Tag-based next generation sequencing: a feasible and reliable assay for EGFR T790M mutation detection in circulating tumor DNA of non small cell lung cancer patients

The current standard work-up of Non Small Cell Lung Cancer (NSCLC) patients includes the search for sensitizing mutations of EGFR gene (Sharma et al. 2007; Riely et al. 2006; Rosell et al. 2010; Mok et al. 2009) that allowed identification of patients eligible for treatment with an EGFR tyrosine kinase inhibitor (TKI) (Singh & Jadhav 2018). Most patients respond to first and second-generation EGFR TKIs, such as gefitinib, erlotinib and afatinib, but acquired resistance is likely to occur, leading to disease progression. EGFR T790M substitution has been indicated as the prevalent molecular event involved and occurs in approximately 50–60% of the cases developing TKI resistance (Yu et al. 2013; Hata et al. 2013; Sequist et al. 2011; Oxnard et al. 2011; Cross et al. 2014). Osimertinib is a third-generation EGFR TKI, designed to overcome resistance due to T790M and representing the current standard treatment for advanced, T790M-positive NSCLC patients progressing after first or second- generation EGFR TKI (Cross et al. 2014; Ramalingam et al. 2018). However, more recently the U.S. Food and Drug Administration (FDA) has approved the use of osimertinib also in first line for advanced NSCLC harboring common EGFR mutations (Mok et al. 2017).

Although T790M can be identified through a new biopsy of the progressing neoplasm, this procedure may be challenging as well as stressful for the patient, and could potentially lead to complications. Several studies have demonstrated the feasibility of assessing EGFR mutational status on circulating cell-free DNA (cfDNA) from plasma (Douillard et al. 2014; Sorensen et al. 2014; Sundaresan et al. 2016; Vanni et al. 2015). The cfDNA is becoming a reliable alternative source to tumor DNA, although the sensitivity of methods using cfDNA is generally lower (Ramalingam et al. 2018; Vanni et al. 2015; Luo et al. 2014; Oxnard et al. 2016). This approach is non-invasive, does not pose limitations to repeated sampling, and provides a sufficiently accurate assessment of intra and inter-tumor heterogeneity (Sundaresan et al. 2016; Murtaza et al. 2013; Diaz and Bardelli, 2014). Because circulating cell-free tumor-derived DNA (ctDNA) is diluted out with normal DNA, ctDNA analysis is technically challenging requiring both sensitivity and accuracy (Murtaza et al. 2013). The current methods for the detection of plasma T790M in clinical practice include digital PCR (dPCR) techniques, Real Time PCR assays and Next Generation Sequencing (NGS) (Thress et al., 2015a; Bartels et al. 2017; Kim et al. 2013; Mayo-de-las-Casas et al. 2017). Variable T790M detection rates have been reported ranging between 31 and 66% for BEAMing (beads, emulsion, amplification and magnetics) digital PCR; 18–52% for droplet digital PCR (ddPCR) and 22–30% for common Real Time PCR assays (Luo et al. 2014; Mayo-de-las-Casas et al. 2017). Improving the reliability of T790M detection in cfDNA would represent a significant achievement, as it would permit the access to an effective therapeutic agent to a larger number of patients in the absence of repeated tissue biopsies. Here, we have studied a commercial NGS panel using molecular tagging of the DNA alleles present in the plasma, and compared the results obtained with those by Real Time PCR.

Validation of EGFR T790M mutations was performed by QX200 ddPCR™ System (Bio-Rad Laboratories, Inc., Hercules, CA, USA) using ddPCR Mutation Detection Assays (FAM-Mutation assay: dHsaCP2000019 and HEX-wild type assay: dHsaCP2000020). For each PCR reaction 10 ng of FFPE DNA or 5–10 μl of cfDNA obtained by QIAamp (Qiagen) was amplified according to the ddPCR mutation protocol (Oxnard et al. 2014). Each PCR run, including samples (FFPE DNA and/or cfDNA in 2–4 replicates to screen at least 3.000 genomes per case) and controls (wild type, T790M-positive and no template controls) was analyzed using QuantaSoft analytical software package (Bio-Rad Laboratories). Threshold was determined according to the signals of no template, wild-type DNA and T790M-positive control. The allele fraction for each sample was calculated as merged of replicates, and the positive sample was called when at least 3 FAM-positive droplets were detected. The limit of detection (LOD) of the ddPCR T790M assay was initially determined and achieved values of 0.1% when at least 10 ng of DNA were analyzed. Since a number of studies described false-positive EGFRT790M mutation in FFPE NSCLC tumors (Ye et al., 2013; Do et al., 2017), we set a cut-off > 0.5% by “in house” experiment using 22 FFPE normal tissue samples (Additional file 3: Table S3).

Statistical evaluation of the data in this study was performed using GraphPad Prism version 6 software and XLSTAT (v.19.03.44845). Threshold for statistical significance was considered to be P < 0.05.

To date, the management of NSCLC patients progressing during EGFR TKIs treatment, includes an initial attempt to identify the T790M mutation in the patient’s plasma. In the case of a negative result, a new biopsy or fine-needle aspiration is indicated, when feasible, in order to exclude or confirm the resistance causing mutation (Normanno et al. 2017; Rolfo et al. 2018). This algorithm increases the chances of detecting T790M mutations while reducing the number of biopsies, but may still result in delayed start of subsequent treatments if the mutation is not detected in plasma. Therefore, any improvement of the sensitivity of tests on cfDNA represents a relevant clinical achievement.

Since the discovery of the importance of sensitizing EGFR mutations in the pathogenesis and treatment of NSCLC, several different Real Time PCR assays have been employed for the identification of EGFR mutations in tissue. When utilized on plasma (Kim et al. 2013; Normanno et al. 2017), it was found that, despite high specificity, these methods were hampered by a low sensitivity. The advent of NGS has opened up new perspectives mostly because of its multiplexed gene approach, although the sensitivity of this method remains challenging, given that a value of 1–5% may be considered an acceptable limit of detection. Attempts to overcome these limits may increase the risks of false positive calls.

The technology used in this study presents several advantages compared to the classical NGS approach, since the molecular tagging step generates single tag DNA molecule that are amplified in a subsequent step. The allelic variants will be called only if two identical molecular tags (referred to as family tags) share the same mutation. The use of single-strand barcodes aided in removing mostly late arising PCR errors as well as sequencer miscalls, while maintaining an optimal sensitivity and specificity. Our analytical tests showed that a 0.1% sensitivity could be readily achieved with 30 ng doses of the standard reference for all EGFR allelic variants considered and that this dose could be lowered at 20 ng while maintaining satisfactory results (https://assets.thermofisher.com/TFS-Assets/LSG/brochures/verification-oncomine-lung-cfdna-ion-s5-white-paper.pdf).

In our study, tag-based NGS improved detection rate for both sensitizing and, perhaps more important, T790M resistance mutations compared to Real Time PCR (85.7 and 42.85% versus 61.9 and 21.4%, respectively). The assay was especially sensitive for L858R variation and the T790M resistance mutations that were not detectable by the Real Time PCR and had low allele frequency down to 0.06 and 0.07%, respectively.

Since the T790M mutation is usually present in cfDNA in quantities lower than those of the sensitizing EGFR mutations (this study and Del Re et al. 2017; Karlovich et al. 2016), it is justified the need of high sensitivity tests but the disadvantage of increasing false positive signals should be taken into account. However, in this study, the presence of T790M in cfDNA by tag-based NGS was an unlikely finding in the absence of a sensitizing EGFR mutation. Therefore, the testing for plasma sensitizing EGFR mutation may serve as internal control that informs the likelihood of falsely negative T790M results and concomitantly, provide an indirect proof for circulating tumor derived DNA in the plasma.

Since ddPCR is known together with BEAMing PCR, to have an high technical sensitivity (down to 0.01%), we used it to confirm and validate our tag-based NGS for the only T790M mutation, and interestingly, we found equivalent sensitivity with a 100% concordance and a similar T790M allelic frequency of the positive cases. However, in our opinion even though the ddPCR deliver satisfying analytical and clinical sensitivity, we believe that the amount of cfDNA needed and the single hotspot detected per reaction, may be limiting factors for its routine application.

In our study, the proportion of patients positive for T790M mutation detected in cfDNA by tag-based NGS was in the 50% range, consistently with data reported for post-TKI biopsy tissues (Yu et al. 2013; Hata et al. 2013; Sequist et al. 2011; Oxnard et al. 2011).

Although the concordance for EGFR sensitizing mutations was almost complete between cfDNA and post-TKI tissues, some discrepancy was noted for the T790M mutation and in line with previous studies (Kim et al. 2013; Normanno et al. 2017). Indeed, our finding is not surprising and lack of concordance between plasma and post-TKI tissues depends on tumor heterogeneity as well as on specific sites of the tissue biopsy (Ilie & Hofman 2016). Interestingly, we found an enrichment of the T790M mutation among cases with exon 19 deletions. No other characteristics were correlated with the T790M-mutated status, even the thoracic versus extra thoracic metastatic sites, supporting that biological rather than clinical features may have a role in the development of T790M mutation (Ke et al. 2017).

Moreover, the tag-based NGS method appears suitable for the detection of new gene variants conferring resistance to osimertinib such as the C797X mutations as well as other mutations with similar functions (Thress et al., 2015b). It is known that cells with EGFR C797S mutation may be still sensitive to first generation TKI (and to osimertinib) when present in trans rather than in a cis configuration with T790M (Niederst et al. 2015; Arulananda et al. 2017). In particular NGS technology compared to standard methods, is able to identify both C797X and T790M mutations on the same amplicon and subsequently their cis or trans configuration. In line with this, two osimertinib treated patients (pts. 24 and 35, Additional file 4: Table S4), progressed under therapy and the tag-based NGS detected a concurrent C797S and C797G resistance mutations in cis configuration (Additional file 7: Figure S3). Therefore this NGS method may prove useful to make further therapeutic decisions.

A limit of our study may be the relative small cohort of patients analyzed, but our patients group is representative of a real life routine in the management of TKI progressed NSCLC patients.

So far, despite NGS still remains a quite expensive method, it may represent a performing test in some diagnostic settings, such as the cfDNA analysis for clinical therapy in advanced NSCLC (Coco et al. 2015). Indeed, firstly NGS technology is highly preferable for the multiplexing ability to parallel screening of different genes. Secondly, NGS approach compared to standard methods, prevents the splitting of a scarce genetic material, such as the cfDNA, in various independent reactions thus reducing additional biases particularly important when low mutational events are investigated.

Lastly, the tag-based NGS technology while increasing sensitivity and concomitantly reducing false positive calls, provide a precise determination of the T790M allelic level helping in the stratification of patients into different groups, that can be subsequently investigated for their osimertinib response (Karlovich et al. 2016; Niederst et al. 2015; Ariyasu et al. 2018).

The search of EGFR T790M mutation in ctDNA rather than in tumor tissue DNA is becoming a reliable alternative procedure. However, ctDNA analysis is technically challenging and consequently, clinical laboratories are required to implement molecular assays in order to provide reliable and accurate EGFR test results.

In conclusion this study shows that a tag-based NGS outperformed in ctDNA of post-TKI progressed NSCLC patients compared to Real Time PCR, especially for detection of the resistance T790M mutation. In this context, we propose an algorithm (Fig. 4) that can be applied for the clinical management of TKI progressed patients with advanced NSCLC. Despite the discussed relative high cost of methodology, in some instances tag-based NGS may help in reducing stressful and invasive procedures.

Open image in new window

Further larger scale studies are needed to corroborate application of tag molecular sequencing in the search of T790M in a clinical diagnostic context.