Different testing algorithms and platforms for EGFR mutations and ALK rearrangements in advanced-stage lung adenocarcinoma exist. The multistep approach with single-gene assays has been challenged by more efficient next-generation sequencing (NGS) of a large number of gene alterations. The main criticism of the NGS approach is the detection of genomic alterations of uncertain significance.
To determine the best testing algorithm for patients with lung cancer in our clinical practice.
Two testing approaches for metastatic lung adenocarcinoma were offered between 2012–2015. One approach was reflex testing for an 8-gene panel composed of DNA Sanger sequencing for EGFR, KRAS, PIK3CA, and BRAF and fluorescence in situ hybridization for ALK, ROS1, MET, and RET. At the oncologist's request, a subset of tumors tested by the 8-gene panel was subjected to a 50-gene Ion AmpliSeq Cancer Panel.
Of 1200 non–small cell lung carcinomas (NSCLCs), 57 including 46 adenocarcinomas and NSCLCs, not otherwise specified; 7 squamous cell carcinomas (SCCs); and 4 large cell neuroendocrine carcinomas (LCNECs) were subjected to Ion AmpliSeq Cancer Panel. Ion AmpliSeq Cancer Panel detected 9 potentially actionable variants in 29 adenocarcinomas that were wild type by the 8-gene panel testing (9 of 29, 31.0%) in the following genes: ERBB2 (3 of 29, 10.3%), STK11 (2 of 29, 6.8%), PTEN (2 of 29, 6.8%), FBXW7 (1 of 29, 3.4%), and BRAF G469A (1 of 29, 3.4%). Four SCCs and 2 LCNECs showed investigational genomic alterations.
The NGS approach would result in the identification of a significant number of actionable gene alterations, increasing the therapeutic options for patients with advanced NSCLCs.
Therapeutic successes with EGFR–tyrosine kinase inhibitors (EGFR-TKIs) in EGFR-mutated advanced lung adenocarcinoma revolutionized the treatment of lung cancer.1–4 Shortly after the discovery of EGFR mutations, ALK rearrangements were reported in lung adenocarcinoma, and it was shown that tumors with ALK rearrangement show 75% response rate to ALK inhibitors.5–7 These 2 discoveries not only changed the treatment approach to lung cancer but also the pathologist's role in the management of patients with lung cancer. Subsequently, pathology practices experienced a shift from the morphologic assessment alone to a comprehensive histologic and genotypic characterization of lung cancers. Mok et al,3 in a phase 3 open-label study of gefitinib versus carboplatin-paclitaxel in advanced lung adenocarcinoma, clearly demonstrated that the patients in whom an EGFR mutation has been identified would benefit most from first-line therapy with gefitinib. The study provided a justification for EGFR-mutation testing before the initial treatment of lung adenocarcinoma. As a result, several practice guidelines for standardized EGFR and ALK testing protocols were published.8–12 The guidelines provided general considerations for testing such as patient population, sample requirements, minimum laboratory quality assurance requirements, reporting, and turnaround time. Most of the guidelines did not recommend specific methodology that should be used for mutation analysis, but assay sensitivity was taken into consideration, particularly for samples with low tumor content.10 Furthermore, it was felt that assays not used in the main clinical trials would require validation to ensure that they provide comparable results.8,10 With this in mind, simple, standardized techniques analyzing a single gene as validated in trials were considered optimal. Although many laboratories adopted testing for EGFR mutations and ALK rearrangements at the time of diagnosis of advanced lung adenocarcinoma, different testing algorithms were used. For example, KRAS and EGFR mutations are considered to be mutually exclusive, and as the first step in the testing algorithm, many laboratories performed a rapid and inexpensive KRAS mutation assay that would exclude KRAS-mutated tumors from further EGFR/ALK testing. However, in a case of a KRAS-negative tumor, EGFR and ALK testing was performed. This approach may be cost-effective and informative in 30% of lung adenocarcinomas with KRAS mutation, but for most lung adenocarcinomas, it decreased testing efficiency and prolonged turnaround time in delivering clinically relevant molecular results. In the meantime, advances in sequencing technology and bioinformatics support have led to the development of multiplex assays with high sensitivity, particularly polymerase chain reaction (PCR)–based massively parallel “next-generation sequencing” (NGS). The number of clinically relevant genomic alterations with already available or newly developed kinase inhibitors is rapidly increasing. Therefore, there is a need for comprehensive testing for a large number of genomic alterations that would provide clinically significant information within an acceptable time frame (usually within 2 weeks). The main critique of this approach is the detection of genomic alterations of uncertain clinical significance for which targeted therapies may not be available at the moment. However, additional concerns also remain including validation of this complex technology, cost, uniformity across laboratories, approaches to proficiency testing, and other quality assurance/quality issues. Our laboratory experienced the same dilemma: whether to continue with a multiple single-gene assay for a limited number of genes or introduce NGS as a test of choice. We report the results of our patient care quality improvement effort to determine the best testing options for patients with lung cancer in our clinical practice.
METHODS
Study Group
The Figure illustrates the testing algorithm for patients with non–small cell lung carcinoma (NSCLC) in our institution before 2015. Of 1200 samples from patients with NSCLCs, 57 were subjected to Ion AmpliSeq Cancer Panel (Ion Torrent, Life Technologies, Thermo Fisher Scientific, Waltham, Massachusetts) analysis from 2012 to 2015, including 46 cases of adenocarcinoma and NSCLC, not otherwise specified (NOS), that had been initially tested for the 8-gene panel for EGFR, KRAS, PIK3CA, and BRAF and fluorescence in situ hybridization (FISH) for ALK, RET, MET, and ROS1. Also, 7 squamous cell carcinoma (SCC) and 4 large cell neuroendocrine carcinoma (LCNEC) cases with unusual clinical features were among the samples subjected to Ion AmpliSeq Cancer Panel analysis. The following data were gathered from each case: clinical history, patient age, sex, specimen specifications (location and procedure type), tumor type, and molecular studies' results.
Molecular triage for non–small cell lung carcinomas at our institution before 2015. *Rare cases with detectable driver mutations by the 8-gene panel; unusual clinical features were also subjected to 50-gene Ion Ampliseq Cancer Panel (Ion Torrent, Life Technologies, Thermo Fisher Scientific, Waltham, Massachusetts). Abbreviation: FISH, fluorescence in situ hybridization.
Molecular triage for non–small cell lung carcinomas at our institution before 2015. *Rare cases with detectable driver mutations by the 8-gene panel; unusual clinical features were also subjected to 50-gene Ion Ampliseq Cancer Panel (Ion Torrent, Life Technologies, Thermo Fisher Scientific, Waltham, Massachusetts). Abbreviation: FISH, fluorescence in situ hybridization.
Single-Gene Assays
Next-Generation Sequencing
DNA sequencing was performed with the Ion AmpliSeq Cancer Panel according to the manufacturer's instructions. Briefly, 10 ng of DNA was amplified by PCR using the AmpliSeq Cancer Panel Primers pool and Ion AmpliSeq Master Mix v2.0. Multiplexed bar-coded libraries were enriched by clonal amplification, using emulsion PCR on Ion Sphere particles (Ion PGM Template OT2 200 Kit or Ion PI OT2 200 Kit v3), and loaded on an Ion 318 Chip or Ion P1 Chip (Thermo Fisher Scientific, Waltham, Massachusetts). Massively parallel sequencing was carried out on a Personal Genome Machine Sequencer or Ion Proton (Thermo Fisher Scientific). The raw signal data were analyzed with Torrent Suite (version 4.0.1) to generate BAM files after signal processing, base calling adapter trimming, and alignment to the reference human genome (hg19). Variants were called with Torrent Suite Variant Caller, which generated a list of detected sequence variations (single nucleotide variants and insertions or deletions [indels]) in a variant calling file. The American College of Medical Genetics and Genomics guidelines for germline variants were used for interpretation of variants of uncertain significance.15
The variant calls were annotated, filtered, and prioritized by using SeqReporter16 with the help of an in-house knowledge base and the following publically available databases: COSMIC v68 (http://cancer.sanger.ac.uk/cancergenome/projects/cosmic/; accessed December 2015), dbSNP build 137 (http://www.ncbi.nlm.nih.gov/SNP/; accessed December 2015), and in silico prediction scores (PolyPhen-2 and SIFT) from dbNSFP light v1.3.17 Sequence variants with at least ×300 depth of coverage and mutant allele frequency of greater than 5% of the total reads were included for analysis. Integrated Genomics Viewer (Broad Institute, Cambridge, Massachusetts) was used to review the sequence read pileups as appropriate.
RESULTS
Forty-six cases of adenocarcinoma and NSCLC, NOS were tested by non-NGS 8-gene panel (Sanger sequencing and FISH) and subsequently subjected to targeted hot-spot Ion AmpliSeq Cancer Panel. Of the 46 cases, 29 cases (29 of 46, 63%), including 25 adenocarcinomas (25 of 29, 86%) and 4 NSCLCs, NOS (4 of 29, 14%), were wild type and 17 (17 of 46, 37%) had genomic alterations by the non-NGS 8-gene panel. Their clinicopathologic characteristics are summarized in Table 1. Only 1 of 29 wild-type cases (1 of 29, 3.4%) showed no genomic alterations by hot-spot–targeted Ion AmpliSeq Cancer Panel. The remaining 28 cases showed a total of 53 genomic alterations (mean, 1.9; range, 1–5). The most commonly mutated gene was TP53 (22 of 29; 76%). Among the 29 cases, Ion AmpliSeq Cancer Panel detected 9 actionable and investigational variants (9 of 29, 31.0%) occurring in the following genes: ERBB2 (3 of 29, 10.3%), STK11 (2 of 29, 6.8%), PTEN (2 of 29, 6.8%), FBXW7 (1 of 29, 3.4%), and BRAF G469A (1 of 29, 3.4%) (Table 2). Overall, 41 variants of uncertain clinical significance were found.
Demographic and Clinical Characteristics of 46 Adenocarcinoma, 7 Squamous Cell Carcinoma, and 4 Large Cell Neuroendocrine Carcinoma Patients
Seventeen of 46 adenocarcinomas (17 of 46, 37%) with genomic alterations identified by non-NGS gene panel (10 EGFR and 7 KRAS mutations) were subjected to Ion AmpliSeq Cancer Panel at the treating physician's request to identify additional potentially actionable genomic alterations (Table 3). All 17 cases showed concordant results with Sanger sequencing. A total of 56 genomic alterations were identified (mean, 3.3; range, 1–6). The most commonly mutated gene was TP53 (11 of 17; 65%). EGFR-mutated tumors showed a total of 39 genomic alterations (mean, 3.9; range, 2–6). In addition to EGFR sensitivity mutations, 4 cases also showed investigational targetable mutations including PDFGRA, PIK3CA (2 cases), and CTNNB1, found in 1 case of adenocarcinoma with EGFR exon 19 mutation, 2 cases with EGFR exon 20, and 1 case with EGFR exon 21 mutation, respectively (Table 3). The KRAS-mutated adenocarcinomas showed a total of 17 genomic alterations (mean, 2.4; range, 2–2). Two cases showed investigational targets (VHL1, STK11).
Seventeen Adenocarcinomas With Genomic Alterations Identified by Both Non-NGS and NGS Gene Panels
Seven SCCs with unusual clinical features were subjected to non-NGS and Ion AmpliSeq Cancer Panel testing (Table 1). While wild type by the non-NGS method, they showed a total of 17 genomic alterations (mean, 2.4; range, 2–4) by Ion AmpliSeq Cancer Panel. Four cases showed investigational targets, summarized in Table 2. Thirteen variants of uncertain clinical significance were identified, and as expected, the most common mutations were in the gene TP53 (7 of 17; 41.1%).
Also, 4 LCNECs with unusual clinical features were subjected to both non-NGS and Ion AmpliSeq Cancer Panel testing and showed a total of 11 genomic alterations (mean, 2.7; range, 1–4), only detected by the Ion AmpliSeq Cancer Panel. The most common mutations were in the TP53 gene (4 of 11; 36%). Two LCNECs demonstrated investigational targets including RET substitution (16 p.M918T c.2753T>C) and FBXW7 mutation (p.Q275X, c.823C>T).
DISCUSSION
In this study, we demonstrated that in advanced lung adenocarcinoma and NSCLC, NOS cases deemed to be wild type by single-gene assays, testing for a larger gene panel by Ion AmpliSeq Cancer Panel would result in the identification of actionable gene alteration in nearly an additional one-third of patients. The major advantage of the NGS approach is to provide the oncologists information about potential therapeutic targets for patients who otherwise have limited systemic treatment options. Our results are similar to those reported by Drilon et al18 who also identified actionable genomic alterations by hybrid-based NGS in 26% of lung adenocarcinomas otherwise negative for such alterations by other non-NGS genomic testing approaches. The NGS approach also reveals multiple gene alterations in a single tumor, although co-occurrence of more than 1 oncogenic driver was not identified in our study population. According to the data from the Lung Cancer Mutation Consortium, of 1007 lung adenocarcinomas, co-occurrence of more than 1 oncogenic driver occurs in only 3% of patients.19 We are uncertain about the significance of some coexisting alterations in our study. However, increasing evidence suggests coexistent gene mutations may alter the extent and duration of response to targeted agents.
All practice guidelines are focused on testing in lung adenocarcinoma because targetable mutations such as mutations in EGFR gene are very infrequent in well-characterized SCCs and LCNECs.20–22 We did not identify any EGFR mutations in the SCCs or LCNECs included in our study. However, both tumor types showed potential therapeutic targets such as mutations in MET and RET genes. These observations support the idea to expand the NGS testing beyond adenocarcinoma and to include other histologic subtypes such as SCC and even LCNEC, particularly for patients with unusual clinical characteristics such as rapid course of progression, younger age, and no history of tobacco smoking. This is in contrast to the study by Ranganchari et al23 that showed only 1 driver mutation in SCC (FGFR1), information that did not change the clinical treatment decision. Therefore, the authors concluded that NGS testing in SCC is less likely to alter current practice.
From a laboratory standpoint, NGS provides a few advantages over single-gene assays. Ion AmpliSeq Cancer Panel and similar NGS assays require as little as 10 ng of DNA, which can be obtained from minimal tumor tissue on core needle biopsies or aspiration material. This similar amount of DNA is needed for many single-gene assays. Nevertheless, the small size of specimens represents a challenge for an adequate sequential approach to biomarker testing, and NGS almost completely eliminates this problem by allowing the simultaneous, multiplex analysis of genomic alterations in numerous genes. Turnaround time is one of the key measurements of laboratory performance, and its significance has been strongly emphasized by the College of American Pathologists/International Association for the Study of Lung Cancer/Association for Molecular Pathology guidelines.10 Although the NGS assay may need a slightly longer time for processing, in the end, it would save time by eliminating a 2-step algorithmic approach after negative single-gene assays, as was the practice in our laboratory. In our experience, average turnaround time for the 50-gene Ion AmpliSeq Cancer Panel is within 10 working days of receiving the specimen in the molecular laboratory, as recommended by the guideline.10 In regard to FISH, the expected turnaround time in our laboratory for 90% of the cases is a maximum of 5 working days. Our practice is to perform reflex molecular testing at the time of diagnosis in all advanced-stage lung adenocarcinomas. We strongly believe that the reflex molecular testing requested by a case pathologist at the time of histopathologic diagnosis is a very important practice element that significantly contributes to clinically acceptable turnaround time.
There are a few shortcomings in our study. First, we did not estimate the cost difference between the Ion AmpliSeq Cancer Panel NGS and non-NGS assays. There are essentially 2 approaches to cost analysis. One is from the laboratory standpoint alone. The NGS cost has declined significantly in recent years and is considered to be less expensive when compared to Sanger sequencing or other single-gene assays owing to the reduced time, manpower, and reagents' cost. The other approach is a comparison of the cost-effectiveness of molecular testing and treatment. Recent studies24,25 showed that EGFR testing alone in combination with gefitinib treatment is cost-effective when compared to chemotherapy and no molecular testing. Furthermore, Romanus et al26 showed the cost-effectiveness of multiplex predictive biomarker testing in non–small cell lung cancer.
The second potential shortcoming is lack of survival analysis. Takeda et al27 clearly demonstrated survival benefits for patients treated with targeted therapies established on the results of amplicon-based NGS. The follow-up in our series would be very short, and most patients were diagnosed and tested at clinical stage-4 disease. It would be interesting to know if treatments based on our NGS data had any impact on the quality of life.
In summary, we understand that approaches to molecular testing in lung cancer vary greatly between different institutions and laboratories and that our study reflects a single institution experience. Although many logistical and administrative obstacles limit the use of NGS, our data further provide the support for comprehensive genomic profiling of NSCLC in routine clinical practice. More clinical studies are necessary to determine effectiveness of multiplex testing in other histologic types of lung carcinoma such as SCC. Overall, our findings support first-line testing of advanced NSCLCs, using NGS as a more comprehensive and efficient strategy as compared to single-gene multistep testing.