Supplemental Methods - Next Generation Sequencing
Tissue processing
Resected samples with abundant tumor tissue were selected for NGS testing. To increase
tumor content, a pathologist (K.S.) marked an H&E stained slide to delineate tumorcontaining regions, and these areas were macrodissected from serial unstained FFPE
sections. gDNA was extracted from FFPE samples using the Maxwell® 16 FFPE Plus
LEV DNA Purification kit. Two to ten slides containing 10μm thick sections were
scraped into a microtube and incubated overnight at 70°C with proteinase K solution and
incubation buffer. Subsequently, each sample was treated with lysis buffer, transferred to
loading cartridges and run in the automated instrument. gDNA quantification was
performed using the Quant-iT™ High-Sensitivity DNA Assay Kit (Life Technologies™).
In order to verify overall quality of
the gDNA, a PCR-based quality control (QC) assay
was applied, and used as a guide to recommend the amount of DNA input in the library
preparation, as recommended by the manufacturer:
(http://www.chem.agilent.com/library/usermanuals/Public/G9900-90050.pdf).
Tumor cellularity was determined by visual inspection of the number of tumor nuclei
compared to stromal background in the areas marked for macrodissection, and samples
were classified as: low cellular (1), containing sparse tumor cells within fibrous or
inflammatory background; moderately cellular (2); highly cellular (3), mostly tumor cells
without significant intervening stroma, inflammation or airspaces. Most cases were
estimated as either highly or moderately cellular (87%). Representative images are
presented in Supplemental Figure 3, and sample classification in Supplemental Table 9.
Data processing
Sequencing reads were aligned to the human genome (hg19 assembly) using SureCall
(Agilent Technologies), and variant calling was processed on the Unified Genotyper
Genome Analysis Tool Kit (GATK). Annotation of the variants was carried out on
GenomOncology’s GenomAnalytics platform (GenomOncology, Cleveland, OH). To
enable a combined analysis with SNaPshot, NGS results were filtered to exclusively
include the same base substitutions covered by that panel, summarized in Supplemental
Table 1, in addition to EGFR exon 19 deletion, EGFR exon 20 insertion, and HER2 exon
20 insertions.
Validation process
In order to validate the sequencing panel and analysis, we sequenced 3 NSCLC cell lines
(A549, HCC827, and H4006) and 12 clinical formalin-fixed paraffin-embedded samples
with known mutation status for EGFR and KRAS, as previously assessed in a Clinical
Laboratory Improvement Amendments (CLIA)-certified laboratory. Our platform proved
highly sensitive, successfully identifying 100% of the known mutations in these cases
(Supplemental Table 10). Moreover, there were no false positives among samples wild
type for EGFR and KRAS. Instead, we detected at least one plausible driver alteration in
an EGFR/KRAS wild type sample. This tumor harbored an exon 20 G776V point
mutation in ERBB2. The G776 position is a hotspot for mutations and insertions in
ERBB2, which have been associated with higher sensitivity to trastuzumab.1,2
Sequencing performance
Sequencing performance was assessed by measuring the number of reads, mapped reads,
and target base coverage. The mean number of reads and mapped reads were 5.2 millions
(range 1.8-7.7) and 5.1 millions (range 1.6-7.6), respectively. The average alignment rate
was 0.98 (standard deviation, std dev 0.015). The mean actual target coverage was 904X
(range 270-1373). The average percentages of target reads covered at least 2-fold (2X)
and 20-fold (20X) were 98.5% (std dev 0.01) and 94.3% (std dev 0.03), respectively.
Tumor cellularity was not associated with number of mutations per sample (p=0.503)
after excluding known germline variants, suggesting that high sequencing depth was
enough to counter low cellularity. The frequency of driver mutations after excluding low
cellular samples is 26.7% (23 cases out of 86 tested), therefore not distinct from the 24%
reported in the total cohort. Data on sequencing metrics are summarized in Supplemental
Tables 11 and 12.
References
1.
Shigematsu H, Takahashi T, Nomura M, et al: Somatic mutations of the
HER2 kinase domain in lung adenocarcinomas. Cancer Res 65:1642-6, 2005
2.
Arcila ME, Chaft JE, Nafa K, et al: Prevalence, clinicopathologic
associations, and molecular spectrum of ERBB2 (HER2) tyrosine kinase mutations in
lung adenocarcinomas. Clin Cancer Res 18:4910-8, 2012