Heads Up! Predictive Gene Signatures in Head and Neck Cancer May Be Coming Soon Free

In this issue of Clinical Cancer Research, Bossi and colleagues (1) report that pretreatment tumor gene expression profiles (GEP) of recurrent/metastatic head and neck squamous cell carcinomas (HNSCC) from patients who experience prolonged response [median progression-free-survival (PFS) of 19 months] to treatment with palliative cetuximab-chemotherapy are characterized by basal subtype traits, and significantly differ from GEP of patients with short responses, thus highlighting the potential opportunity for the future development of a predictive gene signature to select HNSCC patients for cetuximab-chemotherapy. Cetuximab, a chimeric immunoglobulin G1 antibody to the Epidermal Growth Factor Receptor (EGFR) is the only “molecularly targeted” therapy in HNSCC. It was FDA approved in 2006 for locally advanced HNSCC in combination with radiation (2), or as monotherapy in the platinum-refractory, recurrent/metastatic setting. In 2011, cetuximab was FDA approved in combination with platinum-based chemotherapy for the treatment of recurrent/metastatic HNSCC, based on a 2.7-month overall survival (OS) benefit in the “EXTREME” phase III trial of first-line treatment of recurrent/metastatic HNSCC with the addition of cetuximab to cisplatin or carboplatin plus infusional fluorouracil (median OS, 10.1 vs. 7.4 months; HR, 0.8; P = 0.04; ref. 3). The addition of cetuximab to platinum plus fluorouracil resulted in an objective response rate of 36%, and median PFS of 5.6 months; however, only 5% of patients had a PFS over 12 months.

Unfortunately, since the approval of cetuximab, there has been modest progress over the past decade toward the development of more targeted therapies for HNSCC in the metastatic setting, and no additional targeted therapies have been FDA approved. The EXTREME regimen is still a benchmark for comparison in large phase III studies in recurrent/metastatic HNSCC, and cetuximab is offered to virtually all advanced HNSCC patients at some point in their care regardless of underlying tumor biology, or patient risk factors. This modest therapeutic progress in HNSCC can be attributed in part to the lack of validated prognostic and therapeutic biomarkers, and it remains the case that despite substantial effort in this area it is unclear which HNSCC patients are likely to benefit from cetuximab therapy. In the past decade, several prognostic gene expression–based signatures have been proposed in HNSCC; however, none have been prospectively validated or examined to predict response to cetuximab–platinum chemotherapy.

In this study, Bossi and colleagues (1) retrospectively compared GEPs of pretreatment archival formalin-fixed, paraffin-embedded tumor specimens obtained from 14 HNSCC patients with prolonged response (median PFS 19 months) and 26 HNSCC patients with short response (median PFS 3 months) to first-line palliative cetuximab-chemotherapy. Supervised class comparison analysis identified 509 differentially expressed genes and principal component analysis (PCA) distributed patients in two main clusters, distinguishing the prolonged responders from the short-term responders. Tumor tissue origin, either from the primary (n = 31) or recurrence or metastasis (n = 9), did not appear to confound the association of the gene signature with clinical outcomes.

The gene signature identified by Bossi and colleagues (1) was then compared with other datasets including: (i) a publically available dataset of 80 metastatic colorectal cancer patients who had received several lines of treatment followed by cetuximab monotherapy, and (ii) five separate microarray-based signatures associated with outcome in HNSCC patients who had not received cetuximab. The gene expression signature by Bossi and colleagues was able to classify metastatic colorectal cancer cases according to PFS, and among HNSCC patients, long cetuximab-chemotherapy-PFS correlated with high scores in 3 of 5 of the HNSCC prognostic signatures. Further molecular subtype analysis revealed that gene expression patterns derived from the long cetuximab-chemotherapy-PFS patients appeared to be correlated with previously described basal and hypoxia molecular subtypes of HNSCC (4, 5), and enriched for gene sets related to βCAT, E2F3, MYC, and p53. Additional drug sensitivity studies in this report showed that the GEP of long cetuximab-chemotherapy-PFS patients appeared to select for cell lines sensitive to EGFR tyrosine kinase inhibitors (afatinib, gefitinib, erlotinib, lapatinib), whereas GEP of short cetuximab-chemotherapy-PFS patients selected for gemcitabine sensitivity.

This is the first study to specifically utilize expression profiling to evaluate response/sensitivity to cetuximab–platinum chemotherapy in HNSCC patients, and revealed enrichment for previously identified molecular subtypes with prolonged response to anti-EGFR therapy. In contrast, further analysis of the expression data of short cetuximab-chemotherapy-PFS patients revealed enrichment for RAS activation (Fig. 1). Indeed, activation of the RAS–RAF–ERK pathway, as a result of activating mutations in KRAS, is a clearly established mechanism of resistance to cetuximab therapy in metastatic colorectal cancer.

This study highlights the utility of GEP to uncover underlying biologic differences and candidate predictive biomarkers when evaluating response to anti-EGFR and platinum therapy in HNSCC, as prior predictive biomarker studies utilizing less comprehensive methodologies evaluating only one or a few biomarkers in HNSCC, have not led to further validation of a biomarker. Although EGFR is overexpressed by IHC in the majority of HNSCC, multiple studies have failed to identify EGFR gene copy number or IHC as a predictive biomarker of response to cetuximab (6, 7). In other tumor types with high EGFR expression such as non–small cell lung cancer (NSCLC), studies evaluating EGFR IHC as a biomarker to predict response to cetuximab have also yielded variable results and cetuximab does not yet have an established role in NSCLC (8, 9). As absence of RAS mutation is a validated predictor of response to cetuximab-chemotherapy regimens in metastatic colorectal cancer (10), other biomarker studies in HNSCC patients treated with cetuximab have evaluated RAS pathway activation as a potential prognostic and/or predictive biomarker of cetuximab response. ERK1/2 expression was significantly associated with poorer OS and PFS in locally advanced patients treated with cetuximab-chemotherapy in a phase II trial (11). In addition, the germline mutation in the 3'-untranslated region of KRAS (rs61764370, KRAS-variant: TG/GG), which has been associated with drug resistance/sensitivity in various cancers, has also been associated with worse prognosis and cisplatin resistance in retrospective analysis of recurrent/metastatic HNSCC patients (12). Interestingly, KRAS-variant HNSCC patients appeared to have improved disease control with the addition of cetuximab to platinum chemotherapy in a small subset of patients, suggesting KRAS-variant mutation may be a predictive biomarker of treatment response.

Predictive gene expression signatures can be a powerful tool for precision medicine; however, few have been validated in prospective studies of any cancer type. The study by Bossi and colleagues is an important first step in the identification of candidate predictive biomarkers of response to cetuximab–platinum therapy in recurrent/metastatic HNSCC and efforts to prospectively validate this signature in HNSCC are underway and include a European phase II trial of cetuximab-chemotherapy in recurrent/metastatic HNSCC. Despite these efforts, it may be difficult to distinguish the relative contribution of predictors of cetuximab activity from EGFR targeting versus antibody-dependent cell-mediated cytotoxicity, and further, it may also be challenging to predict response to chemotherapy versus cetuximab. Recent genomic characterization and classification of molecular subtypes in HNSCC has improved our understanding of underlying disease biology and identified candidate prognostic biomarkers for the development of novel personalized biomarker driven therapeutic approaches for HNSCC. Taken together with findings from this study, there is suggestion that EGFR-directed therapies should be considered for patients with basal molecular subtypes and that patients with RAS activation may be more suitable for other molecularly targeted agents or chemotherapy.

No potential conflicts of interest were disclosed.

Conception and design: N.G. Chau, P.S. Hammerman

Writing, review, and/or revision of the manuscript: N.G. Chau, P.S. Hammerman