Cross-talk Signaling between HER3 and HPV16 E6 and E7 Mediates Resistance to PI3K Inhibitors in Head and Neck Cancer

Human papillomaviruses (HPV) are double-stranded DNA viruses that infect the basal layer of squamous epithelia. Over 120 types of HPV have been identified, and 15 are considered to be high-risk viruses that promote cellular growth and transformation (1, 2). HPV type 16 infection is implicated in approximately 75% of head and neck squamous cell carcinomas (HNSCC) that arise in the oropharynx (3, 4). Historically, the main risk factors for HNSCC have been alcohol and/or tobacco consumption; however, over the past three decades the incidence of carcinogen-induced HNSCC has been slowly declining while HPV(+) HNSCC have increased by 225% (3–5). This epidemiologic shift to virally induced disease is associated with a younger patient population who demonstrate a generally more favorable prognosis than patients with HPV(−) tumors (6, 7). Nevertheless, approximately 25% of HPV(+) HNSCC are lethal, underscoring the need for more effective treatments (6–9).

HPV(+) HNSCCs are molecularly distinct from HPV(−) tumors due to viral expression of the E6 and E7 oncoproteins (10–12). E6 and E7 promote carcinogenesis by binding to and inducing the degradation of tumor suppressor proteins in epithelial cells (13). Specifically, E6 and E7 cause the proteasomal degradation of p53 and the retinoblastoma protein (pRb), respectively (14, 15). The oncogenic PI3K pathway, which leads to enhanced cell survival, growth, migration, and altered metabolism (16), is activated by oncogenic HPVs at multiple stages of infection and tumorigenesis (17–20). E6 and E7 activate PI3K pathway signaling by increasing receptor tyrosine kinase (RTK) expression and stability (21–25), inhibiting regulatory phosphatases (26, 27), and degrading miRNAs that play a role in PI3K pathway suppression (28, 29). Consistent with these findings, we recently reported that E6 and E7 induce the expression of the HER family receptor HER3 in HPV(+) HNSCCs (25). HER3 binds PI3K and activates the pathway in HPV(+) cells (25). In addition, the two downstream mediators of PI3K signaling, AKT, and ribosomal protein S6 (rpS6), are phosphorylated in greater than 60% of HPV(+) HNSCCs (30). Finally, the PI3K pathway is frequently activated in HPV(+) tumors by mutation and/or amplification of PIK3CA, the gene that encodes the p110α catalytic subunit of PI3K (31). These cumulative findings demonstrate that the PI3K pathway is commonly activated in HPV(+) tumors through multiple mechanisms.

As the PI3K pathway plays a central role in tumorigenesis and is activated in HPV(+) HNSCC, patients with HPV(+) tumors are currently being enrolled on clinical trials testing PI3K inhibitors. In the current study, we compared the efficacy of PI3K inhibitors in HPV(+) versus HPV(−) HNSCC preclinical models. Despite the biologic rationale for targeting PI3K in HPV(+) tumors, we found that HPV(+) HNSCC cell lines and patient-derived xenografts (PDX) exhibit resistance to PI3K inhibitors due to a feedback upregulation of E6 and E7. The clinical relevance of this adaptive response is underscored by the finding that upregulation of E6 and E7 is mediated by HER3, which can be therapeutically targeted to enhance the efficacy of PI3K inhibitors in HPV(+) tumors.

The HPV(+) HNSCC cell lines UM-SCC47, 93-VU-147T, UPCI-SCC90, UD-SCC2, and UM-SCC104 were kindly provided by Dr. Randall Kimple (University of Wisconsin-Madison, Madison, WI). The HPV(+) HNSCC cell line UPCI-SCC152 was purchased from ATCC. All HPV(+) cell lines were maintained in DMEM with 10% FBS (Gemini Bio-Products) and 1% penicillin and streptomycin (Life Technologies) supplemented with 1% nonessential amino acids (Gibco). The HPV(−) cell lines SCC9 and Detroit562 were purchased from ATCC. PE/CA-PJ49 and PE/CA-PJ34 (clone 12) were purchased from Sigma-Aldrich. CAL33 was kindly provided by Dr. Gerard Milano (University of Nice, Nice, France). HSC-2 was obtained from the Health Science Research Resources Bank (Osaka, Japan). UM-SCC4, UM-SCC1, and TU138 were kindly provided by Thomas E. Carey (University of Michigan, Ann Arbor, MI). Normal oral keratinocytes-spontaneously immortalized (NOKSI) were kindly provided by Dr. Silvio Gutkind (University of California, San Diego, San Diego, CA). All HPV(−) cell lines were maintained in DMEM with 10% FBS and 1% penicillin and streptomycin. NOKSI cells were maintained in Keratinocyte SFM and supplemented with defined keratinocyte-SFM growth supplement (Gibco). CAL33, HSC-2, and Detroit562 express mutant PIK3CA (H1047R). All cell lines were authenticated by short tandem repeat testing (Genetica DNA Laboratories).

All antibodies used are indicated below: HER3-XP, p-HER3-Y1197, p-AKT-S473, p-AKT-T308, AKT, p-rpS6-S235/236, rpS6, p-p70S6K-T308, p70S6K, and GAPDH (Cell Signaling Technology), β-Tubulin (Abcam), E6 (2E-3F8; Euromedex), E7 (Invitrogen), E7, horseradish peroxidase (HRP)–conjugated goat–anti-rabbit IgG, and goat–anti-mouse IgG (Santa Cruz Biotechnology Inc). A combination of both E7 antibodies (1:1) was used for immunoblotting. E6 and E7 antibodies were diluted 1:500 in SuperBlock Buffer (Thermo Fisher Scientific).

The PI3K-targeted therapies BYL719, BKM120, and BEZ235 were purchased from Selleckchem (www.Selleckchem.com). All compounds were dissolved in DMSO for in vitro experiments. CDX-3379 and the control IgG1 antibody CDX-0062C, were kindly provided by Celldex Therapeutics under a Material Transfer Agreement.

HPV16-E6 and HPV16-E7 were cloned into the doxycycline-inducible pLVX-Puro vector (Clontech) using the gateway cloning system (Life Technologies). pDONR223-ERBB3 was a kind gift from Drs. William Hahn (Dana Farber/Harvard Cancer Institute, Boston, MA) and David Root (Addgene plasmid #23874; ref. 32) and was subsequently cloned into pLX302 using the gateway system; pLX302 was a gift from David Root (Broad Institute, Cambridge, MA; Addgene plasmid #25896; ref. 33). All vectors were sequence validated. Lentivirus was produced according to the Lenti-X Tet-One Inducible Expression Systems Manual (Clontech) and subsequently used to establish TU138, 93-VU-147T, and UPCI-SCC90 stable cell lines after selection with puromycin (0.5 μg/mL) for several weeks. Validation of TU138-stable clones was performed 48 hours after incubation with doxycycline (1 μg/mL) and subsequently examined by qPCR for E6 and E7 expression. Validation for HER3 overexpression was examined by immunoblot analysis.

For siRNA-mediated silencing, cells were transiently transfected with siRNAs targeting E6 and E7 (Santa Cruz Biotechnology, sc-156008 and sc-270423), HER3 siRNA (ON-TARGETplus, SMARTpool #L-003127; GE Dharmacon) or nontargeting siRNA (siNT; ON-TARGETplus Non-targeting Pool, #D-001810; Dharmacon). Lipofectamine RNAiMAX was used for all siRNA experiments according to the manufacturer's instructions (Life Technologies).

Cells were seeded in 96-well plates at 30%–50% confluency and treated with escalating concentrations of BYL719, BKM120, or BEZ235 (0–20 μmol/L). After 72 hours of treatment with corresponding drug or siRNA transfection, cells were washed with PBS and fixed/stained with 0.5% crystal violet diluted in methanol. Plates were air dried overnight and dye was eluted with 0.1 mol/L sodium citrate (pH 6.0) diluted in 50% ethanol. After brief agitation, the absorbance was measured at 595 nm (A595) to determine cell growth. The percentage cell proliferation was calculated by comparison of the A595 reading from drug-treated or siRNA-transfected cells versus control cells. IC₅₀ values were calculated with GraphPad Prism (GraphPad Software).

The nature of the interaction between BYL719 or BKM120 and the HER3 antagonist CDX-3379 was evaluated via the fractional product method described by Chou and Talalay (34). Briefly, this method utilizes the relative cell density following treatment with each individual agent and calculates the expected (E) effect of combination therapy as a product of the individual responses. The observed (O) effect is the relative cell density following dual treatment. A ratio of the observed to expected (O:E) values was calculated and used to estimate the synergy, additivity, or antagonism. A value less than 1 indicated synergism, greater than 1 demonstrated antagonism, and 1 represented additivity.

HNSCC cells were plated in 6-well dishes and treated with vehicle, BYL719, or BKM120 for 72 hours. Alternatively, HNSCC cells were transfected with nontargeting siRNA (siNT) or siE6E7 and treated with BYL719 for 72 hours. Cells were then harvested with 0.5% trypsin, washed with PBS, and resuspended in 100 μL of binding buffer (BD Biosciences). Cells were stained with 2 μL of FITC-conjugated Annexin-V antibody and 2 μL of propidium iodide (FITC Annexin-V Apoptosis Detection Kit, BD Biosciences). The cells were analyzed by flow cytometry (Calibur DxP8). FlowJo Software (Tree Star, Inc.) was used to analyze the data.

Whole-cell lysates were harvested using RIPA lysis buffer (50 mmol/L Tris-HCl pH7.5, 150 mmol/L NaCl, 0.1% SDS, 0.5% deoxycholate, 0.5% NP-40) supplemented with protease inhibitor cocktail (Sigma P8340) and phosphatase inhibitor cocktails 2 and 3 (Sigma P5726 and P0044) on wet ice. Samples were sonicated for 10 seconds and then centrifuged at 15,000 × g for 10 minutes at 4°C. Bradford assay was used to determine protein concentrations (Bio-Rad Laboratories). Equal amounts of protein were fractionated by SDS-PAGE, transferred to a polyvinylidene fluoride membrane (Bio-Rad Laboratories), and analyzed by incubation with the appropriate primary antibody. Proteins were detected via incubation with HRP-conjugated secondary antibodies and ECL Western Blotting Substrate (Santa Cruz Biotechnology) or SuperSignal West Femto Maximum Sensitivity Chemiluminescent Substrate (Thermo Fisher Scientific).

Total RNA was isolated using RNeasy kit purchased from Qiagen. One microgram of total RNA was used to make cDNA using the iScript Reverse Transcription Supermix Kit (Bio-Rad Laboratories). qRT-PCR was conducted using CFX96 Touch Real-Time PCR Detection System (Bio-Rad). All reactions were performed in triplicate and repeated three times. To determine the normalized value, 2^ΔΔC_t values were compared between E6 or E7 and B-Actin, where the change in crossing threshold (ΔC_t) = C_{t Target} − C_{t Control} and ΔΔC_t = ΔC_{t (Vector)} − ΔC_{t (Stable Clone)}. The sequences of primer sets used for this analysis are; E6-F: 5′-CTGCAATGTTTCAGGACCCA-3′, E6-R: 5′ -TCATGTATAGTTGTTTGCAGCTCTGT-3′; E7-F: 5′ -ACCGGACAGAGCCCATTACA-3′, E7-R: 5′-GCCCATTAACAGGTCTTCCAAA-3′; β-Actin-F: 5′-CAGCCATGTACGTTGCTATCCAGG-3′, β-Actin-R: 5′-AGGTCCAGA CGCAGGATGGCATG-3′.

Luminex PhosphoRTK bead array assays were conducted according to manufacturer's instructions (R&D Systems, custom kit). Briefly, cells cultured in 6-well plates were washed twice in ice-cold PBS then lysed with 100 μL of the supplied lysis buffer 15 supplemented with protease and phosphatase inhibitor cocktails (Sigma). Protein extracts were diluted with diluent buffer RD2-3 and 25 μL, containing 6 μg of protein, was added to 25 μL bead-detection mixture in round-bottom nonbinding 96-well plates (Corning 3605). Plates were sealed and incubated overnight at 4°C. Luminex beads were collected on a 96-well magnet and washed twice with 75 μL of Luminex sheath fluid. Beads were resuspended in 60 μL of sheath fluid and analyzed using a Luminex 200 instrument.

All animal procedures and maintenance were conducted in accordance with protocols approved by the Institutional Animal Care and Use Committee of the University of California, San Francisco (San Francisco, CA). HNSCC PDXs were established from patients with newly diagnosed or recurrent HNSCC upon written consent in accordance with Institutional Review Board approval as described previously (35). Female Hsd:athymic Nude-Foxn1^nu mice were injected bilaterally with UM-SCC47 cells (1 × 10⁶ cells). HNSCC PDXs were established in female NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ (NSG) mice as described previously (35). Once xenografts reached 100 mm³ to 200 mm³, mice were randomized into treatment groups and started on their respective treatments. The dose of BYL719 and BKM120 used was 20 mg/kg [diluted in 0.5% carboxymethylcellulose sodium salt (Sigma) in water] and were administered 5 days/week by oral gavage. The dose of CDX-3379 was 10 mg/kg and was administered biweekly by intraperitoneal injection. For all xenograft experiments examining tumor volume, animals were randomized into respective treatment groups with n = 5–8 tumors per group. For analysis of signaling parameters after 5 days of BYL719 treatment, n = 4 to 6 tumors per treatment group. All tumors were harvested 3 hours after the last treatment. Tumor volume measurements were determined by digital calipers biweekly and calculated using the formula (π)/6 × (large diameter) × (small diameter)². Tumor volumes for all xenograft experiments were calculated as fractions of the average starting volume for each group, and the mean tumor volume ± SEM is shown.

Extracted tumor tissue was collected at the indicated time periods, fixed in 10% neutral-buffered formalin, paraffin-embedded, and cut into sections. The sections were heated in 10 mmol/L citrate buffer (pH 6.0) for Ki67 or in EDTA buffer (pH 8.0) for HER3 and E7 in a decloaking chamber. Staining commenced using the Universal quick kit following the manufactures instructions (Vector Laboratories, Inc., PK-8800). Samples were incubated overnight at 4°C with the following antibodies: Ki67 (Cell Signaling Technology, 1:1,000), HER3 (Cell Signaling Technology, 1:50), and E7 (Santa Cruz Biotechnology, 1:50). Antibody binding was revealed by addition of 3,3′-diaminobenzidine substrate and counterstained with Mayer hematoxylin (Vector Laboratories). Tissues were examined using a Nikon Eclipse microscope and quantitation of staining intensity was performed with Figi software.

For all in vitro assays, at least three replicates were analyzed with the same experimental conditions, and at least three independent experiments were performed. Two-tailed Student t test was used to evaluate differences in cell proliferation and apoptosis. Student t test was also used to evaluate differences in protein abundance in immunoblots and staining intensity for IHC images. For in vivo growth assays, 5 to 8 tumors per treatment group were analyzed. To evaluate the sensitivity of cell line xenografts or PDXs to the different therapies, the mean fractional tumor volumes were compared between the vehicle treated and drug treatment arms at the last time point of the study using two-sample t tests with equal SDs.

Because the PI3K pathway is commonly activated in HPV(+) tumors, we hypothesized that HPV(+) HNSCCs would be more sensitive to PI3K inhibitors than HPV(−) HNSCCs. To test this hypothesis, the antiproliferative effects of several different PI3K inhibitors were examined in a panel of HPV(+) and HPV(−) HNSCC cell lines (see Materials and Methods for molecular characteristics of cell lines examined). We first determined the half-maximal inhibitory concentration (IC₅₀) for BYL719 (a p110α-specific inhibitor that is currently in phase II clinical trials in HNSCC) in six HPV(+) and nine HPV(−) cell lines after 72 hours of treatment. The IC₅₀ concentration for BYL719 in all HPV(+) cell lines tested ranged from 3.5 μmol/L to 15.0 μmol/L, whereas the IC₅₀ concentration in HPV(−) cell lines ranged from 0.75 μmol/L to 5.2 μmol/L (Fig. 1A). Furthermore, HPV(−) cell lines treated with 1.0 μmol/L BYL719 demonstrated significant increases in apoptosis relative to HPV(+) cells, as assessed by Annexin V staining (Fig. 1B). Collectively, HPV(+) cell lines were more resistant to BYL719 than were HPV(−) cell lines in both proliferation and apoptosis assays (P = 0.004 and P = 0.03, respectively; Fig. 1C). Similar results were seen with BKM120, a clinically advanced pan-PI3K inhibitor, and the dual PI3K/mTOR inhibitor BEZ235 (Supplementary Fig. S1).

To characterize the biochemical effects of BYL719 in HPV(+) and HPV(−) cell lines, PI3K pathway signaling was analyzed 72 hours after BYL719 treatment via immunoblotting for AKT and rpS6 (Fig. 1D). BYL719 did not abrogate phosphorylation of AKT-S473 or rpS6-S235/236 in the HPV(+) lines, whereas treatment with BYL719 markedly reduced phosphorylation of these proteins in HPV(−) cell lines (Fig. 1D). These data suggest that p110α inhibition does not effectively block signaling through AKT and rpS6 in HPV(+) cells, which may likely contribute to reduced sensitivity to PI3K inhibitors.

To extend these findings to more clinically relevant models, we examined the antitumor effects of PI3K inhibitors in HPV(+) and HPV(−) patient-derived xenografts (PDXs; see Supplementary Table S1 for clinical characteristics of the PDXs examined). Four HPV(+) and two HPV(−) PDXs were established in NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ (NSG) mice and treated with BYL719 (20 mg/kg by oral gavage) for five consecutive days. In all four HPV(+) PDXs, including the HPV(+)/PIK3CA-mutant PDX (PDX-4), five-day treatment with BYL719 did not significantly reduce the percentage of Ki67-positive nuclei (Fig. 1E). In contrast, the two HPV(−) PDXs treated with BYL719 demonstrated significant reductions in Ki67-positive nuclei compared with vehicle-treated tumors. BYL719 reduced expression of phospho-AKT(S473) and phospho-rpS6(S235/236) in the HPV(−) PDX-5, while the phosphorylation of these proteins were only marginally reduced in HPV(+) PDXs (Fig. 1F). To determine whether these findings were consistent among different PI3K inhibitors, the antitumor effects of the pan-PI3K inhibitor BKM120 were assessed in a representative HPV(+) PDX (PDX-3) and an HPV(−) PDX (PDX-5). Treatment with BKM120 for 21 days (20 mg/kg, 5 days/week by oral gavage) inhibited the growth of the HPV(−) PDX while exerting little effect on the HPV(+) PDX (Fig. 1G). Analysis of PI3K pathway signaling in tumors harvested after the last treatment indicated that phospho-AKT(S473) and phospho-rpS6(S235/236) were markedly reduced in BKM120-treated HPV(−) tumors, but only marginally reduced in HPV(+) tumors (Fig. 1H). Taken together, these results suggest that PI3K inhibitors may be ineffective as single-agent therapies in HPV(+) HNSCCs.

Given our findings that HPV(+) HNSCC preclinical models were relatively resistant to PI3K-targeted therapies, we next explored the role of the E6 and E7 oncoproteins in mediating resistance. Proliferation assays were performed in three HPV(+) cell lines transfected with pooled siRNAs targeting polycistronic E6 and E7 RNAs followed by treatment with 1.0 μmol/L of BYL719 (Fig. 2A). Consistent with previous reports (36, 37), knockdown of E6 and E7 reduced the proliferation of all three HPV(+) cell lines. Immunoblot analysis confirmed the knockdown of E6 and E7 with the siRNAs used. Knockdown of E6 and E7 sensitized HPV(+) cell lines to BYL719 in proliferation assays (Fig. 2A). Furthermore, siRNA depletion of E6 and E7 increased BYL719-induced apoptosis in all three HPV(+) cell lines tested (Fig. 2B). The knockdown of E6 and E7 in combination with BYL719 also reduced phosphorylation of AKT(S473), p70S6-kinase(T389), and rpS6(S235/236) substantially more than E6 and E7 knockdown or BYL719 treatment alone in all HPV(+) lines tested (Fig. 2C).

To confirm the role for HPV oncoproteins in PI3K inhibitor resistance, E6 and E7 lentiviral particles were transduced into the HPV(−) cell line TU138 (Fig. 2D). Compared with vector-transduced control cells, two independent clones overexpressing E6 and E7 displayed increased PI3K pathway signaling, as indicated by enhanced phosphorylation of AKT(S473) and rpS6 (S235/236; Fig. 2D, right). Whereas vector-transduced cells were highly sensitive to 1.0 μmol/L BYL719 and BKM120, exogenous expression of E6 and E7 resulted in reduced sensitivity to both agents (by ∼34%–45%; Fig. 2E). Furthermore, both BYL719 and BKM120 induced apoptosis in vector-transduced control cells, while E6- and E7-expressing clones were more resistant to BYL719- and BKM120-induced apoptosis (by ∼20%; Fig. 2F). Overall, these findings suggest that E6 and E7 can promote resistance to PI3K inhibitors in HPV-associated HNSCCs.

We next sought to elucidate the molecular mechanism by which HPV oncoproteins might contribute to PI3K inhibitor resistance. We recently reported that E6 and E7 induce expression of HER3 in HPV(+) HNSCC models (25). HER3 has also been implicated in resistance to PI3K inhibitors in breast cancer (38–40). To determine whether HER3 mediates resistance to PI3K inhibitors in HPV(+) HNSCC, we analyzed the phosphorylation status and total abundance of HER3 over a 24-hour time course of BYL719 treatment (Fig. 3A). Within 8 hours of BYL719 exposure, HER3 and phospho-HER3(Y1197) protein levels were increased in HPV(+) cell lines UM-SCC47 and 93-VU-147T. This time course also revealed that phosphorylation of AKT, at both serine 473 and threonine 308, was reduced by 15 minutes but was restored within 24 hours of treatment. In the HPV(+) cell line SCC90, HER3 and phospho-HER3 protein levels were increased within 15 minutes of BYL719 exposure, and no detectable change in AKT phosphorylation was observed. A phospho-RTK screen was also performed in HPV(+) cell lines treated with 1.0 μmol/L of BYL719 for 24 hours. The results indicated that HER3 was the only RTK whose phosphorylation increased in all three HPV(+) cell lines following treatment with BYL719 (Supplementary Fig. S2). In contrast to HPV(+) cell lines, HER3 protein levels were either unchanged or reduced in all three HPV(−) cell lines treated with BYL719, and AKT phosphorylation was inhibited up to 24 hours (Fig. 3A). Feedback signaling through HER3/AKT was also observed in HPV(+) cell lines treated with BKM120 and BEZ235 over a 24-hour time course (Supplementary Fig. S3).

As total HER3 was increased in HPV(+) cell lines treated with PI3K inhibitors, we next determined whether this was reflected in expression of HER3 on the plasma membrane (Fig. 3B). Flow cytometric analysis demonstrated a significant increase in HER3 surface levels in HPV(+) cell lines treated with BYL719. We then tested the hypothesis that the reactivation of AKT observed after PI3K inhibitor treatment is dependent on HER3. HPV(+) cell lines were transfected with siRNA targeting HER3 or a nontargeting control (siNT) for 48 hours followed by treatment with BYL719 for up to 24 hours (Fig. 3C). While AKT was reactivated within 8 hours of BYL719 exposure in cells transfected with siNT, reactivation was either reduced or blocked in cells depleted of HER3.

To determine whether similar mechanisms occur in HPV(+) in vivo models, we evaluated the abundance and phosphorylation status of HER3 in HPV(+) PDXs treated with BYL719 for five consecutive days (Fig. 3D). Analysis of tumors harvested from three HPV(+) PDXs (PDX-1, PDX-2, and PDX-3) indicated that HER3 phosphorylation and abundance were increased in tumors after five days of BYL719 treatment. IHC staining also demonstrated strong HER3 staining on the plasma membrane in BYL719-treated tumors (Fig. 3E). Collectively, these results suggest that the HER3/AKT signaling pathway is activated in HPV(+) preclinical models in response to PI3K inhibitor treatment.

Our data indicate that HPV E6 and E7 oncoproteins promote resistance to PI3K inhibitors in HPV(+) models and that HER3/AKT signaling is rapidly activated following PI3K inhibition. We next sought to determine the impact of HER3 signaling on E6 and E7 expression. Similar to HER3, E6 and E7 protein levels were increased in HPV(+) cell lines treated with BYL719 for 24 hours (Fig. 4A). To determine whether this increase in E6 and E7 was dependent on HER3, HER3 expression was knocked down using siRNA and then E6 and E7 protein levels were analyzed. Downregulation of HER3 resulted in reduced E6 and E7 protein levels (Fig. 4B). Conversely, ectopic overexpression of HER3 in the HPV(+) cell lines 93-VU-147T and UPCI-SCC90 led to increased E6 and E7 protein levels compared with vector-transduced control cells (Fig. 4C). Importantly, HER3 downregulation prevented increases in E6 and E7 oncoproteins following BYL719 treatment (Fig. 4D).

To extend these in vitro studies to our in vivo models, E6 and E7 protein levels were analyzed in HPV(+) PDXs treated with BYL719 for five days. As observed with HER3, BYL719 treatment led to increased abundance of E6 and E7 in three HPV(+) PDXs treated with BYL719 (Fig. 4E). IHC staining for E7 confirmed this result in HPV(+) PDX-1, in which more robust nuclear E7 staining was observed in BYL719-treated tumors (Fig. 4F). Together, these data suggest that cross-talk between HER3 and HPV oncoproteins is increased in HPV(+) preclinical models treated with PI3K inhibitors.

As cross-talk between HER3 and HPV oncoproteins was elevated in PI3K inhibitor–treated cells, we hypothesized that suppression of HER3 signaling would reduce the abundance of E6 and E7 and sensitize HPV(+) preclinical models to PI3K inhibitors. To test this hypothesis, proliferation assays were performed in HPV(+) cell lines after suppressing HER3 expression with siRNA and treating with BYL719 (Fig. 5A). The combination of HER3 siRNA and BYL719 reduced the proliferation of three HPV(+) cell lines more robustly than single-agent BYL719 or siHER3 alone. Similar results were observed in HPV(+) cell lines depleted of HER3 and treated with BKM120 (Supplementary Fig. S4A). Next, the anti-HER3 mAb CDX-3379, which is in active clinical development in HNSCC, was tested in combination with BYL719. CDX-3379 locks HER3 in an inactive conformation, inhibiting both ligand-dependent and -independent activation of HER3 (41). In proliferation assays, CDX-3379 significantly enhanced the efficacy of BYL719 in all HPV(+) cell lines examined (Fig. 5B). Although HPV(−) cell lines were sensitive to BYL719, CDX-3379 did not enhance this effect. The combination of CDX-3379 and BYL719 was evaluated for synergy using the fractional product method described by Chou and Talalay (34). HER3 inhibition synergized with BYL719 in each of the HPV(+) cell lines examined (Fig. 5B). CDX-3379 was also found to synergize with BKM120 in HPV(+) cell lines (Supplementary Fig. S4B). Immunoblot analysis indicated that the combination of CDX-3379 and BYL719 suppressed the phosphorylation of AKT and rpS6 more than either treatment alone (Fig. 5C). Furthermore, targeting HER3 with CDX-3379 reduced E6 and E7 levels and prevented their upregulation by BYL719 (Fig. 5C). Together, these results demonstrate that targeting HER3 can enhance the efficacy of PI3K inhibitors and sustain knockdown of the PI3K pathway in HPV(+) cells.

We next assessed the impact of HER3 inhibition on response to BYL719 in vivo. The HPV(+) cell line UM-SCC47 and two HPV(+) PDXs, PDX-1 and PDX-4, were established in athymic nude or NSG mice (Fig. 6A). Tumor-bearing mice were treated with vehicle, BYL719 (20 mg/kg), CDX-3379 (10 mg/kg), or the combination for 19–22 days (n = 5–8 tumors per treatment group; Fig. 6A). While UM-SCC47 tumors were moderately sensitive to BYL719, PDX-1 and PDX-4 exhibited only a minor response to BYL719. All the xenograft models responded to CDX-3379, but the combination of BYL719 and CDX-3379 was significantly more potent than single-agent CDX-3379 therapy (Fig. 6A). Tumors from mice bearing PDX-1 were harvested at the end of the study to analyze PI3K pathway signaling (Fig. 6B). Expression of total and phosphorylated HER3 was increased in tumors treated with BYL719, but reduced in tumors treated with CDX-3379. Compared with single-agent treatment groups, tumors harvested from the combination group had reduced levels of phospho-AKT, phospho-rpS6, and nuclear Ki67 (Fig. 6C and D). Consistent with results in HPV(+) cell lines, IHC revealed robust staining for HER3 and E7 in BYL719-treated PDX-1, while CDX-3379 reduced the staining intensity of both proteins. Taken together, these data demonstrate that cotargeting HER3 and PI3K can overcome PI3K inhibitor resistance and block feedback signaling through HER3 and E7 in vivo.

PI3K-targeted therapies are currently in clinical development for the treatment of HNSCC. Initial results have shown only moderate clinical benefit in unselected patient populations with the rapid emergence of drug resistance (42, 43). As the PI3K pathway is activated in HPV(+) HNSCCs, we hypothesized that HPV(+) preclinical models would be sensitive to PI3K-targeting strategies. Contrary to our original hypothesis, we found that HPV(+) HNSCC models were more resistant to PI3K inhibitors than HPV(−) models. PI3K inhibitor resistance was attributed to feedback signaling emanating from HER3 and HPV oncoproteins, which maintained AKT activity in PI3K inhibitor–treated cells. Blocking HER3 activity abrogated this feedback loop and enhanced the efficacy of PI3K inhibitors. This is the first study to identify a role for HPV oncoproteins in HNSCC drug resistance and provides a rational cotargeting strategy to improve the efficacy of PI3K inhibitors in HPV(+) tumors.

Our results demonstrated that HPV(+) HNSCC cell lines and PDXs were less responsive to PI3K inhibitors than HPV(−) models. This difference in drug sensitivity may be attributed to the reactivation of AKT by HER3 following PI3K inhibitor treatment of HPV(+) cells. Suppression of HER3 reduced this adaptive response and enhanced the efficacy of PI3K inhibitors in HPV(+) cell line and tumor models. A similar observation was reported in preclinical models of HER2(+) breast cancer, in which PI3K inhibition resulted in FoxO1-dependent upregulation of HER3 and reactivation of AKT following treatment (38). Others reported that the RTK AXL mediates acquired resistance to PI3Kα inhibition independent from AKT reactivation in HPV(−) PIK3CA-mutant HNSCC models (44). Consistent with these findings, PI3K inhibitors blocked AKT signaling in HPV(−) HNSCC cell lines and PDXs tested in the current study, and these cell lines were relatively sensitive to PI3K inhibitors in short-term assays. In contrast to the current findings, previous studies demonstrated that UD-SCC2 and UM-SCC47 cells were relatively sensitive to the mTOR inhibitors rapamycin and RAD001 (30, 45). Both inhibitors reduced the phosphorylation of AKT and rpS6 in these models (30, 45). This discrepancy in results may be due to differences in feedback signaling cascades that are activated upon inhibition of different nodes of the PI3K pathway. Further research is necessary to determine the most effective way of suppressing PI3K/AKT signaling in HPV(+) HNSCCs. Taken together, the reactivation of AKT by HER3 may limit the efficacy of PI3K inhibitors in HPV(+) HNSCCs, and therefore cotargeting HER3 and PI3K may be an effective therapeutic strategy to overcome PI3K inhibitor–resistance in HPV(+) tumors.

E6 and E7 are ubiquitously expressed in HPV(+) tumors, where integration of viral DNA into the host's genome results in stabilization and heightened expression of E6 and E7 in the tumor (1, 46, 47). One of the consequences of E6 and E7 expression is increased stabilization of, and signaling through, RTKs (21, 22). Our results show that HER3 can reciprocally increase the abundance of E6 and E7 in HPV(+) cells. Cross-talk between HER3 and HPV oncoproteins was heightened in PI3K inhibitor–treated cells, resulting in maintenance of AKT activity and PI3K inhibitor resistance. This mechanism of PI3K inhibitor resistance is supported by the finding that knockdown of E6 and E7, or blockade of HER3, sensitized HPV(+) cells to PI3K inhibitors. To our knowledge, this is the first study to report reciprocal feedback signaling between an RTK and HPV oncoproteins. This finding suggests that therapeutic targeting of HER3 could be a novel method to decrease the abundance of E6 and E7 in HPV(+) tumors. The precise mechanism underlying HER3 upregulation of E6 and E7 remains incompletely understood; however, we speculate that HER3-mediated activation of AKT could play an important role. Collectively, these data highlight a new adaptive response to PI3K inhibitors that may limit the efficacy of these agents in HPV(+) tumors.

While HPV oncoproteins activate the PI3K pathway in HPV(+) tumors, PIK3CA mutation and/or amplification can also contribute to hyperactivation of the pathway. Several preclinical studies and early-phase clinical trials suggest that PIK3CA mutation is correlated with increased response to PI3K pathway inhibitors (48–50). However, other studies have shown no consistent correlation between PIK3CA mutation status and enhanced response to PI3K-targeted therapies (42, 43, 51, 52). In the current study, HPV(+) PIK3CA-mutant and wild-type PDXs were less sensitive to PI3K inhibitors than HPV(−) PDXs. While these data suggest that HPV(+) tumors may be intrinsically resistant to PI3K inhibitors irrespective of PIK3CA mutation status, the paucity of HPV(+)/PIK3CA-mutant HNSCC cell lines limits our ability to fully test this hypothesis in cell culture models. To date, only a small panel of HPV(+) HNSCC cell lines have been generated, and all express wild-type PIK3CA. Furthermore, HPV(−) PIK3CA-mutant and wild-type HNSCC cell lines were found to be sensitive to PI3K inhibitors, suggesting that additional biomarkers beyond PIK3CA mutation are needed to more broadly identify PI3K inhibitor–responsive tumors.

Targeting HER3 reversed PI3K inhibitor resistance in HPV(+) HNSCC cell lines and PDXs. This finding is in accordance with several preclinical studies demonstrating robust antitumor effects of cotargeting PI3K with RTKs in different tumor models (38, 40, 44). The challenge in moving this strategy into clinical trials will be to predict which RTK to target based on the molecular and genetic underpinnings of each patient's tumor, in addition to managing the likely toxicities of combined approaches. Our recent work demonstrates that HER3 is overexpressed and active in HPV(+) HNSCCs (25, 53), suggesting that HER3 could be an important molecular target in HPV(+) tumors. In comparison with HPV(+) preclinical models, the HPV(−) cell lines tested did not respond to the dual targeting of HER3 and PI3K. Thus, HPV(−) tumors may benefit from alternative cotargeting strategies to enhance responses to PI3K inhibitors. The feasibility of the proposed cotargeting approach is further supported by a recent phase Ib clinical trial demonstrating that the combination of a HER3 antagonist with a PI3K inhibitor was safe and well tolerated by breast cancer patients (54).

In summary, this study suggests that cross-talk between HER3 and HPV oncoproteins maintains signaling through AKT in HPV(+) tumors treated with PI3K inhibitors. We propose that this mechanism could limit the clinical efficacy of PI3K inhibitors in HPV(+) patients. As a result, combined blockade of HER3 and PI3K may prevent this feedback resistance mechanism and be an effective therapeutic strategy for the treatment of patients with HPV(+) HNSCC.

No potential conflicts of interest were disclosed.

Conception and design: T.M. Brand, N.E. Bhola, M.V. Ranall, N.J. Krogan, T. LaVallee, M.A. Ozbun, J.E. Bauman, J.R. Grandis

Development of methodology: T.M. Brand, M.V. Ranall, U. Duvvuri, T. LaVallee

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.M. Brand, S. Hartmann, N.E. Bhola, H. Li, Y. Zang, M.V. Ranall, M. Soucheray, C. Kemp

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.M. Brand, S. Hartmann, N.E. Bhola, M.V. Ranall, U. Duvvuri, T. LaVallee, M.A. Ozbun, J.E. Bauman

Writing, review, and/or revision of the manuscript: T.M. Brand, S. Hartmann, N.E. Bhola, S. Bandyopadhyay, U. Duvvuri, T. LaVallee, D.E. Johnson, M.A. Ozbun, J.E. Bauman, J.R. Grandis

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T.M. Brand, R.A. O'Keefe, S. Bandyopadhyay

Study supervision:T.M. Brand, N.J. Krogan, J.R. Grandis

The authors thank Dr. Danielle Swaney (University of California, San Francisco, San Francisco, CA) for critically reviewing this manuscript. The project described was supported by RO1 DE023685 from the National Cancer Institute (to J. Grandis), the American Cancer Society (to J. Grandis), the V Foundation for Cancer Research (to J.E. Bauman), Career Development Award CDA-2-057-10S from the Department of Veterans Affairs (to U. Duvvuri), grant Z-2/59 from the Interdisciplinary Center for Clinical Research of the University Würzburg (to S. Hartmann), postdoctoral fellowship PF-16-087-01-TBG from the American Cancer Society (to T.M. Brand), by NIH grant T32 CA108462 for Molecular and Cellular Mechanisms of Disease (to T.M. Brand), and U54 CA209891 (to N.J. Krogan).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.