High-throughput compound screening identifies navitoclax combined with irradiation as a candidate therapy for HPV-negative head and neck squamous cell carcinoma

Conventional chemotherapeutic agents are nonselective, often resulting in severe side effects and the development of resistance. Therefore, new molecular-targeted therapies are urgently needed to be integrated into existing treatment regimens. Here, we performed a high-throughput compound screen to identify a synergistic interaction between ionizing radiation and 396 anticancer compounds. The assay was run using five human papillomavirus (HPV)-negative head and neck squamous cell carcinoma (HNSCC) cell lines cultured on the human tumor-derived matrix Myogel. Our screen identified several compounds with strong synergistic and antagonistic effects, which we further investigated using multiple irradiation doses. Navitoclax, which emerged as the most promising radiosensitizer, exhibited synergy with irradiation regardless of the p53 mutation status in all 13 HNSCC cell lines. We performed a live cell apoptosis assay for two representative HNSCC cell lines to examine the effects of navitoclax and irradiation. As a single agent, navitoclax reduced proliferation and induced apoptosis in a dose-dependent manner, whereas the navitoclax–irradiation combination arrested cell cycle progression and resulted in substantially elevated apoptosis. Overall, we demonstrated that combining navitoclax with irradiation resulted in synergistic in vitro antitumor effects in HNSCC cell lines, possibly indicating the therapeutic potential for HNSCC patients.


HTS of 396 anticancer compounds reveals synergistic and antagonistic combinations with ionizing radiation on HNSCC cells.
High-throughput compound screening (HTS) is a widely used method for identifying effective drug candidates targeting cancer cells. We used a compound library of 396 FDA-approved drugs as well as experimental drug candidates and probes together with ionizing radiation to investigate potential synergistic and antagonistic combinations in five locally established HNSCC cell lines with previously characterized mutation profiles 18 .
The compound screening revealed a remarkable variation across the five HPV-negative HNSCC cell lines in their response to most compounds and compound-irradiation combinations ( Fig. 1a and Supplementary Figure S1). However, some compounds were completely inactive (n = 39) both as a single agent and combined with irradiation across all five cell lines (Supplementary Figure S1). The other group of compounds appeared highly effective as a single agent in all five cell lines, but exhibited an additive effect when combined with irradiation (n = 28). The majority of the compounds exhibited a synergistic or antagonistic interaction when combined with irradiation (n = 357), although most lacked consistency across cell lines. Nevertheless, some compounds exhibited a clear synergistic or antagonistic pattern across all cell lines. Based on the average ΔDSS values, we selected the 15 most promising synergistic (n = 12) and antagonistic (n = 3) compounds for further validation (Fig. 1b). Interestingly, 7 of 12 synergistic compounds were classified as differentiating or epigenetic modifiers (BAY 87-2243, talazoparib, tretinoin, lonafarnib, acitretin, CUDC-907 and tipifarnib). Three synergistic compounds were kinase inhibitors (triciribine, omipalisib and afatinib) and two were apoptotic modulators (navitoclax and birinapant). Three antagonistic compounds were PLK1 kinase inhibitors (BI 2536 and GSK-461364) and the metabolic modifier pemetrexed.
Top-hit validation in dose-response matrices revealed strong antagonism and synergy in several drug-irradiation combinations. To validate the most promising combinations, we performed a comprehensive drug combination screen of 15 compounds and multiple irradiation doses measured in 6 × 5 doseresponse matrices using the same HTS protocol used in the initial screen. To test whether the combinations acted synergistically or antagonistically, we compared the observed responses to the expected combination responses based on the ZIP reference model using the SynergyFinder web application (version 2) 20,21 . Based on the obtained synergy scores, we classified the compound-irradiation combinations as additive, antagonistic or synergistic. Combinations with a synergy score > 10 were considered strongly synergistic and those < − 10 were considered strongly antagonistic, while synergy scores between − 5 and 5 were classified as additive. Other synergy scores represented borderline synergistic/antagonistic drug pairs.
The dose-response analysis verified that, among seven tested epigenetic modifiers, only two (talazoparib and tretinoin) exhibited a promising synergy (Fig. 2a). Only one of three kinase inhibitors, afatinib (EGFR inhibitor), was also validated to exhibit synergy (Fig. 2b). Both apoptotic modulators (navitoclax and birinapant) also exhibited a synergy when combined with irradiation similar to the previous screen. Interestingly, the SMAC www.nature.com/scientificreports/ mimetic antagonist birinapant showed synergy only in one cell line (UT-SCC-24B), whereas navitoclax (Blc-2 and Bcl-xl inhibitor) exhibited synergy across all five cell lines ( Fig. 2a and Supplementary Table S1). In addition, we validated that PLK1 kinase inhibitors (BI 2536 and GSK-461364) and the metabolic modifier pemetrexed exhibited a strong antagonism when combined with irradiation, similar to results from the initial screen (Fig. 2a,c).
Navitoclax and irradiation exhibited synergy regardless of p53 mutation status. Navitoclax exhibited the highest synergy when combined with ionizing radiation among the tested compounds (Fig. 2a). A strong synergy (> 10) was observed in three of five cell lines (UT-SCC-42A, UT-SCC-42B and UT-SCC-24B) and moderate synergy (5)(6)(7)(8)(9)(10) in two other cell lines (UT-SCC-24A and UT-SCC-9B; see Fig. 2a, right panel). Next, we performed an extended navitoclax-irradiation dose-response screen on an additional eight HNSCC cell lines (Fig. 3a). In total, we tested 13 cell lines, consisting of six p53-mutated, six wild-type and one uncategorized cell line. The HNSCC cell line mutation profiles were characterized in a previous study 18 . We observed a strong combination synergy across ten cell lines and a moderate synergistic effect in three other cell lines www.nature.com/scientificreports/ (Fig. 3a). Therefore, we concluded that the combination of navitoclax and irradiation exhibited a strong synergy regardless of the p53 mutation status.
Navitoclax triggered apoptosis, which substantially increased after irradiation.. Next, we performed a live cell apoptosis assay on two HNSCC cell lines. Navitoclax induced apoptosis, even as a single agent, dose dependently in both cell lines tested (Fig. 4a). Additionally, navitoclax reduced cancer cell proliferation in a dose-dependent manner (Fig. 4b). As expected, the navitoclax-irradiation combination led to a substantially elevated apoptosis rate and halted cancer cell proliferation in both cell lines (Fig. 4a,b). For the UT-SCC-42A cell line, the apoptotic cell number nearly doubled 48 h after the combination treatment compared with irradiation only (from 10 to 19 apoptotic cells per image) and quadrupled compared with untreated cells (from 5 to 19 apoptotic cells per image; Fig. 4a). For UT-SCC-24B, the apoptotic cell number increased by 43.5% (from 23 to 33 cells per image) following the combination treatment compared with irradiation only, and nearly quadrupled compared with untreated cells (from 9 to 33 cells per image; Fig. 4a). Combining 10,000-nM navitoclax and irradiation completely halted cell proliferation in both cell lines (Fig. 4b). For UT-SCC-42A, 10,000-nM navitoclax resulted in apoptosis in 30.1% of cells, which was six times higher compared with untreated cells (4.8%; Fig. 4c). The navitoclax-irradiation combination resulted in the highest apoptosis index (48.1%), which was 62.5% higher than with irradiation treatment (29.6%; Fig. 4c). For UT-SCC-24B, 10,000-nM navitoclax caused apoptosis in 15.8% of cells, which was three times higher compared with untreated cells (4.9%; Fig. 4c). The combination treatment resulted in an apoptosis index (42.4%) that was more than two times higher than with irradiation treatment (19.3%; Fig. 4c). www.nature.com/scientificreports/

Discussion
High-throughput compound screening (HTS) is a useful method for investigating effective cancer targeting compounds. We performed HTS using a library of 396 FDA-approved and investigational compounds and ionizing radiation as an initial screen to identify synergistic and antagonistic combinations for HPV-negative HNSCC cells cultured on a human tumor-derived Myogel matrix. Our screen revealed wide variation among the HNSCC cell lines tested in their responses to the majority of compounds and compound-irradiation combinations. Most compounds did not interact with irradiation (i.e., exhibited an additive effect), while some exhibited a strong synergy or antagonism. Following the initial screen, the most prominent compound-irradiation combinations were further investigated using five tenfold compound concentrations and four irradiation doses in the dose-response matrix analysis. We further investigated the most synergistic compound, navitoclax, using eight additional cell lines and a live cell apoptosis assay.
Multiple studies confirmed that radiosensitivity and radioresistance can depend on the schedule pattern. Although our study features only one type of schedule, we designed it to mimic the typical HNSCC treatment schedule, whereby radiotherapy begins simultaneously or shortly after drug administration. Furthermore, we www.nature.com/scientificreports/ chose to irradiate cells 24 h after cell seeding and drug administration to ensure complete cell attachment before irradiation. Navitoclax is an orally active Blc-2 and Blc-xL inhibitor, which has exhibited in vitro activity against different tumor types as a single agent and together with chemotherapy 14 . However, only three in vitro studies exist for navitoclax in HNSCC 16,17 , only one of which included navitoclax combined with irradiation in HNSCC cell lines 22 . In that study, Ow et al. found that this combination did not significantly improve the response and yielded only a modest benefit in two of eight cell lines 22 . Experiments were performed using a clonogenic survival assay with only one irradiation dose and two navitoclax concentrations. Several clinical trials have been conducted or are ongoing on navitoclax as a single agent or in combination with other anticancer compounds to treat leukemia and solid tumors. However, the navitoclax-radiotherapy combination remains clinically unexplored. Additionally, to our knowledge, no clinical trials exist for HNSCC. Thus, we report here for the first time a strong synergy between navitoclax combined with ionizing radiation in HNSCC cell lines (Fig. 3). Our dose-response matrix analysis confirmed a strong synergy in 10 HNSCC cell lines and a moderate synergy in another three cell lines. Thus, our findings encourage the clinical investigation of navitoclax in combination with irradiation for the treatment of HNSCC as well. Interestingly, other BH3 mimetics in our compound library tested appeared inefficacious (Supplementary Figure S1). The Bcl-2 selective inhibitor venetoclax appeared ineffective in all five cell lines as a single agent as previously reported 23 and when combined with irradiation. The Bcl-2 and Mcl-1 inhibitor AT-101 exhibited a modest, although less convincing, synergy as the Bcl-2 and Bcl-xL inhibitor navitoclax. This indicates that the dual inhibition of Bcl-2 and Bcl-xL may play a crucial role in triggering apoptosis in HNSCC cells.
An inactivating mutation of p53 (the TP53 gene) is the most frequent genomic alteration in HNSCC, accounting for approximately 50% of cases 12 . A recent study reported that the combination of navitoclax with NOXA induction exhibits efficient anticancer effects in HNSCC cells regardless of the p53 status 17 . We selected six wild-type and six mutated p53 HNSCC cell lines to examine whether p53-mutated cell lines are less sensitive to Bcl-2/Bcl-xL inhibition and radiosynergy. Our screen demonstrated that navitoclax synergizes with radiotherapy regardless of the p53 mutation status (Fig. 3a). This indicates that navitoclax can bypass the survival advantage of the mutated p53 tumor suppressor gene following irradiation damage. However, as a single agent, navitoclax seemed more effective in p53 wild-type cell lines (Supplementary Figure S2).
The Bcl-2 family proteins regulate intrinsic (mitochondria-dependent) apoptosis in damaged cells by releasing a mitochondrial protein to cytosol, leading to the activation of caspase proteases 14 . Navitoclax induces intrinsic apoptosis in human tumor cells 15 . Our findings confirm that this phenomenon also occurs in HNSCC. Using a live cell apoptosis assay, we demonstrated that navitoclax triggers apoptosis in HNSCC cells (Fig. 4). In the presence of navitoclax, the apoptotic effect of irradiation increased markedly (Fig. 4). In addition, navitoclax reduced cancer cell proliferation as a single agent and arrested proliferation when combined with radiotherapy in both cell lines tested (Fig. 4).
In addition to navitoclax, our screen identified other promising radiosensitizers, including afatinib. Previous studies, which reported the radiosensitization of afatinib in HNSCC cell lines, support this finding 24, 25 . The first www.nature.com/scientificreports/ clinical trial to study afatinib in combination with radiation therapy in HNSCC is ongoing (NCT01783587). Furthermore, tretinoin exhibited a moderate synergy with irradiation in our screen, similar to a previous study in which tretinoin modulated radiosensitivity in HNSCC cell lines 26 . We also observed a radiosensitization of talazoparib, a PARP1/2 inhibitor, in two of our HNSCC cell lines, similar to findings among small cell lung cancer (SCLC) and some other solid tumors 27 . Several clinical trials combining talazoparib and radiotherapy are now recruiting lung, gynecological and breast cancer patients (NCT04170946, NCT03968406 and NCT04690855). We also identified radioresistant properties among several drugs in HNSCC cell lines. Interestingly, two PLK1 inhibitors (BI 2536 and GSK-461364) showed strong antagonism when administered 24 h before irradiation. One study reported PLK1 inhibition causing radiosensitization or radioresistance depending on the treatment schedule in osteosarcoma and colorectal cancer cell lines using a clonogenic assay 28 . To date, no in vitro or clinical studies for HNSCC combined with BI 2536 and irradiation exist. Clinical trials for BI 2536 primarily focus on leukemia and solid tumors, such as breast, pancreatic, prostate and lung cancers. A phase II clinical trial for BI 2536 was completed for a panel of solid tumors, including HNSCC (NCT00526149). GSK-461364, an experimental compound, lacks in vitro studies for HNSCC. The only existing clinical trial for GSK-461364 was completed for non-Hodgkin's lymphoma (NCT00536835). Pemetrexed, a dihydrofolate reductase inhibitor, serves as treatment for pleural mesothelioma and non-small cell lung cancer (NSCLC). The pemetrexed-irradiation combination appears to carry schedule-dependent interactions in NSCLC and HNSCC cells 29 . In this study, 24-h incubation with pemetrexed in the HNSCC cell line (CAL-27) followed by irradiation (0-8 Gy) resulted in moderate antagonism 29 . Furthermore, pemetrexed was evaluated as a novel treatment for HNSCC patients.
Although several clinical studies have been completed, the benefit of pemetrexed in HNSCC treatment remains unclear 30 . Surprisingly, based on our data, all cell lines sensitive to pemetrexed as a single agent exhibited a strong antagonism when combined with irradiation. This raises concerns regarding concomitant pemetrexed and radiotherapy in HNSCC.
Taken together, our HTS screen revealed both radiosensitive and radioresistant compounds for HNSCC cells. Pemetrexed and PLK1 inhibitors (BI 2536 and GSK-461364) exhibited a strong antagonism when combined with irradiation, whereas the combinations of irradiation with afatinib, tretinoin or talazoparib exhibited a promising synergy, in agreement with previous studies. The most promising finding-and, to our knowledge, the first such report-was that navitoclax combined with irradiation exhibited the strongest synergy, triggered apoptosis and halted the proliferation of HNSCC cells. Our findings, thus, encourage clinical trials using navitoclax combined with irradiation as treatment for HPV-negative HNSCC.

Materials and methods
Cell lines. We used 13 HPV-negative locally established HNSCC cell lines taken from primary and metastatic sites (Supplementary Table S2 Table S2), a piece of metastatic lymph node was obtained from a patient with oral cancer after they provided their informed consent. The fresh tissue was washed thoroughly in PBS containing 100-U/ml penicillin, 100µg/ml streptomycin and 2.5-mg/ml amphotericin B (all from Sigma-Aldrich) and minced with a razor blade into small (1-2 mm) pieces. After washing with PBS, the tissue was digested in 1:1 Dulbecco's Modified Eagle Medium (DMEM)/Ham´s Nutrient Mixture F 12 (Gibco) supplemented with 100-U/ml penicillin, 100-μg/ml streptomycin, 2.5-mg/ml amphotericin B and 1-mg/ml collagenase (all from Sigma-Aldrich) for about 1 h at 37 °C with stirring. After washing with PBS, single cells were obtained by filtering the samples through a 40-µm cell strainer (BD Biosciences, Franklin Lakes, NJ, USA). Cells were seeded on 12-well culture plates in MEM (Gibco) supplemented with 10% heat-inactivated FBS (Gibco), a 1% nonessential amino acid solution (Gibco), 2-mM glutamine, 100-U/ml penicillin, 100-µg/ml streptomycin and 250-ng/ml amphotericin B (all from Sigma-Aldrich). Fibroblasts were removed through brief exposure to a trypsin-EDTA solution (Sigma-Aldrich) as previously described 32 .
High-throughput compound screen (HTS). We used drug sensitivity and resistance testing (DSRT) adapted from a platform for leukemia cells 33 . We performed DSRT on HNSCC cell lines cultured in Myogelcoated wells on 384-well plates. Myogel was used to provide the TME for cancer cells, which improves the predictability of drug testing 7 . The use of human leiomyoma tissue was approved by the Ethics Committee of both Oulu and Tampere University Hospitals (statement number 2/2017), and all research was performed in accordance with relevant regulations. All prospective liquid handling was performed using an automated reagent dispenser (MultiFlo FX, BioTek, Winooski, Vermont, USA) and a HighRes Biosystems automation platform. Before screening, we performed irradiation dose optimization to determine the suitable dose (EC20 value) for each cell line. Next, we screened using the compound library applying a predefined irradiation dose (EC20 www.nature.com/scientificreports/ with the most synergistic and antagonistic effects were identified and further investigated using multiple doses of irradiation (0, 1, 2, 4 and 8 Gy).

Compound library.
We used a compound library of 396 FDA-approved drugs as well as investigational compounds and probes (Supplementary Figure S3). Most compounds were dissolved in dimethyl sulfoxide (DMSO) except for 19 compounds (e.g., some metabolic modifiers and platinum-based compounds), which were dissolved in water due to poor DMSO solubility or stability issues. The library covers kinase inhibitors and other signal transduction modulators. The majority of the compounds were investigational (46%) and FDA approved (46%; Supplementary Figure S3). A minority of the compounds (8%) were probes. A large proportion of the compounds consisted of kinase inhibitors (n = 208), differentiating/epigenetic modifiers (n = 73) or conventional chemotherapy drugs (n = 50). The library also contained small groups, such as hormone therapy drugs (n = 19), apoptotic modulators (n = 10), immunomodulatory compounds (n = 9) and rapalogs (n = 4). The library consisted of six compound plates (384-well plate, Corning, #3764) containing compounds at five concentrations. The compound plates were stored in pressurized inert nitrogen gas-filled storage pods (Roylan Developments Ltd., Surrey, UK) until used for screening. Each compound was tested over a 10 000-fold concentration range. DMSO at 0.1% served as the negative control and 100-μM Benzethonium chloride (BzCl) served as the positive control in the assay plates. Drug sensitivity testing data analysis. To quantitatively profile the compound effects, we calculated the drug sensitivity score (DSS) using the Breeze software (available at https:// breeze. fimm. fi) 34 . DSS was described in several previous studies 19,33 , combining several parameters (IC50, the curve slope and the minimum and maximum responses) into a single metric. The compound effect was normalized against positive (100-μM Benzethonium Chloride, BzCl) and negative (0.1% DMSO) controls to calculate the dose-response curves for each compound in each cell line and condition (with and without irradiation) separately. Each cell line was screened using two parallel compound sets. One set served as a control (single agent) and the other set was irradiated with an optimized EC20 irradiation dose (compound-irradiation combination). Otherwise, plates were handled identically. To determine the synergistic and antagonistic effects of the compounds and irradiation, we calculated the delta DSS (ΔDSS) for each compound and cell line by subtracting the DSS for the compound-irradiation combination from the single agent DSS (Supplementary Figure S1). Depending on the positivity or negativity of the ΔDSS, the compounds were classified as synergistic or antagonistic, respectively. Based on the ΔDSS values, the most relevant synergistic and antagonistic compounds were selected for the validation experiment (Fig. 1).

Dose-response matrix analysis and synergy scoring.
For the synergy validation experiments, we used 384-well screening plates with 15 compounds in five concentrations in triplicate. Each cell line was seeded onto five plates, which were irradiated with different doses (0, 1, 2, 4 or 8 Gy). Otherwise, we used the same DSRT workflow as described above. To test whether the compound-irradiation combinations acted synergistically or antagonistically, we compared the observed responses to the expected combination responses. We then calculated the responses, and generated the dose-response matrices using the ZIP reference model with the SynergyFinder web application (version 2; synergyfinder.fimm.fi) 20,21 . We further investigated the navitoclax-irradiation combination using eight additional UT-SCC cell lines. Based on the dose-response matrices and ZIP synergy scores, the compound-irradiation combinations were classified as noninteractive, antagonistic or synergistic. Combinations with a score > 10 were considered exhibiting a strong synergy and < − 10 a strong antagonism. Scores between 5 and 10 were considered moderately synergistic and between − 5 to − 10 as moderately antagonistic. Scores between − 5 and 5 were classified as noninteractive combinations.
Live cell apoptosis assay. Two  www.nature.com/scientificreports/ day, we prepared the cell suspension for the apoptosis assay using trypsin/EDTA to detach cells from the flask. The suspension of two cell lines (UT-SCC-42A and UT-SCC-24B) was labeled according to the manufacturer's instructions using CellTrace Far Red (Invitrogen, Carlsbad, CA, USA). One million cells were mixed with 1-ml PBS containing 1-µl CellTrace Far Red. Cells were placed in an incubator for 20 min at + 37 °C. Following incubation, a 5-ml culture medium was added, followed by 5-min incubation. Labeled cells were centrifuged for 5 min at 173×g at RT and suspended in a fresh culture medium. Cell suspension was diluted at a density of 100 000 cells/ml and divided into four groups: control with 0.1% DMSO and navitoclax concentrations (100, 1000 and 10,000 nM, respectively) in the presence of the IncuCyte Caspase-3/7 Apoptosis Assay Reagent (Sartorius; 1:1000). Cell suspension was pipetted onto the wells (1000 cells per well, 100 µl). Plates were placed in the Incucyte S3 live cell analysis system (Sartorius) and imaged every 2 h at objective 20× (9 images per well). Six replicates were used for each condition. After 24 h, plates were transferred to an irradiator room and one plate was irradiated with 8 Gy. After irradiation, the plates were placed back in Incucyte and imaged for 48 h. Cancer cell proliferation (red object count) and apoptotic cells (green object count) were quantified using the Incucyte analysis software. The apoptotic index (percentage of apoptotic cells) was determined by selecting apoptotic cells (green and red objects) divided by the total cell number (red object) multiplied by 100.