. As a single agent, short-term treatment with OSMI-2 caused downregulation of pathways related to cell cycle, in agreement with previous reports on OGT inhibitor effects (ref. 5, 21; Supplementary Fig. S4B). We also noted that OGT inhibition decreases phosphorylation of RNA Pol II (Supplementary Fig. S5).
OGT expression is regulated through a detained intron-dependent mechanism (18, 22) and we assessed whether this can be visualized using RNA-seq. Polyadenylated OGT mRNA still contains one intron, intron 4, termed “detained intron” (DI; ref. 22). This intron is rapidly spliced away to enable production of a mature OGT mRNA in a transcription-independent manner (18). As expected, treatment with OGT inhibitor, OSMI-2, led to a depletion of reads mapping to intron 4 (DI4), representing generation of the productive, translation-competent OGT mRNA and predicting upregulation of the OGT protein (Fig. 3A). Unexpectedly, AT7519 treatment completely blocked OSMI-2–induced depletion of reads mapping to intron 4, implying defects in OGT regulation in response to OSMI-2 treatment. We confirmed that AT7519 blocks OSMI-2–induced processing of the OGT mRNA and also blunts upregulation of the OGT at the protein level in both LNCaP and PC3 cells (Fig. 3B and C; Supplementary Fig. S6). As expected, on the basis of cells' inability to upregulate OGT protein, AT7519 treatment also enhanced the effect of OSMI-2 on total O-GlcNAc, and we moved on to further explore the RNA-seq data.
Single-agent treatment with AT7519 primarily led to downregulation of mRNAs (80% of transcripts; Supplementary Fig. S7A). Combining OSMI-2 with AT7519 increased the number of downregulated genes by almost 400. The top gene ontology (GO) terms under AT7519 treatment were related to gene expression; the same processes were even more enriched in the combination treatment (Fig. 4A; Supplementary Fig. S7B). Notably, these ontologies are similar to those observed upon CDK9 inhibition in prior studies (23, 24). Because AT7519 potently inhibits CDK9 (19), we hypothesized that targeting CDK9 is sufficient to induce synthetic lethality with OGT inhibition.
NVP2 is a CDK9 inhibitor with >700-fold more selectivity for CDK9 than any other kinase (23). Combining either OSMI-2 or OSMI-4, a closely related but more potent OGT inhibitor that became available when we were developing this project (10), with 5–10 nanomolar NVP2 resulted in a complete loss of cellular proliferation, depleted ATP levels, and over doubled the activation of cell death response when compared with any single treatment (Fig. 4B and C; Supplementary Fig. S8). On the other hand, OSMI-2 did not potentiate antiproliferative effects of small-molecule inhibitors targeting CDK1, CDK2, or CDK4/6 (Fig. 4C). NVP2 is a highly specific CDK9 inhibitor; however, to assess whether OGT inhibition sensitizes prostate cancer cells to clinically relevant compounds, we used dinaciclib and SB1317, two CDK9 inhibitors that have been tested in phase I–III clinical trials (25). Importantly, OGT inhibition also sensitized LNCaP and PC3 prostate cancer cells to these CDK9 inhibitors (Fig. 4D; Supplementary Fig. S9).We also used siRNA against CDK9 to confirm the importance of CDK9 for the antiproliferative effects in combination with OGT inhibition (Supplementary Fig. S10). Taken together, targeting CDK9 is sufficient to induce anti-prostate cancer effects in combination with OGT inhibitors.
Targeting OGT potentiates CDK9 inhibitor–induced defects in transcription
We hypothesized that targeting OGT potentiates CDK9 inhibitor effects on cancer cell proliferation due to defects in transcription. OGT glycosylates RNA Pol II CTD on Ser-5 and Ser-7, and OGT inhibition blocks preinitiation complex formation in in vitro assays (26), while CDK9 phosphorylates RNA Pol II CTD on Ser-2 to promote productive elongation (25). When RNA Pol II activity is inhibited, all mRNA species start to decay, but mRNAs with short half-lives are lost more rapidly (27). If our hypothesis is correct, combining OSMI-2 with a low dose of CDK9 inhibitor should lead to a stronger decline in short half-live mRNAs than targeting CDK9 alone. To test whether transcription is decreased globally, we compared previously reported genome-wide mRNA half-life data (17) to our RNA-seq dataset. On the basis of the data reported by Schwanhausser and colleagues (2011; ref. 17), we divided all mRNAs into two groups: mRNAs with half-lives less than 4 hours (the timepoint of RNA collection in our RNA-seq experiment) and those with longer half-lives. AT7519 decreased the abundance of almost all short half-life mRNAs (P < 0.003 between AT7519 and OSMI-2; Fig. 5A), consistent with the importance of CDK9 in promoting RNA Pol II activity (23, 24), and validating our experimental approach. OSMI-2 strongly enhanced the AT7519-induced effect on short half-life mRNAs (P < 0.0009 between AT7519-only and AT7519+OSMI-2; Fig. 5A; Supplementary Fig. S11).
Genes driven by super-enhancers are hypersensitive to decrease in RNA Pol II activity, and the prototypical example of these genes is MYC, a transcription factor that is deregulated in over half of human cancers (28), and known to be downregulated in response to CDK9 inhibition in prostate cancer cells (29). Combining a low dose of any of the three structurally divergent, clinically relevant CDK9 inhibitors with OGT inhibitor led to an up to 90% decline in the levels of MYC mRNA, and further enhanced the effects of any single treatment by more than 50% (Fig. 5B). In addition, all of these CDK9 inhibitors blocked OSMI-2–induced splicing and upregulation of the OGT mRNA (Fig. 5C). Finally, combining a low dose of any of the three structurally divergent CDK9 inhibitors with OGT inhibitor led to a complete loss of RNA Pol II CTD Ser-2 phosphorylation, the site modified by CDK9 (Fig. 5D). To conclude, small-molecule inhibition of OGT augments CDK9 inhibitor effects on transcription, which explains the anticancer effects of this combination (Fig. 5E). However, at the same time, this raises concerns of on-target negative effects on normal cells.
Combined inhibition of OGT and CDK9 is lethal to prostate cancer organoids
We used organoid tissue culture system to evaluate whether combined inhibition of OGT and CDK9 has potential as anticancer therapy. Organoid tissue culture recapitulates structural and functional aspects of the in vivo counterpart (30), and prostate cancer organoids retain both epigenomic and transcriptomic concordance with their corresponding tumors in patients (13). These features make organoids an excellent tool for validation of candidate therapies. Castration-resistant prostate cancer (CRPC) is a major challenge in clinical setting and we used three CRPC spheroid models to show that OGT inhibition potentiates the antiproliferative effects of CDK9 inhibition in these cells (Fig. 6A; Supplementary Figs. S12 and S13). In addition, combination of OGT and CDK9 inhibitors led to an almost complete loss of proliferation of prostate cancer organoids derived from a patient with metastatic prostate cancer (Fig. 6B). Importantly, combined inhibition of OGT and CDK9 was not toxic to normal prostate cells that grow at similar rate as LNCaP-95 and C4-2 spheroids in the 3D tissue culture system (Fig. 6C). We also noted that OGT inhibition shows strong antiproliferative effects on prostate cancer cells as a single agent. These data validate the results generated in the conventional tissue culture system, and confirm that combined inhibition of OGT and CDK9 has antiproliferative effects specifically on prostate cancer, but not on normal prostate cells.
Our entry point to this project was sensitization screen that led to the discovery of synthetic lethal interaction between compounds targeting OGT and CDK9. On the basis of our data and existing literature, we propose that the observed synthetic lethality between inhibitors of OGT and CDK9 is specific to cancer cells that are dependent on high levels of transcription (Fig. 5E). Inhibition of OGT blocks RNA Pol II entry into the promoters (31), and targeting OGA inhibits transcription elongation in nuclear extract system (32). Remarkably, OGA inhibitors are progressing to clinical trials (33), and it may be possible to induce similar synthetic lethality between inhibitors targeting OGA and CDK9, as we report here for simultaneous inhibition of OGT and CDK9. Our model also proposes that inhibitors targeting either CDK7 or CDK12, the two other major CTD kinases, should also be lethal in combination with OGT inhibitors.
OGT inhibition enhances the effects induced by targeting CDK9. High dose OGT inhibition decreases phosphorylation of RNA Pol II on the CDK9-targeted Ser-2 (Supplementary Fig. S5), and by combining low dose OGT and CDK9 inhibitors, we see a 5-fold decrease in the Ser-2 phosphorylated form of RNA Pol II (Figs. 2C and 5D). We show that targeting OGT enhances the effects induced by CDK9 inhibitors on highly transcribed genes such as c-MYC, but single-agent inhibition of OGT has only modest effects on transcription (Fig. 5B). Previously, we showed that OGT also posttranslationally controls the levels of c-MYC (5, 34), and OGT can thereby control the levels of key drivers of proliferation by affecting both transcription and protein stability.
The mechanistic underpinnings of the role of OGT in regulating RNA Pol II activity have not been comprehensively described. OGT may prime CTD phosphorylation during transcription initiation or provide hindrance for unscheduled phosphorylation, as phosphorylation and glycosylation are mutually exclusive on the same residue, and O-GlcNAc removal is an ATP-dependent step during transcription initiation (31, 35). To formally test this, one would have to trap the RNA Pol II complexes at the gene promoters/gene body in a synchronized manner, and detect the presence of glycosylation and phosphorylation in a site- and a location-specific manner, an experiment that goes beyond the scope of this article: discovery of the synthetic lethal interaction of combined inhibition of OGT and other cellular processes. To conclude, OGT and CDK9 act in consecutive steps during transcription initiation, and simultaneous inhibition of both leads to a further decrease in transcription, which we propose as the anticancer mechanism of action for this combination (Fig. 5E).
In characterizing the observed synthetic lethal interaction, we noted that inhibition of CDK9 activity blocks splicing of the OGT DI (Figs. 3A, B and 5C). This is the first time that CDK9 is linked to regulation of DIs, and raises the possibility that global dysregulation of DI-containing genes is a hallmark of cancers exhibiting altered activity of CDK9.
CDK9 inhibitors are being evaluated as cancer therapy, and our work showed that OGT inhibition can potentiate antiproliferative effects of this promising class of therapeutics while sparing noncancerous cells. Specific OGT inhibitors used in this study induce antiproliferative effects selectively on prostate cancer organoids, and we propose that OGT inhibitors can be used to prime cancer cells to other targeted therapies.
Disclosure of Potential Conflicts of Interest
H.M. Itkonen reports grants from Norwegian Cancer Society during the conduct of the study. S.E.S. Martin reports a patent for WO2020047251A1 pending. C.J. Thomas reports a patent for U.S. Patent App. # 61/217,514 pending. S. Walker reports grants from Norwegian Cancer Society and NIH during the conduct of the study, as well as a patent for O-glcnac transferase inhibitors and uses thereof issued. No potential conflicts of interest were disclosed by the other authors.
H.M. Itkonen: Conceptualization, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing. N. Poulose: Validation, investigation, writing-review and editing. R.E. Steele: Data curation, software, formal analysis, investigation, visualization, writing-review and editing. S.E.S. Martin: Investigation, writing-review and editing, synthesis of OSMI-compounds. Z.G. Levine: Data curation, software, formal analysis, validation, investigation, visualization, methodology, writing-review and editing. D.Y. Duveau: Investigation, writing-review and editing, synthesis of OSMI-compounds. R. Carelli: Validation, visualization, writing-review and editing. R. Singh: Validation, writing-review and editing. A. Urbanucci: Data curation, investigation, writing-review and editing. M. Loda: Resources, validation, writing-review and editing. C.J. Thomas: Resources, supervision, validation, investigation, writing-review and editing. I.G. Mills: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, writing-original draft, writing-review and editing. S. Walker: Conceptualization, resources, supervision, funding acquisition, investigation, writing-original draft, writing-review and editing.
Authors are grateful to members of the Walker laboratory and Dr. Brian Lewis (NIH) for the critical feedback on the article. We also thank the ICCB-Longwood Screening Facility for the assistance in the screen. We are grateful for the support that was received to complete this work as follows: HMI for the Norwegian Cancer Society travel Fellowships (ID 159970 – 2014 and ID 181596 – 2016), SESM for the NIH (F32 GM117704), AU for the Norwegian Cancer Society (198016-2018), M. Loda's work was supported by NIH grants (RO1CA131945, R01CA187918, DoD PC160357, DoD PC180582, and P50CA211024), and the Prostate Cancer Foundation, C.J. Thomas received support from the intramural programs of the Center for Cancer Research, NCI, and the Division of Preclinical Innovation, National Center for Advancing Translational Sciences of that NIH, I.G. Mills is a member of the Prostate Cancer UK/Movember Centre of Excellence (CEO13_2–004 to R.E. Steele) and also supported by the John Black Charitable Foundation and the Norwegian Research Council (230559) and Norwegian Cancer Society (Project nr. 4521627), and S. Walker was supported by the NIH R01 GM094263.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.