Toward Molecularly Driven Precision Medicine in Lung Adenocarcinoma

Lung cancer remains the leading cause of cancer-related death in the United States. Non–small cell lung cancer accounts for more than 80% of cases, and adenocarcinoma is the most common type and may be increasing in incidence. Historically, the treatment of advanced adenocarcinoma has consisted of cytotoxic chemotherapy. However, a better understanding of the genomics of lung cancer and the molecular pathways underlying oncogenesis has led to the development of therapies specifically targeting these molecular programs. This approach was first developed and validated in patients whose tumors contained gain-of-function mutations in epidermal growth factor receptor (EGFR), who were found to have increased tumor response and progression-free survival to EGFR tyrosine kinase inhibitors (1). Rearrangements of the ALK gene and the ROS1 gene have proven similarly susceptible to targeted therapies, resulting in FDA-approved inhibitors for all three pathways in routine clinical practice (2, 3). Because of the enhanced efficacy of these drugs in genetically determined patient populations as well as their relatively modest toxicity compared with cytotoxic chemotherapy, it is now standard of care to perform genetic testing for the presence of activating mutations or rearrangements in EGFR, ALK, or ROS1 prior to the initiation of systemic therapy.

In recent years, cancer gene panels using next-generation sequencing techniques have become cheaper and more commonly used in the clinical setting in hopes of discovering a targetable gene alteration or qualifying patients to enroll into clinical trials (4). However, the value of prospectively sequencing patient tumors using a broader gene panel outside of FDA-approved biomarkers with matched targeted therapies has not been robustly established, nor has a uniform framework for using such sequencing results to guide therapy been widely adopted across institutions.

In this issue of Cancer Discovery, Jordan and colleagues report on the experience of the first 860 patients with lung adenocarcinoma at one institution undergoing prospective next-generation sequencing of targeted gene panels of > 300 genes (5). They evaluated whether subsequent therapy was matched to genomic findings, and whether patients derived clinical benefit from this matched therapy—defined as stable or shrinking tumor with symptomatic improvement in two scans at least a month apart. Overall, 747 of 860 patients (86.9%) were found to have a “potentially actionable” somatic alteration, with 319 of 860 (37.9%) receiving therapy matched to their sequencing results, and 249 of 319 (78.1%) deriving clinical benefit as defined.

Several aspects of this study are notable. The clinical feasibility of prospectively using targeted sequencing of a broader panel of genes to guide therapy is addressed, with a mean of 17 days for sequencing results to become available after tumor and matched normal tissue were received. Of 860 patients, 765 (89%) had sequenced tumor tissue ≤1 year old, and 473 of 860 (55%) had sequenced tumor tissue ≤30 days, though it is not reported what percentage required rebiopsy of their tumor since the diagnostic biopsy.

In addition to demonstrating feasibility, Jordan and colleagues found that this approach provides clinical benefit. Using an innovative tiered database of molecular alterations (OncoKB; ref. 6), they stratified their analysis of genetic alterations based on available evidence and approval status of the matched biomarker, drug, and tumor type. For combinations that are level 1 (FDA-approved) and 2A (standard of care based on evidence and adoption into professional society guidelines such as MET exon 14 alterations, BRAF V600E mutations, RET fusions, and MET amplification in lung adenocarcinomas), a majority of patients (92% and 52%, respectively) received matched treatments, with a high proportion of clinical benefit for both groups (level 1: 80%–90%; level 2A: 50%–77%). For molecular alterations with lower levels of evidence (2B and below, e.g., ERBB2 amplifications), there was a much lower proportion of matched therapies (2%–25%), but a substantial proportion (36/69, 52%) received clinical benefit, suggesting that matched therapy for broad gene panel sequencing may provide clinical benefit for a substantial subset of patients.

This analysis illustrates a principled approach to the problem of clinical interpretation and actionability of broad gene panel sequencing (7). There are currently several publicly accessible curated molecular databases from multiple institutions (6, 8, 9); Jordan and colleagues provide an excellent demonstration of how such a database can be used clinically to generate matched treatments for molecular alterations. Further, in their cohort they find 239 of 860 (27.8%) patients with multiple actionable molecular alterations. A tiered list of molecular alterations with matched drug therapy may help determine choice of subsequent therapy, as well as suggest combination matched therapies, an approach that may be applicable across multiple settings.

They also demonstrate how this approach contributes to knowledge of clinical response for rarer mutations and identifies gaps in therapeutic options for patient subsets. For example, the size and structure of this cohort allowed the authors to examine the efficacy of EGFR inhibition in rarer EGFR mutations and confirmed decreased efficacy of first-generation EGFR inhibitors in patients with L861Q mutations and exon 18 deletions, suggesting that first-line therapy with second- and third-generation TKIs may be considered for these patients. In another example, 4 of 10 patients with ERBB2 mutations treated with matched therapy had a clinical response, suggesting that more rigorous evaluation may be warranted. Further, Jordan and colleagues identified a significant subset of patients without clinical trials available despite having standard-of-care biomarkers with matched therapies in different tumor types (e.g., inactivating mutations in BRCA1/2 or TSC1/2), suggesting potential opportunities for new clinical trials. Finally, they also performed an exploratory analysis of the 103 tumors without a targetable oncogenic driver and identified increased frequency of mutations in TP53, STK11, KEAP1, KMT2D, and PDGFRA. Further in silico analysis of the KEAP1 protein structure found clustering of clinically detected mutations around a binding interface to NRF2, suggesting an oncogenic mechanism as well as potential therapeutic options. Although these results require substantial additional evaluation for clinical application, this analysis nicely illustrates how broader genetic sequencing of clinical samples can generate clinically and biologically relevant hypotheses and provide direction for future research.

There are several limitations to this study. First, lung adenocarcinoma is the tumor type with the greatest number of validated targetable molecular alterations with existing therapies. Thus, the efficacy demonstrated here may not readily translate to other tumor types. Additionally, this study was conducted at a single academic medical institution with subspecialty physicians, clinical trial availability, and (as the authors note) a distinct patient population enriched in EGFR mutations, nonsmokers, and Asian patients. Thus, their approach may be challenging to replicate in different settings. Not addressed in this study are genomic predictors of response to immune checkpoint blockade, which is now FDA approved in all lines of therapy for lung cancer, produces a prolonged response in a subset of patients (10), and was received by a large subset of patients in their cohort. Finally, the authors' definition of clinical benefit is limited to symptomatic improvement with short-term radiographic evidence of at least stable disease, whereas considering progression-free survival or overall survival if available would have optimal clinical meaning. Nonetheless, this study by Jordan and coauthors represents an important case example of how broader gene panel testing can successfully be prospectively incorporated in a clinical setting with patient benefit and can pave the way for prospective randomized studies that leverage molecular testing in distinct patient populations.

Importantly, Jordan and coauthors demonstrate how careful clinical annotation combined with genomic sequencing results allows (i) rigorous evaluation of a program of prospective clinical sequencing; (ii) identification of patient subsets without therapeutic options despite targetable molecular markers; (iii) accumulation of data about clinical response for rarer mutations; and (iv) investigation into patient subsets without identified targetable oncogenic drivers to discover potential new targets, thus accelerating the development of drugs and refinement of biomarkers. It is critical that further carefully conducted studies of this type are carried out in different settings and for different tumor types to inform the clinical utility of broad molecular testing to guide therapeutic decision making.

E.M. Van Allen reports receiving commercial research grants from Novartis and BMS and speakers bureau honoraria from Illumina, has ownership interest in Syapse, and is a consultant/advisory board member for Genome Medical. No potential conflicts of interest were disclosed by the other authors.