Head and neck squamous cell carcinomas (HNSCC) are frequently altered along the PI3K/AKT/mTORC signaling axis. Despite excellent preclinical data, the use of compounds targeting this pathway as monotherapy has been underwhelming in initial clinical trials, and identification of predictive biomarkers remains challenging. To investigate mTORC-specific inhibition, we tested catalytic mTORC (AZD8055) and PI3K/mTORC (NVP-BEZ-235) inhibitors ± cetuximab in a panel of HNSCC cell lines and patient-derived xenografts (PDX). Cell lines were assayed for response to all agents and siRNA knockdown of targets by multiple approaches. All cell lines showed similar response to both drug and siRNA inhibition of both PI3K and mTORC pathways, with anti-EGFR combination producing modest additive effect. Five PDX models that presented PIK3CA mutation or intrinsic cetuximab resistance were treated with a combination of cetuximab and AZD8055. In vivo single-agent mTORC inhibition inhibited growth of one PIK3CA-mutant cancer, but had little effect on any PIK3CAWT or a second PIK3CA-mutant model. In all models, the combination therapy showed greater growth delay than monotherapy. The uniform ability of PI3K and mTORC inhibition to suppress the growth of HNSCC cells highlights the pathway's role in driving proliferation. Although single-agent therapy was largely ineffective in vivo, improved response of combination treatment in an array of PDXs suggests the potential for adding a catalytic mTORC inhibitor to cetuximab therapy. Overall, these results add to a growing body of evidence, suggesting that approaches that attempt to match biomarkers to the optimal therapy in HNSCC remain complex and challenging. Mol Cancer Ther; 16(7); 1257–68. ©2017 AACR.
Head and neck squamous cell carcinoma (HNSCC) remains a common and deadly disease with almost 60,000 new diagnoses and 13,000 deaths annually in the United States (1). Although declining tobacco use in the United States has reduced disease incidence, this has been partially offset by increases in human papillomavirus (HPV)–driven cancers (2–5), resulting in the need for improved treatment of HNSCC for the foreseeable future. The high frequency of EGFR overexpression in HNSCC led to the use of the EGFR-targeting monoclonal antibody cetuximab for the treatment of this disease (6, 7). Although cetuximab and related anti-EGFR antibody therapies provide increased survival for patients, recent studies have revealed that the long-term clinical benefit may be limited (8). Despite meaningful responses in a small proportion (<20%) of HNSCC patients treated with immunotherapy (9, 10), most patients fail to benefit from these drugs, suggesting the continued need for alternative or combination approaches. The Cancer Genome Atlas (TCGA) report and related studies have revealed potentially targetable pathways, most prominently increased pro-proliferative signaling along the PI3K/AKT/mTORC axis driven primarily by gain of function activating mutation or amplification of PIK3CA, or loss of function of its negative regulator PTEN (11–15).
These findings have promoted the development of targeted therapeutics against PI3K/mTORC signaling. Over 30 active clinical trials in various phases are investigating PI3K inhibitors for head and neck cancer alone. Two recent entrants BYL719 and BKM120 have resulted in preclinical success with on-going trials in multiple solid tumor indications including HNSCC (16–18). Although promise remains for PI3K inhibitors, to date there have not been any major clinical successes (19). An alternative point at which to impact this pathway is further downstream at the mTORC signaling node. The clinical use of rapamycin and the derivative rapalogs was among the first class of targeted agents to be used in the treatment of cancer. Although there are currently a number of trials evaluating rapamycin, everolimus, and temsirolimus, to date there have been no major successes in the treatment of HNSCC (20). More recent efforts have focused on ATP competitive kinase inhibitors targeting either PI3K/mTORC or the dual mTORC complexes. As these compounds target the catalytic site of mTOR, they affect both the AKT phosphorylation function of the Rictor-containing mTORC2 complex, and the protein translation function of the raptor-containing mTORC1 complex (21). The preferred targeting approach remains uncertain.
Herein, we utilized two hallmark compounds in this field, the PI3K/mTORC inhibitor NVP-BEZ-235 and the dual mTORC1/2 inhibitor AZD8055 (18, 22–27), to investigate their role in improving the response to cetuximab in HNSCC. Although newer generations of both classes with improved toxicology profiles have emerged, the goal here was to evaluate these classes of compounds in the treatment of HNSCC both as single agents and in combination with the standard of care cetuximab. Although initial reports had initially indicated that mutations or other alterations along the PI3K/Akt/mTORC axis strongly predicted sensitivity to inhibitors targeting this pathway (12, 28), more recent evidence suggests that this relationship is more complex (29, 30) and that tumors both with and without such alterations may benefit from the addition of PI3K/mTORC to cetuximab treatment (16). In this work, we test both in vitro and in vivo HNSCC models, containing both PI3K/Akt/mTORC altered and unaltered examples, to determine whether the addition of an ATP competitive kinase inhibitor can provide additional antitumor effects. Given the number of PI3K-specific compounds that have already progressed to clinical trial, we focused our efforts on the dual PI3K/mTORC and dual mTORC classes as a precursor to follow on studies of these trials.
Materials and Methods
Additional details on all methods are included in Supplementary Information.
Cell lines, strains, and drugs
Cell line (HPV-negative: UM-SCC1 and TU-138; HPV-positive: UD-SCC2, UM-SCC47, and UPCI-SCC90) sources and culture conditions are summarized in Supplementary Table S1. UM-SCC1 and TU-138 were received in 2001 and 2006, respectively. All HPV-positive cell lines were obtained in 2010. The identity of all cell lines was confirmed via short-tandem repeat testing within 3 months of use (Supplementary Data). AZD8055 and NVP-BEZ-235 were purchased from LC-Labs. Cetuximab (IMC-225; Erbitux) was purchased from the University of Wisconsin Hospital Pharmacy.
Growth inhibition assay
Cells were plated in 96-well plates and 24 hours after plating treated at indicated doses and incubated for 72 to 96 hours. Every 24 hours, cells per well were determined using a SpectraMax i3 MiniMax 300 Imaging Cytometer plate reader using the optical imaging module and transmitted light setting. The Softmax-Pro v6.4 software was trained to automate counting of each cell type. This timecourse approach enabled the temporal analysis of cell proliferation; at 72 or 96 hours, when control wells were nearing confluency, treated wells were normalized to DMSO controls and IC50 values were calculated with GraphPad Prism v7 using the inhibitor concentration versus response (three-parameter) function.
Clonogenic expansion assay
Cells were plated in 12-well plates, treated at indicated doses 24 hours after plating, and allowed to incubate in media containing the inhibitors for 7 to 21 days, until DMSO control wells had colonies of 50 or more cells. Colonies were fixed, stained with methanol/crystal violet, and colonies of 50 or more cells were counted, surviving fraction was determined, and plotted using Prism. Statistical comparisons were made via one-way ANOVA with the Tukey multiple comparison test.
Cells were plated in 96-well plates, 24 hours after plating treated at indicated doses, and allowed to incubate for 72 to 96 hours. BrdUrd incorporation was measured and normalized to DMSO control using a BrdUrd ELISA according to the manufacturer's instructions (Roche; #11 647 229 001).
Apoptosis was measured by caspase activity with the Caspase-Glo 3/7 assay (Promega #G8091) according to the manufacturer's instructions. Luciferase intensity was normalized to number of cells/well and expressed relative to DMSO controls.
Following treatment, cells were lysed with RIPA buffer supplemented with protease/phosphatase inhibitor cocktail (Cell Signaling Techonologies; #5872) and sonicated. Equal amounts of protein were analyzed by SDS-PAGE (∼50 μg/well), transferred to polyvinylidene difluoride membranes (Immobilon-FL), probed by specific primary antibodies (Supplementary Table S2) by overnight incubation at 4°C, and detected with NIR-conjugated secondary antibodies (LiCOR) imaged on a LiCOR Odyssey FC.
siRNA knockdown experiments
Cells were transfected in suspension at indicated concentrations of siRNA oligos (Supplementary Table S3) using Lipofectamine RNAiMax (Life Technologies). For Western blots, cells were incubated with siRNA for 72 hours prior to harvest and processed as above. For proliferation assays, cells were analyzed as above.
Hotspot mutational analysis
Total genomic DNA was isolated from PDX formalin fixed, paraffin embedded (FFPE) tissue and HNSCC cell lines and sequenced using the Illumina TruSeq Cancer Amplicon panel and analyzed as described previously (31). Sequencing data have been deposited with the sequence read archive under BioProject ID: PRJNA381909.
Six- to 8-week-old female NOD-SCID gamma (NSG, NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ) mice (Jackson Laboratories) were used for PDX revival and tissue amplification; 6- to 8-week-old female Hsd:athymic Nude-Foxn1nu(Harlan Laboratories) were used for therapy studies (32). Mice were kept in the Association for Assessment and Accreditation of Laboratory Animal Care–approved Wisconsin Institute for Medical Research (WIMR) Animal Care Facility. Animals were housed in specific pathogen-free rooms, and their clinical health was evaluated weekly. Studies involving the mice were carried out in accordance with an animal protocol approved by the University of Wisconsin.
PDX cetuximab response studies
HNSCC PDXs were generated and propagated in our group as described previously (32, 33). Ten models were evaluated for response to cetuximab by reanimating cryopreserved PDX tissue in NSG mice and then distributed to nude mice for therapy studies. Tumor volume was assessed twice weekly with Vernier calipers and calculated according to the equation V = (π/6) × (large diameter) × (small diameter)2; when average volume reached 250 mm3, mice were randomized into control or cetuximab treatment, receiving cetuximab 2X/week at 10 mg/kg by intraperitoneal injection for 2 weeks, vehicle mice received saline i.p. injections. Two weeks following treatment end, tumor size was measured and used to calculate treated/control (T/C) ratios using Prism.
PDX combination therapy studies
Mice were randomized into control (vehicle), AZD8055, cetuximab, or combination treatment groups (n = 4–6 mice/10–16 tumors per group). AZD8055 was formulated in 30% Capsitol and delivered once daily at 20 mg/kg by oral gavage as described previously (22). Cetuximab was dosed twice weekly at 10 mg/kg i.p. Vehicle and single treatment mice received 30% Capsitol (orally) or saline (intraperitoneally) as appropriate. At times indicated post-treatment, additional tumor-bearing mice were sacrificed and tumors were harvested for FFPE. Growth curves were statistically evaluated with the nonparametric Friedman test using Prism.
Histology and immunohistochemistry
FFPE-preserved PDX tissue was microtomed for 5-μm sections and stained with hematoxylin and eosin (H&E) or IHC for indicated targets (Supplementary Table S1) as described previously (34).
Characterization of HNSCC PDX and cell line models identifies potential mTORC inhibition targets
We characterized an array of in vivo and in vitro HNSCC models, analyzing potential genetic and protein biomarkers, and evaluating therapeutic response to the current standard-of-care targeted agent, cetuximab. We have previously described the establishment and other characteristics of our cohort of PDX models (31–33); here, we show hotspot mutational profiling using an Illumina cancer-targeted sequencing panel, with key genes and patient demographic data highlighted in Fig. 1A. We identified two PDXs in our cohort containing PIK3CA helical-domain activating mutations, as well as a range of PTEN expression levels as analyzed by IHC (Fig. 1A and Supplementary Fig. S1). We subjected approximately half of the cohort to in vivo cetuximab treatment studies using a flank subcutaneous approach. Response was evaluated as durable growth delay for cetuximab treated relative to control mice (T/C ratio) and was plotted from least to most responsive. These growth delay responses were then highlighted by HPV status and potentially impactful alterations (Fig. 1B). Neither HPV status nor mutation profile correlated with response to cetuximab. This preliminary work was used to select models with a potential mTORC-inhibitor targetable alteration or high intrinsic cetuximab resistance for further investigation. Five established HNSCC cell lines were subjected to the same targeted sequencing panel (Fig. 1C). Although no obvious impactful mutations were identified, UPCI-SCC90 was shown to have amplified PIK3CA and resulting pathway activation including high levels of pAKT_S473 and p-p70S6K_T389 (Fig. 1D), consistent with previous reports (35). This characterization enabled proper selection of model PDXs and cell lines for investigation of mTORC and EGFR inhibition.
mTORC inhibition had greater antiproliferative effect than EGFR inhibition on HNSCC cells
We evaluated the same panel of HNSCC-immortalized cell lines (Fig. 1C), including 3 HPV+ (UD-SCC2, UM-SCC47, and UPCI-SCC90) and 2 HPV− (UM-SCC1 and TU-138), for response to the dual mTORC1/2 inhibitor AZD8055, the mTORC/PI3K inhibitor NVP-BEZ-235, and cetuximab. Growth inhibition was assayed by an automated cell counting–based proliferation assay (Fig. 2A). Cell proliferation was assayed every 24 hours over 72 to 96 hours of growth in drug, the resulting growth curves revealed a slower growth rate with increasing drug concentration, but no loss of cell numbers, suggesting an antiproliferative rather than cell killing effect (Supplementary Fig. S2). The cell lines displayed a fairly uniform response to both AZD8055 (3-fold difference highest to lowest in IC50 value) and NVP-BEZ-235 (3-fold difference highest to lowest in IC50 value) as shown in Table 1. Cell lines identified to have PIK3CA amplification (UPCI-SCC90) or PTEN copy loss (UD-SCC2) had generally lower AZD8055 IC50 values, but did not reach statistical significance (P = 0.16). Differences in inhibition to the dual PI3K/mTORC inhibitor NVP-BEZ-235 were even more muted across the five cell lines (Fig. 2A). Response to cetuximab treatment was barely discernable in this assay: IC50 values could not be calculated for any of the cells lines with a sufficiently good fit (R2 > 0.75). Conversely, we were able to discern a response to cetuximab using a clonogenic expansion assay, leading us to employ this assay for all three therapeutics (Fig. 2B). We observed a similar level of response to both small-molecule inhibitors across the five cell lines, but this assay revealed differences in their ability to grow in the presence of cetuximab; the HPV+ cell lines demonstrated a dose-dependent response to this agent, whereas the HPV− cell lines, UM-SCC1 and Tu-138, presented more modest responses to cetuximab. Immunoblot analysis demonstrated that neither EGFR nor pEGFR expression predicted for cetuximab response (Fig. 1D).
|Cell line .||AZD8055 IC50 (nmol/L) .||95% CI .||NVP-BEZ-235 IC50 (nmol/L) .||95% CI .||Cetuximab IC50 (nmol/L) .||95% CI .|
|Cell line .||AZD8055 IC50 (nmol/L) .||95% CI .||NVP-BEZ-235 IC50 (nmol/L) .||95% CI .||Cetuximab IC50 (nmol/L) .||95% CI .|
Abbreviations: na, not applicable; nc, not calculated.
To investigate the mechanism of the inhibitors' effects, we used a BrdUrd uptake assay to show that cell division correlated with growth inhibition (Fig. 2C). For example, UMSCC-47 was among the more responsive cell lines to AZD8055 in the clonogenic expansion assay, at 10 nmol/L concentration about 10% of cells were able to proliferate, and concordantly the BrdUrd uptake was approximately 20% of control. Conversely at those same concentrations, TU-138 both had more BrdUrd uptake and less growth inhibition. We next investigated the induction of apoptosis using a Caspase 3/7 luciferase assay. Although caspase activity was elevated for some cell lines at high drug concentrations, little if any induction of apoptosis was observed for any of the three drugs (Fig. 2D) at concentrations near their IC50 values. These results demonstrate that both dual mTORC and mTORC/PI3K inhibitors had a greater antiproliferative effect on HNSCC cells than cetuximab and suggest that the mechanism is growth suppression rather than induction of cell death.
Combination of cetuximab with mTORC/PI3K inhibition effectively blocks signaling pathways but provides little additional growth inhibition
We next investigated whether the combination of the FDA-approved targeted therapy for HNSCC, cetuximab, with either AZD8055 or NVP-BEZ-235, showed any additive effects on cell proliferation. We selected the clonogenic expansion assay to perform this test as it was more effective in revealing differences in monotherapy. With the exceptions of TU-138 and UPCI-SCC90, there was little if any additional antiproliferative effect when 100 nmol/L cetuximab was added to either small-molecule agent (Fig. 3). Immunoblots confirmed that each inhibitor was effectively suppressing its molecular target. In a dose-dependent manner, both AZD8055 and NVP-BEZ-235 inactivated the mTORC2 target AKT and the mTORC1 target 4EBP1, as evidenced by decreased phosphorylated protein. Cetuximab inhibited phospho-EGFR and two downstream targets AKT and MAPK1, but had less effect on further members of the AKT/mTORC signaling arm as shown by relatively normal levels of p4EBP1 and pS6. Effective inactivation of downstream members of this signaling arm correlated with decreased cell survival (Fig. 3, immunoblot panels). The combination of the monoclonal antibody with a small-molecule inhibitor produced effective inactivation of both p4EBP1 and pMAPK1, but with the exception of TU-138 and UPCI-SCC90 produced little additional reduction in survival.
mTORC or PI3K plus EGFR knockdown reveals similar antiproliferative effects
In order to confirm the pharmacologic effects, we utilized a specific siRNA knockdown (KD) approach against mTOR, PIK3CA, and EGFR to inhibit these signaling pathways in the same HNSCC cell lines. Effective KD was confirmed by immunoblotting (Fig. 4). mTOR and PIK3CA were also evaluated in combination with EGFR KD. Functional effects of the KD were assessed by a growth inhibition assay. Although results varied across the different cell lines, KD of either mTOR or PIK3CA typically produced greater growth suppression than KD of EGFR. Combination KD of EGFR with either other protein had minimal additive effects, similar to the results with the pharmacologic inhibitors. Overall, the results of the siRNA experiments suggest that suppression of targets in the PI3K/AKT/mTORC arm of EGFR signaling is more effective at reducing HNSCC cell growth and survival than directly targeting EGFR.
EGFR/mTORC combination therapy evaluated in selected PDX models
Previous work has suggested important differences between in vitro and in vivo models in response to EGFR and PI3K inhibitors, where effects seen in one system may not be observed in the other (36, 37). To address these concerns and to expand our efforts to a more clinically relevant model system, we evaluated the combination of these therapeutic agents in five of the ten PDX models. PDXs were selected on the basis of PIK3CA-activating mutation (UWSCC-6 and 13) or for intrinsic cetuximab resistance (UWSCC-1, 17, and 64). Mice were treated with daily administration of the dual mTORC inhibitor AZD8055 by oral gavage and biweekly dosing of i.p. cetuximab or single-agent therapy for comparison for 2 or 3 weeks (Fig. 5A). Pharmacokinetic (PK) analysis performed in the initial characterization of AZD8055 demonstrated that dose we selected (20 mg/kg QD) produces free serum concentrations of drug of approximately 100 nmol/L, similar to the levels we tested in vitro (22). The combination therapy produced a significant growth delay in all five tumor models (repeated measures Friedman test, P value at least <0.05), whereas either agent given alone was often ineffective at slowing tumor growth, suggesting at least an additive effect of the combination therapy. Mouse masses were recorded during and following treatment, and no weight loss or other adverse effects were observed (Supplementary Fig. S3). Molecular target inhibition was confirmed by immunohistochemical analysis of FFPE tissues harvested 2 hours after initial treatment. Primary drug targets were inhibited as expected (Supplementary Fig. S4). The most robust inhibition was observed at downstream signaling indicators pMAPK1 and p4EBP1 which were effectively inhibited by cetuximab and AZD8055, respectively. Although for the majority of tumor models we had only sufficient mice for early posttreatment tumor harvesting, for UWSCC-64, we were able to sacrifice mice at the conclusion of 2 weeks of treatment. These FFPE tissues were analyzed by Ki-67 and cleaved caspase 3 IHC (Fig. 5B). Only the combination therapy produced a statistically significant reduction in Ki-67+ cells (one-way ANOVA, Tukey multiple comparisons test, P < 0.05). No significant difference between the four arms of the study was seen in cleaved-caspase staining, consistent with antiproliferative rather than proapoptotic mechanism of drug action. Overall, we observed a greater benefit of combining mTORC inhibition with anti-EGFR therapy in the in vivo PDX models than we did in cell lines.
In this study, we evaluated in vitro and in vivo models of HNSCC for response to mTORC inhibition both alone and in combination with cetuximab, to establish a preclinical basis for whether a catalytic mTORC inhibitor could be a potentially effective addition to cetuximab in HNSCC. Although both rapalogs and PI3K inhibitors are being evaluated in HNSCC, there has been only limited investigation of the particular combination of an ATP competitive mTORC small-molecule inhibitor with cetuximab. A key component of this work is the identification and validation of predictive biomarkers that would guide the selection of patients to this therapeutic combination. Although the goal of biomarker selection has been extensively investigated, our work here adds to the growing complexity and perhaps tenuous linkage between biomarker and treatment.
We began by evaluating a molecularly characterized panel of HPV+ and HPV− HNSCC cell lines for response to a dual mTORC1/2 (AZD8055) and a PI3K/mTORC (NVP-BEZ-235) inhibitor. Despite the presence of potentially impactful alterations such as PIK3CA amplification (UPCI-SCC90) or PTEN copy loss (UD-SCC2), the response to either inhibitor was fairly uniform with a less than 5-fold difference in IC50 values. Timecourse analysis of the growth inhibition assay used to generate the IC50 calculations demonstrated that increasing drug concentration only slowed cell growth and did not reduce the number of cells, even at micromolar concentrations. We further investigated the mechanism of inhibition of these compounds contrasting a DNA replication assay with a caspase activity–based apoptosis assay, and found little if any caspase activity. Conversely, changes in DNA replication were in the same direction and to the order of magnitude approximated functional growth inhibition, suggesting an antiproliferative rather than pro-death action of these drugs.
We next evaluated combination treatment using the clonogenic expansion assay due to better detection of cetuximab inhibition with this assay. Despite the observation that single and combined treatments inhibited the molecular targets as expected (i.e., AZD8055 reduced phosphorylated AKT and S6; cetuximab reduced pEGFR and pMAPK1), little additive benefit of the combination was apparent. Indeed, suppression along the PIK3CA/AKT/mTORC axis appeared to play the dominant role and was almost sufficient to inhibit growth and proliferation by itself. Specific siRNA KD of mTOR or PIK3CA, either solely or in combination with EGFR, mirrored these effects
Despite minimal additional benefit of combining an mTORC inhibitor with cetuximab in vitro, previous studies have shown that the response to cetuximab can be much more pronounced in vivo (38, 39). In our panel of HNSCC PDXs, we observed a range of responses to cetuximab treatment. Surprisingly, neither HPV status nor putative mutational or protein biomarkers predicted for response. We selected five of these models for an AZD8055/cetuximab combination study, two that contain well-described PIK3CA-activating mutations and three that presented high intrinsic resistance to cetuximab. The PDXs containing PIK3CA-activating mutations were chosen to mimic the approach used in the ongoing NCI-Molecular Analysis for Therapy Choice (NCI-MATCH-NCT02465060) study investigating precision oncology using a large sequencing-based “basket” trial. Combination therapy resulted in improved tumor control in all five models tested and slowed growth even in those models in which single-agent therapy showed no response. Unfortunately, anticipated predictive markers were not confirmed. UWSCC-6 containing the PIK3CA E545K mutation was highly responsive to both the mTORC monotherapy and combination therapy. After completing treatment, this tumor regrew rapidly, suggesting that this mutation was a major driving force in the growth of this tumor. Conversely, UWSCC-13, which was identified with a related PIK3CA helical domain mutation (E452K), did not respond to single-agent AZD8055, but was dramatically suppressed by the combination therapy, with durable growth delay at almost 30 days after end of treatment. Of the three PIK3CA WT tumors, two presented poor response to single agent and had only marginal improvement on combination therapy, whereas the third UWSCC-64 was among our best responders to the dual treatment. Although the combination therapy improved response in all five models, the level of improvement varied from model to model. Although the molecular targets of each treatment were inhibited in all five tumors, there did not appear to be a strong correlation between pretreatment levels of likely protein biomarkers and response.
Taken together, these results highlight the complexity in attempting to match therapy to molecular markers that predict for response, and that the presence of a single expected marker rarely functions as a binary switch. Although early reports in the targeting of the PI3K pathway suggested that such divergent effects could be observed, they perhaps are more the exception than the typical response. In the landmark article, Lui and colleagues found that an HNSCC patient–derived xenograft with the PI3KCA E545K mutation was supremely responsive to NVP-BEZ-235, whereas a WT model was completely nonresponsive (12). However, when evaluated in the context of hundreds of PDX models, PIK3CA or PTEN mutation alone was not predictive of a response to the PI3K inhibitor, BYL719 (29). Although PI3KCA mutation combined with WT PTEN was indicative of higher activity (2/3 responding), a large number of models were required to observe this effect. In the context of the limited number of patients typically evaluated in a phase I or II trial, such differences may be difficult to detect. Although one trial of BYL719 (NCT01219699) and several of BKM120 (NCT01570296, NCT01572727, NCT01833169) have utilized PI3K status as a stated inclusion or stratification criteria, none specific to HNSCC has yet been registered, potentially due to recruitment concerns.
In retrospective analysis of genomic markers in HNSCC clinical trials investigating PI3K inhibitors, the record has also not been compelling. The addition of the pan-PI3K isoform inhibitor PX-866 to cetuximab in patients with recurrent metastatic HNSCC did not confer any response benefit including 4 patients who had PIK3CA helical domain activating mutations (30). Similarly, the addition of the rapalog everolimus to erlotinib treatment did not induce a durable response in a patient with a PIK3CA H1047R mutation (40). Although larger trials with the genomic biomarker selection criteria designed as a critical component of the study, such as the NCI-MATCH, are likely to bear more successful results, in the more limited scale of a typical phase II trial, such conclusions will likely be challenging to draw. Similarly, researchers using PDXs and other clinical sample-derived models must strive to design studies with adequate number of well-characterized models that recapitulate the likely patient population as well as possible. Researchers should also seek to adhere to the reporting standards required in the publication of clinical trials, including releasing the information on all models tested and avoiding highlighting the exceptional responders as indicative of a typical outcome.
Although responses varied from model to model, a general trend observed in our work was the small magnitude of additional growth suppression when cetuximab was added to AZD8055 in the five established HNSCC cell lines tested. In contrast, in all five PDX models, the combination therapy was more effective than either treatment alone. Several factors may in part explain these differences: cell-autonomous effects resulting in greater suppression of downstream signaling, tumor–stroma crosstalk that is not recapitulated using standard in vitro culture approaches, or antitumor effects of antibody-dependent cellular cytotoxicity in vivo (even within an immune compromised model). Although the immunodeficient nude mouse is far from an ideal model for the study of antitumor immunity, these animals do produce natural killer cells (41) and have been used to study the role of ADCC using tumor-specific antibodies (42). We looked for tumor-infiltrating immune cells by morphologic analysis of H&E-stained slides and CD45 IHC in cetuximab and control-treated animals. This revealed few if any immune cells and no differences when comparing treatments groups. Although molecular target inhibition by drug was similar in both in vitro and in vivo models, we were limited in the timepoints after treatment for the PDXs that could be assayed. It is possible that signaling differences arose at later time scales that could explain the additive response in PDXs. More likely however is the more complex biology of the mouse model with interaction of the tumor with the stromal and vascular system of the mouse that contributed to this difference. Systemic administration of these therapeutics may well induce changes to the tumor microenvironment that affect tumor growth, but are completely absent from standard in vitro monoculture. We are currently developing both orthotopic and humanized mouse models which engraft a functional immune system for use with our PDX system, as well as microscale devices that allow for in vitro co-culture of tumor and stromal cells. These developments will enable a more robust investigation of immune and tumor microenvironment–mediated effects.
Although the PDX engrafted on an immunocompromised animal may represent a poor model for the study of immune antitumor biology, it does serve as a more direct approach for the investigation of tumor signaling biology, without the potential confounding effects of the immune response. Our results showed that although suppression of PI3K/AKT/mTORC signaling was largely sufficient to inhibit proliferation of HNSCC cells in vitro, it had only limited effect on xenograft tumors in vivo. However, suppression of both PI3K/AKT/mTORC signaling and RAS/RAF/MAPK1 signaling by the combination of AZD8055 with cetuximab resulted in dramatic growth suppression. Consistent with the function of these compounds, this effect was antiproliferative rather than proapoptotic and highlights the potential for combination with cytotoxic therapy such as radiotherapy or immunotherapy.
Considering the well-trod ground on which this work rests, the true clinical potential of this approach must be carefully considered. Two PI3K-specific inhibitors BKM120 and BYL719 are currently under clinical evaluation for HNSCC as both monotherapy and in combination with cetuximab. The mTORC rapalog everolimus is under similar study. However, several PI3K inhibitors have already been investigated in this setting with only limited clinical success, suggesting that targeting an alternative node in this signaling pathway may be worthy of further study. The catalytic mTORC inhibitors may well prove to be an attractive alternative, and although the drug used in this work, AZD8055, is no longer under active clinical investigation, its next-generation version AZD2014 (43, 44) is under study for multiple solid tumor types (NCT02599714, NCT02583542, and NCT02064608) and could be investigated in HNSCC. Although the ability to match therapy to a given biomarker profile remains challenging, the results of our study show that such a combination may have potential clinical benefit regardless of mutation or other alteration status.
The larger question facing targeted therapeutics using monoclonal antibodies (chimeric or fully humanized) or kinase inhibitors is their value in the context of the growing research and clinical success of checkpoint inhibitors and other immunotherapies. The approval of pembrolizumab and nivolumab for the treatment of recurrent/metastatic HNSCC in the 2016 represents the most dramatic improvement for treatment of these patients in almost a decade, and likely will come to form the standard-of-care treatment for many patients. However, although retrospective analysis of the Keynote-012 trial demonstrated that patients with high PD-L1 expression had greatly improved overall survival (∼10 months) compared with single-agent cytotoxic chemotherapy, those with low PD-L1 had only 5 months overall survival, roughly the same as cetuximab monotherapy in the recurrent/metastatic setting (10, 45, 46). Similarly, in the nivolumab phase III trial (CheckMate 141), PD-L1 low patients had only marginal improvement in survival relative to standard monotherapy options (9). Based on the subset analysis of these patients with low PD-L1 expression, this cohort may well have greater benefit of adding an ATP competitive mTORC inhibitor to a cetuximab course rather than receiving a checkpoint inhibitor. Taken together, the response rate in these trials was 18% to pembrolizumab (10) and 13% to nivolumab (9). Clearly, there is room for improvement. The data presented here suggest that the potential benefits of targeted therapy in the majority of HNSCC patients who do not benefit from immunotherapy remain promising.
Disclosure of Potential Conflicts of Interest
R.J. Kimple reports receiving commercial research grant support from Threshold Pharmaceuticals and Peloton Therapeutics. No potential conflicts of interest were disclosed by the other authors.
Conception and design: A.D. Swick, E. Sampene, R.J. Kimple
Development of methodology: A.D. Swick, A.M. Javaid, R.J. Kimple
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.D. Swick, P.J. Prabakaran, M.C. Miller, A.M. Javaid, M.M. Fisher, R.J. Kimple
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.D. Swick, P.J. Prabakaran, M.C. Miller, E. Sampene, I.M. Ong, R. Hu, R.J. Kimple
Writing, review, and/or revision of the manuscript: A.D. Swick, E. Sampene, I.M. Ong, M. Iida, J.Y. Bruce, D.L. Wheeler, R.J. Kimple
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.D. Swick, P.J. Prabakaran, K.P. Nickel, R.J. Kimple
Study supervision: R.J. Kimple
We thank Aasta Pandey, Grace Blitzer, Molly Smith, Andrew P. Stein, and Dana Gunderson for additional experimental assistance. We would like to thank Ella Ward and University of Wisconsin Research Pathology facility for tissue processing and histology services. We also would like to acknowledge the University of Wisconsin Translational Research Initiatives in Pathology laboratory, in part supported by the UW Department of Pathology and Laboratory Medicine and UWCCC grant P30 CA014520, for use of its facilities and services. The authors thank the University of Wisconsin Biotechnology Center DNA Sequencing Facility for providing the hotspot sequencing facilities and services.
This study was supported in part by R00 CA160639 (to R.J. Kimple), Department of Human Oncology Seed Grant (to R.J. Kimple), PhRMA Foundation Postdoctoral Fellowship in Translational Medicine and Therapeutics (to A.D. Swick), the University of Wisconsin Carbone Cancer Center Support Grant P30 CA014520, the UW School of Medicine and Public Health through the Wisconsin Partnership Program, and Wisconsin Head and Neck SPORE Grant (NIHP50 DE026787).
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.