Abstract
Up to 90% of patients with high-grade serous ovarian cancer (HGSC) will develop resistance to platinum-based chemotherapy, posing substantial therapeutic challenges due to a lack of universally druggable targets. Leveraging BenevolentAI’s artificial intelligence (AI)–driven approach to target discovery, we screened potential AI-predicted therapeutic targets mapped to unapproved tool compounds in patient-derived 3D models. This identified TNIK, which is modulated by NCB-0846, as a novel target for platinum-resistant HGSC. Targeting by this compound demonstrated efficacy across both in vitro and ex vivo organoid platinum-resistant models. Additionally, NCB-0846 treatment effectively decreased Wnt activity, a known driver of platinum resistance; however, we found that these effects were not solely mediated by TNIK inhibition. Comprehensive AI, in silico, and in vitro analyses revealed CDK9 as another key target driving NCB-0846’s efficacy. Interestingly, TNIK and CDK9 co-expression positively correlated, and chromosomal gains in both served as prognostic markers for poor patient outcomes. Combined knockdown of TNIK and CDK9 markedly diminished downstream Wnt targets and reduced chemotherapy-resistant cell viability. Furthermore, we identified CDK9 as a novel mediator of canonical Wnt activity, providing mechanistic insights into the combinatorial effects of TNIK and CDK9 inhibition and offering a new understanding of NCB-0846 and CDK9 inhibitor function. Our findings identified the TNIK-CDK9 axis as druggable targets mediating platinum resistance and cell viability in HGSC. With AI at the forefront of drug discovery, this work highlights how to ensure that AI findings are biologically relevant by combining compound screens with physiologically relevant models, thus supporting the identification and validation of potential drug targets.
Introduction
In 2023, an estimated 19,880 women were diagnosed with ovarian cancer, and 12,810 died, making it one of the top five causes of cancer death in women in the United States (1). Globally, 313,595 women were diagnosed, and 207,252 women succumbed to the disease in 2020 (2). High-grade serous ovarian cancer (HGSC) is the most common and lethal subtype of ovarian cancer, with a ∼35% 5-year overall survival rate for advanced stages, which constitute the vast majority of cases (3, 4). Standard-of-care treatment typically involves cytoreduction surgery coupled with adjuvant, and often neoadjuvant, platinum-based chemotherapy (5). However, ∼70% to 90% of patients with HGSC will recur, and almost all patients with HGSC will eventually become platinum-resistant (6–8). Treatment options for patients with platinum-resistant HGSC are currently limited. Although paclitaxel, gemcitabine, topotecan, and bevacizumab can be used, the overall response rates to these therapies remain low (9). Additionally, the lack of recurrent mutations and high degree of genomic disarray in this cancer results in a lack of targetable proteins (10). Artificial intelligence (AI)–based drug discovery is rapidly gaining momentum, driven by the promise of more cost-effective, impactful, and faster advancements in cancer treatment (11–13). However, the potential to identify genuinely novel targets with these models has been limited by the need to ensure that AI outcomes are rooted in real biological contexts. To combat these issues, we utilized BenevolentAI’s proprietary AI-driven Benevolent Platform coupled with patient-relevant models to identify potential new therapeutic targets for platinum-resistant disease.
Traditional 2D cell line screens can often yield unreliable drug targets as they do not accurately mimic human disease (14, 15). Screening with 3D cell spheroids better represents disease biology, which is crucial for conditions like recurrent HGSC in which tumor cells metastasize as spheres (14, 16). However, these methods are typically low-throughput and not normally suitable as a primary screening approach (15). Here, we took an integrative approach combining disease and platinum resistance–relevant data with AI to generate a shortlist of potentially new therapeutic targets together with compounds identified as modulators of these targets that can be readily screened in 3D. In doing so, this not only allows us to verify the physiological relevance and druggability of these predicted targets but also offers the opportunity to identify a multitargeted approach.
From this approach, we identified NCB-0846, a documented TNIK (TRAF2- and NCK-interacting kinase) inhibitor, as a novel agent that demonstrated efficacy across a wide array of patient-derived in vitro cell and ex vivo organoid platinum-resistant models. NCB-0846 has been shown to alter the conformational structure of TNIK to directly inhibit its enzymatic activity (17). This compound has been utilized in a number of cancers such as colorectal cancer and non–small cell lung cancer, among others, and has been shown to be orally available and have established bioavailability, toxicology, and pharmacology profiles (17–25). TNIK, or MAP4K7, is a serine/threonine kinase that is most well-known for its role in the Wnt signaling pathway where it acts as a coactivator of the TCF4/β-catenin complex to drive transcriptional activation of Wnt target genes (26). Interestingly, TNIK amplification has been shown to be essential for tumorigenesis across multiple different cancers (27, 28). Numerous studies have demonstrated the importance of Wnt activation in driving HGSC stemness and platinum resistance; however, none have identified TNIK as a mediator of these effects (29–31). Here we found treatment with NCB-0846 led to a significant reduction in key Wnt target expression across several in vitro models. However, inhibiting TNIK alone did not consistently match the efficacy of NCB-0846 across all cell lines. Therefore, we employed a combination of cell-free, in silico, and AI techniques, which led to the identification of CDK9 as another key target of NCB-0846 and demonstrated targeted inhibition of both TNIK and CDK9 to be more effective than either one individually.
CDK9 is the catalytic subunit of the positive transcription elongation factor complex and directly phosphorylates RNA polymerase II to drive transcription elongation (32). Interestingly, CDK9 has been previously shown to be an attractive target in numerous cancers with studies in both the preclinical and clinical settings and specifically has been shown to be elevated in recurrent HGSC (33–35). Currently, a CDK9 inhibitor is undergoing a phase I/II clinical trial (NCT04718675) in relapsed, refractory ovarian cancer having one or more of the following alterations: BRCA1/2 mutation, MYC amplification/overexpression, or homologous recombination deficient positive (36). Additionally, CDK9 inhibitors have been shown to work synergistically with BRD4 inhibitors in both wild-type and caspase-8 knockout HGSC cells as well as with olaparib in BRCA1-proficient epithelial ovarian cancer (37, 38). However, the full role that CDK9 could be playing in these contexts requires further investigation. Here, we identified that TNIK and CDK9 are significantly co-expressed in tumors of patients with HGSC, and chromosomal gain in both was shown to be prognostic for poor disease-free and overall survival outcomes. Mechanistic studies revealed that the phenotypes observed with NCB-0846 are mediated in part by CDK9 activity. Moreover, selective CDK9 inhibition alone reduced canonical Wnt activity, directly connecting CDK9 to Wnt activation for the first time. This suggests a novel mechanism for regulating Wnt activity and enhances our understanding of NCB-0846 and CDK9 inhibitor efficacy. Taken together, we report the findings of an AI-driven target prediction and compound screen in which we identified, validated, and mechanistically characterized a novel HGSC target in TNIK, as well as CDK9, presenting a druggable multitargeted approach mediating platinum-resistant HGSC.
Materials and Methods
Cell culture
Cells were cultured in 100 mm plates (75 cm2 flasks for OV-UM32.1, OV-UM50.2, and OV-UM52.1) in a humidified atmosphere (5% CO2) at 37°C. At 70% to 90% confluence, trypsin (0.25%)/EDTA solution was used to split the cells up to passage 30. Cell culture medium (Corning) was supplemented with 10% FBS (Gibco) and 1% penicillin-streptomycin [Gibco; OV81, OV81-CP40, OV231, OV231-CP30, OV-UM32.1, OV-UM50.2, and OV-UM52.1 (DMEM); FT237 and FT194 (DMEM/F12); HEC1B (MEM); and JHUEM7 (DMEM/F12 + 0.1 mmol/L non-essential amino acids; Sigma)]. OV81 (OV81.2) and OV81-CP40 (OV81.2-CP40) are primary HGSC cell lines generated in the DiFeo laboratory as previously described (29, 39). OV231, OV-UM32.1, OV-UM50.2, and OV-UM52.1 were generated in the DiFeo laboratory from re-engrafted tumors that were derived from primary tumors of patients diagnosed with HGSC. Written informed consent was obtained from patients prior to tumor resection, and studies were conducted in accordance with the Declaration of Helsinki, Protection of Human Volunteers (21 CFR 50), and Obligations of Clinical Investigators (21 CFR 312) and approved by the University of Michigan Institutional Review Board. After the successful engraftment of tumor pieces into immunodeficient mice, tumor fragments were isolated and propagated on tissue culture plates in vitro. To confirm tumorigenicity and conserved patient genotype of these primary HGSC cells, 1,000,000 cells were injected subcutaneously into immunodeficient mice, and tumors that formed were subjected to short tandem repeat analysis to allow for comparison with the primary tumors of patients. The OV231-CP30 cell line was generated using a protocol previously described by propagating OV231 in the presence of 2.5 μmol/L cisplatin for 30 passages (29). FT194 (RRID:CVCL_A4AW) and FT237 (RRID:CVCL_UH59) were generous gifts from the Drapkin laboratory in 2016, HEC1B cells (RRID:CVCL_0294) were a generous gift from the Narla laboratory in 2017, and JHUEM7 cells (RRID:CVCL_4658) were purchased from RIKEN BioResource Research Center (#RCB1677) in 2021. All cell lines were tested monthly for mycoplasma infection (SouthernBiotech, #13100-01). In addition, the cell lines have been authenticated using genetic profiling via polymorphic short tandem repeat conducted on December 30, 2022. Experiments involving mice were approved by the University of Michigan Institutional Animal Care and Use Committee.
Sphere formation assay
A total of 2,000 cells were plated in MammoCult media + MammoCult proliferation supplement (STEMCELL Technologies, #05620) in a six-well ultralow attachment plate (Corning). Spheres were imaged blinded 6 days after plating with four representative images for each N. Spheres (greater than 10 cells) were then counted and averaged using ImageJ (RRID:SCR_003070). For cisplatin or NCB-0846 treatments, 2,000 cells were plated in MammoCult media + MammoCult proliferation supplement (STEMCELL Technologies, #05620) in a 24-well ultralow attachment plate (Corning). Treatment was added to spheres 3 days after plating and imaged 7 days after treatment using a GelCount: mammalian cell colony, spheroid and organoid counter (Oxford Optronix; https://www.oxford-optronix.com/gelcount-cell-colony-counter; RRID:SCR_023219) and analyzed using the GelCount software to obtain average sphere number per well, average sphere diameter, and total area of well covered.
Aldehyde dehydrogenase activity assay
Aldehyde dehydrogenase activity was measured using the ALDEFLUOR assay system (STEMCELL Technologies, #01700) according to the manufacturer’s protocol (https://www.stemcell.com/products/aldefluor-kit.html). Cells of 1 × 106 were resuspended in 1 mL of ALDEFLUOR assay buffer. In control samples, 5 μL of diethylaminobenzaldehyde (DEAB) was added, and all cells were incubated in a humidified atmosphere at 37°C with 5% CO2 for 30 minutes. Cells were then centrifuged and resuspended in 0.5 mL of ALDEFLUOR assay buffer. A total of 10,000 cells per sample were analyzed through FACS (Bio-Rad, ZE5 Cell Analyzer, RRID:SCR_019712), and ALDEFLUOR-positive cells were identified by a right shift from the cell-specific DEAB+ control.
Target prediction
Targets were predicted for HGSC using two methods: (i) a relational inference machine learning algorithm on top of the Benevolent Knowledge Graph (the Benevolent Platform; ref. 40) and (ii) a causal reasoning algorithm to infer regulators of a differential expression signature representing cisplatin-resistant tumor biopsies derived from GSE71340 (41). These methods were aggregated, and the targets were annotated with associations with disease-relevant mechanisms contained within the Benevolent Knowledge Graph.
Kinase selectivity profiling (Eurofins Discovery)
Kinase selectivity profiling was performed using the KinaseProfiler service (Eurofins Discovery) against a panel of 377 human wild-type kinases. For most protein kinases, a radiometric assay was used to measure the transfer of 33P from [γ-33P]ATP to specific peptide/protein substrates in standard kinase buffer [8 mmol/L 3-(N-morpholino)propanesulfonic acid (MOPS) pH 7.0, 0.2 mmol/L EDTA, with kinase-specific supplements of MgAc, MnCl2, or CaCl2/calmodulin]. For lipid kinases (PI3K family) and atypical kinases (ATM, ATR/ATRIP, and DNA-PK), a nonradiometric homogeneous time-resolved fluorescence method was employed in HEPES-based buffer systems. All assays were conducted at the ATP Km apparent for each kinase, with reactions carried out at 30°C in kinase-specific buffer conditions for 40 minutes (extended to 120 minutes for slower-reacting kinases). Results are expressed as percent kinase activity remaining compared with the control ligand with complete inhibition (0% activity) determined using kinase-specific reference inhibitors. IC50 values were determined using nine-point dose–response curves with compound concentrations starting from 1 μmol/L using half-log dilutions.
In silico target activity assignment
Deconvolution of which targets drive NCB-0846 efficacy was performed using cheminformatics and structural in silico analysis. Three approaches were taken: (i) Maximum ligand similarity of NCB-0846 to the active compounds in the ChEMBL (RRID:SCR_014042) and Reaxys Medicinal Chemistry databases for each of the selected targets (42). (ii) Binding site similarity was carried out by retrieving a set of sequence-similar proteins using phmmer, aligning these sequences using MAFFT (RRID:SCR_011811), and computing the ratio of identical residues in the binding site of TNIK compared with the selected targets (43, 44). The TNIK binding site is defined as the residues within 6 Å of the ligand in the PDB (RRID:SCR_012820) structure of TNIK (5d7a; ref. 17). (iii) Molecular docking calculations were carried out by Schrodinger Glide (2021-1, RRID:SCR_000187) using a representative crystal structure of each target with NCB-0846 as the ligand (45).
Compound screen
For cell-based viability assays, five cells per well were plated in 100 μL of MammoCult media + MammoCult proliferation supplement (STEMCELL Technologies, #05620) into 96-well, black-walled, ultralow attachment plates (Corning). BenevolentAI-identified test compounds were plated into the medium using TTP Labtech Mosquito instrumentation at 10, 3, and 1 μmol/L final drug concentrations via dilution from a 2 mmol/L stock source plate library. Replicate wells were plated in different locations for independent experiments. Positive control compound cisplatin was plated at drug-testing concentrations onto each plate. Negative control wells comprising DMSO were also included. DMSO concentration was normalized to 0.5% where needed across the entire plate. Plates were centrifuged (1 minute, 1,000 rpm) before incubation for 72 hours at 37°C with 5% CO2. Cell viability was determined using CellTiter-Glo 3D cell viability assays (Promega, #G9681) according to the manufacturer’s protocol in which 50 μL of detection reagent was added to each well (https://www.promega.com/products/cell-health-assays/cell-viability-and-cytotoxicity-assays/celltiter-glo-3d-cell-viability-assay/?catNum=G9681#protocols). Bulk reagent addition was performed by using a Multidrop Combi reagent dispenser (Thermo Scientific). Each plate was then mixed for 5 minutes prior to the endpoint read. The luminescent intensity was detected using a PHERAstar plate reader (BMG Labtech) using an LUM plus optics modules.
Proliferation and colony-forming assays
CellTiter-Glo 3D cell viability assays (Promega, #G9681) were performed according to the manufacturer’s protocol (https://www.promega.com/products/cell-health-assays/cell-viability-and-cytotoxicity-assays/celltiter-glo-3d-cell-viability-assay/?catNum=G9681#protocols). A total of 500 cells were plated in MammoCult media + MammoCult proliferation supplement (STEMCELL Technologies, #05620) in 96-well, black-walled, ultralow attachment plates (Corning). Seven days after plating, spheres were treated, and luminescence was measured after 48 hours. For 2D assays, cells were grown in their corresponding media in clear tissue culture–treated 96-well plates. CellTiter-Glo luminescent cell viability assays (Promega, #G7570) were performed according to the manufacturer’s protocol (https://www.promega.com/products/cell-health-assays/cell-viability-and-cytotoxicity-assays/celltiter_glo-luminescent-cell-viability-assay/?catNum=G7570). Cells were plated in 96-well white-walled plates (Corning). One day after plating, cells were treated, and luminescence was measured after 48 hours. 3-(4,5-Dimethylthiazol-2-yl)-2,5-Diphenyltetrazolium Bromide (MTT) assays (Thermo Fisher Scientific, #M6494) were performed according to the manufacturer’s protocol (https://www.thermofisher.com/order/catalog/product/M6494). Cells were plated in 96-well plates (Corning). One day after plating, cells were treated, and absorbance at 570 nm was measured. The area under the curves (AUC) was generated in GraphPad Prism (RRID:SCR_002798) using the trapezoid rule. Dosage points are represented as log10 unless otherwise stated. Dose–response curves were generated via sigmoidal dose response (variable slope). For combination indexes (CI), CI > 1 = antagonistic, CI = 1 = additive, and CI < 1 = synergistic; formula: (EC50_combination/EC50_cisplatin) + (EC50_combination/EC50_NCB-0846), where cisplatin EC50 = highest dose used in cell viability assay if unable to generate EC50. For colony formation assays, cells were plated in low density (∼250 cells/well) in six-well plates (Corning). Cells were treated 1 and 4 days after plating up to 7 days. After 7 days, cells were fixed and stained with 1% crystal violet solution, and the colony number per well was counted and analyzed via ImageJ (RRID:SCR_003070).
Western blotting
Western blotting was performed as previously described (46). All Western blots were controlled by blot-specific loading controls such as vinculin or GAPDH. Densitometry was performed using ImageJ (RRID:SCR_003070), and all densitometry bar graphs are the result of three biological replicates. Antibodies were purchased from (i) Cell Signaling Technology: cleaved PARP (9541, RRID:AB_331426), vinculin (13901, RRID:AB_2728768), TNIK (32712, RRID:AB_2799027), non-phospho (active) β-catenin (Ser33/37/Thr41; 8814, RRID:AB_11127203), β-catenin (8480, RRID:AB_11127855), Axin2 (2151, RRID:AB_2062432), cMYC (9402, RRID:AB_2151827), cyclin D1 (2922, RRID:AB_2228523), GAPDH (5174, RRID:AB_10622025), CDK9 (2316, RRID:AB_2291505), Rpb1 NTD (14958, RRID:AB_2687876), MCL-1 (94296, RRID:AB_2722740), Survivin (2808, RRID:AB_2063948), and cyclin T (81464, RRID:AB_2799973); (ii) Abcam: anti-RNA polymerase II CTD repeat YSPTSPS (phospho S2; ab5095, RRID:AB_304749); (iii) Santa Cruz Biotechnology: CDK9 (sc-13130, RRID:AB_627245); and (iv) Thermo Fisher Scientific: TCF4 (22337-1-AP, RRID:AB_2879076).
RNA extraction and qRT-PCR
Total RNA was extracted using the RNeasy Plus Kit (Qiagen, #74134) according to the manufacturer’s protocol (https://www.qiagen.com/us/products/discovery-and-translational-research/dna-rna-purification/rna-purification/total-rna/rneasy-plus-kits). For mRNA analysis, cDNA synthesis from 1 μg of total RNA was done using the High-Capacity RNA-to-cDNA Kit (Thermo Fisher Scientific, #4387406) according to the manufacturer’s protocol (https://www.thermofisher.com/order/catalog/product/4387406). qRT-PCR was then performed using the PowerTrack SYBR Green Master Mix for qPCR (Thermo Fisher Scientific, #A46109) according to the manufacturer’s protocol (https://www.thermofisher.com/order/catalog/product/A46109?gclid=CjwKCAjw26KxBhBDEiwAu6KXt7IWDzw7AkRufKC05jh42L-u6z7KEEGXT7TVB3zo7JrwrTMaFIhfZRoCs7UQAvD_BwE&ef_id=CjwKCAjw26KxBhBDEiwAu6KXt7IWDzw7AkRufKC05jh42L-u6z7KEEGXT7TVB3zo7JrwrTMaFIhfZRoCs7UQAvD_BwE:G:s&s_kwcid=AL!3652!3!606132911258!e!!g!!powertrack%20sybr%20green%20master%20mix!17574808712!139287687778&cid=gsd_pcr_sbu_r02_co_cp1491_pjt9631_gsd00000_0se_gaw_rs_pur_&gad_source=1) on a QuantStudio 5 Real-Time PCR System (Thermo Fisher Scientific, #A34322, RRID:SCR_020240).
Bacterial transformation and lentiviral transduction
TOP-GFP (#35489, RRID:Addgene_35489) and FOP-GFP (#35490, RRID:Addgene_35490) plasmids were obtained from Addgene. Stbl3 cells were transformed as previously described and spread onto LB Amp plates to grow at 37°C overnight (47). The following day, a single colony was collected and grown in 5 mL LB broth + 50 μg/mL ampicillin for several hours at 210 rpm at 37°C. This starter culture was added to 200 mL LB broth + 50 μg/mL ampicillin and grown overnight at 210 rpm at 37°C. Purified plasmid DNA was obtained using the Qiagen Maxiprep Kit (#12162) according to the manufacturer’s protocol (https://www.qiagen.com/us/products/discovery-and-translational-research/dna-rna-purification/dna-purification/plasmid-dna/qiagen-plasmid-kits?catno=12162). Concentrations were quantified via NanoDrop. Purified plasmids were sequenced by Plasmidsaurus. shScrambled (#SHC016), shTNIK 17 (#TRCN0000037517), and shTNIK 34 (#TRCN0000234734) plasmids were obtained from MilliporeSigma and packaged into lentivirus by the University of Michigan Vector Core (RRID:SCR_011668).
For packaging lentivirus, a 1:3 ratio of lentiviral packaging mix (psPAx2 and MD2G) was combined with 10 μg of TOP-/FOP-GFP plasmids mentioned above and were cotransfected into 5 × 106 HEK293T (RRID:CVCL_0063) cells with X-tremeGENE HP DNA (Millipore Sigma, #6366236001) transfection reagent. The supernatant-containing lentiviruses were collected after 24 hours, filtered, and stored at −80°C. For transduction, cells were infected with the viral stocks supplemented with 4 μg/mL polybrene for 12 hours and then replaced with standard growth medium.
TOP-GFP and FOP-GFP assay
Cells were plated in 60 mm plates, and treatment was added 1 day later. At each time point, media, PBS wash, and trypsinized cells were collected in a 15 mL conical and spun down at 300 g for 5 minutes. The cell pellet was washed in 1 mL PBS and spun again at 300 g for 5 minutes. After aspirating the PBS wash, cells were resuspended in 500 μL PBS and moved into a new tube for flow cytometry. A total of 10,000 cells per sample were analyzed through FACS (Bio-Rad, ZE5 Cell Analyzer, RRID:SCR_019712) and gated on GFP positivity, specifically the top ∼10% GFP-positive cells. The baseline comparison for the 48-hour time point was water, and the baseline comparison for the 6-hour time point was water for lithium chloride (LiCl) alone and DMSO after the incorporation of either NCB-0846 or NVP-2. Analysis was completed in FlowJo (https://www.flowjo.com/solutions/flowjo, RRID:SCR_008520).
siRNA transfection
Cells were transfected with siControl (#D-001810-01), siCDK9 9 (#J-003243-09), or siCDK9 11 (#J-003243-11) obtained from Horizon Discovery using the DharmaFECT 1 transfection reagent (#T-2001) and according to the manufacturer’s protocol (https://horizondiscovery.com/-/media/Files/Horizon/resources/Protocols/basic-dharmafect-protocol.pdf) at a final concentration of 100 nmol/L.
Organoid drug–response profiling
Organoid cultures were established and cultured as described elsewhere (48). To ensure patient anonymity, cultures from that study were reanonymized and assigned new patient codes. All organoid cultures consisted of only cancer cells and were confirmed to be originating from the matching tumor material of the original patient, as assessed by whole-genome sequencing.
Drug–response profiling was performed as described earlier (48). Organoids were cultured in Cultrex RGF BME type 2 (R&D Systems, #3533-010-02). Fully grown organoid cultures were extracted from BME-2, cell pellets were resuspended in fresh BME-2, and 10 μL of the suspension was plated into each well (omitting outside border wells) of precooled 384-well ultralow attachment microplates. Seeding density was adjusted to the expansion rate of a particular organoid culture using the following formula: Number of organoid culture droplets seeded into a single microplate = culture passaging ratio × 30. Gel droplets were solidified in a humidified cell culture incubator at 37°C for 30 to 45 minutes and covered with 40 μL of culture-appropriate growth medium. Culturing medium was exchanged every 2 to 3 days. After 7 to 10 days of growth postseeding experimental drugs, DMSO (vehicle control) or 10 μmol/L bortezomib (positive control) was added to the microplates using Echo 550 Liquid Handler (Labcyte). Experimental compounds were added in five 10-fold dilutions: carboplatin (MedChemExpress, #HY-17393) at 10 nmol/L to 100 μmol/L, NCB-0846 (MedChemExpress, #HY-100830) at 1 nmol/L to 10 μmol/L, and CDK9 inhibitors such as AZD4573 (ChemieTek, #CT-A4573), BAY1251152 (ChemieTek, #CT-BAY125), and NVP-2 (MedChemExpress, #HY-12214A) at 0.1 nmol/L to 1 μmol/L. Microplates were incubated at 37°C for 96 hours. Subsequently, 25 μL of the culture medium was exchanged for the staining solution, containing CellTox Green reagent (Promega, #G8743, final concentration 1×) and Hoechst 33342 (Sigma, #B2261, final concentration 5 μg/mL). Plates were incubated for 2 hours at room temperature and imaged in an ImageXpress Micro Confocal automated fluorescence microscope (Molecular Devices). A single image per well for each fluorophore was taken using 10X objective. Survival indices were estimated by image analysis using MetaXpress (Molecular Devices, RRID:SCR_016654) software. Nuclei were identified as individual objects based on Hoechst 33342 staining and computationally enlarged. Objects that overlapped with the CellTox Green signal were counted as dead cells and negative as viable cells. The fraction of viable cells was estimated in each well. The survival index in each well was estimated by normalization to negative (100% viability) and positive (0%) controls. Dose–response curves were generated using log(inhibitor) versus response−variable slope (four parameters) with bottom constrained to 0, and the AUC was calculated in GraphPad Prism (RRID:SCR_002798) using the trapezoid rule.
Co-immunoprecipitation
Cells were plated in 182 cm2 tissue culture flasks (Fisher Scientific, FB012939) and left in culture for 48 hours. Media, PBS, and trypsin were collected into a 50 mL conical and spun down. Cells were then weighed and resuspended at a 1:7 ratio of cold M-PER (Thermo Fisher Scientific, #78501) + EDTA-free phosphatase/protease inhibitors (Thermo Fisher Scientific, #A32961) and left to sit on ice for 5 minutes until they were centrifuged at 2,600 × g for 5 minutes at 4°C. The supernatant was then collected, and protein amounts were quantified via the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, #23227) according to the manufacturer’s protocol. One microgram of IgG antibody per 1 mg of Dynabeads (Protein A; Thermo Fisher Scientific, #10006D) was used for preclearing. Beads were incubated at room temperature for 15 minutes and then washed 3 times with 200 μL PBS + 0.02% Tween 20. Input samples were saved after the preclearing steps were complete. Antibodies were coupled at a ratio of 5 μg/mg Dynabeads overnight at 4°C, then washed with 200 μL PBS + 0.02% Tween 20 except for the final wash that was just PBS, and then eluted following the manufacturer’s protocol. Antibodies used are IgG (Santa Cruz Biotechnology, sc-2025, RRID:AB_737182) and CDK9 (Santa Cruz Biotechnology, sc-13130, RRID:AB_627245).
RNA sequencing profiling
The libraries were barcoded, pooled, and sequenced using paired-end 100 bp sequencing. Reads were mapped to GRCh38 using STAR version 2.5.2a (RRID:SCR_004463; ref. 49), and gene quantifications were calculated using StringTie version 2.1.1 (RRID:SCR_016323; refs. 50–52) to quantify refGene annotations. Gene read counts calculated using featureCounts (RRID:SCR_012919; ref. 53) version 1.6.1 (subread) were used to evaluate differential expression using DESeq2 version 1.30.1 (RRID:SCR_015687; ref. 54). The gene set enrichment analysis (GSEA; RRID:SCR_003199) was carried out as previously described (55, 56), using GSEA_4.2.2. software, against selected gene sets from the Molecular Signatures Database version 7.5 (RRID:SCR_016863). GSEA was conducted employing the signal-to-noise ratio ranking metric and 1,000 permutations. Plots indicating an FDR less than 0.25 were deemed significant.
Patient data analysis
The Cancer Genome Atlas (TCGA; RRID:SCR_003193) data used in the article are in whole or part based upon data generated by the TCGA Research Network (https://www.cancer.gov/tcga) and accessed through cBioPortal (https://www.cbioportal.org; RRID:SCR_014555) or https://xenabrowser.net/ (57). Survival plots were generated using TCGA: ovarian serous cystadenocarcinoma (Pan-Cancer Atlas) patient cohort with complete samples (n = 201) and analyzed via log-rank (Mantel–Cox) test.
Statistical analysis
Data were analyzed with GraphPad Prism 9 and 10 software (GraphPad Software Inc., RRID:SCR_002798). The specific statistical analysis used is denoted in figure legends. Significance was defined as P < 0.05. Significance was determined via unpaired t test relative to experimental control unless otherwise stated.
Data availability
RNA sequencing data have been deposited in NCBI’s Gene Expression Omnibus (RRID:SCR_005012) and are accessible through Gene Expression Omnibus Series accession number GSE271224. The remaining data generated in this study are available upon request from the corresponding author.
Results
TNIK identified as a novel target for platinum-resistant HGSC
To uncover novel targetable proteins for platinum-resistant HGSC, we utilized previously generated patient-derived isogenic platinum-sensitive and platinum-resistant cell lines: OV81 and OV81-CP40 and OV231 and OV231-CP30 (39). A characteristic that is frequently observed when cells acquire platinum resistance and become metastatic is their ability to form nonadherent tumor spheres, which are indicative of stem-like or cancer-initiating cells (30, 39, 58). Consistent with this observation, we found that both OV81-CP40 and OV231-CP30 formed significantly more tumor spheres when plated in ultralow attachment plates, had enhanced aldehyde dehydrogenase activity, and when grown in 3D demonstrated significantly less cell death after cisplatin treatment compared with their platinum-sensitive matched pair (Supplementary Fig. S1A–S1C). Thus, these platinum-resistant spheroids can more closely mimic how HGSC cells behave in a patient as peritoneal cancer-initiating cells have been shown to play a crucial role in developing platinum resistance in HGSC (59). Accordingly, we employed these cells grown as spheres to screen for candidate drug targets that are essential for platinum-resistant ovarian cancer cell survival.
To identify and prioritize targets to test experimentally, we utilized the Benevolent Platform (60, 61). The foundation of the platform is a disease-agnostic Knowledge Graph that integrates large quantities of diverse and independent biomedical data sources, including more than 35 million scientific articles and the contents of dozens of structured databases and genomics data sources. The Knowledge Graph can be explored by scientists using interactive tools and reasoned over to infer new relationships between biomedical concepts using proprietary algorithms as described (40). Specifically, the Knowledge Graph was used as input to a tensor factorization machine learning model that ranks protein-coding genes for a given disease or mechanism (40). Here, the model was used to prioritize putative genes that selectively kill platinum-resistant HGSC. A second approach was used to prioritize genes as potential regulators of platinum-resistant signatures from patient biopsy data and RNA sequencing derived from the cell assay. Together, the top 500 possible predicted targets were filtered for target novelty in the context of the disease, relevance to chemo-resistant biology, druggability, and safety for the indication and surfaced using information from the Knowledge Graph. Based on these assessments, 74 targets were selected for experimental validation, which included targets with a known role in driving ovarian cancer for validation purposes (Fig. 1A).
Figure 1.
A compound screen discovered the targeting potential of NCB-0846 in platinum-resistant HGSC 3D cell models. A, Schematic of AI methodology applied to target identification. B, Plot demonstrating the fold change cell viability of OV81-CP40 and OV231-CP30 cells after 10 μmol/L dose treatment for 72 hours of compounds used for screen, n = 4. C, Mutations, copy number changes, and mRNA levels in TNIK across the TCGA: ovarian serous cystadenocarcinoma (Firehose Legacy) patient cohort. The figure was adapted from cBioPortal. D,TNIK amplification across a Pan-Cancer cohort. The violin plot was adapted from the TCGA Pan-Cancer database accessed on xenabrowser.net. Data are analyzed via one-way ANOVA, F = 211.1, P < 0.0001, R2 = 0.3839. E,TNIK amplification significantly positively correlates with RNA expression–based stemness score. The figure was adapted from the TCGA: ovarian serous cystadenocarcinoma database accessed on xenabrowser.net. Data are analyzed via simple linear regression, n = 305, F = 7.596, P = 0.0062, R2 = 0.02510. F, Survival plot with or without gain in TNIK for the disease-free interval. G, Enrichment plot of “KEGG_WNT_SIGNALING_PATHWAY” and “HALLMARK_MYC_TARGETS_V2” generated via GSEA comparing OV81 and OV81-CP40 RNA sequencing with nominal P values of 0.000. ES, enrichment score; NES, normalized enrichment score.
In total, 86 compounds targeting 74 unique predicted targets were screened. Hit compounds were identified by their ability to reduce cell viability by ≥50% initially at a 10 μmol/L dose (Fig. 1B). This was supplemented by additional doses of 3 and 1 μmol/L, including control compounds targeting PLK1, to further determine compound sensitivity and efficacy across both cell lines (Supplementary Table S1; ref. 62). Out of these 86 compounds, six compounds were identified based on their ability to decrease cell viability across both cell lines. These six compounds and their corresponding documented targets are sepantronium bromide/YM155 (birc5/survivin), NCB-0846 (TNIK), EC 144 (hsp90aa1), NVP-BHG712 (epha4), L-798, 106 (PTGER3), and IP7e (NR4a2; refs. 17, 63–67). Given the novelty of the predicted targets for the disease and the mechanism of action of these six compounds, we chose to further investigate TNIK/NCB-0846 (68–74).
NCB-0846 is a documented TNIK kinase inhibitor that was first characterized by Masuda and colleagues (17). NCB-0846 has previously been shown to be effective against a multitude of cancers both in vitro and in vivo, targeting colorectal cancer stem cells and preventing lung cancer metastasis, among others (17, 20–25). However, NCB-0846 or its documented target, TNIK, has never been investigated in ovarian cancer despite Wnt activation playing a critical role in HGSC progression and platinum resistance (26, 29, 30, 75–77). Here, we find that TNIK is amplified in 30% of patients with HGSC, more than any other cancer subtype apart from lung squamous cell carcinoma (Fig. 1C and D; refs. 57, 78–80). This amplification significantly positively correlates with tumor stemness score, and TNIK gain results in a significant reduction in disease-free status in patients with ovarian serous cystadenocarcinoma (Fig. 1E and F; Supplementary Fig. S1D and S1E; ref. 57). Congruent with Wnt activation being crucial for HGSC progression and platinum resistance, we found that in platinum-resistant OV81-CP40 cells, TNIK expression was elevated, and the Wnt signaling pathway was significantly enriched compared to the parental OV81 cells (Fig. 1G; Supplementary Fig. S1F). These studies also found that MYC, an established Wnt target oncogene and driver of platinum resistance in HGSC, and its targets were enriched in OV81-CP40 cells (Fig. 1G; Supplementary Fig. S1G; refs. 81–84). Interestingly, NCB-0846 has previously been shown to significantly decrease the expression of genes enriched in “Myc Targets V2” in synovial sarcoma (22). Given these findings, we sought to further validate and investigate the targeting potential of NCB-084 on platinum-resistant HGSC.
NCB-0846 targets and sensitizes platinum-resistant HGSC across in vitro and ex vivo models
NCB-0846 was further validated with additional doses in OV81-CP40 and OV231-CP30 cells grown in 3D, generating EC50s of 709.8 and 1,588 nmol/L, respectively, while also reducing sphere-forming capabilities (Fig. 2A; Supplementary Fig. S2A and S2B). To assess the cancer specificity of this compound, we treated two distinct immortalized but nonmalignant fallopian tube secretory cell lines, FT237 and FT194, and compared them with OV81-CP40 and OV231-CP30. NCB-0846 treatment generated EC50s of 465 and 584 nmol/L in OV81-CP40 and OV231-CP30 cells, respectively, whereas FT237 and FT194 were minimally responsive (Fig. 2B). Additionally, NCB-0846 only induced apoptosis in the cancer cells as shown by increased cleaved PARP in OV81-CP40 and OV231-CP30 only (Fig. 2C; Supplementary Fig. S2C). Interestingly, NCB-0846 treatment was effective in the OV81-CP40 and OV231-CP30 platinum-sensitive isogenic counterparts (Supplementary Fig. S2D). Furthermore, to demonstrate that the observed effects of NCB-0846 were specific to target engagement, we utilized NCB-0970 that is an inactive diastereomer of NCB-0846. NCB-0970 was less effective in decreasing the cell viability in all cell lines tested compared with NCB-0846 with 48-hour EC50s of 9.92 and 3.64 μmol/L in OV81-CP40 and OV231-CP30 cells, respectively (Supplementary Fig. S2E).
Figure 2.
Validation of NCB-0846 across established and novel platinum-resistant HGSC in vitro and ex vivo models. A, Sphere-forming assay of OV81, OV81-CP40, OV231, and OV231-CP30 cells treated with NCB-0846. Data are represented as percent change sphere number ± SEM relative to DMSO, n = 3. B, CellTiter-Glo cell viability assay of OV81-CP40, OV231-CP30, FT237, and FT194 cells treated with a dose curve of NCB-0846 for 48 hours. Data are represented as mean value ± SEM fold changed to DMSO, n = 3. C, Western blot of OV81-CP40, OV231-CP30, FT237, and FT194 cells treated with 2.5 μmol/L NCB-0846 for 24 hours for cleaved PARP. D, MTT cell viability assay used to generate CI for OV81-CP40 and OV231-CP30 cells (Fig. 2E). EC50s for OV81-CP40: cisplatin = non determinable (ND), NCB-0846 = 0.4635 μmol/L, combo = 0.2313 μmol/L; EC50s for OV231-CP30: cisplatin = ND, NCB-0846 = 0.440 μmol/L, combo = 0.3532 μmol/L. Data are represented as mean value ± SEM fold changed to DMSO, n = 3. E, CI calculated using EC50s generated via MTT cell viability assay of OV81-CP40 and OV231-CP30 cells treated with a dose curve of cisplatin, NCB-0846, or a combination in which the dose was treated was the same between compounds, n = 3. F, Colony formation assay in OV81-CP40 and OV231-CP30 cells treated with DMSO, 0.5 μmol/L cisplatin, 0.125 μmol/L NCB-0846, or a combination of cisplatin and NCB-0846 over 7 days. Data are represented as mean value ± SEM fold changed to DMSO, n = 3. G, Representative image of OV-UM32.1, OV-UM50.2, and OV-UM52.1 cells grown on tissue culture–treated plates in DMEM + 10% FBS. H, qPCR of OV-UM32.1, OV-UM50.2, and OV-UM52.1 relative expression relative to 18S or GAPDH. Data are represented as mean value, n = 2–3. I and J, 3D CellTiter-Glo cell viability assay of OV-UM32.1, OV-UM50.2, and OV-UM52.1 cells treated with a dose curve of cisplatin (I) or NCB-0846 (J) for 48 hours. Data are represented as mean value ± SEM relative to DMSO, bottom constrained to 0 and top constrained to 1, n = 3. K, 3D CellTiter-Glo cell viability assay of OV-UM32.1, OV-UM50.2, and OV-UM52.1 cells treated with a dose curve of cisplatin + NCB-0846 for 48 hours. EC50s: OV-UM32.1 = 2.179 μmol/L, OV-UM50.2 = 1.577 μmol/L, OV-UM32.1 = 1.149 μmol/L. Data are represented as mean value ± SEM relative to DMSO, n = 3. L, CI calculated using EC50s generated via 3D CellTiter-Glo cell viability assay of OV-UM32.1, OV-UM50.2, and OV-UM52.1 cells treated with a dose curve of cisplatin, NCB-0846, or a combination in which the dose was treated was the same between compounds, n = 3. M, Dose–response curves of an array of primary HGSC organoids treated with carboplatin. Results are shown as mean ± SD, n = 2. N, Dose–response curves of an array of primary HGSC organoids treated with NCB-0846. Results are shown as mean ± SD, n = 2. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.
Due to the efficacies demonstrated by NCB-0846 targeting, we wanted to investigate whether treatment with NCB-0846 could sensitize OV81-CP40 and OV231-CP30 cells to cisplatin treatment. Combination indexes (CI) generated from the EC50 values of either NCB-0846 or cisplatin alone or in combination revealed that NCB-0846 and cisplatin work synergistically to eliminate both OV81-CP40 and OV231-CP30 cells with CIs of 0.52 and 0.83, respectively (Fig. 2D and E). This is further demonstrated via colony formation assay in which a combination of NCB-0846 and cisplatin for 7 days produces significantly less colonies than DMSO or single treatment alone (Fig. 2F). We then evaluated whether NCB-0846 could enhance the effects of cisplatin in platinum-sensitive OV81 and OV231 cells. Notably, NCB-0846 increased cisplatin sensitivity by at least fivefold, yielding CIs of 0.69 in OV81 and 1.00 in OV231 (Supplementary Fig. S2F and S2G). This result highlights the potential synergistic or additive capabilities of NCB-0846 targeting in the platinum-naïve context as well.
Next, we generated primary cell lines (OV-UM32.1, OV-UM50.2, and OV-UM52.1) from patients who were given varying cycles (6, 5, or 3 cycles) of neoadjuvant platinum prior to tumor-debulking surgery. These cell lines are indicative of the persistent cell population that survive after platinum chemotherapy and most likely contribute to tumor recurrence. Remarkably, these cells grow naturally as 3D tumor spheres on tissue culture–treated plates in DMEM + 10% FBS without any additional growth factors (Fig. 2G; Supplementary Table S2). We also found them to demonstrate varied expression of several Wnt, EMT, and stem genes and to have an attenuated response to platinum chemotherapy with EC50s of 12.94, 22.21, and 9.63 μmol/L in OV-UM32.1, OV-UM50.2, and OV-UM52.1, respectively, after cisplatin treatment (Fig. 2H and I). Additionally, in comparison with OV81-CP40 and OV231-CP30, all three OV-UM cell lines display elevated levels of TNIK mRNA expression further highlighting the importance of TNIK in driving platinum resistance (Fig. 2H; Supplementary Fig. S2H). To test this further, all three cell lines were treated with a dose response of NCB-0846. This generated EC50s of 4.33, 1.60, and 2.61 μmol/L in OV-UM32.1, OV-UM50.2, and OV-UM52.1, respectively (Fig. 2J). Furthermore, we observed that NCB-0846 could also sensitize these cells to cisplatin. Combination treatment together with the EC50s of cisplatin and NCB-0846 generated CIs of 0.6716, 1.057, and 0.5583 in OV-UM32.1, OV-UM50.2, and OV-UM52.1, respectively, demonstrating that NCB-0846 works synergistically with cisplatin in both OV-UM32.1 and OV-UM52.1 while having an additive effect in OV-UM50.2 (Fig. 2K and L).
Finally, we utilized an ex vivo organoid approach derived from patients with HGSC to determine the potential broad clinical applicability of NCB-0846 targets. These organoids have extensive molecular characterization and correlate with patient-specific genotypes and phenotypes including response to clinical treatment regimens, with many being resistant to carboplatin and carboplatin + paclitaxel treatment (48). Of the 13 organoids tested, all of which were derived from primary lesions, interval debulking surgery material, or recurrent disease, none showed less than 50% survival at 10 μmol/L carboplatin treatment. Five organoids exhibited IC50s in clinically relevant ranges less than 100 μmol/L with only one demonstrating a greater than 70% reduction in survival at 100 μmol/L carboplatin (Fig. 2M; Supplementary Table S3). However, treatment of all but one of the previous organoids tested, including five additional organoids, with NCB-0846 generated IC50s between 100 and 1,000 nmol/L across all organoids apart from three: P12_rPer (4,025 nmol/L), P8_rAsc (>10 μmol/L), and P10_rPIfl (>10 μmol/L), highlighting that patients that do not respond to platinum would be sensitive to NCB-0846 (Fig. 2N; Supplementary Table S3). Additionally, AUCs for NCB-0846 and carboplatin treatment had no correlation, indicating that response to carboplatin had no effect on NCB-0846 response further validating the ability of NCB-0846 to eliminate platinum-resistant HGSC (Supplementary Fig. S2I).
NCB-0846–mediated effects are not due to TNIK alone
Given the known effects of NCB-0846 on TNIK and the Wnt pathway, we investigated the effects this compound had on Wnt targets in platinum-resistant HGSC. First, we found the addition of 20 mmol/L LiCl, which has been shown to increase canonical Wnt activity through the inhibition of GSK3β, in combination with NCB-0846 resulted in a 3.7-fold increase in EC50 from 585 nmol/L in LiCl−-treated cells compared with 2.12 μmol/L in LiCl+-treated cells (Fig. 3A; Supplementary Fig. S3A; ref. 85). Additionally, LiCl+-treated cells had a significantly greater AUC, further demonstrating NCB-0846’s dependence on Wnt activity (Fig. 3B). Treatment of OV81-CP40 and OV231-CP30 cells with 2.5 μmol/L NCB-0846 for 6 and 24 hours compared with the diastereomer, NCB-0970, or DMSO control induced PARP cleavage and reduced the protein expression of Wnt targets cMYC and Axin2 by 6 hours and cyclin D1 by 24 hours (Fig. 3C; Supplementary Fig. S3B). This coincides with significant decreases in AXIN2, MYC, and CCND1 mRNA along with decreases in additional Wnt target genes DKK1 and TCF4 (Fig. 3D). Importantly, we observe these significant decreases in Wnt target genes without decreasing the activation of β-catenin, demonstrating that these observed decreases with NCB-0846 are independent of upstream β-catenin inactivation (Fig. 3C; Supplementary Fig. S3B). Interestingly, TNIK mRNA and protein expression was reduced after NCB-0846 treatment, which is consistent with it being a Wnt target gene and with previous studies (17). Furthermore, treatment of OV-UM32.1, OV-UM50.2, and OV-UM52.1 with NCB-0846 generated a significant induction of cleaved PARP by 6 hours. Like OV81-CP40 and OV231-CP30 cells, we observed loss of MYC RNA and protein and cyclin D1 protein expression but had variable Axin2 responses indicative of its complex role in Wnt signaling (Fig. 3E and F; Supplementary Fig. S3C–S3E).
Figure 3.
NCB-0846 efficacy and reduction of Wnt activity in platinum-resistant HGSC cells is mediated by additional targets besides TNIK. A, CellTiter-Glo cell viability assay of OV81-CP40 cells treated with a dose curve of NCB-0846 ± 20 mmol/L LiCl for 48 hours. Data are represented as mean value ± SEM fold changed to DMSO, bottom constrained to 0, n = 3. B, AUC of OV81-CP40 cells treated with NCB-0846 ± 20 mmol/L LiCl (A). Data are represented as mean value ± SEM, n = 3. C, Western blot of OV81-CP40 and OV231-CP30 treated with DMSO as a control or 2.5 μmol/L NCB-0846 or NCB-0970 for 6 or 24 hours. D, qPCR of OV81-CP40 and OV231-CP30 treated with DMSO as a control or NCB-0846 for 6 or 24 hours. Data are represented as mean value ± SEM fold changed to DMSO, n = 3. E, Western blot of OV-UM32.1, OV-UM50.2, and OV-UM52.1 treated with DMSO as a control or 2.5 μmol/L NCB-0846 for 6 or 24 hours. F, qPCR of OV-UM32.1, OV-UM50.2, and OV-UM52.1 treated with DMSO as a control or NCB-0846 for 6 or 24 hours. Data are represented as mean value ± SEM fold changed to DMSO, n ≥ 3. G, MTT cell viability assay of OV81-CP40 and OV231-CP30 cells stably transfected with shScrambled, shTNIK 17, or shTNIK 34 for 72 hours. Data are represented as mean value ± SEM fold changed to shScrambled, n = 3. H, Sphere forming assay of shScrambled, shTNIK 17, and shTNIK 34 OV81-CP40 and OV231-CP30 cells. Data are represented as percent change sphere number ± SEM relative to shScrambled, n = 3. I, Western blot for OV81-CP40 and OV231-CP30 cells stably transfected with shScrambled, shTNIK 17, or shTNIK 34. J, Western blot of baseline expression for OV81-CP40, OV231-CP30, and JHUEM7 cells. K, MTT cell viability assay of JHUEM7 cells treated with a dose curve of NCB-0846 for 48 hours. Data are represented as mean value ± SEM fold changed to DMSO, n = 3. L, Western blot of JHUEM7 treated with DMSO as a control or 2.5 μmol/L NCB-0846 for 6 or 24 hours. M, qPCR of JHUEM7 treated with DMSO as a control or NCB-0846 for 6 or 24 hours. Data are represented as mean value ± SEM fold changed to DMSO, n ≥ 3. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.
Because our compound screen aimed to identify targetable proteins, we sought to determine how NCB-0846 affects platinum-resistant HGSC cells through direct TNIK inhibition. To do so, we employed two distinct shRNAs (shTNIK 17 and shTNIK 34) to effectively inhibit TNIK in OV81-CP40 and OV231-CP30 cells (Supplementary Fig. S3F–S3H). Targeted inhibition of TNIK resulted in a significant decrease in cell viability in OV231-CP30 cells; however, the effects were modest in OV81-CP40 cells (Fig. 3G). This was further demonstrated through a robust inhibition of the sphere-forming capabilities of OV231-CP30 cells upon targeted TNIK inhibition compared with minimal reduction in OV81-CP40 cells (Fig. 3H). In addition, downstream changes like cMYC reduction or cleaved PARP induction were only consistently altered in OV231-CP30 (Fig. 3I; Supplementary Fig. S3H and S3I). These findings suggested that although TNIK is a critical mediator of NCB-0846 efficacy and Wnt targeting, especially in OV231-CP30 cells, our findings in OV81-CP40 cells suggest that alternative targets may also be involved in NCB-0846 sensitivity.
To investigate this further, we utilized a publicly available endometrial adenocarcinoma cell line, JHUEM7, that contains an R1320-truncating mutation in TNIK, resulting in a TNIK loss of function and total loss of TNIK protein expression (Fig. 3J; ref. 86). These cells were still initially responsive to increasing concentrations of NCB-0846; however, the reduction in cell viability was maintained at ∼50% at maximal doses (Fig. 3K). Notwithstanding the absence of functional TNIK in JHUEM7 cells, a substantial reduction in the levels of MYC and cyclin D1 protein and mRNA, as well as AXIN2 and TCF4 mRNA, was observed (Fig. 3L and M; Supplementary Fig. S3J). Further, an additional documented TNIK inhibitor present in our initial compound screen failed to exhibit efficacy across either OV81-CP40 or OV231-CP30 (Supplementary Table S1; ref. 87). Together, these results led us to investigate whether any additional targets could explain the differential response observed with NCB-0846.
CDK9 is a crucial additional target for NCB-0846
To identify additional targets of NCB-0846, we utilized a combination of cell-free and AI techniques. First, working with Eurofins Discovery, we performed a cell-free radioactive ATP kinase assay using their proprietary KinaseProfiler LeadHunter panel of 377 different human wildtype kinases treated with either NCB-0846 or NCB-0970. This assay identified 175 hits for NCB-0846, 54 of which had an IC50 of less than 100 nmol/L and 34 of which had a lower IC50 than TNIK (Fig. 4A). Of the 54 targets with an IC50 of less than 100 nmol/L, some targets were not prioritized due to having a similar or lower IC50 after NCB-0970 treatment or were not expressed in our assay system. This reduced the list to 20 targets, 13 of which were previously predicted as part of the list of 500 derived from the BenevolentAI platform (Supplementary Table S4). These 13 targets were originally deprioritized from initial screening based on their established association with the disease. However, as we aimed to identify additional drivers of NCB-0846’s effects, we have given priority to these 13 biologically relevant AI-predicted targets. This prioritization was not based on novelty but rather on their enrichment in OV81-CP40 cells, coupled with their significant implications in human cancer and/or therapy resistance. Furthermore, these targets exhibited potential synergistic interactions with TNIK, warranting their heightened consideration.
Figure 4.
CDK9 is a key additional molecular target of NCB-0846 and platinum-resistant HGSC. A, Kinome trees produced using the IC50s generated after treating with a dose response of either NCB-0846 or NCB-0970. Modified plots were generated by CORAL, produced by Phanstiel Lab. B, Docking models using representative crystal structures with NCB-0846 for TNIK and CDK9. C, Survival plots with or without gain in CDK9 for disease-free, disease-specific, progression-free, and overall survival. D, mRNA expression (RNA Seq V2 RSEM) of TNIK and CDK9 from TCGA: ovarian serous cystadenocarcinoma (Firehose Legacy) patient cohort. n = 307. E, Survival plots with or without gain in both TNIK and CDK9 for disease-free and overall survival. F, Western blot of OV81-CP40 and OV231-CP30 treated with DMSO as a control or 2.5 μmol/L NCB-0846 or NCB-0970 for 6 or 24 hours. G, qPCR of OV81-CP40 and OV231-CP30 treated with DMSO as a control or NCB-0846 for 6 or 24 hours. Data are represented as mean value ± SEM fold changed to DMSO, n = 3. H, Western blot of OV-UM32.1, OV-UM50.2, and OV-UM52.1 treated with DMSO as a control or 2.5 μmol/L NCB-0846 for 6 or 24 hours. I, Western blot of JHUEM7 treated with DMSO as a control or 2.5 μmol/L NCB-0846 for 6 or 24 hours. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.
Next, we employed in silico studies to further aid in the deconvolution of which of these additional selected targets of NCB-0846 were key to its efficacy. We employed three different approaches: (i) maximum ligand similarity of NCB-0846 to active compounds for each of the selected targets, (ii) binding site similarity of TNIK compared with the selected targets, and (iii) molecular docking calculations using a representative crystal structure of each target with NCB-0846 as the ligand. Through these analyses, we identified CDK9 as an attractive additional target for NCB-0846 as it generated a lower cell-free IC50 than TNIK, demonstrated favorable binding site similarity and Max Tanimoto, and had a molecular docking score very similar to TNIK (Fig. 4B; Supplementary Table S4). Supplementary to our own findings, CDK9 has been implicated as a potential biomarker and therapeutic target for HGSC; however, the molecular mechanisms of CDK9 in ovarian cancer remain to be fully characterized (33, 37, 38). Here, we show that TCGA data of patients with ovarian serous cystadenocarcinoma demonstrate a significant reduction in disease-free, disease-specific, progression-free, and overall survival in patients with a gain in CDK9 (Fig. 4C). Furthermore, TCGA data of 307 patients with ovarian serous cystadenocarcinoma show that there is a significant positive correlation (Spearman’s correlation = 0.16, P = 0.005756, q = 0.0337) between TNIK and CDK9 mRNA expression and gain in both TNIK and CDK9 results in a significant reduction in disease-free and overall survival (Fig. 4D and E). Additionally, both TNIK and CDK9 have been previously shown to have key overlapping downstream targets in CCND1 and MYC, and previous studies have shown NCB-0846 to decrease the expression of the CDK9 target MCL1 in colorectal cancer; however, the mechanism behind how it was able to do so remained undefined (88, 89).
Treatment of OV81-CP40 and OV231-CP30 cells with 2.5 μmol/L NCB-0846 for 6 hours and 24 hours resulted in a significant decrease in CDK9 kinase activity as observed by a loss of RNA polymerase II phosphorylation compared with DMSO (Fig. 4F; Supplementary Fig. S4A). This is further validated by the significant loss of MCL-1 protein and mRNA expression by 6 hours in both OV81-CP40 and OV231-CP30 (Fig. 4F and G; Supplementary Fig. S4A). We postulate the loss of total CDK9 protein and mRNA to be an intracellular response to the inhibition of CDK9’s kinase activity, resulting in its degradation and downregulation. In addition, we also observed a significant loss in RNA polymerase II phosphorylation and MCL-1 protein expression after NCB-0846 treatment across all three primary OV-UM cell lines along with JHUEM7 cells (Fig. 4H and I; Supplementary Fig. S4B–S4E). This observation could explain the observed loss in MYC and cyclin D1 protein and mRNA along with the partial cell death after NCB-0846 treatment in JHUEM7 cells despite being TNIK null.
CDK9 is a novel regulator of canonical Wnt signaling and cooperates with TNIK to target platinum-resistant HGSC
To further ascertain the role CDK9 has in mediating the effects of NCB-0846, we utilized targeted CDK9 inhibition. Knockdown of CDK9 by ∼75% resulted in a modest decrease in cell viability, which may be partially due to the activity of the residual protein (Fig. 5A and B; Supplementary Fig. S5A and S5B; ref. 33). However, the combination of CDK9 knockdown enhanced the effects of NCB-0846 wherein it reduced the EC50 by ∼2.5-fold from 398.8 nmol/L with siControl to 161.3 nmol/L with siCDK9 9 and 142.5 nmol/L with siCDK9 11, demonstrating NCB-0846’s partial dependency on CDK9 (Fig. 5C).
Figure 5.
TNIK and CDK9 work jointly to mediate Wnt activity and platinum-resistant HGSC survival. A, Western blot of OV81-CP40 transiently transfected with two different siRNAs for CDK9, siCDK9 9 or siCDK9 11, or an siControl, for 48 hours. B, CellTiter-Glo cell viability assay of OV81-CP40 cells transiently transfected with siControl, siCDK9 9, or siCDK9 11 for 48 hours. Data are represented as mean value ± SEM fold changed to siControl; n = 3. C, CellTiter-Glo cell viability assay of OV81-CP40 cells transiently transfected with siControl, siCDK9 9, or siCDK9 11 treated with a dose curve of NCB-0846 for 48 hours. Data are represented as mean value ± SEM fold changed to DMSO; n = 3. D, AUC of organoids treated with NCB-0846 (Fig. 2N), NVP-2, BAY1251152, or AZD4573 (Supplementary Fig. S5C). Data are analyzed via simple linear regression, n = 17: NVP-2 vs. NCB-0846: F = 21.42, P = 0.0003, and R2 = 0.5881; BAY1251152 vs. NCB-0846: F = 91.93, P ≤ 0.0001, and R2 = 0.8597; and AZD4573 vs. NCB-0846: F = 38.29, P ≤ 0.0001, and R2 = 0.7185. E, Western blot of OV81-CP40 and OV231-CP30 treated with DMSO as a control or 100 nmol/L NVP-2 for 6 or 24 hours. F, qPCR of OV81-CP40 and OV231-CP30 treated with DMSO as a control or NVP-2 for 6 or 24 hours. Data are represented as mean value ± SEM fold changed to DMSO, n = 3. G, OV81-CP40 cells stably expressing either TOP- or FOP-GFP were treated for 6 hours with 20 mmol/L LiCl, 2.5 μmol/L NCB-0846, and/or 100 nmol/L NVP-2. Data are represented as mean value ± SEM of TOP/FOP and analyzed via ordinary one-way ANOVA with multiple comparisons, n = 3. H, OV81-CP40 cells treated with 40, 20, 5, 2.5, or 1.25 nmol/L NVP-2 ± 20 mmol/L LiCl for 48 hours with MTT as a readout of cell viability. Data are represented as mean value ± SEM fold changed to DMSO, analyzed via a two-way ANOVA with multiple comparisons comparing LiCl− with LiCl+, n = 3. I, Co-immunoprecipitation and Western blotting with indicated antibodies in OV-UM50.2 cells. J, MTT cell viability assay of OV81-CP40 and OV231-CP30 cells stably transfected with either shScrambled or shTNIK 17 transiently transfected with either siControl or siCDK9 9 for 24, 48, and 72 hours. Data are represented as mean value ± SEM, analyzed via ordinary two-way ANOVA (comparing column means), n = 3. K, Western blot of OV81-CP40 and OV231-CP30 cells stably transfected with either shScrambled or shTNIK 17 transiently transfected with either siControl or siCDK9 for 48 hours. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.
Next, with selective CDK9 inhibitors currently in clinical development, we compared the response of three distinct CDK9 inhibitors (AZD4573, NVP-2, and BAY1251152) with NCB-0846 by conducting dose–response curves across the same array of patient-derived organoids previously used (Supplementary Fig. S5C; Supplementary Table S3). When we compared the AUCs of the three CDK9 inhibitors with those from NCB-0846 treatment, we found a significant positive correlation, indicating that these compounds may target similar pathways and the broad clinical applicability of TNIK and CDK9 inhibition (Fig. 5D).
To ascertain if they had similar mechanisms of action, we compared NCB-0846 with a potent and selective CDK9 inhibitor, NVP-2, across various cell models. Consistent with our findings with NCB-0846, we found NVP-2 to be cancer-specific with an EC50 that was nondeterminable up to 80 nmol/L in FT237 cells compared with OV81-CP40 in which it was 12.68 nmol/L (Supplementary Fig. S5D). In addition, NVP-2 generated an EC50 of 52.08 and 60.32 nmol/L in OV81-CP40 and OV231-CP30 spheroids, respectively, and treatment with 100 nmol/L NVP-2 induced PARP cleavage expression by 6 hours in both OV81-CP40 and OV231-CP30 compared with DMSO (Fig. 5E; Supplementary Fig. S5E). This correlated with a significant decrease in CDK9 direct targets: the phosphorylation of RNA polymerase II coinciding with a decrease in MCL-1 and MYC protein and mRNA expression (Fig. 5E and F; Supplementary Fig. S5F). Interestingly, like with NCB-0846, there was also a significant decrease in Wnt target genes, AXIN2, DKK1, and TCF4 after 6 hours and TNIK and CCND1 after 24 hours of NVP-2 treatment (Fig. 5F). This is independent of upstream β-catenin activation as significant loss was only observed after 24 hours (Fig. 5E; Supplementary Fig. S5F).
Given that CDK9 has not been shown to be associated with canonical Wnt signaling, the effect of CDK9 inhibition on canonical Wnt pathway activation was examined using reporter TOP-GFP. Interestingly, both NCB-0846 and NVP-2 treatments were able to significantly decrease TOP-GFP induction after LiCl addition (Fig. 5G). In addition, the incorporation of 20 mmol/L LiCl significantly elevated the cell viability across multiple doses of OV81-CP40 cells treated with NVP-2 (Fig. 5H). Further, we discovered that Wnt activation resulted in a 1.5-fold increase in the EC50 of NCB-0846; however, this increase was greatly diminished when CDK9 was inhibited, indicating that CDK9 partially regulates the decline in NCB-0846 efficacy after Wnt activation (Supplementary Fig. S5G). To further deconvolute the relationship between TNIK and CDK9, we performed co-immunoprecipitation experiments which discovered that CDK9 is capable of directly binding to TNIK and its well-known binding partner cyclin T, but not β-catenin or TCF4 (Fig. 5I). Together, this demonstrates a novel role for CDK9 in mediating Wnt activity and identifies a new binding partner in TNIK, suggesting a mechanism behind dual TNIK and CDK9 inhibition by jointly targeting Wnt activity.
Expanding upon these findings, we wanted to further investigate the mechanism of dual TNIK and CDK9 inhibition in mediating platinum resistance in HGSC. When combined, TNIK and CDK9 knockdown resulted in a greater reduction in cell viability than either knockdown alone (Fig. 5J). This is further supported by elevated cleaved PARP expression with dual TNIK and CDK9 knockdown (Fig. 5K). Both OV81-CP40 and OV231-CP30 cells demonstrated a significant loss of RNA polymerase II phosphorylation and MCL-1 expression with dual knockdown compared with shScrambled + siControl (Fig. 5K; Supplementary Fig. S5H). Further, cMYC protein expression was significantly more reduced after dual knockdown of TNIK and CDK9 when compared with control, shTNIK alone, or siCDK9 alone, indicating a combinatorial effect between TNIK and CDK9 to attenuate cMYC expression (Fig. 5K; Supplementary Fig. S5H).
Finally, given the significance of reducing MYC expression with dual TNIK and CDK9 inhibition, we tested the efficacy of NCB-0846 on an endometrial adenocarcinoma cell line, HEC1B, which contains an R367 truncating mutation in FBXW7, resulting in its loss of function and MYC being unable to undergo phosphorylation-dependent degradation (86, 90). A dose response up to 10 μmol/L NCB-0846 for 48 hours was unable to generate an EC50 as at no point was the cell viability less than 50% (Supplementary Fig. S6A). Additionally, treament with 2.5 μmol/L NCB-0846 was unable to induce cleaved PARP expression (Supplementary Fig. S6B). Despite the lack of response to NCB-0846, there was still a significant decrease in RNA polymerase II phosphorylation and downstream targets MCL-1, Axin2, cyclin D1, DKK1, and TCF4 (Supplementary Fig. S6B–S6D). However, although we did observe a significant loss of MYC expression at 6 hours, this was not sustained, and the expression of MYC returned to baseline at both the protein and mRNA levels after 24 hours, further indicating that the ability to reduce MYC expression plays a key role in NCB-0846’s and more broadly TNIK and CDK9 dual inhibition’s efficacy (Supplementary Fig. S6B–S6D).
Discussion
Patients with platinum-resistant HGSC suffer from a lack of targeted therapies and are still in critical need of new therapeutic options. However, novel target elucidation has been hampered by the lack of consistent targetable genetic drivers and screening practices that do not faithfully mimic human disease (14, 15). Therefore, innovative approaches are necessary to uncover new and unique oncogenic vulnerabilities to provide novel targetable avenues for patients with platinum-resistant HGSC. Here, we employed a tool compound screen using an AI-based approach that ultimately allowed us to uncover novel targetable approaches for platinum-resistant HGSC. We predicted a prioritized set of 74 targets had the potential to eliminate platinum-resistant HGSC. In this study, we identified NCB-0846, a known TNIK kinase inhibitor with established bioavailability, toxicology, and pharmacology profiles, to show efficacy in targeting platinum-resistant HGSC. This compound was able to effectively sensitize cells to cisplatin and eliminate not only platinum-resistant OV81-CP40 and OV231-CP30 cell lines generated in cultures but also three primary HGSC cell lines derived from patients who were treated with neoadjuvant platinum-based therapy, OV-UM32.1, OV-UM50.2, and OV-UM52.1. Interestingly, although OV-UM cell lines exhibited elevated TNIK expression, they were less sensitive to NCB-0846 than the OV81-CP40 and OV231-CP30 cells. This reduced sensitivity may be attributed to the highly aggressive nature of these primary cell lines, which predominantly grow in 3D without supplemental nutrients, owing to their strong stem-like characteristics, while also highlighting the significance of the compound’s dual targeting mechanism. In addition to our platinum-resistant models, we observed that NCB-0846 also enhanced the effects of cisplatin in a platinum-naïve setting, either synergistically or additively. This suggests that inhibition of NCB-0846s targets could enable lower dosing of platinum in initial therapy, potentially reducing the risk of developing resistance. Furthermore, NCB-0846 demonstrated efficacy ex vivo across a large panel of diverse, clinically relevant patient-derived HGSC organoids, highlighting the comprehensive applicability of its targets. Because TNIK activates the Wnt pathway, which is key in driving platinum resistance in HGSC, we treated several platinum-resistant HGSC cell lines with NCB-0846 and observed a significant decrease in Wnt activity.
Interestingly, upon validating the target specificity of NCB-0846, we found that genetic knockdown of TNIK resulted in inconsistent efficacy and Wnt inhibition across multiple cell models and treatment of NCB-0846 in TNIK null JHUEM7 cells showed response. Together, this demonstrated that although TNIK is an important target in driving the observed phenotypes, it is not the only one. Utilizing a combination of cell-free kinome screens, AI, and in silico binding studies, we uncovered that this compound also inhibited CDK9 activity. Interestingly, elevated CDK9 has been previously associated with shorter recurrence intervals in epithelial ovarian cancer; however, further mechanistic studies have yet to be performed (34). In this study, we find that TNIK and CDK9 chromosomal gains are prognostic for decreased disease-free and overall survival, co-express in patients, and have important downstream target overlap. Specifically, NCB-0846 was able to decrease downstream CDK9 kinase activity in all platinum-resistant models tested as well as in TNIK null JHUEM7 cells. In addition, genetic knockdown of CDK9 further sensitized cells to NCB-0846 treatment highlighting the importance of CDK9 in NCB-0846’s efficacy.
When comparing NCB-0846 with the CDK9-specific inhibitor NVP-2, in which efficacy correlated across the HGSC organoid panel, we uncovered that CDK9 inhibition could decrease Wnt activity and that stimulating β-catenin activation could reduce NVP-2’s efficacy. Additionally, NCB-0846 treatment in JHUEM7 cells reduced downstream Wnt targets without TNIK being present, and NCB-0846 treatment with CDK9 knockdown greatly lessened the reduction in efficacy observed after the addition of LiCl, uncovering CDK9 as a novel regulator of Wnt activity. These findings reach beyond NCB-0846’s mechanism of action and HGSC as CDK9 inhibitors are currently used to treat a variety of cancers, many of which also rely on Wnt signaling for driving treatment resistance and/or cancer progression (91–95). Taken together, the results shown here propose a novel mechanism of action for how CDK9 inhibitors target cancer cells. Additionally, we also discovered that CDK9 can directly bind to TNIK, but not β-catenin or TCF4, which may suggest that along with NCB-0846, any pharmacological inhibition of TNIK may affect CDK9 activity and vice versa. This could provide wide-ranging pharmacologic opportunities for future TNIK and/or CDK9 inhibitors as well as present opportunities for CDK9 inhibitor patient stratification.
Finally, dual knockdown of TNIK and CDK9 showed enhanced efficacy in decreasing platinum-resistant cell viability, with MYC reduction playing a key role. Because we observed a minimal and variable reduction in downstream targets when TNIK or CDK9 were inhibited individually, our findings indicate a possible therapeutic advantage by illustrating how combined incomplete inhibition of TNIK and CDK9 can result in substantial declines in both cell viability and targets. This underscores the benefit of concurrent TNIK and CDK9 inhibition, as it requires less inhibition of each to achieve substantial impact. Given the high toxicity issues associated with CDK9 inhibitors in clinical trials, co-inhibition could potentially mitigate these effects and enhance drug resistance evasion (96). This is particularly notable given that a CDK9 inhibitor is currently undergoing a phase I/II clinical trial for relapsed or refractory HGSC with MYC amplification/overexpression (36).
This work underscores the potential to expedite both the discovery and validation steps of therapeutic target discovery through the incorporation of strong, established AI models. In particular, utilization of the Benevolent Platform allowed for the efficient identification of a short, predicted gene list of likely targetable and druggable proteins. Further target deconvolution uncovered areas for drug optimization and enhancement, leading to possible polypharmacology approaches. Additionally, the results shown here stress the benefits of tool compound screens and having multitargeted approaches when identifying and developing targeted therapies. The use of combination therapy has been a foundation for cancer treatment for years as it can reduce the risk of drug resistance and generate greater reductions in tumor growth and metastasis along with reducing cancer stem cell populations (97). Here, we show that by targeting both TNIK and CDK9 together, we can more effectively decrease key cancer-driving molecular targets, like MYC, and treat platinum-resistant HGSC. This demonstrates that compounds targeting multiple oncogenic targets simultaneously or a combination of specific inhibitors of each individual target could be a clinically relevant therapeutic strategy. In sum, this work put into practice powerful BenevolentAI models to effectively and efficiently identify new therapeutic targets and drug candidates, uncovering novel mechanistic insights and positioning the inhibition of TNIK and CDK9 as a promising targetable approach for platinum-resistant HGSC.
Supplementary Material
Supplementary Figure S1: Characterization of OV81-CP40 and OV231-CP30 cell lines and results of GSEA for OV81-CP40 RNAseq.
Supplementary Figure S2: Validation of NCB-0846 in selectively targeting platinum-resistant HGSC across cell models.
Supplementary Figure S3: Validation of lithium chloride and Figure 3 western blot densitometry
Supplementary Figure S4: Densitometry for Figure 4 western blots
Supplementary Figure S5: CDK9 inhibition shows platinum-resistant HGSC efficacy and specificity and mediates responses to enhanced Wnt activity.
Supplementary Figure S6: Inhibition of MYC is important in TNIK and CDK9-mediated targeting efficacy as observed through HEC1B treatment with NCB-0846
Supplementary Table S1: Compounds from the BenevolentAI library created for screening in spheroid grown OV81-CP40 and OV231-CP30 cells. Results reported as mean fold change cell viability relative to DMSO control with standard deviation; n = 4.
Supplementary Table S2: Patient information and treatment regimen for OV-UM32.1, OV-UM50.2 and OV-UM52.1 cell lines.
Supplementary Table S3: IC50s of patient-derived organoids.
Supplementary Table S4: Artificial intelligence predicted additional target kinases for NCB-0846.
Acknowledgments
The authors would like to acknowledge the Rogel Cancer Center support under award number P30 CA046592. Research reported in this publication was supported by the National Cancer Institutes of Health under award number P30 CA046592 by the use of the following University of Michigan Cancer Center Shared Resources: Vector Core, Flow Cytometry and Proteomics, and Center for Chemical Genomics. The authors also acknowledge the support from BenevolentAI for the 3D drug screen studies and biomarker assessment. Further acknowledgments are extended to the generous donors and foundations who have financially supported this research in the DiFeo laboratory, striving to improve the outcomes of patients with HGSC, including the Silver Family Foundation, the Barbara Robson Foundation, and the Debra A. and Dean A. Frick Ovarian Cancer Research Fund. The authors would like to thank the Benevolent target ID product and technology team: Alix Lacoste, Davy Suvee, Alicia Frame, and Andrew Potterton, who contributed to the AI methods used to predict targets for cisplatin-resistant HGSC. Access to HGSC tumor material for organoid derivation was kindly provided through a collaboration of W. Senkowski with Prof. Sampsa Hautaniemi, University of Helsinki, and Dr. Johanna Hynninen, Turku University Hospital and the European Union’s Horizon 2020 research and innovation programme project DECIDER (grant agreement no. 965193). This work was supported by grants from the NCI, R01CA197780 (A. DiFeo), and Training Program in Translational Research T32 GM141840 (N. Puleo). Research reported in this publication was supported by the NCI of the NIH under award number P30CA046592 using the following Cancer Center Shared Resource(s): Cancer Data Science (K. Bedi). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. This work was supported by the European Union’s Horizon 2020 research and innovation programme under grant agreement no. 845045 for RESIST3D (W. Senkowski) and Danish Cancer Society under grant no. R204-A12322 (W. Senkowski).
Footnotes
Note: Supplementary data for this article are available at Molecular Cancer Therapeutics Online (http://mct.aacrjournals.org/).
Authors’ Disclosures
A. Ray, T.J. Bollerman, J. Dunbar, P. Richardson, A. Taddei, and C. Hudson are employees of BenevolentAI. W. Senkowski reports grants from the European Commission and the Danish Cancer Society during the conduct of the study. No disclosures were reported by the other authors.
Authors’ Contributions
N. Puleo: Data curation, formal analysis, writing–original draft, writing–review and editing. H. Ram: Data curation. M.L. Dziubinski: Data curation. D. Carvette: Data curation. J. Teitel: Data curation. S.C. Sekhar: Data curation. K. Bedi: Data curation, formal analysis. A. Robida: Data curation. M.M. Nakashima: Data curation. S. Farsinejad: Data curation. M. Iwanicki: Data curation. W. Senkowski: Data curation. A. Ray: Formal analysis, writing–review and editing. T.J. Bollerman: Formal analysis, writing–review and editing. J. Dunbar: Formal analysis, writing–review and editing. P. Richardson: Formal analysis, writing–review and editing. A. Taddei: Formal analysis, writing–original draft. C. Hudson: Conceptualization, resources. A. DiFeo: Conceptualization, resources, supervision, funding acquisition, writing–original draft, writing–review and editing.
References
- 1. Siegel RL, Miller KD, Wagle NS, Jemal A. Cancer statistics, 2023. CA Cancer J Clin 2023;73:17–48. [DOI] [PubMed] [Google Scholar]
- 2. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2021;71:209–49. [DOI] [PubMed] [Google Scholar]
- 3. Hu J, Wang Z, Zuo R, Zheng C, Lu B, Cheng X, et al. Development of survival predictors for high-grade serous ovarian cancer based on stable radiomic features from computed tomography images. iScience 2022;25:104628. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Cheng Z, Mirza H, Ennis DP, Smith P, Morrill Gavarró L, Sokota C, et al. The genomic landscape of early-stage ovarian high-grade serous carcinoma. Clin Cancer Res 2022;28:2911–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Matulonis UA, Sood AK, Fallowfield L, Howitt BE, Sehouli J, Karlan BY. Ovarian cancer. Nat Rev Dis Primers 2016;2:16061. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. van Zyl B, Tang D, Bowden NA. Biomarkers of platinum resistance in ovarian cancer: what can we use to improve treatment. Endocr Relat Cancer 2018;25:R303–18. [DOI] [PubMed] [Google Scholar]
- 7. Cooke SL, Brenton JD. Evolution of platinum resistance in high-grade serous ovarian cancer. Lancet Oncol 2011;12:1169–74. [DOI] [PubMed] [Google Scholar]
- 8. Institute of Medicine, National Academies of Sciences E and Medicine . Ovarian Cancers: Evolving Paradigms in Research and Care [Internet]. Washington (DC): The National Academies Press; 2016. Available fromhttps://nap.nationalacademies.org/catalog/21841/ovarian-cancers-evolving-paradigms-in-research-and-care. [PubMed] [Google Scholar]
- 9. Luyckx M, Squifflet JL, Bruger AM, Baurain JF. Recurrent high grade serous ovarian cancer management. In: Ovarian Cancer [Internet]. Brisbane (AU): Exon Publications; 2022. p. 87–103. Available from:https://www.ncbi.nlm.nih.gov/books/NBK585985/. [PubMed] [Google Scholar]
- 10. Ciriello G, Miller ML, Aksoy BA, Senbabaoglu Y, Schultz N, Sander C. Emerging landscape of oncogenic signatures across human cancers. Nat Genet 2013;45:1127–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Wouters OJ, McKee M, Luyten J. Estimated research and development investment needed to bring a new medicine to market, 2009-2018. JAMA 2020;323:844–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Hinkson IV, Madej B, Stahlberg EA. Accelerating therapeutics for opportunities in medicine: a paradigm shift in drug discovery. Front Pharmacol 2020;11:770. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Sun D, Gao W, Hu H, Zhou S. Why 90% of clinical drug development fails and how to improve it? Acta Pharm Sin B 2022;12:3049–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Suenaga H, Kagaya N, Kawada M, Tatsuda D, Sato T, Shin-ya K. Phenotypic screening system using three-dimensional (3D) culture models for natural product screening. J Antibiot (Tokyo) 2021;74:660–6. [DOI] [PubMed] [Google Scholar]
- 15. Booij TH, Price LS, Danen EHJ. 3D cell-based assays for drug screens: challenges in imaging, image analysis, and high-content analysis. SLAS Discov 2019;24:615–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Lisio M-A, Fu L, Goyeneche A, Gao Z-H, Telleria C. High-grade serous ovarian cancer: basic sciences, clinical and therapeutic standpoints. Int J Mol Sci 2019;20:952. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Masuda M, Uno Y, Ohbayashi N, Ohata H, Mimata A, Kukimoto-Niino M, et al. TNIK inhibition abrogates colorectal cancer stemness. Nat Commun 2016;7:12586. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Kim J, Oh J, Peterson HM, Carlson JCT, Pittet MJ, Weissleder R. TNIK inhibition has dual synergistic effects on tumor and associated immune cells. Adv Biol (Weinh) 2022;6:e2200030. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Ma L, Li R, Yao Z, Wang B, Liu Y, Liu C, et al. Computational study on new natural compound inhibitors of Traf2 and Nck-interacting kinase (TNIK). Aging (Albany NY) 2022;14:8394–410. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Hirozane T, Masuda M, Sugano T, Sekita T, Goto N, Aoyama T, et al. Direct conversion of osteosarcoma to adipocytes by targeting TNIK. JCI Insight 2021;6:e137245. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Sugano T, Masuda M, Takeshita F, Motoi N, Hirozane T, Goto N, et al. Pharmacological blockage of transforming growth factor-β signalling by a Traf2- and Nck-interacting kinase inhibitor, NCB-0846. Br J Cancer 2021;124:228–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Sekita T, Yamada T, Kobayashi E, Yoshida A, Hirozane T, Kawai A, et al. Feasibility of targeting traf2-and-nck-interacting kinase in synovial sarcoma. Cancers (Basel) 2020;12:1258. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Lee RS, Zhang L, Berger A, Lawrence MG, Song J, Niranjan B, et al. Characterization of the ERG-regulated kinome in prostate cancer identifies TNIK as a potential therapeutic target. Neoplasia 2019;21:389–400. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Guo J, Liang J, Wang Y, Guo T, Liao Y, Zhong B, et al. TNIK drives castration-resistant prostate cancer via phosphorylating EGFR. iScience 2023;27:108713. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Li Z, Lim SK, Liang X, Lim YP. The transcriptional coactivator WBP2 primes triple-negative breast cancer cells for responses to Wnt signaling via the JNK/Jun kinase pathway. J Biol Chem 2018;293:20014–28. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Mahmoudi T, Li VSW, Ng SS, Taouatas N, Vries RGJ, Mohammed S, et al. The kinase TNIK is an essential activator of Wnt target genes. EMBO J 2009;28:3329–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Torres-Ayuso P, An E, Nyswaner KM, Bensen RC, Ritt DA, Specht SI, et al. TNIK is a therapeutic target in Lung Squamous Cell Carcinoma and regulates FAK activation through Merlin. Cancer Discov 2021;11:1411–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Yu D-H, Zhang X, Wang H, Zhang L, Chen H, Hu M, et al. The essential role of TNIK gene amplification in gastric cancer growth. Oncogenesis 2014;2:e89. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Nagaraj AB, Joseph P, Kovalenko O, Singh S, Armstrong A, Redline R, et al. Critical role of Wnt/β-catenin signaling in driving epithelial ovarian cancer platinum resistance. Oncotarget 2015;6:23720–34. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Belur Nagaraj A, Knarr M, Sekhar S, Connor RS, Joseph P, Kovalenko O, et al. The miR–181a–SFRP4 Axis regulates Wnt activation to drive stemness and platinum resistance in ovarian cancer. Cancer Res 2021;81:2044–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Teeuwssen M, Fodde R. Wnt signaling in ovarian cancer stemness, EMT, and therapy resistance. J Clin Med 2019;8:1658. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Morales F, Giordano A. Overview of CDK9 as a target in cancer research. Cell Cycle 2016;15:519–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Wang J, Dean DC, Hornicek FJ, Shi H, Duan Z. Cyclin-dependent kinase 9 (CDK9) is a novel prognostic marker and therapeutic target in ovarian cancer. FASEB J 2019;33:5990–6000. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Parvathareddy SK, Siraj AK, Masoodi T, Annaiyappanaidu P, Al-Badawi IA, Al-Dayel F, et al. Cyclin-dependent kinase 9 (CDK9) predicts recurrence in Middle Eastern epithelial ovarian cancer. J Ovarian Res 2021;14:69. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Anshabo AT, Milne R, Wang S, Albrecht H. CDK9: a comprehensive review of its biology, and its role as a potential target for anti-cancer agents. Front Oncol 2021;11:678559. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Villalona-Calero M, Mita M, Mita A, Federman N, Rasco D, Spigel D, et al. Abstract B112: a dose escalation and cohort expansion study of the CDK9 inhibitor KB-0742 in relapsed, refractory ovarian cancer and transcriptionally addicted relapsed or refractory solid tumors. Cancer Res 2024;84:B112. [Google Scholar]
- 37. Gasimli K, Raab M, Mandal R, Krämer A, Peña-Llopis S, Tahmasbi Rad M, et al. Synergistic sensitization of high-grade serous ovarian cancer cells lacking caspase-8 expression to chemotherapeutics using combinations of small-molecule BRD4 and CDK9 inhibitors. Cancers (Basel) 2023;16:107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Li J, Zhi X, Chen S, Shen X, Chen C, Yuan L, et al. CDK9 inhibitor CDKI-73 is synergetic lethal with PARP inhibitor olaparib in BRCA1 wide-type ovarian cancer. Am J Cancer Res 2020;10:1140–55. [PMC free article] [PubMed] [Google Scholar]
- 39. Belur Nagaraj A, Kovalenko O, Avelar R, Joseph P, Brown A, Surti A, et al. Mitotic exit dysfunction through the deregulation of APC/C characterizes cisplatin-resistant state in epithelial ovarian cancer. Clin Cancer Res 2018;24:4588–601. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Paliwal S, de Giorgio A, Neil D, Michel J-B, Lacoste AM. Preclinical validation of therapeutic targets predicted by tensor factorization on heterogeneous graphs. Sci Rep 2020;10:18250. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Böhm S, Montfort A, Pearce OMT, Topping J, Chakravarty P, Everitt GLA, et al. Neoadjuvant chemotherapy modulates the immune microenvironment in metastases of tubo-ovarian high-grade serous carcinoma. Clin Cancer Res 2016;22:3025–36. [DOI] [PubMed] [Google Scholar]
- 42. Zdrazil B, Felix E, Hunter F, Manners EJ, Blackshaw J, Corbett S, et al. The ChEMBL Database in 2023: a drug discovery platform spanning multiple bioactivity data types and time periods. Nucleic Acids Res 2024;52:D1180–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Katoh K, Misawa K, Kuma K-i, Miyata T. MAFFT: a novel method for rapid multiple sequence alignment based on fast Fourier transform. Nucleic Acids Res 2002;30:3059–66. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Potter SC, Luciani A, Eddy SR, Park Y, Lopez R, Finn RD. HMMER web server: 2018 update. Nucleic Acids Res 2018;46:W200–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Friesner RA, Murphy RB, Repasky MP, Frye LL, Greenwood JR, Halgren TA, et al. Extra precision Glide: docking and scoring incorporating a model of hydrophobic enclosure for Protein−Ligand complexes. J Med Chem 2006;49:6177–96. [DOI] [PubMed] [Google Scholar]
- 46. Avelar RA, Armstrong AJ, Carvette G, Gupta R, Puleo N, Colina JA, et al. Small-molecule–mediated stabilization of PP2A modulates the homologous recombination pathway and potentiates DNA damage-induced cell death. Mol Cancer Ther 2023;22:599–615. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47. Froger A, Hall JE. Transformation of plasmid DNA into E. coli using the heat shock method. J Vis Exp 2007:253. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48. Senkowski W, Gall-Mas L, Falco MM, Li Y, Lavikka K, Kriegbaum MC, et al. A platform for efficient establishment and drug-response profiling of high-grade serous ovarian cancer organoids. Dev Cell 2023;58:1106–21.e7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 2013;29:15–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50. Kovaka S, Zimin AV, Pertea GM, Razaghi R, Salzberg SL, Pertea M. Transcriptome assembly from long-read RNA-seq alignments with StringTie2. Genome Biol 2019;20:278. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51. Pertea M, Kim D, Pertea GM, Leek JT, Salzberg SL. Transcript-level expression analysis of RNA-seq experiments with HISAT, StringTie and Ballgown. Nat Protoc 2016;11:1650–1667. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52. Pertea M, Pertea GM, Antonescu CM, Chang TC, Mendell JT, Salzberg SL. StringTie enables improved reconstruction of a transcriptome from RNA-seq reads. Nat Biotechnol 2015;33:290–295. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53. Liao Y, Smyth GK, Shi W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics 2014;30:923–30. [DOI] [PubMed] [Google Scholar]
- 54. Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol 2014;15:550. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A 2005;102:15545–50. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 56. Mootha VK, Lindgren CM, Eriksson K-F, Subramanian A, Sihag S, Lehar J, et al. PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes. Nat Genet 2003;34:267–73. [DOI] [PubMed] [Google Scholar]
- 57. Goldman MJ, Craft B, Hastie M, Repečka K, McDade F, Kamath A, et al. Visualizing and interpreting cancer genomics data via the Xena platform. Nat Biotechnol 2020;38:675–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 58. Wiechert A, Saygin C, Thiagarajan PS, Rao VS, Hale JS, Gupta N, et al. Cisplatin induces stemness in ovarian cancer. Oncotarget 2016;7:30511–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 59. Steg AD, Bevis KS, Katre AA, Ziebarth A, Dobbin ZC, Alvarez RD, et al. Stem cell pathways contribute to clinical chemoresistance in ovarian cancer. Clin Cancer Res 2012;18:869–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 60. Richardson P, Griffin I, Tucker C, Smith D, Oechsle O, Phelan A, et al. Baricitinib as potential treatment for 2019-nCoV acute respiratory disease. Lancet 2020;395:e30–1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 61. Sharma A, Tiwari S, Deb MK, Marty JL. Severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2): a global pandemic and treatment strategies. Int J Antimicrob Agents 2020;56:106054. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 62. Pujade-Lauraine E, Selle F, Weber B, Ray-Coquard I-L, Vergote I, Sufliarsky J, et al. Volasertib versus chemotherapy in platinum-resistant or -refractory ovarian cancer: a randomized phase II groupe des investigateurs nationaux pour l’etude des cancers de l’ovaire study. J Clin Oncol 2016;34:706–13. [DOI] [PubMed] [Google Scholar]
- 63. Nakahara T, Kita A, Yamanaka K, Mori M, Amino N, Takeuchi M, et al. YM155, a novel small-molecule survivin suppressant, induces regression of established human hormone-refractory prostate tumor xenografts. Cancer Res 2007;67:8014–21. [DOI] [PubMed] [Google Scholar]
- 64. Shi J, Van de Water R, Hong K, Lamer RB, Weichert KW, Sandoval CM, et al. EC144 is a potent inhibitor of the heat shock protein 90. J Med Chem 2012;55:7786–95. [DOI] [PubMed] [Google Scholar]
- 65. Tröster A, Heinzlmeir S, Berger B-T, Gande SL, Saxena K, Sreeramulu S, et al. NVP-BHG712: effects of regioisomers on the affinity and selectivity toward the EPHrin family. ChemMedChem 2018;13:1629–33. [DOI] [PubMed] [Google Scholar]
- 66. Juteau H, Gareau Y, Labelle M, Sturino CF, Sawyer N, Tremblay N, et al. Structure–activity relationship of cinnamic acylsulfonamide analogues on the human EP3 prostanoid receptor. Bioorg Med Chem 2001;9:1977–84. [DOI] [PubMed] [Google Scholar]
- 67. Hintermann S, Chiesi M, von Krosigk U, Mathé D, Felber R, Hengerer B. Identification of a series of highly potent activators of the Nurr1 signaling pathway. Bioorg Med Chem Lett 2007;17:193–6. [DOI] [PubMed] [Google Scholar]
- 68. Munoz-Tello P, Lin H, Khan P, de Vera IMS, Kamenecka TM, Kojetin DJ. Assessment of NR4A ligands that directly bind and modulate the orphan nuclear receptor Nurr1. J Med Chem 2020;63:15639–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 69. Wang B, Li X, Zhao G, Yan H, Dong P, Watari H, et al. miR-203 inhibits ovarian tumor metastasis by targeting BIRC5 and attenuating the TGFβ pathway. J Exp Clin Cancer Res 2018;37:235. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 70. Hou L-J, Huang X-X, Xu L-N, Zhang Y-Y, Zhao N, Ou R-Y, et al. YM155 enhances docetaxel efficacy in ovarian cancer. Am J Transl Res 2018;10:696–708. [PMC free article] [PubMed] [Google Scholar]
- 71. Zhao G, Wang Q, Wu Z, Tian X, Yan H, Wang B, et al. Ovarian primary and metastatic tumors suppressed by survivin knockout or a novel survivin inhibitor. Mol Cancer Ther 2019;18:2233–45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 72. Chu S-H, Liu Y-W, Zhang L, Liu B, Li L, Shi J-Z, et al. Regulation of survival and chemoresistance by HSP90AA1 in ovarian cancer SKOV3 cells. Mol Biol Rep 2013;40:1–6. [DOI] [PubMed] [Google Scholar]
- 73. Li L, Nan F, Guo Q, Guan D, Zhou C. Resistance to bevacizumab in ovarian cancer SKOV3 xenograft due to EphB4 overexpression. J Cancer Res Ther 2019;15:1282–7. [DOI] [PubMed] [Google Scholar]
- 74. Rodriguez-Aguayo C, Bayraktar E, Ivan C, Aslan B, Mai J, He G, et al. PTGER3 induces ovary tumorigenesis and confers resistance to cisplatin therapy through up-regulation Ras-MAPK/Erk-ETS1-ELK1/CFTR1 axis. EBioMedicine 2019;40:290–304. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 75. Bodnar L, Stanczak A, Cierniak S, Smoter M, Cichowicz M, Kozlowski W, et al. Wnt/β-catenin pathway as a potential prognostic and predictive marker in patients with advanced ovarian cancer. J Ovarian Res 2014;7:16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 76. Boone JD, Arend RC, Johnston BE, Cooper SJ, Gilchrist SA, Oelschlager DK, et al. Targeting the Wnt/β-catenin pathway in primary ovarian cancer with the porcupine inhibitor WNT974. Lab Invest 2016;96:249–59. [DOI] [PubMed] [Google Scholar]
- 77. Huang L, Jin Y, Feng S, Zou Y, Xu S, Qiu S, et al. Role of Wnt/β-catenin, Wnt/c-Jun N-terminal kinase and Wnt/Ca2+ pathways in cisplatin-induced chemoresistance in ovarian cancer. Exp Ther Med 2016;12:3851–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 78. Cerami E, Gao J, Dogrusoz U, Gross BE, Sumer SO, Aksoy BA, et al. The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data. Cancer Discov 2012;2:401–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 79. Gao J, Aksoy BA, Dogrusoz U, Dresdner G, Gross B, Sumer SO, et al. Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal. Sci Signal 2013;6:pl1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 80. de Bruijn I, Kundra R, Mastrogiacomo B, Tran TN, Sikina L, Mazor T, et al. Analysis and visualization of longitudinal genomic and clinical data from the AACR project GENIE biopharma collaborative in cBioPortal. Cancer Res 2023;83:3861–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 81. McAnulty J, DiFeo A. The molecular ‘Myc-anisms’ behind Myc-driven tumorigenesis and the relevant Myc-directed therapeutics. Int J Mol Sci 2020;21:9486. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 82. Testa U, Petrucci E, Pasquini L, Castelli G, Pelosi E. Ovarian cancers: genetic abnormalities, tumor heterogeneity and progression, clonal evolution and cancer stem cells. Medicines (Basel) 2018;5:16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 83. Reyes-González JM, Armaiz-Peña GN, Mangala LS, Valiyeva F, Ivan C, Pradeep S, et al. Targeting c-MYC in platinum-resistant ovarian cancer. Mol Cancer Ther 2015;14:2260–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 84. He TC, Sparks AB, Rago C, Hermeking H, Zawel L, da Costa LT, et al. Identification of c-MYC as a target of the APC pathway. Science 1998;281:1509–12. [DOI] [PubMed] [Google Scholar]
- 85. Stambolic V, Ruel L, Woodgett JR. Lithium inhibits glycogen synthase kinase-3 activity and mimics wingless signalling in intact cells. Curr Biol 1996;6:1664–8. [DOI] [PubMed] [Google Scholar]
- 86. Barretina J, Caponigro G, Stransky N, Venkatesan K, Margolin AA, Kim S, et al. The Cancer Cell Line Encyclopedia enables predictive modelling of anticancer drug sensitivity. Nature 2012;483:603–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 87. Ammirati M, Bagley SW, Bhattacharya SK, Buckbinder L, Carlo AA, Conrad R, et al. Discovery of an in vivo tool to establish proof-of-concept for MAP4K4-based antidiabetic treatment. ACS Med Chem Lett 2015;6:1128–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 88. Jung HR, Oh Y, Na D, Min S, Kang J, Jang D, et al. CRISPR screens identify a novel combination treatment targeting BCL-XL and WNT signaling for KRAS/BRAF-mutated colorectal cancers. Oncogene 2021;40:3287–302. [DOI] [PubMed] [Google Scholar]
- 89. Boffo S, Damato A, Alfano L, Giordano A. CDK9 inhibitors in acute myeloid leukemia. J Exp Clin Cancer Res 2018;37:36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 90. Welcker M, Clurman BE. FBW7 ubiquitin ligase: a tumour suppressor at the crossroads of cell division, growth and differentiation. Nat Rev Cancer 2008;8:83–93. [DOI] [PubMed] [Google Scholar]
- 91. Mandal R, Becker S, Strebhardt K. Targeting CDK9 for anti-cancer therapeutics. Cancers (Basel) 2021;13:2181. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 92. Loh YN, Hedditch EL, Baker LA, Jary E, Ward RL, Ford CE. The Wnt signalling pathway is upregulated in an in vitro model of acquired tamoxifen resistant breast cancer. BMC Cancer 2013;13:174. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 93. Staal FJT, Famili F, Garcia Perez L, Pike-Overzet K. Aberrant Wnt signaling in leukemia. Cancers (Basel) 2016;8:78. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 94. Stewart DJ. Wnt signaling pathway in non-small cell lung cancer. J Natl Cancer Inst 2014;106:djt356. [DOI] [PubMed] [Google Scholar]
- 95. Spranger S, Bao R, Gajewski TF. Melanoma-intrinsic β-catenin signalling prevents anti-tumour immunity. Nature 2015;523:231–5. [DOI] [PubMed] [Google Scholar]
- 96. Wu T, Qin Z, Tian Y, Wang J, Xu C, Li Z, et al. Recent developments in the biology and medicinal Chemistry of CDK9 inhibitors: an update. J Med Chem 2020;63:13228–57. [DOI] [PubMed] [Google Scholar]
- 97. Mokhtari RB, Homayouni TS, Baluch N, Morgatskaya E, Kumar S, Das B, et al. Combination therapy in combating cancer. Oncotarget 2017;8:38022–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary Figure S1: Characterization of OV81-CP40 and OV231-CP30 cell lines and results of GSEA for OV81-CP40 RNAseq.
Supplementary Figure S2: Validation of NCB-0846 in selectively targeting platinum-resistant HGSC across cell models.
Supplementary Figure S3: Validation of lithium chloride and Figure 3 western blot densitometry
Supplementary Figure S4: Densitometry for Figure 4 western blots
Supplementary Figure S5: CDK9 inhibition shows platinum-resistant HGSC efficacy and specificity and mediates responses to enhanced Wnt activity.
Supplementary Figure S6: Inhibition of MYC is important in TNIK and CDK9-mediated targeting efficacy as observed through HEC1B treatment with NCB-0846
Supplementary Table S1: Compounds from the BenevolentAI library created for screening in spheroid grown OV81-CP40 and OV231-CP30 cells. Results reported as mean fold change cell viability relative to DMSO control with standard deviation; n = 4.
Supplementary Table S2: Patient information and treatment regimen for OV-UM32.1, OV-UM50.2 and OV-UM52.1 cell lines.
Supplementary Table S3: IC50s of patient-derived organoids.
Supplementary Table S4: Artificial intelligence predicted additional target kinases for NCB-0846.
Data Availability Statement
RNA sequencing data have been deposited in NCBI’s Gene Expression Omnibus (RRID:SCR_005012) and are accessible through Gene Expression Omnibus Series accession number GSE271224. The remaining data generated in this study are available upon request from the corresponding author.