FAK Drives Resistance to Therapy in HPV-Negative Head and Neck Cancer in a p53-Dependent Manner

Globally, there are more than 800,000 new cases of head and neck squamous cell carcinoma (HNSCC) per year (1), and curative treatment is associated with profound and life-long toxicity. Currently, the most clinically useful biomarker of outcome in HNSCC is human papillomavirus (HPV), which is associated with better responses to therapy and survival (2). However, patients with HNSCC unrelated to HPV (HPV-negative) continue to have poor clinical outcomes with 3-year progression-free survival between 40% and 50% (2). Two of the most critical forms of therapy for this disease are radiation and cisplatin; however, there are no current means by which to personalize therapy using these agents.

One potential avenue for improving response to radiation and chemotherapy is via targeting FAK, a protein frequently overexpressed in HPV-negative HNSCC (3). FAK is a nonreceptor tyrosine kinase involved in cell–cell adhesion (4–6), encoded by gene PTK2. At focal adhesions, FAK dimerization results in autophosphorylation at Y397, making FAK catalytically active (5) and leading to the phosphorylation of downstream proteins. Our group has previously identified FAK as a driver of radioresistance in HNSCC (3), and FAK has also been implicated in radioresistance, chemoresistance, and immunotherapy resistance in other disease sites (7–10). FAK can be targeted pharmacologically by either FAK and/or dual FAK/PYK2 inhibitors targeting the Y397 site (6). Defactinib is a dual FAK/PYK2 inhibitor currently in clinical trials (6). However, despite this potential importance, the role of FAK inhibition as a therapeutic strategy in HNSCC is unknown. Furthermore, FAK inhibition has not yet been linked to specific genomic drivers of response.

Any potential therapeutic strategy in HPV-negative HNSCC must take into account the tumor suppressor TP53, which is the most commonly mutated gene in HPV-negative HNSCC, as well as in many other solid tumors (11, 12). This gene has been associated with negative outcomes in multiple clinical studies of HPV-negative HNSCC, particularly when its predicted function is considered (13–15). FAK and p53 interact at multiple levels. In contrast to its well-known canonical function at focal adhesions, FAK interacts with and regulates p53 in the nucleus where FAK acts as a scaffolding protein and leads to ubiquitation and degradation of p53 (16, 17). Conversely, p53 directly represses FAK transcription (18). Thus, the relevance of FAK in HNSCC may be mediated by the presence or absence of functional p53.

To both examine novel therapeutic sensitizers in HPV-negative HNSCC, as well as assess the impact of TP53 on therapeutic response, we initially performed an in vivo shRNA screen in combination with radiation or platinum-based chemotherapy to identify clinically druggable sensitizing targets in mutant TP53, HPV-negative HNSCC. We then hypothesized that wild-type and mutant p53 would have differential effects on FAK activity, DNA-damage repair, and responsiveness to FAK-mediated therapeutic sensitization. Finally, we examined the relationship between FAK, TP53 mutation, and clinical outcome.

The in vivo shRNA screen method was described previously (19–21). For this analysis, three HPV-negative HNSCC cell lines (UM-SCC-22a, HN31, and Cal27) were used. Briefly, lentiviral particles of previously generated druggable proteins and proteins involved in DNA-damage repair shRNA libraries were infected in vitro using spinfection into the HPV-negative HNSCC cell lines at a low MOI. Cells were selected with puromycin, and 4 million cells were subcutaneously injected into the flank of nude mice, and reference cells from the day of injection were collected and frozen. Once the tumor had reached approximately 100 mm³, the xenografts were treated with 2 Gy/day of radiation or varying doses of carboplatin, with a goal of achieving ∼20% tumor volume reduction. The tumors were subsequently allowed to grow to ∼500 mm³, harvested, and DNA was isolated. Tumor and reference cells were then amplified and sequenced on Illumina sequencers (21). Analysis was performed as detailed in Carugo and colleagues (21) and Kumar and colleagues (19). Briefly, shRNA hairpin counts were normalized to counts per million, and log2 fold-change for each shRNA hairpin was calculated compared with the reference pellet shRNA hairpin level. A modified version of the siRNA activity (RSA) algorithm (22) was used to generate gene-level summary measure per cell line with modifications to ensure that at least two hairpins were used to calculate P values and hairpins that ranked above luciferase controls were excluded. Data were displayed graphically using GraphPad Prism (v8.0).

HN30, HN31, UM-SCC-1 (RRID:CVCL_7707), FaDu, 183, HN5, and UM-SCC-22a cell lines were generously supplied by Dr. Jeffrey Myers via the University of Texas MD Anderson Cancer Center head and neck cell line repository. HN30 and HN31 cisplatin-resistant cell lines were generously supplied by Dr. Vlad Sandulache from Baylor College of Medicine. The HN31 cell line with control and FAK shRNA generation were previously described (3). Detroit562 cells were obtained from ATCC. Before experiments, cell lines were genotyped and tested for mycoplasma. All cell lines except Detroit562 and FaDu were cultured in DMEM/F-12 50/50 medium supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin and incubated at 37°C and 5% CO₂ atmosphere. Detroit562 and FaDu cells were cultured in MEM supplemented with 10% heat-inactivated FBS, 1% penicillin/streptomycin, 1% nonessential amino acid mix, and 1% sodium pyruvate. HN30 and HN31 cisplatin-resistant cell lines were cultured in 4 and 10 μmol/L of cisplatin, respectively. Defactinib (cat. #S7654), cisplatin (cat. #S1166), and carboplatin (cat. #S1215) were purchased from Selleck Chemicals.

Assays were performed as previously described (19). Briefly, single cells were plated in a 6-well plate overnight. The next day, cells were pretreated with DMSO or the indicated dose of defactinib for 24 hours for experiments. Cells were then irradiated at the indicated dose or treated with carboplatin for 24 hours. Cells were re-fed with media. The cells formed colonies over the next 10–14 days. Colonies were fixed with 0.25% crystal violet/methanol solution, and colonies containing greater than 50 cells each were counted (19). Clonogenic survival curves were generated using GraphPad Prism (v8.0).

The MTT assay was performed similarly to a previous study with the following modifications (23). Depending on the cell line, 1,000–4,000 cells per well in 100 μL of media were plated into 96-well plates. After 24-hour incubation, cells were treated with defactinib, carboplatin, cisplatin, and/or appropriate control. After 48 hours, cells were aspirated and re-fed with media. Once control cells were confluent (∼48–72 hours after refeeding), 50 μL of 5 mg/mL Thiazolyl Blue Tetrazolium Bromide solution (Sigma, MTT) was added to cells and incubated for 3 hours. Wells were aspirated, and 150 μL DMSO was added to each well. Plates were agitated for 30 minutes on the shaker, then read using an Epoch Microplate Spectrophotometer (BioTek) at 590 nm and Gen5 v.3.05 software (Gen5; RRID:SCR_017317). Plots were generated using GraphPad Prism (v8, RRID:SCR_002798).

HN30 and HN31 cell lines with pDRGFP for HR assay (Addgene, Plasmid #26475, RRID:Addgene_26475) or EJ5GFP for NHEJ assay (Addgene, Plasmid 44026, RRID:Addgene_44026) were generated as previously described (19). The HN30 and HN31 cell lines were transfected with 2 μg mCherry (Addgene, Plasmid 41583, RRID:Addgene_41583) as a negative control, or with both 2 μg mCherry and 6 μg pCBASceI (Addgene, Plasmid 26477, RRID:Addgene_26477). The cells were incubated overnight in 1 mL of media containing DMSO for the controls, 50 μmol/L defactinib, or 50 μmol/L ATM inhibitor (KU-55933, Selleckchem) for HN30 and 10 mmol/L ATM inhibitor (KU-55933, Selleckchem) for HN31 as the positive controls. The media were replaced with 2 mL of media the following day with the same concentrations of drug added for 24 hours. The cells were then trypsinized, centrifuged into a pellet at 1,200 rpm for 5 minutes, and washed with PBS before being resuspended in 1 mL of FACS buffer containing PBS, 0.1% BSA (Sigma) and 0.1% NaA (Sigma). Flow cytometry was run using the BD Accuri C6 Plus flow cytometer (BD Biosciences) to detect GFP and RFP, with cells positive for mCherry and GFP being gated as positive (19). The data were plotted with GraphPad Prism, RRID:SCR_002798.

Intracellular ROS levels were measured according to a previously published protocol using CM-H2DCFDA (Thermo Fisher Scientific, cat: D399; refs. 14, 24). Briefly, cells were treated with DMSO control or defactinib at indicated doses for 24 hours. Cells were irradiated at the indicated dose. Cell culture media were removed, and cells were stained in a serum-free medium containing 5 μmol/L H₂DCFDA for 30 minutes at 37°C. Positive control (serum-free medium with 5 μmol/L H₂DCFDA and 100 μmol/L H₂O₂) and negative control (no H₂DCFDA) were used for all experiments. Cells were trypsinized and harvested by centrifugation at 400 × g at 5 minutes. Cells were washed once with PBS and centrifuged as above. Cells were suspended in 0.5 mL of PBS and assayed immediately by flow cytometry with fluorescence excitation and emission (492–495 / 517–527 nm).

TP53 constructs (wild-type, G245D, R175H) in pBabe retroviral expression vector (pBABEpuro; Addgene, RRID:Addgene_51124) were a generous gift from Drs. Jeffery Myers and Abdullah A. Osman from the University of Texas MD Anderson Cancer Center (15). Retroviral packaging was performed in GP2+293T cells by transfecting 4.25 μg of TP53 constructs and 1.75 μg of CMV-VSG-G with 18 μL of polyethylenimine hydrochloride in DMEM/F12 media with 1% FBS for 6 hours, and then cells were re-fed. TP53 construct virus-containing media were harvested from GP2-293T cells at 48 hours and passed through a 0.45-μm syringe filter. Three mL of TP53 construct virus-containing media with 5 μL of polybrene was placed on UMSCC1 cells in a 6-well plate for 6 hours, and the virus-containing media were replaced with growth media. This entire process was repeated after 48 hours. TP53 construct UMSCC1 cells were then selected with 2 μg/mL puromycin, and TP53 expression was confirmed via western blot.

The FAK shRNA in vivo study was performed according to all relevant ethical regulations after IACUC approval from the University of Texas MD Anderson Cancer Center similar to previous xenograft models (19). Male athymic nude mice (6 weeks old, Cold Spring Harbor, RRID:RGD_5508395) were injected with HN31 tumor cells with control or FAK shRNA (2 ×10⁶ in 0.1 mL of PBS) subcutaneously in the right dorsal flank of each mouse. Tumor diameters were measured with calipers, and tumor volume was calculated as A × B2 × 0.5 (A = largest diameter, B = shortest diameter). When the tumor volumes reached ∼100–150 mm³, mice were balanced according to tumor size into sham or ionizing radiation (2 Gy daily × 4 days) and subsequently tracked for three weeks for tumor-growth delay experiments. Tumor volume was measured every 2–3 days (calculated as above) and averaged between groups. Group comparisons were performed using two-way ANOVA with a post hoc false discovery rate (FDR). Statistical analysis was performed in GraphPad Prism v8.

Immunoblot analyses were performed as previously described (19). Cell lines were washed with cold PBS and scraped into whole-cell lysis buffer (20 mmol/L HEPES, pH 7.9, 0.4 M NaCl, 0.1 mmol/L EDTA, pH 8.0, 0.1 mmol/L EGTA, pH 7, 1% Igepal, 1× Halt protease-inhibitor cocktail (Thermo Scientific), and 1× Halt phosphatase-inhibitor cocktail (Thermo Scientific). Protein lysates were vortexed and sonicated at 100 amplitude for 2 minutes using QSonica Q700 sonicator. Subsequently, protein lysates were centrifuged for 15 minutes at 4°C, and supernatants were transferred to a microcentrifuge tube. Protein assays were performed using a DC Protein Assay kit (Bio-Rad) and were used to load equivalent total protein on 4%–15% gradient (SDS)-polyacrylamide gel (Bio-Rad). Protein was transferred for 10 minutes onto a polyvinylidene-difluoride membrane using a Transblot Turbo device (Bio-Rad). Membranes were blocked with 5% nonfat powdered milk in TBS (TBS, 0.1 M, pH = 7.4) and incubated with primary antibody at 4°C overnight. Primary antibodies used include phospho-FAK (Tyr397,1:250, #3283, RRID:AB_2173659), total FAK (1:1000, #3285, RRID:AB_2269034), phospho-p53 (Ser20, 1:500, cat. # 9287, RRID:AB_2210834), and p53 (1C12, 1:1,000, #2524, RRID:AB_331743) from Cell Signaling Technology; β-actin (C4, 1:10,000, MAB1501, RRID:AB_2223041) from MilliporeSigma; and p53 (DO-1, 1:1,000, cat. # sc-126, RRID:AB_628082) from Santa Cruz Biotechnology. Membranes were washed three times with TBS + 0.1% Tween and incubated with goat anti-rabbit (1:2,000, #NA934V, RRID:AB_2722659) or anti-mouse (1:2,000, #NA931V, RRID:AB_772210) secondary antibodies conjugated to horseradish peroxidase (GE Healthcare) at room temperature for 1 hour. Membranes were washed as above, and a signal was generated using an ECL2 western blotting substrate (Pierce Biotechnology) on Hyblot CL autoradiographic film (Thomas Scientific). Western blot quantitation was done by densitometry measurements using ImageJ software (National Institute of Health). Ratios were calculated relative to actin.

Cells were grown in chamber slides to 50% confluence, pretreated for 2 hours with 0.5 μmol/L defactinib, and then treated with cisplatin at the indicated doses for 24 hours. Cell death was measured by TUNEL staining with the ApopTag Fluorescein In Situ Apoptosis Detection Kit (EMD Millipore) according to the manufacturer's instructions. Briefly, cells were fixed with 1% PFA and permeabilized in ice-cold 2:1 ethanol:acetic acid prior to TUNEL staining. Stained slides were counterstained with DAPI and mounted with VectaShield (Vector Labs).

An institutional cohort of HPV-negative HNSCC treated with surgery and postoperative radiation was examined, with tumors evaluated via Illumina gene-expression array and targeted sequencing as described previously (3, 25). A total of 94 patients had tumors with detailed clinical outcomes, gene expression, and TP53 status available. Patients were stratified by PTK2/FAK linear copy number as performed previously (3) and separated into wild-type and mutant TP53 groups for analysis. PTK2 mRNA expression for wild-type and mutant TP53 tumors was assessed, and patients were stratified into either PTK2 high (upper tertile) or low (remaining tertile) groups for outcome analysis. Disease-free survival (DFS) curves were generated by using the method of Kaplan–Meier, with log-rank statistics used to determine significance. SPSS (v27, RRID:SCR_002865) was used for clinical analysis, with figures generated using GraphPad Prism (v8). A similar analysis was applied to an additional cohort of 66 HPV-negative HNSCC treated with surgery and postoperative radiation from the TCGA.

In vivo shRNA screen statistical analysis was performed as described in the section above. Clonogenic survival curves were analyzed using two-way ANOVA with post hoc analysis for multiple comparisons. TUNEL, ROS, HR, and NHEJ assays were analyzed using a paired t test. Tumor-growth curves for the HN31 xenograft model were analyzed using a matched mixed-effects model, and the average area under the curve for the HN31 shRNA in vivo experiment was analyzed using ordinary one-way ANOVA. For clinical data, disease-specific survival was generated by using the method of Kaplan–Meier with log-rank statistics, and P < 0.05 was considered significant unless otherwise specified. Statistical analysis was generated using GraphPad Prism (v8.0).

Murine experiments were performed according to all relevant ethical regulations after IACUC approval from the University of Texas MD Anderson Cancer Center. All clinical tumor studies performed adhere to the U.S. Common Rule.

Raw human sequence data were generated via the UT MD Anderson Head and Neck Genomics Service. Human sequence data used in this study are not publicly available due to patient privacy requirements but are available upon request from the corresponding author. Other data generated in this study are available within the article and its supplementary data files.

We performed in vivo shRNA-mediated screens in preclinical HNSCC models to discover potential novel agents that were radio- and chemosensitizers. The study of all models for radiosensitization has been published previously (20, 19). In this current study, we focus on mutant TP53, HPV-negative HNSCC models from this published screen, as well as the same models from a separate unpublished experiment in which tumors were treated with carboplatin. Specifically, UM-SCC-22a, HN31, and Cal27 were infected with two barcode-labeled shRNA libraries, one consisting of druggable targets and the second focused on DNA-damage repair (DDR) proteins (full list of targets in ref. 19 Supplementary Table S2). These cells were then implanted in the murine flank. Mice were treated with appropriate sham, conventional fractionated radiotherapy (XRT), or carboplatin, with the goal of ∼20% tumor volume reduction (20, 19). Subsequently, tumors were harvested, and tumor shRNA bar codes were sequenced. We performed redundant siRNA analysis (RSA), as described previously (19, 20, 26). We then evaluated RSA P values in radiation or carboplatin-treated tumors compared with reference cells (generated prior to tumor implant); genes with RSA log P ≤ −1.3 were considered significant. Of the 502 unique screened targets, 92 genes in both the radiation and carboplatin experiments were significant, with 72 out of the 92 genes being associated with response to both carboplatin and radiation (Fig. 1A–C). We next performed KEGG pathway analysis of targets found to be significant in both radiation and carboplatin-treated tumors, which demonstrated enrichment for genes involved in focal adhesion, proteoglycans in cancer, ErbB signaling pathways, non–small cell lung cancer, cell cycle, and pancreatic cancer (Fig. 1D; ref. 27). Finally, we compared targets in untreated tumors versus those treated with either radiation (Fig. 1E) or carboplatin (Fig. 1F). In this analysis, FAK was preferentially associated with sensitization to both radiation (mean control −1.835 versus mean RT −2.593) and carboplatin (mean carboplatin −2.798).

To validate our in vivo shRNA screen results for radiosensitization, we performed a murine xenograft experiment using a FAK shRNA knockdown tumor cell line (Supplementary Fig. S2A; Supplementary Table S1). The combination of FAK depletion and radiation resulted in a significant decrease in tumor volume compared with FAK inhibition (P < 0.01) or radiotherapy alone (P = 0.01; Fig. 2A). Average area under the curve for HN31 shRNA in vivo experiments demonstrated decreasing tumor volume (Control, Control + XRT, shFAK, and shFAK + XRT; *, P = 0.0135; Fig. 2B).

Because our in vivo shRNA screen was designed to leverage currently available clinically targetable proteins, we next focused on defactinib, a pharmacologic ATP-competitive FAK inhibitor. In HPV-negative HNSCC tumor cell lines (HN31 and HN30), defactinib resulted in decreased phospho-FAK, which is a surrogate for FAK activity (Fig. 2C). Because the vast majority of HPV-negative HNSCC contain mutant TP53, we investigated if FAK inhibition was dependent on TP53 mutational status. Our model system used isogenic HNSCC cell lines from a primary tumor (wild-type TP53 HN30) and lymph node metastasis (mutant TP53 HN31; ref. 28). In the clonogenic survival assay, the mutant TP53 cell line (HN31) was radiosensitized to radiation by defactinib, whereas the wild-type TP53 HN30 cell line survival fraction remained unchanged with increasing doses of defactinib (Fig. 2D and E). Similarly, when we expressed wild-type and mutant (G245C and R282W) TP53 in a TP53-null cell line (UMSCC1; Supplementary Fig. S2B), the combination of FAK inhibition and radiation did not have an additive effect in wild-type TP53 cells but resulted in radiosensitization in the null and mutant TP53 cell lines compared with both radiation and defactinib alone at comparable levels (Fig. 2F). These results are consistent with previous clonogenic survival assays in HPV-negative mutant TP53 cell lines using the FAK inhibitor PF562271, which demonstrated radiosensitization (Fig. 3; ref. 3).

To validate our in vivo shRNA screen results for chemosensitization, we inhibited FAK via shRNA or defactinib in a clonogenic survival assay. FAK inhibition resulted in chemosensitization in the presence of carboplatin (Fig. 3A and B). Immunoblot analysis in HNSCC cell lines demonstrated that FAK inhibition resulted in decreased phospho-FAK in the presence of carboplatin (Fig. 3C). To confirm these findings in other cell lines, we performed TUNEL assays on HN31, HN30, FADU, Detroit562, UM-SCC-22a, 183, and HN5 cell lines with cisplatin and defactinib. Interestingly, HN30 was not sensitized to cisplatin by defactinib, but all the TP53 mutant lines did show increased TUNEL positivity with cisplatin plus defactinib (Fig. 3D and E; Supplementary Fig. S3). To examine the role of platinum-based chemoresistance, TP53 mutational status, and FAK inhibition, we initially performed MTT assays combining carboplatin and defactinib in both wild-type TP53 HN30 and mutant TP53 HN31 cells (Supplementary Fig. S1). Generally, modest sensitization to carboplatin by defactinib was observed in HN31 cells, with no sensitization observed in HN30 cells. We additionally evaluated HN31 and HN30 cell lines that were rendered cisplatin-resistant (HN31-CR and HN30-CR) due to serial passage in cisplatin-containing media (Fig. 3F and G; ref. 29). The HN30-CR cell line still contains wild-type TP53. Phospho-FAK expression was higher in the HN31-CR cells compared with the parental cell line; however, this was not observed in the HN30-CR line (Fig. 3H). Moreover, treatment with defactinib significantly sensitized the HN31-CR cells to cisplatin, although FAK inhibition had no effect on HN30-CR cisplatin response in MTT assays (Fig. 3G).

We next focused on the relationship between TP53 status and FAK inhibition as it relates to oxidative stress and DNA damage, as these processes generally underlie the cellular response to both platinum-based chemotherapy and radiation. Using UMSCC1 cells expressing wild-type or mutant (G245D) TP53, cell lines were treated with defactinib, radiation, or the combination. The mutant TP53 cell line demonstrated increased basal phospho-FAK levels compared with wild-type TP53 (Fig. 4A).

When we examined the combination of radiation and defactinib, we found significantly increased ROS (as measured by DCFDA assay) in mutant TP53 HN31 cells (Fig. 4B), with similar results seen in a separate mutant TP53 cell line (FADU, not shown). Conversely, we found a repression of ROS in the wild-type TP53 HN30 cells using the same combination treatment (Fig. 4B).

We next evaluated the effect of FAK inhibition on homologous recombination (HR) and nonhomologous end joining (NHEJ) DDR via an I-Scel–based reporter assay, which results in DNA break that can be repaired by HR or NHEJ, respectively (19, 30). The negative control represents all plasmids except the SceI restriction enzyme. Comparisons are made against the DMSO control, with an ATM inhibitor acting as a positive control. Pharmacologic FAK inhibition in mutant TP53 HN31 cells resulted in significant decreases in both NHEJ (P = 0.02) and HR (P = 0.04), but FAK inhibition in wild-type TP53 HN30 cells resulted in no significant effects on NHEJ (P = 0.13) and HR (P = 0.42; Fig. 4C and D).

To investigate the clinical relevance of FAK in HPV-negative HNSCC, we evaluated pretreatment tumor specimens from patients treated with postoperative radiotherapy (3, 25). Patients were stratified byTP53 mutational status (wild-type vs. mutant) and either PTK2/FAK copy number (amplification, gain, or no change/deletion) or PTK2/FAK mRNA expression (high vs. low; Fig. 5). FAK amplification (P = 0.008) or gain (P = 0.041) was significantly associated with worse DFS in the mutated TP53 patient cohort, but not in the wild-type TP53 cohort (amplification P = 0.116, gain P = 0.441). Similarly, stratification of patients by high (upper tertile) or low (lower four tertiles) FAK expression and TP53 mutational status demonstrated worse DFS in the patients with high FAK expression and mutated TP53 compared with other groups (mutant TP53/PTK2 low P = 0.008, wt TP53/PTK2 high P = 0.028, wt TP53/PTK2 low P = 0.006 vs. mutant TP53/PTK2 high; Fig. 5; Table 1). We evaluated a second cohort of 66 patients treated with surgery and postoperative radiation with known outcomes from the TCGA, and it demonstrated a similar trend of high FAK expression being associated with decreased DFS (Supplementary Fig. S4; Supplementary Table S2).

Currently, HPV is the most important biomarker in HNSCC, and its presence is associated with improved outcomes compared with its HPV-negative counterpart (2). Thus, biomarkers in HPV-negative HNSCC are desperately needed to improve prognostication and identify patients appropriate for treatment intensification and additional systemic therapies. TP53 mutational status is a natural candidate for this role because it is the most commonly mutated gene in HPV-negative HNSCC and has been previously associated with worse clinical outcomes (13, 14, 31). Unfortunately, the use of TP53 mutation as a biomarker in HNSCC is complicated by the fact that TP53 mutations have differential effects ranging from no effect to a loss or even gain of function (32). This phenomenon has led to the use of empirical methods of classifying TP53 to define clinical outcomes, without a satisfactorily validated biomarker (13, 14, 31). However, building on our previous work linking FAK to radioresistance in HNSCC, we found that by combining FAK with a simple binary classification (wild-type vs. mutant) of TP53, we improved the ability to identify patients resistant to current therapy for HNSCC.

Our observation linking FAK and TP53 status is particularly impactful in that we identified FAK as a highly significant target for sensitization to both radiation and platinum-based chemotherapy in mutant TP53 HPV-negative HNSCC using an in vivo shRNA screen. This screen utilized a medium throughput evaluation of targetable genes in an in vivo setting that allows for prioritization of targets based on their relative effects on therapeutic response, with FAK—and its associated signaling cascade—generally ranking high for both radio- and chemosensitization in most tumor models tested. By design, our screen included genes that are targetable with currently available agents, and our research focused on the most common dose fractionation in head and neck cancer of 2 Gy per fraction. It remains unproven the effect of FAK inhibitors in the setting of other dose fractionation schemes. Although FAK inhibition is a suboptimal treatment in the monotherapy setting, it appears well-tolerated as a single agent in early-phase clinical trials, and combinatorial trials with other antineoplastic agents are ongoing (6, 33–36). Thus, our data link tumors most resistant to therapy with a clinically viable strategy, namely, FAK inhibition, that can sensitize to both radiation and chemotherapy.

On its face, the idea that FAK inhibition might sensitize broadly to therapies that work through DNA damage may be nonintuitive. FAK is a crucial signaling step between β1-integrin and growth factor signaling and the intracellular signaling cascade that leads to reorganization of the actin skeleton, lamellipodia formation, and proliferation (5, 37). Thus, FAK has been approached as a therapeutic target largely on the basis of reducing cellular motility and metastasis. However, previous work has implicated β1-integrin signaling in resistance to radiation in HNSCC, in a manner potentially dependent upon cell–cell interaction (38–40). Our own work, and that of others, also directly implicates increased intracellular ROS and repression of DDR following FAK inhibition with therapeutic sensitization (3, 9, 39). Additionally, given that FAK inhibition can downregulate WT p53 (Fig. 2C; HN30 cells), but not mutant p53 (Fig. 2C; HN31 cells), but only suppresses DNA repair in the mutant HN31 (Fig. 4C and D), it suggests that FAK inhibition may also impinge on some other component of the DNA-repair machinery. Cells with mutant TP53 may be inherently more sensitive to this impingement due to their altered DNA-damage response. The exact mechanism of this phenomenon is likely multifactorial but remains to be completely elucidated.

Separately we found that FAK-mediated therapeutic sensitization was dependent upon the presence of mutated TP53. Wild-type p53 is known to repress the transcription of PTK2 at the level of the promoter in some models, a phenomenon not observed when mutant TP53 was expressed (18). Additionally, FAK can repress the function of p53 via repression of p53 transcription. Therefore, a feedback mechanism exists between these signaling pathways at the transcriptional level (41). However, similar to Boudreau and colleagues, we did not observe a relationship between total FAK levels and TP53 status but did see repression of FAK phosphorylation in cells expressing wild-type p53 versus mutant p53 isoforms (42). Moreover, in the mutant TP53 cells rendered resistant to cisplatin, we observed a dramatic increase in FAK phosphorylation, but not total FAK level, compared with the parental line, which was not observed in the wild-type p53 setting. These lines of evidence suggest that the critical effects of FAK modulation on therapeutic response may lie at the posttranscriptional level. Indeed, it has been shown that wild-type p53 can bind FAK directly, which can potentially repress the function of both proteins (17, 43, 44). We have additionally found that at least some forms of mutant p53 can bind to FAK at levels similar to wild-type, however, without a commensurate decrease in FAK activity. It is likely that in tumors with mutant TP53, high levels of FAK activity are attributable to the absence of functional p53 and make tumors more sensitive to the effects of FAK inhibition.

Given the apparent selectivity of FAK-induced radiosensitization for TP53-mutant cells, it seems unlikely that the combination of FAK and radiation would be particularly toxic to normal tissue. In a phase II clinical trial of patients with non–small cell lung cancer treated with defactinib as monotherapy, the most common side effects were fatigue (35%), nausea (22%), diarrhea (20%), vomiting (18%), hyperbilirubinemia (16%), decreased appetite (11%), and peripheral edema (11%; ref. 36). In a murine model of pancreatic cancer, hypofractionated RT plus FAK inhibition resulted in tumor regression, T-cell priming, and long-term survival in the mice, despite the proximity to the stomach, duodenum, and small intestines which traditionally limit radiation dose (45). These data resulted in the currently accruing clinical trial of defactinib and SBRT in pancreatic cancer patients (NCT04331041). The independent mechanisms of radiation and FAK inhibition will hopefully result in acceptable toxicity and minimal changes in weight or severe dermatitis, as was observed in our in vivo experiments.

Our in vivo shRNA screen using radiation and platinum-based chemotherapy demonstrated FAK inhibition as a potential target for both radiosensitization and chemosensitization. In vivo and in vitro data suggest that the benefit occurs in mutant TP53 HPV-negative HNSCC, which is supported by clinical outcomes.

C.R. Pickering reports grants from NIH during the conduct of the study. V.C. Sandulache reports grants from the National Cancer Institute during the conduct of the study. J.V. Heymach reports personal fees from Genentech, Mirati Therapeutics, Janssen Pharmaceuticals, Regeneron, BerGenBio, Jazz Pharmaceuticals, Curio Science, BioAlta, Sanofi, Novartis, GlaxoSmithKline, EMD Serono, BluePrint Medicine, and Chugai Pharmaceutical; personal fees and other support from Eli Lilly & Co; grants, personal fees, and other support from Boehringer-Ingelheim and Spectrum Pharmaceuticals; and grants and personal fees from Takeda Pharmaceuticals and AstraZeneca Pharmaceuticals outside the submitted work. H.D. Skinner reports grants from NIH and NIDCR during the conduct of the study. No disclosures were reported by the other authors.

P.M. Pifer: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. L. Yang: Conceptualization, data curation, validation, investigation, methodology, writing–review and editing. M. Kumar: Conceptualization, validation, investigation, writing–review and editing. T. Xie: Conceptualization, data curation, investigation, methodology, writing–review and editing. M. Frederick: Conceptualization, supervision, methodology, project administration, writing–review and editing. A. Hefner: Validation, investigation, writing–review and editing. B. Beadle: Methodology, writing–review and editing. D. Molkentine: Investigation. J. Molkentine: Conceptualization, resources, investigation, writing–review and editing. A. Dhawan: Investigation, writing–review and editing. M. Abdelhakiem: Investigation, writing–review and editing. A.A. Osman: Resources, writing–review and editing. B.J. Leibowitz: Data curation, investigation, writing–review and editing. J.N. Myers: Conceptualization, resources, supervision, funding acquisition, investigation, project administration, writing–review and editing. C.R. Pickering: Conceptualization, supervision, funding acquisition, methodology, writing–review and editing. V.C. Sandulache: Resources, writing–review and editing. J. Heymach: Resources, supervision, project administration, writing–review and editing. H.D. Skinner: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, writing–original draft.

This study was supported by (i) the National Cancer Institute R01CA168485-08 (H.D. Skinner), 1U54CA274321 (V.C. Sandulache), and T32 CA060397-26 (P.M. Pifer); (ii) the National Institute for Dental and Craniofacial Research R01 DE028105 (H.D. Skinner), R01 DE028061 (H.D. Skinner, C.R. Pickering), and U01DE025181; (iii) The Cancer Prevention Institute of Texas RP150293 (H.D. Skinner, C.R. Pickering); and (iv) Veterans Administration Clinical Science Research and Development Division Career Development Award 1IK2CX001953 (V.C. Sandulache).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.