Lung cancer is one of the leading causes of cancer-related deaths in the world. (1) In the current era of precision medicine, non-small cell lung cancer (NSCLC) has a wide assortment of targeted therapies linked to genomic biomarkers and signatures. (2, 3) Progress in genomic analysis using next-generation sequencing (NGS) has enabled the comprehensive detection of multiple targetable alterations simultaneously from limited tumor tissue. (4) Despite this, single-variant and -gene approaches are still employed as standard procedure in many settings to detect clinically relevant driver mutations resulting in tissue exhaustion and, potentially, incomplete tumor characterization. Apparent discrepancies between NGS and these conventional approaches, particularly regarding gene rearrangements and amplifications, however, make the interpretation of some of these alterations by NGS technologies challenging (5–8).

ALK gene fusions are important predictive biomarkers for targeted tyrosine kinase inhibitor (TKI) efficacy in NSCLC (9, 10). ALK rearrangements were initially assessed by fluorescent in situ hybridization (FISH); however, the approach is laborious, costly, and highly subjective (5, 6, 11). Additionally, because this methodology relies on the detection of merged or discrete signals from fluorescent probes, certain ALK gene fusions may be undetected or false-positive patterns could also be reported (12). Increasingly, immunohistochemistry (IHC) has been proposed as an improved and simplified alternative to FISH in the detection of ALK fusions (11, 13). However, both techniques are cell-based, microscopic approaches, making the process inherently subjective, with additional uncertainty arising from borderline or inconclusive cases. Moreover, these approaches are heavily affected by tumor heterogeneity, depending on the tissue sections used for analysis, particularly due to the scarce material usually obtained for lung cancer diagnosis. Consequently, FISH or IHC can be discordant and inconclusive for ALK fusions (6, 7, 14).

MET amplification represents a therapeutic target in NSCLC, though the clinical consensus and standardization of its definition are still subject to controversy (15, 16). Currently, TKIs of MET, such as crizotinib initially, and more recently capmatinib and tepotinib, have demonstrated clinical efficacy and safety in lung cancer patients with MET exon 14 skipping mutations (17, 18) and MET gene amplification (19). Due to the availability of targeted therapies, molecular testing of MET is now included in routine clinical workup in the NSCLC setting with comprehensive NGS panels including analysis for both exon 14 skipping mutations and MET amplifications (20). Reasons for equivocal outcomes for MET amplification may be attributed to a lack of clinically defined cutoffs for MET copy number or fold change (21–23). When using FISH, MET amplifications can be measured by comparing the ratio of MET copies to chromosome enumeration probe 7 (CEP7) (19, 23, 24). FISH also enables differentiation of focal amplifications and polysomy; however, the clinical significance for each of these patterns remains unclear. Regarding NGS, the assessment of MET amplification depends on the platform and remains to be validated in a broad clinical context including associations with response to MET inhibitors (25, 26).

The comprehensive nature of NGS has also facilitated the creation of novel cancer signatures, previously not possible with conventional techniques, such as tumor mutation burden (TMB) (20). TMB acts as a composite genomic score, often reported as mutations per exome or mutations per megabase (muts/Mb) of DNA sequenced, which highlights the overall mutational load of a tumor (27, 28). Clinical observations in NSCLC show that TMB-high (TMB-H) tumors are more responsive to immune checkpoint inhibition (29, 30). This observation has been confirmed in other solid tumor types and has led to FDA approval of pembrolizumab in TMB-H cases, making TMB a new pan-solid tumor biomarker for immunotherapy (31).

Given the multitude of clinically relevant genomic biomarkers present in NSCLC, we sought to analytically evaluate the level of concordance between conventional molecular pathology approaches and a DNA-based NGS assay. Furthermore, we aimed to elucidate any associations between concomitant driver mutations, co-alterations, and explore correlations to clinical outcomes in specific cases.

As a referral testing center for lung cancer biomarkers, Hospital del Mar has tested over 3,950 NSCLC samples from 2012 to 2017 for ALK, ROS1, and MET alterations (32, 33). Samples that presented sufficient residual material after diagnosis were deposited in the biobank collection of our Institution (MARBiobanc integrated in the Xarxa de Bancs de Tumors de Catalunya). Only initial diagnostic samples with remaining FFPE tissue from NSCLC patients were selected for this study. There were 12 samples with known ALK fusion events previously identified by FISH (Abbott Molecular, Des Plaines, IL, USA) reviewed to determine the tumor area and screened by IHC (D5F3, Ventana, Tucson, AZ, USA) and RNA-based NGS (Oncomine™ Focus Assay, Thermo Fisher Scientific, Waltham, MA, USA). Furthermore, 26 advanced NSCLC cases screened for MET amplification by FISH using the Cappuzzo score system (MET high gene copy number (GCN) defined as those with a mean of ≥5 copies per cell) (24) were recategorized into the currently accepted FISH criteria as the following: high-amplification (MET/CEP7 ratio ≥4.0), medium-amplification (MET/CEP7 ratio >2.2–<4.0), low-amplification (MET/CEP7 ratio ≥1.8–≤2.2), and non-amplification (MET/CEP7 ratio <1.8) (19) ( Supplementary Figure 1 ). Using these FISH criteria, NSCLC samples considered positive for the presence of MET amplification are those with a MET/CEP7 >2.2 (high- and medium-amplification), classifying low-amplified together with non-amplified as negative as no actionability was seen when using MET inhibitors (19). This project was approved by the local ethics committee (CEIC-PSMAR: 2015/6336/I), and all patients provided written informed consent.

All 38 samples were further evaluated by the Personal Genome Diagnostics (PGDx™) elio™ tissue complete, a research use only (RUO) comprehensive genomic profiling NGS assay (PGDx, Baltimore, MD, USA). This NGS panel assesses 505 genes using a minimum starting input of 50 ng (100 ng recommended) DNA from FFPE tumor tissue and identifies key genomic alterations (single-nucleotide variants (SNVs), insertion and deletions (indels), gene rearrangements, and amplifications) and complex genomic signatures including TMB, reported as muts/Mb (exome equivalent), and microsatellite instability status. ERBB2 amplifications were called when >2.5-fold change on normalized read depth was detected, and all other gene amplifications represented on the panel were called when >3-fold change was observed. Quality control metrics evaluated by PGDx elio tissue complete are shown in Supplementary Table 1 . Sample processing from FFPE tissue, including library preparation, hybrid capture, sequencing, and analysis was performed at PGDx. DNA from FFPE was extracted using the QIAamp® DNA FFPE Tissue Kit (QIAGEN, Hilden, Germany) according to the manufacturer’s protocol. Captured libraries were sequenced on the NextSeq® platform (Illumina, San Diego, CA, USA). Somatic variant identification was performed using the PGDx elio tissue complete bioinformatics software, which incorporates a machine learning-based variant calling algorithm to differentiate between true somatic mutations and artifacts or false positive signal (34). The resultant NGS calls were compared with available data from the orthogonal approaches and with patient clinical data. TMB and microsatellite status were also assessed by the PGDx elio tissue complete assay using previously described methods (35). Finally, cases with driver alterations were analyzed for concomitant aberrations and correlated with clinical features. The Mann–Whitney U test was used for continuous parameters. Statistical tests were conducted at the two-sided 0.05 alpha level of significance, carried out with SPSS Statistics software (SPSS Inc., Chicago, IL, USA).

The use of reliable screening methods for detection of molecular alterations in lung cancer is crucial for selecting the appropriate targeted treatment for eligible patients. Our study demonstrates an excellent concordance between all orthogonal methods and the PGDx elio tissue complete NGS assay for ALK rearrangements. Although IHC and FISH assays can detect ALK fusion events, these methods do not determine ALK fusion partners that may predict clinical benefit of selective TKI more accurately (14, 36). Diversity of fusions and how these may affect response to therapies makes comprehensive NGS analysis increasingly necessary for precision oncology.

EML4::ALK is a canonical structural variant with fusion of exon 13 of EML4 with exon 20 of ALK reported to occur in ~30% of NSCLC cases (37, 38). Interestingly, in studies evaluating the clinical consequence of other ALK fusions, many reports have seen no significant difference between EML4::ALK and non-EML4::ALK variants following treatment with targeted ALK inhibitors (39–41). Differences in clinical outcomes, however, have been reported within distinct EML4::ALK fusion variants, where patients with the canonical EML4::ALK (E13:A20) fusion respond better to crizotinib compared with other EML4::ALK fusion variants (41). More recent studies confirm the prognostic value of different variants, although second- and third-generation ALK inhibitors show better results compared with crizotinib regardless the variant (42). The IRF2BP2 gene identified as a new translocation partner fused with ALK exon 20 codifies for an IFN regulatory factor binding protein that interacts with IFN regulatory factor 2 (IRF‐2). The IRF2BP2 gene has been described as a fusion partner in acute promyelocytic leukemia when fused to RARA gene and in NSCLC is found to be fused to NTRK1 (43, 44). The patient with the IRF2BP2::ALK fusion (Pt. 11) received crizotinib as second-line treatment with lower than expected benefit, although progression was at the central nervous system, confirming the limitation of crizotinib at this anatomical site (10). The information provided by comprehensive NGS, evaluated in the context of clinical response to the different ALK inhibitors, may aid in the selection of the optimal treatment strategy for each patient.

In the detection of MET copy number status via FISH, the fluorescent signal of MET is compared with a CEP7 control and the resultant ratio is reported as a specific level of amplification (19). Through the PGDx elio tissue complete assay, the detected read density for MET is compared with the targeted regions overall through a bulk tumor analysis and interpreted bioinformatically in a more quantitative manner. The discrepancies in what constitutes a genuine MET amplification arise from the lack of clinically defined and meaningful cutoffs in assessing MET amplifications, thereby impeding the adoption of MET assessment in standard-of-care workups (15, 16, 45). After approval of MET targeting therapies, a cutoff of 10-fold change detected by NGS has been accepted as the threshold for prescription. However, this remains controversial as assays typically set cutoffs and thresholds for MET amplifications respective to their test and system, and therefore, may be problematic as results of MET positivity do not translate well across platforms (19, 46).

In our series, of 26 MET high GCN cases reclassified according to current FISH criteria, (19) only eight were considered MET amplified (two high-amplified and six medium-amplified). Analyzing by NGS with the PGDx elio tissue complete assay, we observed that five of these eight were also considered amplified, with only three discordant cases (discussed later). The other 22 MET high GCN cases analyzed in this study were divided by current FISH criteria into 4 low-amplified and 15 non-amplified. The agreement with the NGS panel in these 22 cases was 100% since MET amplification was not detected in any of them. Even so, the copy number values resulting from NGS of Pt. 29 and Pt. 39 stand out, with a MET copy number adjusted to the tumor of 2.6% and 37.8% of regions amplified, and Pt. 39, with a MET copy number of 4.2% and 15.6% of regions amplified, respectively. Both cases displayed copy number variation but did not reach the calling threshold to indicate an amplification by the PGDx elio tissue complete assay (≥3-fold copy number in at least 25% of the queried regions), which is consistent with heterogeneous copy number events such as MET gene gains or chromosome 7 polysomy (47–49). As opposed to gains or polysomy, amplification is thought to represent a state of true biologic selection for MET activation as an oncogenic driver (16). In published results of clinical trials in response to MET inhibitors, clinical benefit was only observed in patients with the so-called “true” amplification (18, 19, 50).

Consistent with this, the prevalence rate of MET amplifications in NSCLC can range from 1% to 5%, with estimates varying depending on the assay used and the fold amplification thresholds implemented. Notably, MET-amplified NSCLC patients have been observed to benefit from certain targeted inhibitors, such as crizotinib; however, the magnitude of benefit is below what is observed with other driver alterations, highlighting the relevance of defining and harmonizing the cutoff (51–55). Interestingly, from the two patients treated with MET inhibitors in our series, the one with the discordant results with PGDx elio tissue complete showed a poor (4-month) performance of TKI treatment compared with the patient with confirmed MET amplification by both techniques.

Whereas detecting rearrangement events via FISH and NGS shows high concordance, detecting copy numbers for genomic amplifications remains more ambiguous. Recent studies are addressing this clinical gap. In the study of capmatinib, patients whose tumors harbored MET GCN of at least 10 exhibited greater therapeutic benefit; however, the response rates were relatively low (40% for treatment naïve vs. 29% for previously treated) compared with patients with exon 14 skipping mutations in MET in the same scenarios (18). In that study, MET GCN alterations were defined by FISH and retrospectively assessed by NGS using the FoundationOne CDx panel. No correlation of both assessments has been reported to our knowledge. In another study with tepotinib, MET amplification assessed by Guardant 360^® technology with a cutoff of CN ≥2.5 in liquid biopsy in the absence of MET mutations, tepotinib showed encouraging results (ORR 71% in treatment naïve vs. 30% in second- and third-line settings) (25). These results might better reflect a more biologically relevant cutoff defined by NGS in this study.

Regarding the inverse relationship between TMB score and the presence of canonical driver alterations, it has been previously described that EGFR mutant lung cancers featured lower overall TMB as compared with EGFR wild‐type cancers (56). Also, potential clinical implications have been seen between EGFR-positive patients with low/high TMB when treated with TKIs (57). One limitation of our study is that we do not have sufficient statistical power to compare the parameters of response to treatment and overall survival because both ALK-fused and MET-amplified cases were pooled together in the driver alteration-detected group. However, this inverse association between high TMB and the presence of an ALK fusion was also observed previously, pointing out its potential interest when designing a precision therapy strategy for these patients (58, 59).

The discordant MET results observed in our study might derive from different technical aspects. The PGDx elio tissue complete assay considers factors such as tumor purity, percent of the sequenced genomic region amplified, and minimum coverage depths in determining whether the observed increase in copy number constitutes a genuine amplification event. Such metrics are not taken into consideration with FISH, instead relying on the subjectivity of the observer when deciding on which areas to assess in identifying copy number changes. Bias generated by FISH count due to tissue heterogeneity could potentially differ between tissue sections as reported to occur in the context of other biomarkers and indication, for example, HER2 assessment in breast cancer (60). Despite this, NGS approaches do not solve this biological phenomenon and MET focal amplifications could require further assessment by FISH, as these might not be captured by NGS if the majority of cells do not harbor the amplification. There is no specific information of clinical responses in these particular cases.

The PGDx elio tissue complete assay demonstrated high overall concordance with conventional diagnostic approaches for ALK gene rearrangements. Additionally, the DNA-based NGS approach provides far greater insight into the underlying genomic events with the ability to identify exact breakpoints of gene fusions and report the specific loci associated with gene amplifications. Moreover, it provides broader information regarding potentially relevant genomic co-alterations that may impact clinical outcomes. The optimal cutoff point for defining MET amplifications remains to be defined and might be specific to the NGS platform. Further studies with clinical outcome data are needed to further refine MET amplification as a predictive biomarker.

The original contributions presented in the study are publicly available. This data can be found here: https://ega-archive.org/datasets/EGAD50000000016. Accession number - EGA50000000010.

Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements. Written informed consent from the patients/participants or patients/participants’ legal guardian/next of kin was not required to participate in this study in accordance with the national legislation and the institutional requirements.

SC, JJ, MS, EK, EW, DN, JS, BB, and EA contributed to conception and design of the study. JJ, JK, KG, EV, EK, EW, JH, GC, DN, and JS organized the database. JJ, EK, and JG performed the statistical analysis. SC and EK wrote the first draft of the manuscript. JJ, MS, JK, KG, EV, LP, BB, and EA wrote sections of the manuscript. MH-W, PR, XR, and ET participate in image acquisition and technical support in immunohistochemistry, fluorescence in situ hybridization and next-generation sequencing. All authors contributed to the article and approved the submitted version.