High OGT activity is essential for MYC-driven proliferation of prostate cancer cells.

O-GlcNAc transferase (OGT) is overexpressed in aggressive prostate cancer. OGT modifies intra-cellular proteins via single sugar conjugation (O-GlcNAcylation) to alter their activity. We recently discovered the first fast-acting OGT inhibitor OSMI-2. Here, we probe the stability and function of the chromatin O-GlcNAc and identify transcription factors that coordinate with OGT to promote proliferation of prostate cancer cells.
Methods: Chromatin immunoprecipitation (ChIP) coupled to sequencing (seq), formaldehyde-assisted isolation of regulatory elements, RNA-seq and reverse-phase protein arrays (RPPA) were used to study the importance of OGT for chromatin structure and transcription. Mass spectrometry, western blot, RT-qPCR, cell cycle analysis and viability assays were used to establish the role of OGT for MYC-related processes. Prostate cancer patient data profiled for both mRNA and protein levels were used to validate findings.
Results: We show for the first time that OGT inhibition leads to a rapid loss of O-GlcNAc chromatin mark. O-GlcNAc ChIP-seq regions overlap with super-enhancers (SE) and MYC binding sites. OGT inhibition leads to down-regulation of SE-dependent genes. We establish the first O-GlcNAc chromatin consensus motif, which we use as a bait for mass spectrometry. By combining the proteomic data from oligonucleotide enrichment with O-GlcNAc and MYC ChIP-mass spectrometry, we identify host cell factor 1 (HCF-1) as an interaction partner of MYC. Inhibition of OGT disrupts this interaction and compromises MYC's ability to confer androgen-independent proliferation to prostate cancer cells. We show that OGT is required for MYC-mediated stabilization of mitotic proteins, including Cyclin B1, and/or the increased translation of their coding transcripts. This implies that increased expression of mRNA is not always required to achieve increased protein expression and confer aggressive phenotype. Indeed, high expression of Cyclin B1 protein has strong predictive value in prostate cancer patients (p=0.000014) while mRNA does not. Conclusions: OGT promotes SE-dependent gene expression. OGT activity is required for the interaction between MYC and HCF-1 and expression of MYC-regulated mitotic proteins. These features render OGT essential for the androgen-independent, MYC-driven proliferation of prostate cancer cells. Androgen-independency is the major mechanism of prostate cancer progression, and our study identifies OGT as an essential mediator in this process.

Introduction

It is well known that metabolites can affect cell fate by regulating transcription. This was initially established for the lactose-regulated lac operon in bacteria and the galactose-regulated gal regulome in yeast 1. Metabolite-dependent regulation of transcription in multicellular organisms has also been established, and as an example, sterol synthesis is regulated by sterol regulatory element-binding protein 1 (SREBP1) 2. In the absence of sterols, SREBP1 is cleaved, and this cleavage generates an active nuclear transcription factor that promotes expression of genes involved in sterol biosynthesis.

Altered metabolite levels are a prominent feature of cancer cells/tumors and have been directly linked to their ability to proliferate rapidly 3, 4. One of the most prominent features of cancer cells is the 'Warburg effect', where cells exhibit increased glucose uptake and fermentation of glucose to lactate even in the presence of fully functional mitochondria and oxygen. Further underlining the increased appetite of cancer cells for glucose, this metabolite is frequently used as a tracer to localize tumors and follow response to therapy using PET-imaging 5. In the case of prostate cancer, which is lipogenic, tumors can also be traced using 11C-Acetate 6, 7. Cancer cells adapt to metabolite availability through transcriptional and translational regulation; however, post-translational modifications provide a rapid response to extracellular stimuli.

O-GlcNAc transferase (OGT), the sole enzyme in the human genome that modifies target protein serine and threonine residues with a single O-GlcNAc sugar, is a major metabolic integration point in cells 8. OGT's substrate, UDP-GlcNAc, is produced via the hexosamine biosynthetic pathway (HBP), which uses glucose, glutamine, acetyl-CoA and UTP as building blocks 9. Flux through the HBP increases in response to glucose, which leads to increased O-GlcNAcylation of nuclear proteins 10. Among the most prominent O-GlcNAc modified factors is the C-terminal domain of RNA-Polymerase II, which regulates pre-initiation complex formation 11-13. Increased expression of OGT has been demonstrated in most cancers, including prostate cancer, where high protein-O-GlcNAc levels have been shown to correlate with poor clinical outcome 14-16.

Prostate cancer is the most common male cancer, and the androgen receptor (AR) is the major driver and drug target in this disease 17, 18. The AR is typically expressed at high levels in prostate cancer and positively regulates the HBP 18-20. One mechanism by which increased HBP flux, and especially OGT activity, benefits cancer cells, is through stabilization of pro-oncogenic proteins such as MYC 14, 21, 22.

MYC is a transcription factor that is essential for the proliferation of most cancer cells 23. The protein binds to MAX to form a complex, which, in general, positively regulates transcription 24. MYC binds to a DNA sequence-motif known as the E-box 25. E-boxes occur every 4kb in the human genome, but MYC binds only to the subset of these sites that are non-methylated and accessible 26. When MYC levels increase, as they do in tumors, there is increased binding to lower affinity E-boxes 27. Increased MYC expression leads to enhanced expression of the cell-specific transcriptional program rather than activation of new genes 28, 29.

OGT and MYC work together to enhance cancer cell proliferation 14, 21, 22. Inhibition of OGT activity down-regulates MYC in prostate cancer cells 30, while increased glucose levels lead to OGT-mediated stabilization of MYC in liver cancer cells 21. Cells infected with human papillomavirus, the causative agent in genital cancers, have high levels of OGT, which promotes MYC stability 31. In activated T cells, MYC affects O-GlcNAcylation through regulation of glucose and glutamine transport 22. These data highlight the complex interplay between OGT and MYC in cancer cells. It is currently not known if OGT affects the MYC-dependent cellular program.

Here, we show that inhibition of OGT activity disrupts the interaction between MYC and HCF-1, a known transcriptional co-activator. Inhibition of OGT enhances MYC's ability to suppress transcription of certain cell cycle genes and at the same time enhances the expression of genes involved in the degradation of cell cycle proteins. We show that OGT is required for MYC-mediated stabilization of proteins involved in mitosis. MYC's dependency on OGT becomes most evident during androgen-deprivation, a condition in which overexpression of MYC can drive proliferation. In addition, we show that OGT regulates super-enhancer-dependent transcription through chromatin compaction. Our study is the first one to promote combination of OGT inhibitor with AR-targeted therapies as a strategy for castration-resistant prostate cancer.

Methods

Cell culture conditions and treatments

LNCaP and PC3 cell lines were obtained from the ATCC and maintained according to ATCC guidelines. LNCaP and PC3 cells were grown in RPMI media supplemented with 10% FBS. For hormone starvation, cells were cultured in phenol red-free RMPI supplemented with 10% charcoal-stripped FBS. LNCaP-MYC cells were maintained as described previously 32 and MYC overexpression was induced using 2μg/ml doxycycline (Sigma). Tamoxifen-inducible OGT knockout mouse embryonic fibroblast (MEF) cell line was obtained from Dr. Natasha Zachara at the CardioPEG CoreC4 (NHLBI P01 HL107153) at Johns Hopkins University School of Medicine 33. Cell viability was assessed using CellTiterGlo-reagent (CTG, Promega) according to the manufacturer's instructions. Each CTG-experiment was done using 384-well plates and consisted of at least 3 technical and 3 biological replicates. PUGNAc, Thiamet-G, cOmplete™ EDTA-free Protease Inhibitor Cocktail and PhosSTOP™ were purchased from Sigma. (Z)-4-Hydroxytamoxifen was obtained from Abcam (ab141943). The second-generation OGT inhibitors OSMI-2 and OSMI-4 have been described 34. Cell cycle distribution was analyzed using the Propidium Iodide Flow Cytometry Kit (Abcam, ab139418) and the BD FACSCanto instrument (BD Biosciences). Transfections for siRNA experiments were performed using RNAiMax reagent (Sigma), and OGT targeting siRNAs were obtained from Life Technologies (s16094 and s16095).

Preparation of cell lysates

Preparation of cell lysates for western blotting (WB) was done as previously described 19. Antibodies were obtained from Cell Signaling Technology: MYC #5605, OGT #24083, CDK1 #9116, Cyclin B1 #4138, PLK1 #4513, p-H3ser10 #9701, H3 #4499, p-S2/5-Pol II #4735, GSK3β #12456; from Sigma: OGA #HPA036141; from Abcam: RL2 #ab2739, Actin #ab49900 and from Bethyl Laboratories: HCF-1 #A301-399a. Reverse Phase Protein Array (RPPA) samples were prepared using the same protocol as for WB, and RPPA was purchased as a service from MD Anderson. Signal from 302 antibodies was detected.

Chromatin immunoprecipitation

Chromatin immunoprecipitation (ChIP) was performed from ~2x107 cells by first cross-linking cells with 1% para-formaldehyde (pFa) at room temperature for 10 minutes. PFa was quenched by addition of 125mM glycine for 10 minutes, and cells were washed twice in ice-cold PBS. Next, cells were solubilized into ChIP lysis buffer (0.1% SDS, 1% Triton X-100, 150mM NaCl, 1mM EDTA, 20mM Tris-HCl pH8.0, protease inhibitor, phosphatase inhibitor and 50µM PUGNAc (the last three added fresh)), and samples were sonicated using Bioruptor (Diagenode) to obtain DNA fragments of 200-500 base pairs in length (confirmed by agarose gel analysis for every experiment). After sonication, samples were centrifuged, and the supernatant was used for ChIP. Antibodies for ChIP were: RL2 #ab2739 (Abcam), MYC AF3696 (R&D systems) and HCF-1 #A301-399a (Bethyl Laboratories). ChIP was performed overnight at +4˚C, and beads were washed as follows: 2x in wash buffer 1 (0.1% SDS, 0.1% Na-deoxycholate, 1% Triton X-100, 0.15M NaCl, 1mM EDTA, 20mM HEPES pH8.0), 1x in wash buffer 2 (0.1% SDS, 0.1% Na-deoxycholate, 1% Triton X-100, 0.5M NaCl, 1mM EDTA, 20mM HEPES pH8.0), 1x in wash buffer 3 (0.25M LiCl, 0.5% Na-deoxycholate, 0.5% NP-40, 1mM EDTA, 20mM HEPES pH8.0) and 2x in wash buffer 4 (10mM EDTA, 20mM HEPES pH8.0). After this, DNA was extracted using phenol-chloroform extraction and used for sequencing or qPCR. RNA for RT-qPCR and RNA-seq was isolated using an illustraMiniSpin kit (GE Healthcare) according to the manufacturer's instructions, and cDNA was synthesized using a qScript cDNA Synthesis Kit (Quanta Biosciences). Primers for ChIP-qPCR and RT-qPCR are provided in Supplementary Table.

Formalin-assisted isolation of regulatory elements

One million LNCaP cells were plated for 3 days and half of media was replaced to fresh media when cells were treated with DMSO or 20µM OSMI-4 for 24 hours. Cells were fixed by adding formaldehyde in 1% final concentration for 10 minutes at room temperature and fixation was stopped by addition of 125mM glycine (final) for 5 minutes at room temperature. Plates were washed with ice cold PBS, cells pelleted by centrifugation (1000rpm, 10 minutes) and lysed in 0.5% SDS, 0.5% TritonX 100, 150mM NaCl, 10mM EDTA, 50mM Tris-HCl pH8.0 containing 1X protease inhibitor. Samples were sonicated using Bioruptor. Input sample was taken and the rest of the sample was subjected to three subsequent phenol:chloroform:isoamyl alcohol (25:24:1) extractions to isolate protein-free DNA in the aqueous phase using columns (#2302830, 5Prime, QuantaBio). All the samples (including input) were treated with proteinase K for 1 hour at +55˚C, DNA de-crosslinked at +65˚C for 4 hours and purified using standard phenol-chloroform extraction. qPCR was used to evaluate chromatin compaction relative to input.

Analysis of the ChIP-seq data

Summary of read counts for ChIP-seq are provided in Supplementary Table. The raw reads from previously published studies and generated for this study (ChIP-seq of MYC and O-GlcNAc) were mapped with novoalign (http://www.novocraft.com) to the human reference genome (build hg19) with default parameters. Only reads with q>20 and a maximum of 5 read duplicates per genomic location were allowed by our filtering. Peak detection (i.e., binding site detection) for both ChIP- and DHS-seq was performed with MACS1.4 35 with default parameters. Overlaps between datasets were performed with bedtools v2.25.0 36. Read distribution analysis around center feature peaks was performed as previously described 37 or using bwtool 1.0 - Data operations on bigWig files 38. To assess the presence of motifs of transcription factors (TFs) in the O-GlcNAc dataset, we looked for overrepresented TF motifs. Prediction of TF binding was performed using “findMotifsGenome.pl” with motif lengths from 6 to 30 nt, and peaks distribution analysis was performed with “annotatePeaks.pl”, both parts of the HOMER package 39.

Analysis of the RNA-seq data

RNA-seq raw paired-end reads from different lanes were concatenated and trimmed with Trimgalore (v.0.4.1). We used Salmon (v0.9.1) 40 to quantify the expression of transcripts from UCSC human genome (hg19). We run Salmon with default parameter and corrected input data for fragment-level GC biases using parameter --gcBias. We performed differential expression analysis using DEBrowser 41 with DESeq2 42 by uploading the read counts retrieved from Salmon. Principal component analysis of RNA-seq data was plotted using the function plotPCA, which is included within the DESeq2 package. The biological replicate samples showed strong correlation for each treatment (Supplementary Fig.

Data availability

The ChIP-seq and RNA-seq data reported in this manuscript has been deposited in the NCBI GEO data repository (GSE121474).

RIME and oligo pulldown

RIME was performed as previously described 43. For oligo pulldown, ~2x107 cells were used. All the steps were performed at 4˚C unless otherwise mentioned. Cells were washed with PBS and collected by centrifugation. Next, cells were solubilized into HKMG-buffer (10mM HEPES pH7.5, 100mM KCl, 5mM MgCl2, 10% glycerol, 0.5% NP-40, 1mM DTT, protease inhibitor, phosphatase inhibitor and 50µM PUGNAc (the last four added fresh) and sonicated using Bioruptor (Diagenode). Samples were centrifuged full speed for 10 minutes and the supernatant collected.

Biotinylated oligos were prepared by mixing 30µL of sense and anti-sense sequences (30mg per ml) with 6.6µL of 10X PCR buffer (100mM TRIS-HCl pH8.0, 15mM MgCl2). Next, this mixture was heated to 100˚C for 5 minutes and allowed to cool down to room temperature on a heat block.

One vial of Streptavidin MagneSphere® Paramagnetic Particles (Z5481, Promega) was prepared for each sample by washing three times with HKMG-buffer, and suspended into 100µL of HKMG-buffer (50µL for pre-clearing of samples, 50µL used to bind to oligo). 10µL of annealed oligos were rotated with 50µL of beads for 15 minutes in room temperature, washed 3 times with HKMG-buffer and stored on ice until use.

Cell extracts were pre-cleared with 50µL of beads for 1 hour. After this, oligo-bead complexes were added to extract and incubation was continued overnight. The next day, beads were washed five times with HKMG-buffer and stored at -80˚C until mass spectrometry analysis.

Liquid chromatography-mass spectrometry (LC-MS)

The beads from RIME and oligo pulldown were incubated with 2 µg trypsin GOLD (Promega, Madison, WI, USA) in 50 µl 25 mM ammonium bicarbonate pH 7.8 for 16 h at 37°C. The resulting peptides were purified using a 10 µl OMIX C18-SPE pipette tip (Agilent, Santa Clara, CA, USA), and dried using a Speed Vac concentrator (Concentrator Plus, Eppendorf, Hamburg, Germany). Subsequently, the tryptic peptides were dissolved in 10 µl 0.1% formic acid/2% acetonitrile and 5 µl samples were analyzed using an Ultimate 3000 RSLCnano-UHPLC system connected to a Q Exactive mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) equipped with a nano-electrospray ion source. For liquid chromatography separation, an Acclaim PepMap 100 column (C18, 2 µm beads, 100 Å, 75 μm inner diameter, 50 cm length) (Dionex, Sunnyvale CA, USA) was used. A flow rate of 300 nL/min was employed with a solvent gradient of 4-35% B over 60 min. Solvent A was 0.1% formic acid and solvent B was 0.1% formic acid/90% acetonitrile. The mass spectrometer was operated in the data-dependent mode to automatically switch between MS and MS/MS acquisition. Survey full scan MS spectra (from m/z 400 to 2,000) were acquired with the resolution R = 70,000 at m/z 200, after accumulation to a target of 1x106. The maximum allowed ion accumulation times were 60 ms. The method used allowed sequential isolation of up to the ten most intense ions, depending on signal intensity (intensity threshold 1.7x104), for fragmentation using higher-energy collisional induced dissociation (HCD) at a target value of 1x105 charges, NCE 28, and a resolution R = 17,500. Target ions already selected for MS/MS were dynamically excluded for 30 sec. The isolation window was m/z = 2 without offset. For accurate mass measurements, the lock mass option was enabled in MS mode.

Data were acquired using Xcalibur v2.5.5 and raw files were processed to generate a peak list in Mascot generic format (*.mgf) using ProteoWizard release version 3.0.331. Database searches for trypsin fragments were performed using Mascot in-house version 2.4.0 to search the SwissProt database (Human, 20279 proteins), allowing a maximum of one missed trypsin cleavage site for any given peptide, fragment ion mass tolerance of 0.05 Da, parent ion tolerance of 10 ppm, with oxidation of methionines and acetylation of the protein N-terminus as variable modifications. Scaffold (version Scaffold_4.4.8, Proteome Software Inc., Portland, OR) was used to validate MS/MS based peptide and protein identifications. Peptide identifications were accepted if they could be established at greater than 95.0% probability by the Scaffold Local FDR algorithm. Protein identifications were accepted if they could be established at greater than 99.0% probability and contained at least two identified peptides.

Results

Global chromatin O-GlcNAc mark is sensitive to OGT inhibition

In order to probe the stability of the O-GlcNAc chromatin mark, we utilized a recently reported fast-acting OGT inhibitor OSMI-2 34 and a specific antibody recognizing the O-GlcNAc epitope 13, 44. First, we used western blotting to assess the effects of OSMI-2 on the total O-GlcNAc levels in AR-positive (LNCaP) and AR-negative (PC3) prostate cancer cell line after 4 hours of treatment (Fig. ). OSMI-2 more efficiently decreased total-O-GlcNAc in PC3 cells, which was therefore selected to probe the stability of the O-GlcNAc mark on chromatin. We performed a biological triplicate ChIP-seq experiment in the presence and absence of OSMI-2 and detected 8,906 peaks in PC3 cells (Fig. ). Strikingly, 4 hours treatment with OSMI-2 decreased the number of O-GlcNAc peaks by 70%, which shows that this chromatin mark is remarkably un-stable and further validates that it is installed by OGT (Fig. ). Strong enrichment of the chromatin-bound O-GlcNAc was quantitatively and reproducibly lost upon OSMI-2 treatment, as exemplified using the promoter for the MAX gene (Fig. ). We validated the O-GlcNAc ChIP-seq data using OGT inhibition and genetic depletion of OGT (Supplementary Fig. ). This is the first study to report that O-GlcNAc chromatin mark is rapidly lost when OGT activity is inhibited.

O-GlcNAc chromatin mark has been reported to localize to transcription start sites (TSS) 13, 44. We therefore evaluated the peak distribution of O-GlcNAc in DMSO and OSMI-2 treated cells. We observed a sharp single peak for DMSO-treated cells, while OSMI-2 treatment led to a remarkable re-distribution of the single O-GlcNAc peak into two separate peaks (Fig. ). Next, we evaluated the peak distribution of the peaks that overlap in the DMSO and OSMI-2 ChIP-seq. In this case, both conditions showed sharp single peak. Despite OGT inhibition, we detected these O-GlcNAc marked sites, and accordingly, these common sites have the most stable O-GlcNAc mark based on the read counts (compare DMSO-treated sample in Fig. to 1E). Importantly, the strength of the meta-peak was decreased by ~50% for the OSMI-2 treatment (Fig. ). These data show that the O-GlcNAc chromatin mark is dependent on high OGT activity. The drastic re-distribution of the O-GlcNAc peaks in response to OSMI-2 suggests that this modification participates in transcriptional regulation.

Most prostate cancers are AR-positive 18, and we therefore evaluated the conservation of the O-GlcNAc chromatin mark localization between AR-positive prostate cancer cell line (LNCaP) and PC3 cells. The majority of the LNCaP O-GlcNAc ChIP-seq sites overlapped with the O-GlcNAc sites in PC3 cells (70%, Supplementary Fig. ). In addition, LNCaP O-GlcNAc sites overlapped significantly with previously reported single-replicate ChIP-seq studies (Fig. ). The high degree of conservation between cell lines of different origin suggests that O-GlcNAc could be a regulator of fundamentally important pathways.

We selected the genes associated with the O-GlcNAc chromatin mark in both LNCaP and PC3 cells and subjected this gene set to pathway enrichment analysis to identify candidate OGT-regulated pathways in prostate cancer cells. Among the top pathways were 'RNA transport', 'ribosome', 'Ubiquitin mediated proteolysis' and 'cell cycle' (Supplementary Fig. ). Our data identifies a mechanism that directly links OGT to biomass production and proliferation, in support of models proposing OGT as a nutrient sensor 45. In order to better understand the role of chromatin O-GlcNAc mark, we moved on to characterize its specific localization throughout the genome.

O-GlcNAc marks transcriptionally active chromatin and super-enhancers

The majority of chromatin-bound O-GlcNAc was found in promoters (57%), as previously reported (Fig. ) 13, 44. Almost all of the O-GlcNAc sites (95%) overlapped with DNase-hypersensitive regions (DHS) and markers of active promoters (h3K4me3 and h3K27ac) but no overlap was observed with a marker of transcriptionally inactive chromatin (h3K27me3, Fig. ). In addition, 95% of chromatin O-GlcNAc overlapped with active transcription (RNA-Pol II and nascent RNA). These data are in agreement with reports showing that RNA-Pol II CTD is O-GlcNAc modified and that O-GlcNAc and RNA-Pol II sites overlap 11-13. Given the prominent overlap between O-GlcNAc and RNA-Pol II chromatin-bound regions, we evaluated if OGT inhibition affects phosphorylation status of RNA Pol-II but did not observe prominent effects (Supplementary Fig. ). These data indicate that OSMI-2 does not exert global defect in RNA-Pol II activity but could affect expression of certain genes driven by a specific class of regulatory elements.

Super-enhancers (SE) are cell-type specific enhancers of highly transcribed genes 46. We assessed if O-GlcNAc mark overlaps with these regulatory elements. SEs have been mapped in LNCaP cells using the h3K27ac histone modification as the surrogate 47. Most (70%) of the SE-genes were associated with the O-GlcNAc mark (Fig. ). OGT inhibition decreased the expression of CDK1 and KLK3, two SE-genes selected for validation (Fig. ). Targeting OGT has been shown to decrease the expression of CDK1, but the reason for this is not understood 30.

O-GlcNAc mark did not always directly overlap with SEs (Supplementary Fig. ), and we therefore assessed if OGT activity is important to maintain chromatin-accessibility using formaldehyde-assisted isolation of regulatory elements (FAIRE). OGT inhibition led to chromatin compaction of both the CDK1 and KLK3 enhancers (Supplementary Fig. ). In order to position our finding into clinical setting, we evaluated the expression pattern of OGT and CDK1 in prostate cancer patient samples. CDK1 expression increases steadily from normal prostate tissue to high Gleason Score prostate cancer and tumors with high CDK1 expression have significantly elevated levels of OGT (Fig. ). These data imply that high OGT activity is important for the expression of CDK1 also in vivo.

To summarize our findings so far, we show for the first time that O-GlcNAc chromatin mark is associated with SE genes and OGT-activity is important for the expression of the genes driven by SEs. In addition, O-GlcNAc is enriched on the transcriptionally active promoters in prostate cancer cells but we do not currently know whether this modification is primarily associated with histones or transcription factors.

MYC co-occupies O-GlcNAc marked chromatin sites

Histone read distributions around chromatin-bound O-GlcNAc show that O-GlcNAc sites have the typical chromatin structure of a region bound by transcription factors (Fig. ). To identify candidate transcription factors binding to this motif, we performed de novo motif discovery using HOMER. The top two motifs for O-GlcNAc ChIP-seq were similar to the ETS (also known as E26 transformation-specific) transcription factor ELK4 and ELK1 motifs (Fig. ). We validated this overlap by confirming that 80% of the O-GlcNAc regions overlap with the prostate cancer-relevant ETS transcription factor, ETV1 (Supplementary Fig. ) 18. However, we did not observe prominent overlap between O-GlcNAc occupied regions and other transcription factors relevant for prostate cancer progression (AR, FOXA1 and HOXB13, Supplementary Fig. ) 18.

OGT has previously been reported to be a critical mediator of MYC-dependent transformation 14, 21, 22. We therefore performed a MYC ChIP-seq experiment and observed that 77% of the MYC binding regions overlapped with chromatin-bound O-GlcNAc in prostate cancer cells (Fig. ). To validate the observed overlap, we used a previously published MYC ChIP-exo dataset 48. ChIP-exo allows smaller peaks to be detected more readily, which leads to a higher number of peaks than the conventional ChIP-seq 49. We observed that ~90% of the O-GlcNAc-marked regions overlapped with the MYC ChIP-exo peaks (Fig. ). O-GlcNAc chromatin mark is clearly associated with MYC binding sites but we do not yet know how OGT affects MYC function.

OGT activity is required for MYC-HCF-1 interaction

In order to further characterize the O-GlcNAc-modified and chromatin-associated protein complex that contains MYC, we used mass spectrometry. First, we performed RIME (Rapid Immunoprecipitation for Mass spectrometry of Endogenous proteins) using the same anti-O-GlcNAc antibody as we used for ChIP-seq. Second, we prepared a biotinylated oligonucleotide of the GFY-sequence, which was the third most enriched motif for O-GlcNAc ChIP-seq (Fig. ). Finally, we overlapped these two datasets with MYC RIME and identified a single protein, Host Cell Factor 1 (HCF-1; also called HCFC1) (Supplementary Table).

We confirmed that MYC interacts with HCF-1 in prostate cancer cells (Fig. ), as previously reported using a different model system 50. HCF-1 is cleaved at one of several central repeats, explaining the multiple bands observed in the western blot; the N- and C-terminal fragments remain associated in cells to form the functional protein 51. Both MYC and HCF-1 can be glycosylated by OGT, and we observed that a short-term treatment (4 hours) with OGT inhibitor OSMI-2 abrogates the interaction between the two (Fig. ) 30, 52. Reciprocally, we immunoprecipitated HCF-1 and observed that OSMI-2 decreases the interaction between HCF-1 and MYC (Supplementary Fig. ). Next, we used siRNAs to knockdown OGT, and observed that this decreases the interaction between MYC and HCF-1 (Supplementary Fig. ). In order to confirm that OGT is important for the interaction between MYC and HCF-1, we used a MEF cell line, in which OGT gene can be deleted from the genome by tamoxifen-inducible Cre-recombinase 33. Importantly, OGT knockout disrupted the interaction between MYC and HCF-1 (Fig. ).

OGT activity might be essential for transcription factor complex formation in general, and the fast-acting OSMI-2 enabled us to assess this. OSMI-2 treatment resulted in the loss of interaction between MYC and SP1, the first transcription factor reported to be O-GlcNAc modified and a known MYC-interaction partner (Supplementary Fig. ) 53, 54. On the other hand, OSMI-2 did not affect the previously reported interaction between AR and HCF-1 (Supplementary Fig. ) 55. This differential sensitivity may be important for the integration of nutrient-availability to engage with proliferative-program (MYC) or maintenance of cell-identity (AR).

Our data shows that the acute OGT inhibition affects MYC interaction with partner proteins. In support of this, treatment with OSMI-2 led to loss of MYC-targets in the O-GlcNAc ChIP-seq data (Supplementary Fig. ). Over half of the MYC-bound chromatin regions are additionally marked by O-GlcNAc and HCF-1, further supporting the functional relevance of the MYC-HCF-1 interaction (Fig. ). Based on these data, we propose that OGT is important for the MYC-dependent transcriptional program.

OGT regulates the MYC-dependent transcriptome and proteome

In order to assess the effect of OGT inhibition on MYC-dependent transcription, we used a cell line model in which MYC can be induced by the addition of doxycycline and performed RNA-seq in the presence and absence of OSMI-2. We first confirmed that OSMI-2 does not prominently affect MYC-induction at the mRNA and protein levels (Fig. ). In order to assess if OSMI-2 affects all MYC-dependent genes in one way or the other, we performed principal component analysis. The effect on gene expression was dominated by MYC-induction, in agreement with previous reports establishing MYC as a strong transcriptional activator (Supplementary Fig. ) 28, 29. OSMI-2 had only a modest effect, which may be explained by the fact that OGT-expression is strongly increased and OGA down-regulated in response to OSMI-2 (Supplementary Fig. ).

Addition of OSMI-2 did not drastically change the MYC-induced expression profile but rather amplified it, for example by enhancing the down-regulation of KLK3 and NKX3.1 or by increasing the induction of RNF187 and DDIT4 (Supplementary Fig. ). Interestingly, overexpression of MYC in the presence of OSMI-2 led to down-regulation of a high number of genes (Supplementary Table). We used RT-qPCR to validate that induction of MYC in the presence of OSMI-2 results in down-regulation of MMP7, ZNF812P, GINS2, ASF1B, DTL, and KLK2 (Fig. ). Pathway enrichment analysis of the genes that were down-regulated specifically upon the combination treatment identified the cell cycle, and more specifically, the transition from G1- to S-Phase, as the most significant process (Supplementary Fig. ), suggesting that OGT might be important for MYC-regulated cell cycle progression.

Next, we used reverse-phase protein array technology (RPPA) to test if OGT affects the MYC-dependent proteome. We focused on the proteins that required OGT activity to be regulated by MYC and identified 28 proteins (10% of the total detected) (Supplementary Table). Pathway enrichment analysis reported a specific point of the cell cycle, the G2-M transition, to be enriched for these proteins (Supplementary Fig. ). MYC-dependent expression of these 28 proteins did not strongly correlate with mRNA levels (Fig. ). On the other hand, for both OSMI-2 on its own and in combination with MYC overexpression, we observed a statistically significant correlation between mRNA and protein levels. We selected two proteins PLK1, and Cyclin B1, to validate the RPPA-data using RT-qPCR and western blotting. MYC-induced expression of these proteins was antagonized by OSMI-2 at the protein level but not affected at the mRNA level (Fig. ).

To identify potential mechanisms for OSMI-2 induced down-regulation of MYC-regulated proteins, we turned our attention to the genes that were induced when OSMI-2 was combined with MYC overexpression. Induction of MYC in the presence of OSMI-2 resulted in a 12-fold increased expression of the lactotransferrin (LTF)-gene (Fig. ). O-GlcNAc, MYC, and HCF-1 were all bound to the enhancer of the LTF-gene (Supplementary Fig. ). LTF acts as a tumor-suppressor through its ability to induce the expression of ubiquitin ligases and the protein is negatively regulated by OGT 56. Based on our RNA-seq data, we identified and subsequently validated that two ubiquitin ligases, FBXO32 and PRKN, are induced by OGT inhibition (Supplementary Table and Fig. ). In order to understand the potential clinical relevance of LTF in prostate cancer, we evaluated its expression in patient samples. Interestingly, LTF is abundantly expressed in normal prostate tissue and in low-grade tumors (Gleason score<7), but its expression is completely shut down in the metastatic tissue (Fig. ).

Taken together, OGT regulates MYC-dependent transcription of certain mRNAs involved in the control of G1-S transition and additionally promotes protein stability of mitotic proteins. MYC-induced expression of ubiquitin ligases adds another layer of complexity to MYC-dependent regulation of cell function; this response will be further characterized in future studies. Our data suggest that OGT should be especially important for MYC-driven proliferation.

OGT is required for MYC-dependent proliferation of prostate cancer cells in the absence of androgens

MYC is required for the androgen-independent proliferation of prostate cancer cells, the major cause of disease progression 48, 57. This can be modeled in vitro using the LNCaP-MYC cell line, which requires androgens to achieve full growth potential; however, MYC-expression can be induced by addition of doxycycline and this is sufficient to promote androgen-independent proliferation 48, 58.

First, we wanted to assess if MYC can promote the expression of mitotic proteins in the absence of androgens and whether OGT activity is important in this setting. We observed an even stronger induction of Cyclin B1, and PLK1, when compared to cells growing in the presence of androgens (Fig. ). OSMI-2 treatment almost completely antagonized MYC-induced expression of Cyclin B1, and PLK1. Overexpression of MYC also increased phosphorylation of the marker of mitosis, p-H3ser10, and this was decreased by OSMI-2. Of note, cells are able to adapt to OSMI-2 treatment by increasing the expression of OGT and decreasing the expression of OGA, and the OSMI-2 effect on total-O-GlcNAc at 24 hours of treatment in the absence of androgens is modest. Using siRNAs against OGT we confirmed that the MYC-induced increase in the expression of Cyclin B1 and PLK1 is dependent on OGT (Fig. ).

Second, we used flow cytometry to assess the role of OGT in MYC-induced cell cycle progression. Androgen-deprivation decreased the number of cells in S-Phase by over 50% (Fig. ). Overexpression of MYC was able to double the number of cells going through the S-Phase in the absence of androgens. Importantly, OSMI-2 antagonized MYC-induced accumulation of cells in the S-Phase.

We next used viability assays to confirm reports that MYC overexpression promotes proliferation of prostate cancer cells in the absence of androgens 48, 57, 58. Overexpression of MYC increased cell viability by 25% in the absence of androgens but did not increase viability in the presence of androgens (Fig. ). The growth-benefit of MYC-overexpression was decreased by treatment with OSMI-2 and completely antagonized by siRNAs targeting OGT.

Here we show that the androgen-independent growth advantage induced by MYC is dependent on OGT. Our data shows that the high expression of certain factors at the protein level, even in the absence of mRNA elevation, might be important for cancer cell proliferation and could be used as biomarkers. To test this, we evaluated the predictive value of Cyclin B1 at the mRNA and protein levels in prostate cancer patients. CCNB1 mRNA is non-significantly associated with disease progression (p=0.203), while increased protein level expression of Cyclin B1 is significantly associated with disease progression (p=0.000014; Logrank test, Fig. ). These results highlight the importance of evaluating both mRNA and protein level data to understand the relevance and predictive value of prognostic markers.

Taken together, OGT is important for the MYC-driven cellular program at two levels: the transcription of genes involved in the G1-S transition and the stabilization of proteins involved in the G2-M transition. This renders high OGT activity indispensable for MYC-dependent proliferation of prostate cancer cells.

Discussion

Mammalian OGT is required for development, normal physiological functions of post-mitotic cells and proliferation of immortalized cells 33, 59-62. Here, we have shown that OGT regulates MYC-dependent transcription and affects levels of mitotic proteins in prostate cancer cells.

In general, the functional role of OGT in the regulation of transcription is poorly understood, but one of the most prominent O-GlcNAc modified proteins is RNA-Pol II itself. Glycosylation of RNA-Pol II CTD has been shown to regulate pre-initiation complex formation using in vitro assays and O-GlcNAc mark is found on the promoters in human cell lines as reported here and shown by others 12, 13, 44. In addition, O-GlcNAc is enriched on the promoters in Caenorhabditis elegans, but the modification is not responsive to nutrients 63. Perturbation of OGT activity in specific cell or organ type has prominent effects on transcription and overall physiology of the animal in mouse studies 62, 64, however, given the high number of OGT-targets, it is not possible to attribute these effects directly to transcription. We observed that prostate cancer super-enhancer (SE) genes are also marked by the O-GlcNAc chromatin mark and that OGT inhibition decreases the expression of SE-mRNAs (Fig. ). This implies that OGT activity may be particularly important for highly transcribed genes. In the future, it is important to establish how inhibition of OGT affects transcription using GRO- and NET-seq type assays.

We show that inhibition of OGT enhances MYC-regulated transcriptional repression of certain cell cycle genes, possibly due to disruption of MYC-containing protein complexes, albeit the exact mechanism(s) for this remain to be determined (Figs. ). In addition, we showed that OGT is important for MYC-regulated expression of mitotic proteins (Figs. ). Based on these data, we propose that the decrease in OGT activity leads to suppression of cell cycle program by a dual mechanism: at the mRNA level to control initiation of the S-Phase and additionally OGT affects the stability of proteins involved in mitosis (Figs. ). OGT would thereby function as a metabolic rheostat and utilize its catalytic activity to halt the cell cycle in the absence of sufficient energy and promote proliferation when nutrients are abundant.

Excessive nutrient uptake is detrimental to health in general, and obesity has been linked to both breast and prostate cancers 65, 66. A high rate of proliferation requires a constant supply of energy, but it is becoming increasingly clear that metabolism can also directly affect cell fate 3, 9, 67, 68. It is known that glucose feeding stimulates OGT-activity in mouse liver and that overexpression of the glucose transporter GLUT3 in nonmalignant breast cells leads to loss of tissue polarity and activation of tumor-promoting pathways 10, 69. We therefore speculate that increased blood glucose, as observed in common conditions such as pre-diabetes, which affects over 80 million Americans (source: American Diabetes Association), could serve as a second hit, and be a variation of the genetic '2-hit' hypothesis 14, 70.

OGT is essentially required in proliferating cells but post-mitotic cells can survive without it, which makes the enzyme a prototypical cancer-target 60. OSMI-2 binds to the active site of OGT, and higher levels of the OGT's substrate, UDP-GlcNAc , in LNCaP cells than PC3 cells, renders total-O-GlcNAc levels less sensitive to this inhibitor (Fig. ) 19, 34. AR promotes the expression of the hexosamine biosynthetic pathway (HBP) enzymes and prostate cancer cells therefore have high levels of UDP-GlcNAc 14, 19. Interestingly, progression to lethal, castration-resistant prostate cancer (CRPC) leads to a prominent down-regulation of the HBP enzymes 71. The development of CRPC is not completely understood, but one of the factors known to be able to confer CRPC phenotype is MYC 48, 57, 72. In this study, we showed that MYC-driven, androgen-independent proliferation of prostate cancer cells requires OGT. Based on these data, OGT inhibitors could be combined with anti-androgens to target the cancer cells that become dependent on MYC.

In conclusion, we show that O-GlcNAc, MYC and HCF-1 are found in transcription start sites that contain the GFY-motif and that OGT activity is required for the robust interaction between MYC and HCF-1. Inhibition of OGT-activity impairs MYC's ability to promote androgen-independent proliferation. Our work suggests that OGT could be a drug-target in MYC-addicted tumors. Identification of new actionable targets for cancer-therapy is urgently needed, and in the future, it will be important to understand potential cancer cell-specific adaptive response(s) that are activated in response to OGT inhibition.

Supplementary figures and table legend.

Click here for additional data file.

Supplementary table.

Click here for additional data file.