Compromised CDK12 activity causes dependency on the high activity of O-GlcNAc transferase Authors: Satu Pallasaho, Aishwarya Gondane, Julia Kutz, Jing Liang, Shivani Yalala, Damien Y Duveau, Helmut Pospiech, Craig J Thomas, Massimo Loda, Harri M Itkonen Published: 2024 ABSTRACT AbstractO-GlcNAc transferase (OGT) coordinates with regulators of transcription, including cyclin-dependent kinase 12 (CDK12), the major transcription elongation kinase. Here, we use inhibitor- and knockdown-based strategies to show that co-targeting of OGT and CDK12 is toxic to prostate cancer cells. OGT catalyzes all nucleocytoplasmic O-GlcNAcylation and due to its essentiality in higher eukaryotes, it is not an ideal drug target. Our glycoproteomics-data revealed that short-term CDK12 inhibition induces hyper-O-GlcNAcylation of the spliceosome-machinery in different models of prostate cancer. By integrating our glycoproteomics-, gene essentiality- and clinical-data from CDK12 mutant prostate cancer patients, we identify the non-essential serine-arginine protein kinase 1 (SRPK1) as a synthetic lethal partner with CDK12-inactivation. Both normal and cancer cells become highly sensitive against inhibitors of OGT and SRPK1 if they have lowered activity of CDK12. Inactivating mutations in CDK12 are enriched in aggressive prostate cancer, and we propose that these patients would benefit from therapy targeting the spliceosome.Keywords: cyclin-dependent kinase 12, O-GlcNAc transferase, prostate cancer, spliceosome, synthetic lethality Introduction Prostate cancer (PC) is the most common cancer in men and after advancing to the castration-resistant PC (CRPC), there are currently no curative treatment options (Sandhu et al. 2021). Practically all prostate tumors are positive for the nuclear hormone-activated transcription factor, androgen receptor (AR), and treatments that decrease the AR-activity halt disease progression in the clinical setting (Itkonen and Mills 2012). CRPC occurs when PC cells proliferate in the presence of the AR-targeting therapy. Better understanding of the CRPC-specific features is likely to reveal cancer cell-selective vulnerabilities, which will eventually lead to discovery of more effective treatments against the currently terminal disease, CRPC. One such feature is the increased activity of O-GlcNAc transferase (OGT) in aggressive PC (Itkonen et al. 2013; Itkonen et al. 2021). OGT is the sole enzyme catalyzing the attachment of N-acetylglucosamine (GlcNAc) moiety to target proteins (Levine and Walker 2016; Hart 2019; Itkonen et al. 2021) and its inhibition reduces proliferation of the CRPC cells (Lynch et al. 2012; Itkonen et al. 2013; Itkonen et al. 2016; Itkonen et al. 2019). Although these studies validate OGT as a therapeutic target, they also show that more than a single vulnerability needs to be exploited in order to achieve therapeutic specificity for CRPC. In line with this, we discovered that inhibition of OGT sensitizes CRPC cells to compounds targeting cyclin-dependent kinase 7 (CDK7) and CDK9 (Itkonen et al. 2020; Pallasaho et al. 2022). CDK7 and CDK9 phosphorylate RNA polymerase II (RNA Pol II) to promote transcription initiation and to release the polymerase from the promoter-proximal pausing, respectively (Jonkers and Lis 2015; Chou et al. 2020; Vervoort et al. 2022). In addition to CDK7 and CDK9, RNA Pol II CTD is phosphorylated by the transcription elongation kinases CDK12 and CDK13 to promote transcription of the long genes (Chou et al. 2020). CDK12 activity is important during transcription, and it is therefore unexpected that inactivating mutations in the CDK12-gene confer a growth-advantage. CDK12 mutant primary tumors metastasize with almost 100% penetrance (Schweizer et al. 2020), and mutations in this gene are 7-times more frequent in the CRPC when compared to the primary tumors (Wu et al. 2018). Overall, CDK12 mutant tumors define an aggressive sub-type of CRPC (Wu et al. 2018; Rescigno et al. 2021). Depletion of CDK12 activity causes DNA damage in vitro, and a characterizing feature of the CDK12 mutant tumors are large genomic deletions and amplifications (Quigley et al. 2018; Viswanathan et al. 2018; Sokol et al. 2019). By understanding the rewiring of the cells due to compromised CDK12 activity, it should be possible to discover actionable synthetic lethal interactions. Here, we show that the combined inhibition of OGT and CDK12/13 is toxic to PC and CRPC cells, and that the short-term CDK12/13 inhibition leads to an OGT-dependent remodeling of the splicing-machinery. We go on to show that the co-targeting of CDK12/13 with five different spliceosome regulators is combinatorial lethal to CRPC cells and identify SRPK1 as the most promising target. Combinatorial targeting of either OGT or SRPK1 with CDK12/13 inhibitor augments cell death-activation. Most inhibitors targeting CDK12 also inhibit CDK13, and we therefore confirmed that knockdown of CDK12 sensitizes PC cells to depletion of OGT and SRPK1 activities. In brief, we propose that the inactivation of CDK12 renders CRPC cells acutely reliant on the spliceosome, which opens an actionable therapeutic vulnerability. Results Combined targeting of OGT and CDK12 is toxic to prostate cancer cells CDK12 inactivating mutations are enriched in the aggressive CRPC (Rescigno et al. 2021) but small molecule inhibitors targeting CDK12/13 halt the proliferation of PC cells (Lei et al. 2021); clearly better understanding on how PC cells adapt to declining CDK12 activity is of a high interest. Using a combinatorial lethality screen, we discovered that co-targeting of OGT and CDK9 is toxic to cancer cells (Itkonen et al. 2020). CDK9 is the transcription elongation kinase that releases RNA Pol II for the productive transcription elongation, while CDK12 maintains the phosphorylation in long genes (Chou et al. 2020). We hypothesized that we could enhance the anti-proliferative effects of the CDK12 inhibition using OGT inhibitors. First, we treated PC cells with increasing doses of THZ531, an inhibitor of CDK12 that also targets the structurally similar kinase CDK13 (Zhang et al. 2016), in the presence and absence of the OGT inhibitor OSMI-4 (Martin et al. 2018). Targeting OGT significantly and dose-dependently sensitized PC cells (LNCaP) and CRPC cells (LN95, C4-2 and 22RV1) to CDK12/13 inhibition as measured by cell viability assays (Fig. 1A and Suppl. Fig. 1A). Second, we selected a low dose of the CDK12/13 inhibitor and used live-cell imaging to follow the growth-kinetics of PC cells. Combining OGT and CDK12/13 inhibitors led to an almost complete growth-arrest of both PC and CRPC cells (Fig. 1B and Suppl. Fig. 1B). Third, we measured cell death activation after CDK12/13 and OGT inhibitor treatments, which revealed a 20-fold increase in cell death in the PC and CRPC cells when compared to two cell lines representing normal prostate cells (PNT1 and RWPE-1; Fig. 1C). A rapid doubling time sometimes explains the increased sensitivity of a given cell line to a particular treatment. In the model systems used here, the cell lines derived from normal prostate cells (RWPE-1 and PNT1) had similar or even more rapid doubling times than the cancer cell lines (LNCaP, LN95, 22RV1 and C4-2; Fig. 1C). Therefore, the length of the doubling time does not explain the robust difference in increase of apoptosis between the normal and cancer cell lines in response to the combined inhibition of OGT and CDK12/13. Combined inhibition of CDK12/13 and OGT is lethal to prostate cancer cells. A) CRPC cell viability was measured using CellTiterGlo-assay after 4 days of treatment with OGT inhibitor (OSMI-4) and/or CDK12/13 inhibitor (THZ531) (n: 4). B) Proliferation of CRPC cells was followed using live-cell imaging. Mean and standard error of mean (SEM) is indicated (n: 4). C) Cell death activation (caspase 3/7 cleavage) in PC, CRPC and normal prostate cells after four days of treatment (100 nM THZ531, 20 μM OSMI-4). Data is presented as average of four biological replicates with SEM. Control sample was set to value of 1, and the other treatments are presented relative to that. The doubling time of each of the cell lines is indicated after the cell line name (obtained from the Cellosaurus database). D) CRPC cells were reverse-transfected with the indicated siRNAs for 4 days after which cell viability was assessed using CellTiterGlo-assay. Data shown is from four biological replicate-experiments. E) OGT inhibition using OSMI-4 decreases the expression of replication proteins. Re-analysis of the proteomics data reported in (Martin et al. 2018). Genes encoding for proteins involved in “DNA replication” and “cell cycle” were selected from Kyoto Encyclopedia of genes and genomes (KEGG). F) CRPC cells were treated with 20 μM OSMI-4, 100 nM THZ531, or combination of the two for 24 h, samples analyzed using western blot and densitometry was used to quantitate signal intensity (n: 2–4, densitometry provided in Suppl. Fig. 1D). THZ531 targets both CDK12 and CDK13 (Zhang et al. 2016), and we therefore used knockdown-experiments to ascertain the importance of CDK12. This approach confirmed that combined knockdown of OGT and CDK12 is significantly more toxic to CRPC cells than the knockdown of either of the targets individually (Fig. 1D and Suppl. Fig. 1C). We hypothesized that the depletion of OGT activity augments DNA damage in cells that have lowered activity of CDK12. This hypothesis was based on earlier data showing that the decrease in CDK12 activity impairs the expression of the genes encoding for the DNA repair- and DNA damage response-machineries (Dubbury et al. 2018; Krajewska et al. 2019). To more systematically evaluate if OGT inhibition also affects the levels of proteins maintaining genomic stability, we re-analyzed a previously generated whole proteome profiling experiment of HEK cells treated with the same OGT inhibitor as we used here (Martin et al. 2018). This approach revealed a systematic downregulation of proteins involved in DNA replication and cell cycle (Fig. 1E). According to our hypothesis, we detected an over 3-fold further increase in the levels of the DNA damage marker p-H2AX in the combination treatment when compared to single agent treatments (Fig. 1F and Suppl. Fig. 1D). In support of apoptotic cell death as the main mechanism of the decreased proliferation, we also note a 5-fold increase in PARP cleavage in cells treated with inhibitors of CDK12/CDK13 and OGT compared to single agent treatments. It is possible that the increase in p-H2AX is explained through activation of cell death; nevertheless, co-targeting of OGT and CDK12 is toxic to CRPC cells. We wanted to establish if the decrease in CDK12 activity sensitizes also normal cells to OGT inhibition. To assess this, we repeated the cell viability, knockdown and western blot experiments in cell lines derived from normal prostate epithelia. Indeed, as we expected, also normal cells become dependent on the high OGT activity when CDK12 is inhibited either using small molecule inhibitor that targets also CDK13 or knockdown of CDK12 (Suppl. Figs 2 and 3). To summarize our major findings so far, depletion of OGT activity sensitizes PC cells and cell lines derived from normal prostate epithelia to CDK12 inhibition, which proposes that the inactivation of CDK12 could be a biomarker of sensitivity. However, as OGT is an essential gene in the higher eukaryotes, it is not an ideal target for clinical development (Itkonen et al. 2021). In addition, OGT has thousands of candidate target proteins in the cell, and depleting its activity will have wide-spread effects. It is well-established that OGT’s target-repertoire is significantly remodeled in response to various stimuli and stress (Martinez et al. 2017; Itkonen et al. 2021; Gondane et al. 2022; Pallasaho et al. 2022), and we therefore hypothesized that CDK12 inhibition renders PC cells dependent on O-GlcNAcylation of specific proteins. Ideally, one could inhibit only one of these proteins to selectively kill the cancer cells with the compromised CDK12 activity. However, we did not yet know which are the key proteins modified by OGT in response to CDK12 inhibition. CDK12/13 inhibition induces OGT-dependent remodeling of the spliceosome machinery We used mass spectrometry to identify the proteins that get hyper-O-GlcNAcylated after short-term treatment with the dual CDK12/13 inhibitor. Combined inhibition of OGT and CDK12 is toxic to both PC and CRPC cells (Fig. 1 and Suppl. Fig. 1). Therefore, potential common proteins increasingly O-GlcNAcylated in response to CDK12/13 inhibition are the most promising candidate mediators of cell survival. LN95 cell line was established through an extended androgen-deprivation of the androgen-dependent LNCaP cell line. We immune-enriched the O-GlcNAcylated proteins from PC and CRPC cells (LNCaP and LN95) after 4 h treatment with CDK12/13 inhibitor. Most of the O-GlcNAcome remained unaffected after 4 h of the CDK12/13 inhibitor treatment, but there were a number of proteins that were increasingly glycosylated (Fig. 2A). We did not identify any protein that would be directly targetable via small molecule inhibitor and to be increasingly O-GlcNAcylated in both model systems (Suppl. Table 1). Therefore, we selected the proteins whose glycosylation increased by at least 20% in all three biological replicates and performed pathway-enrichment analysis. This approach identified the “Spliceosome” as the most significantly enriched process in both models (Fig. 2B). We used lectin pulldown experiments to validate O-GlcNAcylation of one of the splicing-associated proteins, HNRNPU, in 22RV1 prostate cancer cells (Suppl. Fig. 4). Previously, it has been established that the depletion of OGT activity induces phosphorylation of the spliceosomal proteins (Tan et al. 2020). These data propose that CDK12 and OGT compete for the same spliceosome proteins for phosphorylation and glycosylation, respectively, and when the activity of the other one decreases, the reciprocal modification dominates. Inhibition of CDK12/13 induces OGT-dependent remodeling of the spliceosome. A) Prostate cancer (LNCaP) and CRPC (LN95) cells were treated for 4 h with CDK12/13 inhibitor (500 nM THZ531), after which O-GlcNAcylated proteins were enriched using immunoprecipitation and analyzed by mass spectrometry (n = 3, normalized to unspecific immunoprecipitation). B) KEGG-pathway enrichment analysis of the proteins whose O-GlcNAcylation increased by 20% after THZ531 treatment in all three replicates. Significance: Color-coding and size of the circle (adjusted P-value, Benjamin-Hochberg method for correction for multiple hypothesis testing). C) CDK12/13 inhibitor-induced alternative splicing (AS) in LNCaP cells treated with 500 nM THZ531 for 6 h relative to untreated cells (P-value < 0.05; re-analysis of RNA-seq previously reported (Lei et al. 2021)). Left are the major AS-events: IR: Intron retention; A5SS: Alternative 5′ splice sites; A3SS: Alternative 3′ splice sites; ES: Exon skipping; MXE: Mutually exclusive exon. D) RT-qPCR to detect OGT mRNA isoforms. 22RV1 cells were treated with OGT inhibitor (20 μM OSMI-4), CDK12/13 inhibitor (500 nM THZ531), or combination for 4 h and the mRNA levels were measured using RT-qPCR (n = 3). We employed rMATS (Shen et al. 2014) and SUPPA (Alamancos et al. 2015) algorithms to assess if short-term (6 h) CDK12/13 inhibition affects splicing across the entire transcriptome. We detected thousands of alternative splicing (AS) events by comparing CDK12/13 inhibitor treatment to control (Fig. 2C). These data are in agreement with a recent paper revealing significant remodeling of splicing in pancreatic cancer cells in response to CDK12 inhibition (Panzeri et al. 2023). Based on our data, CDK12/13 inhibition particularly affects intron-exon definition, as indicated by the combined high number of events in “exon skipping”- and “intron retention”-categories. Genes that were differentially spliced in response to the dual inhibition of CDK12/13 were significantly enriched for NFκB-signaling (Suppl. Fig. 5 and Suppl. Table 2), which was previously shown to be dependent on CDK12 (Henry et al. 2018). In order to validate the CDK12/13 inhibitor-dependent effects on splicing, we selected OGT mRNA, which represents a dynamic model-system to probe if CDK12/13 activity is required for splicing of a specific mRNA as proposed by the global effects on splicing (Fig. 2C). OGT mRNA is co-transcriptionally spliced as most of the mRNAs, except for one intron, intron number 4 (detained intron 4; DI4) (Boutz et al. 2015). In response to a decrease in the OGT activity, DI4 is rapidly spliced away, which leads to generation of the mature, productive mRNA (Park et al. 2017). We used OGT mRNA as a dynamic model system to probe the importance of CDK12/13 activity on splicing. Depletion of CDK12/13 activity with 500 nM THZ531 blocked the processing of the OGT mRNA in response to OGT inhibition (Fig. 2D). Together with the computational data on alternative splicing, our results confirm that CDK12/13 inhibition remodels splicing in PC cells. So far, we have shown that the dual inhibition of CDK12/13 increases O-GlcNAcylation of the spliceosomal proteins and affects global splicing. However, we did not yet know if the increase in O-GlcNAcylation of the spliceosomal proteins is an adaptive mechanism for cells to respond to the decrease in CDK12/13 activity, or a bystander effect: this can be tested by co-targeting CDK12/13 and the spliceosome. Inhibition of CDK12/13 causes dependency on the non-essential components of the spliceosome Specific inhibitors targeting the spliceosome have been described, which enabled us to probe if the decline in CDK12/13 activity causes dependency on the spliceosome (Fig. 3A). Serine/arginine-rich splicing factors (SR proteins) bind to the phosphorylated RNA Pol II and are transferred to the emerging transcript to stimulate the formation of spliceosome (Zhong et al. 2009; Zhang et al. 2021). SR proteins are particularly important for the recognition of alternative/weak splice sites, and these proteins are regulated via phosphorylation by serine-arginine protein kinases (SRPKs) and CDC-like kinases (CLKs), and via methylation by arginine methyltransferases (PRMT). Inhibition of CLK1/4 or PRMT5 activities selectively affects splicing of the different subsets of mRNAs (Boutz et al. 2015; Tan et al. 2019). In contrast, the core spliceosome components, including SF3B, are required for splicing of most mRNAs (Zhang et al. 2021). Loss of CDK12 causes dependency on the non-essential splicing regulator SRPK1. A) Schematic of core spliceosome along with regulators that we targeted in the combinatorial lethal screen. B) Heatmap of 22RV1 cell number after four-day treatment as indicated (n: 2–3; CDK12/13 inhibitor THZ531: 50 nM). C) CRISPR-mediated depletion of SRPK1, CLK1 and CLK4 has minimal effect on cell number (non-essential gene: 0; median of all pan-essential genes: −1). Data was accessed through the DepMap CRISPR-database (DepMap 21Q4 public+score, Chronos) (Meyers et al. 2017). D) SRPK1 is significantly overexpressed in the CDK12-mutant tumors when compared to CDK12-wild type tumors. Primary prostate cancer (PC): The TCGA dataset (Taylor et al. 2010) and castration-resistant prostate cancer (CRPC): The SU2C/PCF dream team-dataset (Abida et al. 2019); P-values are those reported in the cBioPortal (Cerami et al. 2012; Gao et al. 2013). We employed six different spliceosome inhibitors targeting SRPK1 (SRPIN340) (Fukuhara et al. 2006), CLKs (TG-003 and KH-CB19) (Muraki et al. 2004; Fedorov et al. 2011), PRMT5 (GSK-3326595 developed by GSK) or the core spliceosome, SF3B (Planienolide B and E7107) (Kotake et al. 2007) (Fig. 3A). All of these compounds sensitized CRPC cells to CDK12/13 inhibition (Fig. 3B). Remarkably, targeting SRPK1, CLK1/4 and PRMT5 did not affect cell viability as a single agent, but greatly sensitized the CRPC cells to CDK12/13 inhibitor. In contrast, compounds targeting the core spliceosome factor SF3B were toxic as a single agent treatment, too. Synthetic lethal partner is a factor that becomes essential only in the presence of another anomaly. Using the DepMap-data, we determined that SRPK1, CLK1 and CLK4 are non-essential genes in most cell types, while PRMT5 and SF3B1 are essential (Fig. 3C). These effects are in accordance with the results we obtained using the small molecule inhibitors (Fig. 3B). In aggregate, we identified SRPK1, CLK1 and CLK4 as candidate synthetic lethal partners with CDK12 inactivation. We hypothesized that the increased expression of specific spliceosome regulators would be beneficial for cells with the compromised CDK12 activity. To test this, we evaluated if the genes encoding for SRPK1, CLK1, CLK4, PRMT5 and SF3B1 are overexpressed in the CDK12 mutant tumors. Only SRPK1 was significantly overexpressed in the CDK12 mutant primary tumors and also clearly increased in the CRPC disease (in the case of CRPC, the change was not significant, Fig. 3D). In contrast, the expression of CLK1, CLK4, PRMT5 and SF3B1 was not consistently different in the CDK12 mutant PC and CRPC cases (Suppl. Fig. 6). In aggregate with cell proliferation assays, gene essentiality and clinical data from primary and CRPC tumors, we identify SRPK1 as the most promising combinatorial lethal partner with the loss of CDK12 activity. We used synergy scores to confirm the efficacy of combined targeting of CDK12/13 and SRPK1. The combination treatment with CDK12/13 and SRPK1 inhibitors synergistically suppressed the viability of CRPC cells as demonstrated by the Bliss and HSA synergy calculations (Fig. 4A). Next, we used a pan-CDK inhibitor that also targets CDK12 (Dinaciclib) (Johnson et al. 2016) and another inhibitor of SRPK1 (SPHINX31) (Batson et al. 2017), to confirm the combinatorial anti-proliferative effects (Fig. 4B and Suppl. Fig. 7). Earlier, we discovered that the co-targeting of OGT and CDK12/13 induces increase in phosphorylation of H2AX, indicating increase in DNA damage, and we therefore assessed if this also occurs when inhibitors of CDK12/13 and SRPK1 are combined. Indeed, co-targeting of CDK12/13 and SRPK1 augmented phosphorylation of H2AX (Fig. 4C and Suppl. Fig. 8). The increase in H2AX phosphorylation is likely explained in part through activation of cell death. Indeed, similar to the combined inhibition of OGT and CDK12, depletion of SRPK1 activity augmented CDK12 inhibitor-induced cell death (Fig. 4C and Suppl. Fig. 8). To use an orthogonal approach, we performed knockdown experiments, which confirmed that the co-depletion of CDK12 and SRPK1 is significantly more toxic to CRPC cells than the single knockdowns (Fig. 4D and Suppl. Fig. 9). Combinatorial targeting of CDK12/13 and SRPK1 is synergistically toxic to CRPC cells. A) 3D synergy maps between CDK12/13 inhibitor THZ531 and SRPK1 inhibitor SRPIN340. CRPC cells were treated for 3 days after which cell viability was assessed (n = 2). Bliss- and HSA-score values higher than 10 indicate clear synergistic effects (Ianevski et al. 2020). B) CRPC cells were treated with SRPK1 inhibitor SPHINX31 and CDK12/13 inhibitor THZ531 (100 nM) as indicated for 4 days, and cell viability measured using CellTiterGlo. Paired two-tailed Student’s t-test was used to assess the significance (n: 3). C) Cells were treated as indicated for 24 h and western blot was used to evaluate levels of proteins of interest (n: 3–5; densitometry of all replicates provided in Suppl. fig. 8). D) CRPC cells were reverse-transfected with the indicated siRNAs for 4 days after which cell viability was assessed using CellTiterGlo-assay. Data shown is from four biological replicate-experiments and two-tailed paired Student’s t-test was used to confirm the significance. Please note that the siCDK12_1 and siCDK12_2 alone data are the same as presented in Fig. 1D We wanted to establish if the decrease in CDK12 activity sensitizes also normal cells to SRPK1 inhibition. To assess this, we repeated the cell viability, knockdown and western blot experiments in cell lines derived from normal prostate epithelia. Indeed, also normal cells become dependent on high SRPK1 activity when CDK12 is inhibited either using small molecule inhibitor that targets also CDK13 or knockdown of CDK12 (Suppl. Fig. 10). These data show that decrease in CDK12 activity sensitizes cells to SRPK1 inhibitors, and proposes the inactivation of CDK12 as a biomarker of sensitivity. To summarize the major findings of our manuscript, we reveal that the co-targeting of either OGT or spliceosome is combinatorically lethal to cells with lowered CDK12 activity. We propose that both OGT and SRPK1 are required to maintain the expression of DNA replication and repair factors to suppress DNA damage caused by the decrease in CDK12 activity. Our data identifies the non-essential spliceosome component, SRPK1, as a synthetic lethal partner with CDK12 inactivation. Discussion Here, we discovered a combinatorial lethal interaction between the compromised CDK12 activity and the non-essential components of the spliceosome. This project was initiated based on two notions: defects in CDK12 define an aggressive subtype of PC (Cancer Genome Atlas Research 2015; Wu et al. 2018; Abida et al. 2019; Schweizer et al. 2020), and the combined inhibition of OGT and CDK9, a functionally related kinase to CDK12, induces cytotoxic effects (Itkonen et al. 2020). Development of the highly specific inhibitors against both OGT and the transcriptional kinases has enabled direct dissection of the crosstalk between phosphorylation and O-GlcNAcylation. Inhibition of the transcriptional kinases remodel OGT-activity in a selective manner. Previously, we found that CDK7 inhibition results in a hyper-O-GlcNAcylation of CDK9 in CRPC cells but not in the hormone-dependent PC cells, and renders the CRPC cells addicted to the high CDK9 activity (Pallasaho et al. 2022). In response to CDK9 inhibition, OGT activity is required to recruit the DNA damage response protein MRE11 to chromatin (Gondane et al. 2022). These studies revealed that depleting the activity of CDK9 causes DNA damage, which can be augmented by targeting OGT. In this study, our focus was on the third major transcriptional kinase, CDK12. We discovered that, similar to inhibition of CDK9, OGT inhibition augments the CDK12 inhibitor-induced DNA damage (Fig. 1F). However, CDK12 inhibition does not increase chromatin loading of MRE11 (data not shown), and our data therefore propose that CDK12 inhibition induces a distinct remodeling of OGT activity compared to the targeting of the other two transcriptional kinases. In earlier studies, CDK12 inhibitor-induced DNA damage is explained through defective expression of the DNA repair genes (Dubbury et al. 2018; Krajewska et al. 2019), and we reasoned that OGT may also be important for the expression of the factors involved in maintaining genomic stability. Indeed, here we show that OGT inhibition reduces the expression of DNA replication- and cell cycle-proteins (Fig. 1E). Using glycoproteomics, we discovered that inhibition of CDK12/13 induces hyper-glycosylation of the spliceosomal proteins (Fig. 2B). Spliceosome is a complex machinery, and we performed a combinatorial lethality screen and integrated the results with gene-essentiality data from DepMap database (Meyers et al. 2017) along with clinical data to identify synthetic lethal targets with CDK12 inactivation (Fig. 3). Using this strategy, we identified SRPK1 as a kinase that is non-essential to most cell types (Fig. 3C), is overexpressed in PC (Fig. 3D), and its inhibitor synergizes with compounds targeting CDK12/13 (Fig. 4A). However, both normal and cancer cells become highly sensitive against inhibitors of OGT and SRPK1 if they have lowered activity of CDK12, which means that this biological dependency is conserved. In the future, it is of a great interest to establish if the expression of the DNA damage repair machinery is selectively dependent on the different components of the spliceosome, in particular SRPK1. In brief, SRPK1 forms a synthetic lethal pair with CDK12 inactivation. Integration of the genome-wide CRISPR knockout- and compound-screening data represents an appealing strategy to identify synthetic lethal partners. Normal and cancer cells utilize the same machineries to carry out their “functions”, which in the case of cancer cells is to produce more cells. Cancer cell-specific genetic alterations are a characteristic that segregates the malignant cells from the normal, and the identification of synthetic lethal targets holds great promise. This has led to selective and truly life-changing therapies, one of the most famous examples being Imatinib, the compound that blocks the proliferation of BCR-ABL fusion-dependent cells (Cohen et al. 2021). More thorough utilization of the genome-wide dependency data will lead to novel actionable synthetic lethal interactions. To conclude, here we report a novel combinatorial lethal interaction between the compromised CDK12 activity and the non-essential components of the spliceosome. Our study shows how OGT activity is rapidly remodeled in response to decrease in the CDK12/13 activity. Remodeling of OGT activity, and the abundance of other post-translational modifications, can be established using similar immunoprecipitation-based approach, which should be taken advantage of in the future studies. Materials and methods Cell culture, compounds, preparation of cell lysates and isolation of mRNA LNCaP, 22RV1 and C4-2 cell lines were obtained from the American Tissue Culture Collection (ATCC) and were maintained in RPMI medium supplemented with 10% fetal bovine serum (FBS). LN95 cell line was kindly provided by Professor Stephen Plymate (University of Washington), these cells were maintained in phenol red-free RPMI supplemented with 10% charcoal-stripped FBS. All the cell lines were allowed to adhere for at least one day prior to the treatment with inhibitors. THZ531, OSMI-4, TG003, SRPIN340, GSK3326595, Dinaciclib and SPHINX31 were obtained from MedChemExpress. E7107 was a gift from H3Biosciences/Esai. Transfection of Silencer® Select siRNAs against CDK12 (s28622 and s28623), OGT (16094) and SRPK1 (13450) was achieved using RNAiMax (ThermoFisher Scientific). For the validation of knockdown, RNA was isolated after 24 h of transfection with illustra RNAspin Mini RNA Isolation Kit (Cytiva) and cDNA was synthesized using qScript cDNA Synthesis Kit (QuantaBio). Following primers were used in RT-qPCR: OGT productive mRNA, (F-ACTGTGTTCGCAGTGACCTG, R-CAAATTTCCCCTTGTGCATT), OGT detained intron mRNA (F-ACTGTGTTCGCAGTGACCTG, R-AGTTGAAGACTTGGCAAAAAGT), CDK12 (F-CTAACAGCAGAGAGCGTCACC, R-AAAGGTTTGATAACTGTGCCCA), SRPK1 (F- GTTCGCAATTCAGACCCTAATGA, R-GGATGATACGGCACTTGGTATG), Actin (F-TGGGACGACATGGAGAAAAT, R- AGAGGCGTACAGGGATAGCA). For viability assays, knockdown was performed for 4 days and viability measured using CellTiterGlo-reagent (Promega). For western blot-analysis, the samples were prepared as previously described (except no sonication) (Itkonen et al. 2013). Protein concentration was determined with bicinchoninic acid (Pierce™ BCA Protein Assay Kit, Thermo Scientific). Antibodies used were: from Santa Cruz Biotechnology: p-H2AX (517348), HNRNPU (sc-32315), from Abcam: Actin (ab49900), and from Cell Signaling technologies: PARP (9532). Signal intensity of western blot (densitometry) was determined using Image Lab version 6.0 (Bio-Rad). Bioinformatics The RNA-seq data of LNCaP cells treated with control or 500 nM THZ531 was downloaded from BioProject (Accession: PRJCA004744) (Lei et al. 2021). First, the raw fastq files were aligned to human genome Hg38 using STAR aligner (Dobin et al. 2013; Gondane and Itkonen 2023). To identify the alternatively spliced sites, we used rMATS v4.1.1 (Shen et al. 2014) and SUPPA (Alamancos et al. 2015). The essentiality and the expression level of spliceosome components in patient samples were assessed using the data available through the DepMap CRISPR-database (DepMap 21Q4 Public+Score, Chronos) (Meyers et al. 2017) and cBioPortal (Cerami et al. 2012; Gao et al. 2013), respectively. The volcano- and box-plots were made in R Studio. Proliferation assays Proliferation rate of cells was measured using the Incucyte live cell imaging system following manufacturer’s (Sartorius) instructions. For viability assays, we used CellTiter-Glo® and CellTiter-Glo® 2.0 assays (Promega). Synergy calculations were performed using the tool reported by Ianevski et al. (Ianevski et al. 2020). Crystal violet staining assay was used to assess the relative number of cells as previously described (Barkovskaya et al. 2020), with minor changes: Cells growing on a 96-well plate were fixed first with ice-cold 70% methanol for 2 min and then with 100% methanol for 10 min. Cells were allowed to dry after which they were stained with 0.05% crystal violet solution for 10 min and washed twice with deionized water. After the cells dried out, 50 μL of 10% acetic acid was added to each well, the cells were de-stained for 15 min on a plate shaker and the absorbance was measured. For statistical analysis, paired two-tailed Student’s t-test was used unless otherwise mentioned. Mass spectrometry For O-GlcNAc immunoprecipitation, we combined two antibodies: RL2 (Abcam: ab2739) and CTD110.6 (Cell Signaling Technologies: 9875), and used Pierce Direct IP Kit (ThermoFisher Scientific) according to manufacturer’s protocol to prepare the biological triplicate samples for mass spectrometry (MS). MS was performed by the Weill Cornell Medicine (WCM) Meyer Cancer Center Proteomics & Metabolomics Core Facility. Lectin pulldown To enrich for O-GlcNAcylated proteins for western blot, we performed lectin pulldown experiments as previously described (Itkonen and Mills 2015). In brief, we treated 22RV1 cells with 500 nM THZ531 for 4 h, prepared cell lysates using lysis buffer (300 mM NaCl, 50 mM TRIS–HCl pH 7.4 and 2 mM EDTA), diluted samples to have 100 mM NaCl, 50 mM TRIS–HCl pH 7.4 and 2 mM EDTA, and performed wheat germ agglutinin (WGA) lectin (Vector Laboratories: AL-1023S) enrichment overnight. Next morning, the samples were washed three times using 100 mM NaCl, 50 mM TRIS–HCl pH 7.4 and 2 mM EDTA. For negative control, we used agarose beads. Finally, samples were analyzed using western blotting. Supplementary Material Acknowledgments The authors wish to acknowledge CSC – IT Center for Science, Finland, for the computational resources. Mass spectrometry was performed by the WCM Proteomics and Metabolomics Core Facility. Preliminary data for this project was generated by HMI in Professor Suzanne Walker’s laboratory (Harvard Medical School, supported by National Institutes of Health grant R01 GM094263). E7107 was a gift from H3Biosciences/Esai. Contributor Information Satu Pallasaho, Department of Biochemistry and Developmental Biology, Faculty of Medicine, University of Helsinki, Helsinki 00014, Finland. Aishwarya Gondane, Department of Biochemistry and Developmental Biology, Faculty of Medicine, University of Helsinki, Helsinki 00014, Finland. Julia Kutz, Project group Biochemistry, Leibniz Institute on Aging - Fritz Lipmann Institute, Jena D-07745, Germany. Jing Liang, Department of Biochemistry and Developmental Biology, Faculty of Medicine, University of Helsinki, Helsinki 00014, Finland. Shivani Yalala, Department of Biochemistry and Developmental Biology, Faculty of Medicine, University of Helsinki, Helsinki 00014, Finland. Damien Y Duveau, Division of Preclinical Innovation, National Center for Advancing Translational Sciences, NIH, Rockville, MD 20850, United States. Helmut Pospiech, Project group Biochemistry, Leibniz Institute on Aging - Fritz Lipmann Institute, Jena D-07745, Germany; University Hospital and Medical Faculty of the Heinrich-Heine University Düsseldorf, Life Science Center, Düsseldorf D-40225, Germany. Craig J Thomas, Division of Preclinical Innovation, National Center for Advancing Translational Sciences, NIH, Rockville, MD 20850, United States; Lymphoid Malignancies Branch, Center for Cancer Research, NCI, NIH, Bethesda, MD, United States. Massimo Loda, Department of Pathology and Laboratory Medicine, Weill Cornell Medicine, New York-Presbyterian Hospital, New York, New York 10021, United States; The Broad Institute of Harvard and MIT, Cambridge, MA 02142, United States; The New York Genome Center, New York, New York 10013, United States. Harri M Itkonen, Department of Biochemistry and Developmental Biology, Faculty of Medicine, University of Helsinki, Helsinki 00014, Finland; Department of Pathology and Laboratory Medicine, Weill Cornell Medicine, New York-Presbyterian Hospital, New York, New York 10021, United States. Author contributions Satu Pallasaho (Data curation [supporting], Formal analysis [supporting], Funding acquisition [supporting], Investigation [supporting], Methodology [supporting], Validation [supporting], Visualization [supporting], Writing—original draft [supporting], Writing—review & editing [supporting]), Aishwarya Gondane (Data curation [supporting], Formal analysis [supporting], Funding acquisition [supporting], Investigation [supporting], Methodology [supporting], Software [lead], Validation [supporting], Visualization [supporting], Writing—original draft [supporting], Writing—review & editing [supporting]), Julia Kutz (Formal analysis [supporting], Writing—original draft [supporting]), Jing Liang (Formal analysis [supporting], Writing—original draft [supporting]), Shivani Yalala (Formal analysis [supporting], Writing—original draft [supporting]), Damien Duveau (Investigation [supporting], Writing—original draft [supporting]), Helmut Pospiech (Formal analysis [supporting], Writing—original draft [supporting]), Craig J Thomas (Investigation [supporting], Writing—original draft [supporting]), Massimo Loda (Investigation [supporting], Writing—original draft [supporting]), and Harri Itkonen (Conceptualization [lead], Data curation [lead], Formal analysis [lead], Funding acquisition [lead], Investigation [lead], Methodology [lead], Project administration [lead], Resources [lead], Supervision [lead], Writing—original draft [lead], Writing—review & editing [lead]) Funding This work was supported as follows. S.P. is supported by the Emil Aaltonen Foundation, Waldemar von Frenckells Stiftelse, Otto A. Malm Foundation and University of Helsinki Doctoral Programme in Biomedicine. A.G. is supported in part by the K. Albin Johanssons Foundation, Oskar Öflund Foundation and Magnus Ehrnrooth Foundation. HMI is grateful for the funding from the Academy of Finland (Decision nr. 331324, nr. 335902 and nr. 358112), the Jenny and Antti Wihuri Foundation, Sigrid Juselius Foundation and the Cancer Foundation Finland. The Fritz Lipmann Institute is a member of the Science Association “Gottfried Wilhelm Leibniz” (WGL) and financially supported by the Federal Government of Germany and the State of Thuringia. Conflict of interest statement: None declared. Data availability All the data is presented as part of the manuscript. Raw data is available upon a reasonable request. References Associated Data Supplementary Materials Data Availability Statement All the data is presented as part of the manuscript. Raw data is available upon a reasonable request.