Challenges in Amplicon-Based DNA NGS Identification of MET Exon 14 Skipping Events in Non-Small Cell Lung Cancers

Abstract

Introduction: MET Exon 14 skipping alterations are drivers of non-small cell lung carcinoma (NSCLC) with responses to tyrosine kinase inhibitors. Amplicon-based DNA NGS assays (DNA NGSs) for the detection of METex14 skipping can yield false-negative results. We examined the efficacy of METex14 skipping with a DNA NGS and reflex RNA-based NGS (RNA NGS) strategy. Materials and Methods: Clinical cases with definitive or suspected lung adenocarcinoma (LungCa), lacking driver mutations with targeted DNA NGS, underwent the RNA NGS to identify oncogenic drivers. Samples with METex14 skipping identified on reflex RNA NGSs were confirmed with Sanger sequencing. Results: METex14 skipping events were detected in 22/762 (2.9%) samples by DNA NGS. RNA NGS identified 10 additional samples, for an overall frequency of 32/762 (4.1%). All 22 METex14 DNA variants affected the donor splice site. Sanger sequencing revealed that missed METex14 variants were largely deletions spanning the ~30 bp intronic region upstream of the splice acceptor site. The failure of DNA NGS to detect all METex14 mutants was due to a lack of coverage of the 3′ splice acceptor site of intron 13, branch point, and polypyrimidine tract. Conclusions: Modification of amplicon-based DNA hotspot assays, with primers that cover both donor and acceptor splice sites, can identify a larger number of METex14 skipping events. Clinical data show that patients with advanced NSCLC and METex14 variants responded to targeted therapy.

1. Introduction

Exon 14 skipping mutations in the mesenchymal-epithelial transition (MET) gene are oncogenic drivers in non-small cell lung carcinoma (NSCLC). These mutations are present in approximately 2–5% of NSCLC and have a higher proportional prevalence within adenocarcinoma, adenosquamous cell carcinoma, and sarcomatoid carcinoma subtypes [1,2,3,4,5]. The MET proto-oncogene encodes the hepatic growth factor receptor (HGFR/c-MET), a receptor tyrosine kinase that is produced primarily in cells of epithelial origin [1,6]. Downstream signaling is activated through the phosphoinositide 3-kinase (PI3K) and RAS-RAF pathways, following binding to hepatocyte growth factor (HGF), the stromal ligand [3,6]. Pathologic activation of c-MET signaling has been shown to occur through somatic alterations leading to loss of exon 14, gene copy-number amplification, and occasionally as MET gene fusions [7], resulting in increased cell proliferation, survival, invasion, and metastasis, contributing to tumorigenesis and growth [1]. Among these alterations, NSCLC with MET exon 14 skipping and high-level MET amplification events have been successfully treated with tyrosine kinase inhibitors.

Somatic alterations in splice sites leading to a loss of exon 14 transcription or exon 14 skipping in MET occur through mechanisms including point mutations, insertions or deletions, and large-scale whole-exon deletions that spatially disrupt splicing sites at the donor or acceptor site flanking the exon 14 of the MET gene [1]. This leads to the deletion of the MET juxtamembrane domain, which contains the binding site (Y1003) for the E3 ubiquitin ligase CBL. Single nucleotide variants within exon 14, at Y1003 or D1010, can also result in decreased degradation of the c-MET protein [2,3]. As a result of such alterations, impaired c-MET ubiquitination leads to decreased turnover and a subsequent increase in signaling [1,6,7].

ATP-competitive, small-molecule tyrosine kinase inhibitors that target c-MET have demonstrated clinical efficacy and safety in the treatment of patients with NSCLC with MET exon 14 (METex14) skipping mutations [1]. These include nonselective type 1a inhibitors, such as crizotinib and selective type 1b inhibitors such as capmatinib and tepotinib [1,8,9], all of which are FDA-approved or have a breakthrough designation for use in the treatment of metastatic NSCLC with METex14 skipping. The selective inhibitors tend to have excellent response rates with tolerable toxicity [1,8,9].

With the promise of targeted therapy, there is a need for highly sensitive assays that interrogate the array of mutations that lead to the skipping of MET exon 14. Molecular methods of METex14 detection include targeted sequencing of DNA or RNA. Studies have shown that DNA-based assays (DNA NGSs) alone that utilize amplicon-based primers and probes may be sub-optimal when compared to RNA-based NGS assays (RNA NGS) focused on detecting the fusion of exons 13 and 15 of the MET gene—an alteration that results from METex14 skipping [4,10,11]. NGS panels have been historically designed to sequence the DNA of all genomic variants surrounding the 5′ donor site of exon 14 that leads to an alteration or ablation at the splicing site or the deletion of the entire exon [4,12,13].

We compared the performance of two commercially available targeted amplicon-based DNA- and RNA-based NGS assays for the detection of MET-ex14 variants by reviewing NGS-based profiling results of patients with lung adenocarcinomas or when lung adenocarcinoma was in the differential (LungCa). In addition, we further characterized the clinicopathological characteristics of patients harboring METex14 skipping mutations.

2. Materials and Methods

2.1. Patients

The study was approved by the Institutional Review Board (IRB) at Columbia University Irving Medical Center. The research was conducted in compliance with the U.S. Common Rule. Over a 26-month period (5/2018–7/2020), DNA-based NGS profiling of clinical cases with adenocarcinomas of definitive or NSCLCs of suspected lung origin with our institutional targeted DNA hotspot Lung Cancer Panel was conducted on 762 samples, as shown in Figure 1A. Clinically relevant variants were identified in 587 (77)% of the lung tumors, and 31% (N = 182) of these cases were actionable alterations with FDA-approved therapeutic potential. Per the institutional standard molecular testing algorithm, cases without driver mutations on the DNA hotspot panel, including hotspot variants in KRAS, EGFR, ERBB2, BRAF, and MET genes, were reflexed to RNA-based NGS analysis from the same tissue with the RNA Fusion Panel [14]. Consequently, 175/762 (23%) were interrogated for RNA fusions. These 175 specimens that did not harbor DNA driver mutations were part of the clinical cases submitted for molecular analysis. The samples came from patients with diagnosed or suspected lung cancers. A clinical chart review was conducted with a censoring date of 2/24/2022.

2.2. Specimen Types and Cytomorphological Assessment

Formalin-fixed paraffin-embedded (FFPE) surgical pathology (biopsies and resections) and cytopathology (fine needle aspirations and effusions) were evaluated. For each case, the original diagnosis was confirmed using the 2015 WHO Classification by a thoracic pathologist. The predominant architectural pattern and cytomorphological features for LungCa were noted.

2.3. DNA NGS

DNA extraction from FFPE tissue was performed following the manufacturer’s protocol using the DNeasy Tissue kit on a QIAcube (QIAGEN, Hilden, Germany). Subsequently, the extracted DNA underwent testing with one of two custom-targeted multigene Solid Tumor Panels. From 5/2018 to 11/2019, the extracted DNA underwent library preparation and targeted NGS with the TruSeq Amplicon-Cancer Panel on an Illumina MiSeq. Variant analysis was performed with NextGENe software 2.4.1 (SoftGenetics, LLC, State College, PA, USA). Among other relevant genes, the panel is designed to identify mutations/variants within MET (NM_000245.4) exons 2, 14, 16, and 19. MET exon 14 is covered by the following genomic coordinates Chr7: 116411958–116412127 (GRCh37/hg19) on the panel. From December 2019 onwards, a new multigene Solid Tumor Panel was utilized at our institution. This panel was based on Stem-Loop Inhibition Mediated Amplification (SLIMamp^®, Illumina, San Diego, CA, USA) technology (Pillar Biosciences, MA, USA) with multiplexed sequencing performed on the Illumina MiSeq. While both strategies employ amplicon-based target enrichment, SLIMamp enables amplification of overlapping or tiled amplicons in a single multiplex PCR reaction, whereas other PCR-based target enrichment methods often require separate PCR reactions and amplicon pooling. In SLIMamp, a stem-loop structure is formed by the overlapping amplicon, preventing the annealing of primers and thereby suppressing additional amplification [15]. The novel technology performs well for the detection of variants in solid tumors [16]. Variant analysis was performed with NextGENe software 2.4.1 (SoftGenetics, LLC, State College, PA, USA). The NextGene software was set to analyze samples with a minimum sequence coverage of 500X; variants with a mutation frequency of 3% and above; and insertion and/or deletions (indel) present in at least 5% of the reads [16]. The 50 gene panel is designed to identify mutations within MET (NM_000245.4) exon 2, 14, 16, 18, or 19. On this panel, the MET exon 14 is covered by the following genomic coordinates Chr7: 116411984–116412078 (GRCh37/hg19). The performance of the two DNA panels for the detection of the MET exon 14 is very similar. The second panel was implemented since the TruSeq Amplicon-Cancer Panel was discontinued.

2.4. RNA NGS

RNA NGS analysis was performed using custom primers with Anchored Multiplex PCR technology (ArcherDx, Boulder, CO, USA). The RNA panel targets chromosomal breakpoints to identify known and novel fusions in lung, thyroid, brain, and gastrointestinal tumors. RNA NGS overcomes several limitations of DNA NGS for the detection of structural variants and novel fusion partners, enabling the direct observation of gene fusions and splicing junctions, with the potential to improve the detection of actionable targets, to develop more sensitive diagnostic tests, and to understand prognostic information. RNA was extracted using the ALL-Prep DNA/RNA FFPE kit (QIAGEN) followed by targeted sequencing with Archer FusionPlex reagents (ArcherDx, Boulder, CO, USA). cDNA libraries with molecular barcode adapters were prepared and sequencing was performed on a Illumina MiSeq platform. Identification of fusion transcripts, exons, and breakpoints in 18 genes was performed with Archer Analysis software version 5.2 as previously described [14].

2.5. Sanger Sequencing

Sanger sequencing was conducted on samples with negative results on DNA NGS but with METex14 skipping identified by RNA NGS. Of the 10 samples in this category, 7 had sufficient material for identification of the specific variant. Samples that were positive by DNA NGS were not subjected to RNA NGS or Sanger sequencing. Prior validation studies showed that all METex14 alterations harbored in tested DNA samples were identified by RNA NGS [12]. The following four sets of primers were used in an attempt to capture the coding sequence of the exon, splice donor and acceptor sites as well as the branch point (B; −29), and the polypyrimidine tract (PolyPyr; −9 to −24) in intron 13 with chromosomal coordinates according to the human genome build GRCh37, as in Pruis et al. [4], with gene annotation according to NM_000245.4. Primer sequences are as follows [4]:

• forward 5′-CGATGCAAGAGTACACACTCCT-3′;
• reverse 5′-ACAACCCACTGAGGTATATGTATAGGTATT-3′;
• forward 5′-TCGATTCTTGTGTGCTGTCTTATATGTAG-3′;
• reverse 5′-TGCATCGTAGCGAACTAATTCACT-3′;
• forward 5′-ATGTAGTCCATAAAACCCATGAGTTCTG-3′;
• reverse 5′-GGGCACTTACAAGCCTATCCAA-3′;
• forward 5′-GCTTGTAAGTGCCCGAAGTGTAA-3′;
• reverse 5′-TGCAAAACCAAAAATAAACAACAATGTCAC-3′.

3. Results

3.1. Clinical Characteristics

Over a 26-month period from May 2018 to July 2020, METex14 skipping events were detected in a total of 28 patients; 22/762 (2.9%) samples (from 21 patients) by DNA profiling and an additional 10 samples (from 7 patients) by RNA NGS on driver-negative cases (tumors lacking mutations in KRAS, EGFR, ERBB2 and BRAF) for an overall frequency of 32/762 (4.1%) (Figure 1B). Clinical characteristics and treatment of patients with METex14 skipping events are depicted in

Table 1. Clinical Characteristics of Patients with NSCLC with MET Exon 14 Skipping Alterations Identified by DNA or RNA NGS.

Clinical Characteristics	MET Exon 14 Skipping (N = 28)
Median age at diagnosis Years (range)	78 (57–89)
Gender (n = 28)
Female	14 (50%)
Male	14 (50%)
Smoking (n = 28)
Never smoker	13 (46%)
Ever smoker	15 (54%)
Stage at diagnosis (n = 24) *
I	11 (46%)
II	5 (21%)
III	1 (4%)
IV	7 (29%)

Data are from 28 patients with MET skipping detected by DNA (N = 21) and RNA (N = 7). * Staging information was unavailable for four patients. NSCLC: Non-small cell lung cancer.

. A total of 10 out of 32 (31.3%) METex14 skipping events identified by RNA NGS in accordance with our testing algorithm were missed by DNA analysis. The 10 samples came from seven patients, as two patients had more than one sequenced tumor that harbored MET ex14 skipping. One patient who had tested for three tumors was shown to have two molecularly unrelated tumors with different MET exon14 skipping mutations, as determined by Sanger sequencing. The third tumor in this patient was unavailable for sequencing. Only one of the two tumors from the second patient with multiple tumors was available for confirmation using Sanger sequencing. At the time of diagnosis, the median age of the patients with METex14 mutations was 77 years and 71% of them were females, compared to 71 years and 63% of females, respectively, for cases with other driver mutations.

3.2. Clinical Implications

Six of the twenty-eight patients with METex14 mutations detected by either DNA or RNA NGS and diagnosed with Stage IV cancer were treated at our institution, and five received targeted therapy—crizotinib (n = 3) or capmatinib (n = 2). Clinical data showed that three patients receiving crizotinib and one patient receiving capmatinib achieved a clinical response at the time of censoring. The second patient on capmatinib expired seven months after initiation of therapy (

Table 2. Targeted Therapy in Patients with Stage IV NSCLC.

Targeted Therapy Stage IV Patients (N = 6)	Time on Therapy (Years)	Clinical Response	Detected by DNA or RNA NGS
Crizotinib (n = 3)	3.45	Yes	DNA
	2.01	Yes	DNA
	1.98	Yes	RNA
Capmatinib (n = 2)	1.50	Yes	RNA
	Expired after 7 months	NA	DNA
No Therapy (n = 1)		NA	RNA

Data shown is for 6 patients.

). Of note, our RNA validation data showed that samples positive for METex14 by DNA NGS were also positive by RNA NGS.

3.3. Histological Characteristics

The 32 samples from 28 patients comprised 28 surgical pathology (11 biopsy, 17 resection) and 4 cytology specimens. A morphological analysis of available slides from 30 of the 32 samples with METex14 skipping events detected by DNA and/or RNA NGS showed that the majority of samples 27/30 (90%) were adenocarcinoma, two were adenosquamous carcinoma and one tumor was a combined large cell neuroendocrine/adenocarcinoma. Notably, 17% (5/30) showed spindle/pleomorphic/tumor giant cells and 67% (20/30) had intranuclear inclusions (Figure 2A,B). Immunohistochemistry revealed that all specimens were positive for cytokeratin 7 and showed strong or focal positivity for TTF-1 and Napsin A, supporting the diagnosis of a lung primary (Figure 2C).

3.4. Confirmatory Sequencing

Examination of METex14 sequence coverage of the two DNA NGS panels revealed a potential for the DNA assay to miss METex14 mutants (also referred to as METex14 variants) due to lack of coverage at the 3′ splice acceptor site of intron 13, branch point, and polypyrimidine tract. The 5′ splice donor site of intron 14 showed adequate sequence coverage by both panels (Figure 3). All 22 METex14 DNA variants discovered by DNA NGS in our laboratory were found on the 3′ end of the gene, including the 5′ splice donor site of intron 14, as these were covered by the DNA NGS panel (Figure 4). Of these, the majority affected the canonical splice site, as described by Yu et al. [17], as four cases harbored MET c.3028G > A, four cases harbored c.3028+1G > A/T/C, and four cases harbored c.3028+2T > A/C/G. Five cases harbored canonical splicing site deletions. Similar to Yu et al. [17], all five cases affecting the noncanonical splicing donor site were c.3028+3A > G/T. On the other hand, 6 of the 10 METex14 variants that were missed by DNA NGS were deletions found upstream of the splice acceptor site of intron 13, as determined by Sanger Sequencing (Figure 3B). One sample showed a wildtype (WT) sequence. Interestingly, one case with three tumor foci harbored two different METex14 skipping variants in samples #2 and #3. On further examination, these tumors appeared to be independent primaries that were histologically similar but molecularly unrelated. We were unable to sequence the third biopsy (#4).

4. Discussion

This study investigated the detection of METex14 skipping by current amplicon-based DNA NGS assays and showed that interrogation of DNA alone is often insufficient for the identification of all events. Our results show that a reflex DNA and RNA NGS algorithm detected MET exon14 skipping events in 4.1% (32/762) cases versus 2.9% (22/762) when only DNA NGS was utilized. Notably, 31.3% (10/32) of the total number of METex14 skipping events identified by RNA NGS were missed by DNA analysis, highlighting the importance of using RNA NGS.

4.1. DNA NGS vs. RNA NGS

The importance of selecting the most appropriate technology for the detection of all METex14 variants in DNA- and RNA-based assays has been addressed [3,9]. Hybrid capture-based NGS panels with a probe design that targets regions of interest in DNA, coupled with bioinformatics pipelines to detect large deletions, have proven to be a more reliable method for identifying METex14 compared with amplicon-based methods when using DNA as the input material [2,3,9]. Amplicon-mediated target enrichment for DNA-based detection of MET exon 14 skipping events detects approximately 63% of known MET exon 14 skipping variants [18]. On the other hand, RNA-based assays have been more successful at identifying these variants, primarily because the spectrum of variants that result in METex14 skipping varies considerably in size and are located in multiple areas surrounding MET exon 14. About one-third of the mutations occur between exons 13 and 14, in the area known as the acceptor site of exon 14, and two-thirds occur between exons 14 and 15, in the area known as the donor site [12,19] (Figure 3). In addition, MET exon 14 skipping can result from large deletions that can span not only all of exon 14 but large portions of the intronic sequence. These types of mutations would not be detectable using DNA NGS if the introns flanking MET exon 14 were not targeted at both donor and acceptor splice sites. Reflex testing, as described in this report, could have a limitation with the amount of nucleic acid obtained from paucicellular specimens, particularly those of cytology origin. In this regard, it should be noted that technological advances have generated optimal sequencing data with minimal input of nucleic acids. While performing both DNA NGS and RNA NGS simultaneously is more efficient, it may be cost-prohibitive. A reflex approach of DNA NGS followed by RNA NGS in cases without driver mutations mitigates costs. In fact, our own experience with this algorithm for more than three years has proved it to be reliable and cost-effective [14]. Ideally, all NSCLC specimens should undergo both DNA and RNA analysis whenever possible. NCCN guidelines for NSCLC [20] recommend RNA-based NGS (if not already performed) to maximize detection of fusion events in patients who lack identifiable driver oncogenes on DNA-based NGS analysis. The guidelines acknowledge improved detection of METex14 skipping by RNA NGS and highlight the superiority of RNA NGS for the detection of other gene fusions, such as those occurring in NTRK1/2/3, as well as RET, ROS, and ALK rearrangements, for which testing is recommended in NSCLC patients [20]. Studies have shown that METex14 skipping can co-occur with ROS1 translocations and MET amplifications, as well as mutations in KRAS and BRAF [21]. Although RET rearrangements do not typically overlap with other drivers, a few studies have shown that RET rearrangements can infrequently overlap with EGFR or KRAS mutations [22,23], underscoring the importance of a more comprehensive testing approach for NSCLC, which includes DNA and RNA profiling.

All 22 METex14 variants identified by the amplicon-based DNA hotspot panels in our study were single nucleotide variants (SNVs) that involved either the canonical or noncanonical donor (MET c.3082+3A) splice sites or deletions encompassing the splice donor site of intron 14. In contrast, DNA sequencing of 6 of 10 cases identified by RNA NGS analysis revealed that METex14 variants that were missed by DNA NGS were largely deletions spanning the ~30 bp intronic region upstream of the splice acceptor site. This observation further emphasizes the importance of interrogating both the splice donor and splice acceptor sites flanking METex14 and designing primers that interrogate at least 50 bp of the intronic sequences adjacent to both splice sites. Our study included one sample with METex14 skipping detected by RNA NGS alone. This sample was identified to have a wildtype sequence of the MET gene. It raises the possibility of mutations harbored in cis-acting elements, such as splicing enhancers or silencers, which can influence the recognition of these sites by spliceosomal components. Mutations that disrupt these elements or active cryptic splice sites can lead to aberrant splicing, causing intron retention or exon skipping [24], and maybe missed by Sanger sequencing.

4.2. Clinical and Cytomorphological Assessment

Clinical characteristics of patients with METex14 skipping mutations (n = 28 patients) include an equal number of males and females, 54% having a smoking history (ever-smoker status) and consistent with multiple prior studies, and an advanced age with an average age at diagnosis of 78 years [2.4.24]. Prior studies have varied in demonstrating a predominance of male [4] or female patients [24], and a few, including ours, have shown an association with smoking history [3].

Our morphological assessment of 30 samples revealed that the majority (n = 27) of LungCas were adenocarcinomas, two were adenosquamous carcinomas and one was a combined large cell neuroendocrine/adenocarcinoma. An in-depth pathology review of a subset of adenocarcinoma samples (n = 30) in our cohort revealed that acinar was the most commonly identified adenocarcinoma pattern (69%), consistent with reports in the literature [25,26]. Interestingly, the majority (67%) contained intranuclear inclusions. While no sarcomatoid carcinomas were evaluated, 5/30 had features morphological features associated with them.

4.3. Clinical Implications

A few years ago, crizotinib received the FDA’s breakthrough therapy designation for patients with MET exon 14 alterations having disease progression on or after platinum-based chemotherapy. More recently, the FDA-approved capmatinib and tepotinib in May 2020 and February 2021, respectively. Both capmatinib and tepotinib are indicated for metastatic NSCLC with METex14 skipping. In addition, amivantamab-vmjw, a bispecific antibody directed against epidermal growth factor (EGF) and MET receptors, that was FDA-approved for metastatic NSCLC with EGFR exon 20 insertion mutations has shown some efficacy in clinical trials (NCT02609776) for tumors with METex14 skipping events. Clinical trial data, along with multiple studies supporting the use of MET inhibitors in patients with metastatic METex14 NSCLC, have shown that patients with METex14 NSCLC have better outcomes when receiving a targeted therapy [27,28]. In this regard, 3/4 of patients in our study with Stage IV disease and METex14 identified by RNA alone, who received targeted therapy, experienced a clinical response.

4.4. Limitations

The likely limitation of the study, besides the sample size, is the potential lack of adequate samples to perform reflex RNA testing after DNA NGS. However, our experience with using this algorithm over the past 3–4 years has documented that, in NSCLC, the detection of clinically relevant variants increased from 77% to 87% [14]. The failure rate for the RNA NGS panel due to specimen adequacy is less than 3%. By adopting this process, we were able to increase the overall diagnostic yield with a minimal increase in turn-around time and a reduction in overall cost [14].

5. Conclusions

In conclusion, it is important that assays used to guide clinicians in selecting the optimal therapy detect all METex14 skipping mutations present at the donor and acceptor splice sites adjacent to the MET exon 14. For the widely used amplicon-based DNA NGS assays, the assay design should include primers and probes that target the 3′ acceptor site of the MET exon 14. Furthermore, clinicians are encouraged to consider the limitations and advantages of the available technologies guiding treatment decisions. NCCN Guidelines recommend RNA-based NGS (if not performed concurrently with DNA NGS analysis) for patients who do not have identifiable driver oncogenes on DNA testing (NCCN). As seen in our study, a reflex testing strategy comprised both DNA and RNA NGS on driver-negative samples can detect targetable METex14 events. While securing good quality RNA may be a limitation for analysis, the information obtained from RNA NGS is essential to providing functional information concerning intronic variants whose effect on splicing may otherwise be difficult to determine.